Comprehensive Biotechnology 0444640460, 9780444640468

Citation preview

COMPREHENSIVE BIOTECHNOLOGY THIRD EDITION

This page intentionally left blank

COMPREHENSIVE BIOTECHNOLOGY THIRD EDITION EDITOR IN CHIEF

Murray Moo-Young University of Waterloo, Waterloo, ON, Canada

VOLUME 1

Scientific Fundamentals of Biotechnology VOLUME EDITOR

Michael Butler National Institute of Bioprocessing Research and Training (NIBRT), Dublin, Ireland

Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom 50 Hampshire Street, 5th Floor, Cambridge MA 02139, United States Copyright Ó 2019 Elsevier B.V. All rights reserved No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers may always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library ISBN 978-0-444-64046-8

For information on all publications visit our website at http://store.elsevier.com

Publisher: Oliver Walter Acquisition Editor: Priscilla Braglia Content Project Manager: Michael Nicholls Associate Content Project Manager: Fahmida Sultana Designer: Christian Bilbow

EDITOR IN CHIEF Murray Moo-Young is a distinguished professor emeritus at the University of Waterloo, Canada, where he supervises postgraduate students and others in research on biochemical engineering fundamentals, bioremediation of polluted environments, and biomanufacturing of biopharmaceuticals, biofuels and other bio-products. A Jamaican-born Chinese-Canadian, Murray received his BSc and PhD degrees from the University of London and a MSc from University of Toronto, and a postdoctoral fellowship at University of Edinburgh. He has been an invited visiting guest professor at several prestigious universities worldwide including MIT, UC Berkeley, EPFL Zurich, ECUST China, Osaka Japan, Karlsruhe Germany and Oxford UK. To date, his work has produced 13 books, 10 patents and over 365 journal papers. He is a consultant worldwide to industry, academia and government. He is founder and executive editor of the hi-impact journal Biotechnology Advances (IF>10) and the six-volume reference work Comprehensive Biotechnology, both published by Elsevier. His honors include the premier awards of the Canadian Society for Chemical Engineering and the American Chemical Society, Biochemical Technology Division. He is an elected fellow of the American Institute for Medical and Biological Engineering (FAIMBE), “one of the highest honors for a bio-engineer”, and of the Royal Society of Canada (FRSC), “the highest accolade for a Canadian scholar”.

v

This page intentionally left blank

VOLUME EDITORS Michael Butler is the Chief Scientific Officer (CSO) of the National Institute for Bioprocessing Research and Training (NIBRT), Ireland; Adjunct Full Professor in University College Dublin; and Distinguished Professor Emeritus of the University of Manitoba, Canada. Born in Wales, he gained his BSc degree in Biochemistry at Birmingham University, MSc at the University of Waterloo, and PhD at King’s College, University of London. Following this, he became a Lecturer and subsequently Principal Lecturer in Biotechnology at Manchester Metropolitan University (1974–90). He was appointed in 1990 to an Industrial Chair in Fermentation Technology at the University of Manitoba, where he was awarded the title Distinguished Professor in 2008. He has published more than 150 articles in peer-reviewed journals as well as written and edited 7 books in the area of animal cell technology. He is an editor for Biotechnology Advances and Comprehensive Biotechnology as well as being on the editorial board of Biotechnology and Bioengineering. He has always collaborated closely with industry and is a past recipient of the prestigious Canadian national Synergy Award for University-Industry innovation (2004). He was the director of MabNet, a Canadian sponsored network for monoclonal antibody production. His previous experience has included a period as Associate Dean Science at Manitoba and several periods as Visiting Scientist or Professor at MIT, Animal Virus Institute (Pirbright), and Universities of Oxford and Rio de Janeiro. In the past he has served on several grant awarding committees, including NSERC (Ottawa), NSF (Washington), MHRC (Winnipeg), and Alberta Ingenuity (Edmonton). He was a founding member of Protein Expression in Animal Cells (PEACe) and elected to the executive committee of ESACT. He was a leader of a biotechnology subgroup for a Canadian government Department of Foreign Affairs mission to Brazil; advisory panel to Swedish program on biotechnology; and chair of SFI advisory panel (Dublin). He founded Biogro Technologies Inc., a research spin-off company on media development. His research work focuses on the development of bioprocesses using mammalian cells for the production of recombinant proteins, monoclonal antibodies, and viral vaccines. He is particularly interested in the bioprocess conditions that can be used to control the biochemical structure of glycoproteins and hence the quality of biopharmaceuticals.

Colin Webb is a Professor of Chemical Engineering at the University of Manchester, UK. He graduated in 1976 as a Chemical Engineer and added a PhD in Biochemical Engineering in 1980 both at the University of Aston. He joined UMIST as a Research Associate in 1979, taking up a lectureship in biotechnology in 1983. He was appointed to a newly created industrially sponsored chair in 1994 and established the Satake Centre for Grain Process Engineering (SCGPE) at Manchester in the same year. As Director of the Centre, he raised awareness of the novel uses that cereals could be put to, particularly as sustainable feedstocks for alternative chemicals and was honored, in 1999, as the UK’s first Distinguished Fellow of the International Academy of Food Science and Technology. His research is at the interface between biotechnology and chemical engineering and is largely directed toward the sustainable bioconversion of agricultural raw materials and the development of integrated biorefinery systems. He has supervised to successful completion a total of 130 students for higher degrees, including 41 PhDs, and has more than 300 publications, including 11 books. Colin’s H-index is 51 based on more than 9000 citations. He is Editor of The Biochemical Engineering Journal and is an editorial board member of several other biotechnology related journals. Within UMIST and the University of Manchester, he has been Head of Department, Head of School, and Associate Dean. Colin has been an external advisor to a large number of universities worldwide, including international scientific advisor to Kobe University, Japan (2007–10), and a visiting professor at universities in Spain and Australia. As a practicing chemical engineer, he has been Vice-President of IChemE (2012–18), Chair of Accreditation (2004–12), and recipient of five UK government SMART awards, the IChemE Hanson Medal (2006), Council Medal (2019), Donald Medal (2019), and the IChemE Bioprocessing Prize (2011).

vii

viii

Volume Editors Dr. Antonio Moreira is Vice Provost for Academic Affairs and Professor of Chemical, Biochemical and Environmental Engineering at UMBC (University of Maryland, Baltimore County). He was chairman of the Chemical and Biochemical Engineering Department, associate dean for the College of Engineering, and associate provost for Academic Affairs. Prior to UMBC, he spent nearly 10 years in industry, with senior management positions with International Flavors and Fragrances, Inc. and Schering-Plough Corp. (now Merck, Inc.). He has significant experience with R&D and commercialization of biopharmaceuticals. He holds a BS degree in Chemical Engineering from the University of Porto, Portugal, and MS and PhD degrees in Chemical and Biochemical Engineering from the University of Pennsylvania. He has an active research program in bioprocess engineering, is author or coauthor to more than 200 publications and presentations, has overseen more than $20 million in contracts and grants, and consults with various biotechnology and pharmaceutical companies. He received a NATO Senior Fellowship and the Parenteral Drug Association’s James Agalloco Award. He was founding president for the Chesapeake Bay Area Chapter of the International Society for Pharmaceutical Engineers, chair of the Council for Biotechnology Centers for BIO, and serves on various scientific advisory boards. He is a graduate of Leadership Maryland. He has been recognized with honorary international membership by The Brazilian Academy of Pharmaceutical Sciences and received the Order of Merit in Public Education by the President of Portugal.

Professor Bernard Grodzinski earned his BSc (Toronto), MSc and PhD (York University, 1974) before becoming a Postdoctoral Fellow of Botany, Oxford, UK. Between 1975–79 he was on faculty of the Botany School of the University of Cambridge. In 1979, he returned to Canada where he serves as Professor in the Department of Plant Agriculture of the University of Guelph. He is CoDirector of Guelph’s Closed Environment Systems Research Facility (CESRF) and also heads the Biotron’s Low-Temperature Research team an initiative of Guelph and Western Universities. The unique infrastructure developed at Guelph has helped foster strong collaborative research for colleagues, students and researchers in Canada and internationally. His primary interests remain understanding photosynthesis, respiration, translocation and crop productivity investigating plant growth and homeostasis in both the natural field and artificially controlled environments (CE) that include commercial greenhouses and specialized chambers being tested for manned space programs. His efforts have led to a better understanding of natural ecotype variation and phenotype responses to environmental stresses. Genetic approaches currently being pursued include selecting plants for better light interception and improved carbon and nitrogen metabolism that control source-sink development and enhance crop yield and quality.

Zhanfeng Cui has been the Donald Pollock Professor of Chemical Engineering, University of Oxford, since the Chair was established in 2000. He is the founding Director of the Oxford Centre for Tissue Engineering and Bioprocessing (OCTEB). He was educated in China and got his BSc degree from Inner Mongolia University of Technology (1982) and MSc (1984) and PhD (1987) from Dalian University of Technology. After a postdoctoral experience in Strathclyde University in Scotland, he joined Edinburgh University as a lecturer in Chemical Engineering (1991). He then held academic appointments at Oxford Engineering Science Department as University Lecturer (1994–98) and Reader (1999–2000). He was a Visiting Professor of Georgia Institute of Technology, USA (1999); the Braun Intertec Visiting Professor to the University of Minnesota, USA (2004); and a Chang-Jiang Visiting Professor to Dalian University of Technology, China (2005). He is a Chartered Engineer, a Chartered Scientist, and a Fellow of the Institution of Chemical Engineers. In 2009, he was awarded a Doctor of Science (DSc) by Oxford University to recognize his research achievement. In 2013, he is elected to Fellow of the Royal Academy of Engineering. His research interests include tissue engineering and stem cell technologies, bioseparation and bioprocessing, and membrane science and technology. He and his coworkers have published more than 120 articles in refereed journal papers and filed 7 patent applications in the last 5 years. He is the academic founder of Zyoxel Limited, an Oxford University spin-off in 2009.

Volume Editors

ix

Dr. Hua Ye is an Associate Professor in Engineering Science and Deputy Director of the CRMI Technology Centre for Regenerative Medicine in the Institute of Biomedical Engineering at the University of Oxford. Dr. Ye holds a degree in Chemical Engineering from Dalian University of Technology (1998), China, and a PhD in Biochemical Engineering from the University of Oxford (2005). She spent 1 year exploring the commercial value of her PhD project on hollow fibre membrane bioreactor as an enterprise fellow at Begbroke Science Park of the University of Oxford before taking up the post of postdoctoral research associate at Imperial College London in 2005. Her research interests lie in stem cell bioprocessing (expansion and differentiation) and biomaterials and bioreactors for tissue engineering. Dr. Ye has published more than 50 journal articles, 6 book chapters, and secured research funding of more than £11M both from the Research Councils UK and industry.

Spiros N. Agathos has been a Professor of Bioengineering (since 1993) at the Catholic University of Louvain, Earth and Life Institute, Belgium, and since 2015, he is the Inaugural Dean of Life Sciences and Biotechnology at Yachay Tech, the first research-intensive university in Ecuador. He obtained his Dipl. Eng. in Chemical Engineering from the National Technical University of Athens, his M. Eng. in the same field from McGill University, and his PhD in Biochemical Engineering from the Massachusetts Institute of Technology (MIT). Professor Agathos was a faculty member at the University of Western Ontario (1982–85) and at Rutgers University (1985–93) and served as a visiting professor in Europe and the Americas. He has published more than 200 articles, 4 books, and 4 patents. He has been Editor or Editorial Board member of many journals and on numerous committees for science and technology policy. He is a consultant to governments and industry, while his former students and postdocs have significant academic and industrial positions across the globe. Among his many awards, he is an elected fellow of the American Academy for Microbiology (AAM), the International Water Association (IWA), the American Institute for Medical and Biological Engineering (AIMBE), and the Society for Industrial Microbiology and Biotechnology (SIMB). His research interests include fungal, insect, and mammalian cell cultures in bioreactors, biocatalyst development, bioprocess optimisation, bioreactor design and scale-up, pollutant biodegradation and site bioremediation, microbial ecogenomics, and biotechnology for sustainability.

Ben A. Stenuit is Full Professor of Industrial and Environmental Biotechnology (since 2017) at Polytech Montpellier, University of Montpellier, Joint Research Unit of Agropolymer Engineering and Emerging Technologies (IATE, UMR 1208), Montpellier, France. He is also Invited Lecturer (from 2015) on biological treatment of wastewater at the Louvain School of Engineering, Catholic University of Louvain (UCL), Louvain-la-Neuve, Belgium. He obtained his PhD in 2009 in Agricultural Sciences and Bioengineering from the Faculty of Bioscience Engineering at the Catholic University of Louvain (with Prof. Spiros N. Agathos). From 2010 to 2013, he was a postdoctoral researcher at the University of California, Berkeley, at the Civil and Environmental Engineering Department (with Prof. Lisa Alvarez-Cohen). From 2013 to 2017, he was working as a postdoctoral researcher at UCL (Earth and Life Institute). To date, he has coauthored more than 25 peer-reviewed publications. His research interests include environmental molecular microbiology, molecular systems ecology, biorefinery, industrial and environmental biotechnology, and bioprocess design and modeling.

This page intentionally left blank

SECTION EDITORS Feng-Wu Bai received his BSc and MSc degrees from Dalian University of Technology, China, and PhD from the University of Waterloo, Canada. He has been a visiting professor at world-famous universities, including MIT, and a consultant for government and industry. Currently, he is a Biochemical Engineering Professor at Dalian University of Technology, China, and his research interest comprises the combination of chemical engineering principles with biotechnology advances for the production of biofuels, bioenergy, and bio-based chemicals at large scale as an alternative to petroleum-based products. To date, his academic achievement and technical innovation have produced 2 books, 3 invited book chapters, more than 120 peer-reviewed articles, and 2 patents that have been commercialized in fuel ethanol production. He is an editor of Biotechnology Advances and editorial board member of Journal of Biotechnology and Chinese Journal of Biotechnology.

Pavneesh Madan is an Assistant Professor in the Department of Biomedical Sciences, Ontario Veterinary College, the University of Guelph. He received his DVM and MVSc degrees from College of Veterinary Sciences, Hisar (India), and PhD in Animal Reproductive Biotechnology from the University of British Columbia (UBC), Vancouver, Canada. His research interest is in understanding cellular, molecular, and genetic mechanisms and regulating preimplantation embryo development and arrest.

Massimo Francesco Marcone earned his B.Sc., B.A. and Ph.D., from the University of Guelph and now serves as a full professor of Food Science in the Ontario Agriculture College at the University of Guelph in Ontario, Canada. His institution, as a whole, ranks as the fourth most comprehensive University in Canada, with his internationally renowned academic department ranking fourth in the world in the CWUR World University Rankings. During his academic career, he has specialized in the area of food analysis and holds an active research program with several M.Sc. and Ph.D. students. He has published over 100 peer-reviewed papers in international science journals leading him to be named a Fellow of the Royal Society for Chemistry in the United Kingdom. His opinions and works have made him highly sought after by the media, especially to determine between fact or fiction when it comes to a plethora of exotic foods and delicacies. His research work has allowed him to travel around the world with several documentary journalists examining many foods of public interest and leading to the publication of three creative non-fiction books on his research work. He teaches a variety of courses taken by approximately 1,600 undergraduate students a year. His teaching accomplishments have been well-recognized through numerous prestigious teaching awards received over the years including the OAC Distinguished Teaching Award, G.P. McRostie Faculty Award, OAC Distinguished Extension Award, and University of Guelph GSA Teaching Fellowship Excellence Award, as Professor of the Year. Quality teaching and engagement of students has always been what he has strived for and giving the “A, B, C” (most accurate, balanced, and current) perspective on his subject area.

xi

This page intentionally left blank

CONTRIBUTORS TO VOLUME 1 Daryoush Abedi Isfahan University of Medical Sciences, Isfahan, Iran M Al-Rubeai University College Dublin, Dublin, Ireland B-F Alfonso University College Dublin, Dublin, Ireland PM Alves IBET, ITQB-UNL, Oeiras, Portugal John G Baust Binghamton University, Binghamton, NY, United States John M Baust Binghamton University, Binghamton, NY, United States; and CPSI Biotech, Inc., NY, United States M Bellgard Murdoch University, Murdoch, WA, Australia R Berlemont University of Liège, Liège, Belgium MJ Betenbaugh Johns Hopkins University, Baltimore, MD, United States IK Blaby Brookhaven National Laboratory, Upton, NY, United States TA Bowden Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, United Kingdom H Bozorgmanesh University of California at Irvine, Orange, CA, United States Greg Bridges Department of Electrical and Computer Engineering, University of Manitoba, Winnipeg, MB, Canada Michael Butler National Institute of Bioprocessing Research & Training (NIBRT), Dublin, Ireland

A Campbell Invitrogen, Part of Life Technologies, Grand Island, NY, United States MP Campbell Macquarie University, Sydney, NSW, Australia MJT Carrondo IBET, ITQB-UNL, Oeiras, Portugal Parminder Chahal McGill University, Montreal, QC, Canada B Chapman Murdoch University, Murdoch, WA, Australia P Chen University of Waterloo, Waterloo, ON, Canada SR Chhabra Lawrence Berkeley National Laboratory, Berkeley, CA, United States ABH Choo Bioprocessing Technology Institute, Singapore KM Coombs University of Manitoba, Winnipeg, MB, Canada; Manitoba Centre for Proteomics and Systems Biology, Winnipeg, MB, Canada; and Manitoba Institute of Child Health, Winnipeg, MB, Canada William L Corwin Binghamton University, Binghamton, NY, United States EFJ Cosgrave National Institute for Bioprocessing Research and Training, Dublin, Ireland Doug Cossar PlantForm Corporation, University of Guelph, Guelph, ON, Canada M Crispin University of Oxford, Oxford, United Kingdom V de Crécy-Lagard University of Florida, Gainesville, FL, United States

xiii

xiv

Contributors to Volume 1

T de Kievit University of Manitoba, Winnipeg, MB, Canada

DC James University of Sheffield, Sheffield, United Kingdom

Arnold L Demain Drew University, Madison, NJ, United States

R Jasso-Chávez Instituto Nacional de Cardiología, México DF, Mexico

Aline Do Minh McGill University, Montreal, QC, Canada

N Jenkins University College Dublin, Dublin, Ireland

H Dorai Centocor R&D, Inc., PA, United States

Amine A Kamen McGill University, Montreal, QC, Canada

C Foster, III University of California at Irvine, Orange, CA, United States

JJ Kattla National Institute for Bioprocessing Research and Training, Dublin, Ireland

E Frixione Department of Cell Biology, Center for Research and Advanced Studies IPN (Cinvestav), Mexico City, Mexico

JD Keasling Lawrence Berkeley National Laboratory, Berkeley, CA, United States; and Joint BioEnergy Institute, Emeryville, CA, United States

C Gerday University of Liège, Liège, Belgium

M Khajehpour University of Manitoba, Winnipeg, MB, Canada

SF Gorfien Invitrogen, Part of Life Technologies, Grand Island, NY, United States

M Kilstrup Technical University of Denmark, Kgs. Lyngby, Denmark

Peter P Gray The University of Queensland, Brisbane, QLD, Australia

MY Kim Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea

B Grodzinski University of Guelph, Guelph, ON, Canada M Hernández Department of Cell Biology, Center for Research and Advanced Studies IPN (Cinvestav), Mexico City, Mexico RB Hitchman Oxford Expression Technologies Ltd, Oxford, United Kingdom Benjamin S Hughes The University of Queensland, Brisbane, QLD, Australia NPA Hüner University of Western Ontario, London, ON, Canada A Hunter Murdoch University, Murdoch, WA, Australia

LA King Oxford Brookes University, Oxford, United Kingdom Mauricio Krause University College Dublin, Dublin, Ireland W Kuhtrieber University of California at Irvine, Orange, CA, United States JRT Lakey University of California at Irvine, Orange, CA, United States E Lattová University of Manitoba, Winnipeg, MB, Canada SY Lee Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea

Andres Illanes Pontificia Universidad Católica de Chile, Valparaíso, Chile

A Lewis IRP, NIDA, NIH, DHHS, Baltimore, MD, United States; and Johns Hopkins University, Baltimore, MD, United States

M Jafari University of Waterloo, Waterloo, ON, Canada

F Lin University of Manitoba, Winnipeg, MB, Canada

KK Jain Jain PharmaBiotech, Basel, Switzerland

S Lu University of Waterloo, Waterloo, ON, Canada

Contributors to Volume 1

xv

D Lutz Getter Bio-Med, Inc., Ramat Gan, Israel

NW Owens University of Manitoba, Winnipeg, MB, Canada

TJ Lyons Evolugate, Gainesville, FL, United States

NS Panikov Northeastern University, Boston, MA, United States

T Mamo University of Waterloo, Waterloo, ON, Canada

Maria Papagianni Department of Food Hygiene and Technology, School of Veterinary Medicine, Aristotle University of Thessaloniki, Thessaloniki, Greece

BL Mark University of Manitoba, Winnipeg, MB, Canada J Martinussen Technical University of Denmark, Kgs. Lyngby, Denmark KE McCloskey University of California, Merced, CA, United States BJ McConkey University of Waterloo, Waterloo, ON, Canada SA McKenna University of Manitoba, Winnipeg, MB, Canada J McLeod University of Sheffield, Sheffield, United Kingdom EJ Mead University of Kent, Canterbury, United Kingdom MCM Mellado New University of Lisbon, Institute of Experimental Biology and Technology, Oeiras, Portugal MR Mirbolooki University of California at Irvine, Orange, CA, United States R Moreno-Sánchez Instituto Nacional de Cardiología, México DF, Mexico Trent P Munro The University of Queensland, Brisbane, QLD, Australia D Na Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea A Nahid The University of Melbourne, Melbourne, VIC, Australia N Naidoo University of Pennsylvania School of Medicine, Philadelphia, PA, United States

JY Park Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea Michelle Peckham University of Leeds, Leeds, United Kingdom H Perreault University of Manitoba, Winnipeg, MB, Canada C Perry Chou University of Waterloo, Waterloo, ON, Canada Warren Pilbrough The University of Queensland, Brisbane, QLD, Australia Karen M Polizzi Department of Chemical Engineering, Centre for Synthetic Biology and Innovation, Imperial College London, London, United Kingdom RD Possee NERC Centre for Hydrology & Ecology (CEH) Oxford, Oxford, United Kingdom Michael Pyne University of Waterloo, Waterloo, ON, Canada H Quezada Instituto Nacional de Cardiología, México DF, Mexico S Rodríguez-Enríquez Instituto Nacional de Cardiología, México DF, Mexico U Roessner The University of Melbourne, Melbourne, VIC, Australia A Roldão New University of Lisbon, Institute of Experimental Biology and Technology, Oeiras, Portugal

R Nazarian University of Waterloo, Waterloo, ON, Canada

PM Rudd National Institute for Bioprocessing Research and Training, Dublin, Ireland

Philip Newsholme University College Dublin, Dublin, Ireland

E Saavedra Instituto Nacional de Cardiología, México DF, Mexico

SKW Oh Bioprocessing Technology Institute, Singapore

P Sadatmousavi University of Waterloo, Waterloo, ON, Canada

xvi

Contributors to Volume 1

Elham Salimi Department of Electrical and Computer Engineering, University of Manitoba, Winnipeg, MB, Canada

JD Varner School of Chemical Engineering Purdue, West Lafayette, IN, United States

Sergio Sanchez Universidad Nacional Autonoma de Mexico (UNAM), Mexico DF, Mexico

MC Vemuri Invitrogen, Part of Life Technologies, Grand Island, NY, United States

TM Sauerwald Centocor R&D, Inc., PA, United States

J Wang University of Waterloo, Waterloo, ON, Canada

CN Scanlan University of Oxford, Oxford, United Kingdom

Yuan Wen The Ohio State University, Columbus, OH, United States

F Schweizer University of Manitoba, Winnipeg, MB, Canada David Sharon McGill University, Montreal, QC, Canada AC Silva New University of Lisbon, Institute of Experimental Biology and Technology, Oeiras, Portugal CM Smales University of Kent, Canterbury, United Kingdom M Soltani University of Waterloo, Waterloo, ON, Canada Laura Stenson University College Dublin, Dublin, Ireland WB Struwe University of Oxford, Oxford, United Kingdom

M Willemoës University of Copenhagen, Copenhagen, Denmark MR Wormald University of Oxford, Oxford, United Kingdom D Wu University of Manitoba, Winnipeg, MB, Canada; and Huazhong University of Science and Technology, Wuhan, China James Wynn Bio-Based Technology Derisking, Lansing, MI, United States W Xu University of Waterloo, Waterloo, ON, Canada

Manuela Sulvucci University College Dublin, Dublin, Ireland

Shang-Tian Yang The Ohio State University, Columbus, OH, United States

Rodolfo Sumayao University College Dublin, Dublin, Ireland

H Yoshida University of Hyogo, Hyogo, Japan

M Taniguchi University of Hyogo, Hyogo, Japan

Lin Zhang University of Waterloo, Waterloo, ON, Canada

RA Tasseff Procter and Gamble, Mason, OH, United States

Xudong Zhang The Ohio State University, Columbus, OH, United States

JM Van Emon US Environmental Protection Agency, Las Vegas, NV, United States

CONTENTS OF VOLUME 1 Editor in Chief

v

Volume Editors

vii

Section Editors

xi

Contributors to Volume 1

xiii

Foreword

xxi

Preface

xxiii

1.00

Introduction Michael Butler

1

1.01

Amino Acid Metabolism Philip Newsholme, Laura Stenson, Manuela Sulvucci, Rodolfo Sumayao, and Mauricio Krause

3

1.02

Enzyme Biocatalysis Daryoush Abedi, Lin Zhang, Michael Pyne, and C Perry Chou

15

1.03

Immobilized Biocatalysts Andres Illanes

25

1.04

Lipids, Fatty Acids James Wynn

39

1.05

Structure and Biosynthesis of Glycoprotein Carbohydrates M Crispin, CN Scanlan, and TA Bowden

51

1.06

Nucleotide Metabolism J Martinussen, M Willemoës, and M Kilstrup

69

1.07

Organic Acids Maria Papagianni

85

1.08

Peptides and Glycopeptides NW Owens and F Schweizer

98

1.09

Protein Structural Analysis BL Mark, SA McKenna, and M Khajehpour

116

1.10

Secondary Metabolites Sergio Sanchez and Arnold L Demain

131

xvii

xviii

Contents of Volume 1

1.11

Cell Line Isolation and Design Trent P Munro, Warren Pilbrough, Benjamin S Hughes, and Peter P Gray

144

1.12

Cell Preservation Technology John G Baust, William L Corwin, and John M Baust

154

1.13

Cytoskeleton and Cell Motility Michelle Peckham

166

1.14

Design of Culture Media SF Gorfien, A Campbell, and MC Vemuri

181

1.15

Protein Folding in the Endoplasmic Reticulum N Naidoo

192

1.16

Extremophiles R Berlemont and C Gerday

203

1.17

Metabolic Design and Control for Production in Prokaryotes SR Chhabra and JD Keasling

217

1.18

Microbial Growth Dynamics NS Panikov

231

1.19

Modes of Culture/Animal Cells Xudong Zhang, Yuan Wen, and Shang-Tian Yang

274

1.20

Modes of Microbial Culture IK Blaby, V de Crécy-Lagard, and TJ Lyons

292

1.21

Photosynthesis and Photoautotrophy NPA Hüner and B Grodzinski

305

1.22

Protein Expression in Insect Cells RB Hitchman, RD Possee, and LA King

313

1.23

Stem Cells SKW Oh and ABH Choo

331

1.24

Structural Organization of CellsdThe Cytoskeleton E Frixione and M Hernández

355

1.25

Viruses Produced From Cells KM Coombs

372

1.26

Cell Transfection Aline Do Minh, David Sharon, Parminder Chahal, and Amine A Kamen

383

1.27

mRNA Translation and Recombinant Gene Expression From Mammalian Cell Expression Systems EJ Mead and CM Smales

391

1.28

Posttranslation Modifications Other Than Glycosylation N Jenkins

398

1.29

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories J McLeod and DC James

405

1.30

Glycomics EFJ Cosgrave, JJ Kattla, MP Campbell, WB Struwe, MR Wormald, and PM Rudd

413

1.31

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics U Roessner, A Nahid, B Chapman, A Hunter, and M Bellgard

435

Contents of Volume 1

xix

1.32

Theory and Applications of Proteomics BJ McConkey

448

1.33

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds D Na, MY Kim, JY Park, and SY Lee

456

1.34

Apoptosis: The Signaling Pathways and Their Control TM Sauerwald, A Lewis, H Dorai, and MJ Betenbaugh

468

1.35

Design Principles of Self-Assembling Peptides and Their Potential Applications P Sadatmousavi, M Soltani, R Nazarian, T Mamo, S Lu, W Xu, J Wang, P Chen, and M Jafari

480

1.36

Rational Design of Strategies Based on Metabolic Control Analysis E Saavedra, S Rodríguez-Enríquez, H Quezada, R Jasso-Chávez, and R Moreno-Sánchez

495

1.37

Unfolded Protein Response M Taniguchi and H Yoshida

508

1.38

Cell Migration D Wu and F Lin

521

1.39

Biofilms T de Kievit

529

1.40

Flow Cytometry B-F Alfonso and M Al-Rubeai

541

1.41

Biological Imaging by Superresolution Light Microscopy D Lutz

561

1.42

Biosensors Karen M Polizzi

572

1.43

Dielectric Properties of Cells Elham Salimi and Greg Bridges

585

1.44

Cell Isolation From Tissue MR Mirbolooki, H Bozorgmanesh, C Foster, III, W Kuhtrieber, and JRT Lakey

599

1.45

Nanobiotechnology KK Jain

607

1.46

Effects of Shear Stress on Cells KE McCloskey

624

1.47

Viruses and Virus-Like Particles in Biotechnology: Fundamentals and Applications A Roldão, AC Silva, MCM Mellado, PM Alves, and MJT Carrondo

633

1.48

Mathematical Models in Biotechnology RA Tasseff and JD Varner

657

1.49

Immunoassays in Biotechnology JM Van Emon

668

1.50

Mass Spectrometry H Perreault and E Lattová

679

1.51

Bioprocessing Techniques Doug Cossar

688

This page intentionally left blank

FOREWORD For a long time, biotechnology has been the young stepson of basic biomedical sciences, identified mostly with the food industry and a bit later with development of vaccines. It was not highly regarded by basic researchers who focused on decipheringdamong other problemsdmetabolic pathways, the genetic code, and the underlying pathogenetic mechanisms of diseases. Drug discovery at the beginning of the century was incidental (e.g., aspirin, penicillin, and even insulin), and the idea of curing a genetic disease was not even a dreamdthey were regarded irreversible. Nobody has dreamt on the possibility of genetic manipulation. This landscape has since changed dramatically, and biotechnology has become the respected inseparable Siamese twin of basic biomedical sciences. Efforts to translate fundamental discoveries into useful products are recognized now as an important and integral hallmark in both academia and industry, and the traditional border between the two has lost its sharp lines of demarcation. In many places around the world, we witness a process where the biotechnological industry, and mostly its research and development branches, is growing and developing around universitiesdwhich serve as nuclei of crystallization for their flourishing, generating a new type of collaborative ecosystem we have not known before. MIT and the pharmaceutical industry in Cambridge, Massachusetts, USA, are probably prime examples of this process. The border between science and its technological applications has almost disappeared. The fences among faculties of biomedical basic sciences, medicine, biotechnology, chemical engineering, chemistry, physics, and civil engineering, among many others, have been lowered, and they hire these daysdmore and more frequentlydscientists with similar profilesdactually competing with one another. Industry is investing more and more efforts in research and development, and has tightened its ties with academia. Consequently, it is employing leading and world-renowned scientists that academia would have been proud to have, and is using state-of-the-art, most advanced, and innovative technologies that are far from being just scaled up pilot processes developed in academia. Working nowadays in the biotechnological industry is no longer regarded as a second choice for academics, and the movement between the two enterprises has become bilateral compared to the unilaterality that domineered the relationship between the two for a long time: experience in industry is regarded nowadays by many in academia as an advantage. Not surprisingly, even the basic nomenclature is changing, and biotechnology has lost its defined traditional boundaries and merged into the actively evolving conglomerate of biomedical sciences, along with cell and molecular biology, immunology, biochemistry, genetics, agriculture, biomedical engineering, and pharmaceutical sciences, among other areas. What are the roots of this revolution? When did it all start? Like many other revolutions, including political ones, we can easily identify points of time when they erupted, but find it more difficult to identify the underlying streams and developments that led to the eruption. It is widely accepteddI thinkdthat the seeds of the revolution were planted with the deciphering of the mechanisms that underlie the central dogma of biologydthe discovery of the double helical structure of DNA which immediately disclosed its mechanism of replication, and then the discovery of mRNA, the specific tRNAs, and the mechanisms of protein translation. These discoveries that were made in the 1950s led to the development of an avalanche of subsequent technologies which enabled us to sequence and then manipulate genes, express or silence them in different organismsdfrom viruses to mammalsdto analyze the other elements in the central dogma that are located hierarchically above DNA (RNA, proteins, small moleculesde.g., the different omics). Last but not least, they enabled us to edit genes (CRISPR-Cas) and to generate proteins that are better in many aspects than the natural ones (e.g., synthetic biology). There is currently no field in biomedical and life sciences that is not deeply involved with biotechnology: from the development of vaccines and small molecules, through drug targeting in tissue-specific liposomes or

xxi

xxii

Foreword

nanoparticles; from the development of insects- or fungi-resistant plants to animals that were “cleaned” from viral-hosting genes and that have humanized HLA system (and can serve therefore as an unlimited source of organs for transplantation); from yeast that were engineered to generate antimalarial drugs (e.g., artemisinin), to microorganisms that can make cannabinoids or odd chain fatty acids that are rare in nature; from mammals that secrete protein hormones and/or antibodies in their milk or to their circulation, to microorganisms that dissolve biofilms or clean water from biological or chemical contaminantsdthe list in nonexhaustive. The newly evolving biotechnology will revolutionize our lives in every aspectdfrom preventive and therapeutic medicine, to agriculture, to biomanufacturing and the environment. The third edition of Comprehensive Biotechnology tries to encompass all these developments and in particular the more recent ones. However, due to the exponential growth of the field, the task is daunting, and is becoming more and more difficult. Can we prophesize that future editions will have to devote a separate volume to each of the numerous subfields of what used to be “classical” biotechnologydone devoted to development of vaccines, the other to microorganisms that produce small molecules, and another one to engineered plants that can grow without using insecticides or fungicides. When this will happen, we shall be able to say that the revolution of biotechnology is completed, no need to see it as a separate area.

Aaron Ciechanover The Rappaport Faculty of Medicine and Research Institute Technion Israel Institute of Technology Haifa, Israel

PREFACE Increasingly the life sciences area is impacting virtually all technologies and vice versa; hence, the ongoing importance of biotechnology worldwide. My fellow editors and I believe that it is time to produce a third edition of Comprehensive Biotechnology to bring up to date this unique “one-stop-shopping” authorized reference. It covers the comprehensive knowledge base of this multidisciplinary field and can be readily be accessed online by neophytes as well as veterans in the field, serving a wide range of stake-holders, via the prestigious publisher, Elsevier. This third edition of Comprehensive Biotechnology follows the tradition of the previous two editions: it covers the biotechnology field comprehensively, in six volumes, and is edited by internationally renown biotechnologists for the related subdisciplines. We continue to be responsible for the development of this major reference work (MRW) as the primary source of information with a range of wide value and interest: for teachers, researchers, and administrators in academia, industry, and government. The volume editors, associate editors, and section editors have relevant expertise to handle all essential constituent components of the multiauthored contents indicated below. Since the publication of the second edition, most of the foundation knowledge base has changed to the extent that new material could be incorporated by the expedient of postscripts to several chapters. However, other topics needed new or completely revised chapters. Notably among these are those which dealt with the recent innovations in gene editing, artificial intelligence, digital impaction, renewable resources, and climate change on the field. Comprehensive Biotechnology 3rd edition is published as an electronic online publication, with the following six volumes. Vol 1: Scientific Fundamentals of Biotechnology, edited by Prof. Michael Butler. Vol 2: Engineering Perspectives in Biotechnology, edited by Prof. Colin Webb. Vol 3: Industrial Biotechnology and Commodity Products, coedited by Prof. Antonio Moreira and Prof. Fengwu Bai. Vol 4: Agricultural and Related Biotechnologies, coedited by Prof. Bernard Grodzinski, Prof. Pavneesh Madan and Prof. Massimo Marcone. Vol 5: Medical Biotechnology and Healthcare, coedited by Prof. Zhanfeng Cui and Prof. Hua Ye. Vol 6: Environmental and Related Biotechnologies, coedited by Prof. Spiros Agathos and Prof. Benoit Stenuit Obviously Comprehensvie Biotechnology, third edition, would not have been possible without the combined expertise of the collective editors with their extensive professional connections. As with the previous two editions, this edition provides the following important points.

• All six volumes are published at the same time online, not as a series; this is not a conventional encyclopedia

but a symbiotic integration of brief articles on established topics and longer chapters on new or emerging areas. • Hyperlinks provide sources of extensive additional, related information instantaneously; material authored and edited by world-renown experts in all aspects of the broad multidisciplinary field of biotechnology. • Scope and nature of the work are vetted by a prestigious International Advisory Board including three Nobel laureates. • Most chapters carry a glossary and a professional summary of the authors indicating their appropriate credentials.

xxiii

xxiv

Preface

• An extensive index for the entire publication gives a complete list of the many topics treated in the increasingly expanding field.

• To facilitate the one-stop shopping for rapid customer service, the work is designed as an integration of six

mini-encyclopedias. The first two ones provide the bases of the scientific and engineering principles of biotechnology, and the remaining four volumes treat the four major branches that impact on industry, medicine, agriculture, and the environment. Each volume has several sections which addresses the various aspects of the given branch. For convenience each section identifies the relevant topics in alphabetical order. To accomplish the work with appropriate expertise, many authors have contributed to CB3.

As before, this edition is blessed by a contribution of a Nobel laureate; this time, Prof. Aaron Ciechanover (Technion, Haifa); for Comprehensive Biotechnology, 2nd edition, Prof. Werner Arber (U Basel, Switzerland); for Comprehensive Biotechnology, 1st edition, Prof. Don Glaser (UC, Berkeley, US). Murray Moo-Young, PhD, FAIMBE, FRSC Comprehensive Biotechnology, 3rd edition Editor-in-Chief Distinguished Professor Emeritus, Department of Chemical Engineering Advisory board, Centre for Bioengineering and Biotechnology University of Waterloo, Ontario, Canada Postscript As of possible interest are reviews of the first and second editions of Comprehensive Biotechnology It would be very difficult to thoroughly cover the scope of biotechnology in one book. But this new edition of Comprehensive Biotechnology (1st ed., 1989) accomplishes what the title claims; the six-volume set provides detailed synopses of the applications, instruments, methodologies, and principles of modern biotechnology. Volume 1 provides the science background needed to understand biotechnology; it covers the essential biochemistry, biology, biophysics, chemistry, and computer science used in biotechnology applications and research. An explanation of engineering principles relevant to biotechnology follows in volume 2. The authors focus on engineering concepts appropriate to biotechnology product manufacturing. The third volume builds on the first two volumes in its coverage of biotechnology applications in industry and commodity products, including coverage of food ingredients, clinical products, and specialty chemicals. Summing Up: Highly recommended. Lower-division undergraduates through professionals. CHOICE

Review of the First Edition Murray Moo-Young and his colleagues have brought off a notable success in producing this work... Comprehensive Biotechnology will be an essential purchase for all departments and institutions, academic or industrial, that claim an interest in any aspect of the ill-defined field popularly known as biotechnology. Nature, Volume 321 (1986).

PERMISSIONS ACKNOWLEDGEMENT The following material is reproduced with kind permission of Oxford University Press Figure 3. Sulfur Metabolism in Plants and Related Biotechnologies. www.oup.com The following material is reproduced with kind permission of American Association for the Advancement of Science Figure 5a. Microfluidic Technology and Its Biological Application. www.aaas.org The following material is reproduced with kind permission of Nature Publishing Group Figure 10. Microbial Growth Dynamic Table 4. Microbial Growth Dynamic Figure 4. Integrated Production and Separation. http://www.nature.com

i

1.00

Introduction

Michael Butler, National Institute of Bioprocessing Research & Training (NIBRT), Dublin, Ireland © 2019 Elsevier B.V. All rights reserved.

The first volume of the third edition of Comprehensive Biotechnology covers the scientific fundamentals that underpin the biotechnology area. Most of the chapters in this volume have been updated from earlier editions, and new chapters included reflecting development in areas that have emerged since the early editions of this series. The science described in this volume is fundamental to the rapid expansion of biotechnology and reflects strategic interactions between science and engineering. Whereas the life sciences are dedicated to gaining an understanding of existing life forms, the application of these scientific fundamentals to biotechnology involves manipulation and engineering that leads to useful products that can be used in medicine, agriculture, industry, or environmental control. The scientific fundamentals that underpin biotechnology are explored and explained in this first volume. The stages that lead from scientific discovery to commercial production of a useful biological can be well illustrated by reference to the first antibiotic, penicillin. The process took several developmental stages extended over several years that in many ways parallel the process that goes on in a large biotechnology today, albeit at a much faster pace. It was in 1928 that Alexander Fleming initiated the first stage with his scientific observation that the mold Penicillium notatum grown on a Petri dish could produce an inhibitory zone, preventing bacterial growth. The second stage was the realization that this observation was related to a secreted product of the mold that could have therapeutic value in the treatment of human bacterial infections. The third stage was not accomplished until 1940 when Florey and Chain at Oxford isolated and purified the secreted product that we know as penicillin. The fourth stage involved process development to produce sufficient quantities of penicillin that could be used as the first antibiotic medicine. The fortuitous discovery of Penicillium chrysogenum and subsequent repeated cycles of random mutation enhanced the specific productivity x1000 compared to the original Fleming isolate. The original multiple, small, shallow containers were replaced by deep fermentation tanks with agitators to enable aeration. This took the production from a multiple process to a unit process amenable to scale up to the desired volume required for mass biosynthesis of penicillin. This development process took around 18 years from discovery to large-scale production and is the basis for today’s production level of penicillin: 10 million kg worldwide at a sale price around 5 cents a gram, which would have been unbelievable figures to Fleming. The stages of penicillin development from discovery to production mimic what happens in more modern biotechnology. The stages are shorter because of our greater understanding of the scientific fundamentals, greater ability to manipulate systematically, and the use of platform technologies that are suitable for groups of useful biologicals (such as monoclonal antibodies). Isolated genes can be inserted into common producer cells such as Chinese hamster ovary (CHO) cells that have been adapted to suspension culture and have the robustness to withstand agitation in large-scale stirred-tank bioreactors. The random mutation steps for enhanced specific productivity have been replaced by the systematic use of marker genes that can be amplified or targeted transfection at genomic hot spots. The understanding of basic cell metabolism has enabled the development of fed-batch cultures that allow prolonged culture of cells at high density and the routine production of recombinant proteins in stirred-tank bioreactors at concentrations of 5 g L 1 and higher. There continues to be an increase in the market value of biopharmaceuticals, which is now close to 200 billion dollars per year.3 Monoclonal antibodies dominate this market and represent 8 of the top 10 bioproducts in recent times, an overwhelming majority of these being produced in CHO cells. There are, however, signs that the price of some of the top-value products may be driven down as a result of pressure for sales in a wider market and as patents expire with biosimilars competing for market share. This parallels the developmental history of penicillin with a gradual decrease in price through a combination of demand, market competition, and improved production processes. There are, however, technical challenges that beset biosimilars, through the inherent complexity of biopharmaceuticals.2 These products are not single chemical entities, and the myriad of variants such as glycoforms pose a challenge to demonstrate biological equivalence of new-market biosimilars.4 Some of these challenges may be met by increased bioprocess control with enhanced resolution and high-throughput analysis of final products. This of course requires a good understanding of the scientific fundamentals of the bioprocess, cells, and product. Most would agree that the starting points of modern biotechnology came in the early 1970s with the isolation of enzymes such as restriction endonucleases and ligases that enabled the precise cutting and rejoining of strands of DNA. This led to the techniques of recombinant DNA technology that Cohen and Boyer exploited for the production of selected proteins in single-cell hosts. This was followed by the production of the first human recombinant protein (somatostatin) for bacteria and soon to be followed by human insulin. A notable discovery at this time (1975) was also made by Kohler and Milstein with the ability to fuse cells to make immortal lines for the production of monoclonal antibodies. The distinct commercial nature of biotechnology was made clear in 1980 with the first patent for a life form. This ground-breaking patent was issued to Exxon for a genetically modified Pseudomonas, capable of degrading crude oil with application in treating oil spills. This clearly made the philosophical change from a discovery to an invention. Subsequent major developments have included the establishment of pluripotent human stem cells as a viable cell-based human therapy in regenerative medicine (1998) and the complete sequencing of the human genome (2000). Since then the business of biotechnology has been a dominating force in the market based upon numerous novel life forms and bioprocesses associated with the production of biologicals. Predictions are for further strong economic growth in this area in the future.1

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00001-X

1

2

Introduction

The series of articles presented in this volume have been divided into thematic sections covering the scientific fundamentals that underpin biotechnology. Section 1 includes a series of articles on the biochemical basis of our understanding. This includes the structure of the most important groups of metabolites (amino acids, carbohydrates, fatty acids, organic acids, cholesterol) from which large macromolecules are synthesized. This section also includes articles on some of the techniques associated with the use of selected enzymes and structural analysis, which forms the basis of our knowledge of peptides, glycopeptides, proteins, and glycoproteins. Section 2 under the title of the biological basis focuses on cellular processes, including growth dynamics and an explanation of the various techniques used for the isolation and preservation of some of the cell lines that are important in biotechnology. Cellular metabolic activities such as protein synthesis, photosynthesis, and the importance of intracellular trafficking are also included in this section. Extremophiles have been very important in producing some of the robust enzymes used in biotechnology. Also, one article concerns stem cells which are becoming increasingly important as vehicles for cell therapy. Virus production from selected cells in culture is described as the essential basis of human vaccines. Section 3 is entitled the genetic basis and includes a discussion on molecular biology, the major driver for the revolution that has occurred in recombinant DNA technology. Here the fundamentals of the cellular processes of gene expression, transcription, protein translation, and posttranslational modification are described. In vitro analysis and manipulation of these processes include nucleotide sequencing and polymerase chain reaction (PCR) and are explained in this section. Systems biology is the topic of Section 4 with the focus on pools of biological molecules usually included with the -omics suffix such as genomics, proteomics, glycomics, and metabolomics. The approach of systems biology is to consider the importance of interactions between individual components of these pools and is contrary to the reductionism, which is the basis of classical biochemistry. Use of this approach allows the possibility of metabolic engineering to systematically change cells for increased productivity or performance under bioprocess conditions. Section 5 is entitled the metabolic basis and considers the application of our knowledge of cellular metabolism in designing and controlling producer cells in biotechnological applications. The biophysical basis (Section 6) concerns some of the important physical technologies that have enabled cellular analysis such as flow cytometry and microscopy. The overall behavior and interaction of cells is described through articles on cell migration and biofilms. Here is also described the growing area of nanobiotechnology and the robustness of cells that is important to analyze in large-scale bioprocesses. Process control during production of biopharmaceuticals has become important to ensure consistency of these complex molecules. This is enabled by the use of sophisticated biosensors, which allows online monitoring of the conditions in a bioreactor. Analysis of the state of the cells during production is also improved by virtue of an increased understanding of the dielectric properties of cells that might change in line with different metabolic states. The clear application of this is through the increased use of capacitance probes that can monitor cell growth during a bioprocess run. The computational basis is dealt with in Section 7 and is important in relation to the -omic approaches in biotechnology, which tend to generate an enormous amount of data that is the raw material used in bioinformatics and the mathematical modeling of bioprocesses. The final section entitled analysis and control describes selected applications of some of the techniques developed through the scientific fundamentals for routine analysis such as immunoassays and mass spectrometry. These fundamental techniques have been applied to product characterization to ensure the desired consistent quality of bioproducts that can result from good design and control of the bioprocess as described in the last article on bioprocessing techniques.

References 1. Goodman, M. Market Watch: Sales of Biologics to Show Robust Growth through to 2013. Nat. Rev. Drug Discov. 2009, 8 (11), 837. 2. Kos, I. A.; Azevedo, V. F.; Neto, D. E.; Kowalski, S. C. The Biosimilars Journey: Current Status and Ongoing Challenges. Drugs Context 2018, 7, 212543. https://doi.org/ 10.7573/dic.212543. 3. Walsh, G. Biopharmaceutical Benchmarks 2018. Nat. Biotechnol. 2018, 36, 1136–1145. 4. Hussaarts, L.; Mühlebach, S.; Shah, V. P.; McNeil, S.; Borchard, G.; Flühmann, B.; Weinstein, V.; Neervannan, S.; Griffiths, E.; Jiang, W.; Wolff-Holz, E.; Crommelin, D. J. A.; de Vlieger, J. S. B. Equivalence of Complex Drug Products: Advances in and Challenges for Current Regulatory Frameworks. Ann. N. Y. Acad. Sci. 2017, 1407 (1), 39–49.

1.01

Amino Acid Metabolism

Philip Newsholme, Laura Stenson, Manuela Sulvucci, Rodolfo Sumayao, and Mauricio Krause, University College Dublin, Dublin, Ireland © 2011 Elsevier B.V. All rights reserved. This is a reprint of P. Newsholme, L. Stenson, M. Sulvucci, R. Sumayao, M. Krause, 1.02 - Amino Acid Metabolism, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 3-14.

1.01.1 1.01.2 1.01.3 1.01.3.1 1.01.3.1.1 1.01.3.1.2 1.01.3.1.3 1.01.3.2 1.01.3.2.1 1.01.3.2.2 1.01.3.2.3 1.01.3.2.4 1.01.4 1.01.4.1 1.01.4.2 1.01.4.3 1.01.4.4 1.01.4.5 1.01.4.5.1 1.01.4.5.2 1.01.4.5.3 1.01.4.5.4 1.01.4.5.5 1.01.4.5.6 1.01.4.5.7 1.01.5 References

Introduction General Properties, Classification, and Structure of Amino Acids Biosynthesis of Amino Acids Biosynthesis of the Nonessential Amino Acids Alanine, Asparagine, Aspartate, and Glutamate Are Synthesized From Pyruvate, Oxaloacetate, and 2-Oxoglutarate Proline and Ornithine Are Synthesized From Glutamate Serine and Cysteine Are Derived From 3-Phosphoglycerate Biosynthesis of Essential Amino Acids Lysine, Methionine, Threonine, Asparagine, and Isoleucine Are Synthesized From Oxalacetate and Aspartate Leucine and Valine Are Synthesized From Pyruvate Aromatic Amino Acids Are Synthesized From Phosphoenolpyruvate and Erythrose 4-Phosphate Histidine Is Derived From 5-Phosphoribosyl-a-pyrophosphate and ATP Catabolism of Amino Acids Introduction to Catabolism General Processes in Amino Acid Catabolism Deamination and Transamination The Urea Cycle Metabolic Catabolism of Individual Amino Acids Amino Acids Are Catabolized Into Pyruvate Amino Acids Catabolized to Oxaloacetate Amino Acids Catabolized to 2-Oxoglutarate Amino Acids That Are Catabolized to Succinyl-CoA Amino Acids That Are Catabolized into Acetyl-CoA and Acetoacetate Catabolism of Branched Acids Catabolism of Aromatic Acids Important Biomolecules Synthesized From Amino Acids

4 4 6 6 7 7 7 7 7 7 7 8 8 8 8 9 10 10 11 12 12 12 12 13 13 14 14

Glossary Acid A chemical that may donate a proton in an aqueous environment. Anabolism A set of metabolic reactions that construct larger biomolecules from smaller units. Amino acid A molecule containing an amine group and a carboxylic acid group, and a variable side chain containing carbon and hydrogen atoms that may be supplemented with oxygen, nitrogen, or sulfur atoms. Base A chemical that may accept a proton in an aqueous environment. Catabolism A set of metabolic reactions that breakdown complex biomolecules into simple units, often coupled to the release of energy. Coenzymes A loosely bound cofactor, which is a molecule required for chemical catalysis at the active site of an enzyme. Enzyme A protein that is required to increase the rate of a chemical reaction. Hormones A chemical released by a cell in one part of the body that communicates with other cells in the organism so as to provide an integrated physiologic response. Isoelectric point The pH at which a particular molecule carries no net electrical charge. Nucleotides Molecules that, when combined with other molecular building blocks, make up the structural units of RNA (ribonucleic acid) and DNA (deoxyribonucleic acid). Metabolism A series of linked chemical reactions that occur in living cells to sustain life. pH A measure of the acidity or basicity of a solution. The numerical value of pH represents the negative logarithm (base 10) of the molar concentration of hydrogen ions in a solution.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00002-1

3

4

Amino Acid Metabolism

pKa A quantitative measure of the strength of an acid in solution. It is the equilibrium constant for the reaction in which the proton dissociates from the original undissociated acid. Polarity A chemical concept that indicates how electrons shared between atoms are distributed.

1.01.1

Introduction

Proteins exert essential functions in biology, from structural roles, secreted signaling molecules, ion channels, transport, or catalysts of biochemical reactions (enzymes). The unique characteristics of a protein are dictated by its linear sequence of amino acids, termed its primary structure. This sequence can determine the final conformation of a protein and also its interactions with other proteins or molecules to exert their function inside and outside the cells. It is generally accepted that only 20 proteinogenic amino acids are included in the genetic code and therefore regularly found in proteins. However, it is now accepted that a 21st amino acid, selenocysteine, exists in mammalian proteins. Hence, every mammalian protein is constructed from a set of 21 amino acids.4 Beyond their importance for the synthesis of proteins, amino acids can also be fully or partially oxidized in order to produce energy or to be converted into other compounds such as glucose, fatty acids, ketone bodies, and purine and pyrimidine bases (used for nucleotide synthesis from which RNA and DNA are formed). In this article, we describe the structure, the characteristics, and the metabolism of the key amino acids and also discuss the importance of their availability in health and disease conditions.

1.01.2

General Properties, Classification, and Structure of Amino Acids

All of the amino acids used for protein synthesis have the same general structure (Figure 1). They contain a carboxylic acid group (–COO
) and an amino group (–NH3) attached to the alpha carbon (the carbon atom next to the carboxylate group) in an L-configuration, a hydrogen atom, and a chemical group called a side chain (–R) that is characteristic for each different amino acid. In solution, the free amino acids can exist as zwitterions (a chemical compound that carries a total net charge of zero and is thus electrically neutral but carries formal charges on different atoms), in which the amino group is positively charged and the carboxylate group is negatively charged. In proteins, these amino acids are joined into linear polymers called polypeptide chains through chemical bonding between the carboxylic acid group of one amino acid and the amino group of the next amino acid (Figure 2). As mentioned above, the chemical properties of the side chain determine the types of chemical interactions and molecular functions. Thus, amino acids are often grouped by the polarity of the side chain (e.g., charged, nonpolar hydrophobic, or uncharged polar) or by structural features (e.g., aliphatic, cyclic, or aromatic). The side chains of the nonpolar hydrophobic amino acids cluster together to exclude water in the hydrophobic effect characteristic of the core of globular proteins. The uncharged polar amino acids participate in hydrogen bonding. Amino acids that contain a sulfhydryl group, such as cysteine, form disulfide bonds with other components. The negatively charged acidic amino acids form ionic (electrostatic) bonds with positively charged molecules, such as basic amino acids. The charge on the amino acid at a particular pH is determined by the pKa of each chemical group that is associated with a dissociable proton. A list of the 20 amino acids, their classification, isoelectric point (pIdthe pH at which a particular molecule or surface carries no net electrical charge), and pKa is detailed in Table 1. Amino acids fulfill vital roles throughout biology. Not only do they function as building blocks for protein synthesis but they can also act as precursors to neurotransmitters, signaling molecules, and antioxidants. Amino acids can be combined in a large number of possible arrangements to form a variety of different proteins that can function as monomers or in combination with other peptide chains to form functional multisubunit structures.1 H C

R

COOH

H2N Figure 1

General structure of the amino acids. R ¼ side chain.

peptide bond

H

O

R C

C

H O R

C

C

NH2

O

H

+

H

NH

O

H

H

O

R C

C

NH2

N

H

R

C

C

O OH H

Figure 2 Polypeptide bond between amino acids in proteins. A peptide bond (amide bond) is a covalent chemical bond formed by the reaction between the carboxyl group of one amino acid and the amino group of another amino acid, releasing water.

Amino Acid Metabolism Table 1

5

Structure, name, abbreviation, pKa and isoelectric point of the common amino acids

Side chain (-R) structure Neutral amino acids: nonpolar, aliphatic

Neutral amino acids: polar, aliphatic

Neutral amino acids: aromatic

Neutral amino acids: sulfur-containing

Name

Abbreviation

pKa/a-COOH

pKa/a-NH2

pI

Glycine Alanine Valine Leucine Isoleucine

Gly or G Ala or A Val or V Leu or L Ile or I

2.3 2.3 2.3 2.4 2.4

9.6 9.7 9.6 9.6 9.7

6.0 6.0 6.0 6.0 6.1

Serine Threonine

Ser or S Thr or T

2.2 2.6

9.2 10.4

5.7 6.5

Asparagine

Asn or N

2.o

8.8

5.4

Glutamine

Gln or Q

2.2

9.1

5.5

Phenylalanine

Phe or Fq

1.8

9.1

5.5

Tryptophan

Trp or W

2.4

9.4

5.9

Tyrosine

Tyr or Y

2.2

9.1

5.7

Cysteine Cystine

Cys or C Cys-Cys

1.7 2.3

10.8 9.7

5.0 5.1

Met or M

2.3

9.2

5.8

Pro or P

10.6

6.3

6.3

Aspartate Glutamate

Asp or D Glu or E

2.1 2.2

9.8 9.7

3.0 3.2

Lysine Arginine

Lys or K Arg or R

2.2 2.2

9.0 9.0

9.8 10.8

Histidine

His or H

1.8

9.2

7.6

Methionine Neutral amino acids: containing secondary amino group Proline

Acidic amino acids

Basic amino acids

The specific linear arrangement of amino acids is vital to the function and conformation of a protein. A slight change in the amino acid sequence can be detrimental. A single amino acid substitution in hemoglobin, the protein that transports oxygen via red blood cells, leads to sickle cell disease.1 Red blood cells are disk shaped in healthy individuals. However, abnormal hemoglobin molecules crystallize causing the red blood cells to appear sickle shaped in affected individuals. People with the disease experience sickle cell crises in which the sickle-shaped cells block small blood vessels and obstruct blood flow.1

6

Amino Acid Metabolism Table 2

Essential and nonessential amino acids in humans

Essential

Nonessential

*Arginine *Histidine Isoleucine Leucine Lysine Methionine Phenylalanine Threonine Tryptophan Valine

Alanine Asparagine Aspartate Cysteine Glutamate Glutamine Glycine Proline Serine Tyrosine

*Essential for infants since these amino acids can only be synthesized in older children and adults

The R group distinguishes amino acids from one another and dictates the unique properties of each amino acid.2 The R group is important as it allows for posttranslational modifications that can modulate a protein’s function. Examples of such posttranslational modification include phosphorylation, nitrosylation, and acetylation. Proline is the only cyclic amino acid. It is more conformationally restricted than other amino acids due to its ring structure and has a major influence on protein arrangement.3 The aromatic amino acids have distinct absorption spectra in the near-ultraviolet range due to the presence of the aromatic ring. This allows proteins to be characterized and quantified using a simple laboratory spectrophotometer.2 Amino acids that have side chains carrying a negative charge are classified as acidic as a result of the carboxyl group in the side chain. Amino acids containing amino groups in their side chains are designated basic.1 All amino acids except glycine share an important characteristicdthey show chirality or asymmetry.2 This chirality results from the asymmetric alpha carbon connecting to four different groups. Some amino acids such as isoleucine and threonine have additional chiral centers because each possesses an additional asymmetric carbon.2 Glycine is the only amino acid that lacks a chiral center as the alpha carbon is not asymmetric, binding to two hydrogen atoms. The word chiral is derived from the Greek word meaning hand. An amino acid that possesses a chiral center is labeled either an L-(levo, left) or a D-(dextro, right) stereoisomer according to the direction it rotates polarized light. The L-stereoisomer of an amino acid is a mirror image of the D-form, just as the right hand is a mirror image of the left hand. The L- and D-amino acids of a stereoisomeric pair are known as enantiomers.2 Stereoisomers share the same molecular formula and bond sequence but differ by the arrangement of their atoms in three-dimensional space. Only L-amino acids occur in mammalian proteins. However, 2 D-isomers can be detected in abundance in peptidoglycan bacterial cell walls and also in certain antibiotics.

1.01.3

Biosynthesis of Amino Acids

Amino acids are synthesized in microorganisms, plants, and animals. Some amino acids are synthesized via pathways specific to plants and microorganisms. These amino acids cannot be synthesized de novo by mammals and, therefore, must be obtained from the diet and are known as “essential amino acids.” The amino acids that can be synthesized by mammals from common intermediates derived from metabolic pathways are referred to as “nonessential amino acids.” The essential and nonessential amino acids in humans are listed in Table 2. This classification of amino acids is not really an accurate dichotomy as there are some considerable overlaps between the two. Note that arginine is classified as an essential amino acid despite the fact that it can be synthesized from the urea cycle. This is because most of the arginine synthesized from this route is hydrolyzed to urea and ornithine by arginase and there are not sufficient amounts to sustain the normal growth and development of infants and children. In humans, tyrosine is synthesized by the hydroxylation of phenylalanine and the sulfhydryl group of cysteine is derived from methionine. Thus, both of these amino acids can also be considered nonessential. In addition, different food sources contain different proportions of essential amino acids. Thus, the diet must contain a variety of different protein sources that complement each other so as to supply all the essential amino acids in the required proportion.

1.01.3.1

Biosynthesis of the Nonessential Amino Acids

The biosynthesis of nonessential amino acids involves relatively simple pathways of metabolism compared with the essential amino acids that are generally more intricate. Except for tyrosine, which is synthesized by the hydroxylation of an essential amino acid phenylalanine, all the nonessential amino acids are synthesized from intermediates of common metabolic intermediates: pyruvate, oxaloacetate, a-ketoglutarate, and 3-phosphoglycerate. The reader is referred to Ref. 4 for details of common metabolic pathways and their intermediates.

Amino Acid Metabolism 1.01.3.1.1

7

Alanine, Asparagine, Aspartate, and Glutamate Are Synthesized From Pyruvate, Oxaloacetate, and 2-Oxoglutarate

Pyruvate, oxaloacetate, and a-ketoglutarate are three common a-ketoacids, which can be transaminated in one step to alanine, aspartate, and glutamate, respectively. Thus, the carbon skeletons of these amino acids are traceable to their corresponding a-ketoacid. Asparagine and glutamine, on the other hand, are the products of amidations of aspartate and glutamate, respectively.

1.01.3.1.2

Proline and Ornithine Are Synthesized From Glutamate

Proline and arginine are both derived from glutamate. g-Glutamyl kinase catalyzes the first step in this process, which involves the activation of the g-carboxylate group of glutamate by phosphorylation with adenosine triphosphate (ATP). This forms the g-glutamyl phosphate intermediate that is reduced to glutamate-5-semialdehyde. Glutamate-5-semialdehyde spontaneously cyclicizes to an internal Schiff base. The final reduction to proline is catalyzed by pyrroline-5-carboxylate reductase, which requires the presence of either reduced nicotinamide adenine dinucleotide (NADH) or reduced nicotinamide adenine dinucleotide phosphate (NADPH). The formation of the semialdehyde is a branch point with one branch leading to proline, as previously described and the other leading to the formation of ornithine and arginine. In humans, glutamate-5-semialdehyde is directly transaminated to yield ornithine in a reaction catalyzed by ornithine-d-aminotransferase. Ornithine is then converted into arginine through the urea cycle. The pathway for Escherichia coli likewise involves ATP-dependent reduction of the carboxyl group of glutamate to an aldehyde, which is then converted to its corresponding amine by transamination. Hydrolysis of the acetyl protecting group eventually forms ornithine, which, as previously mentioned, is converted into arginine via the urea cycle.

1.01.3.1.3

Serine and Cysteine Are Derived From 3-Phosphoglycerate

Serine is formed from the glycolytic intermediate 3-phosphoglycerate in a three-step pathway beginning with the conversion of 3-phosphorylglycerate hydroxyl group to a ketone yielding 3-phosphohydroxypyruvate. Transamination of 3-phosphohydroxypyruvate forms phosphoserine that, upon hydrolysis, yields serine. Serine can be directly converted into glycine by serine hydroxymethyl transferase in a reaction that also yields N5,N10-methylenetetrahydrofolate. In animals, cysteine is formed from serine and homocysteine, a breakdown product of methionine. This involves reaction of homocysteine with serine to form cystathionine, which subsequently forms cysteine and a-ketobutyrate. In plants and microorganisms, serine is converted into cysteine in a two-step reaction involving the activation of serine’s hydroxyl group by acetylation followed by the displacement of acetate by sulfide.

1.01.3.2

Biosynthesis of Essential Amino Acids

Unlike the synthesis of nonessential amino acids, pathways involving the synthesis of essential amino acids are present only in plants and microorganisms and usually involve more steps than those of the nonessential amino acids. Essential amino acids are, however, synthesized from common metabolic precursors, as with the nonessential amino acids. The ready availability of essential amino acids in microorganisms that populate the gastrointestinal tract obviated the need for higher organisms to continue to produce them, which probably explains why the enzymes that synthesize essential amino acids were apparently lost early in animal evolution.5

1.01.3.2.1

Lysine, Methionine, Threonine, Asparagine, and Isoleucine Are Synthesized From Oxalacetate and Aspartate

In bacteria, lysine, methionine, and threonine are all synthesized from oxaloacetate-derived aspartate, in pathways whose common initial reaction is ATP-dependent phosphorylation of aspartate catalyzed by aspartokinase. Although lysine, methionine, and threonine originate from the same committed step catalyzed by aspartokinase, the pathways leading to the synthesis of these amino acids are independently controlled. In E. coli, for example, aspartokinase exists in three isoforms which respond differently to the three amino acids both in terms of feedback inhibition of enzyme activity and repression of enzyme synthesis. Moreover, feedback inhibition occurs at the branch points along the pathways leading to the synthesis of two different amino acids. Thus, increased levels of methionine inhibit its own synthesis by inhibiting the O-acylation of homoserine, which is the branch point between threonine and methionine synthesis from aspartate.5

1.01.3.2.2

Leucine and Valine Are Synthesized From Pyruvate

Valine and leucine are synthesized via a five- or nine-step pathway, respectively. The pathways of valine and isoleucine synthesis are comprised of similar chemical interconversions, the only difference being in the first step of the series. The step preceding the branch point between valine and isoleucine synthesis is a thiamine pyrophosphate (TPP)-dependent reaction, in which pyruvate forms an adduct with TPP. This adduct undergoes decarboxylation forming a resonance-stabilized carbanion that reacts either to the keto group of another pyruvate on the way to valine or to the keto group of threonine-derived a-ketobutyrate on the way to isoleucine. Synthesis of leucine, on the other hand, branches off from the valine pathway at the a-ketoisovalerate level.5

1.01.3.2.3

Aromatic Amino Acids Are Synthesized From Phosphoenolpyruvate and Erythrose 4-Phosphate

Phosphoenolpyruvate and erythrose 4-phosphate (an intermediate in the pentose phosphate pathway) are the precursors for the synthesis of the aromatic amino acids. The condensation of these two common intermediates leads to the formation of chorismate, which is the branch point for the synthesis tryptophan or, via a different pathway, tyrosine and phenylalanine. Chorismate is

8

Amino Acid Metabolism

converted either to anthranilate on the way to tryptophan or to prephenate leading to tyrosine or phenylalanine. Although synthesis of aromatic amino acids is exclusive to plants and microorganisms, mammals can synthesize tyrosine directly from the hydroxylation of phenylalanine.5

1.01.3.2.4

Histidine Is Derived From 5-Phosphoribosyl-a-pyrophosphate and ATP

5-Phosphoribosyl-a-pyrophosphate, an intermediate also involved in the biosynthesis of tryptophan and purine and pyrimidine nucleotides, serves as a donor of five carbons for the biosynthesis of histidine. The histidine’s sixth carbon is derived from ATP. In Arabidopsis, recombinant ATP-phosphoribosyltransferase, the first key enzyme in the biosynthesis of histidine, is inhibited by L-histidine, suggesting that histidine biosynthesis is regulated by feedback inhibition as with lysine and methionine. Histidine is special in that the pathway for its biosynthesis is linked to the pathways of nucleotide formation. Histidine residues are often found in enzyme active sites where its imidazole ring acts as a nucleophile and a general acid–base catalyst. We now know that RNA can have catalytic properties (ribozymes); therefore, the presence of the imidazole moiety in purines may play a similar role in ribozyme catalytic activity.5

1.01.4

Catabolism of Amino Acids

1.01.4.1

Introduction to Catabolism

All metabolic processes, whether at cellular, organ, or organism level, can be categorized either as catabolic (from the Greek kata ¼ downward þ ballein ¼ to throw) or as anabolic (from the Greek ana ¼ upward þ ballein ¼ to throw). More specifically, anabolic processes result in the differentiation and growth of cells, the construction of organs and tissues, and an increase in body size by synthesizing complex biomolecules from basic building blocks. The building blocks are obtained by breaking down organic substrates obtained from the environment, through catabolic pathways, so releasing chemically available energy (i.e., ATP) and/or generating metabolic intermediates used in anabolic pathways.6 Although there are at least 30 different amino acids described in nature, 21 are found in mammalian proteins.4 The metabolism and the role played by each amino acid are different, even though some common aspects can be identified. The free amino acids account for only 1% of the total present in mammals and their total intracellular concentration is one order of magnitude higher than in plasma.4 The intracellular and extracellular concentrations of most amino acids must be kept constant for regulation of nitrogen and protein metabolism (e.g., protein synthesis and degradation). Amino acids cannot be stored in the form of a large polymer, which acts in the sole capacity of a ready releasable pool of monomers; thus, surplus amino acid nitrogen is released from the organism through operation of the urea cycle while their carbon skeletons are converted into intermediates of common metabolic pathways (e.g., the citric acid cycle intermediates). Thus, amino acids act as precursors of glucose, glycogen, fatty acids, and ketone bodies and are therefore metabolic fuels.4,7 The common routes of transport and fates of amino acids are described in Figure 3. There are four main sources of amino acids: 1. Dietary intake. Only a very low concentration of free amino acids can be found in food. Amino acids are mainly obtained by hydrolysis of proteins (average intake is around 90 g d
1 in a Western diet). 2. Endogenous protein. Protein turnover releases free amino acids. 3. Intestinal bacteria. Microorganisms produce and release amino acids. 4. Others protein sources. Include secretory cells, mucus, and desquamated epithelial cells.4 The main use of protein and amino acid breakdown is to provide building blocks for the synthesis of nitrogen-based compounds (nucleotide bases, glutathione, creatine, etc.), synthesis of proteins for growth and repair and as metabolic fuels.3 Protein turnover is tightly regulated in order to eliminate and resynthesize damaged proteins (misfolded/mistranslated), to avoid aggregation and cell signaling. Protein turnover is characterized by variable rate (several orders of magnitude) and controlled by three pathways (lysosomal–autophagic, ubiquitin–proteasome, and the calpain–calpastatin systems).4 Protein synthesis can correctly take place if and only if all amino acids are available in the cell. If the concentration of an amino acid is depleted due to disease, trauma, or injury, then the amino acid can be classified as “conditionally essential.” Metabolic disorders that impair the degradation of amino acids are defined inborn errors. Phenylketonuria and maple syrup syndrome are the best known of many hereditary errors of amino acid metabolism. Both conditions lead to a pathologic accumulation of phenylalanine and leucine, isoleucine, and valine, respectively, which in turn results in mental retardation unless the patients are put on a low amino acid diet immediately, during the first months of life.6 Amino acid catabolism takes place mostly in the liver (except branched amino acids), but small intestine, skeletal and cardiac muscle, and kidney participate as well.7

1.01.4.2

General Processes in Amino Acid Catabolism

Amino acids can be broken down through three different processes: 1. Specific catabolic pathways (glycine, lysine, methionine, serine, threonine, and tryptophan)

Amino Acid Metabolism

Peripheral tissues

Purine nucleotide cycle α-KG

α-Keto acid

Amino acid

Glutamate Pyruvate

Amino acid metabolism

NH4

Glutamate NH4

α-KG

9

Circulation

Glutamine

Alanine

Glucose

Glutamine Alanine

Glutamine NH4

+

NH4

Carbon

NH4

Nitrogen α-KG

HCO3–

2ATP 2ADP+Pi

Glucose

CPSI

Carbamoyl phosphate

Urea

Cytosol

CO2 + H2O

Glutamate

Mitochondria

Alanine

Glucose

Urea

H2O

Arginase

Ornithine OTC

Arginine ASL

Ornithine

Fumarate

Argininosuccinate Malate

ASS

TCA

Citrulline

Citrulline

Fumarate

Aspartate

ADP+Pi ATP

Glutamate α-KG

Urea

Fumarate Malate

Kidney

Hepatocytes in liver

Urea

Urine

Figure 3 Fate of amino acid nitrogen. Because the nitrogen of the amino acids can form ammonium, which is toxic to the body, the liver converts ammonium nitrogen to urea and the kidney secretes it in the urine. The breakdown of amino acids and the purine-nucleotide cycle in peripheral tissues generates nitrogen (as ammonium). To avoid the toxic effects of this product, the ammonium is incorporated, by transamination reactions, into new amino acidsdalanine and glutamine. The liver will take up these two amino acids generating a carbon skeleton that can be used to produce energy or other metabolically useful compounds, such as glucose. The nitrogen group generated by the amino acid breakdown is used for the synthesis of urea in the urea cycle. The reactions in this cycle involve mitochondrial and cytosolic enzymes and also are connected to the tricarboxylic acid cycle (TCA) cycle. The nontoxic urea synthesized can be transported to the kidney and eliminated in the urine. a-KG, 2-oxoglutarate; CPS1, carbamoyl-phosphate synthetase 1; ASS, arginine-succinate synthetase; ASL, arginine-succinate lyase.

2. Conversion to another amino acid and subsequently degradated through specific catabolic pathways (arginine, asparagine, glutamine, histidine, phenylalanine, proline, and serine) 3. By transdeamination, that is, a transamination plus a deamination step (alanine, aspartate, isoleucine, leucine, ornithine, serine, tyrosine, and valine)4,5

1.01.4.3

Deamination and Transamination

Amino acids have an amino group (–NH2) attached to the a-carbon. The first step in amino acid catabolism is usually constituted by the removal of its a-amino group resulting in the production of an oxoacid and either the liberation of the nitrogen surplus (as ammonium) in deamination reactions or transfer of the amino group to an oxoacid in transamination reactions.4,7

10

Amino Acid Metabolism

A

B

2-oxoglutarate Amino acid A

α-Keto acid A

Amino acids (–NH2 donors) Metabolism intermediates

α-Keto acid B

Amino acid B

Aspartate (A)

Oxaloacetate (A)

2-oxoglutarate (B)

Glutamate (B)

GDH Transamination

Glutamate α-Keto acids

Urea cycle

NEW amino acid

Urea

2-oxoglutarate

Figure 4 Transamination reactions and its role in amino acid degradation and synthesis. (A) In this reaction (on the top), the nitrogen group from an amino acid is donated to an a-ketoacid, forming a new amino acid and the corresponding a-ketoacid of the amino group donor. An example of this reaction (bottom reaction) shows the two amino acids and their corresponding a-ketoacids. (B) Glutamate, for its participation in transamination reactions, play a key role in amino acid synthesis and degradation. The amino group (–NH2) of an amino acid can donate its group to 2-oxoglutarate to produce glutamate. The fate of the new glutamate is to participate in another transamination reaction donating its amino group to an a-ketoacid to form a new amino acid or to participate on the urea cycle to eliminate the nitrogen as urea.

In transamination reactions, the a-amino group is transferred to a keto acid (oxoacid) according to the following reaction scheme: Amino acid1 þ Oxoacid2 /Oxoacid1 þ Amino acid2 See Figure 4 for further detail. In most of the cases, oxoacid2 is 2-oxoglutarate; thus, the transamination reactions (catalyzed by transaminase enzymes) lead to the production of glutamate.4 In the hepatocyte, glutamate may be converted into 2-oxoglutarate and NH4 in a reaction catalyzed by glutamate dehydrogenase. This reaction requires NADþ or NADPþ as an oxidizing agent and regenerates 2-oxoglutarate for reuse in other transamination reactions8,9: Glutamate þ NADðPÞþ þ H2 O ¼ 2
oxoglutarate þ NHþ 4 þ NADðPÞH Subsequently, urea is synthesized from NH4 and aspartate.

1.01.4.4

The Urea Cycle

As stated above, amino acid catabolism results in the production of nitrogenous molecules (such as NH4) that must be excreted to maintain nitrogen balance. Different living organisms have deployed different mechanisms to achieve this task, which can be classified into three different categories according to nitrogen compound to be excreted: 1. Ammonotelic (ammonia excreting). Typical of aquatic animals. 2. Ureoletic (urea excreting). Less toxic waste that does not require too much dilution in waterdexcreted by terrestrial vertebrates. 3. Uricoletic (uric acid excreting). Produced by birds and terrestrial reptiles. Some animals are able to switch from ammonotelism to either ureotelism or uricoletism accordingly to change in water availability. In this section, we consider only the urea cycle because it is the most important process used by mammals to excrete nitrogen waste. The urea cycle was first described in 1932 by Hans Krebs and Kurt Henseleit. Hans Krebs wrote in 1964, “The concept of the ornithine cycle arose from the observation that ornithine, citrulline and arginine stimulated urea production in the presence of ammonia without themselves being consumed in the process.” The urea cycle is made up of five enzymatic reactions, two of which take place in the mitochondria and the other three in the cytosol [4, 5, 8 and Figure 3]. The end product is urea, NH2–CO–NH2, a remarkably simple molecule that can carry two nitrogen atoms, but which is not further metabolized in mammals but is simply excreted in the urine. For further details, the reader is referred to Ref. 4.

1.01.4.5

Metabolic Catabolism of Individual Amino Acids

The catabolism of amino acids converts their carbon backbone into citric acid cycle intermediates or their precursors; thus, they can be subsequently metabolized to CO2 and H2O releasing ATP or used to produce glucose (gluconeogenesis), see Figure 5 for further detail. Moreover, oxidative breakdown of amino acids typically accounts for 10%–15% of metabolic energy generated by animals.

Amino Acid Metabolism

Tryptophan Glycine

Threonine

Alanine Serine Cysteine

Aspartate Asparagine

Glucose

Alanine

11

Blood

Pyruvate Acetyl CoA

Oxaloacetate

Malate

Citrate

Arginine Proline Glutamine Histidine Glutamate

TCA Cycle 2-oxoglutarate

Fumarate Aspartate Tyrosine Phenylalanine

Succinyl CoA

Propionyl CoA

Valine/isoleucine/threonine/methionine Figure 5

General view of the possible fate of different amino acids during their breakdown.

In this section, we explain how each single amino acid is catabolized. The standard amino acids are characterized by different carbon skeletons, so their conversions to citric acid cycle intermediates often involve transamination with glycolytic or tricarboxylic acid (TCA) cycle intermediates.7–9 The 21 standard amino acids are broken down into one of six metabolic intermediates: pyruvate, 2-oxoglutarate, succinyl-CoA, fumarate, oxaloacetate, acetyl-CoA, or acetoacetate.8 Therefore, amino acids can be categorized into two types depending on their catabolic pathways: 1. Glucogenic amino acids. The carbon skeletons are converted into pyruvate, 2-oxoglutarate, succinyl-CoA, fumarate, and oxaloacetate and they act as glucose precursors. 2. Ketogenic amino acids. The carbon skeletons are catabolized to acetyl-CoA or acetoacetate and can therefore lead to production of fatty acids or ketone bodies. Metabolism of some amino acids may lead to the formation of more than one of the above-listed metabolic intermediates and these amino acids belong to both categories: glucogenic and ketogenic. For example, isoleucine catabolism produces both acetyl-CoA, which makes it a ketogenic amino acid, and succinyl-CoA, which leads to glucose production.7–9 In the following section, the amino acids are grouped according to the citric acid intermediate they are converted into, and a brief description of the reaction involved is given.

1.01.4.5.1

Amino Acids Are Catabolized Into Pyruvate

Amino acids that contain three carbon atoms, such as alanine, serine, glycine (via serine), and cysteine, are converted into pyruvate (the entry point for the citric acid cycle or gluconeogenesis). Transamination enzymes that catalyze key reactions require a pyridoxal phosphate cofactor.6,10 1.01.4.5.1.1 Alanine Alanine is converted into pyruvate by a reaction catalyzed by the enzyme alanine aminotransferase, which reversibly transfers the amino group from the amino acid alanine to 2-oxoglutarate to produce pyruvate and glutamate. Subsequently, 2-oxoglutarate is regenerated by glutamate dehydrogenase, from glutamate. 1.01.4.5.1.2 Serine Serine is converted into pyruvate by a reaction catalyzed by serine dehydratase, which allows the b-elimination of the hydroxyl group of serine to form an amino acrylate intermediate that in turn tautomerizes into the imine, which is then hydrolyzed to produce NH4 and pyruvate. 1.01.4.5.1.3 Glycine Glycine is converted into pyruvate by initial conversion to serine by a reaction catalyzed by the enzyme serine hydroxymethyl transferase, which requires the N5,N10-methylene-tetrahydrofolate cofactor, involving the glycine cleavage system by transfer of a methylene group from glycine.

12

Amino Acid Metabolism

1.01.4.5.1.4 Cysteine Cysteine can be converted into pyruvate via several pathways, for example, the three carbons of cysteine can be converted into cystathionine that in turn is transformed into pyruvate and homocysteine. 1.01.4.5.1.5 Threonine Threonine is an amino acid that is both glucogenic and ketogenic. The most common pathway of degradation involves the formation of acetyl-CoA and glycine. The latter is subsequently converted into serine by serine hydroxymethyl transferase, and then serine is transformed into pyruvate by serine dehydratase.

1.01.4.5.2

Amino Acids Catabolized to Oxaloacetate

Aspartate and asparagine are both readily catabolized to oxaloacetate. 1.01.4.5.2.1 Aspartate Aspartate is converted into oxaloacetate by a reaction catalyzed by the enzyme aspartate aminotransferase, which transfers an amino group from aspartate to 2-oxoglutarate to produce glutamate and oxaloacetate. 1.01.4.5.2.2 Asparagine Asparagine is hydrolyzed into aspartate and ammonia through a reaction catalyzed by the enzyme asparaginase.

1.01.4.5.3

Amino Acids Catabolized to 2-Oxoglutarate

Glutamine, proline, arginine, and histidine are converted into glutamate first and then deaminated by a transaminase reaction to produce 2-oxoglutarate. 1.01.4.5.3.1 Glutamine Glutamine is converted into glutamate by a reaction catalyzed by the enzyme glutaminase. 1.01.4.5.3.2 Proline Proline is oxidized by the enzyme proline oxidase to form pyrroline-5-carboxylate that spontaneously hydrolyzes to produce glutamate g-semialdehyde which is further oxidized to form glutamate by the enzyme glutamate-5-semialdehyde dehydrogenase. 1.01.4.5.3.3 Arginine In the urea cycle, the enzyme arginase converts arginine into urea and ornithine. The enzyme ornithine d-aminotransferase catalyzes the transfer of the d-amino group of ornithine to 2-oxoglutarate to produce glutamate and glutamate g-semialdehyde. 1.01.4.5.3.4 Histidine Histidine is converted into urocanate by a deamination reaction catalyzed by the enzyme histidine ammonia lyase. Subsequently, the enzyme urocanate hydratase adds H2O to produce 4-imidazolone-5-propionate which is then hydrolyzed by imidazolone propionase to form M-formiminoglutamate. The formimino group is then transferred by glutamate formiminotransferase to tetrahydrofolate to produce glutamate and N5-formimino-tetrahydrofolate.4

1.01.4.5.4

Amino Acids That Are Catabolized to Succinyl-CoA

Methionine, valine, and isoleucine are catabolized into propionyl-CoA that is converted into D-methylmalonyl-CoA by propionylCoA carboxylase via fatty acids b-oxidation. D-methylmalonyl-CoA is subsequently racemized into L-methylmalonyl-CoA by methylmalonyl-CoA racemase. The reaction catalyzed by methylmalonyl mutase eventually produces succinyl-CoA. 1.01.4.5.4.1 Methionine The degradation of methionine requires nine steps, one of which involves the synthesis of S-adenosylmethionine (SAM). The first step is catalyzed by the enzyme methionine adenosyl transferase that transfers the adenosyl group of ATP to the sulfur of methionine to produce SAM. Subsequently, the enzyme SAM methylase transfers the activated methyl group to an acceptor to form S-adenosylhomocysteine that is then hydrolyzed by the enzyme adenosylhomocysteinase to form homocysteine. The enzyme cystathionine b-synthase catalyzes the condensation of a serine residue with homocysteine to produce cystathionine. Cystathioniine g-lyase cleaves cystathionine into cysteine and a-ketobutyrate. 2-Ketobutyrate is transformed into propionyl-CoA by a-ketobutyrate dehydrogenase that catalyzes a reaction which is analogous to pyruvate dehydrogenase and 2-oxoglutarate dehydrogenase.5

1.01.4.5.5

Amino Acids That Are Catabolized into Acetyl-CoA and Acetoacetate

Lysine and leucine are the only purely ketogenic amino acids, as they are degraded into the precursors for ketone body synthesis, acetyl-CoA and acetoacetate.

Amino Acid Metabolism

13

1.01.4.5.5.1 Leucine Leucine degradation is similar to the branched amino acids valine and isoleucine (see below). In the first step, leucine is transaminated by branched amino acid aminotransferase to produce a-ketoisocaproate that is in turn oxidatively decarboxylated to form isovaleryl-CoA by the branched chain a-ketoacid dehydrogenase complex. Subsequently, isovaleryl-CoA is dehydrogenated to form b-methylcrotonyl-CoA by the enzyme isovaleryl-CoA dehydrogenase. Subsequently, b-methylcrotonyl-CoA is carboxylated by the enzyme ethylcrotonyl-CoA carboxylase to form b-methylglutaconyl-CoA. b-Methylglutaconyl-CoA is then hydrated by b-methylglutaconyl-CoA hydratase to form b-hydroxy-b-methylglutaryl-CoA that is then cleaved into acetyl-CoA and acetoacetate. The enzyme that catalyzes this last stage is 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) lyase, a familiar enzyme from ketogenesis.5

1.01.4.5.6

Catabolism of Branched Acids

Most of the amino acids are catabolized in the liver. On the other hand, branched chain amino acids are catabolized mainly in the skeletal muscle, adipose tissue, kidney, and the brain. The latter tissues contain the branched amino acid aminotransferase enzyme, which the liver does not.6,10 Branched chain a-ketoacid dehydrogenase is a multienzymatic complex similar to pyruvate dehydrogenase and 2-oxoglutarate dehydrogenase. This enzyme contains a thiamine pyrophosphate cofactor, a lipoamide cofactor, a flavin adenine dinucleotide (FAD) prosthetic group. The chemistry, mechanism, and structure of these enzymes are very similar. Branched chain a-ketoacid dehydrogenase is phosphorylated by a kinase, which inactivates the enzyme similarly to the phosphorylation dependent mechanism of pyruvate dehydrogenase inhibition.5

1.01.4.5.7

Catabolism of Aromatic Acids

The degradation of aromatic amino acids requires molecular oxygen (O2) to degrade the aromatic ring structure. For example, the degradation of phenylalanine starts with phenylalanine 4-monooxygenase, which adds a hydroxyl group to phenylalanine to produce tyrosine. Subsequently, tyrosine aminotransferase deaminates tyrosine to produce 4-hydroxyphenylpyruvate and in turn 4-hydroxyphenylpyruvate dioxygenase catalyzes the formation of homogentisate. Eventually, the enzyme homogentisate 1,2-dioxygenase catalyzes the formation of 4-maleylacetoacetate, which is converted into 4-fumarylacetoacetate by maleylacetoacetate isomerase. Subsequently, 4-fumarylacetoacetate produces fumarate and acetoacetate.4,5

A

B

Diet

Glutamate ATP

Cysteine

γ-GCS

ADP+Pi α-Glutamylcysteine Glycine ATP GSHs ADP+Pi

Citrulline Glutamine

Glutathione

Small intestine

Citrulline

Blood

Arginine

C

Arginine

CO2–

Kidney H3

N+

Glutamate

CO2–

Glutamate decarboxylate H3N+

CH2

CO2–

γ-Aminobutyrate (GABA)

New Nitric oxide Proteins Polyamines Amino acids Figure 6 Amino acid as precursors of different molecules. (A) Amino acid synthesis could include an inter-organ collaboration. In the case shown, arginine synthesis starts using glutamate or citruline. Glutamine from the diet is converted on the small intestine to citruline that can be finally converted into arginine in the kidneys. This amino acid can be taken and used by many tissues as a precursor for many biomolecules such as polyamines (for DNA and cell proliferation), nitric oxide (as vasodilator or for cell signaling), general synthesis of proteins in general, and also as precursor for other amino acids. (B) Glutathione is a tripeptide (g-glutamyl–cysteinyl–glycine) composed of glutamate, cysteine, and glycine, with the amino group of cysteine joined in peptide linkage to the g-carboxyl group of glutamate. It is the most important nonenzymatic antioxidant in the body protecting against oxidative stress. (C) a-Aminiobutyrate (GABA) is a key neurotransmitter that is synthesized from the amino acid glutamate by decarboxylation.

14

Amino Acid Metabolism

1.01.5

Important Biomolecules Synthesized From Amino Acids

In addition to providing building blocks for proteins, amino acids are precursors of specialized biomolecules including hormones, nucleotides, coenzymes, cell-wall polymers, porphyrins, neurotransmitters, and pigments. For example, glycine is a precursor of porphyrins, whereas glycine, arginine, and methionine are used to form creatine. Glutamate, glycine, and cysteine are used to form glutathione. Biological amines are products of amino acid decarboxylation including dopamine, norepinephrine (noradrenaline), and epinephrine (adrenaline), which are derived from tyrosine. From a biomedical and physiologic perspective, L-argine is used to produce nitric oxide, one of the most important regulators of blood pressure and smooth muscle contraction. For further information, the reader is referred to Figure 6 and Refs. 4 and 5.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Campbell, N. A.; Reece, J. B. Biology, 6th ed.; Benjamin Cummings, 2002; pp 71–75. Zubay, G. Biochemistry, 3rd ed.; WCB, 1993; pp 47–53. Berg, J.; Tymoczko, J. L.; Stryer, L. Biochemistry, Freeman, 2006. Newsholme, E. A.; Leech, T. R. Functional Biochemistry in Health and Disease, Wiley, 2009. Nelson, D. L.; Cox, M. M. Lehninger Principles of Biochemistry, 4th ed.; Freeman, 2004. Wu, G. Amino Acids: Metabolism, Functions and Nutrition. Amino Acids 2009, 37, 1–17. Devlin, T. M. Textbook of Biochemistry with Clinical Correlations, Wiley, 2006. Voet, D.; Voet, J. G. Amino Acid Metabolism. In Biochemistry, 3rd ed.; Wiley, 2004 (Chapter 26). Berg, J. M.; Tymoczko, J. L.; Stryer, L. Biochemistry, 6th ed.; Freeman, 2006; pp 27–34. Lixiang, C. P. Catabolism of Nutritionally Essential Amino Acids in Developing Porcine Enterocytes. Amino Acids 2009, 37, 143–152.

1.02

Enzyme Biocatalysis

Daryoush Abedi, Isfahan University of Medical Sciences, Isfahan, Iran Lin Zhang, Michael Pyne, and C Perry Chou, University of Waterloo, Waterloo, ON, Canada © 2011 Elsevier B.V. All rights reserved. This is a reprint of D. Abedi, L. Zhang, M. Pyne, C. Perry Chou, 1.03 - Enzyme Biocatalysis, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 15-24.

1.02.1 1.02.2 1.02.3 1.02.4 1.02.5 1.02.6 1.02.7 References

Introduction to Enzymes Enzyme Kinetics Enzyme Engineering Enzyme Production Immobilized Enzymes Enzyme Applications Conclusions

15 16 19 20 22 23 24 24

Glossary Enzyme A protein macromolecule that can drive a specific chemical reaction. Enzyme engineering The application of modifying an enzyme’s molecular structure to improve its properties or to develop novel catalytic activities. Enzyme kinetics The study of the reaction mechanism and reaction rate of an enzyme-catalyzed chemical reaction. Immobilized enzyme An enzyme physically or chemically attached to an inert solid support. Recombinant DNA technology The technology of cutting and pasting DNA fragments in vitro for various practical applications, such as gene cloning and protein production.

1.02.1

Introduction to Enzymes

Enzymes are protein biopolymers formed in all living cells from the 20 natural amino acids. Within a cell, enzymes act as biocatalysts driving numerous chemical reactions and coordinating various cellular functions. In vivo synthesis of an enzyme is mediated through expression of the enzyme-encoded gene, including transcription, translation, and processing steps at the transcriptional and/or translational level. The amino acid sequence of an enzyme determines the uniqueness of its molecular structure and dictates biological activity. With a seemingly infinite number of combinations of amino acids available to construct an enzyme, nature has created an enzyme kingdom with amazingly diversified molecular structures, reaction kinetics, substrate specificities, and biological functions in order to perform and regulate all types of biochemical reactions. Even for catalyzing the same reaction, multiple enzyme forms (often from different biological sources) are available. To take advantage of nature‘s wisdom, one of the major interests in enzymes is to apply their biological activities to catalyze chemical reactions of practical interest (particularly in applications involving industrial manufacturing and environmental processing). Application of enzyme biocatalysis begins with the search of enzymes from various biological sources. Microbial cells are prevalent and adaptive within harsh environments, such as contaminated soil, industrial effluent, and activated sludge, through the evolution of novel genes and/or pathways that promote their survival. As a result, such a rich microbial community offers a novel enzyme/gene bank for biomining. To date, many enzyme (protein) and gene (DNA) databases have been established for online access.1 Typically, recombinant DNA technology can be utilized to clone the gene encoding an enzyme of interest for characterization and large-scale production, particularly when the DNA sequence is available. If necessary, an enzyme can be engineered to modify its molecular structure, reaction kinetics, and catalytic properties to better suit practical applications. During the past few decades, advanced biotechnologies have been developed to search, identify, characterize, evolve, and even tailor enzymes. This has offered a great advantage for exploring enzyme biocatalysis. For the purposes of industrial applications, it is often advantageous to make an enzyme available in an immobilized form, either through the immobilization of the enzyme molecule or through immobilization of the whole cells expressing the enzyme of interest, so that the enzyme becomes readily recoverable and reusable. Such techniques and other relevant methods associated with enzyme biocatalysis are discussed in the following sections.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00003-3

15

16

Enzyme Biocatalysis

1.02.2

Enzyme Kinetics

As with chemical reactions, the mechanism of an enzymatic reaction can be modeled through the use of kinetics. Enzyme kinetics represents a powerful tool to understand the molecular interactions, reaction routes, and relevant metabolic pathways within a biological system. Precise modeling of enzyme kinetics can be critical for optimizing bioreaction performance. Similar to chemical reaction kinetics, enzyme kinetics is formulated by a series of elementary steps describing the molecular interactions among substrates, inhibitors, products, and enzymes. The basic reaction scheme describes the conversion of a single substrate into a product: E

S%P The mechanism encompasses two reversible elementary steps: (1) the binding of a substrate (S) onto a free enzyme (E) to form an enzyme–substrate complex (ES) and (2) the formation and desorption of the product (P) to regenerate the free enzyme: k1

k2

k1

k2

E þ S % ES % P þ E Notice that the overall reaction is reversible as is often the case with many enzymes, for example, isomerases, dehydrogenases, and transaminases. With the two proposed elementary steps, mass balance for the four species (i.e., S, E, ES, and P) can be conducted to formulate the following four ordinary differential equations (ODEs) with their respective initial conditions: d½S ¼ k1 ½S½E þ k1 ½ES dt d½E ¼ k1 ½S½E þ k1 ½ES þ k2 ½ES  k2 ½P½E dt d½ES ¼ k1 ½S½E  k1 ½ES  k2 ½ES þ k2 ½P½E dt d½P ¼ k2 ½ES  k2 ½P½E dt Initial conditions at t ¼ 0 : ½S ¼ ½S0 ; ½E ¼ ½E0 ; ½ES ¼ 0; ½P ¼ 0 The four ODEs can be solved numerically using an ODE solver to obtain the time profiles for the four species and a typical simulation result (Fig. 1). Notably, due to a much lower [E]0 relative to [S]0, the ES and E concentrations remain relatively unchanged after a brief transient phase. Hence, the ES concentration is assumed to be constant throughout the entire reaction period, indicating that the system is under a quasi-steady-state assumption: d½ES ¼ k1 ½S½E  k1 ½ES  k2 ½ES þ k2 ½P½E ¼ 0 dt

5 4.5 4

3 0.5

[S]

2.5

0.45

[E]

0.4

[ES]

2

[P]

1.5

Concentration

Concentration

3.5

1

0.35 0.3 0.25

[E] [ES] [P]

0.2

0.5

0.15

0

0.05

0.1

0

2

4

6 Time

8

10

0 0

0.2

0.4

0.6

0.8

Time

Figure 1 Time profiles for the four species involved in the two-step enzymatic conversion. The numerical simulation is conducted with the following parameters: k1 ¼ 4, k1 ¼ 2, k2 ¼ 4, k2 ¼ 2, [S]0 ¼ 5, [E]0 ¼ 0.5 (arbitrary units). Time profiles for E, ES, and P at the outset of the reaction are amplified.

1

17

Enzyme Biocatalysis

Together with the conservation of enzyme molecules (i.e., [E]0 ¼ [E] þ [ES]), the instantaneous reaction rate is derived as follows: v¼

d½P k1 k2 ½S  k1 k2 ½P ¼ k2 ½ES  k2 ½P½E ¼ ½E dt k1 ½S þ k2 ½P þ k1 þ k2 0

Initially at t ¼ 0: [S] ¼ [S]0, [P] ¼ 0; hence,   k1 k2 ½S0 ½E0 Vm ½S0 d½P 1 1 Km v0 ¼ or ¼ ¼ ¼ þ dt t¼0 k1 ½S0 þ k1 þ k2 ½S0 þ Km v0 Vm Vm ½S0 where Vm ¼ k2 ½E0

and

Km ¼

k1 þ k2 k1

The above expression of v0 is described as the Michaelis–Menten equation, with Km as the Michaelis constant, which was first proposed by Leonor Michaelis and Maude Menten in 1913. In an approach known as rapid equilibrium hypothesis, the substrate conversion step is designated as rate-determining step since the first step of substrate binding is often a fast equilibrium process relative to substrate conversion (implying k1 [ k2). In this case, Km z k1/k1. Notice that the rapid equilibrium hypothesis is simply the quasi-steady-state assumption applied under the condition of a rate-limiting conversion step. However, the quasi-steady-state assumption may not hold true under certain conditions, such as when [E]0 is only slightly less than [S]0. The estimation of the initial rate v0 can potentially result in a significant error since the quasi-steady-state assumption does not hold true during this transient stage. Based on the simulation result in Fig. 1, the reaction rate (i.e., d[P]/dt) does not remain constant at the outset of the reaction (i.e., t < 0.1). Even though the reaction rate remains relatively constant for most of the reaction (i.e., t > 0.1), it is 10% lower than the estimated v0 according to the Michaelis–Menten equation (1.39 vs. 1.54). This estimation error becomes insignificant when the second step is designated to be limiting (i.e., k1, k1 [ k2, k2), which holds true for most enzymatic reactions. As shown in Fig. 2, the reaction rate remains constant throughout the duration of the reaction. Since the overall reaction rate is much lower, it remains constant for a longer period of time, resulting in a much lower estimation error of 2% using the Michaelis–Menten equation (0.0177 vs. 0.0181). Even with this calculation flaw, the Michaelis–Menten equation is regarded as the most popular model for enzyme kinetics. As some enzymes have more than one substrate-binding site, binding of one substrate can potentially affect (either facilitate or inhibit) the binding of other substrates, a mechanism known as allosteric or cooperative binding. In this case, a modified form of the Michaelis–Menten equation is used to correlate v0 and [S]0:   Vm ½Sn0 1 1 Km or ¼ ; where n ¼ cooperativity coefficient þ v0 ¼ n v0 Vm Vm ½Sn0 ½S0 þ Km

5 4.5

[S] [E]

4

[ES] [P]

3

0.5 0.45

2.5

[E] [ES] [P]

0.4

2

0.35 Concentration

Concentration

3.5

1.5 1

0.3 0.25 0.2 0.15

0.5

0.1

0

0.05

0

50

100 Time

150

200

0 0

2

4

6

8

10

Time

Figure 2 Time profiles for the four species involved in the two-step enzymatic conversion, where the second step is limiting. The numerical simulation is conducted with the following parameters: k1 ¼ 4, k1 ¼ 2, k2 ¼ 0.04, k2 ¼ 0.02, [S]0 ¼ 5, [E]0 ¼ 0.5. Time profiles for E, ES, and P at the outset of the reaction are amplified.

18

Enzyme Biocatalysis

Notice that n > 1 represents positive cooperativity with respect to substrate binding, whereas n < 1 denotes negative cooperativity. On the other hand, enzyme activity can be inhibited by certain molecules, known as inhibitors, which bind to the enzyme (E) and/or enzyme–substrate complex (ES) during the reaction. There are three major models (Table 1) describing enzyme kinetics in the presence of an inhibitor (I), that is, (1) competitive inhibition, in which the inhibitor binds to the enzyme only; (2) uncompetitive inhibition, in which the inhibitor binds to the enzyme–substrate complex only; and (3) mixed (or noncompetitive) inhibition, in which the inhibitor binds to both the enzyme and enzyme–substrate complex. All three mechanisms of inhibition can be modeled by the Michaelis–Menten equation with the use of apparent parameters, that is, app

y0 ¼

Vm ½S0 app ½S0 þ Km

Under certain conditions, substrate overbinding to an enzyme can result in loss of enzyme activity, a phenomenon known as substrate inhibition. The uncompetitive inhibition model can be applied in this situation with overbinding portrayed as the binding of an extra substrate molecule to the enzyme–substrate complex:

E+S

k1 k–1

k2 ES + S

k–2

P+E

Ksi ES2 Hence, a modified Michaelis–Menten equation is required in order to correlate v0 and [S]0:   Vm ½S0 ½S0 1 Km 1 y0 ¼ or ¼ þ þ 2 ½S v0 Vm ½S0 Vm Vm Ksi Km þ ½S0 þ Ksi0 Many enzymatic reactions, particularly organic syntheses, involve more than one substrate and thus must be modeled using multisubstrate enzyme kinetics. Bi Bi reactions in which two substrates (i.e., A and B) and two products (i.e., P and Q) are involved are quite common: E A þ B%P þ Q There are two major types of Bi Bi reaction mechanisms. Those in which both substrates must combine with an enzyme before a reaction occurs are known as sequential reactions. Sequential reactions are subclassified into ordered and random mechanisms depending on whether the substrate binding follows a specific reaction sequence. The corresponding mechanisms are described

Table 1

One-substrate reaction kinetics involving inhibition

Inhibition type

No

E+S E+S +

k1 k–1 k1 k–1

ES ES

k2 k2

P+E

Kmapp

Vm [S]0 Km + [S]0

Vm

Km

Vm [S]0 + [S]0

Vm

αKm

Vm [S]0 Km + αʹ [S] 0

Vm

Km

αʹ

αʹ

Vm [S]0 + αʹ [S] 0

Vm

αKm

αʹ

αʹ

P+E

—

Competitive

Vmapp

v0

Reaction

αKm

Ki EI E+S

k1 k–1

ES

k2

P+E

+ —

Uncompetitive

Kʹ i ESI E+S

—

ES P+E k–1 k2 +

—

Mixed

+

k1

Ki

Kʹ i

EI

αKm

ESI

a ¼ 1 þ ð½I=Ki Þ; a0 ¼ 1 þ ð½I=Ki0 Þ and Ki ¼ ½E½I=½EI represent dissociation constants for the competitive and uncompetitive inhibition reactions, respectively. Ki0 ¼ ½ES½I=½SEI.

Enzyme Biocatalysis

19

below. Enzymatic reaction is represented by a horizontal line, whereas substrate binding and product release are denoted by vertical arrows with respective rate constants. Different enzyme species are specified under the horizontal line: A

P

B

k1 k–1 k2 k–2

k–4

Ordered Bi Bi reaction

Q k4

k–5

k5

k3 EA

E

EAB

A

EPQ

k–3

EQ

P

B

k1 k–1 k2 k–2

E

Q

k–4 k4 k–5 k5

EA

EQ

Random Bi Bi reaction

k3

E EB k2 k–1 k1

k–2 B

EAB

E

EPQ

k–3

k4

A

EP k–4 k5 k–5

P

Q

On the other hand, mechanisms in which different substrates bind to different enzyme species are known as ping pong reactions. In this case, the free enzyme has two forms denoted E and F, with a respective binding specificity to A and B: A

P

k1 k–1 Ping pong Bi Bi reaction E

EA

k–3 k2 k–2

FP

B k3

k4

F

Q k–4

FB

k–6 k5 k–5

EQ

k6

E

Similar to an enzyme reaction with one substrate, a set of ODEs based on mass balance for each species can be formulated and solved to yield the exact solution for the time profiles of species concentrations. According to the rapid equilibrium hypothesis, the initial reaction rate for various two-substrate mechanisms can be expressed in a similar form to the Michaelis–Menten equation.2 Other complex enzyme mechanisms involving more than two substrates and enzyme species can also be developed. For example, iso-mechanisms in which multiple substrates and enzyme conformations are involved can be modeled with hybrid mechanisms, such as ping pong-ordered, ping pong-random, or iso-ping pong reactions.3

1.02.3

Enzyme Engineering

Natural enzymes are often not optimal for practical applications. To better suit practical challenges, enzyme properties, such as environmental tolerance (e.g., to suboptimal temperature and pH), isoelectric point, substrate specificity, reaction mechanism, and molecular stability, need to be modified, improved, or tailored.4 Enzyme engineering is a powerful tool enabling these approaches primarily based on the optimization of amino acid sequence. Currently, it is still technically difficult to design an artificial enzyme in vitro. Hence, a natural enzyme close to the final target form has to be selected first for engineering. The modified amino acid sequence can potentially lead to a minor or major structural change, which subsequently elicits desirable properties. The novel catalytic properties derived herewith can be introduced into biological cells, particularly microorganisms, for genetic and metabolic engineering purposes.5 A typical enzymatic reaction has three major steps occurring at the active site of the enzyme, namely substrate binding, biochemical conversion, and product release. The efficiency of each step can be potentially affected by the amino acids involved in the catalytic mechanism. Theoretically, these amino acids can be identified if the enzyme‘s three-dimensional structure is available whereupon they become rational targets for manipulation. The 20 natural amino acids offer a variety of molecular properties, such as size, charge, acidity (or alkalinity), hydrophilicity (or hydrophobicity), nucleophilicity, and enantioselectivity. Strategies in manipulating amino acid residues near the active site can change the local enzyme structure and reaction environment and subsequently affect the specificity, binding, and selectivity of the substrate. On the other hand, the reaction chemistry can be engineered based on the amino acid manipulation given a general observation that proteins with a similar fold but different catalytic groups can catalyze very different chemical reactions. The catalytic groups can be engineered to possibly develop a new catalytic mechanism, recognize a new substrate (hence, the conversion to a new product), increase the turnover number, and reduce the activation energy. While the amino acids surrounding the active site appear to be legitimate targets for enzyme engineering, those structurally remote from the active site can sometimes impose a significant impact on the catalytic reaction and therefore become hot spots

20

Enzyme Biocatalysis

for manipulation. It is difficult not only to identify these amino acids but also to predict the combinatorial effect associated with the manipulation of multiple hot spots. Technically, searching key amino acids can be performed by constructing a mutant gene library through random mutagenesis followed by screening desirable phenotypes among the derived mutants. There are many standard protocols based on error-prone polymerase chain reaction (ep-PCR) developed for in vitro random mutagenesis.6,7 Shuffling homologous genes is an alternative approach for library construction.8 The key issue is to ensure the generation of a large population of mutant variants for screening and characterization of desirable phenotypes. On the other hand, library screening is rather enzyme dependent and requires a special design. Usually, the screening is conducted when mutant variants are expressed in the bacterium Escherichia coli or via phage display. It is hoped that certain expressed mutant variants can exhibit a distinctive phenotype allowing easy screening. Developments in high-throughput screening and protein assay technologies significantly improve the screening efficiency.9 Successful derivation of desirable mutants will enable the identification of amino acids contributing to phenotypical changes. Site-directed mutagenesis of these key amino acids can be explored to understand the structural and kinetic effects associated with these mutations. Multiple rounds of mutagenesis and screening can be performed to accumulate multiple mutations synergistically leading to a major phenotypical change, a concept called directed evolution.10 Such techniques have resulted in the construction of many man-made enzymes (or genetically modified microorganisms) with novel bioactivities that are practically useful. While directed evolution has been commonly applied for enzyme engineering, it is considered resource intensive and time consuming particularly in light of the requirement for high-throughput screening. Given the availability of three-dimensional protein structures and extensive information associated with the kinetic roles that various structural components play, tailoring a new enzyme based on the structural and kinetic knowledge has become quite feasible. For example, molecular assembling is an approach to graft catalytic machineries, such as functional amino acids forming an active site, into a selectively designed molecular template.11 Techniques in computational bioinformatics, molecular modeling, and multivariate statistics have been applied to understand the protein sequence–function relationship as well as rapidly discover new types of catalytic function.12 These approaches along with directed evolution can form an efficient platform to design protein molecules with improved enzyme properties or novel catalytic activities for industrial applications.

1.02.4

Enzyme Production

Prior to the advent of modern biotechnology, enzymes were obtained primarily via extraction from natural biological sources, such as plant tissues or exudates, animal organs, and microorganisms. With the development of recombinant DNA technology, enzyme production has entered a new era. Theoretically, enzymes from any biological source can be produced using an appropriate gene expression system. Microbial cells (in particular the Gram-negative bacterium E. coli) are commonly adopted as the expression host for key enzyme-encoded genes. As a result, currently over 60% of commercial enzymes are produced using recombinant DNA technology.13 Recombinant DNA technology offers a molecular tool to clone a gene of interest for expression within the same host (homologous expression) or within a different host (heterologous expression). In addition, various high copy number expression vectors are available for gene overexpression. Once constructed, the expression vector containing the gene of interest is delivered into the host cell to generate the recombinant strain, which is then propagated under optimal enzyme production conditions in a well-controlled bioreactor. The resulting cells are harvested and the recombinant enzyme is extracted and purified. Theoretically, an enzyme produced using recombinant DNA technology should have the same molecular structure and catalytic properties as the native one. The enzyme yield and quality for the overall production process can be enhanced with appropriate bioprocessing strategies. If necessary, the enzyme properties, such as substrate specificity and enzyme stability, can be modified or genetically tailored as discussed above. A typical bioprocess for recombinant enzyme production consists of three technical stages, that is, upstream for construction of the host/vector system, midstream for cultivation, and downstream for protein harvesting and purification (Fig. 3). Strategies have been developed both at the molecular and bioprocess levels in order to enhance recombinant protein production in the E. coli expression system, as wild-type E. coli strains are typically not optimized for industrial applications. Developing a bioprocess for enzyme production starts with the construction of the host/vector system in order to optimize the production strain with respect to expression of the recombinant gene(s). Recombinant enzyme production can be limited by in vivo gene expression (including replication, transcription, posttranscriptional processing, translation, and posttranslational processing), cultivation, or downstream processing. These limitations can be overcome or eliminated with an appropriate design of the expression system. For example, the use of a strong promoter system for the regulation of gene expression not only increases the protein yield but also enables two-stage cultivation for mitigating potential physiological impact associated with the heterologously expressed protein.14 Also, the use of a protein fusion tag not only facilitates protein purification but also enhances gene expression.15 Certain genetically modified E. coli strains secrete less growth-inhibiting metabolites, such as acetate, and can therefore outperform wild-type strains for recombinant protein production, in particular during high cell density cultivation (HCDC).16 Hence, genetic manipulation of the host/vector system is considered the most effective and economic approach to enhance recombinant protein production in terms of mediating functional expression, increasing cultivation performance, and facilitating downstream purification. During cultivation of recombinant cells within a bioreactor, the gene of interest is heterologously expressed in the exponential phase of cell growth. Both high-level gene expression and HCDC must be carried out simultaneously in order to optimize culture

Enzyme Biocatalysis

Construction of host/vector system

Upstream

Cultivation for gene expression

Midstream

Intracellular enzyme

Extracellular enzyme

Cell harvest

Extracellular medium harvest

21

Cell lysis

Cell debris removal

Concentration of medium

Affinity chromatography

Precipitation of total protein

Downstream

Chromatography purification

Final product formulation

Figure 3

General bioprocess scheme for enzyme production.

performance. Accordingly, strategies have been developed on the basis of boosting cell growth, alleviating physiological deterioration, and enhancing gene expression. Fed-batch operation is often used for HCDC with the medium recipe and feeding profile being developed in accordance to sustain cell growth and gene expression. Medium overfeeding can result in the accumulation of toxic metabolites and inhibition of cell growth during cultivation. Tuning and control of cultivation parameters, such as pH, temperature, dissolved oxygen, and feeding profile, can substantially affect culture performance, particularly during large-scale production. In addition, HCDC is often limited by oxygen transfer, which results in local anaerobiosis and is especially problematic during large-scale operations. Following cultivation, cells (or cell-free medium) are harvested for the purification of intracellular (or extracellular) enzyme product (Fig. 3). Downstream processing for enzyme purification begins with various pretreatment steps, including cell lysis, removal of cell debris, and recovery of total protein, in order to obtain a concentrated crude protein extract. Most of these steps require high-volume solid–liquid separation, which can be performed by centrifugation or filtration. Cell lysis is typically achieved chemically (e.g., by addition of alkaline), enzymatically (e.g., by treatment with lysozyme), or mechanically (e.g., by sonication or passage through a French press). On the other hand, enzyme purification is predominantly accomplished by chromatography. Depending on the interactive force between the enzyme and the functional group of the chromatographic media, various chromatographies can be applied, including affinity, ion-exchange, reversed phase, and size exclusion, under ambient or high-pressure conditions. While solid beads packed in a column are commonly used as the media for chromatographic operation (i.e., column chromatography), the use of polymeric membranes (i.e., membrane chromatography) has become prevalent recently due to a much higher volumetric binding capacity.17 Several protein fusion tags with their corresponding chromatographic media have been developed for facilitating enzyme harvest and purification.15 It is anticipated that enzyme activity is retained when fused with a protein fusion tag. If necessary, the enzyme moiety can be released with a protease which recognizes the specific protein sequence joining the enzyme and fusion tag. Fusion tag cleavage can be conducted simultaneously with fusion protein binding to the chromatography media in order to facilitate separation of the enzyme and fusion tag, an approach known as on-column cleavage.18 A final polishing step is usually implemented to remove trace contaminant proteins during the late purification stage. Following purification, the enzyme product is formulated for storage. In addition, lyophilization is an alternative approach that can be utilized for long-term enzyme storage. Lastly, several chemical supplements are commonly used to enhance enzyme stability.19

22

Enzyme Biocatalysis

1.02.5

Immobilized Enzymes

Enzymes are widely used in industrial processes due to many technical advantages, such as high specificity, mild reaction conditions, and extensive biocompatibility. However, enzymes also possess particular characteristics that prevent their practicality within certain applications. Specifically, tedious isolation and purification procedures, structural instability, sensitivity to slightly suboptimal process conditions (e.g., pH, temperature, trace elements, and ionic strength), and high solubility in aqueous medium are the major hurdles that hinder efficient enzyme recovery. Enzyme immobilization is generally the preferred method to resolve such inherent problems. Immobilized biocatalysts are enzymes or enzyme-containing cells that are physically or chemically confined within a solid support while preserving maximal biological activity. As a result, immobilized enzymes offer a heterogeneous catalytic system for continuous and repeated operation that is more robust and cost effective. To ensure preservation of bioactivity, the functional group(s) within the enzyme active site must not be involved in the immobilization reaction and the tertiary structure of the enzyme should be minimally disturbed upon immobilization. Enzyme immobilization offers several technical advantages compared to traditional suspended enzyme methods. First, immobilized enzymes are insoluble and can be easily recovered for repeated use in both batch and continuous reactors. Furthermore, process control becomes more simplified with an immobilized catalytic system. Lastly, immobilization provides a biocatalyst with enhanced structural, thermal, and pH stability and, as a result, extends the operation lifetime and shelf-life of the enzyme. Based on the interactive force linking an enzyme to its solid support, enzyme immobilization can be performed physically (via adsorption and entrapment) or chemically (via cross-linking and carrier binding) (Fig. 4). Enzyme adsorption involves the adhesion of an enzyme to the active surface of an adsorbent, such as carbon, ionic-exchange resins, celluloses, and clays. While the immobilization reaction is simple and unlikely to affect enzyme activity, enzyme leakage is a common occurrence due to weak bonding. Immobilization yield can also be quite low since the enzyme is fully exposed and can be sensitive to the reaction environment. Enzyme entrapment is an alternative physical immobilization method typically achieved through gel entrapment or microencapsulation of the biocatalyst. Entrapped enzymes offer physical protection against the reaction environment but still suffer from occasional enzyme leakage. Specifically, gel entrapment is achieved by the formation of a polymer network in the presence of an enzyme solution. Polymeric materials, such as polyacrylamide, gelatin, and alginate, are commonly employed but can potentially affect the physical, chemical, and kinetic properties of the immobilized biocatalyst. Hence, the entrapment matrix should be carefully selected to optimize the immobilization and reaction efficiency while minimizing

Immobilization methods

Chemical methods

Physical methods

Adsorption + + + + + + + + + +

Carrier binding Cross-linking

Carrier cross-linking

CLEs

Figure 4

Carrier-free cross-linking

CLEAs

CLECs

Various methods used for enzyme immobilization.

Gel entrapment

CLSDs

Entrapment

Fiber entrapment

Microencapsulation

Enzyme Biocatalysis

23

negative effects on the enzyme. Yet another method involves confining an enzyme within a semipermeable microcapsule. Liquid microencapsulation entails an organic polymer solution mixing with an aqueous enzyme solution in the presence of a surfactant. A polymer membrane is formed at the liquid–liquid interface and the aqueous phase containing the enzyme is entrapped within the membrane. Accordingly, solid microencapsulation involves the entrapment of an enzyme within a solid membrane, hollow fiber, or nanostructure such as silica sol–gel. Chemical immobilization occurs through the formation of covalent bonds between certain enzyme amino acid residues and the matrix support. While covalent bond formation can prevent enzyme leakage from a biocatalyst matrix, enzyme activity can be affected as the immobilization reaction sometimes results in significant conformational changes in the enzyme structure. In general, various amino acid side chains (e.g., the guanidinyl group of arginine and the carboxyl groups of glutamate and aspartate) are chemically active with respect to the immobilization reaction. Synthetic functional groups can also be included in an enzyme molecule for mediating molecular affinity during the immobilization process. If necessary, an enzyme can be engineered to generate desirable functional groups to be used for immobilization. A common method for enzyme immobilization involves chemical attachment of enzyme molecules to a water-insoluble support. Carriers can be made from inorganic materials (e.g., silica gel and alumina), natural organic materials (e.g., proteins (albumin and collagen) and carbohydrates (alginate, chitosan, and cellulose)), and synthetic organic materials (e.g., polystyrene, polyacrylamide, and polypropylene). Smart polymers (e.g., poly-N-isopropylacrylamide) are novel carriers of particular importance because a dramatic conformational change can be induced by a minor perturbation in environmental conditions, such as pH, temperature, or ionic strength.20 The carrier properties, primarily determined by the characteristics of the functional groups, can affect the efficiency of immobilization, the yield of the biocatalytic reaction, and the stability of the biocatalyst. Other properties, such as size, permeability, surface area, mechanical stability, affinity to proteins, toxicity, and biocompatibility, are also important. With the presence of functional groups on both enzyme and carrier, chemical bonds between enzyme molecules (carrier-free cross-linking methods) or between enzyme and carrier molecules (carrier cross-linking methods) can be generated. Among various multifunctional cross-linking agents, glutaraldehyde is the most commonly used due to its low cost, high efficiency, and superior stability. The carbonyl group of glutaraldehyde is the functional group involved in the immobilization reaction, which occurs rapidly even at room temperature. Because the linkages are irreversible and relatively strong, the cross-linking system is rather stable against environmental challenges. However, the activity of a chemically immobilized enzyme is generally lower due to the occurrence of side reactions during cross-linking and/or by diffusion limitation within the immobilized system. In addition, the carrier must be selected carefully as it can drastically affect the kinetic behavior of the immobilized system. Immobilization efficiency can be affected by several factors, including the nature and concentration of the enzyme, pH, ionic strength, temperature, and the cross-linking agent utilized. For carrier-free cross-linking methods, the immobilized system can be established by direct cross-linking of different enzyme forms. Soluble enzymes are used to form cross-linked enzymes (CLEs), which can suffer from dramatic loss of enzyme activity and poor mechanical stability. Crystalline enzymes are used to form cross-linked crystalline enzymes (CLCEs), which are capable of preserving a higher level of enzyme activity through increased stability against environmental challenges. Insoluble enzyme aggregates, prepared by mixing suitable aggregating agents with enzyme solutions, have also been explored for cross-linking to generate cross-linked enzyme aggregates (CLEAs), which exhibit similar activity and stability to CLECs. Finally, spray-dried enzymes can be cross-linked to form cross-linked spray-dried enzymes (CLSDEs). However, this method is less common since the enzyme is liable to significant deactivation during the spray-drying process.21

1.02.6

Enzyme Applications

Due to several process advantages, including mild reaction conditions, high substrate specificity, and environmentally friendly processing, enzymes have been utilized extensively in the chemical, food, pharmaceutical, agricultural, and fuel industries primarily for manufacturing purposes. As a result, such enzymes are collectively referred to as industrial enzymes and comprise the main focus of this section. In the washing industry, enzymes possessing the ability to degrade various food products, such as proteases (proteins), amylases (starch), and lipases (oils), can be combined with detergents in order to improve washing performance. These detergent enzymes are able to function at low temperatures (hence, reducing energy consumption), reduce dependence on soap, and pose less of a threat to the environment. Similarly, enzymes used in the textile and leather industries have proved to be cost effective, time saving, and environmentally friendly. For example, amylase, pectinase, and glucose oxidase are used together for bleaching and dyeing within the cotton industry. Finer finishing can be attained through the use of cellulases and lactases in denim processing to generate a desirable stone washed appearance. Likewise, cutinases can be adopted for surface modification of synthetic fibers, while lipases and proteases are commonly used to scour and soften wool products within the textile industry as well as to remove unnecessary components from raw materials in the leather industry. Enzymes also find widespread use within the food industry. For example, starch can be enzymatically hydrolyzed to generate other forms of sugar. Glucose can be released from starch through the action of amylases and can be further transformed into fructose via glucose isomerases. Furthermore, complex proteins can be hydrolyzed by proteases and peptidases that are used to enhance the solubilization and digestibility of protein-containing foods. Pectinases find use within the food and beverage industry through the clarification of juices and other beverages. Regarding the fuel industry, several enzymes are involved in the production of clean and renewable biofuels. For example, biodiesel is composed of fatty acid alkyl esters produced by transesterification of oils with short-chain alcohols (e.g., methanol), a process dependent upon

24

Enzyme Biocatalysis

lipases. Additionally, one of the most famous enzyme applications in the pharmaceutical industry involves the production of 6-aminopenicillanic acid and other semisynthetic b-lactam antibiotics through the use of penicillin acylase. While biocatalysts based upon immobilization of individual enzymes and whole cells are used to catalyze chemical reactions in many industrial applications, living cells can also be adopted directly to act as a cell factory for in vivo biosynthesis due to the presence of a desired synthesis pathway(s). Cells can be genetically manipulated by implementing heterologous genes to catalyze unconventional reactions that otherwise cannot be driven and/or by knocking out certain enzyme-encoding genes to block the corresponding biosynthesis pathways, an approach known as metabolic engineering. These genetically engineered cells offer a powerful tool for the production of many high-valued metabolites, such as alcohols, organic acids, vitamins, nucleosides, and amino acids.22 If necessary, the approaches of immobilized biocatalyst and cell factory can be combined. For example, vanillin, largely used in the food industry, can be synthesized via a two-stage process, that is, fermentation of a metabolically engineered E. coli to convert the glucose feedstock to vanillic acid, followed by subsequent transformation to vanillin catalyzed by an immobilized aryl aldehyde dehydrogenase.23

1.02.7

Conclusions

Recombinant DNA technology has offered a complete solution for cloning and expressing a gene of interest. Bioprocessing strategies for fermentation and downstream processing have been well developed during the past few decades to the point where mass-producing an enzyme of interest becomes routine. While novel carriers are still being discovered and immobilization efficiencies often require improvement, immobilization techniques are considered rather mature. Innovative applications in enzyme biocatalysis have been widely extended to various aspects and activities of human life. To take full advantage of this technology, more novel enzymes have to be identified from nature and manipulated within the laboratory. With mature techniques in enzymology, protein engineering, bioinformatics, computational biology, and structural biology currently being developed, it is possible that enzymes can be artificially designed and tailored from scratch to perfectly meet all technical requirements.

References 1. Bansal, A. Bioinformatics in Microbial Biotechnology – A Mini Review. Microb. Cell Fact. 2005, 4, 19. 2. Leskovac, V., Ed.; Comprehensive Enzyme Kinetics, Kluwer Academic/Plenum Publishers: New York, 2003. 3. Bisswanger, H. Enzyme Kinetics. In Enzyme Kinetics. Principles and Methods; Bisswanger, H., Ed., 2nd ed.; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2008; pp 136–138. 4. Luetz, S.; Giver, L.; Lalonde, J. Engineered Enzymes for Chemical Production. Biotechnol. Bioeng. 2008, 101, 647–653. 5. Bernhardt, P.; O‘Connor, S. E. Opportunities for Enzyme Engineering in Natural Product Biosynthesis. Curr. Opin. Chem. Biol. 2009, 13, 35–42. 6. Lanio, T.; Jeltsch, A. PCR-based Random Mutagenesis Method Using Spiked Oligonucleotides to Randomize Selected Parts of a Gene without Any Wild-type Background. Biotechniques 1998, 25, 958–965. 7. Parikh, A.; Guengerich, F. P. Random Mutagenesis by Whole-plasmid PCR Amplification. Biotechniques 1998, 24, 428–431. 8. Stemmer, W. DNA Shuffling by Random Fragmentation and Reassembly: In Vitro Recombination for Molecular Evolution. Proceedings of the National Academy of Sciences of the United States of America 1994, 91, 10747–10751. 9. Becker, S.; Michalczyk, A.; Wilhelm, S.; et al. Ultrahigh-throughput Screening to Identify E. coli Cells Expressing Functionally Active Enzymes on Their Surface. Chembiochem 2007, 8, 943–949. 10. Kaur, J.; Sharma, R. Directed Evolution: An Approach to Engineer Enzymes. Crit. Rev. Biotechnol. 2006, 26, 165–199. 11. Cedrone, F.; Menez, A.; Quemeneur, E. Tailoring New Enzyme Functions by Rational Redesign. Curr. Opin. Struct. Biol. 2000, 10, 405–410. 12. Bannen, R. M.; Suresh, V.; Phillips, G. N.; et al. Optimal Design of Thermally Stable Proteins. Bioinformatics 2008, 24, 2339–2343. 13. Demain, A. L.; Vaishnav, P. Production of Recombinant Proteins by Microbes and Higher Organisms. Biotechnol. Adv. 2009, 27, 297–306. 14. Lee, J.; Parulekar, S. J. Enhanced Production of Alpha-amylase in Fed-batch Cultures of Bacillus Subtilis Tn106 [Pat5]. Biotechnol. Bioeng. 1993, 42, 1142–1150. 15. Arnau, J.; Lauritzen, C.; Petersen, G. E.; Pedersen, J. Current Strategies for the Use of Affinity Tags and Tag Removal for the Purification of Recombinant Proteins. Protein Expr. Purif. 2006, 48, 1–13. 16. Chou, C. P. Engineering Cell Physiology to Enhance Recombinant Protein Production in Escherichia coli. Appl. Microbiol. Biotechnol. 2007, 76, 521–532. 17. Przybycien, T. M.; Pujar, N. S.; Steele, L. M. Alternative Bioseparation Operations: Life beyond Packed-bed Chromatography. Curr. Opin. Biotechnol. 2004, 15, 469–478. 18. Dian, C.; Eshaghi, S.; Urbig, T.; et al. Strategies for the Purification and On-column Cleavage of Glutathione-S-transferase Fusion Target Proteins. J. Chromatogr. B 2002, 769, 133–144. 19. Iyer, P. V.; Ananthanarayan, L. Enzyme Stability and Stabilization – Aqueous and Non-aqueous Environment. Process Biochem. 2008, 43, 1019–1032. 20. Sheldon, R. A. Enzyme Immobilization: The Quest for Optimum Performance. Adv. Synth. Catal. 2007, 349, 1289–1307. 21. Cao, L. Q.; van Langen, L.; Sheldon, R. A. Immobilised Enzymes: Carrier-bound or Carrier-free? Curr. Opin. Biotechnol. 2003, 14, 387–394. 22. Soetaert, W.; Vandamme, E. The Impact of Industrial Biotechnology. Biotechnol. J. 2006, 1, 756–769. 23. Liese, A.; Villela, M. Production of Fine Chemicals Using Biocatalysis. Curr. Opin. Biotechnol. 1999, 10, 595–603.

1.03

Immobilized Biocatalysts

Andres Illanes, Pontificia Universidad Católica de Chile, Valparaíso, Chile © 2011 Elsevier B.V. All rights reserved. This is a reprint of A. Illanes, 1.04 - Immobilized Biocatalysts, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 25-39.

1.03.1 1.03.1.1 1.03.1.2 1.03.2 1.03.2.1 1.03.2.2 1.03.3 1.03.3.1 1.03.3.1.1 1.03.3.1.2 1.03.3.2 1.03.3.2.1 1.03.3.2.2 1.03.4 1.03.4.1 1.03.4.2 1.03.4.3 1.03.5 1.03.5.1 1.03.5.2 1.03.5.3 1.03.5.4 1.03.6 1.03.6.1 1.03.6.2 1.03.6.2.1 1.03.6.2.2 1.03.7 References

1.03.1

Introduction: Definitions and Scope Immobilized Cells Immobilized Enzymes Applications of Immobilized Enzymes Application of Immobilized Enzymes as Industrial Catalysts Other Applications of Immobilized Enzymes Methods of Enzyme Immobilization Immobilization by Chemical Interaction Carrier-Bound Immobilized Enzymes Carrier Free Immobilized Enzymes Immobilization by Containment Entrapment Membrane Retention Properties of Immobilized Enzymes Activity Stability Selectivity and Specificity Evaluation of Enzyme Immobilization Immobilization Yields Specific Activity of Immobilized Enzyme Biocatalyst Stability Optimization of Enzyme Immobilization Heterogeneous Biocatalysis Definitions Mass Transfer Effects External Diffusional Restrictions Internal Diffusional Restrictions Future Prospects for Immobilized Biocatalysts

25 26 26 26 26 27 28 28 28 29 30 30 30 31 31 31 31 32 32 33 33 33 34 34 34 34 36 38 38

Introduction: Definitions and Scope

In broad terms, biocatalysts are biological entities capable of catalyzing chemical reactions, being those isolated enzymes or any other biological superstructure containing them. Strictly speaking, biocatalysts are enzymes because they are the functional proteins that catalyze all reactions of cell metabolism. From a technological perspective, a biocatalyst is a catalyst of biological nature able to perform a useful chemical reaction under controlled conditions. The use of such biological entities as process catalysts usually requires them to withstand harsh conditions far different from physiological. The active principles, whether enzymes or whole cells, can be conveniently structured to perform as process catalysts by modifying them; so, most biocatalysts are actually quite different from the biological entities from which they derive. A crucial issue in biocatalysis is to transform these biological catalysts into robust process catalysts. Biocatalyst stabilization has been attempted by several strategies including chemical modification, immobilization to solid matrices, crystallization, and aggregation. Complementary to them have been the use of modern techniques of protein engineering, namely site-directed mutagenesis, directed evolution by tandem mutagenesis, and gene shuffling, and also the screening of intrinsically stable biocatalysts from novel microbial strains and also from environmental metagenome. A review on the subject has been published by Bommarius and Polizzi.1 Immobilization can be considered as the most relevant strategy for improving biocatalyst performance under process conditions. This article is mainly focused to immobilized enzymes as process catalysts, but comments will be made, when appropriate, to immobilized cells and other potential applications of immobilized biocatalysts. Immobilized biocatalysts have been defined as those physically confined or localized in a certain space. A more precise definition of an immobilized biocatalyst is one which has been attached to a solid matrix or contained within it. Matrices are usually inert

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00004-5

25

26

Immobilized Biocatalysts

but may be active as it occurs in the case of enzymes or cell aggregates and enzyme crystals where the enzyme protein itself constitutes the matrix.

1.03.1.1

Immobilized Cells

Under this designation, different types of biocatalysts can be identified, according to the physiological state of the cells. Metabolically active cells can be immobilized to perform in a medium allowing cell growth; also, metabolically intact immobilized cells can be used in a resting medium to perform a given bioconversion and also damaged cells that preserve the functionality of some of their enzymes can be used in immobilized form to catalyze a specific reaction. The first case represents an alternative to conventional fermentation, offering some interesting potential advantages such as continuous operation at high dilution rates, flexible reactor configuration, reduction of nonproductive growth phases in the case of producing non-growth-associated metabolites, and increased cell density leading to higher volumetric productivity. Worthwhile to mention is the production of organic acids (e.g., citric and lactic acids), alcohols (e.g., ethanol, sorbitol, glycerol, propanediol, mannitol, and xylitol), antibiotics (e.g., penicillin G and oxytetracycline), and several microbial enzymes. Bioconversions with resting cells can also be improved by immobilization, as illustrated by the biotransformation of steroids to produce valuable pharmaceutical compounds such as steroid hormones. Immobilization of damaged (fixed) cells is an alternative to enzyme immobilization when the intracellular enzyme is labile or hard to extract, or when coenzyme regeneration is required. This approach does not differ significantly from immobilization of isolated enzymes so that it will be dealt within the context of enzyme immobilization. More information on immobilized cells has been reported by Chibata et al.2

1.03.1.2

Immobilized Enzymes

It refers to the immobilization of enzyme proteins that have been isolated from the cells producing them. Extracellular enzymes produced by fermentation are in many cases pure enough to be immobilized, because the cell membrane acts as a selective barrier and not many proteins or other intracellular materials are excreted; in this case, the enzyme in the clarified spent fermentation liquor may be too diluted so concentration, and often removal of salts and other low-molecular-weight compounds, is required before immobilization. Even though extracellular enzymes are robust and structurally well conditioned to perform in a nonphysiological environment, immobilization can significantly increase their stability under use conditions. Intracellular or cell-bound enzymes require to be released from the cells to obtain a crude extract containing the desired enzyme that is heavily contaminated with cell proteins and other macromolecular cell components. In this case, purification is required before immobilization not only to remove unwanted contaminants but also to increase the specific activity of the protein solution that will be subjected to immobilization. According to its definition, immobilized enzymes include those physically trapped in a membrane surrounding space as well as enzyme aggregates and enzymes linked to or contained within a solid matrix. Most immobilization methods produce an insoluble biocatalyst and actually most are aimed to generate a heterogeneous system in which the biocatalyst is easily separated from the reaction medium. However, immobilization and insolubilization are different concepts; moreover, enzymes are increasingly being used in nonconventional media, where water solubility has very little meaning and enzyme proteins can be intrinsically insoluble. Most enzymes within the living cell are attached to interfaces or contained within solid-stage assemblages. Even cytoplasmic enzymes are within an environment that more closely resembles a gel than a diluted aqueous solution. Therefore, immobilized enzymes are, to a certain extent, a more close representation of its physiological stage; despite this, the fundamental principles of enzyme kinetics were developed based on their behavior in dilute aqueous solutions, which is an aspect to be considered when analyzing immobilized enzyme kinetics.

1.03.2

Applications of Immobilized Enzymes

Immobilized enzymes are endowed with excellent properties as catalysts, so an ever-increasing number of applications have been developed through the years. They have a great potential as industrial catalysts and also in other areas such as clinical and chemical analysis, biosensors, biomedicine, and research.

1.03.2.1

Application of Immobilized Enzymes as Industrial Catalysts

Soon after the concept of immobilized enzyme was established in the 1960s, its technological potential was realized. By the end of that decade, the large-scale production of L-amino acids for human and animal nutrition had been developed in Japan using aminoacylase from Aspergillus oryzae immobilized into DEAE-Sephadex. This process has the historical record of being the first large-scale process conducted with an immobilized enzyme. Immobilized aminoacylase was used in continuously operated packed-bed reactors, being the carrier recoverable after enzyme exhaustion. Savings in production costs were as high as 40% by using the enzyme in immobilized form because of reduction in enzyme cost and also in labor; the process was promptly established and is still in operation. By the mid-1970s, the production of high-fructose syrups (HFSs) with immobilized glucose isomerase acquired industrial significance and soon developed to a multimillion ton business with production plants in many places all over the world. The technology was easily adopted by the cornstarch industry that traditionally used enzymatic processes for starch liquefaction and

Immobilized Biocatalysts

27

saccharification with a-amylase and glucoamylase; however, glucose isomerase was a more expensive and labile enzyme and the technology was not established until robust immobilized biocatalysts were developed. Annual production of HFS from cornstarch in the USA alone is estimated to be close to 10 million tons, representing about 40% of the caloric sweetener market, and keeps growing as sugar is progressively being replaced by HFS in soft drinks and many other industrial products. This very successful process represents a paradigm for enzyme immobilization. Another successful large-scale process was developed in the early 1980s: the production of 6-aminopenicillanic acid (6APA) by hydrolysis of penicillin G or V with immobilized penicillin acylase, this b-lactam nucleus being the precursor of many of the derived semisynthetic penicillins of pharmacological significance. Penicillin acylase was discovered in 1960, but the enzyme-catalyzed process for 6APA production could not compete in terms of yield and productivity with the chemical Delft cleavage process that was soon established. Improvement in stability by screening and recombinant DNA technology, but mainly by immobilization, allowed the gradual replacement of the Delft process that has been now swept out from the market. b-Lactam antibiotics, mainly penicillin and cephalosporins, represent more than 60% of the total antibiotic market with annual sales estimated in US$15 billion. Penicillin acylase is used for producing not only 6APA but also the cephalosporinic nucleus 7-amino-3-desacetoxi-cephalosporanic acid from cephalosporin G, as intermediate for the production of semisynthetic cephalosporins. Annual consumption of penicillin acylase is estimated around 20 million tons and is being massively produced not only in Western but also in Eastern countries such as India, Korea, and China. Market for this enzyme is increasing because of its use in the reactions of synthesis of semisynthetic penicillins and cephalosporins as well. In fact, under proper conditions, penicillin acylase can catalyze reverse reactions of synthesis. In this case, it has been hard to displace the chemical processes currently in use, but advances in biocatalyst and process engineering, together with more stringent environmental regulations, have paved the way for a gradual replacement of existing chemical processes, as illustrated by the enzymatic production of amoxicillin and cephalexin. Very robust immobilized penicillin acylases have been developed, and being a very flexible enzyme, new applications are arising for it in the pharmaceutical and finechemical industries. Production of acrylamide from acrylonitrile by nitrile hydratase (nitrile hydro-lyase) represents a multiton business, especially in Japan where the production of acrylamide with cells of Rhodococcus rhodochrous containing a mesophilic non-coenzyme-requiring nitrile hydratase exceeded 30,000 ton year
1 a decade ago, representing 40% of the total world market. Production should have increased further because of the advantages of the bioprocess over the conventional chemical process in terms of environmental protection and energy consumption. Major advances have been reported recently in the development of both immobilized cells and enzymes that are quite certainly the type of biocatalysts now used industrially. These three examples illustrate the impact of enzyme immobilization in industrial biocatalysis, but many more cases exist where immobilization was the clue for industrial success, like the large-scale production of aspartate from fumarate using immobilized cells and the production of cocoa butter analogs by interesterification with immobilized lipases. Enzyme biocatalysis has evolved from rather simple reactions of hydrolysis to more complex reactions of organic synthesis. The latter require unstable, intracellular, and usually coenzyme-dependent enzymes, or else hydrolytic enzymes used in reverse in reaction media engineered to depress their hydrolytic potential. In both cases, enzyme inactivation is a serious drawback, so that biocatalysis in organic synthesis is a major niche for immobilized enzymes now. Interestingly, a new input in enzyme immobilization has been produced because supports designed for hydrolytic reactions do not necessarily match the requirements for performing in nonconventional media, so now the term “biocatalyst engineering” has been coined meaning the different strategies of enzyme modification required for its effective use in such conditions. Among such strategies, enzyme immobilization is of paramount importance for increasing enzyme stability and allowing biocatalyst reuse. Even in the case where the enzyme is naturally insoluble in the reaction media (as it occurs in neat hydrophobic solvents) immobilization may help by providing an amplified surface for substrate interaction. The subject of industrial applications of immobilized enzymes has been recently analyzed by Illanes.3

1.03.2.2

Other Applications of Immobilized Enzymes

Properties of immobilized enzymes make them attractive in several fields beyond industrial application. Immobilized enzymes are increasingly being used in chemical and clinical analysis, biomedicine, and basic and applied research. Enzymes are potent analytical tools because of their specificity and sensibility allowing the quantification of a wide range of substances at very low concentrations with minimal interference. Problems arising from the low stability and high cost of enzymes have been circumvented by the use of robust immobilized enzymes that increase the shelf life of the analytical device. Microfluidic analytical systems have been developed with immobilized enzymes and very robust immobilized enzyme electrodes are being used for chemical analysis and process control in the fermentation and food industries. Biocatalytic nanoparticles are novel type of enzyme catalysts whose impact in chemical and clinical analysis is emerging; high stability and insignificant mass transfer limitations in these biocatalysts are desirable properties yet to be fully exploited. Immobilized enzymes are extensively used in clinical analysis, as detectors in immunoassays and diagnostic kits, because of improving enzyme stability and allowing a flexible design of the analytical devices. Enzymes have a great potential in clinical medicine in the treatment of congenital metabolic deficiencies, in the selective elimination of toxic metabolites accumulated by organ malfunction and in the selective nutritional depletion of malignant cells. Enzyme immobilization in solid matrices or by confinement within semipermeable membranes has been used both in extracorporeal (ex vivo) and intracorporeal (in vivo) applications.

28

Immobilized Biocatalysts

In the case of in vivo applications, immobilization to biocompatible matrices, beyond increasing stability and allowing flexibility of design, is highly desirable for reducing the immune response and also by allowing the targeting of the enzyme to the corresponding site of action within the patient’s body. Several systems for enzyme delivery have been used including microencapsulation, liposome entrapment, and red blood cell ghosts. Magnetic nanoparticles carrying functional biomolecules is a great promise for drug delivery in clinical medicine. An updated review on the subject has been published by Swi.4 Enzymes are essential components in the toolbox of biotechnological research, as illustrated by the case of restriction endonucleases used in gene splicing and thermostable DNA polymerases used for amplification of genetic material. These enzymes require to be highly purified, so their cost is high. Immobilization is being used in those cases to improve the efficiency of enzyme use and solid-phase DNA restriction digest reactors have been developed to be used within the context of recombinant DNA technology.

1.03.3

Methods of Enzyme Immobilization

Methods of immobilization can be divided into those that involve the interaction of the enzyme with a matrix (usually through a chemical bond) and those in which the enzyme is contained within a restricted space.

1.03.3.1

Immobilization by Chemical Interaction

The enzyme molecules may be linked to an inert carrier by covalent or noncovalent bonds (in some cases, interaction forces are too weak to establish a chemical bond) in which case a carrier-bound biocatalyst is obtained. If the enzyme molecules are chemically linked among themselves, generally through a bifunctional reagent, without the participation of an inert carrier the biocatalyst produced is termed “carrier free,” being in this case the enzyme protein its own matrix.

1.03.3.1.1

Carrier-Bound Immobilized Enzymes

Both organic and inorganic compounds have been used as inert enzyme carriers. High protein-binding capacity, high surface-tovolume ratio, insolubility in the reaction medium, high mechanical and chemical stability, recoverability after use, and conformational flexibility are desirable properties of a carrier. There is no material that fulfills all these requirements so that many have been tested to suit the particular needs of a given process and no guidelines are available. Materials for enzyme immobilization can be divided into nonporous and porous. The former biocatalyst, in which the enzyme is attached to an impervious surface, is subjected to minimum mass transfer limitations while enzyme loading is rather low. The latter, where the enzyme is contained within a porous network, suffers from significant mass transfer limitations while having a rather high enzyme loading capacity. Impressive advances in carrier design have been obtained in recent years. Nanotechnology has provided powerful tools to develop nanostructures as enzyme carriers where the above compromise can in principle be solved. In fact, these biocatalysts are increasingly being tested to exploit the benefits of high enzyme loadings, large surface area per unit volume, and reduced mass transfer limitations. However, handling of such nanoparticles within a bioreactor is cumbersome and biocatalyst recovery is a complex task; immobilization in magnetic nanoparticles is a plausible solution that has been proved to be effective. An insight on nanostructured biocatalysts has been done by Wang.5 Materials tested at laboratory scale are not necessarily adequate when scaling up the process, so handling, availability, reliable quality, and cost are variables to be considered which may not be relevant at small scale. Immobilized enzymes were originally developed to perform in conventional aqueous medium, so a judicious analysis should be made about the carrier when using the enzyme in nonconventional media as is frequent in reactions of organic synthesis. Enzymes can be linked to carriers by covalent or noncovalent interactions. Comprehensive information on methods and procedures has been compiled by Guisan.6 Covalent bonds may be established between functional groups in the activated carrier and functional groups in the amino acid residues of the enzyme, such as –OH, –SH, –NH2, and –COOH. Covalent immobilization is a rather complex method, the carrier is hardly recoverable after enzyme exhaustion, immobilization yield is comparatively low, and the kinetic properties of the enzyme can be severely altered. However, operational stability is high and it is quite flexible, so that directed immobilization can be done to suit the particular characteristics of the enzyme and the reaction to be catalyzed. Among the many systems for covalent enzyme immobilization, multipoint covalent attachment, where the enzyme is linked to the support through several amino acid residues, is particularly interesting and has been developed using different solid supports such as porous glass, polyacrylamide, cellulose, chitosan, magnetic particles, and many more. Multipoint covalent attachment to glyoxyl-agarose is particularly appropriate rendering very intense enzyme-support attachments; stable covalent linkages between free nonprotonated amino groups in the enzyme surface and aldehyde groups in the activated carrier are established after reduction of the Schiff’s base formed to render a secondary amine. Significant increase in stability has been obtained with many enzymes of industrial importance. The enzyme is attached to the carrier by its surface region having the highest density of lysine residues, but the method has been extended for those enzymes poor in surface lysine residues either by chemical amination or by protein-engineering techniques. Nonprotonated ε-amino groups of lysine are required, which imposes some restrictions because the pK of that group is 10 so that high pH is required for efficient attachment that may be detrimental for enzyme activity. The problem may be overcome by the use of protecting agents and active-site blockage during immobilization. Amino-epoxy carriers are also quite interesting commercially available matrices for enzyme immobilization

Immobilized Biocatalysts

29

by covalent attachment that are prepared by reaction of highly activated amino-supports with butanediol diglycidyl ether. The first step, carried out at neutral pH, is the ionic adsorption of the enzyme through the region having the highest density of negatively charged amino acid residues; in the second step, the reaction between the epoxy groups in the support and the amino groups in the enzyme takes place; in the third step, carried out at high pH, a much more intense multipoint covalent attachment is promoted contributing significantly to enzyme stabilization. Even though the immobilization procedure is more complex than in glyoxylagarose, very stable derivatives can be produced and even hardly adsorbable enzymes can be efficiently immobilized because the few molecules that are adsorbed are also covalently linked, and hence the equilibrium is shifted toward adsorption. Glutaraldehyde is a bifunctional reagent quite useful for developing protocols for covalent immobilization to solid supports. It is a nontoxic generally recognized as safe (GRAS) reagent and enzyme immobilization on amine-containing supports activated with glutaraldehyde is a simple process as illustrated by the case of activated chitosan, whose free amino groups are reacted with one aldehyde group of glutaraldehyde being the other available for linkage to free amino groups in the enzyme molecule. Immobilization of enzymes on glutaraldehyde-activated supports has been thoroughly analyzed by Betancor et al.7 Noncovalent immobilization to solid supports includes all kind of interactions between the enzyme and the support not involving covalent bonds, namely hydrophobic interactions, ionic bonds, and weaker interactions such as van der Waals forces; these methods are referred generically as adsorption. They are simple, consisting in the mere contact of the enzyme with the carrier (whether functionalized for ionic binding, as in the case of anion or cation exchange resins, or not) at appropriate conditions. Immobilization yields are usually high, because no obnoxious reagents are involved and enzyme–carrier interactions are barely distorting. In this case, the carrier can be easily recovered after enzyme exhaustion by promoting protein desorption, which is an important advantage that allows efficient recycling of the support. However, its main drawback is that the enzyme can be easily desorbed by subtle changes in the reaction medium. In the case of ionic binding, the strength of the interaction between the ionic groups in the carrier and the charged amino acid residues in the enzyme surface is weak and most of the protein will desorb at relatively low ionic strengths (200–300 mM) or by mild changes in pH. In newly designed matrices covered with ionic polymers, such as polyethyleneimine (PEI) and dextran sulfate (DS), the strength of the ionic interaction between enzyme and carrier is increased obtaining more stable and reusable enzyme biocatalysts amenable for industrial use. Enzyme desorption is a serious problem in the case of aqueous biocatalysis but not so when using nonaqueous systems as it is frequent in reactions of organic synthesis. Lipases, which are versatile catalysts for organic synthesis, perform quite well in organic media, and in this case immobilization by adsorption onto hydrophobic carriers is a very good option, even more so because in many cases hyperactivation is observed, the hydrophobic surface of the carrier promoting an open reactive configuration of the active site. Important technological developments have improved the stability of adsorbed enzymes being the use of mesoporous carriers outstanding. Mesoporous silica is an interesting material for enzyme immobilization, providing a rigid structure of controllable pore size. The confinement of the enzyme in a pore of similar dimension to it or the crowding of enzyme molecules within the pore prevent their free movement and restrain enzyme unfolding and denaturation. Size matching between pore size and enzyme molecule is crucial in terms of enzyme stability and it is a variable to be optimized. Leaching of the adsorbed enzyme out from the mesoporous material is a crucial problem that can be solved by promoting covalent or ionic interactions, but novel strategies have been proposed, like the use of entrapping agents that partially close the inlet of the mesopores aiding enzyme retention. Stability is also been increased by cross-linking of the adsorbed enzyme in the mesopores with bifunctional reagents.

1.03.3.1.2

Carrier Free Immobilized Enzymes

The carrier is usually an inert material that simply harbors the enzyme not contributing to catalysis and in many cases impairing it. In this sense, it is a dispensable material that dilutes out the catalytic potential of the biocatalyst. Immobilization of the enzyme in its own protein by promoting aggregation may eventually lead to an insoluble matrix where the carrier contains almost no inert material. This reduces costs and also produces biocatalysts of higher specific activity, being the enzyme concentration within the biocatalyst particle close to its theoretical limit of packing. Enzymes can be insolubilized with bifunctional reagents, such as glutaraldehyde, by straight chemical cross-linking of the soluble protein (cross-linked enzyme, CLE), or a crystallized enzyme (CLE crystal: CLEC), or a precipitated enzyme aggregate obtained under nondenaturing conditions (CLE aggregate: CLEA). CLEs are no longer used mainly because of their poor mechanical properties, low activity and low reproducibility. CLECs, on the other hand, are highquality biocatalysts in terms of activity, stability, and recycling, but their cost is for most of the cases exceedingly high because the enzyme requires to be extensively purified to form crystals. However, these biocatalysts are interesting for some reactions of organic synthesis of high added value that can afford a costly biocatalyst if robust enough to withstand the harsh reaction conditions and active enough to efficiently catalyze intrinsically slow reactions. A complete review on CLECs was published by Roy and Abraham.8 CLAs are outstanding biocatalysts because they can be easily produced by cross-linking of protein aggregates obtained by conventional protein-precipitation techniques producing inexpensive catalysts (no extensive purification of the enzyme is required) that share many of the good properties of CLECs. The method, developed at Delft University by Prof. Roger Sheldon’s research group, has been extensively used and proved to be adequate for producing robust enzyme biocatalysts. CLEAs of most enzymes of industrial significance, such as lipases, penicillin acylase, aminoacylase, nitrilases, proteases, and oxidoreductases, have been prepared in the last decade for catalyzing reactions of hydrolysis and synthesis. Limitations of CLEAs as industrial biocatalysts are mainly a consequence of its rather poor mechanical properties and small and variable size that impairs their handling in reactor operation. However, special reactor configurations have been developed to handle them properly and improvements have been achieved by structural modification of the catalyst, either by entrapping

30

Immobilized Biocatalysts

the CLEAs in gel particles or by allocating them within mesoporous matrices. Procedures have been developed for the coimmobilization of different enzymes (combi-CLEAs) for performing complex biotransformations or acting upon complex heterogeneous substrates. Co-immobilization of enzymes and their respective coenzymes is another potential of CLEAs for catalyzing reactions of synthesis requiring coenzyme regeneration. CLEAs have also been used with multimeric enzymes to prevent subunit dissociation, because in this case every subunit becomes cross-linked and does not leak out from the biocatalyst particle. CLEAs with hydrophilic microenvironments have been produced by co-precipitating the enzyme with ionic polymers, such as polyethyleneimine (PEI) and dextran sulfate (DS), which is a sound strategy to improve enzyme stability in harsh organic media, being these CLEAs particularly useful in reactions of organic synthesis. Synthesis of CLEAs in the presence of additives such as crown ethers or surfactants has been claimed to promote favorable configurations resulting in increased activity or selectivity. The potentials of CLEAs have been highlighted by Sheldon et al.9; an appraisal of carrier free enzymes as alternative of conventional carrierbound enzymes has been reported by Cao et al.10

1.03.3.2

Immobilization by Containment

The enzyme is in this case confined to a restricted space either by molecular entrapment, as it occurs inside polymeric gels, or by selective retention by a semipermeable membrane that allows the free passage of small molecular weight compounds including substrates and products of reaction.

1.03.3.2.1

Entrapment

The method is based on the sol–gel transition that entraps the cells or enzymes (dissolved or suspended while in sol) within the inner cavities of a solid polymeric matrix. Immobilization occurs by promoting polymerization, either chemically or physically, of a monomer solution in which the enzyme is dissolved. The first types of gels used to entrap cells or enzymes were organic or bioorganic materials, being alginate, polyacrylamide, polyurethane, polyvinyl alcohol, k-carrageenan, and chitosan the more frequently reported. Entrapment in this type of polymeric gels has been a powerful tool for cell immobilization, but in the case of enzymes, because of their smaller size, leakage from the matrix is a problem that can be circumvented by increasing gel strength, but this in turn increases mass transfer limitations. Immobilization of cells in polyvinyl alcohol gel has been successful because it is innocuous, cheap, and robust and its use has been extended for enzyme immobilization in the form of enzyme–polymer composites and CLEAs. Biocatalysts of this kind are robust and easy to recover during reactor operation. Sol–gel encapsulation of enzymes in inorganic carriers is a very attractive system of immobilization. Particularly important is the case of silica gels that are formed by acid- or base-catalyzed hydrolysis of tetraalkylosilanes, where the silane precursor undergoes hydrolysis and cross-linking condensation to form a silica matrix in which the enzyme is entrapped. Sol–gel encapsulation of enzymes has had more impact in biosensors than in bioprocesses and in that field several robust enzyme analytical devices have been developed. Silica gels can be templated with surfactants to form robust mesoporous structures of controlled pore size, where mass transfer limitations are significantly reduced. Highly ordered mesoporous matrices have also been used as supports in which the silica gel containing the enzyme is deposited within the pores. Hybrid inorganic–organic gels have also been developed to improve the mechanical properties of the purely organic gels and improve the biocompatibility of the purely inorganic gels. A review on the subject has been done by Pierre.11

1.03.3.2.2

Membrane Retention

In this case, the enzyme is not bound to a carrier or insolubilized by aggregation but confined to a certain space by a semipermeable membrane that physically separates the enzyme from the other components of the reaction system. Retention can be attained by microencapsulation or confinement by ultrafiltration membranes. Microencapsulation is produced by promoting a polymerization reaction in the surface of drops of enzyme aqueous solution dispersed in a water-immiscible organic solvent with the aid of a surfactant. One such system is reverse micelles, in which the hydrophilic head of the surfactant is oriented to the inner enzyme aqueous phase while its hydrophobic tail projects to the outer organic phase. Liposomes are a special kind of micelles composed of a double layer of surfactant, so in this case the external solvent is the aqueous enzyme phase. Enzyme clumping inside micelles may promote enzyme stabilization, while mass transfer limitations are negligible; however, poor mechanical properties and the tendency of the enzymes to denature at the water–organic interface are limitations that have precluded their use as process biocatalyst, being its potential more related to biomedical applications. Lipases are a special case because they are structured to perform at interfaces, their immobilization in reverse micelles being consistently reported as a good strategy. Confinement of enzymes within a space delimited by an ultrafiltration membrane has interesting technological potentials in devising enzyme-membrane reactors allowing the free passage of substrates and products while retaining the enzyme. In this way, retention time of the catalyst is independent of fluid retention time, allowing continuous operation. Besides, the membrane may serve as a permeability barrier for macromolecules whose contact with the enzyme is undesirable, as occurs in the continuous hydrolysis of milk with b-galactosidase to produce lactose-free milk for intolerants, where the milk macromolecular components are kept away from the enzyme avoiding alteration of the product. The enzyme may be free in the inner or outer space of the ultrafiltration reactor but may also be attached to either side of the membrane. Drawbacks of enzyme containment in

Immobilized Biocatalysts

31

ultrafiltration devices are the inactivation promoted by interfaces and excessive aggregation and the fouling of the membrane in prolonged operation. However, the impressive advances in materials science and reactor design are providing better and cheaper membranes and improved configurations making this system quite appealing, especially in the case of reactions of synthesis with coenzyme requiring enzymes where such devices can retain the enzyme as well as the derivatized coenzyme, so that the whole biocatalytic system is contained. Reactors of this type have been used at the industrial level, as illustrated by the case of the synthesis of L-tert-leucine, an important building block for the synthesis of pharmaceuticals, where leucine dehydrogenase catalyzes the reductive amination of trimethylpyruvate in a membrane reactor confining the enzyme and the pegylated coenzyme (reduced nicotinamide adenine dinucleotide); coenzyme regeneration is produced by the reduction, with the auxiliary enzyme formate dehydrogenase, catalyzing the conversion of formate to carbon dioxide that, being highly volatile, leads to a favorable shift of the equilibrium and does not contaminate the product.

1.03.4

Properties of Immobilized Enzymes

Immobilization may have a profound impact on essential properties of the enzyme like activity, stability, substrate specificity, and selectivity.

1.03.4.1

Activity

The expression of activity of an immobilized enzyme may be hindered by mass transfer limitations, as will be further analyzed in the section on heterogeneous catalysis. However, immobilization may also produce structural changes in the enzyme molecule affecting its activity. This usually reduces it, but the opposite may occur as is the case of lipases when immobilized onto hydrophobic matrices where hyperactivation is observed as the consequence of a configuration more prone to catalysis; also, a favorable microenvironment may keep deleterious substances apart from the active site so enhancing activity and immobilization may prevent enzyme aggregation and autolysis. Enzyme activity losses and incomplete protein capture during immobilization should also be considered and may vary considerably from one procedure to another.

1.03.4.2

Stability

Stability (preservation of enzyme activity through time) may be considerably affected by immobilization. Even though far from a general rule, immobilization usually improves enzyme stability by making the enzyme molecule more robust and sometimes protected from an aggressive environment. Again, this effect may vary considerably according to the immobilization method used. Whatever the case, increase in stability is a main objective underlying immobilization, so this is a task to be pursued. In fact, most efforts to engineer the immobilization process have been devoted to increase biocatalyst stability. Stability is often assessed in terms of half-life, which is a crude parameter from a mechanistic point of view; in some cases, models based on plausible inactivation mechanisms have been used and their parameters evaluated. However, most data have been gathered under nonreactive conditions, so its predictive value can be argued against when used for evaluating the actual operational stability of the biocatalyst; in fact, significant negative or positive modulation of stability by catalytic effectors such as substrates or products of the reaction has been reported. The assessment of immobilized enzyme stability is not trivial because of the interplay with mass transfer limitations. In fact, immobilized enzymes appear as more stable when mass transfer limited because the reaction moves away from kinetic limitation; then the true stability of an immobilized enzyme should be determined under conditions of negligible mass transfer limitation.

1.03.4.3

Selectivity and Specificity

Enzymes are increasingly being used as catalysts for organic synthesis where substrate specificity and reaction selectivity are major issues. Enzymes are highly chemo-, regio-, and enantioselective, being attractive catalysts for the production of pharmaceuticals and fine chemicals. Chiral drugs are produced as racemic mixtures when chemically synthesized, being this a problem in terms of process efficiency and safety because only one of the isomers (eutomer) is effective whereas the other (distomer) is inert or, even worse, harmful. Lipases, esterases, and proteases have been used for producing several chiral compounds of industrial interest by performing a kinetic resolution of the corresponding racemate. In several cases, substantial improvement in enantioselectivity has been obtained by immobilization. Enzymes are also powerful catalysts for asymmetric synthesis in which an enantiomerically pure compound is formed from a prochiral substrate by a selective reaction in which chirality resides on the catalyst itself. This strategy has been successfully used, mainly with lipases, to produce enantiomerically pure chiral compounds from prochiral precursors. Immobilization, chemical modification of existing immobilized enzymes, and new enzymes created by directed evolution are powerful tools for obtaining significant changes in enantioselectivity and substrate specificity. This is a fertile area of research and significant outputs are expected from enzyme biocatalysis as a powerful technology for asymmetric synthesis. A review on the subject has been published by García-Urdiales et al.12 Improvement of enzyme activity, stability, and selectivity by immobilization has been reviewed by Mateo et al.13

32

Immobilized Biocatalysts

1.03.5

Evaluation of Enzyme Immobilization

Enzyme immobilizations may be considered as the most significant technological development in biocatalysis in terms of process design, serving to several purposes related mainly to the increased stability and recoverability of the biocatalyst. These properties allow the development of more productive continuous or sequential batch reactor operation, easier operation and control, facilitated downstream operations because product is free of catalyst, and high flexibility in reactor configuration to match the particular characteristics of the process. These advantages may translate into significant cost savings that will help in making the process profitable. To do so, such advantages must prevail over the potential disadvantages of enzyme immobilization: losses in activity, mass transfer limitations, and additional costs in biocatalyst production. With the development of more rational protocols for enzyme immobilization, additional advantages may be obtained in terms of kinetic behavior with enhancement of activity and selectivity, especially when the enzyme is to be used on nonnatural substrates or in nonconventional media. Enzyme immobilization is not a panacea. In fact, immobilization has been a major area of research and development since the early 1970s and, at least from an industrial perspective, its impact can be considered yet modest. Reasons underlying are many, and industrial reluctance to modify existing technologies is certainly one of them. However, despite the impressive advances in the field, drawbacks of immobilized enzymes are still prevalent in many cases making it necessary to thoroughly evaluate the process of immobilization to confront their advantages and disadvantages in each particular situation. Most methods of immobilization involve a preliminary stage in which the carrier (and sometimes the enzyme) is activated, and a second stage in which both partners interact with each other to produce the immobilized enzyme. Optimization of the immobilization process is quite important because it may have significant impact, not only in the cost of the biocatalyst but also in the operating cost of the process in which the biocatalyst will ultimately be used. Optimization of enzyme immobilization is a complex task, because it is a multivariable process and a sound, but not obvious, objective function is required for evaluation as will be analyzed in the corresponding section.

1.03.5.1

Immobilization Yields

It refers to the fraction of contacted enzyme that is recovered in the biocatalyst after immobilization. This yield is sometimes expressed in terms of protein and merely represents the fraction of the contacted protein that is retained in the biocatalyst. So defined, protein immobilization yield (YP) is YP ¼

PI PC
PU ¼ PC PC

(1)

where PI is the immobilized protein, PC the contacted protein, and PU the unbound protein. YP is less than one when the binding capacity of the carrier is exceeded. Immobilized protein is not determined directly, being calculated as the difference between contacted and unbound protein. This parameter is not necessarily a reflection of the immobilized enzyme protein since, as it is usual, the enzyme preparation to be immobilized is not pure and contains variable amounts of contaminant proteins that will be immobilized as well. YP gives some insight on the immobilization process but gives no information as to the expression of activity of that protein. A convenient parameter to define is enzyme immobilization yield (YE), which represents the fraction of the contacted activity that is expressed in the biocatalyst YE ¼

EI EI ¼ EC EI þ EU þ EL

(2)

where EI is the activity of the immobilized protein, EC the contacted enzyme activity, EU the activity of the unbound protein, and EL represents the enzyme activity lost during immobilization. The meaning of YE and Eq. (2) has to be analyzed with caution because enzyme is quantified in terms of units of activity and not mass and, being Eq. (2) actually a material balance, the term EL has to be included to close that balance. EL represents the fraction of the contacted enzyme activity which is not expressed either by the biocatalyst or by the unbound enzyme and may be attributed to inactivation of the bound, but mostly of the unbound enzyme, or to mass transfer limitations and steric hindrances of the immobilized enzyme. An insight on the causes underlying the partial expression of the contacted activity may be gained by comparing YP and YE. If YP is significantly higher than YE, then structural changes, steric hindrances, or mass transfer limitations are probably significant. However, immobilization may have some degree of selectivity for the enzyme with respect to the whole protein; some insight about that eventual selectivity, whether positive or negative, can be obtained by analyzing the values of PU and EU with respect to PC and EC. A simple way of evaluating YE has been used in which the immobilized activity is determined as the difference between the activity of a compound sample containing both the immobilized and unbound enzyme and that of a filtered sample containing only the unbound enzyme. Such difference may reflect the activity in the biocatalyst provided that no significant inactivation occurs during immobilization, which can be assessed by a control experiment. YE is a very relevant parameter because it directly relates to the cost of the biocatalyst; it tends to be quite high (0.8 or higher) in the case of immobilization by adsorption and somewhat lower in covalent immobilization, where figures from 0.4 to 0.6 are usual, even though by applying rational design and optimizing condition values of 0.9 and higher have been reported. A remarkable exception is the case of immobilization of lipases in hydrophobic

Immobilized Biocatalysts

33

carriers where YE may be higher than one as a consequence of hyperactivation; however, even in this case, hyperactivation disappears as the carrier is challenged to high protein loads so in practice YE lower than one are also obtained.

1.03.5.2

Specific Activity of Immobilized Enzyme

Protein load (Pload) is an important parameter that reflects the protein-binding capacity of the carrier and it is something to determine at the early stage of its selection. Pload is simply the ratio of immobilized protein per unit mass (Mcat) of biocatalyst Pload ¼

PI PC
PU ¼ Mcat Mcat

(3)

Protein load can be expressed alternatively per unit volume of biocatalyst or by unit of surface area in the case of impervious carriers. Pload is a crude parameter that gives no insight on the catalytic potential of the immobilized enzyme. A very significant parameter of enzyme immobilization is biocatalyst specific activity (Aspec), defined as the amount of immobilized enzyme activity per unit mass (or volume, or surface area) of biocatalyst Aspec ¼

EI Mcat

(4)

Although EI can be approximated by subtracting the total activity in suspension to the activity in solution after immobilization, it is recommendable to be determined directly over the biocatalyst. This parameter is more related to biocatalyst performance than biocatalyst production and will impact the operation cost of the process in which the immobilized enzyme is going to be used. A higher Aspec means a smaller load of biocatalyst mass which has a direct impact on reactor size and also in the fluid regime within the reactor.

1.03.5.3

Biocatalyst Stability

This is a fundamental property of an immobilized enzyme that is seldom considered as a parameter to be evaluated during immobilization development. However, significant differences in biocatalyst stability with respect to immobilization conditions have been reported. For instance, when determining the effect of enzyme–carrier contact time, best values are frequently smaller for maximizing yield than for maximizing stability. Stability, as Aspec, is related to biocatalyst performance and will have a strong impact on reactor operation. For the purpose of evaluating enzyme immobilization, half-life (t1/2), which is mostly used to quantify enzyme stability, may be considered as a suitable parameter

(5) t1=2 ¼ t
e0 e¼ 2

where e0 represents the initial enzyme activity.

1.03.5.4

Optimization of Enzyme Immobilization

Main variables to consider are the ratio of activating agent to support, the ratio of enzyme contacted to activated support, and pH, temperature, time of contact, and agitation rates in both stages of immobilization. Interactions among variables are likely to occur making a sound statistical design required to determine the best conditions for immobilization. However, the more complex aspect is the definition of a proper objective function because, at the end, optimal conditions will be those minimizing costs or maximizing profits, aspects that are not easy to assess during the early stages of research and development. Immobilization parameters as those presented above need to be determined and taken into consideration. YE and Aspec are usually in compromise, being the optimum conditions not coincident. Some guidelines may be proposed to determine which parameter should be given priority: for instance, in the case of not very expensive enzymes Aspec can be used as objective function considering YE as a restriction parameter. However, this is an oversimplified approach. Stability (expressed as t1/2) may also be in compromise with YE, which introduces an additional problem. An objective function adequately weighing these parameters could be a good criterion for optimization. Ideally, a cost-based objective function considering those three parameters weighed according to their impact on the cost of the process of enzyme utilization should be developed. However, this is quite a complex task because the impact of each parameter on process economics is hard to evaluate and, as said before, relates not only to the immobilization process itself but also to the ultimate use of the biocatalyst. YE is strictly related to the immobilization process and will have a strong impact when the cost of the enzyme is high. On the other hand, stability and specific activity are related to the quality of the biocatalyst and their impact is related to bioreactor operation and design. Unfortunately, the proper weighing of these parameters in terms of their impacts in a cost-objective function is hard to do, especially at an early step of process development as the biocatalyst selection is. In addition, a proper parameter to reflect enzyme stability is not a simple choice. A crude parameter like t1/2 may be used for the present purpose; however, it has no mechanistic significance except for the case of first-order one-stage mechanism of enzyme inactivation, which is seldom representative for immobilized enzymes where more complex multistage mechanisms of inactivation are likely to better describe it. Besides, more data on stability are gathered under nonreactive conditions that may poorly represent it under actual operation conditions. Practical experience may guide the weighing of

34

Immobilized Biocatalysts

immobilization parameters and a preliminary evaluation of operating costs of the process in which the enzyme is going to be utilized, in terms of such parameters is highly desirable. Of course, this approach is restricted to the case in which the producer of the immobilized enzyme is its own user. Commercially available immobilized enzymes are not necessarily optimal with respect to a particular process, being attractive to end users to develop their own immobilization strategies or else to work under agreement and in close contact with immobilized enzyme suppliers.

1.03.6

Heterogeneous Biocatalysis

1.03.6.1

Definitions

Enzyme kinetic principles, developed early in the 20th century, were based on homogeneous systems, this is, when the biocatalyst and their substrates and products of reaction are in a single phase where the reaction occurs. In this case, reaction will be solely determined by the activity of the biocatalyst and rates of transport of substrates and products will be irrelevant. Enzyme immobilization introduces heterogeneity because the catalyst phase (usually solid) in which reaction takes place differs from the bulk liquid phase in which substrates and products are dissolved and the reaction is monitored. Even in the case of enzymes immobilized in the surface of impervious carriers, reaction takes place at the solid–liquid interface whose properties differ from those of the bulk liquid medium. Enzyme immobilization may produce both conformational and micro-environmental effects. The former refers to the structural changes produced in the enzyme molecule and steric effects due to its close proximity to the surface of the carrier; the latter refers to mass transfer limitations. Kinetic behavior of the immobilized enzyme will be determined not only by its catalytic potential but also by the rate of mass transport of substrates from the bulk reaction medium to the biocatalyst and products back from it to the bulk reaction medium. Intrinsic kinetics refers to the behavior in the absence of mass transfer limitations and the kinetic parameters so obtained are termed “intrinsic parameters.” Kinetic behavior under the influence of mass transfer limitations is termed “effective” (is what one gets) or “apparent” (from the standpoint of the enzyme) and the parameters determined under such conditions are termed “effective” or “apparent.” In some cases, partition effects at the biocatalyst medium interface may be significant in which case the term inherent is used to designate the behavior in the absence of mass transfer limitations but subjected to partition. This aspect will not be considered but information can be obtained from Illanes.3

1.03.6.2

Mass Transfer Effects

Mass transfer limitation may severely restrain the expression of the catalytic potential of the immobilized enzyme, being necessary to assess its impact to properly evaluate the biocatalyst performance. Mass transfer limitation make the substrate (and product) concentration in contact with the enzyme different from that in the bulk reaction medium producing the corresponding profiles. In the case of enzymes immobilized within the structure of the carrier, those profiles will be developed inside the biocatalyst whereas in the case of impervious carriers a profile may be developed in the vicinity of the biocatalyst surface as a consequence of a stagnant layer of liquid. Mass transfer limitations are expressed as diffusional restrictions because transport of substrates and products, whether inside the biocatalyst particle (internal diffusional restrictions (IDRs)) or through the stagnant layer surrounding it (external diffusional restrictions (EDRs)), occurs solely by molecular diffusion. EDRs may be significant for enzymes immobilized on the surface of an impervious carrier. IDRs occur when the enzyme is contained within a solid matrix, whether a gel or a microporous solid support. IDRs are usually more severe than EDRs because substrates and products should diffuse through a medium in which mass transfer will be even slower than in a stagnant liquid layer. This is schematically represented in Fig. 1.

1.03.6.2.1

External Diffusional Restrictions

At steady-state conditions, the rate of substrate transport through the stagnant layer and the rate of substrate conversion by enzymatic catalysis at the surface of the biocatalyst should equal. Considering that the enzyme is homogeneously distributed over the surface of an impervious support and intrinsic Michaelis–Menten kinetics

A

B

p p0

s0

s0

s0 s

s

0 Figure 1

L

Schematic representation of: (A) external diffusional restrictions and (B) internal diffusional restrictions in enzyme biocatalysts.

35

Immobilized Biocatalysts

hðs0
sS Þ ¼

Vmax $sS Km þ sS

(6)

where h is the volumetric mass transfer coefficient, s0 and sS the concentrations of substrate in the bulk liquid and in the biocatalyst surface, respectively, and Vmax and Km the intrinsic kinetic parameters of the enzyme. Eq. (6) is more conveniently expressed in dimensionless form as b0
bS ¼

a$bS ¼ av 1 þ bS

(7)

where b0 ¼

s0 sS v0 Vmax ; bS ¼ ; v¼ ; and a ¼ Km Km Vmax h$Km

a is the dimensionless Damkoehler number and represents the relative incidence of enzyme catalytic potential and substrate mass transfer rate. A high value of a implies that the system is limited by substrate diffusion so EDRs are relevant. The opposite holds for small values of a where the system is limited by the catalytic potential of the enzyme. Values of a < 1 mean that the system is essentially free of EDRs. bS can be obtained as a function of measurable b0 and a from Eq. (7) qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1 þ a
b0 Þþ ð1 þ a
b0 Þ2 þ 4b0 (8) bS ¼ 2 At low values of a, the behavior is typically Michaelian, whereas at very high values of a correlation between n and b0 becomes linear, as predicted from Eq. (7). The magnitude of EDR can be conveniently expressed by the effectiveness factor, defined as the ratio of effective to intrinsic reaction rate. For simple Michaelis–Menten kinetics h¼

Vmax $sS Km þsS Vmax s0 Km þs0

¼

bS ð1 þ b0 Þ b0 ð1 þ bS Þ

(9)

where h represents the magnitude of diffusional restrictions, being the fraction of the catalytic potential of the enzyme that is expressed at a certain conditions under the influence of mass transfer limitations. From Eqs. (8) and (9),  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  ð1 þ b0 Þ 1 þ a þ b0
ð1 þ a
b0 Þ2 þ 4b0 (10) h¼ 2b0 a Graphical representation of Eq. (10) is presented in Fig. 2. Eq. (10) allows to quantify the behavior of an enzyme catalyst under EDRs provided that a (or h) and the intrinsic kinetic parameters (Vmax and Km) are evaluated. Intrinsic kinetic parameters can be determined using usual kinetic procedures (by linear or nonlinear regression of initial rate data at different bulk substrate concentrations) working at conditions ensuring negligible EDR: in practice, this can be done by increasing agitation over the threshold value below which reaction rate becomes dependent on agitation speed, which could be done easily as long as the integrity of the catalyst is maintained. If initial rate data is gathered under a broad range of substrate concentration, intrinsic kinetic parameters can be obtained conventionally: for instance, in a double reciprocal plot (1/v vs. 1/s0) a straight line of slope Km/Vmax and intercepts 1/Vmax and
1/Km will be obtained at high values of s0 (s0 \ Km) whereas a straight line will also be obtained at low values of s0 (s0 0 Km) with slope Km(1 þ a)/Vmax. In this way, intrinsic kinetic parameters and a can be determined. Alternatively, some empirical correlations have been proposed for h in terms of the dimensionless numbers of Schmidt and Reynolds. 1.0 0.1 1

0.8

5 0.6

10

η

α

20

0.4 0.2

100 0.0

0

5

10 β0

15

20

Figure 2 Effectiveness factor (h) of an enzyme biocatalyst subjected to external diffusional restrictions as a function of dimensionless bulk liquid substrate concentration (b0) and Damkoehler number (a).

36

Immobilized Biocatalysts

The above analysis can be extended to more complex kinetic behaviors. When product inhibition exists not only substrate transport from the bulk medium to the biocatalyst surface has to be considered but also product transport from that surface back into the bulk medium. Under inhibition, the impact of EDR appears reduced, because the system moves away from kinetic control as the activity of the biocatalyst is reduced. An interesting situation occurs in the case of inhibition by high substrate concentration. In this case, a steady-state analysis renders a third-order equation in bS that, for certain values of the kinetic and mass transfer parameters, may give three positive values of bS for one value of b0 and a stability analysis should be made to assess the right value. More indepth analysis of EDR has been reported by Engasser and Horvath.14

1.03.6.2.2

Internal Diffusional Restrictions

In the case of IDRs, substrate and product profiles will be developed within the biocatalyst matrix, as shown in Fig. 1. In this case, each enzyme molecule is subjected to different environmental conditions according to its relative position within the matrix and a differential analysis within the biocatalyst particle is required to properly describe the system; the resulting differential equation should be solved numerically to yield the corresponding profiles. In the case of IDRs, the behavior of the system is highly dependent on particle geometry so the analysis cannot be generalized as in the case of EDRs. IDR has been rigorously analyzed for the flat slab and the spherical geometry, representing extreme cases of particles with infinite and minimum radius of curvature, respectively; geometries that may well represent some immobilized enzyme configurations. Other geometries have been also analyzed, like the case of cylindrical and spheroidal particles. The problem is quite complex for particles of undefined or irregular geometry and in those cases approximate solutions can be obtained by defining an equivalent length for the particle and using the rigorous solution for the defined geometry, which more closely resembles that of the particle. The case of an immobilized enzyme in a carrier of spherical geometry will be analyzed. A differential equation combining enzyme kinetics and mass transfer allows determining the substrate profile within the biocatalyst particle; then a distribution of local effectiveness factor can be determined from such profile and finally a global effectiveness factor is obtained by adequately averaging that distribution. Considering steady-state condition, homogeneous distribution of the enzyme inside the carrier and intrinsic Michaelis–Menten kinetics, a material balance over a differential section of the catalyst yields   d J 0 $r 2 V 0 $s (11) ¼ r 2 $ max Km þ s dr 0 where J0 is the substrate flux, r the variable radius, and Vmax and Km the intrinsic kinetic parameters of the enzyme. Considering that substrate transport inside the carrier occurs by molecular diffusion and Fick’s law is applicable

Deff

0 d2 s 2Deff ds Vmax $s
¼0 þ 2 dr r dr Km þ s

(12)

where Deff is the effective diffusion coefficient of the substrate inside the biocatalyst particle. In dimensionless form, Eq. (12) becomes d2 b 2 db b
9F2 ¼0 þ dr2 r dr 1þb where b ¼

s Km ,

r ¼ Rr and F is the Thièle modulus defined as R F¼ 3

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi V 0max Km ,Deff

(13)

(14)

with boundary conditions 1. r ¼ 1 b ¼ b0 2. r ¼ 0 db ¼0 dr Boundary condition (1) assumes negligible EDR; if not so, it should be replaced by an equation of continuity through the medium– biocatalyst interface. Numerical solution of Eq. (13) can be obtained using available software b ¼ f ðb0 ; r; FÞ

(15)

which represents the substrates profile inside the spherical biocatalyst particle for any value of F as represented in Fig. 3. Local effectiveness factor, by definition (see Eq. 9) is 0 Vmax $s K þs max $s0 Km þs0

h ¼ V 0m

¼

bð1 þ b0 Þ b0 ð1 þ bÞ

(16)

Immobilized Biocatalysts

β /β 0

1

37

Φ = 0.1 Φ = 0.5 Φ =1

0.75

Φ = 2.5 0.5

Φ =5 0.25

Φ = 7.5 Φ = 10 0

Figure 3

1

0.5

0 ρ

0.5

1

Substrate profile inside a spherical enzyme biocatalyst particle as a function of Thiele modulus (F).

So, from Eqs. (15) and (16), the distribution of h inside the biocatalyst particle is obtained h ¼ f ðb0 ; r; FÞ

(17)

A sound average value for the effectiveness factor of the biocatalyst particle size is the mean integral value of the above distribution R1 Z 1 h$r2 dr h$r2 dr (18) h0 ¼ 0R 1 ¼3 2 dr r 0 0 Then, from Eqs. (17) and (18) h0 ¼ f ðb0 ; FÞ

(19)

Eq. (19) is represented in Fig. 4. h0 is a very strong function of F; however, the range of values of F at which such dependence is observed is highly dependent on the value of b0. To assess the impact of IDRs on enzyme kinetics, the value of intrinsic kinetics and mass transfer parameters must be evaluated. 0 and Km) like comminuting Several strategies have been proposed to approximate the value of the intrinsic kinetic parameters (Vmax the support to obtain particles small enough to obtain very low F values. Another strategy is to reduce the protein load in the support to a point in which the mass specific activity of the biocatalyst is low enough to move the system away from mass transfer control. Deff for substrate within the biocatalyst particle can be estimated from the corresponding values in water but can also be determined experimentally by working with labeled substrate or by determining effusion rates from the biocatalyst particle previously saturated with substrate.

η⬘

1

Φ = 0.1

Φ =1 Φ =5

Φ = 10

0.75 Φ = 25 Φ = 50

0.5

Φ = 100

0.25

0

0

25

50 β0

75

100

Figure 4 Global effectiveness factor (h0 ) of a spherical enzyme biocatalyst particle subjected to internal diffusional restrictions as a function of dimensionless bulk liquid substrate concentration (b0) and Thiele modulus (F).

38

Immobilized Biocatalysts

A similar analysis than the one previously presented for simple Michaelis–Menten has been done for more complex kinetics involving reversible Michaelis–Menten reactions and product and substrate inhibition kinetics, as reported by Jeison et al.15 Eqs. (10) and (19) can be used for the evaluation of reactor performance with immobilized enzymes under EDRs and IDRs, respectively, with b0 ¼ b0i $ð1
XÞ

(20)

where X is the substrate conversion (fraction of substrate converted into product in a given moment) and b0i the initial (or inlet) dimensionless substrate concentration. For instance, batch reactor performance under diffusional restrictions is described by Z Z dX dX t ¼ ¼ (21) veffective vðe; XÞ$hðXÞ si where v(e, X) is the expression for the intrinsic reaction rate of the enzyme and h(X), or h0 (X), correspond to the expressions in Eqs. (10) and (19) for EDRs and IDRs, respectively, considering Eq. (20). In a batch reactor, the impact of diffusional restrictions will increase with the progress of the reaction as the magnitude of the substrate gradient is reduced. More information on reactor design under mass transfer limitations can be obtained in Illanes et al.3

1.03.7

Future Prospects for Immobilized Biocatalysts

Because of their excellent functional properties (activity, selectivity, and specificity), enzymes have a great potential as industrial catalysts. However, in most cases, enzymes have to be significantly improved to exploit such potential. The engineering of enzymes as process catalysts is exciting and challenging, and several scientific and technological developments have paved the way to success, namely, the screening of enzymes with improved properties, the improvement of enzyme properties using molecular biology techniques, the improvement of biocatalyst properties via immobilization and postimmobilization, and the improvement in reaction media and reactor engineering. These strategies complement each other in delivering catalysts for a much more sustainable chemical industry. Poor stability and difficult handling of enzymes within a bioreactor are constraints that can be overcome by immobilization. Immobilized enzymes have already a significant role as process catalysts in a number of industrial processes and also as biosensors and in some biomedical applications. The development of more rational and directed protocols of immobilization, complemented with the advances in material sciences and also in genetic and protein engineering, is opening unprecedented opportunities for enzyme biocatalysis, especially in the production of pharmaceuticals and fine chemicals where the significant added value has been a driving force for technological development. Looking back to the near past, immobilized enzymes appear as having suffered the cycle of euphoria-depression, which is rather common in technological breakthroughs. After a very promising start in the 1970s, their limitations took over but, after a couple of decades, a more realistic vision of their potential and constraints emerged. Now, it is clear that their limitations are on the verge to be solved by technological developments in several fields, while their benefits are there to stay. Predictions made by the eminent Ephraim Katchalsky-Katzir 15 years ago have been premonitory: enzymes are to be used when a novel product is obtained by biocatalysis, when competition with an established technology is envisaged in terms of costs or environmental restrictions, and when the reaction is complex and shows regio and chiral specificity. In his own words: “The use of immobilized enzymes and cells in the food, pharmaceutical, and chemical industries will continue to expand in the future, and will represent an increasingly important section of modern biotechnology.”

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

Bommarius, A. S.; Polizzi, K. M. Vonel Biocatalysts. Recent Developments. Chem. Eng. Sci. 2006, 61, 1004–1016. Chibata, I.; Tosa, T.; Sato, T. Biocatalysis: Immobilized Cells and Enzymes. J. Mol. Catal. 1986, 37, 1–24. Illanes, A. Enzyme Biocatalysis: Principles and Applications, Springer: Cambridge, 2008. Swi, T. M. Artificial Cells, World Scientific: Singapore, 2007. Wang, P. Multi-scale Features in Recent Developments of Enzymic Biocatalyst Systems. Appl. Biochem. Biotechnol. 2008, 152 (2), 343–352. Guisan, J. M. Immobilization of Enzymes and Cells. In Methods in Biotechnology Vol. 22, 2nd ed.; Humana Press: Totowa, NJ, 2006. Betancor, L.; López-Gallego, F.; Hidalgo, A.; et al. Different Mechanisms of Protein Immobilization on Glutaraldehyde Activated Supports: Effect of Support Activation and Immobilization Conditions. Enzym. Microb. Technol. 2006, 39, 877–882. Roy, J. J.; Abraham, T. E. Strategies in Making Cross-linked Enzyme Crystals. Chem. Rev. 2004, 104 (9), 3705–3721. Sheldon, R. A.; Sorgedrager, M.; Janssen, M. H. A. Use of Cross-linked Enzyme Aggregates (CLEAs) for Performing Biotransformations. Chem. Today 2007, 25 (1), 62–67. Cao, L.; van Langen, L.; Sheldon, R. A. Immobilised Enzymes: Carrier-bound or Carrier-free? Curr. Opin. Biotechnol. 2003, 14, 1–8. Pierre, A. C. The Sol–gel Encapsulation of Enzymes. Biocatal. Biotransformation 2004, 22 (3), 145–170. García-Urdiales, E.; Alfonso, I.; Gotor, V. Enantioselective Enzymatic Desymmetrizations in Organic Synthesis. Chem. Rev. 2005, 105, 315–354. Mateo, C.; Palomo, J. M.; Fernández-Lorente, G.; et al. Improvement of Enzyme Activity, Stability and Selectivity via Immobilization Techniques. Enzym. Microb. Technol. 2007, 40, 1451–1463. Engasser, J. M.; Horvath, C. Diffusion and Kinetics with Immobilized Enzymes. In Immobilized Enzyme Principles; Wingard, L., Katchalsky, E., Goldstein, L., Eds., Academic Press: New York, NY, 1976; pp 127–220. Jeison, D.; Ruiz, G.; Acevedo, F.; Illanes, A. Simulation of the Effect of Intrinsic Reaction Kinetics and Particle Size on the Behavior of Immobilized Enzymes under Internal Diffusional Restrictions and Steady State Operation. Process Biochem. 2003, 39 (3), 393–399.

1.04

Lipids, Fatty Acids

James Wynn, Bio-Based Technology Derisking, Lansing, MI, United States © 2011 Elsevier B.V. All rights reserved. This is a reprint of J. Wynn, 1.05 - Lipids, Fatty Acids, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 41-52.

1.04.1 1.04.2 1.04.3 1.04.4 1.04.5 1.04.6 1.04.6.1 1.04.6.2 1.04.7 References

Introduction Structure of Fatty Acids Nomenclature Form in the Cell What Do Lipids Do? Biosynthesis of Fatty Acids and Lipids Classical Synthesis Polyketide Route Biochemistry of Lipid Accumulation

39 39 41 42 42 44 44 47 48 50

Glossary Acyl-carrier protein (ACP) A protein that binds fatty acids during synthesis and transport around the cell. Oleaginous A cell or organism that is capable of accumulating triacylglycerol (TAG) lipids as a carbon and energy reserve material. Phospholipid (PL) A lipid with two fatty acids esterified to a glycerol backbone where the third position on the glycerol backbone is occupied by a charged phospho group, also referred to as a polar lipid. Triacylglycerol (TAG) A lipid composed of three fatty acid molecules esterified to a glycerol backbone, also referred to as a neutral lipid.

1.04.1

Introduction

Lipids, alongside proteins and carbohydrates, make up the majority of the living cell. All cells (with the exception of certain obligate parasites that rely on their hosts for their required lipids) have the ability to biosynthesize lipids. Therefore, lipid synthesis can be considered one of the biochemical prerequisites for life. In chemical terms, lipids are defined very broadly, according to their chemical properties rather than their structure. A chemical definition of a lipid is “a compound that is insoluble in water but soluble in organic solvents, such as chloroform, ethers and alcohols”; a definition that encompasses a huge variety of chemicals including sterols, carotenoids, polyhydroxylalkanoates, triacylglycerols (TAGs), and phospholipids. With the exception of their solvation properties, these chemicals share little in common with regard to their chemical structure or biosynthesis. Therefore, for the purposes of this article, the author will revert to a more limited (though ultimately more biologically relevant) definition of lipids as proposed by Bill Christie (one of the true giants of lipid biochemistry): “fatty acids and their derivatives, and substances related biosynthetically or functionally to these compounds”.2 Fatty acids can be considered the structural units of fats or lipids, in much the same way as amino acids are the structural units of proteins. However, unlike in proteins where the structural units (amino acids) are covalently bonded to one another by amide bonds, in lipids the fatty acids are covalently bonded to a glycerol molecule via an ester linkage (Fig. 1). There are a large variety of fatty acids that differ in their chemical structure and, therefore, in their chemical characteristics. The fatty acid composition, often referred to as the fatty acid profile, of a lipid has a large impact on the physicochemical properties of the lipid. Therefore, fatty acids play key roles in the function of lipids in biological cells and certain fatty acids play vital roles in human and animal development and health. Fatty acids, and the lipids they form such an important part of, also have commercially significant applications in the food industry (cooking oils/fats, cocoa butter, etc.) and chemical industry (detergents, surfactants, and lubricants).

1.04.2

Structure of Fatty Acids

Fatty acids are long-chain carboxylic acids (Fig. 2). Although formic acid is structurally the simplest example of this group, this along with short-chain carboxylic acids (fewer than six carbons) is not generally considered as fatty acid because of the polar nature of its acid group, means it lacks the hydrophobic properties associated with lipids. Once the carbon chain length exceeds approximately

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00005-7

39

40

Lipids, Fatty Acids

Glycerol backbone

Triacylglycerol

Phospholipid X O

Hydrophilic polar head group

H

H

H

H

C

C

C

H

H

OH OH OH

H

H

H

C

C

C

O

O

O

H

H

H

H

O

C

C

C

O

O

H

C=O C=O C=O Hydrophobic

C=O C=O

R1

R1

acyl chains

R2

R3

O−

O= P

H

R2

Figure 1 The major forms of fatty acyl lipids in the cell, triacylglycerols (TAGs) and phospholipids (PL) consist of fatty acids bound to a glycerol backbone.

Stearic acid, 18:0 H3C 18

OH

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

C

17

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

β

α

ω

ω−3

ω−6

Oleic acid, 18:1 Δ9 H3C 18

OH

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

C

17

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

β

α

ω

ω−3

ω−6

Linoleic acid, 18:2 Δ9, Δ12, 18:2n-6 H3C 18

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

C

16

15

14

13

12

11

10

9

8

7

6

5

4

3

2

1

β

α

ω−3

ω−6

α-Linolenic acid, 18:3 Δ9, Δ12,Δ15 18:3n-3 H3C 18

ω Figure 2 bonds.

CH2

CH2

17

16

CH2

CH2

CH2

15

14

13

ω−3

O

OH CH2

C

3

2

1

β

α

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

CH2

12

11

10

9

8

7

6

5

4

ω−6

O

OH

17

ω

O

O

The chemical structure of fatty acids. Fatty acids are long-chain carboxylic acids that may contain one or more carbon–carbon double

six carbons, then the hydrophobic nature of the hydrocarbon acyl chain predominates and the molecule takes on the hydrophobic feature characteristics of lipids. With the exception of some specialized fatty acids produced by certain prokaryotes, the fatty acid residues in lipids are straight acyl chains with a carboxylic acid group at one end and a methyl group at the other end. The structure of fatty acids varies in any one (or more) of three ways: (1) the number of carbon atoms in the carbon chain; (2) the number of carbon–carbon double bonds (C ] C) they contain; or (3) the position of these double bonds. As the length of the fatty acid increases, the melting point of the fatty acid increases (Table 1). Conversely, the addition of double bonds into the fatty acyl chain causes the melting point of the fatty acid to decrease. This is because the addition of a double bond introduces a kink into the otherwise linear acyl chain

Lipids, Fatty Acids Table 1

41

Fatty acid nomenclature and physical properties of selected fatty acids

Common name

Systematic name

Numerical designation

Melting point ( C)

Appearance at room temperature

Myristic acid Stearic acid Oleic acid a-Linolenic acid

Tetradecanoic acid Octadecanoic acid Octadeca-9-enoic acid Octadeca-9,12,15-trienoic acid

14:0 18:0 18:1 18:3 n-3

58.8 69.6 14 11

Solid (fat) Solid (fat) Liquid (oil) Liquid (oil)

and decreases the ability of the hydrophobic fatty acyl chains to pack closely together in an ordered crystalline structure. Although both chain length and possession of double bonds determine the melting temperature of the fatty acid, the impact of double bonds is far more pronounced (Table 1) and of greater biological significance. Although the majority of fatty acids found in nature have 16–18 carbon atoms, fatty acids with as few as six and as many as 28 carbons have been described.9 As it is the properties of the constituent fatty acids that determine, to a great extent, the properties of the lipids that they are constituents of, lipids rich in long and saturated fatty acids are solid at room temperature (and commonly described as fat), whereas lipids rich in shorter, and particularly, more unsaturated fatty acids are liquid at room temperature (commonly referred to as oils). Double bonds in fatty acids are inserted by enzymes called desaturases. As is typical for enzymes, desaturases are very specific as to the form and position of the double bonds they introduce into fatty acyl chains. The double bonds are all in the cis configuration (Fig. 3) and are inserted at specific positions along the fatty acid (Fig. 2). Double bonds are spaced at three carbon intervals so that between each double bond there is a carbon atom that possesses no double bonds. This saturated carbon atom is referred to as the methylene carbon, and therefore, the spacing of double bonds in fatty acids is often described as methylene interrupted. This strict spacing in neighboring double bonds means that once the number of double bonds that a fatty acid contains and the position of the final double bond (from either end) are known, the position of the other double bonds can be inferred. The position of the double bond closest to the methyl end of the fatty acid is normally used in fatty acid designation, as the position of this double bond does not alter when the fatty acid undergoes elongation. During fatty acid elongation, the length of the fatty acyl chain is extended by the addition of an acetyl (two carbon) group to the carboxyl end of the fatty acid; as a result, the position of preexisting double bonds moves (by two carbons) relative to the carboxylic acid group. The terminal methyl carbon of a fatty acid is designated the u carbon, as it occupies a position on the acyl chain the furthest from the a carbon (which is adjacent to the carboxylic acid group) (Fig. 2). Based on the position of the double bond closest to the u carbon, unsaturated fatty acids commonly found in nature can be divided into three major series: u-3, u-6, and u-9. In more modern fatty acid nomenclature, the u has been superseded by n, and, hence, fatty acids fall into the n-3, n-6, and n-9 series (Fig. 4). While n-9 fatty acids are of limited importance, the n-6 and n-3 fatty acids are widely distributed in animal, plant, and microbial lipids. Furthermore, some of these fatty acids play key roles in human growth, development, and health. It is important to note that the public awareness of the health benefits of dietary polyunsaturated fatty acid (PUFA) has resulted in a concerted effort on the part of food manufacturers to supplement (or highlight) the n-3 content of many foods. In this context, the older designation remains in place so consumers are informed of, and discuss, the u-3 or omega three content of foods.

1.04.3

Nomenclature

The naming of fatty acids can be a somewhat confusing topic as a number of names can be (and routinely are) given to any given fatty acid. In many cases, these names are used interchangeably by lipid chemists, even within the same document. The names given can be divided into three basic categories: (1) the systematic name (derived from the chemical structure); (2) the common name; and (3) a numeric designation (Table 1). While the common names of fatty acids are generally easy to remember, they do not give, in most cases, any direct information about the chemical nature of the fatty acid. These names are, by tradition, derived from the source from which the fatty acid (or a closely related fatty acid) was originally isolated, for example, palmitic acid (hexadecanoic acid, 16:0) from palm oil and arachidic acid (eicosanoic acid, 20:0) from peanuts (Arachis hypogeal). In contrast, the systematic name of a fatty acid gives all the information about the chemical structure of the fatty acid but these names are often long and complicated, especially when dealing with long-chain PUFAs (i.e., those with multiple double bonds). For the student of fatty acid (bio)chemistry, the numeric designation is undoubtedly the easiest to understand. This nomenclature takes the form X:Y n-Z, where X is the number of carbons in the acyl

trans double bond R1

H

cis double bond H

H

H C=C

C=C R2

R1

R2

Figure 3 cis and trans double bonds. Although carbon–carbon double bonds can be either cis or trans, those found in naturally occurring fatty acids are all in the cis configuration.

42

Lipids, Fatty Acids

Acetyl-CoA

ACC

Malonyl-CoA

Fatty acid synthase

Elongase

16:0

18:0 Δ9 desat

18:3 (n-3) Δ9,Δ12,Δ15

Δ15 desat

Δ6 desat

18:4 (n-3) Δ6,Δ9,Δ12,Δ15 Elongase

20:4 (n-3) Δ8,Δ11,Δ14,Δ17 Δ5 desat

20:5 (n-3) Δ5,Δ8,Δ11,Δ14,Δ17 Elongase

22:5 (n-3) Δ7,Δ10,Δ13,Δ16,Δ19

18:2 (n-6) Δ9,Δ12 Δ6 desat

18:3 (n-6) Δ6,Δ9,Δ12 Elongase

20:3 (n-6) Δ8,Δ11,Δ14 Δ5 desat

20:4 (n-6) Δ5,Δ8,Δ11,Δ14

Δ12 desat

18:1 (n-9) Δ9 Δ6 desat

18:2 (n-9) D6,Δ9 Elongase

20:2 (n-9) Δ8,Δ11 Δ5 desat

20:3 (n-9) Δ5,Δ8,Δ11

Elongase

22:4 (n-6) Δ5,Δ8,Δ11,Δ14

Δ4 desat

22:6 (n-3) Δ4,Δ7,Δ10,Δ13,Δ16,Δ19 n-3 series Figure 4

n-6 series

n-9 series

Overview of polyunsaturated fatty acid (PUFA) biosynthesis.

chain, Y is the number of double bonds contained in the acyl chain, and Z is the number of carbons between the last double bond and the methyl group and therefore informs the reader which series (n-3 or n-6) the fatty acid belongs to.

1.04.4

Form in the Cell

Fatty acids are critical components of the biological cell. However, in the form of free fatty acids, they are toxic as they have the ability to disrupt biological membranes; as a result, the levels of free fatty acids are kept vanishingly low in living cells and fatty acids occur as esters in combination with other molecules. During synthesis and trafficking between cellular compartments, fatty acids exist as thio-esters of either coenzyme A (CoA) or acyl carrier protein (ACP) (Fig. 5). However, the vast majority of the fatty acids in any cell exists as an acyl residue esterified to one of the three hydroxyl groups of a glycerol molecule to form one of two classes of compounds (Fig. 1). If all the hydroxyl groups are occupied by fatty acyl residues, the resulting compound is a TAG. TAG is a hydrophobic molecule that will coalesce with other TAG molecules to form an oil droplet (oil body) within the cell. Alternatively, the third or sn-3 hydroxyl group of the glycerol molecule can be occupied by a polar compound (Fig. 1). Commonly, this polar compound contains a phospho group, called the head group, and the resultant lipids are called phospholipids. Because of the presence of the polar phosphate-containing head group (Fig. 1), phospholipids are not entirely hydrophobic but display amphiphilic properties. Amphiphilic refers to the fact that phospholipids possess a polar region that is hydrophilic (attracted to water) and a region that is nonpolar and hydrophobic (repelled by water). Due to these contrasting portions of the molecule, phospholipids have the propensity to assemble into liposomes in aqueous environments, with the polar head groups on the outside in contact with the aqueous phase and the fatty acyl chains forming a hydrophobic core (Fig. 6).

1.04.5

What Do Lipids Do?

The two dominant forms of fatty acyl lipids in the cell, phospholipids and TAGs, have very different chemical properties and, as a result, have different biological functions. Phospholipids form the lipid bilayer of biological membranes (Fig. 6), the polar groups orienting themselves toward the aqueous environment (the cytosol or the extracellular milieu) while the hydrophobic fatty acyl residues bury themselves in the interior of the biological membrane. As a major part of biological membranes, phospholipids play a number of vital roles in cell function. Many biological processes are membrane associated, taking place either in, on, or across biological membranes (photosynthesis, generation of adenosine triphosphate (ATP), ion transport, membrane receptors, etc.). Such processes depend (for their activity) on

Lipids, Fatty Acids

Acetyl-CoA

Acetyl-ACP

R

R

C

C

O

S

CH2

CH2

CH2

CH2

NH

NH

C Phosphopantetheine

Acylgroup

O

S

O

C

3⬘ ADP

O

CH2

CH2 CH2

CH2

NH2

NH

NH N

C

O

HO

C

H

H 3C

C

CH3O

H2C

O

C

O

HO

C

H

H3C

C

CH3O

N

P

O O

O–

P

N

N O

CH2

O

O–

O

–

P

H2C H

H H

O

P

NH O

O–

CH2

CH C

O Serine residue of holo-ACP

H OH

O O–

O

Figure 5

43

The structure of acyl-CoA and acyl-ACP, a soluble form of fatty acids that are used in many fatty-acid-requiring reactions.

Liposome structure

Phospholipid structure

Polar (hydrophilic ) head group Nonpolar (hydrophobic ) fatty acyl residues

Section of lipid layer

Polar head groups arranged to form a hydrophilic shell in contact with water

Polar head groups on the outside of the lipid bilayer in contact with water

Fatty acyl chains inside the liposome forming a hydrophobic core

Fatty acyl chains inside the bilayer forming a hydrophobic core

Figure 6 Phospholipids. Due to their amphiphilic nature, these lipids form liposomes and as a lipid bilayer play a structural role in biological membranes.

44

Lipids, Fatty Acids

the physicochemical properties of the membrane with which they are associated. As such, the properties of phospholipids are important factors in cell function. The properties of the phospholipids in turn (as mentioned above) are determined by the fatty acids that make up the lipophilic (hydrophobic) portion of the phospholipids. In particular, membrane fluidity plays a key role in regulating the activity of membrane-associated processes, and fluidity of lipid bilayers can be modified significantly by modest changes in the fatty acyl profile of the constituent phospholipids. An archetypal example of this phenomenon is the ability of free-living microorganisms to alter the fatty acid profile of their membrane phospholipid in response to changes in growth temperature.8 When the temperature is decreased, there is often a discernible enrichment in fatty acids with lower melting points (i.e., fatty acids with more double bonds), whereas a shift to higher growth temperatures engenders an opposite response with membrane phospholipids becoming more saturated. The other dominant form of fatty acyl lipids in cells is the TAGs that have three fatty acid residues esterified to the glycerol backbone. In contrast to phospholipids, in most cells, TAGs serve no specific physiological function other than acting as a storage molecule. TAG is particularly suited to the role of an energy and carbon store as it is a very concentrated form of energy; fat stores approximately twice the energy per unit mass as protein or starch. Furthermore, due to its hydrophobic nature, TAG can be stored without the concomitant storage of water. The hump of a camel is not, as the common myth would have us believe, full of water; instead, it is a large deposit of fat, which can weigh over 30 kg, and allows the camel to survive long periods without food. In some instances, deposits of fat (in the form of TAG) can play physiological roles, but these roles tend to be physical rather than biochemical in nature, including the fatty pads that form as a protective layer around the mammalian kidney and the layers of blubber in sea mammals that augment their storage role by also providing insulation and buoyancy to the producing organism. Lipids are stored in cells as oil droplets called oil bodies. Although it was considered for many years that the oil bodies formed due to simple coalescing of the hydrophobic TAG molecules, it is now understood that this is an oversimplification. Oil bodies are surrounded by a phospholipid monolayer and in many species possess oil body-specific proteins associated with their other surfaces.6 The oil body-associated proteins include both structural proteins (the oleosins of plants) that appear to be involved in the formation/degradation and stabilization of the oil bodies and synthetic enzymes associated with the production of storage lipid (e.g., diacylglycerol acyltransferase (DAGAT)). Furthermore, in many (though not all) oil-accumulating cells, oil bodies appear to have a specific size; as the oil content of the cell increases, the number of the oil bodies increases rather than a single huge oil body forming as would be expected if the formation was a purely physical phenomenon. This suggests that the formation of lipid bodies is likely to be, in most cells at least, an ordered and regulated process.

1.04.6

Biosynthesis of Fatty Acids and Lipids

1.04.6.1

Classical Synthesis

Fatty acids are synthesized from the key central metabolite acetyl-CoA, a C-2 unit (Fig.4), and, as such, the vast majority of fatty acids found in nature possess even numbers of carbon atoms in their backbone. However, the initial step in fatty acid is the carboxylation of acetyl-CoA (by acetyl-CoA carboxylase (ACC)) to form malonyl-CoA (Fig.4). This ATP-driven carboxylation further activates the acetyl unit of acetyl-CoA and allows the subsequent condensation reactions involved in fatty acyl chain elongation. The biochemistry of saturated fatty acid is a ubiquitous process in living cells, covered extensively (and comprehensively) in a number of basic biochemistry textbooks. Synthesis of saturated fatty acids (i.e., containing no carbon–carbon double bonds) is catalyzed by a single large enzyme complex, fatty acid synthase (FAS). There are two distinct classes of FAS, type I and type II, which differ in their physical organization rather than the chemistry they carry out. Type I FAS is found in animals and nonphotosynthetic eukaryotic microorganisms (yeast, fungi, and some microalgae) and is composed of seven catalytic domains found on a single protein. Type I FAS commonly forms homodimers with a head-to-tail orientation where the fatty acyl chain intermediate passes between the two protein subunits during synthesis. Prokaryotes and photosynthetic organisms (plants and some algae) possess type II FAS, which is composed of multiple separate proteins, each catalyzing a single step in the fatty acid synthesis process and being analogous to a domain of the type I FAS. Both forms of FAS catalyze the same repeated elongation of a fatty acid chain (from an initial acetyl primer) with acetyl units to produce a saturated fatty acid (the chain length of which can vary from 6 to 16) (Fig. 7). During each round of elongation, the nascent fatty acyl chain is covalently linked (via a thiol ester bond) to a phosphopantetheine arm of an ACP domain within the FAS complex. This phosphopantetheine arm is analogous to the phosphopantetheine moiety of CoA and provides a flexible molecular arm that can deliver the growing acyl chain to each of the active sites involved with each round of fatty acid synthesis. Each cycle of fatty acid synthesis is initiated by condensation reaction between the FAS-bound growing saturated acyl chain (2–14 carbons long) and malonyl-CoA, which results in the release of CO2 and the extension of the acyl chain by two carbons. The resulting keto group-containing compound is reduced (by a keto-reductase) to form a hydroxyl group which is dehydrated (by a dehydratase) to form a trans double bond. The final catalytic step in each round is then the reduction of the trans double bond to form a saturated acyl chain that is transferred to the active site of the FAS condensing enzyme domain to initiate another round of fatty acid synthesis. FAS usually release the growing acyl chain after seven elongation cycles (although fatty acids are released with shorter chain lengths) to yield a 16-carbon saturated fatty acid, palmitic acid (16:0). This is achieved by cleavage of the acyl chain/phosphopantetheine arm by a fatty acyl thioesterase (in mammalian systems) or a fatty acid thiolase (in bacteria/plants) yielding a fatty acyl-CoA or a free fatty acid (which is rapidly esterified to an acyl-ACP), respectively.

Lipids, Fatty Acids

O

O HO

C

O HO

O R

CH2

C

C

CH2

C

O

CoA

Malonyl-CoA transacetylase O CH2

C

O

ACP

CO2

O CH2

C

O

ACP

45

O

Condensing enzyme

R

CH2

O

C

E

Keto-reductase OH CH2

R

C

Transfer of extended acyl chain to active site of FAS condensing enzyme

O CH2

C

O

ACP

H Dehydratase

R

CH2

C

H

O

C

C

O

Enoyl-reductase O

ACP

R

CH2

CH

CH

C

O

ACP

H Figure 7 Fatty acid synthase (type I and type II) catalyzes the synthesis of fatty acids via iterative cycles of acyl chain extension, involving condensation, keto-reduction, dehydration, and enol–reduction reactions.

Once formed, the saturated fatty acid can undergo a series of further conversions (each catalyzed by a specific enzyme) that convert saturated fatty acids into a host of structurally and chemically distinct fatty acids. The enzymes that carry out these postFAS conversions are oxygen-requiring desaturases (which insert cis double bonds; Fig. 3) and elongases (which add further acetyl – two carbon – units to the carboxyl end of the nascent fatty acid). Differences in the possession of different desaturase and elongases result in different organisms having different inherent capacities for synthesizing fatty acids. Immediately post synthesis, the saturated fatty acids exist as either a fatty acyl-CoA ester in nonplant eukaryotic cells or a fatty acyl-ACP in plant and prokaryotic cells. Although plants are clearly eukaryotic (possessing nuclei and organelles), their fatty acid synthetic machinery is undoubtedly prokaryotic in character. Indeed, the de novo synthesis of saturated fatty acids in plants occurs in a specialized organelle, the plastid, which is the evolutionary remnant of a prokaryotic endosymbiont (just like other organelles, mitochondria, and chloroplasts). Irrespective of whether the saturated fatty acid was formed by a type I or type II FAS, the first conversion to an unsaturated fatty acid involves the insertion of a cis double bond between carbons 9 and 10 (counting from the carboxylic acid group) (Fig. 2). This reaction is catalyzed by a so-called D9 desaturase. In plants, this enzyme is plastidal and accepts the fatty acyl-ACP (soluble) substrate. However, in animals and fungi, the D9 desaturase and its substrate (a phospholipid associated fatty acid) are membrane bound; this involves the transfer of the fatty acyl chain from the CoA to the sn-2 position of a phospholipid (a reaction catalyzed by an acyl transferase) prior to desaturation in the nonphotosynthetic eukaryotic cells. The next acyl conversion that typically occurs (predominantly to the C18 fatty acid 18:1, oleic acid) is another desaturation reaction, resulting in the formation of a double bond between carbons 12 and 13 of the fatty acyl chain (Fig. 4). This reaction (the D12 desaturation) is catalyzed in all systems by a membrane-bound enzyme utilizing a phospholipid bound fatty acid. The product of the D12 desaturase reaction is an 18carbon di-unsaturated fatty acid called linoleic acid, 18:2 D9,D12. Further conversion of linoleic acid can progress via one of two routes, as linoleic acid is the last common precursor to both the n-3 and n-6 fatty acid biosynthetic pathways (Fig. 4). In some plants and microorganisms, linoleic acid is converted to alpha (a)-linolenic acid (18:3 D9,D12,D15), the first (and therefore parent) fatty acid of the n-3 series, by a D15 desaturase. Linolenic acid (18:2 D9,D12 n-6) and a-linolenic acid (18:3 D9,D12,D15 n-3) are the parent precursors to the n-6 and n-3 series of fatty acids, respectively, and undergo an identical sequence of transformations to form the range of fatty acids outlined in Fig. 4. While desaturases act on phospholipid bound fatty acyl chains to introduce cis double bonds between adjacent carbon atoms, fatty acid elongases catalyze a more complex set of reactions that are more correctly thought of as a series of partial reactions catalyzed by a series of four enzymes that essentially repeat a single cycle of fatty acid synthesis: (1) condensation; (2) keto reduction; (3) dehydration; and, finally, (4) enol reduction (Fig. 8). This elongation process is complicated by the need for the fatty acyl chain to be removed from the sn-2 position of the phospholipid and esterified to a CoA in order for the condensation catalyzing activity to have access to the carboxylic acid group, and once the elongation process is complete, to transfer the newly extended acyl chain back

46

Lipids, Fatty Acids

R

CH2

O

O

O C

O

HO

PL

CH2

C

C

Acyltransferase

R

CH2

O

O

O C

O

HO

CoA

O CoA Transfer of acyl group from malonyl-CoA to ACP domain of elongase complex

CH2

C

C

O

E O

Condensing enzyme elongase O CH2

R

CH2

O CH2

C

C

O

CH2

R

O CH2

C

C

CH

C

O

PL

Acyltransferase

Keto-reductase OH

CH

E

O O

E

R

CH2

CH

CH

C

O

CoA

H Thiolase/acyl-CoA synthase or thioesterase

Dehydratase

R

CH2

C

H

O

C

C

O

Enoyl-reductase O

E

R

CH2

CH

CH

C

O

E

H Figure 8 Fatty acid elongation is catalyzed by a distinct complex of enzymes that carry out a single fatty acyl elongation cycle analogous to that catalyzed by the fatty acid synthase.

onto a phospholipid (at the sn-2 position) for subsequent desaturation. Despite the complicated reaction series and multiple enzyme activities involved in the fatty acid elongation process, the term elongase is used (in fatty acid biochemistry) to describe solely the enzyme that catalyzes the initial (and substrate specific) condensation reaction. From heterologous expression experiments, it appears that the condensation enzyme plays the critical role in fatty acid elongation and that the other enzyme activities involved often demonstrate a remarkable flexibility in their activities. In several studies, the heterologous expression of a condensation portion of an elongation system is sufficient to obtain expression of that elongation activity in the host cell line.3,4 It is presumed that the host enzymes involved with endogenous fatty acid elongation are capable of catalyzing the remaining partial reactions of the elongation cycle. Due to the toxic nature of free fatty acids to cells, the concentration of free fatty acids in the cell is maintained at very low levels. Fatty acids exist as (and are very rapidly converted between) soluble thio-esters (acyl-CoAs or acyl-ACPs), membrane (phospholipids or glycolipids) lipids, or insoluble TAGs (Fig. 1). While fatty acid desaturases generally utilize a phospholipid-bound substrate, the fatty acid elongases require an acyl-CoA substrate. Consequently, the synthesis of long-chain PUFAs involves the iterative migration of the nascent fatty acyl chain between the membrane-bound phospholipid (the site of desaturation) and the cytosolic acyl-CoAs (the substrate for fatty acid elongation). Animals generally do not possess either a D12 or a D15 desaturase and are therefore incapable of synthesizing either linoleic acid or a-linolenic acid de novo from glucose. However, despite this metabolic deficiency, animals have a critical requirement for long-chain PUFAs of both the n-3 and n-6 series. As such, animals have a dietary requirement for fatty acids containing D12 and D15 double bonds and both linoleic and a-linolenic are described as essential fatty acids. In effect, these two fatty acids are vitamins, as without a sufficient dietary supply a deficiency would result. Under normal circumstances, however, these two fatty acids are sufficiently common in plant tissues (and therefore tissues of herbivores) that no acute fatty acid deficiency is ever manifested in nature. An animal’s diet would need to be so poor to achieve deficiency that the animal would inevitably perish of some other nutrient deficiency before essential fatty acid deficiency was achieved. Despite animals lacking D12 and D15 desaturases, they do possess a range of the so-called frontend desaturases (inserting double bonds at or before the 9–10 carbons in the acyl chain) and a number of elongases that are capable of synthesizing PUFAs with chain lengths from 18 to 22 carbons. As a consequence, as long as animals obtain sufficient levels of 18:2 n-6 and 18:3 n-3 in the diet they are capable of synthesizing a host of long-chain PUFAs, the range and relative composition of which will be determined by both the animal‘s physiological capacity (the enzyme activities they possess) and their diet (the fatty acid precursors they ingest). Although both animal and some microbial cells are capable of producing the longest and most unsaturated fatty acid commonly found in nature (docosahexaenoic acid (DHA), 22:6 n-3), the biosynthetic pathways they use are distinctly different. Several microbial systems have been identified that possess a D4 desaturase activity, capable of direct conversion of DPA n-3 (22:5n-3)

Lipids, Fatty Acids

47

22:5 (n-3) Δ7,Δ10,Δ13,Δ16,Δ19

Cytoplasm Elongase

24:5 (n-3) Δ9,Δ12,Δ15,Δ18,Δ21 Δ6 desat Sprecher pathway (animal route)

24:6 (n-3) Δ6,Δ9,Δ12,Δ15,Δ18,Δ21

Δ4 desat Direct desaturation (microbial route)

22:6 (n-3) Δ4,Δ7,Δ10,Δ13,Δ16,Δ19 Translocation out of peroxisome Peroxisome

Translocation into peroxisome

22:6 (n-3) Δ4,Δ7,Δ1,Δ13,Δ16,Δ19 One round β -oxidation 24:6 (n-3) Δ6,Δ9,Δ12,Δ15,Δ18,Δ21

Figure 9 The Sprecher pathway in animal cells is responsible for the synthesis of DHA (22:6 n-3) from DPA (22:5 n-3) in the absence of a D4 desaturase.

to DHA (22:6n-3) (Fig. 4). This enzyme activity is not found in animals. Animal cells complete the same D4 desaturation via a more elaborate set of reactions often referred to as the Sprecher pathway (Fig. 9).7 In the Sprecher pathway, DPA n-3 (22:5 n-3) is first elongated to 24:5n-3; this elongation allows a D6 desaturase to insert a carboxyl end desaturation to generate 24:6 n-3. Subsequent chain shortening via a single round of the peroxisomal b-oxidation cycle then generates 22:6n-3 DHA. As such, the conversion of DPAn-3, the nominal direct precursor to DHAn-3, is anything but simple in animals. This conversion involves elongation systems, desaturation activity, intermediate transport to the peroxisomes for chain shortening, and then transportation of the DHA back into the cytosol for final incorporation into phospholipid or TAG. The complexity of this cycle results in the formation of DHA from precursor fatty acids being relatively limited in animals (almost nonexistent in humans). As a result, preformed DHA is an important nutrient for most animal species. In contrast to animals, many plant species possess both D12 and D15 desaturases and therefore are capable of synthesizing significant quantities of 18:2 n-6 and 18:3 n-3 (the fatty acids essential for animal health) de novo. However, plants are deficient in elongases required for the synthesis of PUFA with acyl chains greater than 18 carbons. Therefore, plants do not synthesize PUFA with either 20 or 22 carbons and preformed C20 and C22 PUFA, the very long chain PUFA (VLC-PUFA), can only be sourced from animal or microbial lipids. As plants obtain their energy from sunlight and their carbon from fixing CO2 (rather than the ingestion of preformed complex organic compounds such as animals), their fatty acid profiles reflect solely the organism‘s biochemical capacity for fatty acid biosynthesis. This has two consequences: (1) the fatty acid profiles of plant species are simpler, containing fewer different fatty acid residues and (2) are generally rich in one or two particular fatty acids. While most prokaryotic microorganisms do not produce any PUFA in their fatty acids (exceptions being some cyanobacteria and some marine bacteria) and no significant amounts of storage lipid in the form of TAG, many eukaryotic microorganisms (fungi and algae) have been known for many years to produce appreciable amounts of both cell lipid and PUFA. Eukaryotic microorganisms therefore combine the characteristics of plants and animals in that they can display relatively simple fatty acid profile rich in one (or just a few) specific fatty acids and also have the biosynthetic capacity to produce a wide range of PUFA, including VLC-PUFA (up to 28 carbons in length) de novo.

1.04.6.2

Polyketide Route

Due to the oxygen requirement of the fatty acyl desaturases, it was, until relatively recently (the past 10 years), a universally accepted fact that PUFA biosynthesis was an aerobic process and that it was restricted to a large degree to eukaryotic systems, cyanobacteria being the prokaryotic odd balls that produced PUFA. This paradigm was challenged by the discovery in the late 1990s of marine bacteria, isolated from fish intestine that displayed the ability to synthesize some of the longest, most unsaturated fatty acids found in nature, eicosapentaenoic acid (EPA) and DHA, and that these bacteria were capable of this synthesis under anaerobic conditions. Even more astounding was the fact that these bacteria were found to produce EPA and DHA without any accumulation of detectable precursor fatty acids, which are usually detected in other systems. Further study identified a single stretch of DNA that when introduced (via genetic transformation) into the model bacteria Escherichia coli bestowed on this host the ability to reproduce the synthesis of EPA and/or DHA without the accumulation of any fatty acid intermediates. It was assumed that this DNA encoded all the desaturase and elongase enzymes required to produce 20:5 n-3 and 22:6 n-3 from either 16:0 or 18:0. Although nucleotide

48

Lipids, Fatty Acids

sequence analysis did detect sequences with similarity to enzymes associated with fatty acid synthase or elongases, no desaturaselike sequences were identified. Over a few years spanning the millennium, the mystery deepened when a group of marine diatoms (the thraustochytrids) were identified that possessed very similar biosynthetic capability to produce DHA and other VLC-PUFA to the previously isolated marine bacteria. These eukaryotic marine microorganisms were shown to possess biosynthetic machinery very similar to that originally identified in the PUFA-synthesizing marine bacteria. The solution to this apparent conundrum came when it was realized that the bacterial genes encoded not a complex of fatty acid elongases and enigmatic desaturases but, instead, encoded a novel PUFA synthase that more closely resembled a polyketide synthase (PKS).5 This protein complex is capable of synthesizing PUFAs directly from acetyl-CoA or malonyl-CoA without the formation of a saturated fatty acid intermediate. PKSs are closely related to FAS, being large enzyme complexes that catalyze the same series of reactions (see above) and which can be composed of either single multidomain proteins or a complex of separate enzymes. The important differentiating feature between FAS and PKS is that, in the latter, some or all of the condensation cycles omit some or all of the keto/enol reduction and dehydration processing steps, leaving some keto, hydroxyl, or enol groups intact in the final product. Polyketides take their name from the numerous keto groups that are found in many members of this chemical group. Like PKS, PUFA synthases do not carry out the full series of keto reduction/dehydration cycle and as a result leave some of the carbon–carbon double bonds formed during the elongation cycle intact. However, in a process that is far from fully understood, the PUFA synthases do not simply leave the carbon–carbon double bonds intact in the fatty acyl chain during the elongation cycle, as these double bonds would be in the trans configuration and spaced at two (rather than the required three) carbon intervals. Instead, the PUFA synthases carry out a trans/cis isomerization and a bond migration to ensure that the double bonds in the resultant PUFA product are in the correct cis, methyleneinterrupted form that is common for PUFA produced from the classical pathway. The biosynthetic steps catalyzed by the PUFA synthase have not been fully elucidated and the exact nature of all of the intermediates remains to be unequivocally identified. Fig. 10 outlines one reaction scheme that is consistent with the product profile associated with known PUFA synthases, although it is based more on the author‘s knowledge of lipid biochemistry than on definitive experimental data. Sequence analysis of these PUFA synthases from marine bacteria and marine microalgae strongly suggests that the genes for the PUFA synthases have been horizontally transferred from the bacteria to the algae.

1.04.7

Biochemistry of Lipid Accumulation

Although the overwhelming majority of living cells have the ability to produce a range of fatty acids, not all accumulate significant quantities of storage lipid (in the form of TAG/neutral lipid). The production of fatty acids and their incorporation into phospholipids are physiologically necessary for biological membrane synthesis/function and required for cell growth and, therefore, is part of cellular primary metabolism. In contrast, the accumulation of storage lipid is not a physiological necessity, is not universal among biological systems, and is an example of secondary, or overflow, metabolism. While significant storage lipid is relatively restricted (although not unknown) among prokaryotes,1 it is more commonly a feature of eukaryotic cells. Even among eukaryotes, however, the ability to accumulate substantial amounts of cell lipid is not universal. Some species do not elaborate lipid as a storage material. These species are designated nonoleaginous and, irrespective of growth conditions, contain less than 25% (dry weight) as lipid – the majority of which is phospholipid. In contrast, oleaginous cells contain greater than 25% (dry weight) cell lipid under suitable conditions, the majority being in the form of storage TAG. Cell lipid contents of >60% dry weight have been recorded in certain microbial species. In oleaginous cells, biochemical processes have evolved to divert excess carbon into fatty acid synthesis to support the extensive storage lipid accumulation. In accordance with the description of storage lipid as a secondary metabolite, significant lipid accumulation is initiated only after cell growth has been halted due to a nutrient other than carbon (usually nitrogen) being depleted and when a continued supply of carbon is available. Under these conditions, the citric acid cycle (tricarboxylic acid cycle (TCA)) is downregulated, but in many organisms glucose uptake and glycolysis continue requiring a suitable fate for the metabolized sugar. Nonoleaginous organisms convert this carbon into an array of compounds including intracellular carbon/energy storage compounds such as starch or glycogen or secrete it as organic acids, alcohols, or other secondary metabolites (including antimicrobial compounds). Oleaginous organisms convert the majority of the metabolized carbon source into intracellular lipid, which can be utilized as an intracellular source of both carbon and energy, should the environmental carbon source become limiting. In oleaginous yeast and fungi (as in animal cells), lipid synthesis occurs in the cytosol, where the FAS complex is located. Lipid accumulation in yeasts and fungi has been extensively studied and is therefore well documented. The impetus for the study of lipid accumulation in these microbial systems was twofold. First, yeast and fungi are amenable to large-scale cultivation and are known to produce a wide array of fatty acids in large quantities. These qualities make yeast and fungi attractive potential production organisms for commercial manufacture of specialty fats/oils. Second, as the organization of lipid biochemistry is similar to that in higher animals, yeasts/fungi constitute good model systems for the processes that underlie lipid accumulation in animal cells. Fatty acid and lipid synthesis starts with the production of the primary substrate acetyl-CoA, which is produced from pyruvate as part of central metabolism by the action of pyruvate dehydrogenase. However, this reaction occurs in the mitochondria, and, as acetyl-CoA is too large a molecule to pass unaided across the inner mitochondrial membrane, active processes are required to export acetyl-CoA into the cytosol for lipid (and other metabolite) synthesis.

Lipids, Fatty Acids

49

Figure 10 A proposed scheme for the synthesis of polyunsaturated fatty acids by a PUFA synthase, the reaction sequence requires three separate reaction schemes: cycles 1, 2 and 3 to operate in a defined sequence to produce DPA (22:5 n-6), EPA (20:5 n-3), and DHA (22:6 n-3).

Acetyl-CoA is not transported across the mitochondrial membrane directly but instead is metabolized to citrate via its aldol condensation with oxaloacetate (as part of the citric acid cycle) in a reaction catalyzed by citrate synthase (Fig. 11). Under conditions of unrestricted growth (i.e., in the presence of all nutrients required for cell growth and doubling), the citrate formed would be further metabolized, via aconitate to isocitrate (a reaction catalyzed by aconitase) and then to a-ketoglutarate by the nicotinamide adenine dinucleotide (NADH)-generating (and decarboxylating) enzyme isocitrate dehydrogenase (NADHdependent isocitrate dehydrogenase (NAD-ICDH)) then further round the TCA cycle to finally reform oxaloacetate for condensation with another acetyl-CoA molecule. In the absence of a suitable nitrogen source, however, the activity of NAD-ICDH is greatly downregulated, essentially shutting down carbon flux through the citric acid cycle. The restriction of isocitrate conversion results in an accumulation of intra-mitochondrial citrate (rather than isocitrate), as the isomerization of citrate to isocitrate is a reversible reaction with an equilibrium strongly favoring the formation of citrate. Buildup of mitochondrial citrate initiates the export of citrate, via a citrate malate antiport (exchanging mitochondrial citrate for cytosolic malate), across the mitochondrial membrane and into the cytosol. In the cytosol, the citrate is cleaved by ATP citrate lyase, an enzyme that has been shown to be crucial for lipid accumulation in oleaginous eukaryotic microorganisms. The cleavage of citrate regenerates oxaloacetate and acetyl-CoA (but in this case in the cytosol), providing acetyl-CoA for fatty acid biosynthesis

50

Lipids, Fatty Acids

Cytosol

Mitochondrial matrix

Glucose Malate

Fumarate

Glycolysis Oxaloacetate Pyruvate

Succinate

Pyruvate PDH

CS

Succinyl-CoA

Acetyl-CoA ME

α-ketoglutarate

Citrate NADP+

Aconitase MDH

Malate

Isocitrate

NAD-ICDH

1

Oxaloacetate Citrate ACL NADPH

Acetyl-CoA

Fatty acid synthesis Figure 11 Biochemical pathways for the diversion of acetyl-CoA from primary metabolism into lipid accumulation in oleaginous eukaryotes. CS, citrate synthase; NADH ICDH, NADH-dependent isocitrate dehydrogenase; PDH, pyruvate dehydrogenase; MDH, malate dehydrogenase; ME, malic enzyme; ACL, ATP citrate lyase; 1, citrate:malate antiport.

and potentially malate (via the action of cytosolic malate dehydrogenase) for exchange for more mitochondrial citrate. An alternative fate for the cytosolic malate produced by the combined activity of ATP citrate lyase and cytosolic malate dehydrogenase is decarboxylation by cytosolic malic enzyme to generate pyruvate and NADPH, thus providing the reducing power required by the fatty acid synthase (for both keto reduction and enoyl reduction). Although the decarboxylation of cytosolic malate by malic enzyme appears to be important in providing reducing power for fatty acid synthesis,10 the stoichiometry of this process is unable to account for all the NADPH required for fatty acid synthesis, the remainder being supplied, presumably, by the cytosolic hexose monophosphate pathway.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Alvarez, H. M.; Steinbuchel, A. Triacylglycerols in Prokaryotic Microorganisms. Appl. Microbiol. Biotechnol. 2002, 60, 367–376. Christie, W. W. Gas Chromatography and Lipids. A Practical Guide, The Oily Press: Somerset, 1989. Leonard, A. E.; Kelder, B.; Bobik, E. G.; et al. Identification and Expression of Mammalian Long-chain PUFA Elongation Enzymes. Lipids 2002, 37, 733–740. Niu, Y.; Kong, J.; Fu, L.; et al. Identification of a Novel C20-elongase Gene from the Marine Microalgae Pavlova viridis and its Expression in Escherichia coli. Mar. Biotechnol. 2009, 11, 17–23. Metz, J. G.; Roessler, P.; Facciotti. Production of Polyunsaturated Fatty Acids by Polyketide Synthases in Both Prokaryotes and Eukaryotes. Science 2001, 293, 290–293. Sarmiento, C.; Ross, J. H.; Herman, E.; Murphy, D. J. Expression and Subcellular Targeting of a Soybean Oleosin in Transgenic Rapeseed. Implications for the Mechanism of Oil-body Formation in Seeds. Plant J. 1997, 11, 783–796. Sprecher, H. Metabolism of Highly Unsaturated N-3 and N-6 Fatty Acids. Biochimica Biophysica Acta 2000, 1486, 219–231. Suvtari, M.; Liukkonen, K.; Laakso, S. Temperature Adaptation in Yeast: The Role of Fatty Acids. J. General Microbiol. 1990, 136, 1469–1474. Van Pelt, C. K.; Huang, M. C.; Tschanz, C. L.; Brenna, J. T. An Octaene Fatty Acid, 4,7,10,13,16,19,22, 25-octacosaoctaenoic Acid (28:8n-3), Found in Marine Oils. J. Lipid Res. 1999, 40, 1501–1505. Wynn, J. P.; bin Abdul Hamid, A.; Ratledge, C. The Role of Malic Enzyme in the Regulation of Lipid Accumulation in Filamentous Fungi. Microbiology 1999, 145, 1911–1917.

1.05

Structure and Biosynthesis of Glycoprotein Carbohydrates

M Crispin and CN Scanlan, University of Oxford, Oxford, United Kingdom TA Bowden, Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, United Kingdom © 2011 Elsevier B.V. All rights reserved. This is a reprint of M. Crispin, C.N. Scanlan, T.A. Bowden, 1.07 - Structure and Biosynthesis of Glycoprotein Carbohydrates, Editor: Murray MooYoung, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 73-90.

1.05.1 Introduction 1.05.2 Monosaccharide Structure 1.05.2.1 Absolute Configuration 1.05.2.2 Stereochemical Projections of Carbohydrates 1.05.2.3 Structures of Common Monosaccharides 1.05.2.4 Anomericity 1.05.2.5 Stereoelectronic Effects 1.05.3 Oligosaccharide Structure 1.05.3.1 Torsion Angles of Glycosidic Linkages 1.05.3.2 Oligosaccharide Nomenclature 1.05.4 Biosynthesis of Glycoproteins 1.05.4.1 N-Linked Glycans: Glycoprotein Folding and Processing 1.05.4.2 O-Linked Glycans: Glycoprotein Folding and Processing 1.05.5 Glycosylation of Therapeutic Glycoproteins 1.05.5.1 Overview 1.05.5.2 Case Study: Structure of Therapeutic Antibody Glycoforms Acknowledgments References Relevant Websites

51 53 53 53 54 55 56 57 58 58 60 60 62 65 65 65 67 68 68

Glossary Antigen-dependent cellular cytotoxicity (ADCC) A process in which immune cells expressing Fc receptors recognize and kill antibody-coated target cells. Complex-type glycan An N-linked glycan containing three mannose residues, three or more N-acetylglucosamine residues, with the potential to contain other types of residues. Diastereoisomers Stereoisomers that are not enantiomers. Enantiomerism A phenomenon whereby two nonsuperimposable molecules are related like mirror images. Epimers Diastereoisomers that differ in configuration at only one chiral center. Glycan The generic term for a carbohydrate group. Glycosylation The formation of linkages between glycans and proteins or lipids. Mutarotation The interconversion of epimers at the anomeric carbon. N-linked glycan A glycan linked to the nitrogen of the side chain of asparagines in the consensus sequence NXS/TX, where X is any amino acid except proline. Oligomannose-type glycan An N-linked glycan composed of mannose and reducing terminus N-acetylglucosamine. O-linked glycan A glycan linked to the oxygen of the side chain of serine, threonine, hydroxylysine, or tyrosine. Unless otherwise stated, this usually refers to the modification of serine or threonine with a-N-acetylgalactosamine (a-GalNAc).

1.05.1

Introduction

In eukaryotes, the majority of cell surface and secreted proteins are covalently modified with carbohydrates. This type of posttranslational modification, glycosylation, is inherently complex exhibiting extensive chemical and conformational heterogeneity. Despite this complexity, there are common structural and biosynthetic principles. Here, we present an introduction to the different hierarchies of carbohydrate structure from stereochemistry of monosaccharides and their linkages to the structural diversity of complex mammalian glycosylation. These structures are discussed in the context of the glycan biosynthetic pathways, with an emphasis on N-linked glycosylation, and we outline the nomenclature with which these complex structures can be described.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00007-0

51

52

Structure and Biosynthesis of Glycoprotein Carbohydrates

We describe the biosynthesis and crystal structures of antibodies and their glycans to illustrate the conformational properties of glycoprotein carbohydrates and how therapeutic antibodies are being developed by modulating these glycans. The range of posttranslational modifications of proteins is both structurally and functionally diverse. The posttranslational modification of proteins with carbohydrates can have a substantial impact on the function.1–4 They range in size from single monosaccharide modifications of cytoplasmic and nuclear proteins to the biosynthesis in the secretory system of proteoglycans containing repetitive polysaccharides of hundreds of residues destined for the extracellular matrix. The vast majority of cell surface proteins in eukaryotic cells are glycosylated. There are two principal types of glycoprotein glycosylation that can be differentiated according to the chemical linkage to the protein: N-linked through the side-chain nitrogen of asparagines residues and O-linked through the side-chain oxygen of typically serine or threonine residues.5 N-linked glycosylation occurs in the endoplasmic reticulum (ER) through the en bloc transfer of a 14-residue-glucosylated oligomannose-type glycan, Glc3Man9GlcNAc2, to the side chain of Asn residues in the consensus sequence NXS/TX, where X is any amino acid except proline. N-linked glycans are subsequently processed by waves of glycosidases in the ER and Golgi and by glycosyltransferases in the Golgi apparatus. By contrast, O-linked glycosylation is initiated in the Golgi apparatus and the biosynthesis occurs entirely by sequential addition of monosaccharide residues. Many glycosyltransferases of the Golgi apparatus can act on both N- and O-linked glycans, and thus many structural motifs can be present on both types of glycans. Although glycans impart physicochemical stability to proteins by shielding hydrophobic surface regions6,7 (Fig. 1), there are a myriad of biological roles reliant on carbohydrate–protein recognition events. From the first stages of glycoprotein biosynthesis, N-linked glycans can assist in protein folding in the ER through the recruitment of chaperones and can signal the degradation of misfolded glycoproteins. Further, intracellular functions include the trafficking of nascent glycoproteins through the secretory pathway including specific signals for lysosomal delivery. The diverse range of glycans in mammals arises from the extensive panel of glycosyltransferases in the Golgi apparatus, and detectable functions of the resulting oligosaccharides are usually associated with the extracellular roles. Such modifications are often cell specific and are important, for example, in cellular trafficking in the immune system. This is illustrated by the recruitment of neutrophils to the site of inflammation, a process that involves the display of a glycan motif, termed sialyl Lewis x, which is recognized by E-selectin of damaged epithelial cells. In contrast to the chemical complexity of mammalian glycosylation, which typically contains various arrangements of terminal sialic acid, galactose, fucose, and N-acetylglucosamine, the glycans of ‘lower’ eukaryotes are usually much more restricted.1 For example, yeast glycans are composed of an array of mannose residues. Similar mannose structures are present at the early stages of mammalian glycan biosynthesis but are largely hydrolyzed and capped by layers of monosaccharides added by Golgi-resident glycosyltransferases. An example of how these species-specific differences have been exploited by the immune system is the expression of a mannose receptor on macrophages that binds to mannosylated pathogenic material such as yeast glycoproteins but does not recognize host glycoproteins. Knowledge of the biological roles of glycans has led to the development of optimized glycoprotein therapeutics bearing defined types of glycans.8 From the above mentioned example, understanding of the selective uptake of mannosylated material by

1S4P

1JUH

2DTQ

2B8H

Figure 1 Glycan-protein interactions. Structures of glycoproteins containing ordered N-linked glycans as determined by X-ray crystallography. Protein is depicted in cartoon form (b-strands, yellow; a-helices, red; and loops, green) and glycans are depicted as sticks (carbon, green; oxygen, red; and nitrogen, blue). Electron density, of a 2Fo-Fc map, corresponding to the glycan moiety is shown as a blue mesh and is contoured at 1s. Protein Data Bank identification codes are shown.

Structure and Biosynthesis of Glycoprotein Carbohydrates

53

Figure 2 Stereochemical representation of glyceraldehyde. (A) Fisher (left) and wedge (right) representation of D-glyceraldehyde. (B) Enantiomers of glyceraldehyde with mirror plane shown as a dashed line. (C) Application of the Cahn–Ingold–Prelog (CIP) prioritization rules for the assignment of absolute configuration of the chiral center of glyceraldehyde. The CIP assignment relies on the prioritized ranking of substituent groups around the chiral center, the C2 of glyceraldehyde. The clockwise configuration of the priorities (first to third), when viewed with the lowest priority constituent pointed away from the viewer, corresponds to an R (recto) assignment, while anticlockwise corresponds to an S (sinister) assignment.

macrophages has enabled targeting of therapeutic glycoproteins to that cell type. This strategy has been used to improve the efficacy of enzyme-replacement therapy of Gaucher’s disease in which a defective b-glucocerebrosidase activity leads to an accumulation of glycolipids in macrophages. Administration of b-glucocerebrosidase manufactured with glycans terminating with mannose can enhance the uptake of the drug. Further examples of improved pharmacokinetics by glycan engineering include the recombinant expression of highly sialylated erythropoietin (EPO) to reduce the rate of glycoprotein clearance by the liver asialoglycoprotein receptor and optimization of therapeutic antibodies by glycan engineering that can modulate immune effector functions. In this introduction to glycoprotein structure and biosynthesis, we outline the different hierarchies of glycan structure and the nomenclature with which these complex structures can be described. We discuss the biosynthetic pathways of these glycans and the application of glycoprotein engineering to the optimization of therapeutic antibodies.

1.05.2

Monosaccharide Structure

1.05.2.1

Absolute Configuration

The precise stereochemistry of monosaccharides defines their biology9,10; for example, glucose can be converted to mannose by the inversion of only one stereochemical center. Historically, the configuration of carbohydrates has been defined by comparison with the most simple sugar, glyceraldehyde that contains an aldehyde, secondary hydroxyl, and a primary hydroxyl at C1–C3, respectively (Fig. 2). The configuration of functional groups around the C2 that turns the plane of polarized light to the left is referred to as levo (L), while to the right dextro (D). However, for other compounds there is no direct correlation between the direction of polarization and absolute configuration. This L/D nomenclature is used to describe the configuration of the chiral center furthest away from the most reduced carbon, such as the aldehyde of an aldose. In the case of hexoses, such as glucose, this is at C5 (Fig. 3). The configuration of this and the remaining stereochemical centers can also be defined according the Cahn–Ingold–Prelog (CIP) priority system. The CIP system relies on the systematic ranking of substituent groups around a chiral center (and can be extended to other stereochemical systems such as double bonds). Rankings are first applied according to atomic number (e.g., O > C > H); the larger the number, the higher the priority. When at this stage the priority of one or more substituent is equal, the priority is based on the bonding to subsequent atoms connected to the group with double bonds counted as two single bonds to the same atom types (e.g., –COH >
CHOH). The clockwise configuration of the priorities (first to third), when viewed with the lowest priority constituent pointed away from the viewer, corresponds to an R (recto) assignment, while anticlockwise corresponds to an S (sinister) assignment.

1.05.2.2

Stereochemical Projections of Carbohydrates

Several stereochemical projections are commonly used to describe carbohydrates.9 The Fischer projections are often used in the description of open-chain monosaccharides (Fig. 3A). All substituents are shown in the plane of the paper with chemical bonds shown as horizontal or vertical lines, with the carbon chain shown vertically with the terminal with most reduced carbon (C1) at the top of the paper. The horizontal substituents, to the left and right of the carbon chain, are projected toward the viewer, while

54

Structure and Biosynthesis of Glycoprotein Carbohydrates

1

CHO

2

HO

H

3

HO

H

4

H

5 HO

4

OH

5

H

OH 6 O HO

2 OH

OH 3

OH

1

6

CH2OH Fischer

6 5

1

4

2 3

HO

OH

OH

O HO

Haworth

4 OH

6 5

HO HO 3

OH O 2

OH Mills

1 OH

Chair (4C1)

Figure 3 Stereochemical projections of D-mannose. Shown are a Fischer projection of D-mannose (with absolute configuration 2,3,4,5S) and Haworth, Mills, and Chair projections of the cyclic structure, a-D-mannosylpyranose.

vertical substituents flanking a chiral center are projected away from the viewer. This relationship is depicted for glyceraldehyde in Fig. 1A, while the Fischer projection of D-mannose (mannosylpyranose) is shown in Fig. 3. The Haworth projection is related to the Fischer projection and is used to depict the ring structure of monosaccharides. In the Haworth projection, the groups depicted below the plane of the ring correspond to the groups depicted on the right of the carbon chain in a Fischer projection. The Haworth projection does not convey the ring pucker and can become crowded when substituents of monosaccharides, such as acetyl groups, are shown. The Mills projection similarly does not convey ring pucker but contains all the advantages of the Haworth projection while being simpler with the concomitant advantages of space. In the Mills projection, the saccharide ring is in the plane of the paper with the substituent groups shown either above the plane with a solid wedge or below the plane with a dashed wedge. The Fisher, Haworth, and Mills projections simplify monosaccharide structure by omitting the effects of the conformation of the ring. The ring of tetrahedral centers is not in a stable conformation when in one plane. For this reason, in hexoses in the pyranose form, the most stable conformation has two opposing groups located above and below the plane to form a chair conformation (Fig. 3). In this conformation, the substituents of the carbon ring are either projected broadly along the plane of the ring and are described as equatorial, or are perpendicular in a series of parallel projections that are described as axial. This ring conformation is described as 4C1, with the C4 above and C1 below the C5,O5,C2,C3 plane, respectively. However, through bond rotation the chair can flip such that C4 is below and C1 above the C5,O5,C2,C3 plane, to give a 1C4 conformation. This flip causes the groups in the equatorial position to switch to axial, and axial groups to switch to equatorial. As this is achieved solely through bond rotation, this transition does not change the configuration of any stereochemical center. Bulky substituents are not usually stably projected in the axial configuration and are usually found in the equatorial position and dominate the ring pucker. For example, in D-mannosylpyranose the C5 CH2OH group dominates the ring pucker and is in the equatorial position, while the smaller C2 hydroxyl group is in the axial position.

1.05.2.3

Structures of Common Monosaccharides

Based upon their stereochemical centers, ring hexoses have 32 possible configurations and further complexity arises from chemical modifications, such as N-acetylation and sulfation. However, only a small subset of these occurs naturally. Similarly, although there are a multitude of possible structures conceivable when different length monosaccharides are considered, such as pentoses and nonoses, only a few are routinely utilized in a particular biological system. Fig. 4 depicts some of the most common monosaccharides that occur in mammalian glycoproteins. Although glycosidic linkages could conceivably be formed from every hydroxyl, in practice this diversity is massively restricted by evolutionarily conserved biosynthetic mechanisms (Section 1.05.3).

Structure and Biosynthesis of Glycoprotein Carbohydrates

OH

OH O

HO

OH

HO

OH

O

HO

55

HO

OH

HO

OH

OH O HO

NH OH O

β-D-Glucose

OH

β-D-N-Acetylglucosamine

OH

α-D-Mannose

OH

OH

HO O

O

O

OH

HO

HO OH

NH

OH

HO

OH

OH O D-Galactose

α-L-Fucose

D-N-Acetylgalactosamine

OH COO–

OH

OH

O

H N

OH

HO O

α-D-N-Acetylneuraminic acid (sialic acid)

Figure 4 Structure of monosaccharide residues common to mammalian N- and O-linked glycoproteins. The most abundant anomer is shown. Where both anomers are common, the anomer is not defined.

1.05.2.4

Anomericity

Cyclization of monosaccharides occurs through nucleophilic attack of a hydroxyl group to the carbonyl carbon. In aldopyranoses, such as glucose and mannose, the C5 hydroxyl group attacks the C1 aldehyde to form a six-membered ring bearing a C1 hemiacetal. As the hydroxyl can attack from both faces of the trigonal planar carbonyl, two stereoisomers can be formed. In a Fischer projection, the a-anomer arises when both the in-ring hydroxyl and the hemiacetal hydroxyl are on the same side, whereas the b-anomer arises when these groups are on opposite sides. It can be seen from Fig. 5 that in the 4C1 chair conformation of D-hexoses the a-anomer has an axial hydroxyl, whereas the b-anomer has an equatorial configuration. Monosaccharides that are lacking substituents off the C1 hemiacetal hydroxyl can readily interchange anomericity; however, substitutions, such as the formation of glycosidic linkages, prevent interchange. The relative stability of the two anomers is affected by steric and stereoelectronic effects (Section 1.05.2.5). An example to illustrate the use of stereochemical projections and the various stereochemical definitions that occur in carbohydrate structure is the conversion of b-D-glucuronic acid to b-L-iduronic acid by C5-epimerase in the biosynthesis of glycosaminoglycans of proteoglycans (Fig. 6). The saccharide, b-D-glucuronic acid, is a derivative of b-D-glucose and differs only in the oxidation state of C6; the glucuronic acid contains a C6 carboxylic acid residue. The enantiomer of b-D-glucuronic acid is b-L-glucuronic acid, that is, the mirror image with the configuration of every chiral center inverted. The mirror image of b-D-glucuronic acid is still in the b-anomer as, according to the Fischer projection, the C1 and C5 hydroxyl groups are still on opposing sides. However, whereas the configuration of the C5 defines an L series, if only the C5 configuration is stereochemically inverted into its epimer, the resulting monosaccharide is a different configuration to the mirror image of b-D-glucuronic acid at C2, C3, and C4 and is therefore not b-L-glucuronic acid. The hexose with that configuration is given the trivial name idose, or in this case, iduronic acid. Note that b-L-iduronic acid has an equatorial carboxylic acid group and axial C2–C4 hydroxyl groups in the 1C4 conformation, while in the 4C1 conformation the bulky carboxylic acid group is axial and the C2–C4 hydroxyl groups are equatorial. Neither conformation is significantly more stable so when the monosaccharide is unsubstituted there are significant populations of both conformations in solution.

56

Structure and Biosynthesis of Glycoprotein Carbohydrates

H H HO H H

OH

2

5

2

H

OH

H

O

H

5 6

α-D-Glucose

HO

H

4

CH2OH

H

OH

3

HO

H

4

HO

1

OH

3

6

H

O

1

OH

H

OH

H

1 2

OH

3

H

4

OH

5 6

CH2OH

H

O CH2OH

β-D-Glucose

D-Glucose

Figure 5 Chemistry of anomer interchange of D-glucose. Hydrogens from C–H groups are only shown in the Fischer projections. Nucleophilic addition of the C5 hydroxyl to the aldehyde produces the pyranose form, while attack with C4 produces the furanose.

A

1 CHO 2

H

OH

3

HO

5

H

2

HO

OH

HO

OH

HO

5

–OOC

H

H

H

HO

acid

L-Glucuronic

O OH

acid

2

1 OH

β-D-Glucuronic acid

5

O 1

2 HO

OH

3

H

4

OH

5

H

L-Iduronic

4

HO

2

6 COO–

OH OH

3

β-L-Glucuronic acid

acid

OH OH 6 COO– 5 4 O

6 COO–

5 3

HO

6 COO–

6

HO 4 HO

1 CHO H

OH

4

6 COO– D-Glucuronic

H

3

H

H

4

H

B

1 CHO

HO 1

2

3

OH β-L-Iduronic acid

Figure 6 Stereochemistry of C5 defines the L/D series. The conversion of b-D-glucuronic acid to b-L-iduronic acid by C5-epimerase in the biosynthesis of glycosaminoglycans.

1.05.2.5

Stereoelectronic Effects

Carbohydrate conformation and relative stabilization of anomers is influenced by stereoelectronic effects around the ring.10 For example, it may be predicted from steric considerations that the b-form of glucose and mannose would predominate as this maximizes spacing between hydroxyl groups. Although there is a larger b-anomer than a-anomer population in glucose (62%–38%, respectively), the difference is smaller than that may have been expected. By contrast, the a-anomer is the predominant form of

Structure and Biosynthesis of Glycoprotein Carbohydrates

57

the mannose monosaccharide (64% a). These relative proportions may be explained by consideration of the relative orientations of the bonding, nonbonding, and antibonding orbitals (Fig. 7). The in-ring O5 has two nonbonding orbitals that in a simplified sp3 hybridization model are positioned in the axial and equatorial positions. The axial component can overlap in-phase with the antibonding orbital of an axial C1 hydroxyl, stabilizing that configuration over the equatorial b-anomer. This anomeric or Edward– Lemieux effect can also be rationalized by considering the relative dipole vectors of the O5 nonbonding orbitals and the dipole of the polarized C1–O1 s-bond. In the a-anomer, these dipoles are in opposing orientations and thus in a low-energy state, while in the equatorial position, the dipoles are more aligned and therefore less energetically favorable. In mannose, which is the C2 epimer of glucose, the C2–O2 s-bond is also positioned in the axial configuration and similarly preferentially stabilizes the axial anomer.

1.05.3

Oligosaccharide Structure

In oligosaccharides, residues are joined through the reducing terminus of one saccharide to a nonreducing hydroxyl group. This chemistry defines the overall direction of the carbohydrate from nonreducing termini to reducing termini. The residue at the reducing termini can perform mutarotation, while the anomericity of the other saccharides is fixed upon the formation of the glycosidic linkage. The conformation of glycosidic linkages is determined by stereoelectronic effects between the ring oxygen and the high-energy lone pair harbored by the exocyclic oxygen.10 This acts to stabilize the gauche conformation and functions in both a- and b-linkages (Fig. 8). In a-linkages, the antibonding orbital of the lone pair of the exocyclic oxygen is parallel with the antibonding orbital (s*)

OH

OH

OH

O HO

O

HO HO

HO OH OH

OH α-D-Mannose

α-D-Glucose

Figure 7 Stereoelectronics of the anomeric effect in hexoses. The hexoses, a-D-glucose and a-D-mannose, are shown in the preferred 4C1 chair conformation. The axial and equatorial nonbonding p-orbitals of the ring oxygen are depicted in red and dashed circles, respectively. The two opposite phases of the s* antibonding orbital of the C1–O1 bond are depicted by unfilled and filled black circles.

A

OH

OH

OH

O

HO HO

O

R

R

OH

B

OH

OH

O

HO HO

O

R

R

Figure 8 The exoanomeric effect. Preferential stabilization of particular C1–O torsion angles by the alignment of orbitals in (A) a-mannose and (B) b-mannose.

58

Structure and Biosynthesis of Glycoprotein Carbohydrates

between O5 and C1. An equivalent overlap of antibonding orbitals can also occur in gauche conformations in b-linkages. In an analogous manner to the conformational restriction that occurs in polypeptide chains by the planar secondary amide bonds, the stereoelectronic effect in carbohydrates stabilizes particular conformations.

1.05.3.1

Torsion Angles of Glycosidic Linkages

Saccharide residues with six-membered rings do not usually exhibit significant deviations away from energetically stable chair conformations. However, there is significant rotational freedom around glycosidic linkages, and linkages through the C6 carbon also have an additional degree of freedom as a result of rotations around the C6–C5 bond. These torsion angles are defined by four-atom dihedral angles (Table 1). A range of different definitions have emerged due to the different experimental techniques employed to study carbohydrate structure.11 For example, nuclear magnetic resonance (NMR) studies of carbohydrate structure often use resonance arising from the hydrogen atoms. However, hydrogen atoms do not contribute significantly to X-ray scattering and are, therefore, usually not directly detected in X-ray crystallographic studies. Instead, X-ray crystallographic studies have generally adopted definitions relying on the measurable carbon groups. However, this has an added complexity in that dihedrals can be defined along either direction of the ring giving rise to the C
1 and C þ 1 systems (Table 1). A typical glycosidic linkage obtained by X-ray crystallographic analysis of a glycoprotein is illustrated in Fig. 9. Structures of carbohydrates can be found in the Protein Data Bank (http://www.rcsb.org) and the Cambridge Crystallographic Data Center (http://www.ccdc.cam.ac.uk), and much structural data have been collated in the glycosciences.de web portal (http://www.glycosciences.de). Data mining of these databases has enabled the collation of preferred torsion angles for a wide range of naturally occurring glycosidic linkages. These angles are obtained by plotting multiple angles of a single linkage in a 4/j plot akin to the Ramachandran plot of peptides. In a direct analogy to protein folding, it is found through such plots that carbohydrates also adopt preferred torsion angles (Fig. 10). The torsion angles between particular saccharide residues can be determined by structural characterization using NMR, X-ray crystallography, and by computational approaches.12 An example of the preferred conformations adopted by glycans is illustrated by the analysis of the X-ray crystal structure of the N-linked glycans of an antibody Fc region bearing oligomannose-type glycans (Fig. 10). It can be seen that the torsion angles of glycosidic linkages, as measured by X-ray crystallography, are tightly clustered, suggesting that glycans do contain some internal structures. However, as X-ray crystallography measures atom positions in the solid state, this will underestimate conformational freedom.

1.05.3.2

Oligosaccharide Nomenclature

The nomenclature of carbohydrates has been extensively outlined by the Joint Commission on Biochemical Nomenclature of the International Union of Pure and Applied Chemistry (IUPAC) and the International Union of Biochemistry and Molecular Biology (IUBMB). For a full explanation, readers are referred to the online reference at http://www.chem.qmul.ac.uk/iupac/2carb/. The Table 1

Definition of torsion angles for NMR and X-ray crystallography11 Crystallography

Angle 4 j j (a1 / 6) u

C
1

Cþ1 0

NMR 0

O5–C1–O–C x C1–O–C0 x–C0 x
1 C1–O–C0 6–C0 5 O–C0 6–C0 5–C0 4

H1–C1–O–C0 x C1–O–C0 x–H0 x C1–O–C0 6–C0 5 O–C0 6–C0 5–H0 5

O5–C1–O–C x C1–O–C0 x–C0 xþ1 C1–O–C0 6–C0 5 O–C0 6–C0 5–O0 5

R + R′

ψ C1

O

Cx+1′ Cx–1′

O5

φ

Cx′

Figure 9 Glycosidic torsion angles illustrated for Mana1 / 2Man disaccharide. The polarity of the dihedral is illustrated by the inset box. Angles over þ180 can be described either as þ180 to þ360 or as
180 to 0. This structure was obtained from Protein Data Bank, identification code 2WAH. The torsion angle plot of the complete glycoprotein is shown in Figure 10 and the full structure is shown in Figure 20.

59

Structure and Biosynthesis of Glycoprotein Carbohydrates β-D-GlcpNAc-(1–4)-ASN

β-D-GlcpNAc-(1–4)-β-D-GlcpNAc

240

240

180

Psi

240

180

360

300

240

360

120

E

0

α-D-Manp-(1–6)-α-D-Manp Phi

180

0 60

0

360

0

300

60

240

60

180

60

120

120

60

120

0

120

360

α-D-Manp-(1–6)-β-D-Manp Phi

F

360

300

300

300

240

240

240

Psi

Psi

360 300

180

180

180

360

300

240

120

360

60

Phi

180

0

0

0

360

0

300

60

240

60

180

60

120

120

60

120

0

360

300

240

180

120

60

0

120

α-D-Manp-(1–2)-α-D-Manp G

Phi

C

300

360

300

240

180

120

60

0

D

360

300

180

α-D-Manp-(1–3)-α-D-Manp Phi

Phi

B

Psi

Psi

360

Psi

Phi

A

β-D-Manp-(1–4)-β-D-GlcpNAc

α-D-Manp-(1–3)-β-D-Manp Phi

H

360

e g

300

f

300

d 240

240

Psi

Psi

g 180

c

b

a Asn

h

180

120

60

60

0

0

c

b

a Asn

360

300

240

180

120

60

0

360

300

240

180

120

60

0

120

Figure 10 Carbohydrate Ramachandran Plot (CARP) for PDB entry 2WAH using the crystallographic C þ 1 definition (http://www.glycosciences.de/ tools/carp/carp.php). The crystal structure is illustrated in Figure 20. Terminal residues that are disordered in the crystal structure are labeled with an asterix. The torsion angles from the 2WAH glycans are labeled with red crosses. The cartoon nomenclature is defined in Section 1.05.3.2 and Figure 11.

disaccharide depicted in Fig. 8 has the full name, a-D-mannopyranosyl-(1 / 2)-a-D-mannopyranose. However, this can be abbreviated to a-D-manp-(1 / 2)-a-D-manp, and in the biochemical literature this is often abbreviated further to Mana1 / 2Man. Due to complexity of oligosaccharides, graphical systems have been developed to present the structure of glycans that is visually less complicated than the full written IUPAC description.1,13 Currently, there are two predominant systems in use, but no system has been officially adopted by the IUPAC and IUBMB. These systems are commonly referred to as the Oxford and CFG system, the latter referring to its development by the Center of Functional Glycomics. The Oxford system is depicted in Fig. 11. There are two main distinctions between the systems. First is the graphical display of linkage position and anomericity in the Oxford system. Linkage position is indicated by the angle of the bond, while anomericity is indicated by a dashed line for a-linkages and a solid line for b-linkages. By contrast, these details are written above the bond in the CFG system. The Oxford system is black and white, while

60

Structure and Biosynthesis of Glycoprotein Carbohydrates

A

6

8

α-Anomer

4

β-Anomer 3

2

B

Sialyl-Lewisx

α-Neu5Ac-(2→3)-β-D-Galp-(1→4)-[α-L-Fucp-(1→3)]-β-D-GlcNAcpC

Sialyl-Lewisa

α-L-Fucp -(1→4)-[α-Neu5Ac-(2→3)-β-D-Galp-(1→3)]-β-D-GlcNAcpD

Protein

α-Neu5Ac-(2→6)-β-D-Galp-(1→4)-β-D-GlcNAcp-(1→2)]-α-D-Manp-(1→6)-[αNeu5Ac-(2→6)-β-D-Galp-(1→4)-β-D-GlcNAcp-(1→2)]-α-D-Manp-(1→3)-β-DManp-(1→4)-β-D-GlcNAcp-(1→4)-[α-L-Fucp-(1→6)]-β-D-GlcNAcp

E

Protein

β-D-Galp-(1→4)-[α-L-Fucp-(1→3)]-β-D-GlcNAcp-(1→3)-β-D-Galp-(1→4)-β-DGlcNAcp-(1→3)-β-D-Galp-(1→4)-β-D-GlcNAcp-(1→6)-[α-Neu5Ac-(2→3)-β-DGalp-(1→4)]-β-D-GlcNAcpFigure 11 Symbolic representation of glycans using the Oxford system.13 (A) Template for defining linkage position and anomericity. The linkage position is shown by the angle of the lines linking the sugar residues (vertical line ¼ 2-link, forward slash ¼ 3-link, horizontal line ¼ 4-link, and back slash ¼ 6-link). Anomericity is indicated by full lines for b-bonds and broken lines for a-bonds. Symbols used for the structural formulae: A ¼ Gal, A ¼ GalNAc, ¼ Glc, - ¼ GlcNAc, B ¼ Man, * ¼ sialic acid, and ◈ ¼ Fuc. Examples of glycan structures together with IUPAC nomenclature: (B) sialyl Lewisx motif, (C) sialyl Lewisa motif, (D) an N-linked complex-type glycan, and (E) an O-linked glycan. Proposal for a standard system for drawing structural diagrams of N- and O-linked carbohydrates and related compounds. Adapted from Harvey, D.J.; Merry, A.H.; Royle, L., et al. Proposal for a standard system for drawing structural diagrams of N- and O-linked carbohydrates and related compounds. Proteomics 2009, 9, 3796–3801.

the CFG system uses a variety of colors (or a gray scale) to distinguish monosaccharides. Full description of the CFG system can be found at http://www.functionalglycomics.org/static/consortium/Nomenclature.shtml and in Varki et al.,1 and a full comparison of the systems is discussed by Harvey et al.13

1.05.4

Biosynthesis of Glycoproteins

Glycans are assembled by glycosyltransferases that use nucleotide-activated monosaccharide donors. Following activation, glycosyltransferase substrates are transported from the cytoplasm to various points along the secretory system by specific membrane transporters. In N- and O-linked glycosylation, glycosyltransferases usually act on the nascent glycoprotein in the secretory system. However, N-linked glycosylation is initiated by the transfer of 14-residue precursor, Glc3Man9GlcNAc2. This glycan is present in the ER conjugated via a pyrophosphate (PP) linkage to the polyisoprenol lipid, dolichol (Dol). This precursor is formed initially in the cytoplasmic face of the ER membrane until being flipped when the biosynthesis has reached approximately Man5GlcNAc2–PP–Dol and the remaining Man and Glc residues are assembled in the ER face of the membrane.

1.05.4.1

N-Linked Glycans: Glycoprotein Folding and Processing

Nascent ER-targeted proteins can be modified by N-linked glycosylation co-translationally as the protein enters the lumen of the ER5. The N-linked glycans have two synergistic effects on protein folding. Firstly, they can stabilize hydrophobic regions on the

Structure and Biosynthesis of Glycoprotein Carbohydrates

61

protein surface (Figs. 1 and 12), and secondly, they can recruit chaperone-mediated protein folding and signal to the cell the folding status of the protein.6,14 The oligosaccharyltransferase recognizes the signals within the primary structure of the extended peptide and transfers Glc3Man9GlcNAc2 to Asn side chains with the glycosylation sequon. This glycosylation sequon is Asn–X–Ser/Thr–X, where X is any amino acid except proline. Proline in the þ4 position can be glycosylated but at significantly lower percentage occupancy. Similarly, it is possible for Asn–X–Cys sequons to be glycosylated but at much reduced efficiency – typically in the region of 1% occupancy. The composition of the remaining positions proximal to the Asn residue can influence oligosaccharyltransferase efficiency and therefore the occupancy of the glycan at a particular glycosylation site. Occupancy can be predicted using the NetNGlyc server (http://www.cbs.dtu.dk/services/NetNGlyc/), which assesses the probability of occupancy of a glycosylation site based on sequence comparison with a panel of sequences with experimentally determined occupancy (Fig. 13). The N-linked glycans of newly synthesized glycoproteins are involved in a quality-control checkpoint in glycoprotein folding6,14. The nonreducing terminal residue of the Glc3 cap (Glca1 / 2Glca1 / 3Glca1 / 3) is hydrolyzed by a-glucosidase I, and the a1 / 3 residues are subsequently hydrolyzed by a-glucosidase II. However, the Glc1Man9GlcNAc2 intermediate is recognized by the ER chaperones, calnexin and calreticulin, and these also recruit other proteins involved in folding such as the disulfide bond isomerase, ERp57. Hydrolysis of the monoglucose cap eliminates the glycan-dependent interaction with these ER chaperones and is permissive for glycoproteins to exit the ER to the intermediate compartment and the Golgi apparatus. However, misfolded glycoproteins are recognized by glucosyltransferase, presumably through exposure of the hydrophobic core, of glycoproteins, which regenerates the monoglucosylated ER-retention signal facilitating further folding events. This cycle of glucose hydrolysis and transfer continues for misfolded glycoproteins until slow a-mannosidase processing creates structures that signal ER-associated degradation. However, folded glycoproteins that are not reglucosylated can exit the ER and are not retrieved to the ER unless an additional protein-based ER retention signal is present. Following protein folding, glycans can be extensively remodeled during secretion5. During transit from the ER, processing by ER a-mannosidase I hydrolyses the D2 mannose (Fig. 14). The arm specificity of the Golgi a1 / 2 mannosidases (Golgi a-mannosidase IA-C) is directly complementary to that of ER a-mannosidase I, showing highest rates of hydrolysis of the D1 and D3 arms. Together, these act to efficiently hydrolyze the terminal Mana1 / 2Man residues to create Man5GlcNAc2. This intermediate is the substrate for UDP-N-acetyl-D-glucosamine:a-3-D-mannoside b1,2-N-acetylglucosaminyltransferase I (GnT I), which transfers GlcNAc, to the 3-arm mannose residue of the oligomannose substrate, Man5GlcNAc2 to form a b1 / 2 linkage (uridine diphosphate (UDP)). The transfer of this GlcNAc is the essential step for the subsequent generation of complex-type glycans that form after the subsequent cleavage of the two 6-arm mannose residues by Golgi a3,6-mannosidase II. Thus, formation of complex glycans proceeds via the hybrid intermediate, GlcNAcb1 / 2Man5GlcNAc2. It is thought that GnT I activity is absolutely required for the action of other glycosyltransferases, including the mammalian core a1 / 6-fucosyltransferase, which catalyzes the addition of a1 / 6-fucose to the reducing end GlcNAc, and GnT III that transfers ‘bisecting’ GlcNAc b1 / 4 linked to the central b-mannose. Thus, hybrid-type glycans bearing the GlcNAcb1 / 2Man5GlcNAc2 core can be further elaborated in a similar manner to the antennae of complex-type glycans. Less common is the formation of biantennary hybrid-type glycans created by the addition of b1 / 4 GlcNAc on the 3-arm mannose. Following hydrolysis of the 6-arm mannose by Golgi a-mannosidase II, biantennary glycans can be synthesized by GnT II addition of b1 / 2 GlcNAc on the newly exposed 6-arm mannose. Further, antennae can be formed by the action of GlcNAc transferases and each arm can be further elongated by galactosylation that can be capped by sialic acid in either a2 / 3 or a2 / 6 linkages or the chain may be further extended by repeating polylactosamine residues. Finally, a further possible capping reaction includes fucosylation. These terminal structures are dependent on the tissue-specific expression of processing enzymes, for example, by the choice of cell line in the heterologous expression of recombinant glycoproteins and can be influenced by the protein structure surrounding the glycosylation site.15 For example, the N-linked glycans of typical serum glycoproteins, like those of the complement

Figure 12 The amphipathic nature of GlcNAc enables stacking against aromatic residues. These interactions cover hydrophobic regions of the protein surface preventing aggregation.

62

Structure and Biosynthesis of Glycoprotein Carbohydrates

N-Linked glycosylation potential

A

Potential Threshold 1

0.75

0.5

0.25

0 0

50

100 Sequence position

150

200

N-Linked glycosylation potential

B

Potential Threshold 1

0.75

0.5

0.25

0 0

50

100

150

200 250 Sequence position

300

350

400

Figure 13 NetNGlyc server prediction of N-linked glycosylation sites position and probability of occupancy. (A) IgG1 Fc domain, the predicted site corresponds to Asn297 in the full-length antibody (the structure of the Fc is shown in Figures 18 and 20). (B) Receptor-binding domain of the Nipah virus attachment glycoprotein. The sites correspond to Asns 306, 378, 417, 481, 529 in the full-length glycoprotein.

components, are largely similar biantennary complex-type glycans, thus reflecting their common origin in the liver.16 However, the glycan sites of complement components C3 are buried in the protein, show very limited processing, and are oligomannose-type glycans.16 Antibodies are a further case of protein-directed glycosylation.15 The glycosylation of antibodies modulates their effector functions and various expression systems have been developed that enable the production of antibodies with specific glycan structures (Section 1.05.3).

1.05.4.2

O-Linked Glycans: Glycoprotein Folding and Processing

Mucin-type O-linked glycosylation corresponds to the GalNAc modification of serine and threonines and is termed ‘O’-linked glycosylation due to the modification of the oxygen of the primary and secondary alcohol groups of serine and threonine, respectively. In addition to the ubiquitous mucin-type O-GalNAc glycosylation, there are other types of O-linked glycans, not discussed further here, which include O-fucose (e.g., as observed in the Notch signaling pathway), O-mannose (e.g., as observed in dystroglycan), reversible cytoplasmic and nuclear O-GlcNAc, and initiator O-xylose of some glycosaminoglycans of proteoglycans. Initiation of mucin-type glycosylation (here simply referred to as O-linked glycosylation) is catalyzed by a range of GalNAc transferases; some act on unmodified peptides, while others utilize a lectin domain to catalyze the modification of serines and threonines in the vicinity of existing O-linked glycans. In contrast to N-linked glycosylation, the biosynthesis of O-linked glycosylation is initiated in the Golgi apparatus and occurs through the stepwise addition of monosaccharide residues. As initiator GalNAc transferases require extended peptide conformation, mucin-type glycosylation generally occurs in peptide regions that do not exhibit secondary structure and are generally characterized by high proline content; also, mucin domains are consequently termed STP domains (Ser/Thr/Pro-rich domains). Proline acts as a secondary structure breaker due to the cyclization of the side chain to the nitrogen of the peptide amide, which prevents canonical hydrogen bonding of secondary structural elements and exhibits a restricted Ramachandran 4/j torsion angle resulting in an extended conformation. The high proline content of STP domains makes them readily identifiable by both disorder prediction algorithms, such as Regional Order Neural Network (RONN; http://www.strubi.ox.ac.uk/RONN)17 and the O-link prediction server, NetOGlyc (http://www.cbs.dtu.dk/services/NetOGlyc/).

Structure and Biosynthesis of Glycoprotein Carbohydrates

63

ER α-Glucosidase I and II D3 D2 D1

ER α-mannosidase I

D3

D1

IC/cis-Golgi

Golgi α-mannosidases IA-C

medial-Golgi GnT I FT Golgi α-mannosidase II FT

trans-Golgi Cell-specific processing

Figure 14 N-linked glycan processing in the secretory system following addition of Glc3Man9GlcNAc2 to a nascent glycoprotein. The processing of N-linked glycans from Glc3Man9GlcNAc2 follows a mostly linear pathway in the ER, intermediate compartment (IC), and cis-Golgi compartments, prior to the diversification in the medial- and trans-Golgi compartments.

RONN and NetOGlyc prediction servers are both based on neural network algorithms following training with a known data set of, in the case of RONN, protein sequences exhibiting structural order, or in the case of NetOGlyc, chemically characterized protein sequences exhibiting O-GalNAc modifications. Together, RONN and NetOGlyc can be used as independent but complementary assessment methods of the likelihood of O-linked glycosylation of candidate glycoproteins. An example of the outputs of these prediction algorithms for the human complement control glycoprotein, CD55, is provided in Fig. 15, which reveals the intrinsically disordered C-terminal STP domain. These O-GalNAc modifications can occur outside of STP domains but are usually on extended loops and are less readily identified by prediction algorithms but have been identified in some crystal structures of globular domains of glycoproteins. For example, the crystal structure of C-cadherin revealed multiple O-GalNAc modifications. O-linked glycosylation is not linked to the folding checkpoint; however, O-linked glycans can contribute to the solubility of the proteins and the stabilization of extended proline-rich regions (Fig. 15). Following the addition of GalNAc to peptide, the processing pathway is highly branched. The range of structures that can be formed have been classified into eight core structures in addition to the initiator GalNAc (termed the Tn antigen) and its sialyl derivative (Fig. 16). These core structures can be further elaborated upon to form an extensive range of possible structures. A comparably simple case of O-linked glycan structures is provided by our example, above, of the human complement regulator CD55 from

64

Structure and Biosynthesis of Glycoprotein Carbohydrates

N-glucosylation potential

A

NetNGlyo 1.0: predicted N-Linked glucosylation sites in sequence

D

N

Potential Threshold

1 0.75

N-Linked glycan

0.5 0.25 0

0

O-glucosylation potential

B

50

100

150 200 250 300 Sequence position NetOGlyo 3.1: predicted O-Linked glycosylation sites in sequence Potential Threshold

1

O-Linked glycans 0 0

C

50

100

150 200 Sequence position

250

300

Probability of disorder

1.0

C

0.5

0.0

GPI-anchor

20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 Residue position

Figure 15 Sequence analysis of the human complement regulator CD55 for the presence of N- and O-linked glycosylation. (A) Site and occupancy analysis by NetNGlyc reveals one putative N-linked glycosylation sequon. (B) Sequence analysis with NetOGlyc reveal a putative C-terminal STPdomain. (C) Disorder prediction with RONN reveals a disordered C-terminus consistent with a high proline content. (D) Model of the CD55 glycoprotein with N- and O-linked glycans displayed as sticks while the protein moiety is displayed in cartoon (rainbow).18 A glycosylphosphatidylinositol (GPI)-anchor is depicted at the C-terminus.

S/T Tn

S/T Core 1 (T Antigen)

S/T Core 5 Figure 16

S/T Core 2

S/T Core 6

S/T Sialyl Tn

S/T Core 3

S/T Core 7

Core structures of O-linked glycosylation. Symbols are defined in Figure 7.11.

S/T Core 4

S/T Core 8

Structure and Biosynthesis of Glycoprotein Carbohydrates

65

erythrocytes. The glycosylation of erythrocyte CD55 closely matches that of erythrocyte membranes and is predominantly NeuNAca2 / 3Galb1 / 3[NeuNaca2 / 6]GalNAca-S/T (Fig. 17). Common terminal motifs include sialylation, but the core structures are also scaffolds for fucosyl motifs of the blood group antigens.

1.05.5

Glycosylation of Therapeutic Glycoproteins

1.05.5.1

Overview

Glycoproteins have a broad range of therapeutic potential and emerged as a major class of therapeutic agents. These include antibodies, blood factors, anticoagulants and thrombolytics, hormones, interferons, EPO, and granulocyte–macrophage colonystimulating factor.8 The precise structure of their glycans can have significant impact on the pharmacological properties of the glycoprotein such as influencing ligand binding, pharmacokinetics, and even immunogenicity. Glycoproteins purified from natural sources (e.g., serum) can have the disadvantage of low or inconsistent yields, potential contamination with viral or bacterial pathogens, and poor purity and high heterogeneity. Recent technological advances in methods for recombinant protein expression have provided a means by which to circumvent these problems and allow the reproducible generation of pure, homogeneous glycoforms.19 Common recombinant glycoprotein expression platforms extend across the eukaryotes: yeast (e.g., Pichia pastoris), insect (e.g., Sf9/baculovirus system, Drosophila melanogaster SC2 system), plant (e.g., tobacco), and mammalian systems (e.g., Chinese hamster ovary (CHO) cells and human embryonic kidney (HEK) 293T cells). By contrast, prokaryotic expression platforms, such as Escherichia coli, are unsuitable as they lack appropriate glycosylation machinery. To control the glycosylation of the eukaryotic expression systems, the cells can either be treated with glycosidase inhibitors to stall the glycan processing at particular stages,20 or more commonly, the cell lines can be genetically modified.21–23 A panel of CHO cell lines has been generated by mutagenesis and lectin selection.21 An example is CHO Lec3.2.8.1, which is deficient in a number of factors including GnT I that stalls the processing of the N-linked glycans at Man5GlcNAc2. However, the genetic flexibility afforded by the P. pastoris expression system avoids the need for lectin selection and has enabled the complete remodeling of its glycosylation pathway including the incorporation of mammalian monosaccharide biosynthetic pathways (e.g., sialic acid) and mammalian glycosyltransferases. Screening of catalytic domains of glycosyltransferases from different species together with a panel of Golgi localization signals has led to the development of a panel of yeast cell lines that yield largely homogeneous complex-type glycans.19

1.05.5.2

Case Study: Structure of Therapeutic Antibody Glycoforms

Antibodies such as the immunoglobulin G (IgG) class are composed of a disulfide-linked dimer of heavy- and light-chain heterodimers forming epitope-specific Fab domains and an Fc domain that mediates immune effector functions (Fig. 18). Recombinant monoclonal antibodies (rMAbs) can, in principle, be targeted to any antigen, such as a known pathogen or tumor epitope. This

1108.4

100

+ 2AB + Na

+ 2AB + H

Relative intensity (%)

+ 2AB + H

2AB + Na

+

+ 2AB + H

+

+

2AB + 2Na – H

1086.4

2AB + Na

+ 1130.4

817.3

2AB + 3Na – 2H

795.3 1152.4 *

*

0 750

800

850

900

950

1000

1050

1100

1150

1200

m/z Figure 17 Matrix-assisted laser desorption/ionization mass spectrometry of O-linked glycans released from human erythrocyte CD55 by hydrazinolysis. Mass spectrometry courtesy of Prof. David J. Harvey (University of Oxford).

66

Structure and Biosynthesis of Glycoprotein Carbohydrates

A

B

Fab

Fab

Fab

Vγ

Fab Cγ 1

Cγ 1 S–S Cγ 2

Cγ 2 Fc

N Fc

Cγ 3 Cγ 3 C

β1,2-GlcNAc

Cγ 2 α1,3-Man Cγ 3

Figure 18 Structural organization of IgG. (A) Crystal structure of the anti-HIV antibody, IgG1 b12 (PDB ID 1HZH) depicted in cartoon representation (b-strands, yellow; a-helices, red; and loops, green).24,25 (B) Schematic representation of (A) showing the immunoglobulin domains (dark yellow circles, heavy chain and light yellow circles, light chains), interchain disulphide bond (S–S) and N-linked glycans of the Fc region (green wedge). (C) Crystal structure of an isolated IgG1 Fc region (PDB ID 2DTQ) colored as (A) with the electron density of a 2Fo-Fc map contoured at 1s around the biantennary N-linked glycans shown as a blue mesh. The carbohydrate is depicted as sticks (carbon, green; oxygen, red; and nitrogen, blue). The orientation of the close-up panel has been inverted to match the orientation of glycans in Figure 1.

versatility and specificity have led to the emergence of rMAbs as a major platform for therapeutics. The engineering of the N-linked glycans of the Fc region can be used to modulate the immunological properties of the antibody, for example, by increasing antigen-dependent cellular cytotoxicity (ADCC). The Fc region is formed from a dimer of the Cg2 and Cg3 constant domains of the heavy chain. In contrast to the Cg3-Cg3 interaction, which is entirely a protein–protein interface, the Cg2 domains interact through the glycans of Asn297. Cellular Fc receptors bind to the tips of both the Cg2 domains and this binding is sensitive to the precise structure of the glycans. Preclinical antibodies bearing nonfucosylated complex-type glycans display an increased affinity for activating the Fc receptor, FcgRIIIa, enhancing ADCC and may improve the efficacy of anticancer rMAbs.26 The production of antibodies devoid of core fucosylation may be achieved using cell lines deficient in a1 / 6 fucosyltransferase, FUT8, or by the use of a-mannosidase inhibitors such as kifunensine.27,28 The change in IgG glycosylation induced by kifunensine is shown in Fig. 19. The crystal structure of the resulting Fc domain bearing oligomannose-type glycosylation reveals substantial rearrangements of the Cg2 domains in the detailed architecture of the large glycans (Fig. 20). The induction of branching structure in the 6-arm of the N-linked glycan results in substantial changes in the packing of the arm against the Cg2 domain. In one chain, the crystal structure reveals a large, bulged oligomannose-type glycan that protrudes into the interdomain space. By contrast, the other glycan is largely disordered. This can be rationalized as a consequence of the domain reorganization that occurs to accommodate the glycans in the interdomain space; while the chain bearing the ordered glycan adopts a similar inter Cg2-Cg3 domain angle to the Fc structure containing complex glycans, the Cg2-Cg3 angle of the opposing chain is larger. This shift disrupts the glycan–protein anchor points and the glycan is consequently more disordered and does not give well-defined electron density.

Structure and Biosynthesis of Glycoprotein Carbohydrates

Relative abundance (%)

B

A2G1F

100

100

1647.4

A2G0F

1485.4

A2G2F

1809.3 50

A1G1F

1282.3 0

1300

A2G1

A2G2

1501.3 1663.3 1825.3 1850.4 1500

m/z

1700

1900

Relative abundance (%)

A

67

1905.6

50 1743.6 0 1200

1600

m/z

2000

2400

D3 Asn

Asn D2 D1

Figure 19 Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry of antibody Fc N-linked glycosylation.29 (A) Spectrum of recombinant Fc expressed in HEK 293T cells revealing a series of complex-type glycans annotated according to the number of GlcNAc antennae of the nonreducing terminus (e.g., A2), the number of galactoses (e.g., G2), and the presence of core a1 / 6 fucose (F). (B) Spectrum of recombinant Fc expressed in HEK 293T cells in the presence of the a1 / 2 mannosidase inhibitor, kifunensine, revealing a major Man9GlcNAc2 species (m/z 1905.6) and a minor Man8GlcNAc2 species (m/z 1743.6).

Figure 20 Crystal structure of IgG1 Fc with oligomannose-type glycosylation (PDB ID 2WAH29) depicted in cartoon representation (b-strands, yellow; a-helices, red; and loops, green) with the electron density of a 2Fo-Fc map contoured at 1s around the oligomannose-type N-linked glycans shown as a blue mesh. The carbohydrate is depicted as sticks (carbon, green; oxygen, red; and nitrogen, blue). Comparison of the structure with the mass spectrometry of the glycans is shown in Figure 19B, which reveals that some saccharide residues are disordered and only Man7GlcNAc2 and Man1GlcNAc2 could be modeled, shown in the panels. The orientation of the close-up panel has been inverted to match the orientation of glycans in Figure 1. The carbohydrate torsion angles of this structure are shown in Figure 10.

Upon binding to FcgRIIIa, the Fc adopts an asymmetric conformation, similar but not as extensive as that observed in the crystal structure of the Man9GlcNAc2 glycoform of Fc. However, as Fc with complex-type glycans devoid of fucosylation also exhibit enhanced ADCC, it is likely that the induction of asymmetry in the oligomannose-glycoform of Fc does not fully account for the increased binding to the FcgRIIIa. The modulatory effect of Fc fucosylation can be eliminated by the site-directed mutagenesis of a glycan site on the receptor.30 This receptor glycan occupies a cavity between the receptor and the Fc that is lined by the Fc glycan and into which the Fc fucose protrudes. Thus, glycans on both the Fc and the receptor influence ADCC. Consideration of the structure and biosynthesis of glycoprotein carbohydrates has led to the development of expression platforms with controlled glycosylation, the exploration of the effect of glycan structure on protein function, and the optimization of therapeutic glycoproteins.

Acknowledgments The authors wish to thank Raymond Dwek, Pauline Rudd, Simon Davis, Veronica Chang, Ian Wilson, Yvonne Jones, David Stuart, David Harvey, Mark Wormald, Thomas Lutteke, Martin Frank and Radu Aricescu for many helpful discussions. The laboratory of CNS is supported by the International AIDS Vaccine Initiative, and TAB is supported by a Sir Henry Wellcome post-doctoral Fellowship by the Wellcome Trust.

68

Structure and Biosynthesis of Glycoprotein Carbohydrates

See Also: 1.15 Protein Folding in the Endoplasmic Reticulum; 1.28 Posttranslation Modifications Other Than Glycosylation; 1.30 Glycomics.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

23. 24. 25. 26. 27. 28. 29. 30.

Varki, A., Cummings, R. D., Esko, J. D.; et al., Eds.; Essentials of Glycobiology, 2nd ed.; Cold Spring Harbour Laboratory Press: New York, NY, 2009. Cabius, H.-J., Ed.; The Sugar Code: Fundamentals of Glycoscience, Wiley: Weinheim, 2009. Fukada, M., Hindsgaul, O., Eds.; Molecular and Cellular Glycobiology, Oxford University Press: New York, NY, 2000. Garg, H. G., Cowman, M. K., Hales, C. A., Eds.; Carbohydrate Chemistry, Biology and Medical Applications, Elsevier: Oxford, 2008. Kornfeld, R.; Kornfeld, S. Assembly of Asparagine-linked Oligosaccharides. Annu. Rev. Biochem. 1985, 54, 631–664. Wormald, M. R.; Dwek, R. A. Glycoproteins: Glycan Presentation and Protein-fold Stability. Structure 1999, 7, R155–R160. Owens, R. J., Nettleship, J. E., Eds.; Functional and Structural Proteomics of Glycoproteins, Springer: New York, NY, 2010. Walsh, G.; Jefferis, R. Post-translational Modifications in the Context of Therapeutic Proteins. Nat. Biotechnol. 2006, 24, 1241–1252. Hecht, S. M., Ed.; Bioorganic Chemistry: Carbohydrates, Oxford University Press: New York, NY, 1999. Sinnot, M. L. Carbohydrate Chemistry and Biochemistry, Royal Society of Chemistry: Cambridge, 2007. Lutteke, T. Analysis and Validation of Carbohydrate Three-dimensional Structures. Acta Crystallogr. D Biol. Crystallogr. 2009, 65, 156–168. Wormald, M. R.; Petrescu, A. J.; Pao, Y. L.; et al. Conformational Studies of Oligosaccharides and Glycopeptides: Complementarity of NMR, X-ray Crystallography, and Molecular Modelling. Chem. Rev. 2002, 102, 371–386. Harvey, D. J.; Merry, A. H.; Royle, L.; et al. Proposal for a Standard System for Drawing Structural Diagrams of N- and O-linked Carbohydrates and Related Compounds. Proteomics 2009, 9, 3796–3801. Parodi, A. J. Protein Glucosylation and its Role in Protein Folding. Annu. Rev. Biochem. 2000, 69, 69–93. Rudd, P. M.; Dwek, R. A. Glycosylation: Heterogeneity and the 3D Structure of Proteins. Crit. Rev. Biochem. Mol. Biol. 1997, 32, 1–100. Crispin, M. D. M.; Ritchie, G. E.; Critchley, A. J.; et al. Monoglucosylated Glycans in the Secreted Human Complement Component C3: Implications for Protein Biosynthesis and Structure. FEBS (Fed. Eur. Biochem. Soc.) Lett. 2004, 566, 270–274. Yang, Z. R.; Thomson, R.; McNeil, P.; Esnouf, R. M. RONN: The Bio-basis Function Neural Network Technique Applied to the Detection of Natively Disordered Regions in Proteins. Bioinformatics 2005, 21, 3369–3376. Lukacik, P.; Roversi, P.; White, J.; et al. Complement Regulation at the Molecular Level: The Structure of Decay-accelerating Factor. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 1279–1284. Beck, A.; Cochet, O.; Wurch, T. GlycoFi’s Technology to Control the Glycosylation of Recombinant Therapeutic Proteins. Expet Opin. Drug Discov. 2010, 5, 95–111. Chang, V. T.; Crispin, M.; Aricescu, A. R.; et al. Glycoprotein Structural Genomics: Solving the Glycosylation Problem. Structure 2007, 15, 267–273. Patnaik, S. K.; Stanley, P. Lectin-resistant CHO Glycosylation Mutants. Meth. Enzymol. 2006, 416, 159–182. Reeves, P. J.; Callewaert, N.; Contreras, R.; Khorana, H. G. Structure and Function in Rhodopsin: High-level Expression of Rhodopsin with Restricted and Homogeneous Nglycosylation by a Tetracycline-inducible N-acetylglucosaminyltransferase I-negative HEK293S Stable Mammalian Cell Line. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 13419– 13424. Crispin, M.; Chang, V. T.; Harvey, D. J.; et al. A Human Embryonic Kidney 293T Cell Line Mutated at the Golgi A-mannosidase II Locus. J. Biol. Chem. 2009, 284, 21684– 21695. Saphire, E. O.; Stanfield, R. L.; Crispin, M. D.; et al. Crystal structure of an Intact Human IgG: Antibody Asymmetry, Flexibility, and a Guide for HIV-1 Vaccine Design. Adv. Exp. Med. Biol. 2003, 535, 55–66. Saphire, E. O.; Stanfield, R. L.; Crispin, M. D.; et al. Contrasting IgG Structures Reveal Extreme Asymmetry and Flexibility. J. Mol. Biol. 2002, 319, 9–18. Niwa, R.; Sakurada, M.; Kobayashi, Y.; et al. Enhanced Natural Killer Cell Binding and Activation by Low-fucose IgG1 Antibody Results in Potent Antibody-dependent Cellular Cytotoxicity Induction at Lower Antigen Density. Clin. Cancer Res. 2005, 11, 2327–2336. Zhou, Q.; Shankara, S.; Roy, A.; et al. Development of a Simple and Rapid Method for Producing Non-fucosylated Oligomannose Containing Antibodies with Increased Effector Function. Biotechnol. Bioeng. 2008, 99, 652–665. van Berkel, P. H.; Gerritsen, J.; van Voskuilen, E.; et al. Rapid Production of Recombinant Human IgG with Improved ADCC Effector Function in a Transient Expression System. Biotechnol. Bioeng. 2009, 105, 350–357. Crispin, M.; Bowden, T. A.; Coles, C. H.; et al. Carbohydrate and Domain Architecture of an Immature Antibody Glycoform Exhibiting Enhanced Effector Functions. J. Mol. Biol. 2009, 387, 1061–1066. Shibata-Koyama, M.; Iida, S.; Okazaki, A.; et al. The N-linked Oligosaccharide at Fcg RIIIa Asn-45: An Inhibitory Element for High Fc Gamma RIIIa Binding Affinity to IgG Glycoforms Lacking Core Fucosylation. Glycobiology 2009, 19, 126–134.

Relevant Websites http://www.cbs.dtu.dk – Center for Biological Sequence Analysis; NetNGlyc 1.0 Server. http://www.cbs.dtu.dk – Center for Biological Sequence Analysis; NetOGlyc 3.1 Server. http://www.eurocarbdb.org – European Carbohydrate Database portal; EUROCarbDB. http://www.functionalglycomics.org – Functional Glycomics Gateway. http://www.glycosciences.de – Glycosciences.de. http://www.pdb.org – Protein DataBank. http://www.strubi.ox.ac.uk – Division of Structural Biology; Regional Order Neural Network.

1.06

Nucleotide Metabolism

J Martinussen, Technical University of Denmark, Kgs. Lyngby, Denmark M Willemoe¨s, University of Copenhagen, Copenhagen, Denmark M Kilstrup, Technical University of Denmark, Kgs. Lyngby, Denmark © 2011 Elsevier B.V. All rights reserved. This is a reprint of J. Martinussen, M. Willemoës, M. Kilstrup, 1.08 - Nucleotide Metabolism, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 91-107.

1.06.1 1.06.2 1.06.3 1.06.3.1 1.06.3.2 1.06.3.3 1.06.4 1.06.4.1 1.06.4.2 1.06.5 1.06.6 1.06.7 1.06.7.1 1.06.7.2 1.06.8 1.06.9 1.06.9.1 1.06.9.2 1.06.10 1.06.10.1 1.06.10.2 1.06.10.3 References

Introduction Synthesis of Phosphoribosyl Diphosphate (PRPP) Purine Biosynthesis The Formation of IMP Compartmentalization of Purine De Novo Biosynthesis Purine Interconversion – the ATP and GTP Branch Pyrimidine Biosynthesis The Formation of UMP Compartmentalization of Pyrimidine De Novo Biosynthesis Nucleoside Triphosphate Formation Deoxyribonucleotide Biosynthesis Nucleotide Salvage Salvage Pathways Uptake of Nucleosides and Nucleobases Purine and Pyrimidine Catabolism Regulation of Gene Expression in Bacterial Nucleotide Synthesis Nucleotide Regulation at the Mechanistic Level Nucleotide Regulation at the Systems Level Exploitation of the Knowledge of Nucleotide Metabolism in Biotechnology Production of Insulin Growth Factor, IGF-1, by E. coli Production of Riboflavin in B. subtilis Inosine Production in E. coli

69 70 71 71 75 75 75 75 76 76 76 77 78 79 80 80 80 83 83 83 84 84 84

Glossary dNTP Building blocks for DNA. N Nucleobase nucleoside and nucleotide are called as “Nucleobase”. NdR Deoxyribonucleoside – nucleobase attached to deoxyribose. NR Ribonucleoside – nucleobase attached to ribose. NTP Nucleoside triphosphate – building blocks for RNA. Nucleotide biosynthesis The formation of nucleotides de novo. Nucleotide salvage Utilization of exogenous purine and pyrimidines. Uptake Transport across the cellular membrane.

1.06.1

Introduction

All cellular processes are either directly or indirectly dependent on nucleotides. Often these metabolites are considered mainly as building blocks for RNA and DNA, but they are far more than that. Adenosine triphosphate (ATP) and guanosine triphosphate (GTP) are carriers of energy in metabolic processes. Purine and pyrimidine nucleotides act as cofactors in polysaccharide and lipid biosynthesis, thus being directly involved in the formation of the cellular envelope, and purine nucleotides are precursors for the formation of coenzyme A, nicotinamide adenine dinucleotide (NAD), and flavin adenine dinucleotide (FAD). On top of that, nucleotide derivatives serve as internal signal molecules, as found for cyclic adenosine 50 -monophosphate (cAMP) and guanosine tetraphosphate (ppGpp). However, also, the individual nucleotide pool sizes and relative levels of the mono-, di-, and triphosphate forms are used by the cell as token for a given metabolic state. Significant changes in nucleotide pool sizes may have severe effects on cell physiology, so obtaining constant nucleotide pool sizes is a major issue for the cell. This homeostasis is achieved by extensive

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00008-2

69

70

Nucleotide Metabolism

control circuits at both the enzymatic and the genetic level. Nucleotides, nucleosides, and nucleobases and their derivatives are industrially important as biotechnological products. Many different compounds are utilized as pharmaceuticals (e.g., 5-fluorosubstituted pyrimidines in anticancer therapy) or as food additives (inosine 50 -monophosphate (IMP) and guanosine 50 -monophosphate (GMP) are important taste enhancers). The nucleotide metabolism is in total made up by up to 100 different reactions that can be divided into four different classes (Fig. 1): (1) de novo pathways, in which the nucleoside monophosphates IMP and uridine 50 -monophosphate (UMP) are formed; (2) interconversion, where IMP and UMP and the other phosphorylated derivatives are converted into different nucleoside triphosphates (NTPs) and dNTPs; (3) salvage, in which the different nucleosides and nucleobases are converted into NMPs; and (4) degradation, where the sugar and base moieties are utilized as carbon and nitrogen sources. RNases should also be regarded as part of the nucleotide metabolism, especially because it was found that the major source of nucleotides for DNA synthesis in Escherichia coli was derived from RNA. Nucleotides required for growth can be obtained in two different ways: either in the biosynthesis pathways, where glycine, glutamine, aspartate, 5-phosphoribosyl-a-1-pyrophosphate (PRPP), carbon dioxide, C-1 units from tetrahydrofolate, and reduction equivalents are used as precursors, or by utilizing nucleotides, nucleosides, and nucleobases in the surroundings (Fig. 1). As phosphorylated compounds are not readily taken up, many cells relay on external dephosphorylation of the nucleotides before uptake. In addition, recycling of nucleotides is central in the cell metabolism due to the high mRNA turnover (Fig. 1). Most organisms have the capability to synthesize nucleotides de novo, but exceptions among species living in fastidious environments are abundantly observed (e.g., many lactobacilli are auxotrophic for both purines and pyrimidines), relying on salvage of exogenous nucleosides and nucleobases, whereas nucleotides need to be dephophorylated by extracellular phosphatases before uptake. All cells seem to have interconversion and salvage pathways, but in contrast to the highly similar purine and pyrimidine biosynthetic pathways in all organisms, interconversion and salvage varies from organism to organism. The objective of this article is to provide an overview of the nucleotide metabolism and its regulation for the practical biotechnologist who does not need all the details from a comprehensive review. Reviews on nucleotide metabolism in specific groups of organisms are available for E. coli and Salmonella,3 Lactic acid bacteria,4 Bacillus sp.,9 and plants.11 Recent, comprehensive reviews on nucleotide metabolism in other eukaryotes are to our knowledge not available, except for an overview on purine metabolism in yeast6 and a review of nucleotide metabolism in Leishmania.1 The latter is devoid of a purine biosynthetic pathway. A comprehensive review on regulatory mechanisms in pyrimidine biosynthetic gene expression in prokaryotes is available.10 For pathways in relevant organisms, an overview is available at the KEGG database (http://www.genome.jp/kegg/pathway.html). This article is not a complete overview of the current, original literature, but the relevant original literature can be found in the reviews mentioned above.

1.06.2

Synthesis of Phosphoribosyl Diphosphate (PRPP)

A common precursor in the biosynthesis of purines and pyrimidines is 5-phospho-D-ribosyl-1,a-diphosphate (PRPP). The phosphoribosyl moiety of nucleotides is derived from PRPP through the action of phosphoribosyltransferases that, therefore, play an RNA RNA synthesis

DNA Ribonucleotide reductase (anaerobic)

NTPs RNA breakdown and anabolic reactions

DNA synthesis

dNTPs DNA breakdown

Nucleoside diphosphate kinase* Ribonucleotide reductase

dNDPs

NDPs Specific nucleoside monophosphate kinases

NMPs

sis nthe

y

Bios

dNMPs Nucleoside kinases/ nucleotide phosphatases

Ribonucleosides

Deoxyribonucleosides

Nucleoside phosphorylases/hydrolases Phosphoribosyltransferases

Nucleobases Figure 1 Nucleotide metabolism. The figure shows the faith of ribonucleotides formed de novo at the monophosphate level, except CTP formed at the triphosphate level. The open arrows indicate reactions associated primarily with nucleotide metabolism (except where mentioned) and thin arrows represents nucleic acid degradation by hydrolysis (NMPs) or phosphorylysis (NDPs). The phosphorylation of NDP to NTP* is achieved either by the nucleoside diphosphate kinase, by substrate level phosphorylation primarily in glycolysis or oxidative phosphorylation.

Nucleotide Metabolism

71

important role in both salvage and de novo pathways as described below. In addition, PRPP also plays an important role in regulating the expression of genes in purine metabolism as discussed later. The compound PRPP is synthesized by PRPP synthase that transfers the b,g-diphosphoryl group from ATP to the 1-position of ribose 5-phosphate to give PRPP: ATPþRibose5-phosphate/PRPPþAMP PRPP synthases are divided into three classes depending on differences in which effectors influence the activity of the enzyme. Class I is activated by phosphate ions and inhibited by adenosine 50 -diphosphate (ADP) and sometimes guanosine 50 -diphosphate (GDP). Class II enzymes, so far only found in plants, are independent of phosphate ions and the purine nucleoside diphosphates are simple competitive inhibitors. A third class of PRPP synthase was discovered in the archaean Methanocaldococcus jannaschii and was found to only be activated by phosphate ions but not allosterically inhibited by nucleotides. Apart from small molecule effectors, PRPP synthases from yeast and mammals are regulated through the formation of high-molecular-weight oligomeric structures composed of different subunits. In mammals, inactive subunits that resemble PRPP synthase at the sequence level are termed PAPs (PRPP synthase-associated proteins). The exact role that these supposedly regulatory subunits play in controlling the activity of PRPP synthase in these organisms is not clear at the moment, but for yeast it appears that among five different subunits only certain combinations are active.

1.06.3

Purine Biosynthesis

Purine nucleotides, like pyrimidine nucleotides, can be obtained by the action of two different metabolic pathways – de novo and salvage. IMP can be considered as the end product of the purine biosynthesis that is subsequently converted by two different routes into AMP and GMP (Fig. 2). Overall, the purine biosynthetic pathway is considered to be conserved in all organisms, because the intermediates in the biosynthetic pathway are the same. However, some steps are catalyzed by alternative enzymes using different reactants.

1.06.3.1

The Formation of IMP

As shown in Fig. 2, the biosynthesis of IMP goes through a gradual assembly of the purine base on the ribose-5-phosphate moiety of PRPP. The abbreviations of metabolites and enzymes used below are depicted in Fig. 2. More information on the reactions, enzymes, and the corresponding genes in different organisms are presented in Table 1. In the first step, pyrophosphate on position one of PRPP is substituted with an amide group from glutamine, resulting in the formation of 5-phosphoribosyl-1-amine (PRA). This reaction is catalyzed by the PRPP amidotransferase. The enzymatic activity is subject to feedback inhibition by AMP and GMP. Moreover, it has been shown that the enzyme is unstable at low AMP/GMP ratios in the presence of oxygen. In the following reaction catalyzed by 5-phophoribosyl-1-N-glycinamide (GAR) synthase, a glycine molecule is attached to the amino group of PRA to obtain GAR. Both enzymes are universally conserved. In the next reaction, where 5-phophoribosyl-1-N-formylglycinamide (FGAR) is formed by transferring a C1 unit to GAR, alternative enzymes have been discovered. In most cases, the C1 unit is obtained from formyl-tetrahydrofolate catalyzed by GAR transformylase but, in some organisms, formate can be utilized directly as formyl donor catalyzed by an alternative GAR transformylase. In the latter case, the reaction is driven by ATP hydrolysis. Some archea are dependent solely on formylation using formate as donor, whereas other organisms such as E. coli have both enzymes. Also in the following reaction, alternative FGAR amidotransferase enzymes have been discovered leading to 5-phophoribosyl-1-N-formylglycinamidine (FGAM). This ATP-dependent glutamine amidotransferase reaction is in Bacillus subtilis and lactic acid bacteria catalyzed by a three-subunit enzyme complex consisting of PurL, PurQ, and PurS, where PurQ forms the glutamine amidotransferase domain, and the PurS protein is forming a hinge between PurQ and PurL. In most other organisms, including both eukaryotes and prokaryotes, the homologs of PurL and PurQ are fused into a single polypeptide chain as seen in most organisms, including both eukaryotes and prokaryotes. In the following reaction, the formation of the imidazole ring in 50 -phophoribosyl-10 -N-(5-amino) imidazole (AIR) is finalized by the ATP-driven ring closure catalyzed by AIR synthase. The building of the second ring structure is initiated by an ATP-driven carboxylation of AIR resulting in formation of 50 -phophoribosyl-10 -N-(5-amino) imidazole-4-carboxylate (CAIR). In eukaryotes, this is a one-step reaction catalyzed by AIR carboxylase, whereas in prokaryotes a carboxyl AIR is first added to obtain the intermediate N5–CAIR catalyzed by N5–AIR carboxylase, and then subsequently converted into CAIR by N5–CAIR mutase in two separate reactions. Driven by ATP hydrolysis, aspartic acid is added by 50 -phophoribosyl-10 -N-(5-amino) imidazole-4-N-succinocarboxamide (SAICAR) synthase, and subsequently, fumarate is cleaved off by the adenylosuccinate lyase to obtain AICAR. The last enzyme in the pathway is the bifunctional enzyme 50 phophoribosyl-10 -N-(5-amino) imidazole-4-N-carboxamide (AICAR) transformylase/IMP cyclohydrolase. In this reaction, the remaining carbon in the purine ring is incorporated by transfer of the formyl group from formyl-tetrahydrofolate to obtain 50 phophoribosyl-10 -N-(5-formylamino) imidazole-4-N-carboxamide (FAICAR), which subsequently is dehydrated, resulting in ring closure to form the end product of the biosynthetic pathway, IMP. An alternative route for obtaining IMP is coupled to histidine biosynthesis. In the fifth reaction of the histidine biosynthesis pathway, an AICAR molecule is formed as a byproduct, which subsequently is converted into IMP through the last two steps of

72

Nucleotide Metabolism

A

B

HCO3–+ (NH3 or glutamine)

NH3 or glutamine –

IMP

+

1

2

12

– 5-Phosphoribosyl-1amine (PRA)

Carbamoylphosphate –

5′-Phosphoribosyl-1′N-(5-amino)imidazole4-carboxylate (CAIR)

+ ATP

CarbamoylN-aspartate 3 Dihydroorotate 4, 5 Orotate 6 Orotidine 5′monophosphate (OMP) 7 Uridine 5′monophosphate (UMP)

20

13

5′-Phosphoribosyl-1′N-(5-amino)imidazole-4 N-succinocarboxamide (SAICAR)

5-Phosphoribosyl-1N-glycinamide (GAR) 14, 15

21 19

5-Phosphoribosyl-1N-formylglycinamide (FGAR)

5′-Phosphoribosyl-1′N-(5-amino)imidazole-4N-carboxamide (AICAR)

16 5-Phosphoribosyl-1N-formylglycinamidine (FGAM)

22

5′-Phosphoribosyl-1′N-(5-formylamino)imidazole-4 N-carboxamide (FAICAR)

17 5′-Phosphoribosyl-1′N-(5-amino)imidazole (AIR)

23 8, 9 Uridine 5′diphosphate (UDP) 10 Uridine 5′triphosphate (UTP) 11 Cytidine 5′triphosphate (CTP)

18 Inosine 5′monophosphate (IMP)

5′-Phosphoribosyl-1′N-(5-amino)imidazole-5N-carboxylate (N 5-CAIR)

26

Adenylosuccinate (sAMP) 27 Adenosine 5′monophosphate (AMP)

24 Xanthosine 5′monophosphate (XMP) 25 Guanosine 5′monophosphate (GMP)

Figure 2 Purine and pyrimidine biosynthetic pathways. Main intermediates in (A) pyrimidine and (B) purine nucleotide de novo biosynthesis is shown. The numbering of reaction arrows refers to Table 1 where the detailed reactions are described and the enzymes and genes are listed. Main sites of allosteric enzyme regulation and their effectors (dotted arrows) are indicated: þ, activation; , inhibition.

the purine biosynthetic pathway. This route does not contribute to the purine de novo synthesis because ATP is the first metabolite in histidine biosynthesis. Instead, it can be considered as an interconversion of adenine nucleotides to IMP. The 10 enzymatic activities are in higher eukaryotes linked to six polypeptides including a trifunctional protein encoded by GART encompassing the GAR synthase, GAR transformylase, and AIR synthase activities; a bifunctional enzyme harboring the AIR carboxylase and SAICAR synthase activities; and a second bifunctional polypeptide with the same activities as the encoded AICAR transformylase/IMP cyclohydrolase from prokaryotes. In lower eukaryotes and yeast, the enzymes are organized as in prokaryotes except for the AIR carboxylase as discussed above.

Table 1

Enzymes in purine and pyrimidine de novo biosynthesis and deoxy-ribonucleotide biosynthesis Gene name b

No. a

Reaction

Pyrimidine de novo biosynthesis Glutamine + H2O / Glutamate + NH3 1

HCO3 + 2ATP + NH3 / Carbamoylphosphate + 2ADP + Pi

2

Carbamoylphosphate + Aspartate / N-carbamoyl-L-aspartate + Pi

3 4 5

N-carbamoyl-L-aspartate / Dihydro-orotate + H2O Dihydro-orotate + Ox. / Orotate + Red. (O2, fumarate or O2, NAD+ dependent) Dihydro-orotate + Ox. / Orotate + Red. (quinone dependent)

6

Orotate + PRPP / OMP + PPi

7 8 9

OMP / UMP + CO2 ATP + UMP / ADP + UDP ATP + UMP / ADP + UDP

10

(d)N1DP + (d)N2TP / (d)N1TP + (d)N2DP

Enzyme

Bacteria

Carbamoylphosphate synthase (CPS) Glutaminase subunit, EC 6.3.5.5 Carbamoylphosphate synthase (CPS) Synthase subunit, EC 6.3.5.5 Aspartate transcarbamoylase (ATCase) Catalytic subunit, EC 2.1.3.2 Aspartate transcarbamoylase (ATCase) Regulatory subunit, EC 2.1.3.2 Dihydro-orotase (DHOase), EC 3.5.2.3 Dihydro-orotate dehydrogenase (DHODH), EC 1.3.1.1 Dihydro-orotate dehydrogenase (DHODH), EC 1.3.5.2 Orotate phosphoribosyltransferase (OPRTase), EC 2.4.2.10 OMP decarboxylase (ODCase), EC 4.1.1.23 Uridylate kinase, EC 2.7.4.22 Uridylate/cytidylate kinase, EC 2.7.4.14

carA

CARA

carB

CARB

pyrB

PYRB

pyrC pyrDa, pyrKDbc

URA5 (URA10)

pyrF pyrH

URA3 URA6

PRPP amidotransferase, EC 2.4.2.14

13

GAR synthase, EC 6.3.4.13

ndk

YNK1

DHODH

PYR6

CMPK1, CMPK2 mitochondrial NDK4, NDPK2 NME1-7

CAD URA2 PYRE-F

UMPS

URA7, URA8

3 genes annotated

CTPS, CTPS2

purF

ADE4

PPAT

purD

ADE5

Atase ( ¼ PUR1) PUR2

GART (Continued)

Nucleotide Metabolism

PRA + glycine / GAR

Human

PYR4

PYRD

pyrE

Purine de novo biosynthesis 12 PRPP + glutamine + H2O / PRA

UTP + ATP + Glutamine + H2O / CTP + ADP + Pi + Glutamate

URA4 URA1

pyrD

pyrG

11

Plant

pyrI

Nucleoside diphosphate kinase, EC 2.4.7.6 (will use any canonical nucleoside triphosphate/diphosphate pair) Multifunctional CPS-ATCase-DHOase (CAD), EC 6.3.5.5/2.1.3.2/3.5.2.3 Bifunctional CPS-ATCase (CA), EC 6.3.5.5/ 2.1.3.2 Bifunctional OPRTase-ODCase (UMP synthase), EC 2.4.2.10/4.1.1.23 CTP synthase (CTPS), EC 6.3.4.2

1,2,3 HCO3 + 2ATP + Glutamine + H2O + Aspartate / Dihydroorotate + 2ADP + 2Pi + Glutamate 1,2 HCO3 + 2ATP + Glutamine + H2O + Aspartate / N-carbamoyl-Laspartate + 2ADP + 2Pi + Glutamate 6,7 Orotate + PRPP / UMP + CO2 + PPi

Yeast

73

Enzymes in purine and pyrimidine de novo biosynthesis and deoxy-ribonucleotide biosynthesisdcont'd

74

Table 1

No. a

Reaction

Enzyme

Bacteria

Yeast

Plant

Human

14 15 16 17 18 19 20 21 22,23

GAR + N10-formyl-tetrahydrofolate / FGAR + Tetrahydrofolate GAR + formate + ATP / FGAR + ADP + Pi FGAR + glutamine + H2O + ATP / FGAM + Glutamate + ADP + Pi FGAM + ATP / AIR + ADP + Pi AIR + HCO3 + ATP / N5-CAIR + ADP + Pi N5-CAIR / CAIR CAIR + aspartate + ATP / SAICAR + ADP + Pi SAICAR / AICAR + fumarate AICAR + N10-formyl-Tetrahydrofolate / IMP + H2O + Tetrahydrofolate

purN purT purL (purQS) purM purK purE purC purB purH

ADE8

PUR3 PUR4

GART

ADE1 PUR7 ADE13 PUR8, PUR12 ADE16, ADE17 PUR9, PUR10

PAICS ADSL ATIC

24

IMP + H2O + NAD+ / XMP + NADH + H+

GAR transformylase, EC 2.1.2.2 GAR transformylase (not assigned) FGAR amidotransferase, EC 6.3.5.3 AIR synthase, EC 6.3.3.1 N5-AIR carboxylase, EC 4.1.1.21 N5-AIR mutase, EC 5.4.99.18 SAICAR synthase, EC 6.3.2.6 Adenylosuccinate lyase, EC 4.3.2.2 Bifunctional AICAR transformylase, IMP cyclohydrolase, EC 2.1.2.3/3.5.4.10 IMP dehydrogenase, EC 1.1.1.205

guaB

IMD2-3

IMPDH1, IMPDH2

25 26 27

XMP + glutamine + H2O + ATP / GMP + glutamate + AMP + PPi IMP + GTP + aspartate / N6-(1,2-Dicarboxyethyl)-AMP + GDP + Pi N6-(1,2-Dicarboxyethyl)-AMP / AMP + fumarate

GMP synthase, EC 6.3.5.2 Adenylosuccinate synthase, EC 6.3.4.4 Adenylosuccinate lyase, EC 4.3.2.2

guaA purA purB

GUA1 ADE12 ADE13

Ribonucleotide reductase “catalytic subunit”, EC 1.17.4.1 (EC 1.17.4.2) Ribonucleotide reductase “radical generating subunit”, EC 1.17.4.1 (EC 1.17.4.2) dUTPase, EC 3.6.1.23 Thymidylate synthase, EC 2.1.1.45 (or EC 2.1.1.148, NADPH dependent)

nrdA, nrdE (nrdD)

RNR1, RNR3

RNR1

RRMi

nrdB, nrdF (nrdG)

RNR2, RNR3

RNR2, TSO2

RRM2, RRM2B

dut thyA, thyX (NADPH dependent)

DUT1 CDC21

nn THY-1

DUT TYMS

tmk dcd dcd

CDC8

ZEUS1

DTYMK

comEB

DCD1

nn

DCTD

Deoxyribonucleotide biosynthesis 28, NDP + Red./ dNDP + Ox. or NTP + Red. / dNTP + Ox. 29 30 31 32 33 34

dUTP + H2O / dUMP + PPi dUMP + N5,N10-methylene- tetrahydrofolate / dTMP + dihydrofolate or dUMP + N5,N10-methylenetetrahydrofolate + NADPH + H+ / dTMP + tetrahydrofolate + NADP+ dTMP + ATP / dTDP + ADP dCTP + H2O / dUTP + NH3 dCTP + 2H2O / dUMP + NH3 + PPi

35

dCMP + H2O / dUMP + NH3

a

Numbering refers to Figure 2 and Figure 3. As provided by the database “http://www.genome.jp/kegg/” (KEGG PATHWAY). c pyrDa (fumarate dependent), pyrKDb (NAD dependent). d nn, sequence has no official name. b

Thymidylate kinase, EC 2.7.4.9 dCTP deaminase, EC 3.5.4.13 Bifunctional dCTP deaminase, dUTPase, EC 3.5.4.30 dCMP deaminase, EC 3.5.4.12

ADE6 ADE7 ADE2

PUR5 PUR6

2 genes annotated nnd ATPURA 2 genes annotated

PFAS GART PAICS

GMPS ADSSL, ADSSL1 ADSL

Nucleotide Metabolism

Gene name b

Nucleotide Metabolism 1.06.3.2

75

Compartmentalization of Purine De Novo Biosynthesis

In plants, the formation of IMP occurs in mitochondria as well as plastids. IMP is further metabolized in the organelles to AMP and ATP. These compounds are subsequently transported to the cytosol by a uniporter. In the cytosol, AMP is deaminated to IMP, which then can be converted into GMP as described in further detail below. This compartmentalization of purine biosynthesis seems to be restricted to plants, as in other eukaryotes studied so far the synthesis of IMP is localized to the cytoplasm.

1.06.3.3

Purine Interconversion – the ATP and GTP Branch

IMP is the major branch-point in the formation of all purine nucleotides. In one branch, IMP is irreversibly converted into AMP in two enzymatic steps. First, succinyl-AMP (sAMP) is formed by condensation of aspartic acid and IMP catalyzed by adenylosuccinate synthase. This reaction is driven by GTP hydrolysis. The enzyme activity is subject to inhibition by AMP and GDP. In the following reaction, sAMP is converted into AMP by the removal of fumarate by adenylosuccinate lyase. This is the same enzyme utilized in the biosynthetic pathway converting SAICAR into AICAR. Alternatively, IMP is oxidized to xanthosine 50 -monophosphate (XMP) by IMP dehydrogenase, where NADþ acts as an electron acceptor. The enzyme is subject to feedback inhibition by GMP. Subsequently, XMP is aminated to GMP in a reaction where the amino group is donated by glutamine catalyzed by GMP synthase. This reaction is driven by ATP hydrolysis to yield AMP and pyrophosphate. This pathway is also unidirectional. However, GMP can be converted into IMP by an alternative route. GMP is directly converted into IMP by GMP reductase. The enzyme activity is subject to inhibition by ATP, but this is prevented by excess GTP.

1.06.4

Pyrimidine Biosynthesis

Like the purine nucleotides, the pyrimidine nucleotides are supplied to the synthesis of nucleic acids and other nucleotidecontaining metabolic components via two sets of reaction pathways. Preformed nucleobases or nucleosides derived from degradation of nucleic acids or taken up from the environment allow the cell to reuse these compounds, either in case of the nucleobases via a set of enzymes termed phosphoribosyltransferases or as for nucleosides by phosphorylation to monophosphates via specific kinases. Alternatively, the cell has the option to synthesize the nucleotides from the de novo pathways involving a series of catalytic steps dedicated to synthesize the pyrimidine ring and finally form the precursor of all pyrimidines in the cell by de novo synthesis, UMP. As found in purine metabolism, the pyrimidine biosynthetic pathway is universally conserved.

1.06.4.1

The Formation of UMP

Unlike the biosynthesis of purines, where the purine base is formed directly on the 5-phospho-ribosyl moiety of the final nucleoside monophosphate, the pyrimidine biosynthesis first forms the nucleobase orotate, which is subsequently transferred to the activated ribosyl 5-phosphate, PRPP (Fig. 2). Pyrimidine biosynthesis is initiated by the formation of carbamoyl-phosphate, a compound shared with the arginine biosynthesis. In plants, bacteria and probably also archaea, the enzyme carbamoyl phosphate synthase is a multicatalytic enzyme composed of two subunits: the synthase subunit that first catalyzes the formation of carbonyl phosphate from bicarbonate and ATP and the glutamine amidotransferase subunit that at the same time hydrolyzes glutamine to glutamate and ammonia that is channeled to the site of carbamate formation and release of phosphate. Lastly, the carbamoyl phosphate is formed at the cost of one more ATP g-phosphoryl. The carbamoyl phosphate synthase is controlled allosterically by both the inhibitor UMP and the activator IMP, and in agreement with carbamoyl phosphate being part of the arginine biosynthesis, the activator ornithine. Carbamoyl phosphate synthase will incorporate ammonia directly if present in high concentrations, without the need for glutamine hydrolysis and some enzymes use only ammonia, which in some higher eukaryotes seems to have a physiological relevance for arginine biosynthesis in mitochondria. The next step in pyrimidine biosynthesis is catalyzed by the aspartate carbamoyltransferase that catalyzes the condensation of aspartate and carbamoyl to yield carbamoyl aspartate. The aspartate carbamoyltransferase is considered to be the first committed step in pyrimidine biosynthesis, and consistent with this, the enzyme is allosterically regulated by cytidine 50 -triphosphate (CTP) (uridine 50 -triphosphate (UTP)) and ATP that increases or decreases the affinity of the enzyme for aspartate, respectively. One very important inhibitor of this enzyme often referred to in the literature is the compound N-(phosphonacetyl)-L-aspartate (PALA), a mimic of the condensed product/transition state for which the enzyme has a high affinity. Following the aspartate carbamoyltransferase catalyzed reaction is the ring closure by dihydroorotase to yield the reduced pyrimidine. Characteristic of some eukaryotes, and mammals in particular, is that the above three reactions are catalyzed by a multienzyme that incorporates the three enzymes as domains in a single polypeptide named the CAD enzyme (from a fusion of the names of the individual enzymes: carbamoyl phosphate synthase, aspartate carbamoyltransferase, and dihydroorotase). Similarly, in yeast this polypeptide is found in addition to an independent dihydroorotase as known from bacteria and archaea, because in the yeast CAD polypeptide the dihydroorotase is inactive. In agreement with CAD initiating pyrimidine biosynthesis, this fusion enzyme is also regulated like carbamoyl phosphate synthase and aspartate carbamoyltransferase. The conversion of dihydroorotate to orotate, which is the first pyrimidine derivative, is catalyzed by the dihydroorotate dehydrogenase that carries out the only redox reaction in pyrimidine biosynthesis. In this reaction, dihydroorotate is oxidized where

76

Nucleotide Metabolism

the electron acceptor is fumarate (class Ia), NADþ (class Ib), or the quinones of the respiratory electron transport chain (class II). All dihydroorotate dehydrogenases are FMN-containing enzymes and share a similar subunit structure. Class II enzymes are associated with the cell membrane in bacteria such as E. coli and Salmonella typhimurium or with the mitochondrial membrane in higher eukaryotes. The class Ia enzyme is found in yeast lactococci and related Gram-positive bacteria (Streptococci, Bacilli, etc.). The class Ib is the characteristic form of dihydroorotate dehydrogenase to be found. It has an additional subunit, the k-subunit that exchanges electrons from FADH2 with the dissociating NADþ. Orotate is incorporated into the first pyrimidine nucleotide, Orotidine monophosphate (OMP), by action of the orotate phosphoribosyltransferase. Next, OMP is decarboxylated by OMP decarboxylase to yield UMP. Whereas yeast resemble bacteria and archaea with respect to the enzymes performing the final two reaction steps of pyrimidine biosynthesis, higher eukaryotes have a bifunctional enzyme, UMP synthase that performs the orotate phosphoribosyltransferase and OMP decarboxylase reactions within two distinct domain homologs to the individual enzymes mentioned above. Thus, the chemical reactions are identical to that of the two monofunctional enzymes. UMP is after two rounds of phosphorylation converted to UTP that serves as a substrate for CTP synthase. This enzyme catalyzes the amination of the uracil base by first activating the pyrimidine at the 4-position by phosphorylation using ATP. The intermediate 4-phosphoryl UTP is then subject to subsequent attack by ammonia released from hydrolysis of glutamine produced at the glutamine amidotransferase domain. In CTP synthase, the glutamine hydrolysis is under allosteric control by GTP, a positive allosteric effector. Also, the product CTP is an allosteric inhibitor of CTP synthase. The enzymes from bacteria and the two isozymes from both yeast and human are displaying very similar characteristics with the exception that the yeast and human enzymes are also regulated by phosphorylation.

1.06.4.2

Compartmentalization of Pyrimidine De Novo Biosynthesis

In yeast and mammals, pyrimidine biosynthesis takes place in the cytosol just as for bacteria and archaea, except that dihydroorotate dehydrogenase as mentioned above in higher eukaryotes is located in the mitochondria in association with the membrane and close to the final electron acceptor, the respiratory quinones. This means that metabolites will have to enter (dihydro-orotate) and leave (orotate) the mitochondrion. However, in yeast, although a eukaryote, the dihydroorotate dehydrogenase is cytosolic presumable as a consequence of the preferred fermentative state of this organism. In plants, the situation appears more complicated as nucleotide biosynthesis occurs in plastids except for the dihydroorotate dehydrogenase that is also found here in the mitochondria. PRPP synthase has been shown to be transported into chloroplasts as also expected for other enzymes involved in nucleotide metabolism.

1.06.5

Nucleoside Triphosphate Formation

The phosphorylation of (d)AMP, (d)GMP, (d)CMP, dTMP(dUMP), and UMP and to nucleoside diphosphates is catalyzed by dedicated nucleoside monophosphate kinases: AMP kinase, GMP kinase, CMP kinase, dTMP kinase, and UMP kinase, respectively. In eukaryotes, a shared UMP/CMP kinase is found. All these kinases constitute a structural family of enzymes except the bacterial and archaean UMP kinase that is structurally unique. GTP allosterically inhibits the UMP kinase from bacteria. In yeast, a bifunctional kinase (CDC8) has been demonstrated that phosphorylates both thymidine and dTMP. This bifunctional enzyme was hitherto only found with the Herpes simplex virus. Once the nucleoside diphosphates are formed, the phosphorylation to triphosphates occurs by nonspecific kinases that phosphorylate both ribonucleosides and deoxyribonucleosides diphosphates. ATP is formed from ADP either by substrate level phosphorylation in glycolysis or by oxidative phosphorylation. Most organisms possess a general nucleoside diphosphate kinase that equilibrates the pools of di- and triphosphates. The nucleoside diphosphate kinase operates via a phosphorylated enzyme intermediate, and any (deoxy)ribonucleoside triphosphate can serve as donor and any diphosphate as acceptor according to the reaction scheme: (d)N1TPþ(d)N2DP/(d)N1DPþ(d)N2TP Pyruvate kinase is also a multisubstrate kinase in terms of nucleoside diphosphates and should be considered an important player in the g-phosphoryl donation to nucleoside diphosphates: PhosphoenolpyruvateþNDP/PyruvateþNTP In anaerobically grown E. coli, nucleotide diphosphate kinase is downregulated and pyruvate kinase is shown to take over the synthesis of nucleoside triphosphates. This observation may explain that in the genome of the obligate fermentative lactic acid bacteria Lactococcus lactis, no gene encoding nucleotide diphosphate kinase is found and pyruate kinase is therefore the best candidate for performing the general nucleoside triphosphate synthesis.

1.06.6

Deoxyribonucleotide Biosynthesis

Deoxyribonucleotides are all derived de novo from one enzyme system, the ribonucleotide reductase. The enzyme is a heterotetramer with two catalytic and regulatory subunits (R1) and two subunits dedicated to the generation of the important radical

Nucleotide Metabolism

77

(R2). The enzyme exists in three different forms or classes that are primarily distinguished by the mechanism by which the enzyme generates the radical needed to reduce the 20 -hydroxyl group on the ribosyl moiety of the substrate. The regulatory pattern is very similar for all three classes in which the regulation that determines the specificity of the active site for the nucleobase is identical, whereas the regulation at the activity site, if present, varies slightly. Class I ribonucleotide reductase is the prominent aerobic enzyme present in most organisms, whereas class II and class III enzymes are only found in microorganisms. Class I and class II enzymes usually have ribonucleoside diphosphates as substrates (ADP, GDP, CDP, and UDP), whereas the class III enzyme acts on the triphosphates. Several reviews are available that describe the similarities and differences in ribonucleotide reductase family of enzymes. Also, more knowledge about the regulation of this central enzyme, apart from the allosteric component, is emerging in the literature. The fourth nucleotide substrate for DNA synthesis, dTTP, is synthesized by what can be envisioned as an add-on module to the ribonucleotide reductase reaction (Fig. 3). The substrate for dTMP synthesis is dUMP that may be obtained from the reduction of UDP or UTP and subsequent cleavage of dUTP by the ubiquitous dUTPase leaving dUMP and diphosphate as products. Alternatively, dUMP is derived from a cytosine nucleotide (Fig. 3). In eukaryotes and classic Gram-positive bacteria belonging to the bacilli or streptococci genera, dCMP is deaminated by the dCMP deaminase. In Gram-negative bacteria such as E. coli and S. typhimurium, dUTP may be derived from dCTP by deamination. The significance of these two alternative routes leading to dUMP is emphasized by the fact that 50%–80% of the dTMP synthesized de novo is derived by these pathways. In archaea, as well as in mycobacteria, a bifunctional dCTP deaminase–dUTPase is found. This enzyme is structurally closely related to the trimeric dUTPase and the dCTP deaminase but performs both the deamination and the diphosphorylysis reactions to yield dUMP directly. Both dCMP deaminase and the dCTP deaminase, as well as the bifunctional enzyme, are inhibited by dTTP. In addition, dCMP deaminase is activated by dCTP. Two different thymidylate synthases are known that both perform the reductive methylation at the 5-position in the pyrimidine moiety of dUMP. The classical ThyA enzyme simultaneously oxidizes N5,N10-methylene-tetrahydrofolate to dihydrofolate and methylates dUMP. The dihydrofolate is then reduced by dihydrofolate reductase to tetrahydrofolate and N5,N10-methylenetetrahydrofolate is subsequently regenerated in this cycle by serine hydroxymethylase that in the same step produces glycine from serine. More recently, an FAD-dependent thymidylate synthase (ThyX) was found that used NADH for the reductive part of the reaction leaving out the dihydrofolate reductase reaction. ThyX is so far only found in bacteria and archaea and the slime mold Dictyostelium discoideum. In some cases, both the ThyA and ThyX enzymes are present as seen in mycobacteria. Despite different catalytic mechanisms, both enzymes are sensitive to the potent thymidylate synthase-specific inhibitor 5-fluoro-uracil deoxyribonucleoside monophosphate. Once formed, dTMP is converted into dTTP via the reactions shared with the pathways described above.

1.06.7

Nucleotide Salvage

The salvage pathways are the conversion of nucleobases and nucleosides into nucleoside monophosphates. They are primarily important in the utilization of exogenous sources but are also important in the recycling of intracellular nucleosides and nucleobases arising from nucleic acid turnover. Whereas the biosynthetic pathways, when present, are essentially conserved in all organisms, the salvage pathways are varying both with respect to their presence and the substrate specificity at the individual enzymes even among related organisms.

UTP

29

ADP ATP

dUTP

33

dCTP

dTTP

ADP 10

UDP

ATP 28

ADP 10

10

ATP

dUDP

30

dTDP

34

ADP ATP

ADP 32

8,9

ATP

UMP

dUMP

35

dCMP

dTMP ADP

31

36 ATP

Biosynthesis

MethyleneTHF (NADPH)

Dihydrofolate (NADP)

Thymidine Pi 37 dR-1-P

Thymine Figure 3 Generic presentation of dTTP formation. The numbering of reaction arrows in the biosynthetic part (numbers up to 35) refers to Table 1, where the detailed reactions are described and the enzymes and genes are listed. In addition, the salvage of thymine and thymidine that is dependent on thymidine kinase (36) and thymidine phosphorylase (37) is shown. THF, Tetrahydrofolate; dRib, deoxyribose; dR-1-P, deoxyribose-1-phosphate; Pi, phosphate.

78

Nucleotide Metabolism

1.06.7.1

Salvage Pathways

In most organisms, the major strategy for salvage of purine ribonucleosides and deoxyribonucleosides (Fig. 4) is the phosphorolytic cleavage by purine nucleoside phosphorylases, and a subsequent phosphoribosylation by phosphoribosyl transferases to obtain the corresponding ribonucleoside monophosphate. The number and substrate specificities vary. In E. coli, one purine nucleoside phosphorylase converts all ribonucleosides and deoxyribonucleosides except xanthosine that is phosphorolytically cleaved by the xanthosine phosphorylase. In B. subtilis, two different enzymes, an adenosine phosphorylase and an inosine/guanosine phosphorylase, are found, whereas xanthosine cannot be utilized. A number of different organisms also possess an alternative strategy for purine nucleoside salvage; namely, a direct phosphorylation of the nucleoside to the corresponding monophosphate, catalyzed by a nucleoside kinase. Escherichia coli can phosphorylate both guanosine and inosine by a guanosine kinase but has no activity with purine deoxyribonucleosides or adenosine. Among the Gram positives, L. lactis and B. subtilis are devoid of any purine ribonucleoside kinase but do possess all deoxyribonucleoside kinase activities. In humans and yeast, the situation is the opposite as found in E. coli as only adenosine, and not guanosine, and inosine are directly phosphorylated. Plants seem to encode all purine nucleoside kinase activities. With respect to pyrimidine nucleoside metabolism, the phosphorolytic cleavage is not the preferred reaction. Instead, uridine and cytidine are phosphorylated to their monophosphates by uridine/cytidine kinase (Fig. 5). This kinase is very relaxed with respect to the phosphoryl-donor substrate and almost any ribonucleoside or deoxyribonucleoside triphosphate will do, except UTP and CTP that are inhibitors acting as bisubstrate analogs. Uridine may also be converted into uracil and ribose

ATP

47

Biosynthesis

10

GTP 10

GDP

ADP

37

36 27

AMP

26

sAMP

ADP

ADP

ATP

ATP

41 38

45

H2O

39

H2O

42

Rib

R-1-P

43

40

ADP

ATP

ATP

41

Xanthosine

Pi

H2O

R-1-P

Rib

42

44

PRPP

H2O

Pi 43

42

Rib

R-1-P

Pi 43

42 R-1-P

46

Hypoxanthine

39

Guanosine

PRPP

Adenine

GMP

ADP 41

Inosine

Pi 43

25

XMP

41

Adenosine

Rib

24

IMP

Xanthine

PRPP Guanine

Figure 4 Generic presentation of purine salvage and interconversions in prokaryotes. Enzymes catalyzing the individual reactions are shown by numbers. Different organisms carry different complements of reactions. Detailed information of the genes and enzymes in the formation of AMP and GMP from IMP (reactions 24 to 27) and ADP/GDP to ATP/GTP (reaction 10) is given in Table 1. 10, nucleoside diphosphate kinase; 24, IMP dehydrogenase; 25, GMP synthase; 26, adenylosuccinate synthase; 27, adenylosuccinate lyase. The remaining reactions are catalyzed by the following enzymes: 36, adenylate kinase; 37, guanylate synthase; 38, adenine phosphoribosyltransferase; 39, hypoxanthine/guanine phosphoribosyltransferase; 40, xanthine phosphoribosyltransferase; 41, purine nucleoside kinase; 42, purine nucleoside phosphorylase; 43, nucleoside hydrolase; 44, adenine deaminase; 45, adenosine deaminase; 46, guanine deaminase; 47, GMP reductase; PRPP, 5-phospho-D-ribosyl-1,a-diphosphate; Rib, Ribose; R-1-P, Ribose-1-phosphate; Pi, phosphate.

11

CTP

UTP

10

10

Biosynthesis

UDP

CDP 52

53

CMP

UMP ADP

ADP

ATP

ATP

51

Cytidine

51 50 H2 0

H20

Pi

43

43 Rib

37 R-1-P

Rib

Cytosine

49

48

Uridine

PRPP

Uracil

Figure 5 Generic presentation of the salvage reactions of pyrimidine ribonucleosides and nucleobases in prokaryotes. The numbering of reactions are as follows: 10, nucleoside diphosphate kinase (see Fig. 1 and Table 1 for details); 11, CTP synthase; 48, uracil phosphoribosyltransferase; 37, pyrimidine nucleoside phosphorylase; 43, nucleoside hydrolase; 49: cytosine deaminase; 50: cytidine deaminase; 51: uridine/cytidine kinase; 52: cytidylate kinase; 53: Uridylate kinase; PRPP, 5-phospho-D-ribosyl-1,a-diphosphate; Rib, ribose; R-1-P, ribose-1-phosphate; Pi, phosphate.

Nucleotide Metabolism

79

1-phosphate by the action of uridine phosphorylase. However, this catabolic pathway serves mainly to release the ribose moiety as a carbon source. Nucleoside hydrolases are abundantly found among all living cells. Most nucleobases are salvaged by the action of phosphoribosyl transferases (Figs. 4 and 5). These enzymes catalyze the formation of the N-glycosyl bond, between the nucleobase and the ribose phosphate moiety, using PRPP as substrate. With respect to the purine phosphoribosyl transferases, all common bases can be salvaged by these enzymes (Fig. 4), but the substrate specificity and number of enzymes vary from species to species. Three groups of purine phosphoribosyl transferases (hypoxanthine phosphoribosyl transferase, adenine phosphoribosyl transferase, and xanthine phosphoribosyl transferase) have been identified. Whereas adenine phosphoribosyl transferase and xanthine phosphoribosyl transferase are dedicated to one base, the specificity of hypoxanthine phosphoribosyl transferase is broader. In most cases, this enzyme also catalyzes the salvage of guanine and, in a few cases, xanthine as well. For the pyrimidines, only uracil can be utilized directly via the action of uracil phosphoribosyltransferase (Fig. 5), whereas organisms that encode the bifunctional UMP synthase can incorporate uracil into UMP via the orotate phosphoribosyltransferase activity of this enzyme. In E. coli and Toxoplasma gondii the uracil phosphoribosyltransferase is activated by GTP, and in the archaea S. solfataricus uracil phosphoribosyltransferase is regulated by the activator GTP and the inhibitor CTP, respectively. In plants, a bifunctional uridine kinase/uracil phosphoribosyltransferase has been identified harboring the two salvage enzymes that produce UMP in one polypeptide. In many organisms, any purine nucleoside or nucleobase may serve as sole general purine source. As the formation of GMP and AMP from IMP in biosynthesis is irreversible, other pathways are active. Whereas all organisms seem to encode a GMP reductase that directly converts GMP to IMP, only eukaryotes harbors an AMP deaminase activity. Instead, prokaryotes rely on deamination of either adenine or adenosine. Examples of both adenosine and adenine deaminases have been found. Bacillus subtilis deaminates adenine to hypoxanthine, L. lactis deaminates adenosine to inosine, whereas E. coli possesses both activities. Guanine deaminase activities converting guanine into xanthine have been found in, for example, B. subtilis, but no deamination of guanosine has been reported in any organism. The cytidine deaminase, which deaminated cytidine to uridine, is together with uridine/cytidine kinase present in most characterized living organisms. Some bacteria and archaea may also deaminate cytosine to uracil via cytosine deaminase and, in this way, salvage this pyrimidine source after release of ammonia. Organisms devoid of cytosine deaminase cannot utilize cytosine and are resistant to the cytotoxic 5-fluoro-cytosine. The deoxyribonucleosides may be salvaged through dedicated kinases that show specificity for one or more of these. Exceptions are Gram-negative bacteria and archaea where a single TK1 type kinase, specific for thymidine and deoxyuridine, is the only deoxyribonucleoside kinase found. In the Gram-positive bacteria B. subtilis, apart from the TK1-like thymidine kinase, a deoxyadenosine/ deoxycytidine kinase and a deoxyguanosine kinase are found that phosphorylate deoxyadenosine/deoxycytidine and deoxyguanosine, respectively. In humans, four deoxyribonucleoside kinases are found. TK1 and two kinases that have specificities similar to the nonthymidine kinases of B. subtilis listed above and one, TK2, a thymidine kinase specifically located in the mitochondria. In addition to these enzymes, bifunctional kinases that phosphorylate both thymidine and dTMP have been found as described below. In cells lacking thymidylate synthase the requirement for thymidine can be met by the reverse reaction of thymidine phosphorylase (Fig. 3). In such strains the dUMP concentration is elevated and readily dephosphorylated to deoxyuridine, which is cleaved to deoxyribose 1-phosphate1 and subsequently used to convert thymine into thymidine.2 deoxyuridineþPi/uracilþdeoxyribose 1-phosphate

[1]

thymineþdeoxyribose 1-phosphate/thymidineþPi

[2]

Cells lacking thymidylate synthase require even less thymine when an additional mutation in the catabolic pathway of deoxyribose 1-phosphate is introduced that maintains a higher intracellular concentration of this compound.

1.06.7.2

Uptake of Nucleosides and Nucleobases

To utilize exogenous nucleotide precursors, the cell needs to transport these compounds across the cytoplasmic membrane. The uptake of nucleotide precursors has been studied in different organisms, and a common feature is that in each organism a number of transport systems with different and overlapping affinities are found. Normally, nucleotides cannot be transported, and in order to be utilized as nucleotide source, they have to be dephosphorylated by exogenous phosphatases before uptake. In eukaryotes, nucleotide transporters are found in mitochondria and chloroplasts but will not be addressed in this article. In Gram negatives, the precursors have to be transported across both the inner and outer membrane. In E. coli, an outer membrane transporter, Tsx, was found, showing affinity toward a large number of nucleotide precursors including nucleosides, nucleobases, and a number of different drugs derived from natural nucleosides and bases. With respect to base transporters located in the cytoplasmic membrane, they are characterized by transporting either purines or pyrimidines, but within the two different kinds, the specificity varies. In B. subtilis and L. lactis, two purine base transporters PbuO and PbuX have been found. PbuO is responsible for all common purine bases except xanthine, which uptake is facilitated by PbuX. The pyrimidine base transporters seem to be more restrictive. In L. lactis, a uracil permease PyrP and an OroP orotate transporter have been identified experimentally. In organisms able to utilize cytosine (e.g., E. coli), a specific cytosine transporter (codB) has been identified.

80

Nucleotide Metabolism

Nucleoside uptake specificity is highly divergent. In L. lactis, two different nucleoside transporters have been identified: a broad specificity transporter shown to facilitate the uptake of all purine and pyrimidine nucleosides and a transporter exclusively devoted to the transport of uridine. In E. coli, five different nucleoside transporters with different and overlapping affinities have been identified. With one exception characterized transporters from prokaryotes all belong to the group of electrochemical potential-driven permeases, which is a large and diverse group of secondary transporters that includes uniporters, symporters, and antiporters. The exception is the broad specificity transporter from L. lactis that turned out to be an ABC transporter.

1.06.8

Purine and Pyrimidine Catabolism

Excess purines can be degraded. Nucleoside monophosphates are subject to hydrolysis by 50 -nucleotidases to obtain the corresponding nucleoside. In most cases, the nucleoside is phosphorolytically cleaved into the nucleobase and ribose-1-phosphate. Some organisms harbor a nucleoside hydrolase activity instead, resulting in the formation of ribose. The sugar moiety may then serve as carbon source. Some organisms degrade the purine base utilizing it as a nitrogen source. The common degradation pathway as described for B. subtilis starts with the formation of xanthine through the salvage and interconversion pathways. Subsequently, xanthine is converted into uric acid, over allantoin to allantoic acid, which forms ureidoglycolic acid and ammonia. Ureidoglycolic acid spontaneously degrades to urea that in the final step is converted into ammonia and carbon dioxide. Similar pathways are commonly present in plants. A few of pathways have been characterized that lead to the degradation of uracil and thymine. The most common and well characterized is the reductive degradation leading to b-alanine. The sequence of enzymes (dihydropyrimidine dehydrogenase, dihydropyrimidinase, and b-ureidopropionase) is for the first two activities similar to steps four and three in pyrimidine biosynthesis (Fig. 2) but with the reverse function.

1.06.9

Regulation of Gene Expression in Bacterial Nucleotide Synthesis

Genes of the nucleotide pathways are highly regulated in microorganisms and respond to a variety of intracellular signals, related to nucleotide synthesis and interconversion. For a review, the reader is referred to Turnbough and Switzer.10 At the systems level, the differences can be exemplified by the feedforward activation of the purine de novo genes in L. lactis by the precursor PRPP, to the repression of the same pathway in E. coli by the purine bases, guanine and hypoxanthine, which signify salvageable purine sources. Mechanistically, the genetic regulation of the nucleotide pathways is also diverse, from UTP modulation of RNA polymerase promoter clearance at the carA promoter of E. coli and the guanine-modulated riboswitch of the xpt gene in B. subtilis to the simple repression of purM by the PurR repressor. Knowledge of nucleotide gene regulation is important in biotechnology for optimization of industrial production of nucleotides or nucleobases, or of products whose synthesis pathway is coregulated with nucleotide genes, such as folate. Aside from these direct benefits of knowing the relevant mechanisms of genetic regulation, the nucleotide genes can be used as highly responsive biosensors that measure and report the intracellular concentrations of the signal molecules and thus offers important clues about changes in the environment. Transcriptomic analysis often detects temporal changes in nucleotide genes, as these are extremely responsive. However, to obtain knowledge about cellular physiology from the upregulation of a certain gene or group of genes, one needs to know what signal the particular regulatory system responds to in the particular species. Regulation of the expression of genes in nucleotide metabolism has only been analyzed in detail in two groups of bacteria: the enteric bacteria (exemplified by E. coli) and the low GC Gram-positive bacteria (exemplified by B. subtilis and L. lactis). Only scarce information is available on regulation of nucleotide genes in eukaryotes except for the cell-cycle control of deoxynucleotides and the purine biosynthesis in yeast. In the following, the known mechanisms used by enteric and Gram-positive bacteria are briefly explained, with emphasis on the relations between metabolome status (concentrations of signal molecules) and gene expression. Table 2 is added as a suggestive code table for deciphering gene expression data from these bacteria into changes in concentrations of signal molecules. Nucleotides and their derivatives serve as important signal molecules in living organisms. The ATP/ADP ratio serves as a measure of the energy status, and the GTP level is a major determinant in the decision about entering sporulation in B. subtilis and aerial mycelium formation in streptomycetes. In the lactic acid bacterium L. lactis, a G nucleotide is involved in multistress resistance. Secondary metabolites as ppGpp and pppGpp, which have great impact on cellular metabolism through the stringent response, and derivatives such as cAMP and cGMP are likewise important signals derived from nucleotides. However, these are not discussed here.

1.06.9.1

Nucleotide Regulation at the Mechanistic Level

From Table 2, it is evident that the major sensor module and the major regulatory mediator of nucleotide availability is the RNA polymerase. Because the RNA polymerase recognizes all four ribonucleotide triphosphates, it is very well suited as a sensor for these compounds. Aside from this sensor function, the polymerase can function as a mediator of signals from other sensor molecules, by responding to the formation of rho-independent transcriptional terminators. In attenuation control, in its broad definition,5 premature transcription termination or translation initiation can be regulated by competition between alternative RNA structures. Efficient

Nucleotide Metabolism Table 2

81

Comparison of regulatory mechanisms responsible for nucleotide regulation in enteric and low-GC Gram-positive bacteria

Organism

Genes or gene products

Signal molecule

Sensor or mediator

Mechanism

Escherichia coli E. coli E. coli

pyrB, pyrE carAB, upp-uraA, codBA purA, hflD-purB, purC, purHD, purF, cvpA-purH-ubiX, purMN,prsA,codBA, carAB, pyrC,pyrD, guaBA,glnB, glyA, gcvTHP, speAB carAB, pyrB-pyrF,

UTP () UTP () G/Hx ()

RNA polymerase,Ribosome RNA polymerase PurR

Attenuation Transcriptional U-stuttering Repression, corepressor

UMP ()

Pyr RNA polymerase

RNA-binding protein, attenuation

pyrG

CTP ()

RNA polymerase

purEKB-purCQLF-purMNH(J)-purD, purR, purA, glyA, guaC, pbuO, pbuG, yqhZ-folD, and xpt-pbuX purDEK, purCSQLF, purMN, purH, purR, fold purEKB-purC(orf)QLF-purMNH(J)purD, ydhL, nupG, pbuG (only in B.s.)xpt, pbuX (both) guaB purA guaB, guaA PyrC, PyrD

PRPP (þ)

PurR

Transcriptional G-stuttering, attenuation Repressor, inducer

PRPP (þ)

PurR

Activator, inducer

guanine or hypoxanthine ()

Riboswitch, RNA polymerase

Aptamer attenuation

GTP/ATP () ATP/GTP () G nucleotide () CTP ()

Unknown Unknown Unknown RNA polymerase Ribosome

Unknown Unknown Unknown Transcript þ1 selection, Translation control by SD sequestering

Bacillus subtilis L. lactis B. subtilis L. lactis B. subtilis

L. lactis B. subtilis L. lactis

B. subtilis B. subtilis L. lactis E. coli

transcriptional termination or sequestering of ribosome binding can result from the formation of smaller secondary RNA structures including less than 20 bases, if it is followed by a row of U-nucleotides (terminator) or if it includes the ribosomal-binding site (translation initiation control), respectively. In attenuation, the formation of these negative structures can be prevented by formation of alternative RNA structures. The attenuation control is determined by a sensor molecule that either stabilizes or destabilizes one of the two alternative structures (e.g., a terminator þ anti-anti-terminator structure (leading to termination) or an antiterminator structure (leading to read-through)). This mediator function of RNA polymerase is exemplified by the PyrR regulation in Gram-positive bacteria, where the choice between alternative RNA conformations is governed by the stabilization of an anti-anti-terminator þ terminator structure through binding of the PyrR protein to the former. The result of PyrR binding is transcription termination. From comparisons of the PyrR protein sequence to other proteins, the RNA-binding PyrR protein was proposed to have originated from an ancient uracil phosphoribosyltransferase that obtained the ability to recognize the specific RNA structure of the anti-anti-terminator. PyrR from B. subtilis is, in accordance with this hypothesis, able to catalyze the uracil phosphoribosyltransferase reaction: UracilþPRPP/UMPþPPi The protein has the ability to recognize uracil, PRPP, and UMP, but apparently only UMP is important for the RNA binding of PyrR. In the presence of UMP, the PyrR protein binds and stabilizes the anti-anti-terminator þ terminator in the mRNA leader of the pyr genes, thereby preventing formation of an antiterminator, and thus leads to premature termination. In the absence of UMP, the PyrR protein is unable to bind RNA, and the antiterminator structure is formed leading to read-through transcription. The RNA polymerase has a similar mediator function in the riboswitch control of the purine operon of B. subtilis. In this system, the choice between alternative RNA conformations is governed by the stabilization of an anti-anti-terminator þ terminator conformation by binding of the purine base guanine or hypoxanthine. The anti-anti-terminator is here a large and intricate structure, called an aptamer, which is able to assume a specific three-dimensional structure with stabilizing hydrogen bonds to a purine base in a binding pocket. In the absence of purine bases, the aptamer is not stable, and instead an antiterminator is formed. Thus, in the presence of high concentrations of guanine or hypoxanthine, the aptamer þ terminator is formed and terminates transcription. In the absence of purine bases, the antiterminator is formed and transcription proceeds. A more direct involvement of the RNA polymerase as a sensor of nucleotide concentrations is seen, for the regulation of the pyrB and pyrE genes in E. coli. The genes are likewise regulated by transcriptional attenuation but by a different and more intricate mechanism. As for the widely known tryptophan and histidine attenuators of E. coli, the formation of a terminator structure in front of the pyrB and pyrE genes is determined by a race between the RNA polymerase and the leading, translating ribosome. If the RNA polymerase is faster than the ribosome, the resulting naked (nontranslated) RNA is able to assume a secondary structure. However, this is not possible if the leading ribosome moves faster, close to RNA polymerase. In contrast to the amino acid attenuators, the formation

82

Nucleotide Metabolism

of naked RNA in pyrB and pyrE attenuation leads to termination, because the regulatory system simply consists of an open reading frame for a leader peptide, followed closely by a terminator structure. The race between RNA polymerase and ribosome is primarily determined by the availability of UTP because the RNA polymerase has a very low affinity toward UTP compared with GTP, CTP, and ATP. Thus under UTP shortage, the RNA polymerase moves slowly, and the leading ribosome that synthesizes the leader peptide catches up and prevents formation of the terminator. Under UTP excess, RNA polymerase is much faster than the ribosome, and upon exposure of the nascent naked RNA, the terminator structure forms and results in termination of pyrB and pyrE gene transcription. It has also been found that the amino acid attenuators may be modulated by UTP. The amino acid responsive attenuation system is a step upward in complexity from the pyrimidine attenuators, and exposure of naked RNA leads to antitermination. This is obtained by an antiterminator structure that prevents formation of the terminator. The antiterminator structure is overlapping the last part of the gene for the leader peptide, and the terminator is located further downstream outside the realm of the ribosome. The response of an amino acid attenuator is thus reversed compared to a pyrimidine attenuator by inclusion of the alternative antiterminator structure. Accordingly, limitation for UTP, which results in a close coupling of the RNA polymerase and the ribosome, leads to increased termination under amino acid excess. When, however, translation of the leader peptide is slowed down by severe limitation of the specific responsive aminoacyl-tRNA, the RNA polymerase is faster than the ribosome, and the resulting exposure of naked RNA leads to antitermination and read-through into the amino acid operon. Exploitation of basic functions of the RNA polymerase during elongation and termination has resulted in the above mentioned mechanisms in nucleotide regulation. Three different mechanisms, distinct from the above, exploit basal functions in RNA polymerase initiation, during promoter clearance. The mechanisms in pyrimidine gene expression have been described in the comprehensive review by Turnbough and Switzer,10 but a short summary of the basic similarities is included here. Most bacterial promoters contain an A or a G nucleotide as the starting (þ1) nucleotide, followed by a stretch of nucleotides that usually show no specific features. Promoters subjected to control by the three different mechanisms are exemplified by the pyrC, the carA, and the pyrG promoters. Table 3 contains a small summary of the mRNA 50 ends that are synthesized under conditions of excess or limitation of a particular nucleotide triphosphate. The presence of two C’s in the start of the pyrC transcript (underlined) results in a low transcription initiation rate from these sites under CTP limitation. Because the choice of starting nucleotide is determined by the elasticity of the RNA polymerase when it is bound to the promoter, there is a certain probability that it skips the C’s and initiate at the G at þ 3. In the pyrC and pyrG genes of E. coli, this possibility is exploited in an ingenious way, as the full size 50 end of the transcripts form a secondary structure that sequesters the ribosomal-binding site and prevents translation. The smaller transcript starting from þ3 is not able to form the secondary structure, so the gene is translated under CTP limitation. The presence of three or more U’s from þ2 on the carA transcript results in transcriptional slippage during initiation under normal and elevated UTP levels. When the 30 -CAAA sequence of the coding DNA strand is used as template for the synthesis of 50 -GUUU, the nascent RNA chain frequently slides backwards because of the low-binding energy of the A–U base pairing. The relocation of the mRNA opens space for insertion of another U. If this happens in consecutive rounds of slippage and U-insertion, a large poly-U transcript is formed, and initiation is aborted. The genes carAB, codBA, and upp-uraA share this type of mechanism, where excess of UTP results in abortion of transcription initiation. The presence of a C nucleotide at þ 4 in the pyrG gene in B. subtilis and L. lactis results in infrequent insertion of C at this point under CTP-limiting conditions. During the idling without CTP, the RNA polymerase has an increased chance of performing the infrequent back-slippage of the preinitiation transcript on the 30 -CCC template and the insertion of an extra G. Unlike the Ustuttering, the poly-G template is not aborted and if more than three G’s are inserted, followed by the normal C, the resulting 50 -GGGGGGC transcript is able to form a stable secondary structure with downstream RNA. This structure functions as an antiterminator structure that prevents formation of a terminator. Thus, CTP excess leads to synthesis of a normal transcript that folds into a terminator structure and transcription is terminated. Under CTP limitation, the extended transcript is formed, which folds into the more stable antiterminator and leads to read-through of the terminator and transcription of the pyrG gene. In comparison with the complex regulatory mechanisms described above, the PurR repressor of E. coli is almost a textbook example of simplicity. It has a close resemblance to the (LacI) lac operon repressor and has a simple dimeric structure. It recognizes a palindromic sequence with the consensus GCAAACGTTTNC. The repressor only binds its DNA targets in a complex with its corepressor, which is either guanine or hypoxanthine. Thus, transcription is repressed under purine excess.

Table 3

Differential synthesis of pyrC, carA and pyrC mRNA ends under nucleotide excess and limitation Transcript 50 end formation (A)

Gene Consensus: Consensus: pyrC (Escherichia coli) carA (E. coli) pyrG (Bacillus subtilis)

Nucleotide sequence GNNNNNN ANNNNNN CCGGCNN GUUUGNN GGGCNNN

Under conditions of All conditions All conditions CTP excess UTP limitation CTP excess

Transcript 50 end formation (B) Nucleotide sequence GNNNNNN ANNNNNN GGCNN GUUUUUUUU GGGGGGCNN

Under conditions of All conditions All conditions CTP limitation UTP excess CTP limitation

Nucleotide Metabolism

83

PurR proteins from low-GC Gram-positive bacteria form a large family that is distinct from the PurR/LacI family from Gramnegative bacteria. The mechanisms of PurR-mediated regulation have been elucidated in B. subtilis (PurRB) and L. lactis (PurRL) and the results are very interesting from an evolutionary point of view. Both PurRL and PurRB are allosterically regulated by binding to PRPP, the precursor for all purine nucleotides. At the systems level, both PurR proteins mediate feedforward regulation of high PRPP levels into high levels of pur gene expression. However, while PurRL functions as a transcriptional activator PurRB is a transcriptional repressor. This apparent paradox is, however, solved because PurRL is bound to DNA at all times but activates transcription only when bound to PRPP, whereas PurRB binds to DNA only in the absence of PRPP. Thus, PRPP is an inducer of both PurRL and PurRB. It has not yet been fully understood how two highly similar proteins have evolved convergently following very different regulatory schemes. Also, it is not yet known whether the Gram-positive relatives carry PurRL- or PurRB-like regulators. When the primary sequence of the PurR protein from L. lactis is compared with the complement of protein sequences from more than 675 totally sequenced bacteria, homologs of the full protein sequence appears within the Gram-positive bacteria only. The C-terminal of these proteins show striking homology to xanthine phosphoribosyltransferases, and it appears that the Grampositive purR gene originated from duplication of the xpt gene encoding a xanthine phosphoribosyl transferase in the Grampositive ancestor.

1.06.9.2

Nucleotide Regulation at the Systems Level

Most of the regulatory systems shown in Table 2 follow a clear biological logic at the systems level. UMP and UTP are end products of the pyrimidine biosynthesis and salvage pathways, and negative feedback regulation of these pathways by their end product makes perfectly sense. Likewise, the negative feedback of the pyrG gene by the product of the PyrG reaction, CTP, is also optimal. The use of GTP/ATP ratio in B. subtilis and a G-nucleotide (GMP, GDP or GTP) in L. lactis as a signal for negative feedback regulation of the branch leading from IMP to GMP is also understandable. Somewhat less clear is the use of CTP as signal in negative feedback of the pyrC and pyrD genes from E. coli. However, CTP can be envisioned as the ultimate end product of the pyrimidine nucleotide synthesis (Fig. 2), which makes it a fair candidate as feedback signal. Logically, the use of the precursor PRPP as signal for the positive feedforward regulation of purine genes by PurR in Gram positives is justified, although PRPP should function as a signal for both pyrimidine and purine nucleotide synthesis. As previously discussed, the PRPP level is inversely correlated to the ADP level because the PRPP synthase is inhibited by ADP. Thus, the PRPP signal is functionally a representation of the ADP level, so that the apparent positive feedforward regulation by PRPP is in effect of a negative ADP-modulated end product feedback regulation. It is also interesting how the purine bases guanine and hypoxanthine can function as signals for regulation of purine biosynthetic genes as it is seen in E. coli mediated by the PurR repressor and in B. subtilis mediated by the riboswitch in front of the purine operon. The most straightforward explanation is that the regulatory system was evolved for the detection of purine bases that are taken up from the environment by transporters. However, it has been found that transport of purine bases is inefficient unless it is coupled to the cognate phosphoribosyl transferase reaction that converts the base to a nucleotide monophosphate. Free purine bases are neither formed by the biosynthesis pathway nor produced from nucleotides by salvage enzymes. They could, however, be considered as end products if they were formed as part, or as a byproduct, of the normal purine recycling pathway. That this could be the case is suggested by the fact that an hpt gpt double mutant of E. coli, which is unable to convert intracellular guanine and hypoxanthine to the corresponding nucleotides but is otherwise genetically wild type, excrete purine bases and frequently acquire purR mutations to alleviate them from the repression of their biosynthetic pathway. It thus appears that purine bases are frequently formed products of the normal synthesis pathway in wild-type strains, and that the PurR repressor, therefore, mediates a negative feedback control on the supply pathway. The enzymes responsible for the production of purine bases under normal conditions have not yet been identified, but they could involve mRNA recycling.

1.06.10 Exploitation of the Knowledge of Nucleotide Metabolism in Biotechnology Knowledge about the regulatory logic at the systems level can be used to convert a living bacterium into a multivalent biosensor, using transcriptomic data as clues. The use of purine biomarkers has proved somewhat useful for improvement of production yields in both Gram-negative and Gram-positive bacteria. Also, the knowledge of the allosteric properties of the enzymes in nucleotide metabolism has proved to be useful in designing new strains with improved capacity to produce high-value compounds. Short descriptions of three of these applications will round off this section on nucleotide metabolism and its regulation.

1.06.10.1 Production of Insulin Growth Factor, IGF-1, by E. coli One of the most well-known reports on the use of transcriptomics to increase product yield is on the production of insulin-like growth factor 1 (IGF-1) in E. coli using high cell-density fed-batch fermentation. Here, Choi et al.2 observed that the prsA gene, encoding the PRPP synthetase, was among the most severe downregulated genes in a transcriptome analysis of the strain containing the gene for the IGF-1 protein, compared with the production strain with an empty expression plasmid. From the report, it is evident that the remaining members of the PurR regulon in E. coli were not severely affected, which signifies that the prsA downregulation was not determined by feedback control by the PurR corepressors, guanine or hypoxanthine. From the literature, no other

84

Nucleotide Metabolism

information is available about the genetic regulation of prsA, but if another regulatory system downregulated the gene, this would most likely signify that the gene product was not in demand. After transformation of the production strain with the prsA gene transcribed from an Isopropyl-b-D-thiogalactoside (IPTG) inducible promoter, the authors were, however, able to increase the yield of IGF-1 by 2.3-fold. This significant increase cannot be explained from the known regulatory properties of the prsA gene, and the effect of the increased gene dose could suggest that the gene is subjected to a type of regulation that does not correlate with the need of the gene in the specific experimental setting.

1.06.10.2 Production of Riboflavin in B. subtilis In a more recent study on the production of riboflavin in B. subtilis, the researchers found that a large number of the PurR regulon members were downregulated between two- and fourfold when multiple copies of the riboflavin biosynthesis genes were present in the production strain.7 Riboflavin synthesis requires 1 mol of each GTP, ribulose 5-phosphate, and NADPH. The system is especially complicated because ribulose 5-phosphate serves a double role in the riboflavin synthesis as it is also required in the formation of PRPP, which is used for GTP synthesis. As the PurR protein responds positively to the intracellular PRPP concentration, the authors reasoned that the lowered PurR regulon members signified that the PRPP level was low. Accordingly, the authors inserted a prs gene under the control of a strong promoter. Analysis of the relevant metabolites in the Prs-overproducing strain showed that the PRPP level was unaltered while the substrate ribose 5-phosphate for its synthesis was low and ribulose 5-phosphate was normal. To increase the conversion of ribulose 5-phosphate to ribose 5-phosphate, the ywlF gene encoding a ribose 5-phosphate isomerase was cloned and coexpressed with the prs gene in the production strain. The result was a fourfold increase in both ribose 5phosphate and PRPP levels, accompanied by upregulation of the PurR regulon genes and a 25% increase in riboflavin titer from fed-batch cultures, compared with the production strain without the inserted prs and ywlF genes. Although the increase was modest, the study shows promises for the use of production strains as biosensors in the optimization of production yields.

1.06.10.3 Inosine Production in E. coli The sodium salt of IMP is used as a taste enhancer and the corresponding nucleoside inosine is used as the precursor in the biochemical production of IMP. Inosine is produced in fermentations using strains mutated in different genes in nucleotide metabolism leading to an increase in inosine yields. The first step in the purine biosynthetic pathway catalyzed by PRPP amidotransferase is subject to feedback inhibition by AMP and GMP, and a mutant in purF encoding PRPP amidotransferase in which the allosteric sites have been destroyed were constructed. The substrate PRPP for this reaction is obtained by the action of the PRPP synthase encoded by prs, which is subject to feedback inhibition by ADP. An ADP insensitive PRPP synthase derivative was introduced in the production strain together with the mutated purF allele, and a significant increase in inosine production was observed.8

References 1. Carter, N. S.; Yates, P.; Arendt, C. S.; et al. Purine and Pyrimidine Metabolism in Leishmania. Adv. Exp. Med. Biol. 2008, 625, 141–154. 2. Choi, J. H.; Lee, S. J.; Lee, S. J.; Lee, S. Y. Enhanced Production of Insulin-like Growth Factor I Fusion Protein in Escherichia coli by Coexpression of the Down-regulated Genes Identified by Transcriptome Profiling. Appl. Environ. Microbiol. 2003, 69, 4737–4742. 3. Jensen, K. F.; Dandanell, G.; Hove, J. B.; Willemoes, M. Nucleotides, Nucleosides, and Nucleobases. In Escherichia coli and Salmonella: Cellular and Molecular Biology; Böck, A., Curtiss R. III, Kaper, J. B.; et al., Eds., ASM Press: Washington, DC, 2008. https://doi.org/10:1128/ecosal.3.6.2. 4. Kilstrup, M.; Hammer, K.; Ruhdal, J. P.; Martinussen, J. Nucleotide Metabolism and its Control in Lactic Acid Bacteria. FEMS Microbiol. Rev. 2005, 29, 555–590. 5. Nudler, E.; Mironov, A. S. The Riboswitch Control of Bacterial Metabolism. Trends Biochem. Sci. 2004, 29, 11–17. 6. Rolfes, R. J. Regulation of Purine Nucleotide Biosynthesis: In Yeast and beyond. Biochem. Soc. Trans. 2006, 34, 786–790. 7. Shi, S.; Chen, T.; Zhang, Z.; et al. Transcriptome Analysis Guided Metabolic Engineering of Bacillus Subtilis for Riboflavin Production. Metab. Eng. 2009, 11, 243–252. 8. Shimaoka, M.; Takenaka, Y.; Kurahashi, O.; et al. Effect of Amplification of Desensitized PurF and Prs on Inosine Accumulation Escherichia coli. J. Biosci. Bioeng. 2007, 103, 255–261. 9. Switzer, R. L.; Zalkin, H.; Saxild, H. H. Purine, Pyrimidine and Pyridine Nucleotide Metabolism. In Bacillus subtilis and its Closest Relatives: From Genes to Cells; Sonnewald, U., Hoch, J. A., Losick, R., Eds., ASM Press: Washington, DC, 2002; pp 255–270. 10. Turnbough, C. L.; Switzer, R. L. Regulation of Pyrimidine Biosynthetic Gene Expression in Bacteria: Repression without Repressors. Microbiol. Mol. Biol. Rev. 2008, 72, 266–300. 11. Zrenner, R.; Stitt, M.; Sonnewald, U.; Boldt, R. Pyrimidine and Purine Biosynthesis and Degradation in Plants. Annu. Rev. Plant Biol. 2006, 57, 805–836.

1.07

Organic Acidsq

Maria Papagianni, Department of Food Hygiene and Technology, School of Veterinary Medicine, Aristotle University of Thessaloniki, Thessaloniki, Greece © 2019 Elsevier B.V. All rights reserved. This is an update of M. Papagianni, 1.09 - Organic Acids, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 109-120.

1.07.1 1.07.2 1.07.2.1 1.07.2.2 1.07.2.2.1 1.07.2.2.2 1.07.2.2.3 1.07.2.2.4 1.07.2.2.5 1.07.2.2.6 1.07.2.3 1.07.2.4 1.07.3 1.07.3.1 1.07.3.2 1.07.3.3 1.07.4 1.07.4.1 1.07.4.2 1.07.4.3 1.07.5 1.07.5.1 1.07.5.2 1.07.5.3 1.07.6 1.07.6.1 1.07.6.2 1.07.6.3 1.07.7 References

Introduction Citric Acid Microorganisms and the Biochemistry of Citric Acid Accumulation The Citric Acid Fermentation by A. niger Carbon Sources Nitrogen and Phosphate Limitation pH Aeration Trace Metals Fungal Morphology Industrial Production of Citric Acid Recovery of Citric Acid Lactic Acid Microorganisms and the Biochemistry of Lactic Acid Accumulation Lactic Acid Production Processes Recovery of Lactic Acid Gluconic Acid Microorganisms and the Biochemistry of Gluconic Acid Accumulation Gluconic Acid Production Processes Recovery of Gluconate and Gluconic Acid Itaconic Acid Microorganisms and the Biochemistry of Itaconic Acid Accumulation Itaconic Acid Production Processes Recovery of Itaconic Acid Succinic Acid Microorganisms and the Biochemistry of Succinic Acid Accumulation Succinic Acid Production Processes Recovery of Succinic Acid Perspectives

86 86 87 88 88 89 89 89 89 89 90 91 91 91 92 92 93 93 94 94 94 95 95 95 96 96 96 96 97 97

Glossary Anaplerotic reaction Enzyme catalyzed reactions that form intermediate metabolites in various metabolic pathways such as the tricarboxylic acid cycle. Biotransformation The chemical conversion of a substance from a chemical to another chemical that is mediated by a living organism or enzyme preparations derived from them. Productivity The total production capacity required in a fermentation plant is worked out by comparing the required production rate with the productivity of the process. The productivity of the fermentation stage is the amount of product produced per unit time per unit volume. Cell-recycle continuous culture techniques Continuous systems in which the feed to the bioreactor contains recycled microorganisms. Fermentation yield The total amount of produced product divided by the total amount of consumed substrate.

q

Change History: November 2017. M. Papagianni rewrote the article. New figures have been added, as well as one table. New references were also added.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00009-4

85

86

Organic Acids

1.07.1

Introduction

Organic acids are widely distributed in nature as they occur in animal, plant and microbial sources. They contain one or more carboxylic acid groups, which may be covalently linked in groups such as amides, esters, and peptides. Several fungi, yeasts and bacteria produce organic acids. In aerobic bacteria and fungi, organic acid accumulation occurs as a result of incomplete substrate oxidation, due to imbalances in essential nutrients, such as mineral ions.1 The use of manganese deficient media for overproduction of citric acid by Aspergillus niger is a well-known example of this case.1 In anaerobic bacteria, formation of organic acids is rather a means to regenerate NADH and their accumulation strictly parallels growth.2 Depending on their metabolic origin, microbially produced organic acids can be distinguished between those derived from a main metabolic pathway of aerobic microorganisms, such as the tricarboxylic acid cycle and glycolysis, and those derived from the direct oxidation of glucose.1,2 The second group, produced by one or two enzymatic steps from glucose, is in this respect relevant to biotransformations.1 Citric, lactic, itaconic, and malic acids belong to the first group, while gluconic and acetic acids to the second group. The number of acidic compounds isolated from microorganisms exceeds 100. They range from simple, unsubstituted acids such as formic, to complex, glycosylated acids such as pyolipic acid. The monocarboxylic acids range from the simple formic acid to acetic, propionic, stearic and the most complex acids containing branch chains, such as the isovaleric acid.1,2 More than a dozen di- and tricarboxylic acids are known, while the number of known hydroxy-(or keto-) acids of microbial origin exceeds 50, with the commercially important citric, gluconic, itaconic, ascorbic, and lactic acids included in this category. A small number of organic acids, namely the citric, lactic, gluconic, and itaconic acids, are produced industrially solely by microbial processes (Table 1). Large quantities of acetic acid (
190 kt year1) are also produced by fermentation but the main way of production is chemical synthesis. For many organic acids the market is still small. It is expected, however, that it will increase in the future as new markets arise from new applications. Production of organic acids by fermentation processes has become one of the targets of an emerging bioindustry that uses renewable materials as substrates for the production of bulk chemicals. The functional groups of organic acids can serve as excellent natural starting materials for the chemical industry in applications best suited to a sustainable industrial society.3–6 Apart from the use of low-cost substrates, it is essential if a bioprocess is to be cost-effective, the process itself to be improved in order to give higher yields and product titers. There are always two options for process improvement. The first comprises improvements on the biochemical engineering part, while the other, improvements on microbial physiology suited to a particular process. Highly sophisticated culture techniques, such as continuous culture, cell-recycle, or retention techniques, may offer successful alternatives to classic batch operations. Genetic engineering, genomic information, and metabolic engineering of the producer microorganisms may offer robust industrial strains, otherwise highly effective microbial cell factories.7–9 The organic acids described in this article are produced in large volumes through bioprocessing. Producer microorganisms, the biochemistry of production, production details, recovery processes as well as applications, and economic aspects are the subjects of discussion throughout the article.

1.07.2

Citric Acid

Citric acid (2-hydroxy-propane-1,2,3-tricarboxylic acid, Fig. 1) is an intermediate of the tricarboxylic acid cycle and as such, it occurs in every living organism. It was first described as a constituent of lemons from which was produced until the development of the fermentation process by A. niger in the 1920s.

Table 1

Annual production of major organic acids produced by fermentation and microorganisms used in processes

Organic acid

Annual production (kilotonnes)

Microorganisms

Citric acid Lactic acid Gluconic acid Itaconic acid Succinic acid

1600 150 90 50 40

Aspergillus niger, yeasts Lactic acid bacteria A. niger, Gluconobacter oxidans Aspergillus itaconicus, Aspergillus terreus Recombinant Escherichia coli, yeasts

OH O Figure 1

Citric acid.

O

OH

OH

O OH

Organic Acids

87

Citric acid is commercially the most important of organic acids. Commercial production from Italian lemons commenced in 1826 by John and Edmund Sturge, in England. Lemons remained the only source of the acid until 1919 when the first industrial process using A. niger was established in Belgium. However, a virtual monopoly of production was maintained by Italians for the next 100 years and the product remained expensive.1 In search for alternative sources of production, chemical and microbiological ways of synthesis were investigated. Chemical synthesis of the acid was first reported in 1880, using glycerol as the starting material. Later, the acid was synthesized from dichloroacetone. A number of other methods were reported but have so far proved economically uncompetitive with fermentation. In 1917, Currie discovered that some strains of A. niger were able to grow and produce large amounts of citric acid in an acidic medium (initial pH 2.5–3.5) with a high concentration of sugar and mineral salts. This observation formed the basis of the industrial production of citric acid which established by Chas. Pfizer & Co. Inc. in the United States in 1923. Citric acid is used extensively by the food, pharmaceutical and chemical industries. Among its properties, acidity, flavor, salt formation and low toxicity make the basis of most of its applications. The use of citric acid as flavoring is due to its acidity that reduces aftertaste, and its ability to act as flavor enhancer. Citric acid forms a wide range of metallic salts and this is the reason for its extensive use as sequestering agent in many industrial processes. Some of its salts make important pharmaceutical products. Trisodium citrate, for example, is used as anticoagulant blood preservative because it prevents blood cloating by complexing calcium. The salts are also the basis for its antioxidant properties in fats and oils since metal-catalyzed reactions are reduced by chelating metals.

1.07.2.1

Microorganisms and the Biochemistry of Citric Acid Accumulation

Apart from A. niger, many other filamentous fungi have been found to produce citric acid including strains of A. nidulans, A. awamori, A. wentii, A. flavus, A. fonsecaeus, A. luchensis, A. phoenicus, A. saitoi, Penicillium janthinellum, P. restrictum, Eupenicillium sp., Absidia sp., Acremonium sp., Botrytis sp., Mucor piriformis, Talaromyces sp., Ustulina vulgaris, and Trichoderma viride. In addition to filamentous fungi, yeasts are able to produce citric acid and several strains produce large amounts from n-alkanes and carbohydrates substrates. During the 1960s and 1970s citric acid was produced commercially from yeasts using hydrocarbon substrates. The process is no longer economical but yeasts are used to produce the acid from carbohydrate feed stocks. Yeastproducers of citric acid include Candida, Hansenula, Pichia, Debaromyces, Torula, Torulopsis, Kloekera, Saccharomyces, Zygosaccharomyces, and Yarrowia spp. Several strains of the Candida species, such as C. tropicalis, C. catenula, C. guilliermondii, and C. intermedia, have been used industrially. Another yeast, that might have been used industrially for citric acid production, is Y. lipolytica. This is able to produce citric acid from carbohydrate sources which are efficiently assimilated by A. niger, e.g., glucose and sucrose, but in addition, it utilizes efficiently n-paraffins and fatty acids that are not used by A. niger. The main drawback of the yeast fermentations is that substantial quantities of the carbon substrate are routed and therefore wasted-in the formation of isocitric acid. Attempts have been made to reduce isocitric acid production by selecting mutants with very low aconitase activity. Aspergillus niger remains today the main industrial producer of citric acid.1,2,10 Certain strains of A. niger produce large amounts of the acid in various types of fermentation processes. The yield of citric acid often exceeds 70% of the theoretical yield on the carbon source. Industrial strains of citric acid producer A. niger are among the best secretly kept microorganisms in the biotech industry. The critical parameters that result in citric acid overproduction in the A. niger process include a high carbohydrate concentration in the medium, nitrogen and certain trace metals in limited amounts, maintenance of high dissolved oxygen concentrations, and the low pH.10–12 Maintaining the above parameters allowed the development of highly productive industrial processes. The biochemical pathways of citric acid biosynthesis involve the glycolytic catabolism of glucose to two molecules of pyruvate, and their conversion to the precursors of citrate, oxaloacetate and pyruvate (Fig. 2). A key step in this process is the use of one molecule of pyruvate and the carbon dioxide released during the formation of acetyl-coA to form oxaloacetate (anaplerotic CO2 fixation). In addition, the pyruvate carboxylase reaction has an important role in citric acid biosynthesis. In A. niger, pyruvate carboxylase is localized in the cytosol, and the oxaloacetate formed is further converted to malate by the cytosolic malate dehydrogenase – the reaction generating 50% of the glycolytically produced NADH. The end product malate is the co-substrate of the mitochondrial tricarboxylic acid carrier in eukaryotes and it is therefore of great importance to citric acid overproduction. Citrate is one of the best-known inhibitors of glycolysis and the paradox of the ability of A. niger to overproduce citrate by an active glycolytic pathway attracts substantial interest. Under particular nutrient conditions, citrate inhibition is counteracted from the accumulation of compounds that act as positive effectors of the phosphofructokinase gene (pfk 1). Ammonium (NH4þ) is such a positive effector since citrate inhibition of pfk 1 seems in vivo to be antagonized by it.13 This antagonism is functionally linked to the well-known effect of trace metal ions, particularly manganese ions, on citric acid accumulation. A critical role in citric acid fermentation has been attributed to manganese ions. Influence of manganese ions on protein synthesis was considered to be of major importance because cycloheximide, an inhibitor of de novo protein synthesis, was found to antagonize the effect of manganese addition. Cellular anabolism of A. niger is impaired under manganese deficiency and/or nitrogen and phosphate limitation. The protein breakdown under manganese deficiency results in a high intracellular NH4þ concentration (the “ammonium pool”), that causes inhibition of the enzyme phosphofructokinase, an essential enzyme in the conversion of glucose and fructose to pyruvate. This leads to a flux through glycolysis and the formation of citric acid.10,12 The high glucose and NH4þ concentrations strongly repress the formation of 2-oxoglutarate dehydrogenase and thus inhibit the catabolism of citric acid within the tricarboxylic acid cycle.

88

Organic Acids

Figure 2

Biochemical pathways of citric acid formation.

1.07.2.2

The Citric Acid Fermentation by A. niger

Citric acid accumulation is strongly influenced by the composition of the medium, particularly in the submerged process. The concentrations of sugar, protons, and oxygen, should be in excess, while those of nitrogen and phosphate, have to be limiting. Concentrations of certain trace metals and especially manganese, should be kept below critical levels. Fungal morphology has early been found to be of importance and the development of certain micro- and macro-morphological forms, is a prerequisite for high yield acid production.14,15

1.07.2.2.1

Carbon Sources

In general, only sugars that are rapidly metabolized by the fungus allow a high final yield of citric acid. The carbon source used in industrial fermentations is typically beet molasses. Other substrates, such as cane molasses, fruit pulps, polysaccharides and sugars

Organic Acids

89

are used if local conditions permit their economic use. However, these low-grade carbon sources usually contain elevated amounts of cations from prior processes which need to be removed. Apart from the type of the carbon source, its concentration is also very important in the citric acid fermentation by A. niger. Concentrations of the carbon sources are usually very high: 100 to over 200 g l1.1 The final yield of the acid increases with increasing initial sugar concentration and the highest production rates are usually achieved at 14%–22% of sugar.

1.07.2.2.2

Nitrogen and Phosphate Limitation

Synthetic laboratory media are supplemented with ammonium salts (ammonium nitrate or ammonium sulfate) to provide the necessary nitrogen. Complex media, e.g., molasses, which are naturally rich in nitrogen-containing compounds, do not require additional nitrogen supplementation. Other sources of nitrogen include yeast/malt extract, urea and nitrates. Balanced concentrations of nitrogen and phosphate are required for citric acid accumulation. Research reports, however, on nitrogen and phosphate limitations are often contradictory and they seem to largely depend on the type of culture. For example, nitrogen limitation was found to be essential for citric acid production in continuous culture, while in batch culture acid accumulation takes place whenever phosphate is limited even when nitrogen is not.

1.07.2.2.3

pH

A low initial pH of the medium is necessary for increased yields in the citric acid fermentation by A. niger. The pH should fall below 2.0 within a few hours of the initiation of the process. This is achieved by the uptake of ammonia by the germinating spores and the subsequent release of protons. Accumulation of citric acid further lowers the pH because of its pK value, at levels around 1.8, even in the absence of other buffering agents. The low pH of the process is inhibitory to the production of other acids, such as oxalic and gluconic acids, through inactivation of the enzymes catalyzing their biosynthesis. Increasing the pH to 4.5 during the production phase reduces the final yield of citric acid by up to 80%. A low pH also reduces the risk of contamination.

1.07.2.2.4

Aeration

The citric acid fermentation process by A. niger has a certain requirement for high dissolved oxygen (DO) concentrations. Increased DO concentrations increase the acid yield, while even a short interruption in aeration has detrimental effects on the yield as it reroutes metabolism toward biomass formation. The biochemical basis of the high DO tension requirement lies on the induction of an alternative respiratory pathway which is required for the re-oxidation of the amounts of NADH produced in glycolysis. NADH recycling is an essential step for maintaining a high flux through the glycolytic pathway. On the other hand, dissolve oxygen tension should be strictly controlled and maintained at the required levels because exceedingly high levels may have the opposite effect by lowering the partial pressure of the dissolved CO2. CO2 is the substrate of pyruvate carboxylase which replenishes the supply of oxaloacetate for citrate synthesis. Sufficient CO2 is produced by the pyruvate decarboxylase reaction to satisfy the stoichiometric demand of the pyruvate carboxylase reaction, but excessive aeration results in losses.

1.07.2.2.5

Trace Metals

Aspergillus niger requires certain trace metals for growth and strictly limited amounts of others for citric acid production, especially in submerged fermentation. The concentration of metal ions below which citric acid is accumulated in appreciable amounts is not absolute and it depends on their relative proportion to other nutrients. Concentrations of Zn, Fe, Cu, Mn, heavy metals and alkaline metals must be limiting and the optimal levels of Zn and Fe are at 0.3 and 1.3 ppm, respectively. The levels of Fe and Zn are probably related to the diversion of carbon between biomass and citric acid, while Cu acts on the growth mechanisms rather than those leading to acid production. Cu ions reduce the effects of Fe ions and this could be due to its antagonistic effect on the uptake of Mn ions. The biochemical basis of the requirement for trace metal limitation and especially that of manganese, has been studied intensively. Manganese plays an important role in various cell functions, including cell wall synthesis, sporulation, and production of secondary metabolites. The influence of Mn ions on protein synthesis was considered to be of major importance for citric acid overproduction. It has been mentioned earlier that cellular anabolism of A. niger is impaired under manganese deficiency and/or nitrogen and phosphate limitation. The protein breakdown under manganese deficiency results in a high intracellular ammonium concentration which inhibits phosphofructokinase leading to a flux through glycolysis and the formation of citric acid. On the other hand, the combination of high glucose and ammonium concentrations represses the formation of 2-oxoglutarate dehydrogenase, inhibiting the catabolism of citric acid within the tricarboxylic acid cycle. Because of these particular conditions of production, high flux rates upstream and reduced flux rates downstream of the accumulation point, citric acid is regarded as an “overflow end product”.1,10,16

1.07.2.2.6

Fungal Morphology

The main factors that affect A. niger morphology in submerged culture are the same factors that influence the productivity of the process: the agitation intensity, broth pH, nutritional factors, the type of the inoculum (spores or vegetative), the amount of the inoculum and the growth rate of the fungus.14 In submerged culture, A. niger morphology varies between two extreme forms: pelleted and free filamentous growth, depending on culture conditions and the genotype of the strain. Whether the pelleted or free filamentous form (macro-morphology) is more desirable for citric acid production is the subject of much controversy. In all cases, however, the mycelium (micro-morphology) of the acidogenic A. niger has been found to conform to the morphological type first

90

Organic Acids

Figure 3

The acidogenic mycelium of Aspergillus niger (scanning electron microscopy).

described by Snell and Schweiger17 in 1951, that of short and swollen hyphal branches with swollen, bulbous tips. This morphological type is induced by intensive agitation, low pH (
2.0), omission of Mn ions, and phosphate limitation.14 Fig. 3 shows the typical acidogenic mycelium of A. niger in submerged fermentation. Each distinctive morphological form has both advantages and disadvantages with regard to overall process productivities. The pelleted form results in less viscous broths and increased mass and heat transfer rates. The cores of pellets, however, become subjected to autolysis and as a result, an often large proportion of the mycelium becomes inactive. The other form, of the free filamentous mycelium, results in viscous and often pseudoplastic broths that reduce mass and heat transfer rates inside the fermenter. Most common in practice, however, is an intermediate morphological type, that of mycelial “clumps” which are stable aggregates of filaments around a small core. Under the appropriate fermentation conditions, citric acid yields exceed 85% with the clumped morphological form.

1.07.2.3

Industrial Production of Citric Acid

There are two different types of fermentation processes for the production of citric acid: the surface and the submerged process. Small amounts are also produced by the Koji process, which is the solid-state version of the surface process. Citric acid production by yeasts is carried out exclusively in submerged processes.1,18 The surface process – introduced around 1920 – is the first method employed on large scale. It is labor-intensive though still in use, even by some major producers of citric acid. It is carried out in shallow aluminum or stainless steel trays, of an area of about 5 m2, filled with medium to a depth of 5–20 cm. The medium is sterilized usually by boiling. Following cooling, the medium is pumped into the trays stacked in racks, in chambers operated in a way so that microbiological conditions are nearly aseptic. Spores are distributed over the surface and sterile air is passed over providing oxygen and cooling. The surface temperature is maintained at 28–30  C. The mycelium grows as a felt on the surface of the medium. Depending on the employed strain and the initial sugar level, the process lasts from 8 to 15 days. Aeration, humidity, temperature, pH, medium depth, nutrient concentrations and trace metal levels are the factors that influence productivity (usually about 1 kg m2 day) and the final yield (
75%). Upon the end of the process, trays are emptied and the broth is filtered to remove the mycelium. The mycelium is then washed and the extracted fluid may contain up to one sixth of the yield of citric acid. The details of the surface process, despite its long history, are not well documented because of restriction of information by the manufacturers. Because of its set up, the particular process is less susceptible to various trace metal effects and variations in dissolved oxygen levels. The surface process remains in use because of its lower power requirements and higher reproducibility compared to the submerged process. The Koji process was originally developed in Japan and it is used in some East Asian countries in small-scale operations. The solid substrate consists of agricultural wastes and the carbohydrate source is usually starch and cellulose. The solid substrate is sterilized by steam, spores are sprayed over it, and the mass is incubated at 30  C. The process lasts 4–5 days. Yields are low due to the difficulty of controlling the process parameters and the concentrations of trace metals. The submerged process is the method of choice in industrialized countries. It requires less labor; it can be largely automated and gives higher production rates. The process is usually carried out in aerated stirred tank bioreactors, although reactors with different

Organic Acids

91

configurations are also used, such as tower and air-lift reactors. Citric acid is produced continuously for several decades in bulk amounts by A. niger in batch processes. Processes have been described for continuous production of citric acid although without commercial application. Fed-batch processes have been used industrially. Production in immobilized systems has also been reported but they did not gain acceptance by the industry. A typical plant that employs the submerged process consists of five parts: media preparation, inoculum reactor, process reactor, broth separation, and product recovery. The low pH of the process, the ability of citric acid to solubilize metal ions, and the content in Mn ions make the requirement for high-grade materials especially important, therefore the metallic parts of equipment are constructed of high-grade stainless steel. The smaller-scale inoculum reactor is inoculated with spores and the process conditions should be adjusted to give rapid growth rather than product formation. The inoculum grows for up to 30 h and the correct stage for inoculation of the process fermenter is judged by the pH levels. The initial phase of the process is critical. An important requirement is the provision of an aeration system that can maintain high DO tension levels. Upon completion of the initial stage, the main reactor is inoculated in a ratio of 1:10. The process is carried out until the production rate is no more economical. This is usually achieved in 5–10 days, depending on the applied methods. Although very high yields are possible, productivity – and not the maximum yield – is a more important consideration on industrial basis.

1.07.2.4

Recovery of Citric Acid

With completion of fermentation, the broth is filtered to remove mycelium. Another step of polishing filtration can be applied to remove remaining mycelia, antifoam and oxalate. Calcium hydroxide is added to the filtrate and citric acid is converted into calcium citrate. The precipitated salt is separated by filtration and the filter cake is treated with sulfuric acid. The produced calcium sulfate precipitates while the filtrate containing the acid is purified, by passing over activated carbon, and demineralized by ion exchanges. The resulting solution, which contains pure citric acid, is evaporated and the acid is obtained as crystals that are easily removed by centrifugation. Citric acid is marketed as anhydrous crystalline chemical, as crystalline monohydrate, and as crystalline sodium salt.

1.07.3

Lactic Acid

Lactic acid (2-hydroxypropanoic acid, C3H6O3, Fig. 4) was first isolated from sour milk in 1798. It occurs in two enantiomeric forms, L-lactic acid, and D-lactic acid, and as a racemic mixture (DL-lactic acid). Almost 70% of the annually produced amount of lactic acid is produced from fermentation processes, while the remainder from chemical manufacture. Synthetic production gives a mixture of equal amounts of the two isomers, L- and D-lactic acid, while the fermentation product could be either of them.18 For most applications, production of lactic acid through fermentation is preferential. Lactic acid is a chemical with extensive use by the food industry which uses solely the L-form since the D-isomer is toxic to humans. Lactic acid is also used by the chemical industry, including polymer technology and it is lately of increased interest for the production of biocompatible and decomposable polymers, such as polylactic acid (PLA).4 The racemate product of chemical synthesis is again difficult to use since it is harder to polymerize than the single isomer forms.2,18 Lactic acid is an expensive chemical when produced by fermentation processes and lactic acid bacteria. It is produced as lactate (a salt of lactic acid, most often a calcium salt) and has to be recovered. To counteract the costly recovery methods, research today is focused on improving the production at cell level. The biochemistry and genetics of lactic acid bacteria has been the subject of indepth research and a significant number of reports are dealing with metabolic engineering of lactic acid bacteria.

1.07.3.1

Microorganisms and the Biochemistry of Lactic Acid Accumulation

Commercial production processes use homolactic lactic acid bacteria, such as Lactobacillus delbrueckii, L. bulgaricus and Lactococcus lactis. Lactic acid can be also produced by other bacteria, e.g., Bacillus coagulans, and fungi, e.g., Rhizopus nigricans, and R. oryzae; however, these are not used in industrial production processes. The biochemical pathway for lactic acid formation by homofermentative lactic acid bacteria and fungi involves the catabolism of glucose or other hexoses through the glycolytic hexose bisphosphate route and subsequent regeneration of NADH by reduction of pyruvate. The homolactic fermentation for lactic acid formation can be summarized as: C6H12O6 þ 2NADþ þ 2ADP þ 2Pi / 2C3H6O3 þ 2NADþ þ 2ATP OH OH

CH3 O Figure 4

Lactic acid.

92

Organic Acids

In this pathway, two molecules of lactic acid can be formed from one molecule of hexose and therefore, the theoretical yield equals 1 g of lactic acid per gram of hexose.

1.07.3.2

Lactic Acid Production Processes

Lactobacillus delbrueckii is the mostly used lactic acid bacterium for lactate production. Industrial strains are characterized by acid tolerance and phage insensitivity besides high yields of lactic acid. The process is carried out in stainless steel stirred tank bioreactors, of up to 100 m3 working volume, at a temperature >45  C (usually 50  C), pH 5.5–6.0 (maintained so with addition of sterile calcium carbonate), and under gentle agitation. The process is micro-aerophilic. The carbon source could be glucose (supplied as glucose syrup), maltose, or sucrose. The concentration of the carbon source should be between 120 and 180 g L1. The medium contains nitrogen and phosphate salts, micro-nutrients, amino acids and vitamins of the B-complex. The raw materials used should be of high purity as this strongly assists the purification procedure. Fermentation lasts until the sugar is metabolized, usually within 48–96 h, with conversion yields between 85% and 95% of the theoretical. Fed-batch and continuous fermentation systems, as well as processes with immobilized cultures, have been described in the literature, but as yet industrial applications have not been realized.

1.07.3.3

Recovery of Lactic Acid

Lactic acid is a highly hygroscopic, syrupy liquid which is commercially available in several different grades, i.e., technical, plastic, food and pharmacopoeia grades. Several approaches can be used for the recovery of lactic acid, depending on the intended use of the commercial product. The basic steps of downstream processing are shown in the diagram of Fig. 5. Plastic-grade lactic acid can be obtained by esterification with methanol after concentrating. Other methods include solvent extraction, ion-exchange separation, vacuum distillation and membrane separation. It is important when higher purity acid is to be obtained that the residual sugar concentration in the fermentation broth has dropped to levels below 0.1% (w/v). Broths containing raw materials of lower quality, as well as accumulation of various fermentation by-products, require more extensive purification steps. For the production of food-grade lactic acid a fermentation medium with a higher-grade sugar source and low protein content is required. The calcium present in the broth is precipitated as calcium sulfate. The solution is then washed, filtered, and treated with activated carbon. Following an evaporation step to about 25% solids and another treatment with activated carbon, the solution is finally evaporated to 50%–65% total acidity.

Carbohydrate substrate

Calcium lactate dissolved by heating

Calcium sulfate prcipitation

Filtration

Concentration

Hexacyanoferrat treatment for the removal of heavy metals

Purification Figure 5

Diagram for recovery of lactic acid from fermentation broth.

Organic Acids OH

OH OH

OH OH Figure 6

Gluconic acid.

1.07.4

Gluconic Acid

93

OH

O

Gluconic acid (pentahydroxycaproic acid, C6H12O17, Fig. 6) is derived from D-glucose by an oxidative reaction catalyzed by glucose oxidase. The products of oxidation of the aldehyde group on the C-1 of b-D-glucose to a carboxyl group are D-glucono-d-lactone and hydrogen peroxide. D-glucono-d-lactone is further hydrolyzed to gluconic acid.2,18 Gluconic acid can be produced by a purely chemical process, by an enzymatic process (in the absence of cells), and by fermentation. Commercially, gluconic acid is produced by fermentation, mainly in the form of gluconates. The main properties of gluconic acid are its extremely low toxicity and low corrosivity, as well as its ability to form water-soluble complexes with a variety of divalent and trivalent ions. Due to these, it is used widely in the food, pharmaceutical and chemical industries. Gluconic acid is a common food and beverage additive although in several food applications such as baking powders for instant products, sausages and pickles, gluconic acid 1,5-lactone is preferred since it enables acid conditions to be reached gradually over a longer period. Applications of gluconic acid in the chemical industry include industrial cleaning (removing rust and scale from metals and glass), metal surface treatment, textile bleach stabilizer and aluminum processing.19 The main product among gluconic acid derivatives is sodium gluconate.20 It is used in detergents, in metallurgy (alkaline derusting), in the paper industry in mixtures with gelatin as sizing agent, and in concrete manufacture in concentrations of 0.02%– 0.2% wt to produce highly resistant concrete to cold weather conditions. Gluconate salts of iron and calcium are used in medicine as carriers of the respective ions in treatments of anemias and osteoporosis. These salts offer low toxicity, stability, and solubility, properties that provide safety and a high concentration of the desired ion.

1.07.4.1

Microorganisms and the Biochemistry of Gluconic Acid Accumulation

Gluconic acid is produced by several prokaryotic, e.g., Pseudomonas, Vibrio, Acetobacter and Gluconobacter spp., and eukaryotic microorganisms, e.g., Aspergillus, Penicillium and Gliocladium spp. Industrially, only A. niger and Gluconobacter oxidans are used as producer organisms in fermentation processes. Production of gluconic acid by A. niger is either the traditional calcium gluconate process or the sodium gluconate process. The alternative process, carried out by G. oxidans, gives 2-oxogluconic acid, 5-oxogluconic acid, and tartaric acid. The initial step in the A. niger fermentation is carried out by the enzyme glucose oxidase that is secreted into the medium. The enzyme uses molecular oxygen in its reaction while its substrate is the b-form of glucose, which is about 150 times more effective than the a-isomer. Glucose oxidase removes two hydrogen atoms from b-D-glucopyranose to produce lactone and hydrogen peroxide. During the conversion, the enzyme is itself reduced by the removal of two hydrogens. It is reoxidized later by oxygen to give the other product of the reaction, hydrogen peroxide. Hydrogen peroxide is cleaved by catalase, while lactone is hydrolyzed either spontaneously or by glucono-d-lactonase (Fig. 7). Aspergillus niger produces all the enzymes that are necessary for the conversion of glucose to gluconic acid. External pH regulates glucose oxidase$ the enzyme is induced by glucose at pH values above 4.0, while it is inactivated at pH values below 2.0. Gluconate is therefore the predominant reaction product at higher pH levels. The other important parameter in the process is oxygen, while lactone accumulation represents a limiting step in the reaction as it reduces the rate of glucose oxidation.

β -D-Glucose Oxygen

glucose oxidase H2O2

D-Glucono-δ-lactone

gluconolactonase

D-Gluconic acid

Figure 7

Gluconic acid formation.

94

Organic Acids

Among bacteria, Pseudomonas and Gluconobacter (Acetobacter) spp. are the main producers of gluconic acid. A membrane-bound dehydrogenase converts extracellular glucose into extracellular gluconic acid. Gluconic acid is not an end product, but it is transported into the cell to be catabolized in the pentose phosphate pathway (PP). At extracellular glucose concentrations above 15 mM, and at pH values below 3.5, the PP pathway is repressed and gluconic acid is accumulated in the medium. D-glucose

1.07.4.2

Gluconic Acid Production Processes

The first process for gluconic acid production by A. niger was developed in the 1930s. To avoid pH decrease, fermentation broth is neutralized by calcium carbonate, and the process is known as the “calcium gluconate process”. Production medium contains 120– 150 g L1 glucose or alternatively dextrose syrup, and limiting amounts of nitrogen and phosphorus for growth restriction. Elevated dissolved oxygen levels have been shown to be advantageous. Fermentation is carried out at 30–33  C and pH 6.5. Yields approximating 95% of the theoretical are obtained by using a stirred tank bioreactor operated at 5.5 vvm aeration (volume air/volume broth/min) and 2 bars pressure in 19 h fermentation. The process involving sodium gluconate production by A. niger differs in many aspects from the above, as for example, the use of higher glucose concentrations and the use of NaOH for pH control. Sodium gluconate is readily soluble in water (39.6% at 30  C) and this is an important point that makes the particular process preferable since the solubility of calcium gluconate in water is low (4% at 30  C). Fed-batch processes with intermittent glucose feedings are also employed by the industry for the production of sodium gluconate. Processes with Gluconobacter species have quite different requirements regarding pH and oxygen concentration. The pH should be below 3.5, while full aeration, without any requirement for increased oxygen pressure, is necessary for increased yields. Glucose concentration in the media should be above 150 g L1. Several different bacterial gluconic acid fermentation processes have been described in the literature including continuous and solid-state fermentation systems, only a few, however, are carried out industrially, all of them being submerged batch fermentations.

1.07.4.3

Recovery of Gluconate and Gluconic Acid

The recovery process depends on the type of final product, the method of broth neutralization and the nature of the carbon sources used. The main manufactured form of gluconic acid is sodium gluconate. The recovery procedure includes filtration of broth to remove biomass and subsequent treatment with activated carbon to decolorize. Carbon is removed by filtration, and the product sodium gluconate is slightly concentrated by evaporation. Crystallization is achieved by cooling the solution. The mass of crystals is then separated with centrifugation and dried. Crystallization at different temperatures is used in order to separate gluconic acid and d-gluconolactone in aqueous solutions. Gluconic acid is crystallized at the temperature range of 0–30  C, while d-gluconolactone at 30–70  C. Ion exchange can also be used for separation. In the case of calcium gluconate product, the broth is concentrated to a hot supersaturated solution of calcium gluconate. The next steps involve cooling at 20  C and addition of water miscible solvents and often a treatment with activated carbon to facilitate the crystallization process. The mass of crystals is separated by centrifugation and the final product is delivered after washing and drying at 80  C. Gluconic acid is marketed as a 50% wt solution. It is readily soluble in water and the equilibrium of lactone and the free acid in solution is dependent on pH and temperature.

1.07.5

Itaconic Acid

Itaconic acid (methylene butanedioic acid, C5H6O4, Fig. 8) was originally known as a thermal decomposition product of citric acid. Chemically, itaconic acid is a functionalized analog of acrylic acid and the simplest conjugated alkenoic acid.18 In the 1930s, it was found that itaconic acid can be produced by Aspergillus itaconicus in fermentation of carbohydrates substrates. Itaconic acid is produced solely by fungal fermentation. It is an expensive product, produced in small quantities, with main uses in the manufacturing of styrene butadiene copolymers.21 For many years there seem to be almost no research interest for the production of itaconic acid and the process remained unchanged since its introduction. The situation, however, has progressively changed. The U.S. Department of Energy lists itaconic acid among the 12 building blocks with the highest potential to be produced by industrial biotechnology. The trifunctional O

O OH

OH O Figure 8

Itaconic acid.

Organic Acids

95

SUBSTRATE Glucose

Itaconate Malate

GLYCOLYSIS

CO2

Oxaloacetate

cis-Aconitate

Malate

Pyruvate CO2

Oxaloacetate

Acetyl-CoA

cis-Aconitate mitochondrion

CELL

Figure 9

Main metabolic steps involved in the biosynthesis of itaconic acid.

structure of itaconic acid allows for the synthesis of new polymers like polyitaconates or related co-polymers. Also polyesters derived from itaconic acid are gaining increased attention lately and their application fields could be numerous. Itaconic acid current production levels are rather low and its uses are kept limited still today. However, the development of new fermentation technologies along with new knowledge on the genetics of its biosynthesis, have led to process improvements and higher yields.6,15,22,23

1.07.5.1

Microorganisms and the Biochemistry of Itaconic Acid Accumulation

Apart from A. itaconicus, itaconic acid is also produced by Aspergillus terreus and both species are used in commercial production.23,24 The biochemistry of itaconic acid formation shares high similarity to that of citric acid. The general pathway from hexose to tricarboxylic acid intermediate is the same (Fig. 9): carbon catabolism via glycolysis and anaplerotic formation of oxaloacetate by CO2 fixation. The difference, however, is made by the presence of the enzyme aconitate decarboxylase in A. terreus which catalyzes the conversion of cis-aconitate to aconitate. The reaction takes place in the cytosol. Aspergillus terreus transports cis-aconitate rather than citrate, in exchange with malate out of the mitochondria.25

1.07.5.2

Itaconic Acid Production Processes

Itaconic acid is produced in batch fermentations. The carbon source in the production medium should be glucose syrup, molasses, or starch hydrolysates, diluted to approximately 10% wt. Phosphate limitation is required for growth restriction while some trace metals should be included in limited amounts. The process pH should be maintained at 2.8–3.2, since lower values favor the formation of itatartaric acid. Other by-products of the fermentation include succinic and citramalic acids. Yields of 50%–60% of the theoretical are obtained in 8–10 days.

1.07.5.3

Recovery of Itaconic Acid

Recovery is usually performed by broth filtration, activated carbon treatment, evaporation, and cooling to induce crystallization. When industrial-grade itaconic acid is required, treatment with activated carbon is omitted.

96

Organic Acids

1.07.6

Succinic Acid

Succinic acid is a four-carbon dicarboxylic acid (butanedioic acid or amber acid, C4H6O4, Fig. 10). It has mostly been produced chemically from the maleic anhydride.18 Succinic acid is used by the chemical industry in the production of polyurethanes, paints and coatings, adhesives, sealants, industrial lubricants, phthalate-free plasticizers, dyes and pigments. It is also used by the food industry as a food additive and flavoring agent and by the cosmetics and pharmaceutical industries in the production of cosmetics and personal care products, and pharmaceutical compounds, respectively. Succinic acid is included in the U.S. Department of Energy’s Top Value Added Chemicals from Biomass due to its potential to become an important building block for deriving both speciality and commodity chemicals. An important area of current market growth is the manufacture of the biodegradable polymer polybutylene succinate (PBS). If produced at sufficiently low cost, succinic acid produced in fermentation can be used as a replacement of maleic anhydride and as a precursor for 1,4 butanediol (BDO), tetrahydrofuran (THF) and g-butyrolactone (GBL).

1.07.6.1

Microorganisms and the Biochemistry of Succinic Acid Accumulation

The acid can be synthesized biologically as an intermediate of the tricarboxylic acid cycle and also as one of the mixed-acid fermentation products of bacteria (anaerobic metabolism).2,18 The anaerobic bacterium Anaerobiospirillum succiniciproducens, and the facultative anaerobes Actinobacillus succinogenes and Mannheimia succiniciproducens, produce succinic acid as a major fermentation product.26 Escherichia coli also produces succinic acid anaerobically, but as a minor fermentation product.27 Anaerobiospirillum succiniciproducens and A. succinogenes produce relatively more succinic acid than other organisms but the yield is kept low by concomitant production of acetic, formic, and lactic acids. Formation of succinic acid anaerobically involves CO2 fixation by carboxylation reactions. The biologically synthesized fourcarbon platform and the assimilation of CO2 during the process, makes the bio-production of succinic acid an attractive process for optimization. Strain improvement, genetic engineering and engineering of the metabolic routes of the producer bacteria, as well as bioprocessing strategies, have been employed with promising results.

1.07.6.2

Succinic Acid Production Processes

At present, succinic acid is mostly produced commercially by the chemical process from maleic anhydride derived from petroleum. Reports of the decade 2000–10 were showing that A. succinogenes is able to produce amounts of succinic acid of approximately 66 g L1, by consuming 98 g L1 glucose after 84 h of batch fermentation. Substrate components included yeast extract, corn steep liquor and MgCO3 to prevent pH drop. Continuous and fed-batch fermentations increased productivity; the yield, however, of succinic acid did not exceed 50% wt which was rather low for commercialization. A decade later, following the huge progress made in genetic and metabolic engineering, the facts have changed and microbial production of succinic acid is an established commercial process with the first plant operated in Italy in 2012. Commercial bio-production of succinic acid is achieved through fermentation of carbohydrates or glycerol substrates using genetically engineered bacteria (mainly E. coli) or yeasts. Details, however, of the industrial process are yet not available in the literature.

1.07.6.3

Recovery of Succinic Acid

The downstream processing for the purification of succinic acid is critical for the cost of production. Separation of by-products (acetic, formic, lactic and pyruvic acids) and the need for recovery of the free acid from its salt form, in which it normally exists in the broth due to the addition of the buffering agent for pH control, are the main steps which can be very costly and not always efficient. Several purification schemes, including electrodialysis, acidification and extraction have been developed. Improved purification processes include those employing precipitation of succinate salts by calcium dihydroxide or reactive extraction using hydrophobic tertiary amines. More recently, an integrated succinic acid recovery process composed of reactive extraction, vacuum distillation and crystallization was developed and produced a final product with the purity of 99.76% wt and the yield of 73.09% wt from the actual fermentation broth of M. succiniciproducens. This process is simpler and more cost-effective than those mentioned above.

O OH Figure 10

Succinic acid.

O OH

Organic Acids

1.07.7

97

Perspectives

Currently the demand for microbially produced organic acids is increasing. The drive behind the trend lies in the emerging attractive applications and in the need to replace expensive hydrocarbon starting materials being used in conventional chemical synthesis. Production of organic acids from bioprocesses can be increased only if it is cost-effective. This means that in many cases longestablished processes may need to be improved or immature processes to be further developed using the tools that joint forces in science offer today.

References 1. Mattey, M. The Production of Organic Acids. Crit. Rev. Biotechnol. 1992, 12, 87–132. 2. Kubicek, C. P. Organic Acids. In Basic Biotechnology; Ratledge, C., Kristiansen, B., Eds., 2nd ed.; Cambridge University Press: Cambridge, 2001; pp 305–324. 3. Cok, B.; Tsiropoulos, I.; Roes, A. L.; Patel, M. K. Succinic Acid Production Derived from Carbohydrates: An Energy and Greenhouse Gas Assessment of a Platform Chemical toward a Bio-based Economy. Biofuels Bioprod Biorefin. 2014, 8, 16–29. 4. Dusselier, M.; Van Wouwe, P.; Dewaele, A.; Maksina, E.; Sels, F. B. Lactic Acid as a Platform Chemical in the Biobased Economy: The Role of Chemocatalysis. Energy Environ. Sci. 2013, 6, 1415–1442. 5. Luskin, L. S. Acidic Monomers. In Functional Monomers; Yokum, R. H., Nyquist, E. B., Eds., Marcel Dekker: New York, 1974; pp 465–500 [Chapter 2]. 6. Sauer, M.; Porto, D.; Mattanovich, D.; Branduardi, P. Microbial Production of Organic Acids: Expanding the Markets. Trends Biotechnol. 2008, 26, 100–108. 7. Knuf, C.; Nielsen, J. Aspergilli: Systems Biology and Industrial Applications. Biotechnol. J. 2012, 7, 1147–1155. 8. Meyer, V.; Fiedler, M.; Nitsche, B.; King, R. The Cell Factory Aspergillus Enters the Dig Data Era: Opportunities and Challenges for Optimising Product Formation. Adv. Biochem. Eng. Biotechnol. 2015, 149, 91–132. 9. Park, H. S.; Jun, S. C.; Han, K. H.; Hong, S. B.; Yu, J. H. Diversity, Application, and Synthetic Biology of Industrially Important Aspergillus Fungi. Adv. Appl. Microbiol. 2017, 100, 161–202. 10. Papagianni, M. Advances in Citric Acid Fermentation by Aspergillus niger: Biochemical Aspects, Membrane Transport and Modeling. Biotechnol. Adv. 2007, 25, 244–263. 11. Papagianni, M.; Wayman, F.; Mattey, M. Fate and Role of Ammonium Ions during Fermentation of Citric Acid by Aspergillus niger. Appl. Environ. Microbiol. 2005, 71, 7178–7186. 12. Saha, B. C. Emerging Biotechnologies for Production of Itaconic Acid and its Applications as a Platform Chemical. J. Ind. Microbiol. Biotechnol. 2017, 44, 303–315. 13. Habison, A.; Kubicek, C. P.; Röhr, M. Phosphofructokinase as a Regulatory Enzyme in Citric Acid Accumulating Aspergillus niger. FEMS (Fed. Eur. Microbiol. Soc.) Microbiol. Lett. 1979, 5, 39–42. 14. Papagianni, M. Fungal Morphology. In Citric Acid Biotechnology; Mattey, M., Kristiansen, B., Linden, J., Eds., Taylor and Francis: London, 1998; pp 69–84 [Chapter 5]. 15. Tate, B. E.; Berg, R. G. Synthesis of Citraconic Anhydride. US Patent 3,835,162 (To Pfizer, Inc.), 1974. 16. Röhr, M.; Kubicek, C. P. Regulatory Aspects of Citric Acid Fermentation by Aspergillus niger. Process Biochem. 1981, 16, 34–37. 17. Snell, R. L.; Schweiger, L. B. Improvements in or Relation to Production of Citric Acid by Fermentation. U.K. Patent. No. 653,808, 1951. 18. Cornils, B.; Lappe, P. Ullmann’s Encyclopedia of Industrial Chemistry, 7th ed.; Wiley-VCH Verlag GmbH: Weiheim, 2007. 19. Ramachandran, S.; Fontanille, P.; Pandey, A.; Larroche, C. Gluconic Acid: Properties, Applications and Microbial Production. Food Technol. Biotechnol. 2006, 44, 185–195. 20. Lu, F.; Li, C.; Wang, Z.; Zhao, W.; Chu, J.; Zhuang, Y.; Zhang, S. High Efficiency Cell-recycle Continuous Sodium Gluconate Production by Aspergillus niger Using On-line Physiological Parameters Association Analysis to Regulate Feed Rate Rationally. Bioresour. Technol. 2016, 220, 433–441. 21. Huang, X.; Lu, X.; Li, Y.; Li, X.; Li, J. J. Improving Itaconic Acid Production through Genetic Engineering of an Industrial Aspergillus terreus Strain. Microb. Cell Factories 2014, 13, 119. 22. Shekhawat, D.; Jackson, J. E.; Miller, D. J. Process Model and Economic Analysis of Itaconic Acid Production from Dimethyl Succinate and Formaldehyde. Bioresour. Technol. 2006, 97, 342–347. 23. TNO Quality of life, T. N. O. Identification of Key Genes in Itaconic Acid Production by Aspergillus terreus. TNO Newslett. 2008, 9. 24. Okabe, M.; Lies, D.; Kanamasa, S.; Park, E. Y. Biotechnological Production of Itaconic Acid and its Biosynthesis in Aspergillus terreus. Appl. Microbiol. Biotechnol. 2009, 84, 597–606. 25. Steiger, M. G.; Blumhoff, M. L.; Mattanovich, D.; Sauer, M. Biochemistry of Microbial Itaconic Acid Production. Front. Microbiol. 2013, 4, 23. 26. Lee, S. J.; Song, H.; Lee, S. Y. Genome-based Metabolic Engineering of Mannheimia succiniciproducens for Succinic Acid Production. Appl. Environ. Microbiol. 2006, 72, 1939–1948. 27. Lee, S. J.; Lee, D. Y.; Kim, T. Y.; Kim, B. H.; Lee, J.; Lee, S. Y. Metabolic Engineering of Escherichia coli for Enhances Production of Succinic Acid, Based on Genome Comparison and in Silico Gene Knockout Simulation. Appl. Environ. Microbiol. 2005, 71, 7880–7887.

1.08

Peptides and Glycopeptides

NW Owens and F Schweizer, University of Manitoba, Winnipeg, MB, Canada © 2011 Elsevier B.V. All rights reserved. This is a reprint of N.W. Owens, F. Schweizer, 1.10 - Peptides and Glycopeptides, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 121-138.

1.08.1 1.08.2 1.08.2.1 1.08.2.2 1.08.3 1.08.3.1 1.08.3.2 1.08.4 1.08.4.1 1.08.4.1.1 1.08.5 1.08.6 1.08.6.1 1.08.6.2 1.08.6.2.1 1.08.6.2.2 1.08.6.2.3 1.08.7 1.08.7.1 1.08.7.2 1.08.8 1.08.8.1 1.08.8.2 1.08.8.3 1.08.9 1.08.9.1 1.08.9.2 1.08.9.3 1.08.9.4 1.08.10 1.08.10.1 1.08.10.2 1.08.10.2.1 1.08.10.2.2 1.08.11 References

Introduction Peptide Hormones Angiotensin II and Bradykinin Oxytocin and Vasopressin Neuropeptides Substance P Neuropeptide Y Antibacterial Peptides Defensins Cecropins Glycosylation Is a Common and Important Posttranslational Modification of Peptides Common Glycosidic Linkages N-Glycosylation O-Glycosylation Threonine O-Glycosylation in Drosocin 5R-Hydroxy-L-lysine O-Glycosylation in Collagen 4R-Hydroxy-L-proline O-Glycosylation in Hydroxyproline-Rich Glycoproteins Peptide Synthesis Solid-Phase Peptide Synthesis Recombinant Peptide Synthesis Glycopeptide Synthesis Controlling Regio- and Stereoselectivity Formation of the Glycosidic Linkage Strategies for Glycopeptide Synthesis Peptides and Glycopeptides as Models of Proteins and Glycoproteins Mucin Glycoprotein Model Peptides Antifreeze Glycoprotein Model Peptides Collagen Glycoprotein Model Peptides HRGP Model Peptides Application of Synthetic Peptides and Glycopeptides for the Treatment of Disease A Peptide-based Malaria Vaccine Glycopeptide-based Cancer Vaccines A Glycopeptide-based Vaccine With Multiple Tumor-Associated Carbohydrate Antigens A Glycopeptide Vaccine Based on the MUC1 Glycoprotein Summary

99 100 100 100 101 101 101 102 103 103 103 104 104 104 104 104 105 105 106 107 107 107 107 108 109 109 110 110 110 111 111 112 112 112 113 114

Glossary Antibody A protein produced by humans or higher animals in response to exposure to a specific antigen and characterized by specific reactivity with its complementary antigen. Antigen A substance, usually a protein or carbohydrate, that, when introduced into the body of a human or higher animal, stimulates the production of an antibody that will react specifically with it. Chromatography A process of separating gases, liquids, or solids in a mixture or solution by adsorption as the mixture or solution flows over the adsorbent medium, often in a column. The substances are separated because of their differing chemical interaction with the adsorbent medium. Conformation The three-dimensional structure of a molecule. A change in conformation is caused by bond angle rotation, and not from bond breaking.

98

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00010-0

Peptides and Glycopeptides

99

Glycan A carbohydrate structure attached to another biomolecule, such as a peptide, protein, or lipid. This usually consists of O-linked monosaccharides in a linear or branched structure. Glycobiology The study of the structure, biosynthesis, and biology of saccharides and glycoconjugates in nature. Glycosylation The attachment of carbohydrate onto a molecule, such as a peptide or protein. Immunization The process in which the immune system creates antibodies against a particular foreign pathogen in response to vaccination or infection. Oligosaccharide A short polymer of saccharides, usually composed of 3–10 residues. Protecting group A reversible chemical modification used to render a functional group unreactive in order to obtain chemoselectivity. Receptor A protein that is embedded either in the plasma membrane or in the cytoplasm of a cell, to which a mobile signaling molecule may attach. A molecule which binds to a receptor is referred to as a ligand.

1.08.1

Introduction

Our appetite, blood pressure, and ability to feel pain rely on biochemical processes that involve short oligomers of amino acids, called peptides. This fact has only been recently discovered, and most of our knowledge of peptide and protein structure and function has arisen in the last 50 years. We now know that peptides and proteins, and their glycosylated counterparts, are critical for the proper functioning of cells and organisms.1 The mechanism by which peptides carry out their life-sustaining function is through diverse biological roles, which include acting as hormones, neurotransmitters, and antibiotics. Peptide function is often compared with that of proteins, and while they share many of the same physical characteristics, relative to proteins, peptides have a narrower range of activity. However, much diversity exists within each of these roles. Although there is no definitive distinction between the two, peptides are, in effect, short proteins. They can be generally characterized as having a linear structure in which 2–50 amino acids are linked through amide bonds. Because of their relatively short length, peptides might be expected to be conformationally disordered in solution; however, they can adopt distinct secondary structures, such as a-helices, b-strands, and reverse turns.1 In some cases, peptides are linked, typically through disulfide bridges, to form cyclic or multimeric structures. Regardless of their type of structure, many peptides exert their biological function through interaction with a specific receptor, but can also act in a nonspecific manner through association with lipid membranes. In some cases, the low molecular weight of peptides allows them to be able to cross cellular membranes by passive diffusion, which expands their functional capability.1 In general, peptides can undergo the same posttranslational modifications as proteins, such as phosphorylation and glycosylation. The naturally occurring modification of peptides by glycosylation has diverse effects on their function and physical characteristics, such as altering their conformation, hydration, ability to cross membranes, and stability.2 The importance and scope of the role of peptides in biology became apparent only in the 1950s as the technology for isolating peptides from natural sources improved. Peptides are present in very minute quantities in vivo. For example, femto- (10 15) to pico(10 12) molar concentrations of a given peptide per milligram of tissue is not unheard of.1 The use of size-exclusion and countercurrent chromatography provided a means for isolating peptides and led to numerous discoveries in the field. Advancements in the chemical synthesis of peptides have resolved many of the problems that arise from their isolation from natural sources. In order to establish structure–activity relationships, access to synthetic peptides provides important structural evidence of a given peptide isolated from a natural source in order to correlate its biological activity. Much of our understanding of the biological roles of peptides as hormones, neuropeptides, and antibiotics has arisen because of the ability to chemically synthesize them. It was in the early 20th century that Emil Fischer established the first successful synthesis of a peptide (the dipeptide glycyl-glycine), which laid the foundation for future advancements in peptide synthesis. More recently, advancement in peptide chemistry, such as the development of solid-phase and chemoenzymatic synthesis, has revolutionized the field by lowering the cost and time requirements for peptide synthesis. Similarly, advancements in carbohydrate chemistry have provided access to synthetic glycopeptides, which has unleashed a flurry of activity in the field of glycobiology for understanding the role of specific glycopeptides. Furthermore, the field of protein chemistry has benefited from access to synthetic peptides and glycopeptides, as they have also emerged as invaluable models for determining the function and characteristics of larger and sometimes unwieldy proteins and glycoproteins. In many cases, proteins exert their function through short regions of their sequence, and provided the native secondary structure is maintained, short model peptides can be used to replicate and model the function of the native protein. In other circumstances, short peptides can be used to understand some physical aspects of protein structure, such as how glycosylation affects the conformation of a particular protein. In this way, much information about proteins and glycoproteins has been garnered from the use of carefully selected model peptides. Finally, access to peptides and glycopeptides with well-defined structures has allowed for the development of peptide- and glycopeptide-based therapeutics. For example, as peptides also play an important role in the adaptive immune response, they can be used to generate vaccines against many harmful pathogens, including viruses, bacteria, and even tumors. Currently, a large amount of research is focused on generating glycopeptide-based vaccines for the treatment of cancer and has produced some promising lead candidates. Here, we will explore the topic of peptides and glycopeptides, beginning with a brief introduction to the biological roles of several important classes of peptides.

100

1.08.2

Peptides and Glycopeptides

Peptide Hormones

Peptides are used by most animal species, including humans, as hormones to induce various biological responses.3 This broad evolutionary selectivity may have arisen because peptides can be rapidly degraded through the action of proteases after they have performed their task. This allows for tight control and regulation of their function.

1.08.2.1

Angiotensin II and Bradykinin

Despite having short sequences of amino acids, peptide hormones have varied functions and act with high specificity. For example, angiotensin II (Asp-Arg-Val-Tyr-Ile-His-Pro-Phe-His-Leu) (Figure 1) and bradykinin (Arg-Pro-Pro-Gly-Phe-Ser-Pro-Phe-Arg) (Figure 2) are peptide hormones that are released by the liver to cause the constriction and expansion of blood vessels, respectively.4,5 The body will constrict blood vessels in response to cold or to prevent blood loss from a wound. Angiotensin II, together with the protein renin and the steroid aldosterone, constitutes one of the most important hormonal systems for controlling blood pressure. It is one of a few mechanisms by which the body can cause vasoconstriction, and acts by binding to several angiotensin receptors on the smooth muscle that surrounds veins and arteries. Its production is controlled through enzymes that act on protein and peptide precursors, angiotensinogen and angiotensin I. It is now believed that angiotensin II is expressed in nearly every organ, and is implicated in multiple physiological processes, including cognitive function, aging, and reproduction.4 Bradykinin also acts on smooth muscle by binding to specific receptor proteins. The body causes vasodilation to increase blood flow throughout the body or only to specific organs in response to various situations.5 Finally, both angiotensin I and bradykinin are extremely flexible in solution, but have been shown to adopt distinct conformations when bound to their protein receptors found on cell surfaces, which is typical of most peptide hormones. In 1992, Garcia and coworkers attempted to uncover the bioactive conformation of angiotensin II through the analysis of the X-ray crystal structure of the peptide bound to a monoclonal antibody, acting as a surrogate receptor.6 They confirmed that angiotensin II binds to the protein receptor through a turn region involving the Ile, His, and Pro residues (Figure 1). The center of the turn was lodged in the deepest region of the binding site, which reflects its importance in the binding process. Similarly, extensive molecular modeling and nuclear magnetic resonance (NMR) studies of the vasodilator bradykinin by Kyle and coworkers found that the four terminal residues Ser-Pro-Phe-Arg in fact form a b-turn (Figure 2).7 This turn involves a hydrogen bond from the arginine backbone amide proton to the serine backbone carbonyl group. Therefore, both peptides adopt specific conformations when bound to their respective receptors.

1.08.2.2

Oxytocin and Vasopressin

In some cases, peptides are less conformationally flexible because they are cyclized through covalent bonds, which are typically disulfide linkages. Oxytocin (Cys-Tyr-Ile-Gln-Asn-Cys-Pro-Leu-Gly-NH2) (Figure 3) and vasopressin (Cys-Tyr-Phe-Gln-AsnCys-Pro-Arg-Gly-NH2) are cyclic nonapeptides in which the first and sixth amino acids are linked through a disulfide bridge. Proline

Angiotensin II

Amino acids found to be directly involved in the binding event Arg

Asp O

Val

Ile

Tyr

N

O O

H3N

NH

O H N

H N N H

N H

O

O

O O O

OH H2N

His

H N

N

Pro

HN

NH

O HN

O NH2 NH O

Phe

HN O N

His

Leu Figure 1 Structure of the peptide hormone angiotensin II, which causes blood vessel constriction. The residues Ile, His, and Pro have been found to be buried deepest in the receptor-binding pocket.

Peptides and Glycopeptides

101

Bradykinin Phe Pro

Arg Pro

NH2 H2N

Gly

Ser O

H N

H N O

O O

O NH3

Four terminal residues form a turn, which is involved in the receptor binding interaction

O

OH

O

N

Pro N

N H

N

NH

O

NH H N

Phe

HO O

HN

H2N

Arg NH2

Figure 2 Structure of the peptide hormone bradykinin, which causes blood vessel expansion. While it is flexible in solution, its C-terminal residues form a b-turn conformation when bound to its target receptor. The intramolecular hydrogen bond between arginine and serine is shown.

is the first amino acid in a tripeptide ‘tail’ attached to the ring. Despite similar amino acid sequences, these two peptide hormones have very different functions. Oxytocin is present in most mammals and acts during pregnancy to cause uterine contraction and milk production, whereas vasopressin mainly acts on the kidneys to conserve water.8 Both peptide hormones are released from the pituitary gland and have specific and high-affinity receptors in their targeted tissues. Because of significant interest in developing synthetic analogs of oxytocin, efforts have been made to further understand its biologically active conformation, especially the conformation of the proline-linked tripeptide appendage. It is now known that the N-terminal amide conformation of proline affects the function of oxytocin. It seems that when proline is in a cis-amide conformation (Figure 3(A)), the peptide displays very strong binding affinity for the oxytocin receptor, but must isomerize into a trans-amide conformation to exert its function (Figure 3(B)).9 These examples serve to illustrate that peptides have diverse roles as hormones, where knowing the conformation of the peptide has been found to be important for understanding and replicating their biological activity.

1.08.3

Neuropeptides

Peptides also transmit information through the action of neurons – an effect that was first discovered in the late 1960s, and which has only recently become widely accepted. To date, more than 100 peptides have been discovered that are involved in the function of the nervous system.10 Many neuropeptides are neurotransmitters, which are chemicals that relay a signal between neurons or from a neuron to a different type of cell. Neuropeptides regulate many functions in the body: growth, water and salt metabolism, temperature control, water and food intake, cardiovascular, gastrointestinal and respiratory control, memory, and behavior. Their presence in unicellular organisms, plants, invertebrates, and vertebrates indicates their importance as chemical messengers in the biosphere. The structure of neuropeptides has been well conserved throughout evolution, where many neuropeptides have the same sequence in different species, but may carry out different functions. For example, human insulin is extremely similar to porcine and bovine insulin.1

1.08.3.1

Substance P

The peptide substance P (Arg-Pro-Lys-Pro-Gln-Gln-Phe-Phe-Gly-Leu-Met), named after the powder it forms after extraction, acts as a neurotransmitter.10 It plays an important role in pain perception and in the transmission of pain information in the central nervous system. The peptide is released in the sensory neurons in skin and muscle tissue to signal a painful or cell-damaging event to the brain. Substance P is formed through the cleavage of a precursor protein and is shipped in vesicles to the axon terminals. Once released into the synaptic cleft, the peptide binds to neurokinin class receptors. The peptide itself does not adopt any distinct secondary structure in solution, but rather is simply characterized by an extended conformation. Because of its propensity to increase inflammation, substance P is now implicated in diseases and conditions associated with inflammation, including asthma and chronic bronchitis.

1.08.3.2

Neuropeptide Y

Neuropeptide Y (NPY), named after its two terminal tyrosine residues, contains 36 amino acids, and has been implicated in several physiological processes, including controlling our stress level and regulating our food intake.10 The peptide is secreted by the

102

Peptides and Glycopeptides

Oxytocin Gln

A H2N

Asn

O

O

Oxytocin has its highest binding affinity when proline is in a cis-amide conformation

NH2

O

O N H

Pro

O

NH

HN

Cys

O

N

Leu

S S HN

Gly

O H N

Ile

HN

O O

O

NH

NH3 O

Cys

Tyr

NH2

OH

Gln H2N

Asn

O

O

B

NH2

O

Pro

Leu

O

O

N H NH

HN

H N

N

O

NH2 N H

Cys

O

Gly

O

O

S S HN H N

Ile O

NH3

O

Oxytocin in its bioactive conformation in which proline is in a trans-amide conformation

Cys

Tyr OH

Figure 3 Oxytocin is a cyclic peptide hormone. The N-terminal amide conformation of proline affects the binding affinity and agonistic effect of the hormone. (A) The cis-amide conformer has a strong binding affinity to the target receptor; (B) however, isomerization to the trans-amide conformer is required for maximum biological activity.

hypothalamus, a region of the brain that connects the nervous and endocrine systems, and signals for an increase in appetite and a decrease in physical activity. The peptide binds to membrane-bound protein receptors known to cause changes in metabolism. Essentially, NPY signals for an increase in energy storage. Because of current concerns about obesity, the neuropeptide has attracted significant attention as a mechanism by which food intake is controlled. Furthermore, NPY seems to act in response to stress. In mouse model studies, NPY levels increased in an environment of high stress, as well as in a high-fat and high-sugar diet, which caused weight gain specifically in the abdominal regions of the mice. Therefore, it appears that NPY plays a role in linking stress with weight gain. More recently, NPY has also been implicated in modulating the immune response, specifically affecting macrophage, T-cell, and cytokine release, which indicates that NPY also plays a role in linking our stress levels with our immune response. Substance P and NPY are only two examples of neuropeptides, but are indicative of the broad role that peptides play in many aspects of physiology, and the interactions between the nervous and endocrine systems. Besides their propensity to bind to specific receptors, some peptides also have the ability to bind and disrupt bacterial membranes.

1.08.4

Antibacterial Peptides

Bacteria remain a constant threat not only to humans but also to all organisms on Earth. Incredibly, almost every species relies on peptides as a defense mechanism against invasive bacteria to some degree. Plants, mammals, insects, birds, crustaceans,

Peptides and Glycopeptides

103

amphibians, fungi, and even bacteria themselves use peptides to thwart bacterial infections.11 The widespread presence of antibacterial peptides in the plant and animal kingdoms suggests that they have played an important role in the evolutionary development of complex multicellular organisms. These peptides are usually composed of fewer than 60 amino acids and are characterized by having a net positive charge from a surplus of basic amino acids. The fundamental structural principle underlying all classes of antimicrobial peptides is the ability to adopt a shape that clusters hydrophobic and cationic amino acids in discrete regions, leading to amphipathic character. Although the mode of action of these peptides is not fully understood, most antibacterial peptides increase the permeability of the bacterial cell membrane, which causes lysis and cell death.11 Generally, the net positive charge of the peptide allows for binding to the negatively charged phospholipid head groups of the outer leaflet of the bacterial membrane. After initial electrostatic binding, the hydrophobic region of the peptide inserts itself into the outer leaflet of the membrane, thereby causing a relative strain in the outer leaflet relative to the inner leaflet. This strain is relieved by a phase transition in which peptides reorient across the membrane, and it is through the association of multiple peptides that pores are formed in the bacterial membrane. As more and more peptides bind to the surface of the outer leaflet, the membrane will begin to disintegrate as vesicles coated with the cationic peptides are formed. Eukaryotic cells are typically not affected by these peptides because of the fundamental differences in the composition of the membranes. Primarily, cholesterol, present only in eukaryotic cells, has the effect of making the cell membrane more fluid and adaptive to pore formation from the peptides. Furthermore, the outer leaflet of the membranes of plants and animals is mostly composed of lipids with no net charge. Because of the lack of negatively charged groups on the exterior of human cell membranes, antibacterial peptides discriminate human cells from bacterial cells and selectively bind to the latter.

1.08.4.1

Defensins

The defensins represent a well-studied class of antibacterial peptides. They are present in both vertebrate and invertebrate species and are active against both Gram-positive and Gram-negative bacteria, as well as fungi, viruses, and protozoa.12 The defensins are typically composed of 29–34 residues and contain 3 disulfide bridges, which give them a relatively rigid antiparallel b-sheet structure in which patches of cationic and hydrophobic residues are aligned to form an amphipathic configuration. Subtypes of the defensins, which includes a-, b-, and the less common b-variety, vary in several respects: the spacing and pattern of the disulfide bridges, the segment of the precursor molecule which is cleaved off, and the position and composition of the gene that codes for the peptides. In mammals, the peptides are expressed in cells of the immune system, such as neutrophils, as well as in the epithelial cells of the skin, tongue, esophagus, and respiratory tract. This indicates that these peptides are used as a first line of defense against invasive bacteria.

1.08.4.1.1

Cecropins

Another class of antibacterial peptides, the cecropins, named after the discovery of their presence in the cecropia moth, were among the first antibacterial peptides to be isolated.12 The cecropins are characterized by having 35–39 amino acids and an amidated C-terminus. They are linear peptides that adopt an a-helical secondary structure upon contact with a bacterial membrane. Both the defensins and cecropins have a broad-spectrum antibacterial activity against both Gram-positive and Gram-negative bacteria, fungi, and other parasites, which is not always the case with antibacterial peptides. However, several classes of antibacterial peptides are typically present in a single organism to cover all potential sources of infection. For example, the fruit fly Drosophila melanogaster produces at least seven distinct groups of peptide antibiotics, each peptide assuming a completely different activity spectrum.13 Because of their broad-spectrum activity against bacteria, which includes emergent antibiotic-resistant bacteria such as methicillin-resistant Staphylococcus aureus and a relatively low risk of bacterial resistance, antibacterial peptides are currently targets for the development of new antibiotics.

1.08.5

Glycosylation Is a Common and Important Posttranslational Modification of Peptides

Peptides are synthesized and modified through various biological mechanisms.1 Some peptides are synthesized in a similar fashion to proteins, through translation of messenger RNA in the endoplasmic reticulum. In other cases, bioactive peptides are created by cleavage of larger precursor proteins. Typically, peptides and proteins are modified by phosphorylation and glycosylation during and after the process of translation. The modification by glycosylation has various effects on the function and characteristics of peptides and proteins, many of which are still being understood.2 Until recently, the role of carbohydrate structures, or glycans, attached to peptides and proteins was unknown. Carbohydrates, or saccharides, were thought of as playing a role in only energy storage. However, in the past 20 years, the field of glycobiology has demonstrated that carbohydrates are responsible for many critical biological processes, from protein trafficking to cellular communication to organism development.14 Many of these events are predicated on binding interactions between a glycan and a protein receptor. However, glycosylation also affects important aspects of peptide and protein structure, including their three-dimensional shape, thermal stability, and their hydration levels. Finally, glycosylation can also affect how quickly peptides and proteins are digested, and their ability to cross cell membranes. Glycopeptides and glycoproteins are usually categorized based on the predominant type of glycosidic linkage present in their structure, in other words, how the sugars are linked to the chain of amino acids. Typically, carbohydrates are N-linked to the amide

104

Peptides and Glycopeptides

nitrogen atom of L-asparagine or O-linked to the oxygen atom of the amino acids L-serine and L-threonine.14 Less common examples include O-glycosidic linkages to 5R-hydroxy-L-lysine and 4R-hydroxy-L-proline.

1.08.6

Common Glycosidic Linkages

1.08.6.1

N-Glycosylation

The N-linked glycans are characterized by a conserved core structure of sugars in which an N-acetylglucosamine residue is b-linked to an L-asparagine residue.14 The core glycan structure is synthesized in a type of biological assembly line located in the lumen of the endoplasmic reticulum. Enzymes called glycosyltransferases are responsible for transferring carbohydrates in a stereo- and regioselective manner onto a lipid scaffold, called dolichol. At the appropriate moment, the glycan structure is transferred to asparagine residues in polypeptides containing the sequence Asn-Xaa-Ser/Thr, where Xaa is any amino acid except proline. After the transfer of the conserved core structure, the glycopeptides and glycoproteins are transferred to the Golgi apparatus where sugars are selectively trimmed by endoglycosidases and added by glycosyltransferases, leading to a very diverse group of N-glycan structures. The resulting glycopeptides and glycoproteins either are transferred to the plasma membrane and are secreted outside the cell or remain within the cell to carry out their function.

1.08.6.2

O-Glycosylation

In contrast to the N-linked glycans, the biosynthesis of O-linked glycans occurs by the consecutive attachment of carbohydrate residues in the Golgi apparatus to serine or threonine residues in a peptide or protein.2 There is no special sequence of amino acids that signal for O-glycosylation. However, this modification tends to occur in sequences rich in Ser, Thr, Pro, Gly, and Ala. While many diverse glycan structures are formed, the O-linked glycans are typically b-linked to serine or threonine through an N-acetylgalactosamine residue. O-Linked glycoproteins typically have much larger glycan structures than their N-linked counterparts, and, as such, these highly hydrated structures perform the role of forming lubricating and protective barriers at the surface of tissues. In some cases, however, the glycans are much smaller in size, and as is the case for some antibacterial peptides, their role is less clear.

1.08.6.2.1

Threonine O-Glycosylation in Drosocin

Several O-glycosylated antibacterial peptides are known to exist in nature. For example, drosocin is a peptide composed of 19 amino acids and is rich in the cationic amino acids lysine and arginine, as well as the nonpolar cyclic amino acid proline.11 Drosocin has been found to be active against Gram-negative bacteria such as Escherichia coli and Klebsiella pneumoniae, but is poorly active against Gram-positive bacteria such as Pseudomonas aeruginosa. It is O-glycosylated at a single threonine residue to the disaccharide galactosyl-b-1,3-N-acetylgalactosamine, which has been found to enhance its antibacterial activity. However, the reasons behind the effect of glycosylation on antibacterial activity remain unclear. It has been speculated that glycosylation may affect the interaction of the peptide with the membrane and associated receptors, association with the peptidoglycan that surrounds the bacterial cell membrane, provide protection against degradation, and/or act as a competitive inhibitor of enzymes responsible for peptidoglycan synthesis.15 In 2002, Gobbo and coworkers demonstrated that the antibacterial activity of drosocin can be affected by the characteristics of the glycan attached to the threonine residue, in particular, the size of the glycan (mono-vs. disaccharide), the type of sugar (D-galactose (Gal) vs. D-N-acetylgalactosamine (GalNAc)), and the type of glycosidic linkage (a-O or b-O) (Figure 4).15 It was found that removal of the outer galactose residue from the native disaccharide glycan structure only slightly decreases the activity of the peptide. In addition, in peptides containing just a single carbohydrate residue attached to threonine, only a small decrease in antibacterial activity was observed if the native N-acetylgalactosamine monomer was replaced with a galactose monomer. Also, there seemed to be very little difference in activity between the a-O- and b-O-linked galactose sugars. Therefore, the outer glycan residue, N-acetyl group, and the stereochemistry at the anomeric carbon do not seem to be responsible for the antibacterial activity of drosocin. Interestingly, the activity of drosocin was tested in the presence of a high concentration of Gal or GalNAc as possible competitive inhibitors of the glycosylated peptide, and no difference in activity was observed. This suggests that a direct interaction between the glycan and a receptor is not responsible for its antibacterial activity. Therefore, it remains unclear as to how the native glycan of drosocin affects its biological activity.

1.08.6.2.2

5R-Hydroxy-L-lysine O-Glycosylation in Collagen

The O-glycosylation of 5R-hydroxy-L-lysine residues, while relatively uncommon, is known to occur to collagen molecules. Collagen is the predominant structural protein in humans and animals, where it forms extended fibrous networks that give tissues their structural integrity.16 These collagen fibers are composed of bundles of tropocollagen molecules, which have a characteristic tertiary structure of three polypeptide strands, each in a left-handed polyproline II (PPII) conformation, twisted into a right-handed triple helix. The polypeptide strands have a repeating Xaa-Yaa-Gly sequence, where Xaa is often L-proline and Yaa is often 4R-hydroxyL-proline (Hyp). Typically, there are nearly 300 repeats of the Xaa-Yaa-Gly sequence per strand in the collagen molecule. The glycans galactose and glucosyl-a-1,2-galactose b-O-linked to 5R-hydroxy-L-lysine have been implicated in the secretion and assembly of collagen fibrils,17 embryonic development and cell viability,18 and the interaction of collagen with protein receptors19; however, their role is still not fully understood.

Peptides and Glycopeptides

R

MIC (μM) against selected bacterial strains E. coli ML-35

OH OH O

OH OH O

HO

105

O HO

AcHN

K. pneumoniae 79

P. aeruginosa ATCC 27853

0.5

4

>128

1

8

>128

2

32

>128

1

8

>128

2

64

>128

O

OH OH O HO AcHN

R Gly-Lys-Pro-Arg-Pro-Tyr-Ser-Pro-Arg-Pro-Thr-Ser-His-Pro-Arg-Pro-Ile-Arg-Val

O

OH OH O HO HO

O

OH OH HO

OH O O

H

Figure 4 Antibacterial activity of glycosylated analogs of drosocin. (From top to bottom) Native disaccharide glycan Gal-b-1,3-GalNAc-a-O-Thr, monosaccharides a-O-GalNAc, a-O-Gal, and b-O-Gal, and unglycosylated drosocin. The different glycan structures showed different minimum inhibitory concentration (MIC) against various strains of bacteria. Variation of the glycan did not greatly affect the MIC values of the drosocin analogs.15

1.08.6.2.3

4R-Hydroxy-L-proline O-Glycosylation in Hydroxyproline-Rich Glycoproteins

Interestingly, the O-glycosylation of Hyp is not known to occur in humans at all, but is widespread in the plant kingdom in the form of hydroxyproline-rich glycoproteins (HRGPs). The O-glycosylation of Hyp is widespread in the plant kingdom and occurs in HRGPs that are associated with the cell walls of algae and flowering plants.20 By comparison, the stereoisomer of Hyp, 4S-hydroxy-L-proline, is rarely found in nature – having only been isolated from extracts of the sandalwood tree Santalum album, several species of fungi, and the cyanobacteria Lyngbya majuscule.21 The reasons as to why nature prefers the 4R-stereoisomer remain unclear. HRGPs are associated with forming the protective extracellular network of plant cell walls for algae and flowering plants. They are characterized by a PPII conformation and short homooligomers of L-arabinofuranose and larger heteropolymers of L-arabinose (Ara) and Gal.20 The functional consequences of glycosylation are still unclear, but indications are that glycosylation contributes to the stability of the PPII conformation in HRGPs. Early work by Derek T. A. Lamport using molecular modeling indicated that b-linked tetrasaccharides of Ara attached to Hyp in polyproline peptides would ‘nest’ with the polyproline II type helix and form three hydrogen bonds to the carbonyl groups of the peptide backbone.22 This indicated that the carbohydrate moiety might provide HRGPs with a rod-like structure and high tensile strength, in a similar fashion to the stability provided to collagen from the triple helix structure. The glycosylation of HRGPs has also been implicated in increasing their solubility and resistance to proteolytic degradation through steric hindrance to proteases. These attributes may have been the selection pressure responsible for the evolutionary origin of HRGPs. Evidence for this was exemplified by the work of Esquerré-Tugayé and Maxau, who found that both Hyp and glycosylation levels in plant cell walls increase up to 10-fold after bacterial infection.23 Although it has been accepted that glycosylation is important for the integrity of the plant cell wall and contributes to its stability, few studies have explored in detail the contributions to stability of HRGPs garnered upon Hyp glycosylation. In 2001, Ferris and coworkers compared the circular dichroism (CD) spectra of the glycosylated and nonglycosylated forms of the HRGP GP1 isolated from Chlamydomonas reinhardtii.24 They found that upon chemical deglycosylation, analysis by CD showed that there was a significant decrease in the net intensities of the spectrum extrema, which was correlated to a transition from a structured to a less structured molecule. The authors concluded that the carbohydrate side chains reinforce the PPII conformation of GP1.

1.08.7

Peptide Synthesis

Our current understanding of the various biological functions of peptides has largely relied on advances in chemical synthesis. The creation of synthetic peptides with well-defined sequences allows for the investigation of their role in cell- or animal-based models. Synthetic peptides can also have tremendous economic value. For instance, the artificial sweetener aspartame is a dipeptide 25 L-aspartyl-L-phenylalanine methyl ester and has an annual worldwide production of more than 16,000 metric tons. Peptides are formed through the condensation of carboxylic acid and amino groups of amino acids; however, chemical protecting groups are required to provide control over this reaction. For example, in the coupling reaction of two amino acids, in order to react only the carboxylic acid group, the amine must be protected to prevent self-condensation (Figure 5). Similarly, the acid group of the amino acid in which the amine is to react must be protected. Further, reactive side-chain functional groups are protected in order to prevent the occurrence of side reactions.

106

Peptides and Glycopeptides

R1

O X

H2N

+

H2N

O

Complex mixture of products

OH R2

O

R1

P

X N H

+

O

Only

O

H N

P

O O

= protecting group

P

N H

R2

P

O

R1

P

H2N

R2

X = activating group

Figure 5 Coupling of two amino acids without the use of protecting groups leads to a complex mixture of products involving self- and cross-condensation, as well as uncontrolled oligomerization. However, appropriate use of protecting and activating groups allows for controlled formation of the target amide linkage.

1.08.7.1

Solid-Phase Peptide Synthesis

Early approaches to the chemical synthesis of peptides relied on solution-phase synthesis, in which amino acids are coupled in a reaction vessel and purified by chromatography after each coupling step. This method works well for short peptide sequences, but quickly becomes costly for longer peptide sequences in terms of loss of product, amounts of solvent used, and time. Current methods focus on using solid-phase peptide synthesis (SPPS) in which an insoluble polymer resin is used as a chemical substrate for building the amino acid chain (Figure 6).1 Amino acids located at the C-terminus of the target peptide are coupled to the resin via their carboxyl group, while being N-protected. The N-protecting group is then removed and the coupling step is repeated with the next amino acid in the chain in a C- to N-terminal direction. Typically, amino acids are coupled one at a time; however, in some cases, several amino acids are first coupled together using solution-phase chemistry to form a peptide fragment, and then the peptide fragments are coupled on-resin; this approach is called fragment condensation.

HO

O H N

Couple first amino acid

O

P R2

P

O

N -Deprotection

H2N

O

= resin solid support HO

R2

R1

Couple second amino acid

= protecting group

O H N

P N H

O O

R2

R1

N -Deprotection and cleavage from the resin

O H N

H2N

OH O

R2

Figure 6 The use of a resin solid support facilitates peptide synthesis. Amino acid building blocks are added via their C-terminus to the resin. Sequential coupling and N-deprotection steps build the target peptide. The peptide is released from the resin in a final chemical cleavage step.

Peptides and Glycopeptides

107

The advantage of SPPS is that by attaching the peptide to a solid resin support, the peptide can be purified after each coupling step by filtration rather than chromatography, and a large excess of reagents can be used to drive the reaction to completion. The filtration of the resin is accomplished by placing it in a reaction vessel in which a polyurethane or Teflon filter is used as a barrier to retain the resin, while allowing for the solvent to pass through. After each coupling and N-deprotection step, the resin can be washed of excess reagents, leaving only the peptide-linked resin in the reaction vessel. As a final step in the synthesis, the peptide is chemically cleaved from the resin and collected for purification (Figure 6). There are two conventional approaches to SPPS based on the choice of N-protecting group of the amino acid building blocks. The Fmoc-strategy relies on using 9-fluorenylmethoxycarbonyl (Fmoc)-protected amino acids, in which the N-Fmoc group can be deprotected under mildly basic conditions in the coupling cycles.26 Therefore, under the Fmoc-strategy, a polymer resin that has an acid-labile link to the peptide is used. A strong acid, such as trifluoroacetic acid, is used to cleave the peptide from the resin. By contrast, the Boc-strategy relies on the acid sensitivity of the tert-butyloxycarbonyl (Boc)-protecting group, which allows for using a resin that requires a very strong acid, such as hydrogen fluoride, to cleave the peptide–resin linkage.26 SPPS revolutionized the synthesis of peptides, and now, biotechnological approaches to peptide synthesis are having a similar effect on the field.

1.08.7.2

Recombinant Peptide Synthesis

Peptides can also be generated by biochemical methods using recombinant DNA techniques, genetic engineering, and the application of enzymes in chemical synthesis to control the regio- and stereospecificity of the peptide coupling reactions.1 The large-scale production of pharmacologically active peptides can be accomplished by the recombination and expression of genetic material in bacteria. Because of the low cost of production and scalability, many therapeutic peptides are obtained by the fermentation of recombinant cells. However, one drawback of this approach is that the posttranslational modification of peptides is poorly controlled, as it is not directly genetically controlled, and therefore many different glycoforms are produced. Thus, the synthesis of glycopeptides, which has become a growing area of research, remains reliant on synthetic chemical approaches.

1.08.8

Glycopeptide Synthesis

As the science of glycobiology has matured, so has a desire to understand the roles of specific glycopeptides and glycoproteins. Synthetic glycopeptides are required for several reasons. First, glycopeptides and glycoproteins are present in very minute quantities in vivo, making isolation of a given compound in sufficient quantities quite difficult.27 Second, as previously stated, each glycopeptide or glycoprotein can exist in many different glycoforms, which exacerbates the problem of isolation even more so than for peptides. For example, erythropoietin is a clinically useful stimulant of red blood cells (RBCs) for treatment of anemia, but recombinant expression in Chinese hamster ovary cells produces 13 different glycoforms; this complicates the isolation and characterization process.28 Thus, in a similar fashion for peptides, emphasis has shifted from isolation to the chemical synthesis of glycopeptides and glycoproteins; this can ensure the production of sufficient quantities of glycopeptides with well-defined glycan structures, in order to correlate a specific glycan with a physiological outcome.

1.08.8.1

Controlling Regio- and Stereoselectivity

The synthesis of O- and N-glycopeptides focuses on controlling the regio- and stereoselectivity of glycosidic linkages. In order to accomplish this task, current methods rely on chemically protecting carbohydrate and amino acid functional groups to form only the desired glycosidic linkage. As carbohydrates have many functional groups, oligosaccharide and glycopeptide synthesis remains very time consuming, as purification is required after each successive protection, deprotection, and coupling step. Examples of O-protecting groups used in glycopeptide synthesis include benzylether, silylether, and acetate ester groups (Figure 7).27 The glycosylation reaction follows a reaction pathway in which the protected building block, termed a glycosyl donor, has a suitable leaving group, which in the presence of a suitable activating group forms a reactive oxocarbenium ion intermediate (Figure 8(A)). Examples of glycosyl donors include trichloroacetamides, glycals, glycosyl halides, glycosylphosphates, and thioglycosides, which can be activated by Lewis acids such as boron trifluoride diethyletherate.27 Another common method of activation of a glycosyl donor is using N-iodosuccinimide with catalytic amounts of triflic acid, which creates a highly reactive donor. The choice of protecting group on the carbohydrate has a large influence on the stereoselectivity of the glycosylation reaction. Acetate protecting groups can participate in the glycosylation reaction by forming an acetoxonium ion intermediate, which favors the incoming amino acid nucleophile to form a b-type linkage (Figure 8(B)). By contrast, benzyl protecting groups have no such participating effect and therefore cause both a- and b-linkages to form (Figure 8(C)). The a-linkage would be expected to predominate because of stereoelectronic reasons.

1.08.8.2

Formation of the Glycosidic Linkage

The chemical synthesis of glycosylated building blocks relies on different approaches for forming N- versus O-glycosidic linkages.27 Methods for forming the asparagine to N-acetylglucosamine linkage include coupling a protected N-acetylglucosamine residue

108

Peptides and Glycopeptides

OR

OR O

R=

RO

OR

Benzyl

RO

Triisopropylsilyl

Si

O

Acetate Figure 7

Common carbohydrate protecting groups used in glycopeptide and oligosaccharide synthesis.

A

OP

OP O

PO

OP

Activator

OP

OP

P = protecting group

PO

PO

LG

OP O

O

OP

OP

OP

Oxocarbenium ion

LG = leaving group

B

OP

OP

OP

OP

OP

O

O

OP O

PO O

O

OP O

PO

PO O

OP

O

O

PO

R O

O O

O

R

O

H

Acetoxonium ion C

OP

OP

OP

Incoming nucleophile favors the formation of only the β-linkage OP

OP

O

OP O

O

PO

PO

PO O

R = any sugar

O

R O

O

O

R

H

No participating effect

Incoming nucleophile can form both α- and β-linkages

α-Linked product is expected to predominate

Figure 8 Formation of the glycosidic linkage type depends on several factors. (A) The addition of an activating group to a glycosyl donor causes the formation of a reactive oxocarbenium ion. (B) A 2-position acetate protecting group can participate in the glycosylation reaction by forming an acetoxonium ion, resulting in the exclusive formation of the b-kage. (C) Without a participating effect, both a- and b-linkages are formed, but the a-linkage predominates for stereoelectronic reasons.

with L-aspartic acid to form exclusively the b-amide linkage (Figure 9(A)). The formation of the serine/threonine O-linkage to a carbohydrate such as galactose can be achieved by reaction of L-serine or L-threonine with a glycosyl donor in the presence of a Lewis acid such as boron trifluoride diethyletherate (Figure 9(B)).

1.08.8.3

Strategies for Glycopeptide Synthesis

Typically, the synthesis of peptides and glycopeptides is carried out using solid-phase synthesis either by incorporation of the glycosylated building blocks as individual amino acids or by using fragment condensation. However, the strongly acidic conditions used to cleave the peptide from the resin can potentially hydrolyze O-glycosidic linkages present in oligosaccharide structures. The direct coupling of an oligosaccharide and fully formed peptides using a convergent-type approach has been successfully applied to the synthesis of large and complex glycopeptides.1 However, the efficiency of the key coupling step can be problematic. A third

Peptides and Glycopeptides

109

A OP

O

O

P

O

PO PO

NH2

+

OP

NHAc

H N

PO

O

HO

O

Coupling reagent PO

O

NHAc

NH

O P P

P

N H

O

β-N-GalNAc-Asn linkage B OP

OP OP

OP O

PO

BF3 O

O

OP

O

P

PO

O

HO PO

O PO

O

PO

OP

PO

O O P

NH P P

P = protecting group

NH

O

α-O-Gal-Ser linkage

Figure 9 Representative examples of the formation of N- and O-linked glycosylated amino acid building blocks. (A) Reaction of a protected N-acetylglucosamine residue with a protected L-aspartic acid residue using a coupling reagent will lead to the desired b-N-GalNAc-Asn building block. (B) Reaction of a protected glycosyl donor with boron trifluoride diethyletherate will form an oxocarbenium ion, which will react with a protected serine residue to form the target a-O-Gal-Ser building block.

strategy is to use a combination of enzymatic oligosaccharide synthesis together with a conventional peptide synthesis approach.27 In effect, glycosyltransferases or endoglycosidases are used to transfer mono- or oligosaccharides, respectively, onto monosaccharide-labeled polypeptides. This can occur in aqueous solution without the need for protecting groups on the sugars, which greatly reduces the cost, time required, and environmental impact of the synthesis. Glycosyltransferases typically are used to add sugar residues one at a time, whereas endoglycosidases have been used to transfer large oligosaccharides onto a GlcNAc residue attached to a polypeptide.27 Endoglycosidases vary in their specificity for glycan structures. For example, Endo-A, isolated from the bacterium Arthrobacter protophormiae, is specific for transferring high-mannose-type N-glycans, whereas Endo-M isolated from the fungus Mucor hiemalis can act on high-mannose, complex, and hybrid-type N-glycans. However, overall, the use of endoglycosidases suffers from low trans-glycosylation efficiency and the problem of product hydrolysis.

1.08.9

Peptides and Glycopeptides as Models of Proteins and Glycoproteins

The ability to synthesize chemically well-defined peptides and glycopeptides allows for them to be used as models of larger proteins and glycoproteins. Most proteins are relatively large biomolecules, but exert their function only through small regions of their structure. In addition, some aspects of protein structure require the study of only short regions of their sequence. These facts position peptides and glycopeptides, which have relatively manageable synthetic requirements, as valuable models of proteins and glycoproteins.

1.08.9.1

Mucin Glycoprotein Model Peptides

The mucin glycoproteins are characterized by large, serially O-a-N-acetylgalactosamine-linked oligosaccharides, and, as previously mentioned, are used primarily by the body to retain water at surfaces exposed to the environment for lubrication and protection.14 Coltart and coworkers used model glycopeptides to help understand the effects of these serially O-linked glycans on mucin conformation.30 The authors synthesized pentapeptides of the sequence Ser-Thr-Thr-Ala-Val with a-O-GalNAc-linked mono- and disaccharides at the serine and threonine residues (Figure 10). It was found by a combination of NMR experiments and computational calculations that the clustering of these O-linked glycans induced an extended conformation in the peptide. This was based on the large number of nOe contacts between the proximal a-O-GalNAc residue and the peptide backbone, a reduction in GalNAc amide proton chemical shift temperature dependencies, and 3JNH-Ha coupling constants inconsistent with extensive conformational averaging. Computational calculations indicated that the structural rigidity could be attributed to hydrogen-bonding interactions between the sugar and the peptide backbone. The authors postulated that the extended structure induced by a-O-GalNAc facilitates recognition events involving the glycan. This study also provided insight into the different influences from the inner and outer glycan residues and the native a- and abnormal b-linked sugars on peptide structure and stability (Figure 10). Interestingly, it was found that sugars attached beyond the first a-O-GalNAc residue had no influence on the structure of the peptide. The distinct nOe fingerprint did not change as

110

Peptides and Glycopeptides

R

R O

O

O H N

N H

O

O

O

H N

N H

O

O

OH

N H

O

R

OH

R=

OH

OH O

HO

HO AcHN

O

α-O -GalNAc

OH

OH O

OH

OH O HO

O HO

AcHN

O

Gal-β1,4-GalNAc-α-O -

OH

OH O

OH O

O HO

O AcHN

Gal-β1,4-GalNAc-β-O -

Figure 10 Model pentapeptides of the mucin glycoproteins. Different glycan structures were used to understand the effects of glycosylation on the conformation of the peptide. Each peptide had only one type of glycan attached. The natural a-linked glycans instilled an extended conformation in the peptide, while the unnatural b-linked glycan induced no such conformational preference.30

the number of sugar residues was increased from one to three. Also, while a-O-GalNAc was found to be critical for structural organization, the b-O-linked sugars did not seem to induce any rigidity in the peptide. Based on the uniformity of 1H and 15N NMR chemical shifts, temperature dependence of amide proton chemical shifts, and comparison of backbone and side-chain coupling constants, the b-O-linked model peptide resembled the nonglycosylated form with a high degree of flexibility. Therefore, the anomeric linkage has a large influence on the structural properties of the peptide. This seems to reflect a broader trend in glycobiology, where the nature of the glycosidic linkage influences not only the presentation of the carbohydrate moiety but also peptide backbone conformation.14

1.08.9.2

Antifreeze Glycoprotein Model Peptides

The antifreeze glycoproteins (AFGPs) are a type of O-linked glycoproteins found in the blood plasma of several species of deep-sea polar fish.31 They are used to protect the fish from the cellular damage that can arise from ice crystal formation in subzero water. The proteins are characterized by having a repeating Thr-Ala-Ala tripeptide sequence, where threonine is b-N-linked to a galactosyl-a-1,3-N-acetylgalactosylaminyl disaccharide unit. Czechura and coworkers have found that smaller glycopeptides can exhibit similar cryoprotective properties of the AFGPs.32 The exact nature of how the glycopeptides prevent ice formation remains unclear; however, the reasons behind the cryoprotective nature of the glycopeptides have been attributed to their ability to bind to the surface of initially formed ice crystals and prevent the cascade of larger ice crystal formation. The ice–glycopeptide interactions seem to rely on hydrogen-bonding interactions between the hydrophilic sugar hydroxyl groups and the ice crystal, and also the b-methyl group of threonine seems to cause important hydrophobic disruptions of water alignment.

1.08.9.3

Collagen Glycoprotein Model Peptides

Model peptides were also useful for understanding that O-glycosylation of threonine residues can stabilize the collagen triple helix. The cuticle collagen of a deep-sea worm, Riftia pachyptila, which lives near hydrothermal vents, has a b-O-galactosylated threonine residue in place of hydroxyproline in the Yaa position of the collagen (Gly-Xaa-Yaa) repeat.33 A study carried out by Bann and coworkers demonstrated that the carbohydrate was essential for the formation of the collagen triple helix.34 The model peptide Ac-(Gly-Pro-Thr(b-D-Gal))10-NH2 was found to have a thermal transition at 41  C as determined by thermal melting curves using CD, which is indicative of a triple helix to a single-strand transition (Figure 11). This was also confirmed by analytical centrifugation studies. By contrast, the model peptide Ac-(Gly-Pro-Thr)10-NH2 did not exhibit such a thermal transition, indicating that the triple helix does not form in the absence of Thr O-glycosylation. Therefore, it seems that the glycosylated threonine residue somehow stabilizes the triple helix in place of hydroxyproline. The authors proposed that in a similar manner to other O-glycosylated peptides, such as the mucins, the galactose residue instills an extended, rigid structure on the polypeptide strands. The galactose residue may also stabilize the triple helix through hydrogen bonding to the polypeptide backbone. This theory was supported by low exchange rates of backbone amide NH protons as measured by NMR experiments, which indicates that the sugar may shield the polypeptide backbone from the solvent. Bann suggested that in order to clarify the mechanism of stabilization incurred through glycosylation, further studies were required to determine if glycosylation affects the cis–trans isomerization of the neighboring proline residue and/or affects the conformation of the individual collagen strands.

1.08.9.4

HRGP Model Peptides

Toward a greater understanding of the effects of the glycans on HRGP conformational stability, in 2001, Shpak and coworkers used synthetic peptides to probe the effects of the short oligoarabinosides and larger arabinogalactan polysaccharides that are

Peptides and Glycopeptides

111

OH OH

HO

O HO O NH O

O N

NH2

O

NH N H

Tm = 41 °C

O

O

10

O N

HO O

NH2 N H

O

10

No transition observed

Figure 11 Model peptides of the collagen glycoproteins found in Riftia pachyptila. Glycosylation of threonine in the Gly-Pro-Thr repeat sequence displayed a thermal transition (at 41  C), which is characteristic of a triple helix to a single-strand conformational transition. No such transition was found for the unglycosylated model peptide. The authors concluded that glycosylation is critical for maintaining the native triple helical structure of R. pachyptila.34

characteristic of HRGPs, on PPII conformation.35 Instead of conventional chemical synthesis, these peptides were synthesized using the cellular machinery of model cells. Synthetic genes were delivered into cultured tobacco cells (Nicotiana tabacum) to produce peptides with sequences of (Ser-Pro1–4)n, which is characteristic of extensin-type HRGPs. These sequences were hydroxylated by endogenous prolyl-4-hydroxylase to form contiguous and noncontiguous hydroxyproline residues. It was found that the (Ser-Pro)n and (Ser-Pro3)n sequences led to noncontiguous hydroxylation and were glycosylated with arabinogalactan polysaccharides. Similarly (Ser-Pro2)n sequences also led to noncontiguous hydroxylation, but were glycosylated with both arabinogalactan polysaccharides and arabinose trisaccharides. Finally (Ser-Pro4)n sequences led to contiguous hydroxylation and were glycosylated predominantly with arabinose trisaccharides. The effects of the carbohydrates on the stability of the model peptides were measured using CD by comparing the intensity of the extrema of the glycosylated and deglycosylated peptides. It was found that the polysaccharides intermittently attached to (Ser-Hyp/ Pro)n (Ser-Hyp/Pro2)n, and (Ser-Hyp/Pro3)n sequences caused decreases in the intensity of the CD maxima. The authors correlated these decreases in intensity to destabilization of the polyproline conformation by intermittent O-D-galactose-linked polysaccharides. By contrast, the arabinose oligosaccharides attached to the sequence (Ser-Hyp4)n caused a marked increase in the CD maxima, which was correlated with a stabilizing effect on the polyproline conformation. The authors concluded that the polypeptide sequence determines the type of glycan attached, which in turn affects the stability of the polyproline conformation of the HRGP and, perhaps by extrapolation, the integrity of the plant cell wall. What remains to be investigated in well-defined model peptides is whether it is the first arabinose or galactose residue that is responsible for the change in polyproline conformation or the larger glycan structure. In all of these examples, model peptides and glycopeptides were instrumental in understanding the physical properties of larger proteins and glycoproteins. This provides more information on the function of these biomolecules. An advantage of understanding what takes place in nature is being able to apply this knowledge in the treatment of disease.

1.08.10 Application of Synthetic Peptides and Glycopeptides for the Treatment of Disease Access to synthetic peptides and glycopeptides has also allowed for their use in medicinal applications. Because of their important role in triggering the adaptive immune response, peptides and glycopeptides are being used for the creation of vaccines against many diseases, including malaria and cancer. Vaccines work by triggering our immune system into preparing antibodies not only against harmful viruses but also against bacteria and tumor cells. Vaccine development relies on our body‘s adaptive immune response, in which peptides actually play an important role.1 Peptides isolated through the destruction of larger proteins from normal and abnormal cells are collected and presented at the surface of cells through the major histocompatibility complex (MHC), which allows for the discrimination of destructive entities (viruses, bacteria, and tumor cells) from healthy cells. White blood cells then coordinate or carry out the destruction of the pathogen or infected/malfunctioning cells bearing those peptide antigens. In this way, foreign entities, such as bacteria or viruses, are tagged as foreign, allowing for the formation of antibodies specific for the antigenic peptide region. This process can be exploited by chemically synthesizing or isolating peptides that are specific to pathogenic cells.

1.08.10.1 A Peptide-based Malaria Vaccine Malaria is a disease which continues to cause millions of deaths every year, especially of children in developing countries.36 It is predominantly caused by the protozoa Plasmodium falciparum which resides in the gut of the Anopheles mosquito. The microbes are transferred into humans during the blood meal of an infected female Anopheles mosquito. After migrating to the liver, P. falciparum proceeds to infect human RBCs, which causes various symptoms including fever, nausea, anemia, and, in severe cases, coma and death. There are ongoing efforts to develop a peptide-based vaccine that would prevent the spread of malaria, which have focused on various points in the infection process.37 The merozite surface protein-1 (MSP-1) is one of the most abundant

112

Peptides and Glycopeptides

membrane proteins of P. falciparum, and is known to play a crucial role in the infection of RBCs. It is therefore considered to be a promising vaccine candidate. Thousands of peptides derived from MSP-1 have been synthesized by SPPS, with the aim of identifying those peptide ligands responsible for binding to RBCs. Espejo and coworkers identified a peptide sequence Glu-Val-Leu-Tyr-Leu-Lys-Pro-Leu-Ala-Gly-Val-Tyr-Arg-Ser-Leu-Lys-Lys-GlnLeu-Glu derived from MSP-1, which is conserved in all parasite strains.38 It binds to RBCs with an affinity constant of 180 nanomolar (nM), where five residues (highlighted in bold) were found to be critical for binding. Interestingly, the peptide itself exhibits no capacity to induce an immune response. Espejo and coworkers proceeded to screen derivatives of the lead peptide in the hope of inducing protective antibodies as determined by the immunization of Aotus monkeys.39 Immunized and control monkeys were then challenged with active parasites to measure the effectiveness of immunization. It was found that replacement of several of the key binding amino acids produced peptides that exhibited strong antibody production and full protection in 16%–50% of the immunized monkeys. Some modifications produced peptides that remained nonimmunogenic. Differences in immunological behavior were speculated to reflect different bioactive structural conformations of the peptides. A combination of 1H-NMR, CD, and molecular modeling experiments was used to compare the solution conformations of protective and nonprotective peptides. Both types of peptides exhibited significant a-helical structure; however, they varied in the location and extent of this type of secondary structure. The more strongly immunogenic peptides had more flexibility in their N-terminal regions, which may allow for a better-induced fit into the immune system molecules and, as a result, stronger protective immunity. Although these peptides exhibit the ability to induce antibodies against the causative agent of malaria, vaccine development remains problematic as P. falciparum is very effective at avoiding detection by the immune system through various mechanisms. Therefore, more work is required in the development of peptide-based malaria vaccines.

1.08.10.2 Glycopeptide-based Cancer Vaccines Another emerging area of research involves using glycopeptides for the development of cancer vaccines.40 Cancerous cells exhibit different glycan structures than healthy cells, which include the over- and underexpression of naturally occurring glycans, and the expression of glycans normally restricted to embryonic tissues.41 These changes have been attributed to the altered expression of glycosyltransferases in the Golgi apparatus. These abnormal glycans have been targeted as antigens for the creation of cancer vaccines. As previously discussed, the isolation of tumor-associated antigens from biological sources suffers from the production of glycopeptides with heterogeneous glycans, which leads to nonspecific immune responses. Therefore, most focus lies on the generation of synthetic glycopeptides with well-defined sequences as vaccines to combat cancer and other diseases. However, these tumor-associated antigens produce relatively weak immune responses.40 Therefore, current approaches combine the antigens with immunostimulants, such as keyhole limpet hemocyanin (KLH) glycoprotein, to boost antibody and T-cell production. KLH is a widely employed immunostimulant for the production of antibodies in biotechnology and therapeutic research and applications and is effective because of its high molecular weight and the distinctive epitopes it presents.

1.08.10.2.1

A Glycopeptide-based Vaccine With Multiple Tumor-Associated Carbohydrate Antigens

Abnormal mucin glycosylation is characteristic of many human carcinomas, and has therefore attracted significant attention.41 Furthermore, the mucin-like glycopeptides have been shown to cause a strong antibody response in cancer cell lines. Ragupathi and coworkers developed a glycopeptide-based vaccine with multiple mucin-type carbohydrate antigens attributed to colon and breast cancer tumor cells (Figure 12).42 The carbohydrate antigens used are all well-known tumor-associated carbohydrate antigens (TACAs): TN, TF, STN, LeY, and Globo-H. The five different carbohydrate antigens were attached to a pentapeptide containing repeating 6-hydroxy-L-norleucine residues. The glycopeptide conjugate was found to be a more potent immunostimulant than the corresponding peptides with only a single TACA at producing immunoglobulin G and M (IgG and IgM) antibodies as determined by enzyme-linked immunosorbent assay (ELISA). The antibodies generated from the multivalent glycopeptide antigen were found to be active against human breast and colon cancer cell lines, MCF7 and LSC, respectively. Therefore, this study indicated that glycopeptides could be effective initiators of antibody production against human cancer cell lines toward the development of cancer vaccines. It also demonstrated the power of a multivalent approach in the presentation of carbohydrate antigens, which has become a well-established approach in carbohydrate chemistry.

1.08.10.2.2

A Glycopeptide Vaccine Based on the MUC1 Glycoprotein

The MUC1 mucin glycoprotein is known to exhibit abnormal glycan structures within a 20-amino-acid tandem repeat domain that is associated with tumor cells.40 The glycan structures are characterized as being very small with terminal sialic acid carbohydrate residues. An MUC1 glycopeptide analog developed by Dziadek and coworkers bearing a sialic acid-N-acetylgalactosamine glycan (STN antigen) on a decapeptide sequence Gly-Val-Thr*-Ser-Ala-Pro-Asp-Thr-Arg-Pro-Ala-Pro derived from MUC1 (Figure 13).43 It was attached to an immunogenic peptide sequence derived from the egg-white protein ovalbumin and was found to elicit a strong IgG antibody response in transgenic mice that express the MUC1 glycoprotein. Interestingly, the antibodies did not exhibit high specificity for the unglycosylated peptide sequence or the STN antigen alone. It was speculated that this indicates that the carbohydrate seems to induce a conformational change in the MUC1 glycopeptide, which becomes an important structural element required for an immunogenic response. Therefore, as previously described, glycosylation has an impact on peptide conformation, which, in this case, affects the interaction of the glycopeptide with components of the immune system. These examples serve to show that synthetic glycopeptides are becoming increasingly important for understanding and treating diseases such as cancer.

113

Peptides and Glycopeptides

STN antigen OH

OH

OH

OH

OH

OH

OH CO2H

TN antigen

O

O

HO

HO

O

O

O

HO

AcHN

HO

HO

O

O

AcHN

OH

O

HO

O

Globo-H antigen

OH

HO

O OH

OH

HO

O

HO

O

HO

OH O HO

AcHN

O

AcHN

O

O

HO

O OH

N H

OH

HO

O

OH

OH

O O OH

O O

O

AcHN

HO

N H

O OH

OH

O

O

H N

N H

O

O OH O

O

H N

O

HO

R O

OH O

O

H N

O

AcHN

O HO

OH

OH HO

Ley antigen

OH

HO O

R=

H N

N S O

TF antigen

O

KLH

O

Figure 12 Glycopeptide-based cancer vaccine developed by Danishefsky and coworkers, which displays multiple carbohydrate-associated tumor antigens. The glycans were attached to 6-hydroxy-L-norleucine residues, and the entire glycopeptide was tethered to keyhole limpet hemocyanin (KLH). The vaccine effectively produced IgG and IgM antibodies that were active against breast and colon cancer cell lines MCF7 and LSC, respectively.42

HO HO AcHN

OH

CO2H O OH

STN antigen

O OH O

HO AcHN

O

Gly-Val-Thr-Ser-Ala-Pro-Asp-Thr-Arg-Pro-Ala-Pro------Ile-Ser-Gln-Ala-Val-Ala-Ala-His-Ala-Glu-Ile-Asn-Glu-Ala-Gly-Arg

MUC1- derived glycopeptide sequence

Immunogenic peptide sequence derived from ovalbumin

Figure 13 Glycopeptide-based cancer vaccine developed by Kunz and coworkers, which was derived from the MUC1 glycoprotein. The STN antigen was attached to a peptide sequence characteristic of MUC1. The glycopeptide was tethered to an immunogenic peptide sequence derived from ovalbumin. The vaccine was found to elicit a strong IgG antibody response in mice, which represents a promising candidate for vaccine development against cancer.43

1.08.11 Summary Peptides and glycopeptides have come to be recognized for their roles in biological processes from the cellular to the organism level of function. Initially, there was much skepticism about the role of peptides in the nervous and endocrine system functions because of their relatively simple structure. However, it has become evident that peptides have been selected in most multicellular organisms

114

Peptides and Glycopeptides

for the transmission of information, whether it is between neurons in the form of neuropeptides, or within and between organs in the form of peptide hormones. The peptides angiotensin II and bradykinin are just two of many examples of peptide hormones that control vital biological functions, having opposite effects on physiological processes. Moreover, they exemplify that many peptides adopt a preferred bioactive conformation when bound to their target receptor. Similarly, substance P and NPY are both critical neuropeptides required for our survival. Synthetic analogs of these peptides may be of therapeutic use for understanding and treating various disorders and diseases related to pain perception and obesity. Peptides also play a fundamental, and sometimes overlooked, role in the innate immunity of most species on the Earth as antibacterial agents. This suggests that in the wake of the emergence of antibiotic-resistant bacteria, antibacterial peptides present a basis for developing new antibiotics. Although more potent antibiotics exist, there are definite advantages to these peptides: the ability to kill target cells quickly, unusually broad activity spectra, activity against some of the more serious antibiotic-resistant bacteria, and a unique mode of action that reduces the chance of resistance development. The effects of glycosylation on peptide structure and function continue to be understood, and, within the field of glycobiology, remains an exciting area of discovery. It has been found that glycosylation affects peptide trafficking and biological recognition events, and can also change the conformation, stability, and hydration of peptides. Advances in the synthesis of peptides and glycopeptides have been provided for understanding the function of specific peptides and glycopeptides, such as antibacterial peptides, peptide hormones and neuropeptides, and also access to valuable models for understanding the function of larger proteins and glycoproteins, such as plant and animal structural glycoproteins and AFGPs. Synthetic peptides and glycopeptides are also being used in therapeutic applications for understanding and treating disease. The development of peptide- and glycopeptide-based vaccines relies on our adaptive immune response and may lead to treatments for diseases such as malaria and cancer.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32.

Sewald, N.; Jakube, H.-D. Peptides: Chemistry and Biology, Wiley-VCH: Weinheim, 2002. Varki, A., Cummings, R. D., Esko, J. D.; et al., Eds.; Essentials of Glycobiology, 2nd ed.; Laboratory Press: Cold Spring Harborny: Cold Spring Harbo NY, 2009. Kastin, A. J., Ed.; Handbook of Biologically Active Peptides, Academic Press: London, 2006. Izumi, Y.; Iwao, H. Angiotensin II and its Related Peptides. In Handbook of Biologically Active Peptides; Kastin, A. J., Ed., Academic Press: London, 2006; pp 1169–1174. Campbell, D. J. Bradykinin and its Related Peptides. In Handbook of Biologically Active Peptides; Kastin, A. J., Ed., Academic Press: London, 2006; pp 1175–1179. Garcia, K. C.; Ronco, P. M.; Verroust, P. J.; et al. Three-dimensional Structure of an Angiotensin II–fab Complex at 3 Å: Hormone Recognition by an Anti-idiotypic Antibody. Science 1992, 257, 502–507. Kyle, D. J.; Blake, P. R.; Smithwick, D.; et al. NMR and Computational Evidence that High-affinity Bradykinin Receptor Antagonists Adopt C-terminal b-turns. J. Med. Chem. 1993, 36, 1450–1460. Caldwell, H. K.; Young, W. S., III Oxytocin and Vasopressin: Genetics and Behavioral Implications. In Handbook of Neurochemistry and Molecular Neurobiology: Neuroactive Proteins and Peptides; Lajtha, A., Lim, R., Eds., 3rd ed.; Springer: Berlin, 2006; pp 573–607. Wittelsberger, A.; Patiny, L.; Slaninova, J.; et al. Introduction of a Cis-prolyl Mimic in Position 7 of the Peptide Hormone Oxytocin Does Not Result in Antagonistic Activity. J. Med. Chem. 2005, 48, 6553–6562. Strand, F. L., Ed.; Neuropeptides: Regulators of Physiological Processes, MIT Press: Cambridge, 1999. Devine, D. A.; Hancock, R. E. W. In Mammalian Host Defense Peptides, Cambridge University Press: New York, NY, 2004. Otvos, L., Jr. Antibacterial Peptides Isolated from Insects. J. Pept. Sci. 2000, 6, 497–511. Otvos, L., Jr. The Proline-rich Antibacterial Peptide Family. Cell. Mol. Life Sci. 2002, 59, 1138–1150. Taylor, M. E.; Drickamer, K. Introduction to Glycobiology, Oxford University Press: New York, NY, 2003. Gobbo, M.; Biondi, L.; Filira, F.; et al. Antimicrobial Peptides: Synthesis and Antibacterial Activity of Linear and Cyclic Drosocin and Apidaecin 1b Analogues. J. Med. Chem. 2002, 45, 4494–4504. Nimni, M. E., Ed.; Collagen, CRC Press: Boca Raton, FL, 1988. Notbohm, H.; Nokelainen, M.; Myllyharju, J.; et al. Recombinant Human Type II Collagens with Low and High Levels of Hydroxylysine and its Glycosylated Forms Show Marked Differences in Fibrillogenesis in Vitro. J. Biol. Chem. 1999, 274, 8988–8992. Wang, C.; Kovanen, V.; Raudasoja, P.; et al. The Glycosyltransferases Activities of Lysyl Hydroxylas 3 (LH3) in the Extracellular Space Are Important for Cell Growth and Viability. J. Cell Mol. Med. 2009, 13, 508–521. Myllylä, R.; Wang, C.; Heikkinen, J.; et al. Expanding the Lysyl Hydroxylase Toolbox: New Insights into the Localization and Activities of Lysyl Hydroxylase 3 (LH3). J. Cell. Physiol. 2007, 212, 323–329. Lamport, D. T. A. Structure, Biosynthesis and Significance of Cell Wall Glycoproteins. Recent Adv. Phytochem. 1977, 11, 79–115. Mauger, A. B. Naturally Occurring Proline Analogues. J. Nat. Prod. 1996, 59, 1205–1211. Lamport, D. T. A. Structure and Function of Plant Glycoproteins. In The Biochemistry of Plants, Vol. 3, Miflin, B. J., Ed.; Academic Press: New York, NY, 1980; pp 501–540. Esquerré-Tugayé, M. T.; Maxau, D. Effect of a Fungal Disease on Extensin, the Plant Cell Wall Glycoprotein. J. Exp. Bot. 1974, 25, 509–513. Ferris, P. J.; Woessner, J. P.; Waffenschmidt, S.; et al. Glycosylated Polyproline II Rods with Kinks as a Structural Motif in Plant Hydroxyproline-rich Glycoproteins. Biochemistry 2001, 40, 2978–2987. Global Industry Analysts. Artificial Sweeteners: Global Strategic Business Report, Research and Markets: Dublin, 2008. Green, T. W.; Wuts, P. G. M. Protective Groups in Organic Synthesis, Wiley-Interscience: New York, NY, 1999. Gamblin, D. P.; Scanlan, E. M.; Davis, B. G. Glycoprotein Synthesis: An Update. Chem. Rev. 2009, 109, 131–163. Delorme, E.; Lorenzini, T.; Griffin, J.; et al. Role of Glycosylation on the Secretion and Biological Activity of Erythropoietin. Biochemistry 1992, 31, 9871–9876. Wang, L.-X. Chemoenzymatic Synthesis of Glycopeptides and Glycoproteins through Endoglycosidase-catalyzed Transglycosylation. Carbohydr. Res. 2008, 343, 1509–1522. Coltart, D. M.; Royyuru, A. K.; Williams, L. J.; et al. Principles of mucin architecture: Structural studies on synthetic glycopeptides bearing clustered mono-, di-, tri-, and hexasaccharide glycodomains. J. Am. Chem. Soc. 2002, 124, 9833–9840. Knight, C. A.; Wen, D.; Laursen, R. A. Nonequilibrium Antifreeze Peptides and the Recrystallization of Ice. Cryobiology 1995, 32, 23–24. Czechura, P.; Tam, R. Y.; Dimitrijevic, E.; et al. The Importance of Hydration for Inhibiting Ice Recrystallization with C-linked Antifreeze Glycoproteins. J. Am. Chem. Soc. 2008, 130, 2928–2929.

Peptides and Glycopeptides

115

33. Gaill, F.; Herbage, D.; Lepescheux, L. A Discrete Helicoid of Collagenous Fibrils: The Cuticle of Deep-sea Hydrothermal Vent Worms. Matrix 1991, 11, 197–205. 34. Bann, J. G.; Peyton, D. H.; Bächinger, H. P. Sweet Is Stable: Glycosylation Stabilizes Collagen. FEBS Lett. 2000, 473, 237–240. 35. Shpak, E.; Barbar, E.; Leykam, J. F.; Kieliszewski, M. J. Contiguous Hydroxyproline Residues Direct Hydroxyproline Arabinosylation in Nicotiana Tabacum. J. Biol. Chem. 2001, 276, 11272–11278. 36. Baird, J. K. Effectiveness of Antimalarial Drugs. N. Engl. J. Med. 2005, 352, 1565–1577. 37. Lozano, J. M.; Bermudez, A.; Patarroyo, E. Peptide Vaccines for Malaria. In Handbook of Biologically Active Peptides; Kastin, A. J., Ed., Academic Press: London, 2006; pp 515–526. 38. Urquiza, M.; Rodriguez, L. E.; Suarez, J. E.; et al. Identification of Plasmodium Falciparum MSP-1 Peptides Able to Bind to Human Red Blood Cells. Parasite Immunol. 1996, 18515–18526. 39. Espejo, F.; Cubillos, M.; Salazar, L. M.; et al. Structure, Immunogenicity, and Protectivity Relationship for the 1585 Malarial Peptide and its Substitution Analogues. Angew. Chem. Int. Ed. 2001, 40, 4654–4657. 40. Liakatos, A.; Kunz, H. Synthetic Glycopeptides for the Development of Cancer Vaccines. Curr. Opin. Mol. Therapeut. 2007, 9, 35–44. 41. Brooks, S. A.; Carter, T. M.; Royle, E.; et al. Altered Glycosylation of Proteins in Cancer: What Is the Potential for New Anti-tumor Strategies? Anti Canc. Agents Med. Chem. 2008, 8, 2–21. 42. Ragupathi, G.; Koide, F.; Livingston, P. O.; et al. Preparation and Evaluation of Unimolecular Pentavalent and Hexavalent Antigenic Constructs Targeting Prostate and Breast Cancer: A Synthetic Route to Anticancer Vaccine Candidates. J. Am. Chem. Soc. 2006, 128, 2715–2725. 43. Dziadek, S.; Hobel, A.; Schmitt, E.; Kunz, H. A Fully Synthetic Vaccine Consisting of a Tumor-associated Glycopeptide Antigen and a T-cell Epitope for the Induction of a Highly Specific Humoral Immune Response. Angew. Chem. Int. Ed. 2005, 44, 7630–7635.

1.09

Protein Structural Analysis

BL Mark, SA McKenna, and M Khajehpour, University of Manitoba, Winnipeg, MB, Canada © 2011 Elsevier B.V. All rights reserved. This is an update of B.L. Mark, S.A. McKenna, M. Khajehpour, 1.11 - Protein Structural Analysis, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 139-153.

1.09.1 1.09.2 1.09.2.1 1.09.2.2 1.09.2.3 1.09.2.4 1.09.2.5 1.09.2.6 1.09.3 1.09.3.1 1.09.3.2 1.09.3.3 1.09.3.4 1.09.4 1.09.4.1 1.09.4.2 1.09.4.3 1.09.4.4 1.09.4.5 1.09.4.6 1.09.5 References

Introduction Protein X-ray Crystallography Protein Requirements for X-ray Crystallography Protein Crystallization Scattering of X-rays by a Crystal Structure Determination Model Building and Refinement Applications of Protein Crystallography to Biology NMR Spectroscopy Principles of NMR Spectroscopy Sample Requirements Structure Determination by NMR Beyond High-Resolution Structure Structure Analysis Using Intrinsic Protein Fluorescence Overview Quantum Mechanical Description Instrumentation Steady-State Fluorescence Spectroscopy Tryptophan Fluorescence as Probe of Protein Structure Tryptophan Fluorescence as Probe of Ligand Binding Conclusions

117 117 117 117 119 120 121 121 122 122 123 125 125 126 126 127 127 128 128 129 130 130

Glossary Chemical shift perturbation experiments Nuclear magnetic resonance (NMR) technique that exploits the extreme sensitivity of the chemical shift phenomenon to its surrounding environment as a probe of ligand binding. Dipole moment Every polar molecule can be approximately represented by an electric dipole (i.e., a partial positive and a negative charge separated by a distance). The dipole moment is the product of this partial charge by the distance; the larger the dipole moment, the more polar the molecule. Electronic excited state and electronic ground state The electronic ground state of a given molecule is the state in which that molecule has minimum electronic energy; any electronic state that has more energy than the ground state is called an electronic excited state. Extinction coefficient A molecular property of matter defined by Beer’s law; this property indicates how strongly a substance at 1 M concentration absorbs light at a given wavelength. Nuclear magnetic resonance (NMR) Spectroscopic technique in which the analysis of a protein sample in a strong magnetic field is used to obtain structural and dynamic information. Protein X-ray crystallography A method of determining the three-dimensional arrangement of atoms in a crystal that is comprised of a biological macromolecule such as a protein or an enzyme. Structural biology A field of science dedicated to analyzing the structural basis for how biological macromolecules carry out biochemical functions. Structure–activity relationships by NMR (SAR-by-NMR) Technique for screening and improving the affinity of pharmaceuticals for proteins through chemical linkage of nearby weak-binding ligands to develop a high-affinity analog. Synchrotron A particle accelerator that can produce brilliant X-ray beams of tunable wavelength for use of protein X-ray crystallography research.

116

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00011-2

Protein Structural Analysis

1.09.1

117

Introduction

The biochemical function of a protein is based on its three-dimensional (3D) molecular structure. Thus, to truly understand how a protein works, knowledge of its molecular structure must be obtained. We provide an introductory overview of the sample requirements and technical procedures necessary to determine and analyze protein structure by X-ray crystallography and nuclear magnetic resonance (NMR), which are the preeminent methods in use today to examine the 3D structure of biological macromolecules. Although these methods can provide unparalleled insight into the structural basis of protein function, they can be enhanced greatly through the use of complementary spectroscopic techniques that probe the thermodynamic properties of proteins, and their interaction with natural or man-made ligands. Fluorescence spectroscopy is particularly useful in this respect, as it exploits the intrinsic fluorescence proteins to rapidly provide insight into protein folding and stability and the interaction of proteins with ligands. Together, these techniques provide a powerful tool set with which to analyze and understand protein structure.

1.09.2

Protein X-ray Crystallography

Protein X-ray crystallography and protein NMR remain unsurpassed as techniques for analyzing protein structure, because they can be used to determine experimentally the 3D structures of biological macromolecules at or near atomic resolution. They are complementary techniques, each with their own strengths. Whereas NMR can provide information about the structural dynamics of a protein in solution, which protein X-ray crystallography does not, current NMR technology tends to limit successful 3D structure determinations to proteins 5–10 mg is usually necessary to commence trials), and the system can also be used to produce recombinant selenomethionyl protein for use in multiwavelength anomalous dispersion (MAD) phasing techniques (Section 1.09.3.4).4,12 Expression hosts such as yeast, insect cell systems, and mammalian cell culture can also be used to produce recombinant protein; however, they are typically reserved for proteins that require posttranslational modifications for proper expression. Alternatively, in cases where the expression of a large eukaryotic (or prokaryotic) protein fails in E. coli, individual domains of the protein can often be expressed with success. Eukaryotic protein expression in E. coli can be enhanced with the use of helper plasmids encoding tRNAs that are rare to E. coli, and the induction of gene expression at low temperature (16–28
C) can promote proper protein folding and reduce inclusion body formation. The simplicity of the E. coli system also facilitates the rapid production of multiple variants of a protein or protein domain. It is sometimes necessary to produce amino- and carboxy-terminal truncation variants before a well-folded and structurally ordered variant is found that will crystallize. Limited proteolysis experiments in combination with mass spectrometry can be used to guide terminal truncation experiments and to identify mobile loop regions that can be deleted during the process of designing the optimal expression construct. Affinity purification tags are commonly used to simplify recombinant protein purification from E. coli and other expression hosts; however, care must be taken to remove a tag before crystallization screening because it may impede crystallization. Following affinity purification, ion-exchange, gel filtration, and additional chromatographic methods are often necessary to purify the protein to a state of homogeneity that is sufficient to allow crystallization to occur. Unless it can be demonstrated that freezing does not damage the purified protein (resulting in unwanted heterogeneity), samples are usually filter sterilized and stored at 4
C.

1.09.2.2

Protein Crystallization

A number of protein crystallization techniques exist as described elsewhere.1 The most common of these is the vapor diffusion method of crystallization (Fig. 2A). Classically, a 1–4-mL drop of a highly pure, homogeneous protein sample (typically at a concentration of 2–10 mg mL1 and stabilized in a buffer of appropriate pH and salt) is placed in the center of a microscope cover slip. To the drop is added an equal volume of mother liquor, an aqueous solution containing a protein-precipitating agent such as ammonium sulfate or polyethylene glycol and buffered at a constant pH. The mother liquor may also include additional

118

Protein Structural Analysis

Gene of interest

Redesign expression construct to remove regions of disorder

Determine crystallographic space group, collect complete data set and asses data quality

Construct recombinant protein expression plasmid

Crystal structure of a homologous protein available?

Yes

Express and purify soluble protein (~5 − 10 mg) Screen crystallization space (commercial screens or inhouse screens)

Obtain phases via molecular replacement (MR) search

Continue screen

No

Crystallization condition(s) found?

No

Yes Optimize growth of single crystals with in-house reagents and test for X-ray diffraction

Correct solution with interpretable electron density?

Yes Acceptable diffraction? (d min< 3 Å)

No

Yes

No

Purify and crystallize selenomethionine substituted protein, or prepare isomorphous heavy atom derivative crystals Obtain phases and electron density maps via multi- or singlewavelength anomalous dispersion (MAD/SAD) measurements or by multiple isomorphous replacement (MIR)

Build and refine 3D model and investigate structural basis of biological function Deposit in Protein Data Bank

Figure 1

General steps involved in determining the crystallographic structure of a protein.

A

Hanging drop

Sitting drop Cover slip (sealed with grease) or clear tape Drop of protein/ mother liquor (e.g., 1µl of each) Reservoir of mother liquor (e.g., 100–500 µl)

B

Increasing pH

Increasing [precipitant]

Figure 2 (A) Experimental setup for the vapor diffusion method of protein crystallization. (B) Two-dimensional search to optimize pH and precipitating agent concentration in a 24-well format. An optimal condition is shown in black with semioptimal conditions in gray.

components such as salts, metal ions, and other additives that promote crystallization. The precipitating agent and protein concentration are just low enough so that protein does not immediately precipitate out of solution as an amorphous aggregate when first mixed with the mother liquor. Lack of precipitation is achieved in part by the twofold dilution of both the protein solution and mother liquor that occurs upon mixing them in equal parts. The cover slip is inverted and placed onto a reservoir containing 500–1000 mL of undiluted mother liquor. Lining the lip of the reservoir with vacuum grease onto which the cover slip is placed seals the system. As the mother liquor in the drop is less concentrated than it is in the reservoir, water will diffuse from the drop to the excess undiluted mother liquor in the reservoir. This equilibration slowly concentrates both the protein and precipitant in the drop, promoting nucleation and the growth of protein crystals. Protein crystallization is largely an empirical process, requiring numerous conditions to be screened before one is found that promotes crystallization of the protein of interest. The time spent searching for crystallization conditions can be reduced greatly through the use of commercially available crystallization screens. Available in 96-well format, they consist of mother liquor

Protein Structural Analysis

119

solutions that sample precipitating agents, salts, buffers (pH), and additives that commonly give rise to protein crystals. They can be rapidly screened using the vapor diffusion method to find initial crystallization conditions. Once one or more conditions are identified, they are recreated using in-house reagents and the parameters of the condition (precipitant concentration, pH, additives, etc.) can be optimized in 24-well format to grow single crystals of sufficient quality and size for X-ray diffraction analysis (Fig. 2B). Protein crystals usually grow to a diameter of less than 0.5 mm, often much shorter, and must be manipulated under a microscope. As water is integral to protein structure, and protein crystals are porous (typically 50% solvent), they are also very fragile and must not be allowed to dry. To avoid desiccation, and to maximize crystal integrity during exposure to ionizing X-rays, single crystals can be transferred into mother liquor containing a cryoprotectant such as glycerol, then retrieved from this solution using a small nylon loop and quickly flash cooled in liquid nitrogen.1 Rapid cooling in the presence of cryoprotectant stops the formation of ice crystals while leaving the protein crystal intact. Using protective tongs that have been precooled in liquid nitrogen, the loop containing the crystal is mounted onto an adjustable spindle (goniometer), where it is held at cryogenic temperature via a stream of cooled nitrogen gas. The crystal is then aligned to the X-ray beam for X-ray diffraction analysis (Fig. 3A).

1.09.2.3

Scattering of X-rays by a Crystal

A number of books detailing the fundamentals of protein X-ray crystallography are available,4,12 including an exceptional introductory book written by Rhodes.11 Briefly, crystals are a 3D, periodic array of molecules, or X-ray scattering elements, held together by noncovalent interactions. The smallest repeating unit within a crystal is defined by the unit cell, a parallelepiped with edges a, b, c related by angles a, b, g (Fig. 3B). When copies of the unit cell contents are stacked together in three dimensions,

A

Lowresolution reflection (largest interplanar distance (dmax))

Diffracted X-rays

Direct X-ray beam

X-ray detector plate

Direct X-ray beam stop Crystal held in a nylon loop

High-resolution reflection (shortest interplanar distance (dmin))

X-ray source C

B

a

c b Figure 3 (A) Diagram of the experimental apparatus used to collect X-ray diffraction data from a protein crystal. (B) An empty unit cell showing the hkl family of planes 222, where each unit cell axis is intersected twice by planes of the family. (C) A typical X-ray diffraction image showing numerous reflections, each of which can be indexed using the Miller indices hkl.

120

Protein Structural Analysis

it completely describes the entire crystal. The choice of unit cell edge lengths and angles depends on the symmetry relationships that are adopted by the molecules, or asymmetric units, in the crystal. An asymmetric unit, which can be a single protein, a portion of a protein, or multiple proteins, is the smallest part of a crystal structure to which crystallographic symmetry operators (rotations and translations) can be applied to generate the contents of one complete unit cell. The simplest unit cell contains only one asymmetric unit. However, more often asymmetric units crystallize in a symmetrical fashion (related by rotations and translations) to generate a unit cell with internal symmetry. The space group of a crystal describes the symmetry of its unit cell. There are only 65 possible space groups that chiral molecules, such as proteins, can adopt. X-rays are used to determine the actual arrangement of atoms in the unit cell because the wavelength (l) of this radiation is sufficiently short to resolve interatomic distances. Unfortunately, X-ray refraction from known materials is very weak and so a lens cannot be made to recombine scattered X-rays into a visible image. X-rays scattered from crystals are instead recombined using Fourier mathematics to visualize the electron density distribution within the crystal as a 3D contour map, and into this map is built and refined a 3D molecular model of the asymmetric unit. X-ray diffraction from crystals can be interpreted as the reflection of X-rays from families of equally spaced, parallel planes within the crystal that are populated by numerous atoms (Fig. 3B).4,12 Many thousands of such families of planes can be imagined passing though the crystal in different directions, with each family of planes separated by some distance d (measured in angstroms (Å)). An X-ray reflected from an individual family of planes is captured as a discreet spot on a 2D X-ray imaging plate and its intensity measured (Fig. 3A). The reflection is identified using the Miller indices h, k, l, where h, k, and l specify how many times the planes of the family intersect the a, b, and c axes of the unit cell, respectively (Fig. 3B). Due to the geometrical constraints that are required for a family of planes to produce an X-ray reflection, the crystal must be rotated about an axis in order to measure as many reflections as possible. Rotation of the crystal correctly positions families of planes within incident X-ray beam to satisfy the geometrical condition necessary for them to reflect X-rays and, thus, allows for numerous reflections to be acquired. Reflections are captured in batches on separate images as the crystal is rotated (Fig. 3C). Intense X-ray reflections arise from families of planes that are populated by many atoms, whereas weak reflections are from those that are populated by fewer atoms. Reflection intensities thus provide some, but not all, of the information about the distribution of atoms with the unit cell (see Section 1.09.2.4). Families of planes separated by short interplanar distances will reflect X-rays at a higher angles with respect to the incident X-ray beam than a families of planes that are separated by a larger interplanar distances. Thus, low-angle reflections (captured close to the center of the image plate) provide low-resolution information, such as the gross position of proteins within the unit cell, whereas high-angle reflections (captured on the outer edges of the image plate) provide high-resolution information about the positions of atoms within the unit cell (Fig. 3).11 The shortest interplanar distance (dmin) that can be measured, or resolved, is reported as the maximum resolution to which the crystal diffracts. The average dmin for protein crystal structures that have been determined is between 1.5 and 2 Å, which is very near atomic resolution (https://www.rcsb.org/). It should be noted that the collected reflections provide information about the contents of an average unit cell, in other words, a single cell that represents all unit cells of the crystal. The dimensions and space group symmetry of this representative unit cell can be determined directly from the X-ray diffraction patterns that are recorded.

1.09.2.4

Structure Determination

Each X-ray reflection is a wave, with an amplitude and phase, and is the sum of the X-ray scatter from all atoms in the unit cell. The amplitude of a reflection, jFhklj, can be determined directly from its measured intensity and, together with an associated phase ahkl, is known as a structure factor (Fhkl). Structure factors, in turn, are Fourier terms that can be used to compute a Fourier series that represents the unit cell contents as a 3D electron density map into which a molecular model of the asymmetric unit can be built. The problem, however, is that only reflection intensities are directly measured; the phase information for each reflection is lost and must be determined through indirect means. This is referred to as the phase problem,13 and it can pose a significant challenge to the successful determination of a protein structure. Three commonly used methods to solve the phase problem are multiple isomorphous replacement (MIR), MAD phasing, and molecular replacement (MR).13 MIR phasing involves soaking a crystal with a heavy atom compound to introduce a small number of heavy atoms into the unit cell without altering the unit cell dimensions; that is, the soaked crystal must remain isomorphous to the native crystal. The bound heavy atoms will scatter X-rays with greater intensity than those of the protein; thus, reflection intensities collected from a heavy atom derivative crystal will be altered relative to the same reflections collected from the native crystal. The differences in intensity measured for each reflection hkl can be used to locate the positions of a small number of heavy atoms in the unit cell in the absence of phase information. The determined heavy atom substructure can then be used to the estimate phases for the structure factors of the native protein crystal. Due to ambiguity in the phase estimates that are obtained from a single isomorphous heavy atom derivative, multiple derivatives are usually required to provide phases that are accurate enough to generate an interpretable electron density map. A popular alternative to MIR phasing is MAD phasing, which exploits the anomalous X-ray scattering that occurs from special atom types when they are exposed to X-ray energies that are near their absorption edge. Selenium is one such atom type and can be easily incorporated into proteins before crystallization by substituting methionine with selenomethionine during recombinant protein expression. Unlike MIR phasing, where differences in reflection intensities are measured by comparing a heavy atom derivative crystal to a native crystal, differences in reflection intensities arising from anomalous scattering are measured from reflections that are collected from a single crystal. This allows for the selenium atom substructure to be determined using a single

Protein Structural Analysis

121

crystal, thus avoiding problems of nonisomorphism and greatly improving the quality of initial phase estimates. Moreover, modern synchrotron sources allow for the precise tuning of X-ray wavelength, where collection of diffraction data at multiple wavelengths from a single crystal can be used to resolve the phase ambiguity that results from collecting data at only one X-ray wavelength. In cases where selenomethionyl protein cannot be made, anomalous scattering atoms can be soaked into protein crystals and the above procedure carried out on this single derivative. In either case, phase estimates generated from a MAD phasing experiment often produce electron density maps of exceptional quality. MR can be used if the 3D structure of a similar protein is available.13 Once the unit cell dimensions and space group have been determined for the crystal of a protein whose structure is unknown, the available protein structure can be used to model the unknown asymmetric unit of the crystal by rotating and translating it into the unit cell in accordance with the space group symmetry. Structure factor phases can then be estimated from the placement of the available structure in the unit cell. An electron density map is calculated by combining the MR phase estimates with the corresponding structure factor amplitudes that were obtained from X-ray diffraction measurements of the crystal of the unknown protein structure.

1.09.2.5

Model Building and Refinement

Analysis of the protein structure begins by building a molecular model of the polypeptide chain into an asymmetric unit that has been identified within the electron density map. This is carried out with the use of a computer graphics workstation that can rotate and translate the map in three dimensions. Prior knowledge of the amino acid sequence aids in building the polypeptide chain into the density map. During the building process, model parameters, such as bond lengths and angles, are refined to help minimize the difference between experimentally observed structure factor amplitudes and those calculated from the molecular model. To avoid overfitting the molecular model to the experimental diffraction data, refinement programs apply stereochemical restraints that regularize adjustable model parameters to acceptable stereochemical values.12 Improved phase estimates can be calculated using the refined model to generate electron density maps of enhanced quality. Using the improved maps, additional rounds of model building and refinement are carried out until the difference between the observed and calculated structure factor amplitudes has been minimized as much as possible. In addition to ensuring that the model is stereochemically and conformationally reasonable, an overall quantification of the agreement between the model and experimental data is given by the crystallographic R-factor.4 The general formula is P jjFobs j  jFcalc jj hkl P R¼ jFobs j hkl

Two R-factor indices are reported for the final refined model. The free R-factor is calculated using 5%–10% of randomly chosen observed reflections that are set aside and not used during model refinement, thus providing a measure of agreement between observed and calculated structure factor amplitudes that is less biased by the refinement process. The working R-factor is calculated using reflections against which the model was refined. Both indices should decrease during the proper refinement of a structure. For a model refined to 2.5-Å resolution, for example, the final working R-factor should not be much greater than 0.2.11 The free R-factor should not be much greater than the working R-factor and it should not diverge from the working R-factor during refinement, as this indicates possible overfitting of the model to the experimental data.4

1.09.2.6

Applications of Protein Crystallography to Biology

Analyzing the 3D structure of a protein by X-ray crystallography provides direct insight into the structural basis of protein function. In addition to providing detailed insight into molecular structure of a single protein, crystallographic structures of biological multimolecular complexes can reveal the molecular basis for how proteins interact with other biological molecules such as DNA, RNA, small molecule ligands, and other proteins. This information can guide a plethora of biochemical experiments, including site-directed mutagenesis studies aimed at understanding the molecular basis for how binding sites on protein specifically recognize and associate with other biological molecules, or to understand the catalytic function of individual amino acid side chains within the active site of an enzyme. Structure-guided functional studies such as these continue to enhance greatly our understanding of the molecular interactions of biological molecules and enzyme catalysis in both health and disease. From an industry perspective, analysis of protein structure by X-ray crystallography can provide extremely precious information for the development of small molecule inhibitors of proteins that are involved in disease. The crystallographic structure of a protein drug target can clearly reveal the 3D shape and electrostatic charge distribution of a potential drug-binding site (usually an enzyme active site). This information accelerates the design potent and selective inhibitors because the possible choices of inhibitor shape and electrostatic charge distribution can be focused to those that are complementary to the binding site of the targeted protein. Having already determined the native structure of the target protein, MR can be used to quickly determine the structure of the protein bound to, for example, a lead compound that was identified from a chemical library screen. The complex can reveal chemical changes that can be made to the lead compound to enhance its potency and specificity toward the target protein. This is an iterative process, where multiple rounds of chemical synthesis and structure determination are carried out during optimization of the drug candidate.

122

Protein Structural Analysis

Given the tremendous advances in protein X-ray crystallography over the last decade, it has become more accessible to researchers in the life sciences who are interested in understanding the molecular processes of life. The Protein Data Bank (http://www.rcsb.org/) is now experiencing exponential growth in the total number of crystallographically determined biological macromolecular structures. As more researchers continue to become interested in using protein structure to understand biochemical processes, the growth in the number of new protein structures deposited to the PDB is expected to increase and with this will come ever larger, more complex, and interesting biological structures that will further enhance our understanding of the molecular basis of life.

1.09.3

NMR Spectroscopy

Although the most powerful and widespread technique for the high-resolution determination of protein structure is X-ray crystallography, a major obstacle remains that numerous proteins are simply not amenable to crystallization. As a result, NMR spectroscopy has emerged as an important complementary technique to expand the breadth of potential protein structures that can be determined. As NMR is performed on a concentrated solution of the protein of interest, it also offers the significant advantage of examining protein structure under conditions that more closely mimic the physiological state. Furthermore, NMR spectroscopy is an extremely versatile technique that is capable of expanding the repertoire of the structural biologist beyond simple high-resolution structural determination to include the rapid examination of ligand binding, protein folding, and protein dynamics. As a result, NMR spectroscopy and X-ray crystallography are no longer viewed as competitive techniques but as a tandem set of approaches that can yield a spectrum of information ranging from ligand binding to high-resolution structural determination. This section focuses on understanding the theoretical underpinnings of NMR spectroscopy, sample preparation requirements, the structure determination process, and approaches to effect structural analysis beyond high-resolution structure.

1.09.3.1

Principles of NMR Spectroscopy

One of the major barriers preventing an inexperienced user from adopting NMR spectroscopy approaches is the perceived difficult theoretical underpinning of the technique. With recent advances in instrumentation, software, and experimental design, NMR spectroscopy has become partially automated and accessible to the novice user. The goal of this section is to provide a rudimentary outline of the physical principles upon which NMR spectroscopy experiments are based. More rigorous treatment of the physical basis for NMR spectroscopy is found elsewhere.3,10 NMR spectroscopy relies on the observation that most atomic nuclei possess an intrinsic property, nuclear spin, which generates a magnetic dipole. In the presence of a powerful external magnetic field (i.e., an NMR spectrometer), these magnetic dipoles align in one of two states: either with the external magnetic field or opposed to it. The quantized energy difference separating these two states can be determined, as can the frequency associated with transitions between the two energy levels. With absorption of radiofrequency (RF) pulses of the appropriate wavelength, the populations occupying the two states can be specifically perturbed. It is the measurement of the time dependence of the return to equilibrium configuration (relaxation) from this perturbed state that ultimately results in the NMR resonance signal.10 Only specific atomic nuclei are amenable to solution-state NMR spectroscopy due to their magnetic properties, for proteins most notably 1H, 13C, and 15N (spin quantum number of 1/2).10 These nuclei possess a characteristic absorption frequency or resonance that can be detected and is dependent upon the strength of the magnetic field in which it is situated. For example, in a spectrometer with an 18.8 T magnetic field, protons in isolation achieve resonance at approximately 800 MHz. In theory, all resonance frequencies for a given NMR-observable nucleus should be identical. The power of NMR spectroscopy is that the observed resonance frequencies in a complex sample are extremely dependent on the chemical environment surrounding the nuclei being examined. The resultant chemical shift from ideality allows for the dispersion of resonance signals and identification (assignment) of individual nuclei within a protein sample based on chemical environment. The magnitude of chemical shift is dependent upon the applied magnetic field strength, and for this reason, the frequency axes on NMR spectra are typically reported as the chemical shift of a particular nucleus relative to that nucleus in isolation in parts per million (ppm). Chemical shift ranges for protons (0–11 ppm), 15N (100–130 ppm), and 13C (40–200 ppm) are typically observed in proteins.3 In addition to dispersion created by chemical shift, magnetic coupling between nearby covalently bonded nuclei alters the electronic shielding a given nucleus experiences.10 Known as spin–spin coupling, this effect is smaller than typical chemical shifts and is therefore manifested on NMR spectra as resonance peak splitting. Analysis of spin–spin coupling provides detailed information about the through-bond covalent geometry of a protein. Additionally, certain NMR experiments allow nuclei to transfer magnetization through space when they are sufficiently close (within 5 Å) through a process known as the nuclear Overhauser effect (NOE).10 This transfer process is extremely sensitive to the distance between the nuclei (inverse of the sixth power of the separation distance) and can therefore identify atoms that are spatially proximate in the 3D structure of a protein. Despite the observed dispersion of chemical shifts from ideality, the raw number of similar nuclei present in a protein sample in combination with the observed spin–spin coupling between nuclei results in severely overlapped resonance peaks in 1D NMR spectra, making assignment of individual resonances from 1D spectra unfeasible.14 To increase dispersion such that individual resonances can be identified, multidimensional NMR techniques are required. At the heart of all multidimensional approaches

Protein Structural Analysis

123

is the predictable ability to transfer nuclear magnetization to specific nuclei through bonds (spin–spin coupling) or space (NOE) by manipulating the timing and nature of a series of RF pulses. These highly versatile pulse sequences can, therefore, be tuned to probe the covalent and noncovalent environment of nuclei in a protein structure.3,10 In 2D NMR, experiments are designed such that resonances from a 1D experiment are dispersed by sorting the signals onto a second frequency axis, typically via the chemical shift of a coupled nucleus. For example, a commonly used 2D NMR experiment in proteins, the heteronuclear single quantum coherence (HSQC) experiment, sorts backbone amide 1H resonances as a function of their connected 15N partner3 (Fig. 4). The result is a 2D array of peaks corresponding to the chemical environment of each backbone amide in the protein. 2D NMR spectroscopy can be easily expanded to higher order experiments (i.e., 3D and 4D) in order to increase the ability to identify specific nuclei within a protein sample.3 One of the major limitations of structure determination by NMR spectroscopy is the direct relationship between the relaxation processes that give rise to the NMR signal and molecular size. Essentially, the more rapidly a molecule tumbles in solution, the sharper the NMR signal. Larger proteins tumble slowly, which results in comparatively broader, less-intense signals. This has limited the study of relatively small proteins ( 800 proteins, shows a sixfold or greater change in expression. The stages are (1) reversible attachment; (2) irreversible attachment (clusters develop, motility is lost, and the las quorum sensing regulon is activated); (3) maturation I (rhl quorum sensing system becomes active); (4) maturation II; and (5) dispersion (pores and channels form which release planktonic bacteria). Upregulated proteins during the process include those of anaerobic processes, denitrification, efflux pump, and quorum-sensing proteins. Despite the pessimism existing about the future of antibiotic discovery, some scientists still have hope in the future. The most promising areas are 1. using high-throughput screening against targets such as fatty acid biosynthesis, peptide deformylase, lipid A biosynthesis, and tRNA synthetases; 2. employing genomic information revealing new targets which are present in pathogens but not in humans (also for “genome mining”); 3. cloning secondary metabolite pathways from uncultivated environmental microbes into culturable bacteria, that is, the metagenomic approach; 4. exploiting the vast microbial presence in the marine environment; 5. increasing the isolation of endophytic microbes, that is, microorganisms living in plants, for discovery of those antibiotics that protect against plant pathogens and against human disease; 6. continuing the exploration of techniques to culture uncultivated microorganisms; 7. combinatorial biosynthesis of new antibiotics; 8. using novel screening techniques; and 9. using natural products as scaffolds for combinatorial chemistry.

Secondary Metabolites

1.10.3

139

Other Applications of Secondary Metabolites

An extremely important concept for the further development of natural products is that compounds, which possess antibiotic activity, also possess other activities. Some of these activities had been quietly exploited in the past, and it became clear in the 1980s that such broadening of scope should be encouraged. Thus, a broad screening of antibiotically active molecules for antagonistic activity against organisms other than microorganisms, as well as for activities useful for pharmacological or agricultural applications, was proposed in order to yield new and useful lives for “failed antibiotics”. This resulted in the development of a large number of simple in vitro laboratory tests, for example, enzyme inhibition screens7 to detect, isolate, and purify useful compounds. Microbial secondary metabolites are now being used for applications other than antibacterial, antifungal, and antiviral infections. For example, immunosuppressants have revolutionized medicine by facilitating organ transplantation. Other applications include antitumor drugs, enzyme inhibitors, gastrointestinal motor stimulator agents, hypocholesterolemic drugs, ruminant growth stimulants, insecticides, herbicides, coccidiostats, antiparasitics versus coccidia, helminths, and other pharmacological activities. Further applications are possible in various areas of pharmacology and agriculture, developments catalyzed by the use of simple enzyme assays for screening before testing in intact animals or in the field.

1.10.3.1

Anticancer Agents

In the year 2000, approximately 10 million new cases of cancer were diagnosed in the world, resulting in 6 million cancer-related deaths. The tumor types with the highest incidence were lung (12.3%), breast (10.4%), and colorectal (9.4%). Of the 140 anticancer agents approved since 1940 and available for use, over 60% can be traced to a natural product. Of the 126 small molecules among them, 67% are natural in origin. In their review on the use of microbes to prescreen potential antitumor compounds, Newman and Shapiro8 concluded that microorganisms have played an important role in identifying compounds with therapeutic benefit against cancer. Most of the important compounds used for chemotherapy of tumors are microbially produced antibiotics. Approved antitumor agents from microorganisms are actinomycin D (dactinomycin), anthracyclines (including daunorubicin, doxorubicin (adriamycin), epirubicin, pirirubicin, idarubicin, valrubicin, amrubicin), glycopeptides (bleomycin, and phleomycin), the mitosane mitomycin C, anthracenones (mithramycin, streptozotocin, and pentostatin), and the endiyne calcheamycin attached to a monoclonal antibody (Mylotarg). Other antitumor products in use are vinblastine, vincristine, etoposide, taxol, mithramycin, deoxycoformycin, and L-asparaginase. A new anthracycline, 11-hydroxyaclacinomycin A, was produced by cloning the doxorubicin resistance gene and the aklavinone 11-hydroxylase gene dnrF from the doxorubicin producer, S. peucetius ssp. caesius, into the aclacinomycin A producer. The hybrid molecule showed greater activity against leukemia and melanoma than aclacinomycin A. Additional new anthracyclines were made by introducing DNA from S. purpurascens into S. galilaeus, both of which normally produce known anthracyclines. Novel anthracyclines were also produced by cloning DNA from the nogalomycin producer, S. nogalater into S. lividans and into an aclacinomycin-negative mutant of S. galilaeus. Cloning of the actI, actIV, and actVII genes from S. coelicolor into the 2-hydroxyaklavinone producer, S. galilaeus 31,671 yielded the novel hybrid metabolites, desoxyerythrolaccin, and 1-O-methyldesoxyerythrolaccin. Similar studies yielded the novel metabolite aloesaponarin II. Epirubicin (40 -epidoxorubicin) is a semisynthetic anthracycline with less cardiotoxicity than doxorubicin. Genetic engineering of a blocked S. peucetius strain provided a new method to produce it. The gene introduced was avrE of the avermectin-producing S. avermitilis or the eryBIV genes of the erythromycin producer, S. erythraea. These genes and the blocked gene in the recipient are involved in deoxysugar biosynthesis. The extremely toxic enediyne antitumor drug calicheamicin was attached to a humanized monoclonal antibody, and was approved for use against acute myeloid leukemia. The monoclonal antibody is designed to direct the antitumor agent to the CD 33 antigen, which is a protein expressed by myeloid leukemic cells. The conjugate is called Mylotarg (orgemtuzumab, and ozogamicin). This was a very exciting development, which serves as a model for the application of toxic antitumor agents in the future. Taxol (paclitaxel) has been a very successful antitumor molecule. It was originally discovered in plants by Wall and Wani,9 and has also been found to be a fungal metabolite.10 It is a diterpene alkaloid, approved for breast and ovarian cancer, which acts by blocking depolymerization of microtubules. In addition, taxol promotes tubulin polymerization, and inhibits rapidly dividing mammalian cancer cells. Taxol was originally isolated from the bark of the Pacific yew tree (Taxus brevifolia), but it took six trees of 100 years of age to treat one cancer patient. It is now produced by plant cell culture or by semisynthesis from taxoids made by Taxus species. The use of cells of the plant T. chinensis to produce taxol became the industrial means to make the compound. The addition of methyl jasmonate, a plant signal transducer, increased production from 28 to 110 mg/L. Fungi such as Taxomyces adreanae, Pestalotiopsis microspora, Tubercularia sp. and Phyllosticta citricarpa produce taxol. The highest level reported is 265 mg/L produced by P. citricarpa. Taxol has sales of $1.6 billion/year. A good source of secondary metabolites are the myxobacteria, relatively large Gram-negative rods, which move by gliding or creeping. They form fruiting bodies and have a very diverse morphology. Over 400 compounds had been isolated from these organisms by 2005, but the first in clinical trials were the epothilones, potential antitumor agents, which act like taxol, but are active against taxol-resistant tumors. They are 16-member ring polyketide macrolide lactones produced by the myxobacterium Sorangium cellulosum, which were originally developed as antifungal agents against rust fungi, but have found their use as antitumor

140

Secondary Metabolites

compounds. They contain a methylthiazole group attached by an olefinic bond. They are active against breast cancers, including those that are resistant to taxol and other forms of chemotherapy. They bind to and stabilize microtubules essential for DNA replication and cell division, even more so than taxol. One epothilone, ixebepilone, produced chemically by Bristol Myers-Squibb from epothilone B, has been approved by FDA. Epithilone polyketides are more water soluble than taxol. The producing microbe is a very slow grower (16 h doubling time) and low producer (20 mg/L). The epithilone gene cluster was cloned, sequenced, characterized, and expressed in the faster-growing S. coelicolor. The recombinant produced epithilones A and B, but unfortunately still at the low level of 50–100 mg/L. Camptothecin is a modified monoterpene indole alkaloid produced by certain plants (angiosperms) discovered in the 1960s by Wall and Wani.9 It also is produced by an endophytic fungus (Entrophospora infrequens) from the plant Nothapodytes foetida. It is used for recurrent colon cancer, and has unusual activity against lung, ovarian, and uterine cancer. Colon cancer is the second-leading cause of cancer fatalities in the USA and the third most common cancer among US citizens. Camptothecin is known commercially as Camptosar (Pharmacia) and Campto (Aventis and Yakult) and achieved sales of $1 billion in 2003. Its water-soluble derivatives irinotecan and topotecan are used clinically. In view of the low concentration of camptothecin in tree roots and poor yield from chemical synthesis, the fungal fermentation is very promising for industrial production of camptothecin. Its cellular target is type I DNA topoisomerase. When patients become resistant to irenotecan, its use can be prolonged by combining it with the monoclonal antibody Erbitux (cetuximab) from ImClone/BMS. Erbitux blocks a protein that stimulates tumor growth and the combination helps metastatic colorectal cancer patients expressing epidermal growth factor receptor (EGFR). This protein is expressed in 80% of advanced metastatic colorectal cancers. The drug combination reduces invasion of normal tissues by tumor cells and the spread of tumors to new areas. Metastatic testicular cancer, although rather uncommon (1% of male malignancies in the USA; 80,000 in the year 2000), is the most common carcinoma in men aged 15–35 years. Whereas the cure rate for disseminated testicular cancer was 5% in 1974, today, it is 90% mainly due to combination chemotherapy with the natural products bleomycin and etoposide, and the synthetic agent cisplatin. Angiogenesis (recruitment of new blood vessels) is necessary for tumors to obtain oxygen and nutrients. Tumors actively secrete growth factors which trigger angiogenesis. Antiangiogenesis therapy is now known as one of four cancer treatments. The other three are surgery, radiotherapy, and chemotherapy. By the end of 2007, 23 antiangiogenic drugs were in Phase III clinical trials and more than 30 were in Phase II. Fumagillin, produced by A. fumigatus, was one of the first agents found to act as an antiangiogenesis compound. Next to come along for angiogenesis inhibition were its oxidation product ovalacin and the fumagillin analog TNP-470 (¼AGM-1470). TNP-470 binds to and inhibits type 2 methionine aminopeptidase (MetAP2). This interferes with amino-terminal processing of methionine, which may lead to inactivation of enzymes essential for growth of endothelial cells. In animal models, TNP470 effectively treated many types of tumor and metastases. The marine environment offers new opportunities for the discovery of antitumor agents. The new genus Salinispora and its two species, S. tropica and S. arenicola, have been isolated around the world. These require seawater for growth. S. tropica makes a new bicyclic b-lactone b-lactam called salinosporamide A, which is a proteosome inhibitor, and is in clinical trials against cancer. Further, the genus Marinophilus contains species that produce novel polyenes, which have no antifungal activity but display potent antitumor activity. The symbionts of marine invertebrate animals continue to reveal interesting natural products. Variants of the toxic dolastin from the sea hare Dolabella auricalaria seem promising against cancer. These include soblidofin (T2F 1027), which completed Phase II against soft-tissue sarcoma, and synthadotin (¼tasidotin¼ 1LX 651), which is at the same clinical stage against melanoma, prostate, and nonsmall-cell lung cancers. These are thought to be produced by cyanobacteria sequestered by the marine invertebrates in their diet.

1.10.3.2

Immunosuppressants

Cyclosporin A was originally discovered in the 1970s as a narrow-spectrum antifungal peptide produced by the mold, Tolypocladium niveum (previously T. inflatum). Discovery of its immunosuppressive activity led to its use in heart, liver, and kidney transplants, and to the overwhelming success of the organ transplant field. Sales of cyclosporin A reached $1 billion in 1994. Although cyclosporin A had been the only product on the market for many years, two other products, produced by actinomycetes, provided new opportunities. These are rapamycin (¼sirolimus),11 and the independently discovered FK-506 (¼tacrolimus).12 They are both narrow-spectrum polyketide antifungal agents, which are 100-fold more potent than cyclosporin as immunosuppressants and less toxic. Tacrolimus and sirolimus have both been used clinically for many years. Tacrolimus has been used for transplants of liver, kidney, heart, pancreas, lung, and intestines, and for prevention of graft-versus-host disease. Recently, a topical preparation has been shown to be very active against atopic dermatitis, a widespread skin disease. Sirolimus does not exhibit the nephrotoxicity of cyclosporin A and tacrolimus and is synergistic with both compounds in immunosuppressive action. Sirolimus has been the basis of chemical modification to yield chemically important products such as everolimus, temsirolimus (CCI-779), and deforolimus (A23573). Studies on the mode of action of these immunosuppressive agents have markedly expanded current knowledge of T-cell activation and proliferation. Sirolimus, tacrolimus, and cyclosporin A act by interacting with an intracellular protein (an immunophilin), thus forming a novel complex, which selectively disrupts the signal transduction events of lymphocyte activation. The targets of

Secondary Metabolites

141

cyclosporin A, tacrolimus, and sirolimus are inhibitors of signal transduction cascades in microorganisms and humans. In humans, the signal transduction pathway is required for the activation of T cells. A previously unknown protein called mTOR (¼target of rapamycin) is a member of the family of lipid/protein kinases and part of the sirolimus-sensitive signal transduction pathway. Sirolimus, combined with its immunophilin, inhibits TOR kinase, which normally transduces growth-promoting signals that are sent in response to nutrients (e.g., amino acids) and growth factors. TOR has phosphatidylinositol lipid kinase activity, which is involved in cell cycle regulation. A broad-spectrum antibiotic discovered in 1893, actually the first antibiotic ever discovered, is mycophenolic acid produced by Penicillium brevicompactum. Mycophenolic acid has antibacterial, antifungal, antiviral, antitumor, antipsoriasis, and immunosuppressive activities. It was never commercialized as an antibiotic because of its toxicity, but its 2-morpholinoethylester was approved as a new immunosuppressant for kidney transplantation in 1995, and for heart transplants in 1998.

1.10.3.3

Hypocholesterolemic Compounds

Only 30% of the cholesterol in the human body comes from the diet. The remaining 70% is synthesized by the body, mainly in the liver. Many people cannot control their cholesterol at a healthy level by diet alone, but must depend on hypocholesterolemic drugs. Statins inhibit de novo production of cholesterol in the liver, the major source of blood cholesterol. High blood cholesterol leads to atherosclerosis, which is a causal factor in many types of coronary heart disease, a leading cause of human death. Statins were a success because they reduced total plasma cholesterol by 20%–40%, whereas the previously used fibrates only reduced it by 10%–15%. The statins are microbially produced enzyme inhibitors, inhibiting 3-hydroxy-3-methylglutaryl-coenzyme A reductase, the regulatory and rate-limiting enzyme of cholesterol biosynthesis in liver. Statins were first discovered in fungi in England and Japan in 1976, with mevastatin (compactin) from P. brevicompactum being reported as an antifungal agent by Brown et al.13 of Glaxo. Earlier in 1968, Akira Endo, at Sankyo in Japan, screened 6000 fungal extracts for inhibition of cholesterol biosynthesis by rat liver membranes and found two actives, that is, ML-236A and B, produced by P. citrinum. ML-236B was named compactin (¼mevastatin) and found to inhibit 3-hydroxy-3-methylglutaryl coenzyme A reductase.14 In 1976, Sankyo prepared a patent application on compactin, but it never became a commercial drug. At that time, Alberts’ group at Merck started to screen for new inhibitors and discovered lovastatin to be produced by A. terreus, which they called Mevacor. It had a structure similar to compactin, but contained a methyl group. At about the same time, that is, 1979–80, Endo reported the discovery of lovastatin from Monascus ruber, which he named monocolin K (¼mevinolin). It was patented in Japan but without structural elucidation. Merck filed for a patent containing their findings including the structure of lovastatin. The company received a US patent in 1980, and lovastatin became the first statin on the market. Further clinical tests on lovastatin went into full speed in 1982, and the drug was finally approved by FDA in 1987 after clinical tests in humans had shown a lowering of total blood cholesterol of 18%–34%, a 19%–39% decrease in low-density lipoprotein cholesterol (“bad cholesterol”), and a slight increase in high-density lipoprotein cholesterol (“good cholesterol”). Merck later produced simvastatin (Zocor) in which the 2-methylbutanoate side chain of lovastatin was chemically modified to 2,2-dimethylbutanoate; it was launched in 1988. Although compactin was never used medically, Sankyo bioconverted it by hydroxylation, yielding pravastatin, which became commercial in 1989, and was then licensed to Bristol–Myers Squibb. The bioconversion is carried out using actinomycetes. In 1985, Bruce D. Roth was able to chemically synthesize one of the statins that Endo had isolated from his fungus in the 1970s. Two years later, he headed a group at Parke–Davis working on synthesis of a synthetic statin called atorvastatin (Lipitor). They compared atorvastatin in a further clinical trial versus fluvastatin, lovastatin, pravastatin, and simvastatin; atorvastatin showed the best results. FDA approved it in January 1997. Parke–Davis decided to comarket it with Pfizer in 1996. By mid-1998, atorvastatin had 18% of the statin market as compared to simvastatin’s 37%. Pfizer, in 2000, purchased Warner-Lambert, the parent of Parke– Davis, and became the sole owner of atorvastatin, which became the leading drug in the world. The largest segment of the pharmaceutical business is for cholesterol-lowering drugs, amounting to about 30% of the market. Simvastatin reached sales of over $7 billion and pravastatin attained sales of $5 billion. Atorvastatin became the world’s leading drug at $12 billion/year. Natural statins are produced by many fungi: A. terreus and species of Monascus, Penicillium, Doratomyces, Eupenicillium, Gymnoascus, Hypomyces, Paecilomyces, Phoma, Trichoderma, and Pleurotis. Although pravastatin is commercially made by bioconversion from compactin, certain strains of Aspergillus and Monascus can produce pravastatin directly. The titer of lovastatin in A. terreus reached 7–8 g/L while that of compactin in P. citrinum has been reported to be 5 g/L. Simvastatin has traditionally been made by a semisynthetic multistep process starting with lovastatin. This can now be avoided by use of an E. coli strain overexpressing LovD in the presence of a cell-membrane permeable thioester, that is, dimethylbutyryl-5methylmercaptopropionate. Statins reduce cardiovascular events including myocardial infarction, stroke, and death. Not only are they active against atherosclerosis, the most common cause of death in Western countries, but also improve endothelial function, and are antiinflammatory. The antiinflammatory effect is over and above their action in lowering cholesterol. They are also reported to lower the occurrence of Alzheimer’s disease and reduce inflammation. Statins also lower elevated C-reactive protein (CRP) levels independent of their effect on cholesterol. This is important as half of all myocardial infarctions occur in patients with normal low-density lipoprotein (LDL) levels. Statins can also prevent stroke, and reduce development of peripheral vascular disease.

142 1.10.3.4

Secondary Metabolites Antihelmintic Agents

A major agricultural problem has been the infection of farm animals by worms. The predominant type of screening effort over the years was the testing of synthetic compounds against nematodes, and some commercial products did result. Although Merck Laboratories had developed a commercially useful synthetic product, thiobenzole, they had enough foresight to also examine microbial broths for antihelmintic activity. The S. avermitilis culture, which was isolated by Omura and coworkers at the Kitasato Institute in Japan and sent to Merck, was found by Campbell and others to produce a family of secondary metabolites having both antihelmintic and insecticidal activities. These killed the intestinal nematode, Nematospiroides dubius, in mice and were named “avermectins”. They are disaccharide derivatives of macrocyclic lactones with exceptional activity against parasites, that is, at least 10 times higher than any synthetic antihelmintic agent known. Despite their macrolide structure, avermectins lack antibiotic activity, do not inhibit protein synthesis nor are they ionophores; instead, they interfere with neurotransmission in many invertebrates. They have activity against both nematode and arthropod parasites in sheep, cattle, dogs, horses, and swine. A semisynthetic derivative, 22, 23-dihydroavermectin B1 (Ivermectin) is 1000 times more active than thiobenzole, and is a commercial veterinary product. Ivermectin is made by hydrogenation at C22–C23 of avermectin B1a and B1b with rhodium chloride acting as catalyst. By genetic engineering of S. avermitilis in which certain PKS genes were replaced by genes from the PKS of S. venezuelae (the pikromycin producer), ivermectin could be made directly by fermentation. The addition of the S-adenosylmethionine analog, sinefungin, to the avermectin producer yielded eight new demethylated derivatives and shifted fermentation from the A þ B mixture to predominantly the B (less methylated) forms, which are more desirable. A new avermectin, called Doramectin (¼cyclohexylavermectin B1), was developed at Pfizer by the technique of mutational biosynthesis. Structural analogs of avermectin are used as endectocides in certain species. Emamecin is used for salmon and eprinomectin for cattle. Semisynthesis of 400 -amino-400 -deoxyavermectin B1 from avermectin B1 yielded a major (1500-fold) increase in activity against lepidopterin insect larvae such as Spodoptera eridania, the southern armyworm. The 1998 market for endectocides (avermectins) was over $1 billion, divided among livestock ($750 million) and pets ($330 million). The avermectins are closely related to the milbemycins, a group of nonglycosidated macrolides produced by S. hygroscopicus ssp. aureolacrimosus. These compounds, discovered by Sankyo researchers, possess activity against worms and insects. A fortunate fallout of the avermectin work was the finding that ivermectin has activity against the black fly vector of human onchocerciasis (“river blindness”). It interferes with transmission of the filarial nematode, Onchocerca volvulus, to the human population. Because 40 million people were affected by this disease, the decision by Merck to supply ivermectin free of charge to the World Health Organization, for use in humans in the tropics, was met with great enthusiasm and hope for conquering this parasitic disease. Thus, Merck has given ivermectin free of charge to over 200 million people in 33 countries since 1987 to cure river blindness.

1.10.3.5

Other Applications

There are many more uses for microbial secondary metabolites. These are described in Table 2. Table 2

Additional applications of secondary metabolites

Application Enzyme inhibition a-amylase a-glucosidase Pancreatic lipase protease Gastrointestinal motor stimulation Ruminant growth stimulation Herbicidal against broad-leaved weeds Antiparasitic against coccidiosis Malaria Agricultural fungicidal Bioinsecticidal

Compound

Producer microorganism

Paim Acarbose Lipstatin Chymostatin

Streptomyces corchorushi Actinoplanes sp. SE50 Streptomyces toxytricini S. hygroscopicus

Erythromycin

Saccharopolyspora erythraea

Pleuromutilins Quinoxalines Virginiamycin

Pleurotus mutilis Streptomyces echinatus Streptomyces virginiae

Bialaphos

Streptomyces viridochromogenes

Monensin Lasalocid Salinomycin Fosmidomycin

Streptomyces cinnamonensis Streptomyces albus S. albus Streptomyces lavendulae

Strobilurin A

Strobiluris tenacellus

Spinosyns

S. spinosa

Secondary Metabolites

143

References 1. Zasloff, M. Magainins, a Class of Antimicrobial Peptides from Xenopus Skin: Isolation, Characterization of Two Active Forms, and Partial CDNA Sequence of a Precursor. Proc. Natl. Acad. Sci. U. S. A 1987, 84, 5449–5453. 2. Fleming, I. D.; Nisbet, L. J.; Brewer, S. J. Target Directed Antimicrobial Screens. In Bioactive Microbial Products: Search and Discovery; Bu‘lock, J. D., Nisbet, L. J., Winstanley, D. J., Eds., Academic Press: London, 1982; pp 107–130. 3. Woodruff, H. B.; Hernandez, S.; Stapley, E. D. Evolution of Antibiotic Screening Programme. Hindustan Antibiot. Bull. 1979, 21, 71–84. 4. Baltz, R. H. Antibiotic Discovery from Actinomycetes: Will a Renaissance Follow the Decline and Fall? SIM News 2005, 55, 186–196. 5. Gloer, J. B. Applications of Fungal Ecology in the Search for New Bioactive Natural Products. In The Mycota, Vol. IV, .; 2nd ed.; The Mycota, Vol IV, Kubicek, C. P., Druzhinina, I. S., Eds.; Springer: New York, NY, 2007; pp 257–283. 6. Brennan, M. B. Drug Discovery: Filtering Out Failures Early in the Game. Chem. Eng. News 2000, 78 (23), 63–74. 7. Umezawa, H. Low-molecular-weight Inhibitors of Microbial Origin. Annu. Rev. Microbiol. 1982, 36, 75–99. 8. Newman, D. J.; Shapiro, S. Microbial Prescreens for Anticancer Activity. SIM News 2008, 58, 132–150. 9. Wall, M. E.; Wani, M. C. Campothecin and Taxol: Discovery to Clinic. Canc. Res. 1995, 55, 753–760. 10. Stierle, A.; Strobel, G.; Stierle, D. Taxol and Taxane Production by Taxomyces Andreanae, an Endophytic Fungus. Science 1993, 260, 214–216. 11. Vezina, D.; Kudelski, A.; Sehgal, S. N. Rapamycin (AY 22,989), a New Antifungal Antibiotic. 1. Taxonomy of the Producing Streptomycete and Isolation of the Active Principle. J. Antibiot. 1975, 28, 721–726. 12. Kino, T.; Hatanaka, H.; Hashimoto, M.; et al. FK-506, a Novel Immunosuppressant Isolated from Streptomyces. 1. Fermentation, Isolation and Physico-chemical and Biological Characteristics. J. Antibiot. 1987, 40, 1249–1255. 13. Brown, A. G.; Smale, T. C.; King, T. J.; et al. Crystal and Molecular Structure of Compactin, a New Antifungal Metabolite from Penicillium brevicompactum. J. Chem. Soc. Perkin Trans. I 1976, 1165–1170. 14. Endo, A.; Kuroda, M.; Tsujita, Y. ML-236B and ML-236C, New Inhibitors of Cholesterogenesis Produced by Penicillium Citrinin. J. Antibiot. 1976, 29, 1346–1348. 15. Bumann, D. Has Nature Already Identified All Useful Antibiotic Targets? Curr. Opin. Microbiol. 2008, 11, 387–392.

1.11

Cell Line Isolation and Design

Trent P Munro, Warren Pilbrough, Benjamin S Hughes, and Peter P Gray, The University of Queensland, Brisbane, QLD, Australia © 2011 Elsevier B.V. All rights reserved. This is a reprint of T.P. Munro, W. Pilbrough, B.S. Hughes, P.P. Gray, 1.13 - Cell Line Isolation and Design, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 169-178.

1.11.1 1.11.2 1.11.2.1 1.11.2.2 1.11.2.2.1 1.11.2.2.2 1.11.3 1.11.3.1 1.11.3.2 1.11.3.2.1 1.11.3.2.2 1.11.3.3 1.11.3.4 1.11.4 1.11.4.1 1.11.4.2 1.11.4.3 1.11.4.4 1.11.4.5 1.11.4.6 1.11.5 1.11.5.1 1.11.5.2 1.11.5.3 References

Introduction Clone Selection and Isolation Manual Clone Isolation Automating Clone Selection and Isolation Robotic Colony Picking Flow Cytometry and Fluorescence-Activated Cell Sorting Automating Clone Screening Manual Clone Screening Automated Clone Screening Microbioreactors Cell-Imaging Devices High-Throughput Productivity and Product Quality Assessment Assessing Clone Stability Designer Cell Lines for Bioproduction Regulatory Considerations Traditional Versus Designer Cell Lines The Contribution of Vector and Host Cell Classical Cell Line Development Novel Vector Design Strategies Host Cell Engineering: The New Frontier Future Perspectives and Conclusions High ProductivitydAre We There Yet? Future of Bioproduction Summary and Conclusion

144 145 145 145 145 147 147 148 148 148 149 149 149 149 149 150 150 150 151 151 152 152 152 152 152

Glossary Aneuploidy An aberrant state where the number of chromosomes is not a multiple of the haploid number, usually reflecting genomic instability. Biopharmaceuticals Therapeutic products derived from the biotechnology industry. Cell cloning Isolation and expansion of a single cell to produce a population of cells with similar properties. Cell line development Creation of characterized mammalian cell strains for the production of biopharmaceuticals or vaccines. Continuous cell lines Cell lines that have been immortalized due to mutation or other changes, and no longer undergo senescence. Epigenetics Heritable effects on phenotype not associated with DNA sequence. Normally, chemical modifications of histones and DNA, which affect the degree of expression.

1.11.1

Introduction

Biopharmaceuticals are a rapidly growing segment of the pharmaceutical industry, representing about 10% of the drugs on the market. These drugs also make up a quarter of all blockbusters (sales > US$1 billion) and nearly half of new entities in development. Global biopharmaceutical sales exceeded US$100 billion in 2009. Major classes of biopharmaceuticals include: recombinant vaccines, hormones, growth factors, cytokines, blood factors, therapeutic enzymes, and monoclonal antibodies (mAbs). Among these, sales of mAbs have shown a staggering 35% average growth rate over the past 3 years. Three mAbsdEnbrel, Remicade, and Rituxandare also in the top 10 of all drugs on the market by sales. Much of this article is focused on mAb production, but similar considerations apply to the development of other biological drugs or vaccines. For a detailed review of the biopharmaceutical marketplace, see Ref. 1.

144

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00013-6

Cell Line Isolation and Design

145

The production of this class of drugs has only been possible through advances in biotechnology, genetic engineering, and bioprocessing. Unlike small molecule drugs, which are synthesized chemically and can be thoroughly characterized at the molecular level, many therapeutic proteins are large and complex, requiring correct folding and posttranslational modifications for biological activity and must be produced using living cells. Although simple prokaryotic hosts such as Escherichia coli are sufficient to produce certain drugs such as insulin and human growth hormone, the focus of this article is on therapeutics derived from mammalian cells. The creation of a stable recombinant mammalian cell line for bioproduction is a complex and multistep process that takes considerable time and resources. Initially, a plasmid vector containing DNA encoding the therapeutic protein must be transferred into a cell and stably inserted into the genome. This usually involves random insertion mediated by chemical, viral, or physical means followed by application of selection pressure to kill off cells not stably integrating the plasmid DNA with a corresponding resistance gene. Depending on the protocol and selection agent chosen, gene amplification may also be induced to increase the level of expression, usually by inhibiting the selection marker. The most common amplifiable selection markers used in industry at present are glutamine synthetase, inhibited by methionine sulfoxide, or dihydrofolate reductase (DHFR), inhibited by methotrexate (MTX) (for a review, see Ref. 2). Although this classical approach can be very effective, it results in a heterogeneous distribution of protein expression levels in the transfected cell population. To isolate a production cell line, single-cell clones must be chosen from the mixed initial pool, which possess requisite characteristics for bioproduction. These characteristics include sustained high-level expression of the transgene, rapid cell growth, and appropriate product quality attributes. The cloning step is critical for product development and has been the focus of intense study aimed toward increase in the probability of selecting the best clone for subsequent scale-up and manufacturing.3 The occurrence of suitable clones is rare, and clone selection, isolation, and screening are major bottlenecks in drug development and accelerating this process would reduce time to market, bringing forward future cash flows and extending the period of patent protection for a product. This article explores the state of the art in single-cell isolation, highthroughput clone screening, and cell line engineering, and the role of these disciplines in cell line development and design.

1.11.2

Clone Selection and Isolation

As described above, because most transfections involve random integration and the widely used selection techniques mediate gene amplification, expression levels within stable pools tend to be highly variable.2 The goal of clonal isolation is, thus, to select a small number of rare, high-expressing clones with suitable bioprocess and product quality attributes from a large number of unsuitable clones with diverse phenotypes.

1.11.2.1

Manual Clone Isolation

In the past, cell clones were typically isolated by basic manual methods, the simplest of these being limiting dilution cloning, in which cells are serially diluted in multiwell plates to a theoretical concentration of less than one cell per well. After incubation, cells from wells at the lowest dilution showing growth are expanded and screened for expression. Unfortunately, this technique is nonselective, needs many plates to isolate even a few candidate clones, and must be repeated several times to assure clonality. This is because Poisson’s effects, cell clumping, and limited plating efficiency frequently result in growing wells, which originate from more than one cell. Scoring clonality by direct microscopic observation at the time of plating is theoretically possible, but rarely practical, and requires an experienced operator. A variation of limited dilution cloning is known as capillary-aided cell cloning, which increases the likelihood of monoclonality by spotting cells at a dilute concentration in 1-mL droplets into wells of a plate. Clumps cannot pass through the capillary and the small droplets are less likely to contain more than one cell. Direct manual colony picking is an alternative method for adherent cells. It is laborious and also requires a high degree of operator skill. Cells are grown under selection until a suitable number of colonies appear that are still easily distinguished from their neighbors. Cells are observed repeatedly under the microscope and healthy colonies of around 500 cells that appear compact and polygonal are chosen. Small, branched, or irregular colonies are avoided, as these may not be clonal or may have undesirable growth characteristics. Once a colony has been identified, it can be manually picked using a custom-made cloning cylinder or modified pipette tip. The picked colony is then transferred to fresh selection media for characterization and scale-up. The cloning techniques discussed so far involve random selection and do not discriminate by productivity. Thus, a greater number of clones must be isolated and screened to find a high producer (typically hundreds to thousands of clones), making manual clone selection a very tedious exercise indeed.

1.11.2.2 1.11.2.2.1

Automating Clone Selection and Isolation Robotic Colony Picking

A number of companies have developed specific equipment to automate clonal isolation. Both the CellCelector from The Automation Partnership (www.automationpartnership.com) and the ClonaCell EasyPick from Stem Cell Technologies (www. stemcell.com) and Hamilton (www.hamiltonrobotics.com) couple precise liquid-handling robotics with automated microscopic image acquisition. Transfected stable pools are grown in semi-solid methylcellulose-based media and imaging software is used to identify cells based on predetermined morphology. The robotic liquid handler is then used to pick and dispense colonies into individual wells for further analysis and outgrowth. Although this method does eliminate much of the labor associated with manual cloning, it fails to provide any information on potential cellular productivity, with colony morphology being the only available criterion for cell selection. A more advanced technology for intelligent colony isolation is the Genetix ClonePix FL (www.genetix.com) (Fig. 1). This self-contained system not only incorporates automated image acquisition and precise colony picking but also utilizes a detection

146

Cell Line Isolation and Design

A

C

E

G

B

D

F

H

Figure 1 Genetix ClonePix FL workflow. (A) The Genetix ClonePix FL instrument. (B) Screenshot from the ClonePix FL software showing cell colonies growing in semisolid media. Antibody expression can be seen in green with secreted protein creating a fluorescent halo around the colonies. (C) Picking pins inside the ClonePix FL instrument and low-magnification fluorescence and white light images representing antibody expression and cell growth, respectively. (D) Ranking plot showing mean fluorescence intensity of clones after analysis. The highest ranked clones are highlighted and targeted for picking. (E) Image of the Genetix Clone Select Imager (CSI) instrument. (F) Output from the CSI software showing cell growth in individual wells over time and thumbnails of cell confluence. (G) CSI imaging data showing cell growth across a 96-well plate. This plate received picked colonies from the ClonePix instrument. (H) CSI software interpretation of the cell growth seen in (G).

Cell Line Isolation and Design

147

reagent that provides a fluorescent readout of the amount of product secreted by each colony. Using the inbuilt software, thousands of individual colonies can be scored for growth and morphology by means of white light illumination, and the corresponding product secretion level can be scored by fluorescence. User-defined criteria are then used to identify colonies expected to be high producers. Colonies meeting these selection criteria are automatically picked and transferred to destination plates for outgrowth and further analysis. It should be noted that colony clonality is not guaranteed, meaning additional rounds of cloning may be required. A similar system has also been developed by Cyntellect known as CellXpress (www.cyntellect.com). CellXpress is powered by a technology known as Laser-Enabled Analysis and Processing (LEAP) and also relies on fluorescent-detection reagents for productivity assessment, but instead of picking colonies of interest, uses a laser to ablate nonproducing cells. This may represent a potential limitation as incomplete killing and clone contamination are a possibility. In addition, because it isolates individual cells or small numbers of cells in the beginnings of a colony, it is more subject to clonal variation than the ClonePix, which averages expression at the colony level after a longer period of growth. Nonetheless, both CellXpress and ClonePix FL offer significant advantages over traditional cell cloning methodologies.

1.11.2.2.2

Flow Cytometry and Fluorescence-Activated Cell Sorting

Flow cytometry is a powerful technique for single-cell analysis. In combination with immunodiagnostic reagents, a broad variety of intracellular and membrane-bound markers can be measured simultaneously in millions of individual cells. Instruments such as the Becton Dickinson FACSAria (www.bd.com) analyze up to 70,000 events/s allowing detection of very rare events in a cell population. Many such instruments also have the capability to capture cells according to a particular fluorescent or scatter profile and sort them as single cells or defined subpopulations into a tube or well plate. Several approaches incorporating flow cytometry into clone isolation for biopharmaceutical cell line development have been described.3–5 In industry, years of vector design refinements have led to highly stringent platforms for stable pool generation using chemical selection. In such cases, the average expression level may be relatively high, and flow cytometry combined with single-cell deposition may be used simply as a tool for delivering single cells into wells. Gating by light scattering helps to eliminate doublets and cell clumps. This method, although simpler and more reliable than the manual methods described above, still relies on secondary imaging of plates to confirm clonality of each well, because doublets or stray cells may sometimes be delivered despite proper gating. Several more advanced flow cytometry and fluorescence-activated cell sorting (FACS)-based techniques have also been developed that provide at least some readout of cell-specific productivity for selection of high producers. But as mentioned above, selection based on a single cell can be prone to clonal variation.6 For the DHFR system, fluorescent MTX can be used as a staining reagent to allow identification of cells with high levels of DHFR activity that have undergone gene amplification and are more likely to be high producers. Similarly, direct fusions between DHFR and a fluorescent protein, such as green fluorescent protein, can help identify high producers by intracellular fluorescence. Unfortunately, amplification of selection markers does not necessarily correlate well with higher expression of product, and clones with a high degree of amplification may be unstable. In a more direct assay, a fluorescent protein may be co-expressed with the transgene on a dicistronic transcript.7 In this way, reporter expression is directly proportional to product expression. This method, however, only provides a transcriptional readout and does not address potential translational or cell secretion bottlenecks. Nevertheless, readings can be made directly on cells from culture without any exogenous label. It can also be used in a multicolor assay with other labeling strategies or in conjunction with instruments such as the ClonePix FL mentioned above. Secreted proteins such as recombinant antibodies can be measured directly as they transit through the cell membrane. This method has been termed cold capture, as cells are placed on ice to slow secretion and allow for detection of membraneassociated antibody.8 Live cells labeled with fluorescent secondary antibodies are then analyzed by flow cytometry and selected based on fluorescent signal, which correlates to the amount of membrane-associated antibody product. Studies have shown good agreement between this signal and the actual rate of secretion and the method is simple to perform. A hybrid of the preceding two methods has also been developed in which the membrane protein CD20 is expressed on a dicistronic messenger RNA (mRNA) with the gene of interest. CD20 is trafficked to the cell membrane and can be detected using a fluorescent anti-CD20 antibody. This method provides a readout representative of the cellular transcription rate and potentially also of the secretion rate, although a membrane protein may be subject to somewhat different dynamics than a fully secreted protein. Advantages of this method include that the reporter molecule remains stably inserted in the membrane for a higher signal output, and the same detection molecule can be used for a range of expression targets. Other techniques also exist to measure secretion rate by FACS, including gel microdrop and affinity matrix. Both of these methods use a primary biotin–avidin immobilization, which in turn captures secreted product and is read after subsequent addition of a fluorescently conjugated detection antibody. In practice, these methods are relatively cumbersome and technically challenging, involving more steps than cold capture, requiring specialized equipment in the case of gel microdrop, and extensive assay optimization. For details, see Ref. 3.

1.11.3

Automating Clone Screening

After clone selection and isolation (Section 1.11.2), hundreds or even thousands of clones must be screened at various expansion stages to narrow the candidates down to a final production clone. This generates a need for representative high-throughput clone screening and analysis tools.

148

Cell Line Isolation and Design

1.11.3.1

Manual Clone Screening

Clones are typically cultured initially in large numbers of 96- or 384-well plates. A first round screen is performed to reduce the number of cultures to a manageable level so that they can be expanded. Supernatants are collected from wells showing good growth, and a primary screen by enzyme-linked immunosorbent assay (ELISA; or equivalent assay) is conducted. A question in early-stage screening is how predictive titer measurements are of a clone’s true potential at scale. This is because of the different culture format (static/adherent), the difficulty in quantifying cell numbers to assess per cell productivity, and the unknown suspension growth characteristics that ultimately contribute to volumetric yield in the final production system. The top-ranked clones by ELISA (typically the top 5%–30%) are expanded into 24-well plates with respective supernatants removed for further ELISA screening once the plates reach sufficient confluence (typically >50%). The screening process is repeated with cultures expanded to 6-well plates, at which point the top candidates (generally 6–24) are more carefully assessed in suspension culture screens. These studies are usually performed in noninstrumented culture vessels such as spinners or shake flasks. An alternative small-scale model that lends itself to higher-throughput evaluation of clones is the “tubespin” reactor system,9 consisting essentially of vented 50 mL tubes. Each clone is periodically sampled to compare cell growth rates, cell-specific productivity, and volumetric yield (titer). If cells are not preadapted to suspension growth (and/or serum-free media), this would need to occur here, which can lengthen the process considerably. It is also important during screening to assess clones for product quality and eliminate those that show unsatisfactory profiles. Lastly, the top four to six clones are evaluated in bioreactors under fully controlled conditions to choose the final bioproduction clone. To better predict performance, generic or platform fed-batch protocols may be implemented as part of the clonal selection process, although typically batch processes are used for early-stage screening.

1.11.3.2

Automated Clone Screening

Screening and expanding clones can be logistically challenging. For example, different growth rates can put wells out of synchronization and require cherry-picking of wells. However, there are automated platforms that can increase the capacity for carrying clones forward to later screening steps where they can be better evaluated. The Cello system (The Automation Partnership) is such an example, handling tens of thousands of clones in static plates. It minimizes operator workload by monitoring clonality with a microscope, removing samples for online antibody titer determination, and handling all culture expansions through screening up to 6-well plates. Researchers using the system estimate the Cello can increase project throughput threefold, without increasing manual workload.10 However, the system is purely for static culture, which is broadly indicative but not ideal for predicting productivity under suspension growth. The Automation Partnership also makes an automated shake flask handling system, Sonata, which uses up to 1 L flasks, but is not designed specifically for clone screening.

1.11.3.2.1

Microbioreactors

Traditionally, clone screening has been challenged by timeline pressure and the need to deliver manufacturing-relevant clones. Clones that perform well under initial screening conditions often show low correlation to fed-batch performance in stirred-tank bioreactors (STBRs). Increasingly, industry is looking to utilize high-throughput microbioreactors to reduce timelines and build in additional quality to the clonal screening process. Applications range from evaluating clones for “manufacturability” and improved growth characteristics such as low lactate synthesis, to determine product quality and glycosylation profiles, which may be affected by variations in culture format. Evaluating clones that will ultimately be grown under suspension conditions places increased significance on introducing shaking as early as possible in the screening process. Various labs have developed their own high-throughput “mini” or microbioreactor systems utilizing shaken 96-well deep-well plates and using liquid-handling robots to expand clones. There are also commercially available systems such as m-Flask by Duetz (Applikon). The advantages of such systems are that they provide preliminary information better approximating STBRs compared to growth under static conditions, but the mode of mixing and lack of control, as compared to STBRs, means these systems often fail to be representative of performance under scaled-up conditions. There do, however, exist multiple microbioreactor options offering improved control that have been successfully demonstrated as accurate scale-down models.11 The M24 microreactor (manufactured by Microreactor Technologies Inc., distributed by Pall) provides noninvasive online monitoring and the ability to individually control culture pH, dissolved oxygen (DO), and temperature of up to 24 cultures in parallel. The custom 24-well deep-well plate with 7 mL culture volumes offers the control and data quality of lab-scale bioreactors, though the sparging rates often require the addition of antifoam. This is to be addressed soon with a new plate design. Alternate commercial offerings include the Cellstation high-throughput bioreactor system (Fluorometrix), which utilizes twelve 35-mL working volume glass bioreactors. This system has stirrers and mixing comparable to a STBR and has been successfully demonstrated by identifying process-relevant characteristics of SP2/0 clones. The ambr (The Automation Partnership) is a bioreactor mimic that offers 24 or 48 disposable 10- to 15-mL working volume reactors with robotic feeding and sampling. Finally, the SimCell (Seahorse Bioscience) is a microbioreactor platform with a working volume of only 700 mL but a capacity for up to 1260 parallel cultures. The fully automated robotic system utilizes optical measurement of DO, pH, and total cell density, and control of temperature, pH, DO, and glucose levels.

Cell Line Isolation and Design 1.11.3.2.2

149

Cell-Imaging Devices

Many robotic automation packages mentioned above utilize in-built imaging software and microscopes to track cell growth. Stand-alone devices for high-throughput imaging of plates during clonal selection include the CloneSelect Imager (Genetix) (Fig. 1), and the newer fluorescent-capable CellReporter (Genetix) and Cellavista (Roche, www.roche-applied-science.com). These devices can also be integrated into robotic platforms and allow for rapid (
3 min) scanning of plates for cell-confluence measurements. When the same plates are imaged over time, the in-built software can track clonality and growth rates. Furthermore, optimal culture viability can be maintained during clonal expansion by ensuring cultures are passaged prior to exceeding >80% confluence for extended periods. A number of other automated cell-imaging devices are now also on the market, which could be used to analyze clonal outgrowth. Some examples include the Essen IncuCyte (www.essenbioscience.com), the Celigo adherent cell cytometer (www. cyntellect.com), and even high-content imagers such as the BD Pathway (www.bdbiosciences.com), PE Operetta (www. perkinelmer.com), and GE Healthcare IN Cell analyzer (www.gehealthcare.com).

1.11.3.3

High-Throughput Productivity and Product Quality Assessment

Productivity is determined from antibody concentration, which is traditionally measured by ELISA or high-performance liquid chromatography (HPLC). Today, robotics and 96-well filter plates, either loaded with protein A, or utilizing magnetic protein A bead assays, now enable high-throughput analysis of clone supernatant samples. Also, multiple technology offerings also exist for label-free high-throughput immunoglobulin G quantitation, such as the Biacore from GE (www.biacore.com) or Octet platform from ForteBio (www.fortebio.com). Quality attributes must also be assessed prior to final clone choice to ensure commercial relevance. The need for product quality data is underscored by the fact that most quality attributes are clone dependent. These may include molecular integrity, aggregation, glycosylation, and charge heterogeneity. Due to constraints of time and throughput, not all attributes can be evaluated during initial screening; however, companies are increasingly looking to perform more intelligent screens earlier in the clone evaluation process. This will require the future development and integration of new high-throughput assays for product quality and manufacturability.

1.11.3.4

Assessing Clone Stability

An essential component of clone screening is verifying the stability of clones. Expression levels can frequently decline after clone isolation. This may be due to any one of several factors such as genetic rearrangement, loss of transgene copies, epigenetic transgene silencing, or the transient effects of clonal variation. Typical stability studies are performed by passaging the top-ranked clones for a period of up to 60 generations. Samples are taken to monitor growth characteristics (doubling time and cell morphology), specific productivity (expressed in pg/cell/day), and overall volumetric yield (titer). Product quality attributes may be tested, as well as mRNA expression and DNA copy number.12 The number of population doublings should cover the likely production life of the cells. Cell lines intended for perfusion processes may require longer study durations, for example, up to several months.2,12 Although stability studies are routinely performed over long periods and many cell generations, a rapid stability screen can be done in semi-solid media by imaging on the ClonePix FL, to easily and quantitatively assess population heterogeneity. It is assumed that heterogeneity reflects instability. Colonies can also be picked if subcloning is required. Similarly, flow cytometry can be used to assess population heterogeneity in order to predict instability.

1.11.4

Designer Cell Lines for Bioproduction

Development of designer “engineered” cell lines for bioproduction is a promising avenue to reduce the burden of clone selection and screening. An ideal cell line for bioproduction must be productive and robust. This is achieved through a combination of a high cell-specific productivity and a sustained high viable cell concentration. Additionally, the cells must grow rapidly to minimize cycle time, maintain product quality, and be able to withstand the increased stresses associated with mixing and gas transfer at production scale.

1.11.4.1

Regulatory Considerations

Beyond technical requirements, a cell line must also meet defined regulatory criteria to be suitable for producing human medicines. Cell line suitability is judged through a risk–benefit approach that accounts for, among other things, cell substrate history, the production and purification processes, quality control, intended patient population, and route of administration. Information on the original donor should include consideration of the species of origin, tissue type, age, gender, health history, and testing performed for adventitious agents. Also, relevant to assess risk is the method of isolation and the passaging history of the cell line with respect to raw materials and any possible exposure to adventitious agents. If genetically modified, the sequence of any DNA used must also be determined and verified to be stable beyond the expected end of production. Cloning of the cell line is desirable to reduce heterogeneity and because it represents a breakpoint from the early history of the cell line, where gaps in traceability or other risks may exist.

150

Cell Line Isolation and Design

In general, characteristics that improve cell line manufacturability, such as suspension growth and freedom from senescence, may conversely pose increased safety risks as they are often associated with neoplastic transformation. Thus, manufacturability must be balanced against the intended use of the biological product. For highly purified therapeutic proteins and antibodies, where adventitious agents and cellular residuals can be cleared to vanishingly low levels, continuous cell lines may often be used. For attenuated live virus vaccine substrates, which undergo minimal purification and are used for prophylaxis in healthy subjects (including young children), the requirements are necessarily more strict. In these cases, normal diploid cell strains such as MRC-5 or WI-38 are used, because they are nontransformed and have an established safety record. Important characteristics of normal diploid cells are finite lifespan, stable diploid karyotype, and freedom from tumorigenicity. Chemical inactivation of vaccines and less direct routes of administration, such as oral or mucosal routes, can incrementally reduce the risk from adventitious agents and cell residuals, though this assessment will be at the discretion of regulators and is likely to be case dependent. Such mitigating factors, along with more rigorous and better understood in vivo safety testing, is leading to ever greater willingness on the part of regulators to consider continuous cell lines not just for therapeutic protein production but also for vaccines.

1.11.4.2

Traditional Versus Designer Cell Lines

The traditional host cell lines for stable protein expression, Chinese hamster ovary (CHO) and NS0, came about in large part for historical reasons. In the case of CHO cells, dhfr knockout strains such as DG44 and DUXB11 were available in academia, which facilitated convenient gene amplification for improved productivity. NS0 cells, on the other hand, were commonly used as myeloma fusion partners for B cells in the creation of hybridomas for antibody production and, when recombinant DNA technology was introduced, were an obvious choice as hosts for stable transfection with antibody genes. More recently, “designer” cell lines have been specifically created to meet particular bioproduction needs. For example, PER.C6 cells (Crucell) were isolated from retinal tissue and transformed with the adenovirus E1A gene to complement propagation of E1-deleted adenovirus vectors in culture, eliminating overlapping sequences to reduce the risk of recombination generating replication-competent adenoviruses, as occurs in 293 cells. PER.C6 have since also been found useful for producing recombinant proteins due to their robust cell culture performance and high-level expression from single gene copies, which may be due in part to transactivation of the common Cytomegalovirus (CMV) promoter by E1A. EB66 cells (Vivalis) are another recent example of a designer cell line. These cells are of duck embryonic stem cell origin and grow continuously in suspension culture yet maintain a stable diploid karyotype and remain nontumorigenic. This makes them an ideal substrate for vaccines, combining both manufacturability and safety. As an avian host, the duck is preferred over chicken due to lack of infectious endogenous retroparticles or proviral sequences, yet the cells still propagate a broad selection of viruses normally grown in chicken eggs. Both PER.C6 and EB66 cells have very well-documented histories, consistent with their development specifically for bioproduction. Other examples of designer cell lines include those marketed by ProBiogen (www.probiogen.de) and Cevec Pharmaceuticals (www. cevec-pharmaceuticals.com).

1.11.4.3

The Contribution of Vector and Host Cell

For recombinant cell lines producing biologics, productivity is a function of both the host cell background and the transgene expression vector.13 Expression vectors are designed for maximum transcription through selection of efficient promoters, enhancers, and transcription termination sites. The stability of transcripts can be improved with suitable recognition sequences for RNA-binding proteins and through removal of cryptic splice sites and mRNA-destabilizing motifs. Other RNA signals such as the 50 and 30 UTR, polyA tail, introns, and splice sites can be appropriately designed to improve mRNA transcription, processing, and transport. Translation can be improved through codon optimization, eliminating interfering mRNA secondary structure, enhanced ribosome binding, and removal of translational pause signals. Vector design typically has less impact posttranslationally, except perhaps through selection of appropriate signal sequences. The host cell has arguably a greater role here as it provides all the machinery, energy, and building blocks for growth and recombinant protein expression, mediates stress resistance and robustness, and is responsible for correct posttranslational processing (e.g., glycosylation) and trafficking of the protein through the secretory pathway. The host cell can have different effects on posttranscriptional and posttranslational regulation depending on the expression of transacting factors due to genetic background.

1.11.4.4

Classical Cell Line Development

As described in Sections 1.11.2 and 1.11.3, classical cell line development is a numbers game, involving transfection, selection, and screening of large numbers of clones to find high producers with desired characteristics. Random integration of vector sequences and gene amplification can contribute to considerable variation in copy number and chromosomal context of the transgene, which in turn can affect expression level and susceptibility to epigenetic silencing. These steps can also disrupt or dysregulate endogenous genes as well as impacting genome-wide epigenetic imprinting. The random diversity generated as a consequence of stable transfection is exploited to select improved production lines but is nevertheless an uncertain and uncontrolled process that must be restarted for each new cell line development campaign. Various techniques have been introduced recently to improve predictability of vector expression and more rationally engineer host cells.

Cell Line Isolation and Design 1.11.4.5

151

Novel Vector Design Strategies

“Position’s effects,” or the variation in expression level with genomic integration site, can be moderated by including chromatinremodeling elements in the 50 and/or 30 UTR of the transgene (reviewed in Ref. 14). The stability of resulting clones is also improved as a result, through minimizing expression variability and epigenetic silencing. Examples of such elements include locus control regions, insulators, scaffold/matrix attachment regions, ubiquitous chromatin-opening elements, stabilizing and anti-repressor elements, and expression augmenting sequence elements. Although called by many names, all essentially isolate the expression cassette in a region of open chromatin and limit the spread of repressive epigenetic marks to the region of the transgene. The average expression level of clones in a pool may greatly improve through the use of chromatin remodeling, the number of clones that need to be screened to find a high producer may be significantly reduced, and expression stability may be more predictable. It has been reported, however, that the expression level of the best clone is not typically higher than achieved by traditional methods. Chromatin remodeling does not address the issue of insertional mutagenesis. Two other techniques, however, result in consistent expression with limited insertional effects: namely, targeted integration and episomal expression. Targeted integration involves first the random insertion of marker sequences in the host genome and then screening for integrations that have occurred in transcriptional hotspots. The marker sequences can then be swapped with the desired transgene by recombinase-mediated cassette exchange.15 A suitable selection marker or reporter allows the isolation of successful targeted integrations. Though this process is not completely efficient, expression consistency is much improved over traditional methods. Random integrations sometimes occur as well and must be eliminated by screening. A major limitation of targeted integration is that single copy insertions are required, due to the potential for incomplete exchange and unwanted chromosomal rearrangements with multiple copies. Fortunately, high specific productivities have been demonstrated with single copy insertions.16 Episomal expression is another novel strategy with the potential to improve cell line development. Stable expression from small episomally replicating plasmids has been technically proven but to our knowledge the specific productivities are still too low to make bioproduction cell lines. However, artificial chromosomes, which are much larger but still manageable for cell line development, have been shown to accept highly productive expression construct inserts, and be stably maintained in host cells.17 The production cells can be “cured” of the expression chromosome to restore the original host cell, and the chromosome can be isolated and transferred to other host cells if desired. A major strategic advantage of exchangeability and minimal impact on the host cell is the ability to perform host cell engineering in production cell lines. Thus, productivity and product quality can be screened directly, and the production line reverted to a clean host cell. Equally, an already successful production line could be used as the basis for future production lines, minimizing repeated development time. With increasing emphasis on platform processes, production cell characteristics could be effectively controlled without introducing variation inherent in classical cell line development.

1.11.4.6

Host Cell Engineering: The New Frontier

There have been a variety of targets explored for constructive metabolic engineering in mammalian cells.18 In cell proliferation control, cyclin-dependent kinase inhibitors have been expressed in inducible fashion to promote a biphasic production process involving distinct growth and production stages. In nutrient metabolism, lactose dehydrogenase has been knocked down and pyruvate carboxylase overexpressed to divert carbon flux away from lactate and toward the tricarboxylic cycle. In apoptosis control, anti-apoptotic proteins such as Bcl-2 and Bcl-XL or caspase inhibitors such as XIAP, CrmA, and Aven have been overexpressed, or caspases directly knocked down with antisense techniques. In glycosylation engineering, various approaches have been taken to produce more human-like glycosylation in typical rodent lines, such as knockdown of immunogenic N-glycolylneuraminic acid (NGNA), expression of human alpha-2,6-sialyltransferase, overexpression of GnTIII for bisecting GlcNAc, and knockdown of fucosyltransferase. In engineering higher productivity, ER chaperones such as GRP78, PDI, calnexin, and calreticulin have been overexpressed, along with the transcription factor XBP-1 to promote expansion of the rough ER, and more recently secretory pathway proteins Sly1 and Munc18c. Despite the many targets and general enthusiasm for the concept of host cell engineering, industry uptake has been slow which may relate in part to the lack of stability seen with overexpression or knockdown techniques. Another route is inverse metabolic engineering, where genome-wide analysis tools (transcriptomics, proteomics, etc.) are used to generate lists of differentially expressed genes or proteins. Theoretically, these could be used as targets for cell engineering. Ultimately, however, while many differential expression lists have been generated, few studies have followed through to target validation. One nice example is the identification and validation of siat7e and lama4 in HeLa and MDCK cells as modulators of attachment-dependent cell growth.19,20 In CHO, the lack of a sequenced genome has limited the power of these methods, though completion of the CHO-K1 sequence was announced in early 2011. Even with the genome sequence, the challenge of expression stability or complete knockdown remains, along with other technical challenges such as overcoming pathway redundancies and off-target effects. Clearly, a global systems-based approach will be needed. New techniques for gene-editing enable probably the most powerful means to effect host cell engineering strategies: total knockout of genes. Once a gene is deleted, the effect is complete and irreversible, and not subject to the instability of knockdown or over expression. This was demonstrated with fucosyltransferase knockouts that resulted in proteins that were totally afucosylated, compared to two other cell lines isolated by conventional means which only showed partial response.21 Another key example is the knockout of bax and bak genes in an industrially relevant CHO cell line, which essentially eliminated the ability of these cells to

152

Cell Line Isolation and Design

undergo apoptosis.22 This study was conducted using zinc finger nucleases to target the deletions with very high efficiency (over 10%). Zinc fingers and meganuclease technology from Cellectis are showing considerable promise for complex genetic engineering in mammalian cells. This should prove to be the beginning of a fruitful new era in mammalian cell engineering, as tools become more available, cost effective, and technically feasible; thus, bringing mammalian cells more inline with the ease of genetic modification achieved with prokaryotes and lower eukaryotes. Cell line engineering will no doubt continue to be an area of intense study. This article by no means covers every example of cell engineering and further details can be found in recent review articles such as Ref. 23.

1.11.5

Future Perspectives and Conclusions

1.11.5.1

High ProductivitydAre We There Yet?

Over the past 20 years, there has been exponential growth in the amount of therapeutic product produced on a per cell basis, and until recently there has been an almost exclusive focus on titer improvements. For many current therapeutics, titers of 2–5 g/L are sufficient and there is no longer a serious under capacity in worldwide mammalian cell culture capabilities. The current focus in the biopharmaceutical industry is on developing streamlined processes and identifying clones that exhibit predictable expression levels and desirable product quality attributes. Further improvements in intelligent clone isolation and analysis will be essential for accelerating the time to market for biologic therapies of the future.

1.11.5.2

Future of Bioproduction

Clone selection and cell line engineering will remain key innovation drivers for biopharmaceutical research and development. Perhaps, the biggest change likely comes as the result of genome-scale modeling and systems biology, which will be used to integrate the recent explosion in omics data into meaningful industrially relevant outcomes. This has already been realized in some prokaryotic-based production platforms in which efficiency has been improved by several orders of magnitude. Our increased understanding and ability to modify and control complex systems should allow for the development of the next generation of therapies.

1.11.5.3

Summary and Conclusion

Clone selection and analysis remains a rate-limiting step during biologic drug development. Crucially, the selection of the final manufacturing clone also carries far-reaching impacts upon all downstream aspects of the drug-development process. However, novel and improved engineering approaches, coupled with a number of emerging high-throughput techniques, are allowing for intelligent selection of quality clones with significant time and cost savings.

See also: 1.05 Structure and Biosynthesis of Glycoprotein Carbohydrates; 1.19 Modes of Culture/Animal Cells; 1.26 Cell Transfection; 1.29 Engineering Protein Folding and Secretion in Eukaryotic Cell Factories; 1.33 Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds; 1.40 Flow Cytometry.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.

Walsh, G. Biopharmaceutical Benchmarks 2010. Nat. Biotechnol. 2010, 28, 917–924. Wurm, F. M. Production of Recombinant Protein Therapeutics in Cultivated Mammalian Cells. Nat. Biotechnol. 2004, 22, 1393–1398. Browne, S. M.; Al-Rubeai, M. Selection Methods for High-Producing Mammalian Cell Lines. Trends Biotechnol. 2007, 25, 425–432. Carroll, S.; Al-Rubeai, M. The Selection of High-Producing Cell Lines Using Flow Cytometry and Cell Sorting. Exp. Opin. Biol. Ther. 2004, 4, 1821–1829. Mattanovich, D.; Borth, N. Applications of Cell Sorting in Biotechnology. Microb. Cell Fact. 2006, 5, 12. Pilbrough, W.; Munro, T. P.; Gray, P. Intraclonal Protein Expression Heterogeneity in Recombinant CHO Cells. PLoS One 2009, 4, e8432. Sleiman, R. J.; Gray, P. P.; Sunstrom, N. A. Splitting the Difference: When Near Enough Is Not Close Enough. Cytom. 2006, 69A, 426. Brezinsky, S. C.; Chiang, G. G.; Szilvasi, A.; et al. A Simple Method for Enriching Populations of Transfected CHO Cells for Cells of Higher Specific Productivity. J. Immunol. Methods 2003, 277, 141–155. De Jesus, M. J.; Girard, P.; Bourgeois, M.; et al. TubeSpin Satellites: A Fast Track Approach for Process Development with Animal Cells Using Shaking Technology. Biochem. Eng. J. 2004, 17, 217–223. Lindgren, K.; Salmen, A.; Lundgren, M.; et al. Automation of Cell Line Development. Cytotechnology 2009, 59, 110. Amanullah, A.; Otero, J. M.; Mikola, M.; et al. Novel Micro-Bioreactor High Throughput Technology for Cell Culture Process Development: Reproducibility and Scalability Assessment of Fed-Batch CHO Cultures. Biotechnol. Bioeng. 2010, 106, 57–67. Barnes, L. M.; Bentley, C. M.; Moy, N.; Dickson, A. J. Molecular Analysis of Successful Cell Line Selection in Transfected GS-NS0 Myeloma Cells. Biotechnol. Bioeng. 2007, 96, 337–348. Kim, H.; Laudemann, J.; Stevens, J.; Wu, M. Expression Vector Engineering for Recombinant Protein Production. Cell Line Dev. 2009, 6, 97–108. Kwaks, T. H.; Otte, A. P. Employing Epigenetics to Augment the Expression of Therapeutic Proteins in Mammalian Cells. Trends Biotechnol. 2006, 24, 137–142. Oumard, A.; Qiao, J.; Jostock, T.; et al. Recommended Method for Chromosome Exploitation: RMCE-Based Cassette-Exchange Systems in Animal Cell Biotechnology. Cytotechnology 2006, 50, 93–108. Hartman, T. E.; Sar, N.; Genereux, K.; et al. Derivation and Characterization of Cholesterol-independent Non-GS NS0 Cell Lines for Production of Recombinant Antibodies. Biotechnol. Bioeng. 2007, 96, 294–306.

Cell Line Isolation and Design

153

17. Kennard, M. L.; Goosney, D. L.; Monteith, D.; et al. Auditioning of CHO Host Cell Lines Using the Artificial Chromosome Expression (ACE) Technology. Biotechnol. Bioeng. 2009, 104, 526–539. 18. Lim, Y.; Wong, N. S.; Lee, Y. Y.; et al. Engineering Mammalian Cells in Bioprocessing – Current Achievements and Future Perspectives. Biotechnol. Appl. Biochem. 2010, 55, 175–189. 19. Chu, C.; Lugovtsev, V.; Golding, H.; et al. Conversion of MDCK cell line to suspension culture by transfecting with human siat7e gene and its application for influenza virus production. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 14802–14807. 20. Jaluria, P.; Betenbaugh, M.; Konstantopoulos, K.; Shiloach, J. Enhancement of Cell Proliferation in Various Mammalian Cell Lines by Gene Insertion of a Cyclin-Dependent Kinase Homolog. BMC Biotechnol. 2007, 7, 71. 21. Kanda, Y.; Yamane-Ohnuki, N.; Sakai, N.; et al. Comparison of Cell Lines for Stable Production of Fucose-Negative Antibodies with Enhanced ADCC. Biotechnol. Bioeng. 2006, 94, 680–688. 22. Cost, G. J.; Freyvert, Y.; Vafiadis, A.; et al. BAK and BAX Deletion Using Zinc-Finger Nucleases Yields Apoptosis-Resistant CHO Cells. Biotechnol. Bioeng. 2010, 105, 330–340. 23. Kramer, O.; Klausing, S.; Noll, T. Methods in Mammalian Cell Line Engineering: From Random Mutagenesis to Sequence-Specific Approaches. Appl. Microbiol. Biotechnol. 2010, 88, 425–436.

1.12

Cell Preservation Technology

John G Baust and William L Corwin, Binghamton University, Binghamton, NY, United States John M Baust, Binghamton University, Binghamton, NY, United States; and CPSI Biotech, Inc., NY, United States © 2011 Elsevier B.V. All rights reserved. This is a reprint of J.G. Baust, W.L. Corwin, J.M. Baust, 1.14 - Cell Preservation Technology, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 179-190.

1.12.1 1.12.2 1.12.3 1.12.4 1.12.5 1.12.5.1 1.12.6 1.12.7 1.12.8 1.12.9 References Further Reading

Introduction Hypothermic Storage Hypothermic Continuum Cryopreservation Modes of Cell Death Apoptosis Cell Death Continuum Preservation-Induced Cell Death Targeted Control of Molecular-Based Death Concluding Thoughts

154 155 155 156 157 158 159 159 160 161 162 162

Glossary Apoptosis A highly conserved process of programmed cell death that involves specific biochemical changes that lead to cellular changes that include blebbing, loss of cell membrane asymmetry and attachment, cell shrinkage, nuclear fragmentation, chromatin condensation, and chromosomal DNA fragmentation. Cell death continuum The concept that apoptosis and necrosis are not separate entities but, instead, extreme opposite ends of an array of ways a cell can die. Cryoprotective agents (CPAs) Compounds that protect cells against the formation of damaging intracellular ice during the freezing process. Delayed-onset cell death The occurrence of continued cell death events (i.e., apoptosis and necrosis) several hours after the initial cold exposure. Glass transition temperature (Tg) The temperature at which the transition from liquid to amorphous or glassy solid state occurs. Mitochondrial permeability transition pore (mPTP) A protein pore formed in the membrane of the mitochondria due to some pathology or molecularly regulated event that can trigger a cell death response. Secondary necrosis The process through which a cell commits to the activation of an apoptotic cell death response but, because of adenosine triphosphate (ATP) depletion, is shunted to a necrotic mode of death. Supercooling The process of lowering the temperature of liquid below its freezing point without it becoming a solid.

1.12.1

Introduction

Biopreservation is a diverse scientific specialty, which integrates the efforts of many different fields, such as cryobiology, bioengineering, computer sciences, molecular biology, and cellular biology, to name a few.1 As a field, biopreservation aims to effectively maximize the methods associated with the preservation of cells, tissues, and organs and their subsequent return to prestorage functionality.2 Currently, this field is growing at a rapid pace, as advances in other related areas of interest, such as cell therapy, stem-cell research, personalized medicine, cell banking, and cancer research, drive the need for optimized storage protocols for the biologics utilized in these fields.3 However, even with the advances that have been made during this rapid expansion, there still remain significant problems with the current preservation techniques. These issues include, but are not limited to, loss of function, utilization of nonhuman components in storage solutions, and activation of cellular stress pathways, which can result in changes in gene expression and protein levels.4 Historically, efforts to preserve biologics, such as cells and organs, have focused primarily on the use of low temperatures to facilitate the need for cells “on demand.” Methods of cold storage are typically divided into two categories: hypothermic storage and cryopreservation. Hypothermic storage is the use of temperatures ranging from below normothermic temperatures for mammals

154

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00014-8

Cell Preservation Technology

155

(32–37
C) to usually no lower than 0
C to store biologics for a certain time period. Generally, this type of storage regime is utilized in the field of transplantation medicine. Cryopreservation is use of ultralow temperatures at or below 80
C and, most commonly, below 140
C, as this is below the reported nominal glass transition temperature (Tg) of water (an important point which will be addressed further in later sections). While these storage regimes have proved to be useful, they are increasingly revealing limitations as advances in other fields continue to change the demands on protocol outcome.3,4 As such, the focus of these storage regimes has begun to change with identification of molecular responses to cold (i.e., apoptosis) and the need to mitigate this change by incorporating a molecular-based approach into the development of next-generation cryopreservation methodologies.

1.12.2

Hypothermic Storage

Hypothermic storage is the use of temperatures below normothermic conditions to preserve biological material. While the protective effects of cold have been documented for centuries and utilized by mankind for countless years, a true understanding of the effects of cold exposure on biologics has only begun within the last 60 years. This modern era of low-temperature cell preservation began in the 1950s,3,5 following important discoveries by Carrel. His investigations concerned normothermic perfusion of organs prior to transplantation, in which he described the fundamental characteristics of a perfusion medium leading to the advancement of the cold perfusion technique. This modern period of cold preservation saw the development of both hypothermic and cryopreservation techniques as effective means for increasing storage intervals while limiting the negative effects associated with the removal of cells and tissues from their in vivo environments, such as ischemia and hypoxia.1 The principle behind the use of cold as a preservation agent lies in its ability to slow cellular functions in a manner which can then be reversed following return to normothermic conditions.2 This reduction in function allows for the cells to be held in a suspended animation-like state that reduces metabolic needs, by slowing cell death that may result from the accumulated effects of related stress factors.3,4,6 While cryopreservation utilizes ultralow temperatures that can effectively bring a cell’s metabolism to a halt and, thus, allows for an indefinite storage period, the current state of this technique proves effective only for single-cell suspensions or simple tissues.1 The cryopreservation of complex tissue or whole organs results in the accumulation of severe cold injury, or cryoinjury, resulting in cell death and, in turn, loss of higher-order functions.3 Thus, hypothermic storage currently serves as the most effective paradigm for preservation of these complex biologics prior to their utilization.1,5 Typically, for whole organs, the process involves an initial step of cold perfusion, where a cold solution is flushed through the vasculature either just prior to or just after its removal from the body. This solution is left in and around the organ as it is then stored hypothermically to allow for its transport prior to utilization. It then undergoes a warm reperfusion process to flush the hypothermic solution from the organ and return it to normothermic temperature prior to transplantation. While this method has proven to be far superior to warm perfusion and storage, it is still a tradeoff of more profound, faster acting stresses for less dramatic but, nonetheless, significant stresses associated with cold exposure. To this end, there have been numerous investigations into the metabolic, biochemical, and physical characteristics of cells both in their normal, normothermic environment and also how these factors change in response to exposure to cold.3–6 Many details of normal cellular function have become common knowledge within cell biology, such as how cells regulate the flow of ions to create gradients and electron potential across membranes to drive reactions or how specific biochemical processes such as the citric acid cycle work to create energy for the cell. In understanding how a cell functions normally, it then becomes possible to understand how cold exposure causes changes to these aspects of a cell and where improvements can be made. This initiative of intelligent design based on the progression of our understanding is best exemplified by the advancements in solution design. Through the application of knowledge gained about cell systems and their response to cold stresses, the first intracellular-like cold perfusion solution, University of Wisconsin solution (ViaSpan), was developed and has since become the gold-standard preservation solution for many organ systems.7 While this approach has continued into recent years with more success,1,5 there still remains a significant void that solution design alone cannot bridge. With advancements in our understanding of the molecular response of cells to cold, a more specifically targeted approach can be achieved. This targeted approach has already begun with reports showing the benefits of adding components to hypothermic solutions to control molecular responses to cold.1 Moving into the future, this approach will need to incorporate what is already known with what is discovered to achieve a protocol that optimizes all steps of the preservation process, before, during, and after, in a cell system-specific manner to achieve methods better suited for today’s demands on the biopreservation field.

1.12.3

Hypothermic Continuum

There are four intensities of hypothermia as referenced in medical literature, generally referred to as mild (32–35
C), moderate (27–32
C), deep or profound (10–27
C), and ultraprofound ( green , stoichiometric ratio (CO2 evolved per mass unit of substrate consumed); blue-green curve, stoichiometric NH3/O2 (uptake of ammonium per unit consumed oxygen). From Erickson, L. E.; Minkevich, I. G.; Eroshin, V. K. Application of mass and energy balance regularities in fermentation. Biotechnol. Bioeng. 2000, 67(6), 748–774.

The yield on conserved substrates varies mainly because of alterations in the biomass chemical composition expressed by the parameter ss, the intracellular content of a deficient element or cell quota. For most known cases, the content ss increases in parallel with growth acceleration (Fig. 16). As yield and cell quota are inversely related to each other (Eq. 45), Y values decrease with growth rate. The physiological mechanisms of this variation are as follows. The intensive growth requires higher internal concentrations of some conserved limiting substrates that preserve their chemical identity after uptake (Kþ, Mg2þ, and vitamins). Other conserved substrates (sources of P, N, S, etc.) are incorporated into macromolecular cell constituents (mainly nucleic acids and proteins) whose intracellular content also should be kept high at a high growth rate. Both types of changes in cellular composition are manifested as s increase in parallel to growth acceleration, and both require additional maintenance energy. The observed m-dependent variation in s is therefore a compromise between biosynthetic requirements and energy conservation. However, it would be erroneous to interpret m as a factor affecting the s level. In fact, both growth intensity (m) and the intracellular content of biogenic elements (s) are controlled by common independent variables, the most essential being the limiting substrate concentration in the medium, s. Let us assume that, under steady-state conditions, m and s are hyperbolic functions of s: m¼

q 1 Qs ¼ ; s s Ks þ s

s ¼ s0 þ

ðsm
s0 Þs Ks þ s

(46)

where m is the specific growth rate and q is the specific substrate uptake rate; s0 and sm are, respectively, the lower and upper limits of s variation; the low limit s / s0 is attained when s / 0 and the upper limit s / sm when s / N. By excluding s from both these equations and under the assumption Ks z Ks, we arrive at the following relationship between Y, s, and m ¼ D: s¼

s0 ; 1
ð1
s0 =sm Þm=mm

Y ¼ Ym
ðYm
Y0 Þ

m ; mm

(47)

where Ym and Y0 are, respectively, the upper and lower limits of yield variation (Ym ¼ 1/s0 and Y0 ¼ 1/sm). As we can see, the linear relationship between Y and m is normally observed in chemostat culture (Fig. 17). Eqs. (46) and (47) can also be obtained from the SCM by assuming that the intracellular content of a conserved element (N, P, K, Fe, etc.) is proportional to the r value. It is quite justifiable because all listed elements are part of ribosomes or enzymes (polypeptides and/or cofactors) supporting intensive cellular growth. SCM provides reasonable agreement with steady-state and transient data and resolves one of the paradoxical observations shown in Fig. 18: the phototrophic culture consumed all available phosphate the first day of batch growth; then, cell growth did not cease but rather accelerated with a following gradual decline. By the end of the

Microbial Growth Dynamics

263

Figure 17 Variation of stoichiometric parameters of cell yield Y and cell quota s as a function of dilution growth rate in the chemostat culture of Chlorella vulgaris limited by nitrogen source. Black circles: yield YN (left axis). Grey circles: cell quota [Greek sigma] (right axis). The curves are calculated using Eqs. (46) and (47).

Cell biomass Phosphate Phosphate (mg P/L)

Figure 18 Batch growth of Selenastrum capricornutum limited by phosphate. The curves are calculated using the synthetic chemostat model, supplemented with expressions for conserved substrates Eqs. (45)–(47). Note the lack of correlation between cell growth and uptake of the limiting substrate. Experimental data are redrawn from Nyholm, N. Kinetics of phosphate limited algal growth. Biotechnol. Bioeng. 1977, 19(3), 467–492.

5-day experiment, cell density was increased by 100 times, mostly during a period when extracellular P was absent. The observed remarkable decoupling between growth and P uptake is explained by the fact that cells are able to redistribute the deficient conserved element (here P; but the same pattern can be seen for N, K, Fe, etc.) between mother and daughter cells. Even in the absence of an extracellular P source, cells are able to grow at the expense of intracellular P reserves (mostly RNA, while polyphosphate remained low) with a simultaneous decline in the total cellular P.

1.18.3.3

Effect of Environmental Factors

Here, we will analyze only a limited number of the most important and general environmental factors, such as pH, temperature, and effects of tonicity/free water. Effects of pH. It is not easy to get good data on growth rate versus pH because of numerous side effects associated with the use of buffer, drift of pH during batch experiments, changes in the availability and toxicity of metals, etc. Probably, the best technique for this purpose is the pH-auxostat type of continuous culture or exponential phase of batch culture at a series of instrumentally controlled pH values. The chemostat data at a single D and a series of different pH levels are ambiguous because cells respond to at least two factors: pH and substrate concentration. When accurately measured, the plot of m or q versus pH is always a symmetric

264

Microbial Growth Dynamics

bell-shaped curve. In enzymology, this kind of pH effect is explained by the ionization of several functional groups within the active center of an enzyme. These groups should have a pKa of 5–9 (imidazole group of histidine and thiol sulfhydryl group of cysteine) to K1 K2 change their ionization state. If we assume that the enzyme is a dibasic acid ðH2 E4H2 E
4E2
Þ and only the singly ionized
complex, HE , is active, then, the rate of reaction (v) depends on pH as follows:

K1 Hþ v ¼ Vm þ 2 (48) ½H  þ K1 ð½Hþ  þ K2 Þ This equation produces a symmetrical bell-shaped curve with a maximum at (pK1 þ pK2)/2 (Fig. 19). Undoubtedly, the molecular basis of pH effects on microbial cells should be wider and reflect many other effects, such as homeostatic mechanisms maintaining constant intracellular pH. However, for practical purposes, Eq. (46) works reasonably well. Effects of temperature. In chemical kinetics, the dependence of the reaction rate on temperature is explained by the transition-state theory developed by Eyring as early as in 1930–35. It is based on the use of thermodynamics and principles of quantum mechanics. The reaction proceeds through a continuum of energy states and must surpass the state of maximum energy, when a transient activated complex is formed. Then, the dependence of the reaction rate constant k on the absolute temperature, T, is expressed as follows: d ln k DH þ RT ¼ dT RT 2

(49)

where R is the gas constant and DH* is the enthalpy of activated complex formation. The classic Arrhenius equation may be obtained from Eq. (47) under a simplified condition DH* þ RT z DH ¼ Ea (where Ea is the activation energy). Most often, the Arrhenius equation is used in its integrated form: ln k ¼ ln A


Ea RT

k ¼ Ae
Ea =RT

or

(50)

where A is the integration constant, interpreted as the frequency of collisions of reacting molecules. There are also several purely empirical expressions relating rates to the Celsius temperature, Tc, for example, the exponential formula: ln k ¼ ln A þ aTC

or

k ¼ AeaTC

(51)

Here, the parameter a is a constant related to the widely used temperature coefficient Q10 ¼ exp(10a). All presented mathematical expressions predict exponential or almost exponential increases of chemical reaction rates with temperature. However, all bioprocesses, including microbial growth, deviate from this relationship at high temperature because of the thermal denaturation of enzymes. Assume that denaturation is reversible, with equilibrium constant KT ¼ [E0 ]/[E], where E represents active enzyme molecules and E0 is the inactive ones. Then, the combination of Eq. (50) with the van‘t Hoff relationship for KT (
RT ln KT ¼ DGo ¼ DHo
TDSo) results in: v¼

Ae
Ea =RT 1þ

Ds 0
DH 0

eR

RTC

¼

Ae
Ea =RT   B 1
CT C 1þe

(52)

where DG , DS , and DH are the standard Gibbs free energy, enthalpy, and entropy of the denaturation reaction, respectively, which can be simplified into aggregated constants B and C. A similar temperature-denaturing term can be added to the exponential Eq. (51) with Celsius temperature as an independent variable: o

o

o

AeaTC 

v¼ 1þe

Figure 19



C B 1
273þT

Effect of pH on microbial activity as calculated using Eq. (48).

C

(53)

Microbial Growth Dynamics

265

Figure 20 The effect of temperature on microbial activity in a methanogenic bioreactor (Panikov, unpublished data). Experimental (Experim) data were obtained from a fluidized bed reactor containing a microbial community enriched from a Sphagnum peat bog. The gas mixture N2:CO2:H2 ¼ 90:5:5 was continuously run through the reactor with off-gas mass spectroscopic analysis for methane. After stabilizing at 3 C for several days, the reactor was warmed at the rate of 1 grad h
1 and then 10-min averaged data points were plotted versus respective temperature. Experimental data points on methane formation rates were fitted to the modified Arrhenius equation (red line, Eq. 52) and modified exponential equation (blue line, Eq. 53).

Within a biologically meaningful range of temperatures (0–60 C), the empirical Eq. (53) does not differ significantly from the more accurate Eq. (50) and is more stable in solving the inverse problem when it is fitted to experimental data. Fig. 20 shows an example of the temperature effect on microbial activity described as a curve peaked at 43 C. In contrast to enzymatic and chemical reactions, microorganisms and other living cells demonstrate significant hysteresis in their temperature response (and probably in response to many other environmental factors). If temperature is subject to cyclic changes (Fig. 21), then the rates of microbial reactions (growth, respiration, protein synthesis, etc.) are different, being dependent on the way temperature varied in the past; say, whether temperature was increasing or decreasing over time. To explain the hysteresis phenomenon, we should remember that adaptation to any temperature extreme requires the synthesis of specialized defending proteins, for example, heat-shock proteins preserve cell integrity under hot conditions, while adaptation to cold is controlled by desaturating enzymes and chaperones.109 In terms of the SCM, the adaptation process should be driven by the difference between optimal and current temperature (DT) and instant growth rate in a way analogous to Eq. (36). The transition is expected to take about 1/m time units. We see from the isothermal segment at 30 C (Fig. 21, top) that adjustment to a warm regime indeed takes about 6 h, which agrees well with the growth intensity of methanogenic microorganisms. Therefore, biological inertia is the most probable reason for observed hysteresis.

1.18.3.4

Self-inhibition of Growth: Metabolic Acidification and Alkalization

In biotechnology, the formation of products may be the most important part of the process of microbial cultivation. Here, we discuss to what degree metabolites produced by growing microorganisms affect their own growth dynamics. One of the greatest challenges in microbial growth theory was to predict a spontaneous formation of by-products and overcome their possible inhibitory effects to reach maximal productivity of microbial culture. Let us start from metabolic acidity. It often prevents smooth cultivation and fast production of cell mass or other target products. Common opinion is that acidity results from the accumulation of organic acids, and, sometimes, acidification is explained as an immediate result of the generation of a transmembrane proton gradient coupling respiration to ATP production. This is incorrect. First, acidification is very frequent but not the unique result of growth. In fact, sometimes, culture liquid remains neutral, or the pH may even rise. It depends on the medium composition and what particular compounds serve as the energy, C, and N sources (account for more than 90% of the total uptake). Second, acidification is unlikely to be caused by pumping Hþ ions outside a membrane because Hþ (or “hydronium cation,” H3Oþ) binds to OH
to form water. The change in pH in a growing microbial culture can be explained and predicted by analysis of the global ion balance aimed at quantifying all ion-exchange reactions across the membrane of growing cells. The electroneutrality conditions require that the sum of all ions be zero: Naþ þ Kþ
Cl

NO3
þ Hþ
OH
. ¼ 0


(54)

Here, all cations are added with a “þ” sign and anions with a “
” sign. The OH ion can be expressed through the ion product of water, K0 w ¼ [Hþ][OH
], while all strong ions can be grouped as SID, the strong ions difference (SID ¼ Naþ þ Kþ
Cl
– NO3
), then:





(540) SID þ Hþ
Kw0 Hþ ¼ 0

266

Microbial Growth Dynamics

Figure 21 The effect of temperature cycling on microbial activity. Incubation temperature was controlled by a programmable circulating thermostat between þ1 C and 30 C as shown by the red line. Methanogenic activity (the same community as shown in Fig. 20) was recorded as instant methane production rate during five cycles. (Top) Pooled average data for all measured temperature cycles; the methanogenic activity (q) was normalized to the maximal activity Q observed within each individual cycle. (Bottom) The plot of metabolic rate versus temperature demonstrates hysteresis; red arrows indicate direction of the temperature change.

The simplified Eq. (54)0 is already sufficient to make the statement that the pH of the medium is not controlled immediately by H fluxes because of extremely low K0 w value; the components of SID have a much higher impact on pH and even a small difference between fluxes of strong cations and anions can be of great significance. For example, if the membrane ATPase transfers Hþ from inside the cell to the outside, the pH will not change, unless movement of other ions is involved. Extracellular alkalization often is caused by Kþ/NH4þ efflux or Cl
and SO42
influx.110 Acidification is most often related to Kþ and NH4þ uptake. Accurate calculation of the global ion balance accompanying chemoorganotrophic growth111,112 led to the conclusion that the main source of metabolic acidity is uptake of ammonium sulfate or chloride. Irrespective of biophysical mechanisms of its transport through the membrane, the net uptake of NH4þ is much higher compared with the strong anion SO42
. Thus, SID declines and pH drops. To further confirm this conclusion, Fig. 22 shows the rate of base titration (equivalent to the rate of metabolic acidification), which closely correlates with other processes associated with growth and reflects the process of NH4þ uptake. For example, aerobic cultivation of yeasts on 10 g L
1 glucose (assuming Y ¼ 0.5 and sN ¼ 0.1) requires consumption of 18 mmol NH4þ, and it takes usually 15–20 mmol of KOH to neutralize growth-associated acidification. The replacement of (NH4)2SO4 with urea prevents acidification, while the use of KNO3 likely results in alkalization. The opposite process of medium alkalization is observed when microorganisms are grown on Na or K salts of organic acids (e.g., sodium pyruvate or acetate) as a single C source; in this case, the organic moiety of the C substrate is catabolized to CO2 with concomitant release of strong cations into the medium (Naþ and Kþ). þ

1.18.3.5

Growth and Formation of Intermediate Product: Simulation of the Crabtree Effect

A comprehensive survey of product formation is available in Pirt.1 In this section, we give an example of quantitative descriptions of growth dynamics accompanied by the formation of intermediary products such as organic acids and alcohols, which could be toxic. Such products are excreted under aerobic conditions and should be distinguished from fermentation products, which are the end products of microbial metabolism under anoxic conditions. Well-known examples of such intermediates of aerobic growth are acetate (E. coli and other enterobacteria) and ethanol in S. cerevisiae, among others. It is called the Crabtree effect, named after the English biochemist Herbert Grace Crabtree. In the most general sense, the Crabtree effect describes the phenomenon of metabolic competition between fermentation (substrate-level phosphorylation) and aerobic metabolism of glucose via the TCA cycle

Microbial Growth Dynamics

267

20

Rates (mmol–1h or ml–1h)

KOH 15

H2 CO2

10

5

0

0

5

10

15

20

25

30

35

40

Time (h) 2.5

10

2 Cellulose Cells

6

1.5

4

1

2

0.5

0

0

5

10

15

20

25

30

35

40

Cell mass (g l–1)

Residual cellulose (g l–1)

8

0

Time (h)

Figure 22 Example of fermentation dynamics demonstrating correlation of base titration rate with growth dynamics. Batch culture of Clostridium thermocellum was grown on mineral media with cellulose, 10 g L
1.116 Note the synchrony between KOH titration rates and gas production and immediate cessation of acid production after the cell mass stopped increasing.

and the electron transport chain observed in the presence of high, external glucose concentrations. Thus, the Crabtree effect is similar to the well-known Pasteur effect, the only difference being that fermentation is induced not by a lack of oxygen but by an excess of glucose. Usually, the Crabtree effect is observed in a glucose-limited chemostat culture run at a series of dilution rates, D. Step-by-step increases in D eventually bring the culture to the state at which glucose is converted to 2C products (acetate, acetaldehyde, and ethanol), while yield and respiratory activity decline. The model simulating the Crabtree effects assumes the following flux flow: Glucose (S) is utilized in two well-known steps: (1) glycolytic conversion of S to pyruvate (skipped in our scheme) and then to a fermentation product M (acetate or ethanol and CO2) and (2) M is oxidized to CO2. Growth of cells (X) is supported at both steps and, usually, glycolytic growth is faster, although less efficient. The main challenge is to explain and correctly predict the switch from glycolytic to oxidative catabolism. The flow diagram is translated into the set of ODEs: Glucose

Acetate= ethanol

ds ¼ Dðsr
sÞ
qx; dt

dm ¼ lqx
qm x
Dm; dt Cell mass

CO2

q¼Q

s Ks ð1 þ amÞ þ s

qm ¼ Qm

m 1  Km þ m 1 þ bsg

dx ¼ Y1 qx þ Y2 qm x
Dx dt

dp ¼ ð1
l
YÞ1 qx þ ð1
YÞ2 qm x
Dp dt

where l is a metabolic “splitter”; a, b, and g are inhibition constants; and the rest of the symbols are common.

(55)

(56)

(57)

(58)

268

Microbial Growth Dynamics

Figure 23 Demonstration and simulation of the Crabtree effect. (Left) Batch culture of microorganisms grown on glucose (S) fermented into 2C-compound M (acetate, ethanol, etc.) under aerobic conditions. Depletion of glucose releases repression of oxidative catabolism of the intermediate product; thus, a diauxic cell growth pattern occurs. (Right) Behavior of the same organism in chemostat culture. At low dilution rates, both glucose and fermentation products are consumed; at high dilution rates, the oxidative pathway is suppressed by the elevated concentration of residual glucose. Green: glucose concentration. Brown: cell biomass. Orange: intermediate product representing 2C-compound like, most frequently acetate or ethanol. Both sets were calculated from the same model (Eqs. 55–58).

The minimal kinetic model assumes that (1) excess of glucose inhibits the uptake of M and (2) the fermentation product M competitively inhibits the uptake of S. Results of the simulation agree with observations (Fig. 23). To simulate an abrupt transition from an oxidative to a fermentative pattern in chemostat culture, we need to assume a nonlinear glucose inhibition pattern: lack of effect at low residual glucose with the following steep acceleration of inhibition at a glucose concentration of about 10 mM. As a reminder (see Section 1.18.2.5), the same effect is possible to simulate by genome-scale models based on the principle of optimality.

1.18.3.6

Growth and Production of Antibiotics and Other Secondary Metabolites

Secondary metabolites are defined as those organic compounds that are “not directly involved in the normal growth, development, or reproduction of organisms” (Wikipedia). In growth models, the rate of secondary metabolite production, including antibiotics, is usually expressed as: dp ¼ Yp=x x dt

(59)

where Yp/x is no longer constant but is a nonlinear function of internal and environmental factors. As a general trend, expression of the secondary metabolites occurs when intensive growth stops because of partial depletion of one or more nutrient substrates (Fig. 24). The first phase is called the trophophase (generating cell mass), and the secondd the idiophase. The described pattern agrees with ecological interpretation: in natural habitats, synthesis of antibiotics is beneficial (suppressing competitors) only under chronic starvation observed in an overcrowded community, while under a plentiful supply of nutrients this process is meaningless. The SCM reproduces the pattern (Fig. 24) by setting the quotient Yp/x in Eq. (59) as a product (1
r)  (1 þ as), the first term playing the role of the slow controlling factor analogous to translational control of enzyme synthesis (r declines slowly), and the second term is a kind of instant regulation via substrate inhibition. In biotechnological practice, the production of secondary metabolites should be realized as a high biosynthetic activity of half-starved cells. It can be achieved in chemostat with retention of cell biomass: the total cell mass in the bioreactor continuously increases, bringing the population to a state of chronic starvation without intoxication by products or decay of biosynthetic machinery.

1.18.3.7

Population Dynamics: Mutations, Autoselection, and Plasmid Transfer

In this section, we turn our attention to the possible genetic heterogeneity of cellular populations growing in continuous culture. Spontaneous mutation can be detrimental or beneficial, leading to more competitive progeny. Mutation followed by autoselection is subject to quantitative analyses, described next. Description of mutation and autoselection. Continuous culture turned out to be a very efficient tool to study mutation and autoselection.113 Let N be the total cell concentration, M the concentration of mutants, m the specific growth rate of the main nonmutated part of the cell population, and h the specific growth rate of neutral mutants, then: dM ¼ lmN þ hm
DM; dt (60) dN ¼ mN
lmN
DN dt

Microbial Growth Dynamics

269

Figure 24 Antibiotic production pattern in a batch culture simulated by the synthetic chemostat model. Note that antibiotic production occurs after cessation of the active growth.

where l is the mutation rate expressed as the ratio of the numbers of mutants to the total number of cells formed. If l m, then the original strain will be displaced by the mutant; otherwise, if h < m, M will tend to a lower limit M* ¼ lN/(1
h/D). Experimental studies of phage-resistant mutants in a tryptophan-limited chemostat culture of E. coli showed that the period of linear M increase in accordance with Eq. (61) was fairly short. Every 20–100 generations, there was an abrupt fall in the number of mutants, after which linear growth resumed at the same rate (Fig. 25). The observed sawtooth dynamics in M were explained by Moser12 as a combined effect of mutation and selection. The original wild clone gives not a single but a whole array of mutations with subsequent reversions. Let us denote the total cell population in a chemostat culture as N, which is the sum of all

Figure 25 Change in frequency of neutral mutants (phage T5-resistant cells) in a tryptophan-limited chemostat of Escherichia coli B/r/1 trp at a generation time of 2.8 h. From Novick, A. Experimentation with chemostat. In Recent Progress in Microbiology; Blachwell Science Publication: Oxford, 1959; pp. 403–415.

270

Microbial Growth Dynamics

P subpopulations including original and emerging variants, N ¼ Ni. All possible transitions between variants are given by the matrix, li/j ( j ¼ 1, ., n; i ¼ 1, ., n; j s i). Then, the chemostat model takes the following form: n dNj dN X ¼ ¼ mðsÞN
DN; dt dt j¼1

m¼

n 1 X m Nj ; N j¼1 j

n n X X dNj ¼ mj ðsÞNj
DNj
lj/i Ni ; þ li/j Ni dt isj isj n X

ds ¼ Dðs0
sÞ
mj ðsÞNj Yj ; dt j¼1

mj ðsÞ ¼ mm

(62)

s Ksj þ s

Every drop in the sawtooth dynamics of neutral mutants detected by their phage resistance can be interpreted as the appearance of other types of spontaneous mutants with higher growth capabilities. In a chemostat culture, such mutants overcompete and displace all other cells by virtue of their higher affinity to limiting substrate (saturation constant for mutant is smaller than for the rest of cells in the population, i.e., Ksk < Ksj). In the process of displacement, the growth-limiting substrate concentration decreases from s1 ¼ DKsj =ðmm
DÞ to s2 ¼ DKsk =ðmm
DÞ, and the culture density will rise by Yðs1
s2 Þ. Autoselection in turbidostat and pH-auxostat. The affinity to a substrate is not the only driving force of selection outcome. In a turbidostat and pH-auxostat, autoselection is in favor of mutants with higher maximum specific growth rates, s ¼ mmkd mmj > 0. Because the population density is kept constant instrumentally and the dilution rate is allowed to vary, autoselection results in an increase in D from mmj to m mk. Note that pH-auxostat is one of the most convenient research tools to select speedy strains for many biotech applications. ðm
m Þs Mutation toward a higher growth efficiency, Yk > Yj, will lead to the same result as an increase in mm: s ¼ mk
mj ¼ mkKsj þsmj as soon as mk ¼ qsYk and mj ¼ qsYj. The growth of a mutant with a higher resistance to inhibitory metabolic products can be described by Eq. (30) with Kpk > Kp. A mutant with higher resistance will completely displace the original population, and the product concentration will reach a higher steady-state level: P¼

mm Kpk s
Kpk ðKs þ sÞD

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.

Pirt, S. J. Principles of Microbe and Cell Cultivation, Blackwell Sci: Oxford, London, Edinburgh, Melbourne, 1975. Wang, C. H.; Koch, A. L. Constancy of Growth on Simple and Complex Media. J. Bacteriol. 1978, 136 (3), 969–975. Rahn, O. Physiology of Bacteria, P.Blackiston’s Son and Co: Philadelphia, 1932, 438p. Buchanan, R. E. Life Phases in a Bacterial Culture. J. Infect. Dis. 1918, 23 (1), 109–125. Verhulst, P. F. Notice sur la loi que la population suit dans son accroissement. Corresp. Math. et Phys. 1838, 10 (1), 113–121. Monod, J. Recherches sur la croissancu des cultures bacteriennes, Hermann & C 210: Paris, 1942. Monod, J. La technique de culture continue: theorie et applications. Ann. Instit. Pasteur. 1950, 79, 390–410. Panikov, N. S.; et al. Near-zero Growth Kinetics of Pseudomonas Putida Deduced from Proteomic Analysis. Environ. Microbiol. 2015, 17 (1), 215–228. Button, D. K. Kinetics of Nutrient-limited Transport and Microbial Grpowth. Microbiol. Rev. 1985, 49 (3), 270–297. Panikov, N. S. Microbial Growth Kinetics, Chapman & Hall: London, 1995, 378 p. Teissier, G. Kinetic Behaviour of Heterogeneous Population in Completely Mixed Reactors. Ann. Physiol. Physicochem. Biol 1956, 12, 527–586. Moser, H. The Dynamics of Bacterial Populations Maintained in the Chemostat, Carnegie Inst. Wash. Pub: Washington DC, 1958; p 136, 614. Contois, D. E. Kinetics of Baterial Growth, Relationship between Population Density and Specific Growth Rate of Continuous Culture. J. Gen. Microbiol. 1959, 21, 40–49. Powell, E. O. The Growth Rate of Microorganisms as a Function of Substrate Concentration. In Continuous Cultivation of Microorganisms, H.M. Stat. Office: Salisbury, 1967; pp 34–55. Tempest, D. W.; Neijssel, O. M. The Status of YATP and Maintenance Energy as Biologically Interpretable Phenomena. Annu. Rev. Microbiol. 1984, 38, 459–486. Pirt, S. J. Maintenance Energy: a General Model for Energy-limited and Energy-sufficient Growth. Arch. Microbiol. 1982, 133 (4), 300–302. Andrews, J. F. A Mathematical Model for the Continuous Culture of Microorganisms Utilising Inhibitory Substrates. Biotechnol. Bioeng. 1968, 10, 707–723. Pawlowsky, U.; Howell, S. A.; Chi, C. T. Mixad Cultura Biooxidation of Phenol. III. Existence of Multiple Steady States in Continuous Culture with Wall Growth. Biotechnol. Bioeng. 1973, 15 (5), 905–916. Campbell, A. Synchronization of Cell Division. Bacteriol. Rev. 1957, 21 (2), 263–272. Cooney, C. L. Ann. N. Y. Acad. Sci. 1979, 326, 295–314. Ramkrishna, D.; Fredrickson, A. G.; Tsuchiya, H. M. Dynamics of Microbial Population: Models Considering Inhibitors and Variable Cell Composition. Biotech. Bioeng. 1967, 9 (2), 129–170. Domach, M. M.; et al. Computer Model for Glucose-limited Growth of a Single Cell of Escherichia coli B/r-A. Biotechnol. Bioeng. 1984, 26 (2), 203–216. Jeong, J. W.; Snay, J.; Ataai, M. M. A Mathematical Model for Examining Growth and Sporulation Processes of Bacillus Subtilis. Biotech. Bioeng. 1990, 35, 160–184. Land, M.; et al. Insights from 20 years of Bacterial Genome Sequencing. Funct. Integr. Genom. 2015, 15 (2), 141–161. Shabalina, S. A.; Spiridonov, N. A. The Mammalian Transcriptome and the Function of Non-coding DNA Sequences. Genome Biol. 2004, 5 (4), 105. Palsson, B. O. Systems Biology: Properties of Reconstructed Networks, Cambridge University Press, 2006. Agren, R.; Otero, J. M.; Nielsen, J. Genome-scale Modeling Enables Metabolic Engineering of Saccharomyces cerevisiae for Succinic Acid Production. J. Ind. Microbiol. Biotechnol. 2013, 40 (7), 735–747.

Microbial Growth Dynamics 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84.

271

Dias, O.; et al. Reconstructing Genome-scale Metabolic Models with Merlin. Nucleic Acids Res. 2015, 43 (8), 3899–3910. Henry, C. S.; et al. High-throughput Generation, Optimization and Analysis of Genome-scale Metabolic Models. Nat. Biotechnol. 2010, 28 (9), 977–982. Magnusdottir, S.; et al. Generation of Genome-scale Metabolic Reconstructions for 773 Members of the Human Gut Microbiota. Nat. Biotechnol. 2017, 35 (1), 81–89. Feist, A. M.; et al. A Genome-scale Metabolic Reconstruction for Escherichia coli K-12 MG1655 that Accounts for 1260 ORFs and Thermodynamic Information. Mol. Syst. Biol. 2007, 3, 121. Karr, J. R.; et al. A Whole-cell Computational Model Predicts Phenotype from Genotype. Cell 2012, 150 (2), 389–401. Macklin, D. N.; Ruggero, N. A.; Covert, M. W. The Future of Whole-cell Modeling. Curr. Opin. Biotechnol. 2014, 28, 111–115. Shuler, M. L.; Foley, P.; Atlas, J. Modeling a Minimal Cell. Meth. Mol. Biol. 2012, 881, 573. Gibson, D. G.; et al. Complete Chemical Synthesis, Assembly, and Cloning of a Mycoplasma genitalium Genome. Science 2008, 319 (5867), 1215–1220. Hartwell, L. H.; et al. From Molecular to Modular Cell Biology. Nature 1999, 402, C47. Edwards, J. S.; Ibarra, R. U.; Palsson, B. O. In Silico Predictions of Escherichia coli Metabolic Capabilities Are Consistent with Experimental Data. Nat. Biotechnol. 2001, 19 (2), 125–130. Kauffman, K. J.; Prakash, P.; Edwards, J. S. Advances in Flux Balance Analysis. Curr. Opin. Biotechnol. 2003, 14 (5), 491–496. Becker, S. A.; et al. Quantitative Prediction of Cellular Metabolism with Constraint-based Models: the COBRA Toolbox. Nat. Protoc. 2007, 2 (3), 727–738. Orth, J. D.; Thiele, I.; Palsson, B. O. What Is Flux Balance Analysis? Nat. Biotechnol. 2010, 28 (3), 245–248. Palsson, B.Ø. Systems Biology, Cambridge University Press, 2015. Palsson, B.Ø. Systems Biology: Simulation of Dynamic Network States, Cambridge University Press, 2011. Schuetz, R.; et al. Multidimensional Optimality of Microbial Metabolism. Science 2012, 336 (6081), 601. Waldherr, S.; Lindhorst, H. Optimality in Cellular Storage via the Pontryagin Maximum Principle. IFAC-Papers OnLine 2017, 50 (1), 9889–9895. Wang, Z.; et al. Using Evolutionary Genomics, Transcriptomics, and Systems Biology to Reveal Gene Networks Underlying Fungal Development. Fungal Biol. Rev. 2018. Tarlak, F.; Sadkolu, H.; Akr, T. The Role of Flexibility and Optimality in the Prediction of Intracellular Fluxes of Microbial Central Carbon Metabolism. Mol. Biosyst. 2014, 10 (9), 2459–2465. Beste, D.; et al. GSMN-TB: a Web-based Genome Scale Network Model of Mycobacterium tuberculosis Metabolism. Genome Biol. 2007, 8 (5), R89. Neidhardt, F. C.; Ingraham, J. L.; Schaechter, M.; vol. 20; Sinauer Associates Sunderland, MA, 1990. Dikicioglu, D.; Kõrdar, B.; Oliver, S. G. Biomass Composition: the “elephant in the Room” of Metabolic Modelling. Metabolomics 2015, 11 (6), 1690–1701. Pramanik, J.; Keasling, J. D. Stoichiometric Model of Escherichia coli Metabolism: Incorporation of Growth-rate Dependent Biomass Composition and Mechanistic Energy Requirements. Biotechnol. Bioeng. 1997, 56 (4), 398–421. Pramanik, J.; Keasling, J. D. Effect of Escherichia coli Biomass Composition on Central Metabolic Fluxes Predicted by a Stoichiometric Model. Biotechnol. Bioeng. 1998, 60 (2), 230–238. Yuan, H.; et al. Flux Balance Analysis of Plant Metabolism: The Effect of Biomass Composition and Model Structure on Model Predictions. Front. Plant Sci. 2016, 7 (537). Cankorur-Cetinkaya, A.; Dikicioglu, D.; Oliver, S. G. Metabolic Modeling to Identify Engineering Targets for Komagataella phaffii: The Effect of Biomass Composition on Gene Target Identification. Biotechnol. Bioeng. 2017, 114 (11), 2605–2615. Xavier, J. C.; Patil, K. R.; Rocha, I. Integration of Biomass Formulations of Genome-scale Metabolic Models with Experimental Data Reveals Universally Essential Cofactors in Prokaryotes. Metab. Eng. 2017, 39, 200–208. Chan, S. H. J.; et al. Standardizing Biomass Reactions and Ensuring Complete Mass Balance in Genome-scale Metabolic Models. Bioinformatics 2017, 33 (22), 3603–3609. Stephanopoulos, G.; Aristidou, A. A.; Nielsen, J. Metabolic Engineering: Principles and Methodologies, Elsevier, 1998. Varma, A.; Palsson, B. O. Stoichiometric Flux Balance Models Quantitatively Predict Growth and Metabolic By-product Secretion in Wild-type Escherichia coli W3110. Appl. Environ. Microbiol. 1994, 60 (10), 3724–3731. Mahadevan, R.; Edwards, J. S.; Doyle, F. J., III Dynamic Flux Balance Analysis of Diauxic Growth in Escherichia coli. Biophys. J. 2002, 83 (3), 1331–1340. Covert, M. W.; Palsson, B. O. Transcriptional Regulation in Constraints-based Metabolic Models of Escherichia coli. J. Biol. Chem. 2002, 277 (31), 28058–28064. Covert, M. W.; et al. Integrating Metabolic, Transcriptional Regulatory and Signal Transduction Models in Escherichia coli. Bioinformatics 2008, 24 (18), 2044–2050. Jensen, P. A.; Lutz, K. A.; Papin, J. A. TIGER: Toolbox for Integrating Genome-scale Metabolic Models, Expression Data, and Transcriptional Regulatory Networks. BMC Syst. Biol. 2011, 5, 147. Kostromins, A.; Stalidzans, E. Paint4Net: COBRA Toolbox Extension for Visualization of Stoichiometric Models of Metabolism. Biosystems 2012, 109 (2), 233–239. Thorleifsson, S. G.; Thiele, I. rBioNet: A COBRA Toolbox Extension for Reconstructing High-quality Biochemical Networks. Bioinformatics 2011, 27 (14), 2009–2010. Jamialahmadi, O.; Motamedian, E.; Hashemi-Najafabadi, S. BiKEGG: a COBRA Toolbox Extension for Bridging the BiGG and KEGG Databases. Mol. Biosyst. 2016, 12 (11), 3459–3466. Fleming, R. M.; Thiele, I. von Bertalanffy 1.0: a COBRA toolbox extension to thermodynamically constrain metabolic models. Bioinformatics 2011, 27 (1), 142–143. Mao, L.; Verwoerd, W. S. ORCA: a COBRA Toolbox Extension for Model-driven Discovery and Analysis. Bioinformatics 2014, 30 (4), 584–585. Hoffner, K.; Harwood, S. M.; Barton, P. I. A Reliable Simulator for Dynamic Flux Balance Analysis. Biotechnol. Bioeng. 2013, 110 (3), 792–802. Flassig, R. J.; et al. Dynamic Flux Balance Modeling to Increase the Production of High-value Compounds in Green Microalgae. Biotechnol. Biofuels 2016, 9, 165. Gomez, J. A.; Barton, P. I. Dynamic Flux Balance Analysis Using DFBAlab. Meth. Mol. Biol. 2018, 1716, 353–370. Gomez, J. A.; Hoffner, K.; Barton, P. I. DFBAlab: a Fast and Reliable MATLAB Code for Dynamic Flux Balance Analysis. BMC Bioinf. 2014, 15, 409. Kompala, D. S.; Ramkrishna, D.; Tsao, G. T. Cybernetic Modeling of Microbial Growth on Multiple Substrates. Biotechnol. Bioeng. 1984, 26 (11), 1272–1281. Dhurjati, P.; et al. A Cybernetic View of Microbial Growth: Modeling of Cells as Optimal Strategists. Biotechnol. Bioeng. 1985, 27 (1), 1–9. Kompala, D. S.; et al. Investigation of Bacterial Growth on Mixed Substrates: Experimental Evaluation of Cybernetic Models. Biotechnol. Bioeng. 1986, 28 (7), 1044–1055. Turner, B. G.; Ramkrishna, D.; Jansen, N. B. Cybernetic Modeling of Bacterial Cultures at Low Growth Rates: Mixed-substrate Systems. Biotechnol. Bioeng. 1988, 32 (1), 46–54. Straight, J. V.; Ramkrishna, D. Complex Growth Dynamics in Batch Cultures: Experiments and Cybernetic Models. Biotechnol. Bioeng. 1991, 37 (10), 895–909. Varner, J.; Ramkrishna, D. Application of Cybernetic Models to Metabolic Engineering: Investigation of Storage Pathways. Biotechnol. Bioeng. 1998, 58 (2–3), 282–291. Song, H.-S.; Ramkrishna, D. Prediction of Metabolic Function from Limited Data: Lumped Hybrid Cybernetic Modeling (L-HCM). Biotechnol. Bioeng. 2010, 106 (2), 271–284. Kim, J. I.; Varner, J. D.; Ramkrishna, D. A Hybrid Model of Anaerobic E. coli GJT001: Combination of Elementary Flux Modes and Cybernetic Variables. Biotechnol. Prog. 2008, 24 (5), 993–1006. Schuster, S.; Hilgetag, C. On Elementary Flux Modes in Biochemical Reaction Systems at Steady State. J. Biol. Syst. 1994, 2 (02), 165–182. Schuster, S.; Dandekar, T.; Fell, D. A. Detection of Elementary Flux Modes in Biochemical Networks: a Promising Tool for Pathway Analysis and Metabolic Engineering. Trends Biotechnol. 1999, 17 (2), 53–60. von Kamp, A.; et al. Use of CellNetAnalyzer in Biotechnology and Metabolic Engineering. J. Biotechnol. 2017, 261, 221–228. Song, H. S.; Ramkrishna, D. Cybernetic Models Based on Lumped Elementary Modes Accurately Predict Strain-specific Metabolic Function. Biotechnol. Bioeng. 2011, 108 (1), 127–140. Young, J. D.; Ramkrishna, D. On the Matching and Proportional Laws of Cybernetic Models. Biotechnol. Prog. 2007, 23 (1), 83–99. Young, J. D.; et al. Integrating Cybernetic Modeling with Pathway Analysis Provides a Dynamic, Systems-level Description of Metabolic Control. Biotechnol. Bioeng. 2008, 100 (3), 542–559.

272

Microbial Growth Dynamics

85. Song, H. S.; et al. Dynamic Modeling of Aerobic Growth of Shewanella oneidensis. Predicting Triauxic Growth, Flux Distributions, and Energy Requirement for Growth. Metab. Eng. 2013, 15, 25–33. 86. Song, H.-S.; Liu, C. Dynamic Metabolic Modeling of Denitrifying Bacterial Growth: The Cybernetic Approach. Ind. Eng. Chem. Res. 2015, 54 (42), 10221–10227. 87. Song, H.-S.; Ramkrishna, D. Prediction of Dynamic Behavior of Mutant Strains from Limited Wild-type Data. Metab. Eng. 2012, 14 (2), 69–80. 88. Ramkrishna, D.; Song, H.-S. Dynamic Models of Metabolism: Review of the Cybernetic Approach. AIChE J. 2012, 58 (4), 986–997. 89. Droop, M. R. Vitamin B 12 and Marine Ecology. IV. The Kinetics of Uptake, Growth and Inhibition in Monochrysis lutheri. J. Mar. Biol. Ass. 1968, 48 (3), 689–733. 90. Panikov, N. S. Mechanistic Mathematical Models of Microbial Growth in Bioreactors and in Natural Soils: Explanation of Complex Phenomena. Math. Comput. Simulat. 1996, 42, 179–186. 91. Tempest, D. W.; Herbert, D.; Phipps, P. J. Studies on the Growth of Aerobacter aerogenes at Low Dilution Rates in a Chemostat. In Continuous Cultivation of Microorganisms, H.M. Stat. Office: Salisbury, 1967; pp 240–253. 92. Skichko, A.; Kol’tsova, E. Mathematical Model for Describing Oscillations of Bacterial Biomass. Theor. Found. Chem. Eng. 2006, 40 (5), 503–513. 93. Blagodatsky, S. A.; Heinemeyer, O.; Richter, J. Estimating the Active and Total Soil Microbial Biomass by Kinetic Respiration Analysis. Biol. Fertil. Soils 2000, 32 (1), 73–81. 94. Kravchenko, L.; Shvytov, I. Mathematical Simulation of the Dynamics of Interacting Populations of Rhizosphere Microorganisms. Microbiology 2004, 73 (2), 189–195. 95. Pagel, H.; et al. Micro-scale Modeling of Pesticide Degradation Coupled to Carbon Turnover in the Detritusphere: Model Description and Sensitivity Analysis. Biogeochemistry 2014, 117 (1), 185–204. 96. Resat, H.; et al. Modeling Microbial Dynamics in Heterogeneous Environments: Growth on Soil Carbon Sources. Microb. Ecol. 2012, 63 (4), 883–897. 97. Wang, G.; et al. Representation of Dormant and Active Microbial Dynamics for Ecosystem Modeling. PLoS One 2014, 9 (2), e89252. 98. Wutzler, T.; et al. Soil Microbial Biomass and its Activity Estimated by Kinetic Respiration Analysis–Statistical Guidelines. Soil Biol. Biochem. 2012, 45, 102–112. 99. Yakushev, A.; Bubnov, I.; Semenov, A. Estimation of the Effects of Earthworms and Initial Substrates on the Bacterial Community in Vermicomposts. Eurasian Soil Science 2011, 44 (10), 1117. 100. Panikov, N. S. Understanding and Prediction of Soil Microbial Community Dynamics under Global Change. Appl. Soil Ecol. 1999, 11, 161–176. 101. Peebo, K.; et al. Proteome Reallocation in Escherichia coli with Increasing Specific Growth Rate. Mol. Biosyst. 2015, 11 (4), 1184–1193. 102. Fazio, A.; et al. Transcription Factor Control of Growth Rate Dependent Genes in Saccharomyces cerevisiae: A Three Factor Design. BMC Genom. 2008, 9 (1), 341. 103. Wu, J.; et al. Global Analysis of Nutrient Control of Gene Expression in Saccharomyces cerevisiae during Growth and Starvation. Proc. Natl. Acad. Sci. U.S.A. 2004, 101 (9), 3148–3153. 104. Hengge-Aronis, R. Signal Transduction and Regulatory Mechanisms Involved in Control of the SigmaS (RpoS) Subunit of RNA Polymerase. Microbiol. Mol. Biol. Rev. 2002, 66 (3), 373–395. 105. Esener, A. A.; Roels, J. A.; Kossen, N. W. F. Theory and Applications of Unstructured Growth Models: Kinetic and Energetic Aspects. Biotechnol. Bioeng. 1983, 25 (12), 2803–2841. 106. Erickson, L. E.; Minkevich, I. G.; Eroshin, V. K. Utilization of Mass-energy Balance Regularities in the Analysis of Continuous-culture Data. Biotechnol. Bioeng. 1979, 21 (4), 575–591. 107. Erickson, L. E.; Minkevich, I. G.; Eroshin, V. K. Application of Mass and Energy Balance Regularities in Fermentation. Biotechnol. Bioeng. 2000, 67 (6), 748–774. 108. Mayberry, W. R.; Prochazka, G. J.; Payne, W. J. Growth Yields of Bacteria on Selected Organic Compounds. Appl. Microbiol. 1967, 15 (6), 1332–1338. 109. Strocchi, M.; et al. Low Temperature-induced Systems Failure in Escherichia coli: Insights from Rescue by Cold-adapted Chaperones. Proteomics 2006, 6 (1), 193–206. 110. Ullrich, C.; Novacky, A. Recent Aspects of Ion-induced pH Changes. Curr. Topic Plant Biochem. Physiol. 1992, 11, 231–241. 111. Minkevich, I. G. Ionic Balance During Cultivation of Cell Populations in Spatial Homogenous Medium. Biophysics 1979, 24 (4), 712–716. 112. Minkevich, I. G.; Eroshin, V. K. Influence of Cell Metabolism on the Rate of Medium Addition during the pH-stabilized Cultivation. Biophysics 1979, 24 (5), 879–884. 113. Novick, A.; Szilard, L. Experiments with the Chemostat on Spontaneous Mutations of Bacteria. Proc. Natl. Acad. Sci. U. S. A. 1950, 36 (12), 708–719. 114. Varma, A.; Boesch, B. W.; Palsson, B. O. Stoichiometric Interpretation of Escherichia coli Glucose Catabolism under Various Oxygenation Rates. Appl. Environ. Microbiol. 1993, 59 (8), 2465–2473. 115. Nyholm, N. Kinetics of Phosphate Limited Algal Growth. Biotechnol. Bioeng. 1977, 19 (3), 467–492. 116. Panikov, N. S.; Lynd, L. R. Physiological and Methodological Aspects of Cellulolytic Microbial Cultures. In Manual of Industrial Microbiology and Biotechnology; Baltz, R., Davies, J. E., Demain, A. L., Eds.; 2010, 3rd ed.; 2010 (MIMB3). 117. Novick, A. Experimentation with Chemostat. In Recent Progress in Microbiology, Blackwell Sci. Publication: Oxford, 1959; pp 403–415. 118. Edwards, J. S.; Palsson, B. O. Systems Properties of the Haemophilus influenzae Rd Metabolic Genotype. J. Biol. Chem. 1999, 274 (25), 17410–17416. 119. Edwards, J. S.; Palsson, B. O. The Escherichia coli MG1655 In Silico Metabolic Genotype: its Definition, Characteristics, and Capabilities. Proc. Natl. Acad. Sci. U. S. A. 2000, 97 (10), 5528–5533. 120. Reed, J. L.; et al. An Expanded Genome-scale Model of Escherichia coli K-12 (iJR904 GSM/GPR). Genome Biol. 2003, 4 (9), R54. 121. Schilling, C. H.; et al. Genome-scale Metabolic Model of Helicobacter pylori 26695. J. Bacteriol. 2002, 184 (16), 4582–4593. 122. Thiele, I.; et al. Expanded Metabolic Reconstruction of Helicobacter pylori (IIT341 GSM/GPR): an in Silico Genome-scale Characterization of Single- and Double-deletion Mutants. J. Bacteriol. 2005, 187 (16), 5818–5830. 123. Famili, I.; et al. Saccharomyces cerevisiae Phenotypes Can Be Predicted by Using Constraint-based Analysis of a Genome-scale Reconstructed Metabolic Network. Proc. Natl. Acad. Sci. U. S. A. 2003, 100 (23), 13134–13139. 124. Duarte, N. C.; Herrgard, M. J.; Palsson, B. O. Reconstruction and Validation of Saccharomyces cerevisiae IND750, a Fully Compartmentalized Genome-scale Metabolic Model. Genome Res. 2004, 14 (7), 1298–1309. 125. Mahadevan, R.; et al. Characterization of Metabolism in the Fe(III)-reducing Organism Geobacter Sulfurreducens by Constraint-based Modeling. Appl. Environ. Microbiol. 2006, 72 (2), 1558–1568. 126. Becker, S. A.; Palsson, B. O. Genome-scale Reconstruction of the Metabolic Network in Staphylococcus aureus N315: an Initial Draft to the Two-dimensional Annotation. BMC Microbiol. 2005, 5, 8. 127. Hong, S. H.; et al. The Genome Sequence of the Capnophilic Rumen Bacterium Mannheimia succiniciproducens. Nat. Biotechnol. 2004, 22 (10), 1275–1281. 128. Oh, Y. K.; et al. Genome-scale Reconstruction of Metabolic Network in Bacillus subtilis Based on High-throughput Phenotyping and Gene Essentiality Data. J. Biol. Chem. 2007, 282 (39), 28791–28799. 129. Suthers, P. F.; et al. A Genome-scale Metabolic Reconstruction of Mycoplasma genitalium, i PS189 ( In Silico M. genitalium Metabolic Reconstruction). PLoS Comput. Biol. 2009, 5 (2), e1000285. 130. Sainz, J.; et al. Modeling of Yeast Metabolism and Process Dynamics in Batch Fermentation. Biotechnol. Bioeng. 2003, 81 (7), 818–828. 131. Luo, R. Y.; et al. Dynamic Analysis of Optimality in Myocardial Energy Metabolism under Normal and Ischemic Conditions. Mol. Syst. Biol. 2006, 2 (1). n/a–n/a. 132. Hjersted, J. L.; Henson, M. A. Optimization of Fed-batch Saccharomyces cerevisiae Fermentation Using Dynamic Flux Balance Models. Biotechnol. Prog. 2006, 22 (5), 1239–1248. 133. Pizarro, F.; et al. Coupling Kinetic Expressions and Metabolic Networks for Predicting Wine Fermentations. Biotechnol. Bioeng. 2007, 98 (5), 986–998. 134. Hjersted, J. L.; Henson, M. A.; Mahadevan, R. Genome-scale Analysis of Saccharomyces cerevisiae Metabolism and Ethanol Production in Fed-batch Culture. Biotechnol. Bioeng. 2007, 97 (5), 1190–1204. 135. Anesiadis, N.; Cluett, W. R.; Mahadevan, R. Dynamic Metabolic Engineering for Increasing Bioprocess Productivity. Metab. Eng. 2008, 10 (5), 255–266.

Microbial Growth Dynamics

273

136. Luo, R.; et al. Photosynthetic Metabolism of C3 Plants Shows Highly Cooperative Regulation under Changing Environments: a Systems Biological Analysis. Proc. Natl. Acad. Sci. U. S. A. 2009, 106 (3), 847. 137. Oddone, G. M.; Mills, D. A.; Block, D. E. A Dynamic, Genome-scale Flux Model of Lactococcus lactis to Increase Specific Recombinant Protein Expression. Metab. Eng. 2009, 11 (6), 367–381. 138. Gaspard, L.; et al. Dynamic Metabolic Flux Analysis Demonstrated on Cultures where the Limiting Substrate Is Changed from Carbon to Nitrogen and Vice Versa. J. Biomed. Biotechnol. 2010, 2010. 139. Salimi, F.; Zhuang, K.; Mahadevan, R. Genome-scale Metabolic Modeling of a Clostridial Co-culture for Consolidated Bioprocessing. Biotechnol. J. 2010, 5 (7), 726–738. 140. Meadows, A. L.; et al. Application of Dynamic Flux Balance Analysis to an Industrial Escherichia coli Fermentation. Metab. Eng. 2010, 12 (2), 150–160. 141. Zhuang, K.; et al. Genome-scale Dynamic Modeling of the Competition between Rhodoferax and Geobacter in Anoxic Subsurface Environments. ISME J. 2011, 5 (2), 305–316. 142. Vargas, F. A.; et al. Expanding a Dynamic Flux Balance Model of Yeast Fermentation to Genome-scale. BMC Syst. Biol. 2011, 5, 75. 143. Nolan, R. P.; Lee, K. Dynamic Model of CHO Cell Metabolism. Metab. Eng. 2011, 13 (1), 108–124. 144. Hanly, T. J.; Henson, M. A. Dynamic Flux Balance Modeling of Microbial Co-cultures for Efficient Batch Fermentation of Glucose and Xylose Mixtures. Biotechnol. Bioeng. 2011, 108 (2), 376–385. 145. Hanly, T.; Urello, M.; Henson, M. Dynamic Flux Balance Modeling of S. cerevisiae and E. coli Co-cultures for Efficient Consumption of Glucose/xylose Mixtures. Appl. Microbiol. Biotechnol. 2012, 93 (6), 2529–2541. 146. Muthuraj, M.; et al. Flux Balance Analysis of Chlorella sp. FC2 IITG under Photoautotrophic and Heterotrophic Growth Conditions. Photosynth. Res. 2013, 118 (1-2), 167–179. 147. Sanchez, B. J.; Perez-Correa, J. R.; Agosin, E. Construction of Robust Dynamic Genome-scale Metabolic Model Structures of Saccharomyces cerevisiae through Iterative Reparameterization. Metab. Eng. 2014, 25, 159–173. 148. Fernandes de Sousa, S.; et al. Dynamic Metabolic Flux Analysis Using a Convex Analysis Approach: Application to Hybridoma Cell Cultures in Perfusion. Biotechnol. Bioeng. 2016, 113 (5), 1102–1112. 149. Zhao, X.; et al. Dynamic Flux Balance Analysis with Nonlinear Objective Function. J. Math. Biol. 2017, 75 (6–7), 1487–1515. 150. Upton, D. J.; McQueen-Mason, S. J.; Wood, A. J. An Accurate Description of Aspergillus niger Organic Acid Batch Fermentation through Dynamic Metabolic Modelling. Biotechnol. Biofuels 2017, 10, 258. 151. von Wulffen, J.; et al. Transition of an Anaerobic Escherichia coli Culture to Aerobiosis: Balancing mRNA and Protein Levels in a Demand-directed Dynamic Flux Balance Analysis. PLoS One 2016, 11 (7), e0158711. 152. Saitua, F.; et al. Dynamic Genome-scale Metabolic Modeling of the Yeast Pichia pastoris. BMC Syst. Biol. 2017, 11 (1), 27. 153. Antoniewicz, M. R. Dynamic Metabolic Flux Analysisdtools for Probing Transient States of Metabolic Networks. Curr. Opin. Biotechnol. 2013, 24 (6), 973–978. 154. Lloyd, C. J.; Ebrahim, A.; Yang, L.; King, Z. A.; Catoiu, E.; O’Brien, E. J.; Liu, J. K.; Palsson, B. O. COBRAme: A computational framework for genome-scale models of metabolism and gene expression. PLoS Comput Biol 2018, 14 (7), e1006302. https://doi.org/10.1371/journal.pcbi.1006302.

1.19

Modes of Culture/Animal Cells

Xudong Zhang, Yuan Wen, and Shang-Tian Yang, The Ohio State University, Columbus, OH, United States © 2011 Elsevier B.V. All rights reserved. This is a reprint of X. Zhang, Y. Wen, S.T. Yang, 1.21 - Modes of Culture/Animal Cells, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 285-302.

1.19.1 Introduction 1.19.2 Batch Culture: The Basis for All Cell Culture Systems 1.19.2.1 Batch Culture Kinetics 1.19.2.2 Bioreactor and Process Control 1.19.3 Fed-Batch Culture: Dominator of Industrial-Scale Processes 1.19.3.1 Feed Medium Design 1.19.3.2 Feeding Strategy 1.19.3.3 Process and Bioreactor Control 1.19.4 Perfusion Culture: The Most Sophisticated Process 1.19.4.1 Kinetics 1.19.4.2 Perfusion Culture Processes 1.19.4.2.1 Cell Retention Through Immobilization 1.19.4.2.2 Cell Retention Systems for Suspension Culture 1.19.4.3 Scale-Up and Optimization 1.19.5 Concluding Remarks on the Selection of Culture Mode References Relevant Websites

274 275 278 278 280 281 282 282 283 284 284 284 285 288 288 290 291

Glossary Batch culture The cell culture process in which all basal medium components required for cell growth and production are provided at the beginning. Cell-specific perfusion rate (CSPR) A concept generally used in perfusion cell culture, which is measured in the unit of nL/ cell day, representing the volume of feed medium required per cell per day. Continuous culture A continuous culture provides fresh medium in a nonstop fashion to the culture vessel as cells continuously utilize nutrients and there is also a continuous harvest of medium to maintain the same culture volume in the bioreactor. Depending on whether there is cell retention as medium flows through, continuous culture can be categorized as chemostat, where cells are washed away with medium, and perfusion culture, where cells are retained or recycled through a certain device inside or outside of the culture vessel. Current good manufacturing practice (cGMP) Guidelines issued by government regulations to cover the manufacturing and testing of pharmaceuticals, medical devices, and foods. Design of experiment (DOE) A structured strategy to achieve empirical knowledge by systematically varying relevant factors in exercises to study how factors affect the process outputs. Fed-batch culture The cell culture process in which one or more concentrated feed media are provided following a specialized feeding schedule to meet nutritional needs of an extended production period, achieving higher product titers than batch cultures. Flow injection analysis (FIA) An automated chemical analysis approach by injecting a plug of liquid sample into a moving continuous carrier stream and recording the physical/chemical changes when the zone of the sample passes through a detector. Integral of viable cell density (IVCD) The integration of viable cell density along with the cell culture process time. Population doubling time (PDT) In cell culture, the period of time required for the viable cell density to double. Process analytical technologies (PAT) A regulatory framework initiated by Food and Drug Administration aiming at motivating the pharmaceutical industry for process improvement by monitoring critical process parameters (CPP), which have significant impacts on critical quality attributes (CQA).

1.19.1

Introduction

The current worldwide biopharmaceutical market revenue is about 80–90 billion US dollars with an annual growth rate of 13%. Majority of the biologic drugs on the market and in the development pipeline are produced in mammalian cell culture systems. Compared with bacterial, yeast, and insect cell culture systems, animal cell cultures provide efficient protein folding, assembly,

274

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00020-3

Modes of Culture/Animal Cells

275

posttranslational modifications, and secretion and are superior for the production of therapeutic glycoproteins. Currently, processes using mammalian cells supply more than half of Food and Drug Administration (FDA)-approved biological products and most of the biopharmaceutical blockbusters. Table 1 lists some biopharmaceutical therapeutic recombinant proteins produced from mammalian cell cultures, along with the product type, therapeutic applications, and production cell lines, media, and manufacturing process. These manufacturing processes share some common features. They all use either murine myeloma (NS0 or SP2/0) or Chinese hamster ovary (CHO) cells, with CHO being the principal choice (11 out of 14) by most of the manufacturers. Also, driven principally by concerns of potential human pathogen introduction and process variation, serum-free and animal component-free media are commonly used today in both seed train and production processes. Except for the early animal cell culture product, Epogen that is produced by anchorage-dependent CHO cells in roller bottles, all of today’s major animal cell culture products are produced in suspension cultures in stirred-tank bioreactors operated predominantly in either batch or fed-batch mode. However, perfusion cultures are also used in the production of some important animal cell culture products, including Xyntha and Remicade. The development of a mammalian cell culture process usually starts from cell-line generation, which involves transfection to deliver foreign genes to cell genome and the selection of stable high-producer clones. Once a single cell clone is selected, it is expanded and dispensed as aliquots in vials with a high cell density, which is called a master cell bank (MCB), to supply all developmental, clinical, and commercial production cultures. They are stored in a quiescent state at a low temperature in liquid nitrogen. Typically, new cell banks are made from MCB to supply cells for reactor runs, which is called working cell bank (WCB). After thawing a vial, cells are expanded in small-scale vessels such as T-flasks, shake flasks, spinner flasks, and different scales of bioreactors sequentially until enough cells are made to provide the inoculum for the production bioreactor.1 Fig. 1 shows a typical cell culture process starting from WCB through seed train expansion to the final production stage in a large-scale bioreactor. In this illustrated process, cell culture is scaled up with a split ratio of
5 and cell density in the seed train maintains within a range
0.3–1.5 million cells/mL. At the early stages with small-scale culture vessels (T-flasks and spinner flasks), temperature and gas exchange are maintained in a controlled environment (37 C, 5% CO2 incubator). As the cultures propagated and increased in scale in the seed train, bioreactors with fully controlled process parameters and automatic sampling capability are widely used in industry. These bioreactors are usually operated in the batch mode. In the manufacturing of recombinant protein therapeutics, the production-scale bioreactor is usually operated in the fed-batch mode, with a few exceptions in the (continuous) perfusion mode. It should be noted that the design and operation mode of an animal cell culture process is determined by the production needs and facility capacity, which vary with applications and product- and process-development stages. For example, the time taken to provide the first batch material is essential for preclinical study and early-stage clinical trials, which can be minimized by adopting existing platform technologies, including using commercially available media and standard culture vessels (bioreactors) and operating conditions. In late-stage trials and for large-scale commercial production, establishing a robust and reproducible process with high product quantity and quality is crucial to profit margin.2,3 For applications, such as in tissue engineering, cell-based biosensors, and high-throughput bioprocess optimization and drug screening, involving small-scale cell cultures, the bioreactor and process are predominately operated at a batch mode, although fed-batch and perfusion cultures are often the subjects of academic research. In this article, we focus on different cell culture modes used in the manufacturing of recombinant protein therapeutics. In this context, cell culture process development and general design considerations, including feed media design and feeding strategy, and bioreactor design and process control affecting the choice of different culture modes in the manufacturing process are also discussed.

1.19.2

Batch Culture: The Basis for All Cell Culture Systems

In batch culture, all basal medium components and nutrients required for cell growth are provided at the beginning for the cell culture process. A typical cell growth curve in batch culture is composed of four phases: lag phase, exponential or log phase, stationary phase, and death or decline phase. The lag phase may be profoundly dependent on the age of the inoculum and the adaptation of cells to the new culture environment. The inoculated cells should be at early or middle log phase from the maintenance culture to prevent the lag phase. If it is a routine production in an already adapted environment with a similar culture vessel and culture medium, the lag phase is usually minimal. In the log phase, cells exhibit the fastest growth rate of all phases. The duration of the log phase and the peak viable cell density (VCD) are highly cell-line and medium dependent. With the proper medium, the log phase can last up to 5–6 days and peak VCD can reach over 5  106 –10  106 cells/mL. After the active growth phase, cells may enter a brief stationary phase, where VCD moves along a plateau for hours up to a couple of days, followed by declines in cell viability (the ratio between VCD and total cell density) and VCD into the last phase. When major nutrient and energy sources, such as glucose, are depleted from the basal medium, the culture enters the death phase and VCD declines rapidly. The batch culture usually ends when cell viability falls below 50%. Fig. 2 shows typical batch culture kinetics with time-course profiles of cell density, glucose concentration, and the protein product, IgG. Repeated-batch culture is similar to batch culture with an additional operation to partially replace cell suspension with fresh medium to achieve a new culture cycle. Peak cell density in a repeated-batch culture is usually lower than the counterpart batch. A highly viable cell suspension is prerequisite for the success of repeated-batch cultures.4 This culture mode is often used to subculture cells and each cycle of the subculture can be considered as the early part of a batch culture well before cells enter a crisis due to nutrient depletion, with each passage usually lasting from 2 to 4 days.

276

Table 1

Top selling biopharmaceutical products from mammalian cell cultures Company

Approval date

Type of product

Disease or medical use

Cell line and medium

Manufacturing process

Epogen/Erythropoietin

Amgen

June-1989

Anemia

CHO

Cerezyme/Imiglucerase

Genzyme

May-1994

Gaucher disease

CHO

Avonex/Interferon beta-1a Rituxan/Rituximab

Biogen Idec

May-1996

Glycoprotein hormone/cytokine Human b-glucocerebrosidase Interferon

Multiple sclerosis

CHO

Large number of roller bottles, batch 2000 L stirred-tank suspension Stirred-tank suspension

Genentech, Roche, Biogen Idec, Chugai Pharmaceutical

November-1997

Lymphomas, leukemias, autoimmune disorders

CHO with serumfree medium

12,000 L stirred-tank suspension

Synagis/Palivizumab

MedImmune, AstraZeneca

June-1998

mAb, chimeric murine/human antiCD20 IgG1-k mAb, humanized murine IgG1-k

Respiratory syncytial virus

NS0 with medium containing bovine products

Remicade/Infliximab

August-1998

Sp2/0 with medium containing bovine derived materials CHO-DXB11

Herceptin/Trastuzumab

Genentech, Roche

September-1998

TNKase/Tenecteplase

Genentech, Roche

June-2000

mAb, mouse/human chimeric antiTNFa IgG1-k Fusion protein, TNFR2-IgG1 Fc mAb, humanized IgG1-k specific for human HER-2 Human tPA

Crohn‘s disease, Rheumatoid arthritis

Enbrel/Etanercept

Centocor, Johnson and Johnson, Schering-Plough, Mitsubishi Tanabe Pharma Amgen/Wyeth

10,000 L or more stirred-tank suspension, fed-batch, productivity 1 g/L, 18–22 days production culture Stirred-tank, continuous perfusion (spin filter)

Humira/Adalimumab

Abbott Laboratories

December-2002

Xolair/Omalizumab

Genentech, Roche, Novartis, Tanox Genentech, Roche

June-2003

Imclone, Eli Lilly, Bristol-Myers Squibb Wyeth, Pfizer

February-2004

Avastin/Bevacizumab

Erbitux/Cetuximab Xyntha/Antihemophilic Factor (Recombinant)

November-1998

February-2004

February-2008

Rheumatoid arthritis Breast cancer

CHO with serum-free medium

Thrombolysis

CHO-DHFR

mAb, Human antiTNFa IgG1-k

Rheumatoid arthritis

mAb, humanized antiIgE IgG1 mAb, humanized antiVEGF IgG1

Asthma

mAb, mouse/human chimeric antiEGFR IgG1 Recombinant factor VIII

Colorectal cancer, head and neck cancer Hemophilia A, bleeding episodes and surgical prophylaxis

CHO with medium containing no animal or human-derived components CHO-K1 with serum-free medium CHO-K1 DUX B11 with serum-free medium containing hormones and protein hydrolysates Sp2/0 with serum-free medium CHO with chemically defined medium containing recombinant insulin without albumin

All stirred-tank suspension processes without a specified culture mode are operated in fed-batch. From www.fda.gov; www.emea.europa.eu; www.biopharma.com; www.wikipedia.org.

Colorectal cancer

Up to 20,000 L stirred-tank suspension 12,000 L stirred-tank suspension 12,000 L stirred-tank suspension 12,000 L stirred-tank, Extended batch process Batch-fed suspension 12,000 L stirred-tank suspension, fed-batch 10,000–12,000 L stirred-tank suspension 500 L, continuous perfusion, 500 L fresh media is exchanged daily

Modes of Culture/Animal Cells

Brand generic name

Modes of Culture/Animal Cells

Working cell bank (1 ml) vial thaw

250 ml spinner flask

T-flask

Working volume: 20 ml Seed density: ~3 E5 cells/ml Final density: ~15 E5 cells/ml 5% CO2 culture incubator Temperature: 37 ⬚C

3 l spinner flask

Working volume: 100 ml Seed density: ~3 E5 cells/ml Final density: ~15 E5 cells/ml 5% CO2 culture incubator Temperature: 37 ⬚C

Glutamine Feed medium

15 l spinner flask

Working volume: 500 ml Seed density: ~3 E5 cells/ml Working volume: 2.5 l Final density: ~15 E5 cells/ml Seed density: ~3 E5 cells/ml 5% CO2 culture incubator Final density: ~15 E5 cells/ml Temperature: 37 ⬚C 5% CO2 culture incubator Temperature: 37 ⬚C

Glucose

Initial volume: 8000 l Seed density: ~3 E5 cells/ml Final density: ~15 E5 cells/ml DO: 50% pH: 7.0 Agitation: 15 rpm Temperature: 37 ⬚C

Working volume: 1800 l Seed density: ~3 E5 cells/ml Final density: ~15 E5 cells/ml DO: 50% pH: 7.0 Agitation: 30 rpm Temperature: 37 ⬚C

12 000 l bioreactor

Working volume: 350 l Seed density: ~3 E5 cells/ml Final density: ~15 E5 cells/ml DO: 50% pH: 7.0 Agitation: 50 rpm Temperature: 37 ⬚C

20 l bioreactor

Working volume: 14 l Seed density: ~3 E5 cells/ml Final density: ~15 E5 cells/ml DO: 50% pH: 7.0 Aigtation: 120 rpm Temperature: 37 ⬚C

Working volume: 70 l Seed density: ~3 E5 cells/ml Final density: ~15 E5 cells/ml DO: 50% pH: 7.0 Agitation: 100 rpm Temperature: 37 ⬚C

500 l bioreactor

2500 l bioreactor

277

100 l bioreactor

Figure 1 A typical cell culture process starting from working cell bank through seed train expansion to the final production stage in a large-scale bioreactor that is usually operated at the fed-batch mode. All bioreactors in the seed-train expansion are usually operated at the batch mode.

Majority of animal cells including many established cell lines are anchorage dependent, which means their in vitro growth requires the attachment to a support surface. This growth requirement of solid surface for cell adherence is difficult to scale-up in conventional bioreactors. Erythropoietin as the first blockbuster biologics from mammalian cell culture was initially produced by CHO cells in a labor-intensive and time-consuming process involving a large number of roller bottles each with a very low volumetric productivity. This process was later replaced with suspension culture processes with cells, after adaptation to anchorage independence, grown in large stirred-tank bioreactors with gentle agitation for mixing and continuous gas sparging for aeration. The stirred-tank bioreactor with precise controls on the dissolved oxygen (DO), pH, and temperature is the common platform in today‘s commercial biologics production processes because its well-understood engineering principles are more operation friendly and

10

1.4 Batch culture

1.2 VCD Glucose

1.0

IgG

6

0.8 0.6

4

IgG (g l−1)

VCD (106 cells/ml) Glucose (g l−1)

8

0.4 2 0.2 0

0.0 0

Figure 2

1

2

3

4

5 Day

6

7

8

9

10

Typical batch culture kinetics with viable cell density (VCD), glucose concentration, and IgG titer profiles.

278

Modes of Culture/Animal Cells

easier to scale-up. Modern erythropoietin production processes use suspension cultures with significantly improved productivities that are at least one order of magnitude higher than that in roller bottles.

1.19.2.1

Batch Culture Kinetics

A batch culture can be characterized by the specific growth rate m (h1) and population doubling time (PDT). The specific growth rate is defined as follows: m¼

1 dX X dt

(1)

where X is the VCD and t the time. Integration of Eq. (1) yields an expression for the specific growth rate: m¼

lnðX2 =X1 Þ t2  t1

(2)

Cells are usually maintained at early and middle log phase for subculture. Therefore, the two points of VCD should be both within the log phase. Then PDT can be determined as follows: PDT ¼

ln 2 m

(3)

Subculture of cells is a fundamental technique in cell culture technology. It is not only useful for maintaining cells but also crucial for cell adaptation to new media, during which PDT and viability are closely monitored. However, the number of passages should be minimized in subculture in order to avoid cell-line instability. Subculture is also used throughout the seed train process to provide a sufficient cell number for inoculating the production bioreactor.3 Glucose is usually provided at a low level as the carbon and energy source that could exhaust in batch culture. Excessive glucose in an imbalanced nutritional environment can result in the accumulation of waste metabolite, lactate. Glutamine, which is used as both carbon and nitrogen sources for cell growth, is catabolized to ammonia. Both lactate and ammonia in excessive amounts are potentially detrimental to animal cells, although the sensitivity of cells to these metabolites varies with cell lines. Through improvements in cell-line and medium development, the accumulation of these metabolites is not the major cause for cell number decline in batch culture. In some cases, lactate or ammonia can even be utilized by cells in further metabolism to generate more energy or amino acids. In batch cultures, the final product titer depends on cell line and medium, ranging from several milligrams to over 500 mg/L or even exceeding 1000 mg/L. The cell-line and media dependence of product titer can be demonstrated by the concept of cell-specific productivity, which is defined as the amount of product formation by a single cell in unit time, usually measured in pg/cell day (pcd). The calculation of cell-specific productivity involves the concept of integral of VCD (IVCD), expressed as Z t Xdt (4) IVCD ¼ 0

where X is the VCD and t the culture time. It would be helpful to understand IVCD as the production capacity of the cell factory: How many workers are there? How long do they work? The summated product of these two factors indicates the production capacity of a cell culture process. With IVCD, cell-specific productivity qp over the entire culture period can be calculated as follows: qp ¼

P IVDC

(5)

where P is the product titer. Based on the growth and production profiles (Fig. 2), Fig. 3 illustrates a typical relationship between P and IVCD with their slope being qp, which, in this hypothetical example, is 14 pcd. Obviously, this is under the assumption that the cell-specific productivity is constant throughout the culture process. Both qp and IVCD are determined by the genotype (genetic constitution of cells) and the environmental influence. The genotype can be controlled during cell-line generation, while bioreactor design, medium composition, control strategy, and process conditions determine the culture environment. If the cell-specific productivity for a certain culture period is the interest of study, it can be calculated as follows:   1 DP P2  P1 2 ¼ qp ¼ (6) Xave Dt t2  t1 X2 þ X1 q0 p is useful to evaluate how cell-specific productivity varies at different phases of the cell culture process and support kinetic analyses to characterize cell growth and product formation.

1.19.2.2

Bioreactor and Process Control

Stainless-steel stirred-tank bioreactors are commonly used in batch culture manufacturing of biologics. Although mature platform technologies for mammalian cell culture have been well established, the optimal process conditions can be quite different because each cell line has its own unique characteristics. A comprehensive list of these conditions should include medium, hydrodynamics, culture vessel, temperature, DO, etc. and these conditions typically need to be specified and are crucial for a particular cell line. It is

Modes of Culture/Animal Cells

279

600

−1

lgG titer, p (mg l )

500 P = 14IVCD 400 300 200 100 0 0

Figure 3

10

20 30 IVCD (1e6 cells day/ml)

40

50

Correlation between product titer P and IVCD in a typical batch culture.

difficult to develop a generic process for all cell lines. Therefore, important process parameters for each cell line need careful studies to achieve high process performance with great product quantity and quality. Table 2 lists important parameters that need to be monitored either online or offline in a cell culture process. To facilitate product-development and manufacturing processes, many new techniques have been developed, among which online process analytical technologies (PATs) become more and more popular. Industries have been urged by regulatory organizations to adopt PAT in order to control process variations and build more repeatable, robust, and even operatorindependent processes. Although a daily offline sampling is popular in industrial operation practice to collect essential parameters, such as cell density, nutrition and metabolite concentrations, oxygen consumption rate, and product titer, the real-time monitoring and control of these parameters would be realized by online measurement and closed-loop feedback controls. Time/labor and contamination risk associated with sampling can be largely reduced. Real-time growth quantification in bioreactors is one of the most important online PATs, so a robust and reliable online detection system easily adapted to most cultivations is highly desired. Optical density and the capacitance of the culture are probably the easiest to detect, but their sensitivity is low especially during the

Table 2

Parameters monitored in bioreactors

Detection

Parameters

Comments

Usually online

Temperature

Controlled with a water jacket or heating blanket; typically at 37 C. Lower temperature can depress cell metabolism, reduce cell growth and death rates, and improve cell specific productivity (qp). Cells die  rapidly at > 42 C in general. Controlled by base addition and CO2 sparging; typically at 7.0. Higher or lower pH can adversely affect cell growth, cell-specific productivity, product quality, and lactate production. Controlled by O2/air sparging; typically at 20%–50%. High or low DO can change cell growth, cellspecific productivity, and product quality. Key parameters to identify cell growth, measured by hemocytometer-based trypan blue exclusion methods or automatic cell-counting instruments. Measured by ELISA, HPLC, etc. Measured by SDS-PAGE gel, enzyme reactions, HPLC, and mass spec with focus on glycosylation and other protein properties. Main carbon/energy source. Usually several grams per liter in the culture medium. Glucose limitation causes apoptosis, but high glucose concentration can also inhibit oxidative metabolism, lead to high lactate and alanine production, and inhibit cell growth. Main carbon/energy and nitrogen sources, contributing to ammonium production. Normally it is maintained at a low level. Its requirement can be eliminated for cells such as CHO K1 with glutamine synthetase systems. As a potentially toxic compound, it increases osmolality, decreases pH and causes adverse effects on cell growth and productivity. It is produced with high glycolysis rate. As a potentially toxic chemical, it permeates through cell membrane, disrupts local pH inside cells, and inhibits cell growth and productivity. It is produced after glutamine utilization and decomposition. Usually 280–350 mOsm/kg. High osmolality inhibits cell growth but can increase cell specific productivity. CO2 level higher than 15%–20% can cause growth inhibition and low cell specific productivity, while a certain level of CO2 (at least 0.5%–1%) is required for cellular functions such as fatty acid synthesis. It can also provide pH buffering. Removal of CO2 is mainly controlled by air and O2 sparging in bioreactors.

pH Dissolved oxygen (DO) Usually off-line

Cell density/viability Productivity Biologics quality Glucose Glutamine Lactic acid Ammonium Osmolality CO2

280

Modes of Culture/Animal Cells

death phase of culture. Newly developed in situ microscopes have proved to be suitable for cell mass measurement, whereas the acoustic system is still under development and not available even for academic uses. Besides cell density, nutrients such as glucose and glutamine and metabolites such as ammonium, lactate, and CO2 can also be monitored using online instrumentations. Raman spectroscopy has been used as an online sensor to detect process-related analytes and even protein conformation of therapeutic products because of its robustness and well-resolved footprints for different medium components. An alternative of online monitoring is flow injection analysis (FIA) based on an automatic sampling and quick detection system. Detection results such as glucose and glutamine concentrations from FIA can return to the control loop to adjust process parameters such as feeding rate and, thus, it provides a possibility to eliminate all human labor except initiation. Disadvantages of FIA are low sensitivity and selectivity, which are hard to meet the control requirement.4 Using disposable bioreactors with sterile bags, instead of stainless-steel vessels, is a new trend for process performance improvement because disposable bioreactors do not require time/capacity consuming cleaning in place (CIP) and sterilization in place (SIP) systems. Therefore, the usage of stainless piping and steam can be eliminated, and the batch turnaround time, plant space, and capital expenses can be greatly reduced. At the same time, it can eliminate cross-contamination from batch to batch and increase process safety. Although sizes of disposable bioreactors (up to 2000 L) are not as large as industrial large-scale stainlesssteel bioreactors (up to 20,000 L), they are more time and cost efficient in early stages of drug-development campaigns, such as animal experiment and clinical trial I/II. They can also be used in seed trains for large-scale bioreactors. High-throughput (HTP) media and process development allows wide and in-depth exploration of the design space. However, due to the complexity of cell culture processes, it requires right tools to be feasible for completing the screening and optimization experiments within a competitive time frame. SimCell MicroBioreactor Array is an excellent example of HTP media and processdevelopment tools with the capability of handling a large number of fully controlled microbioreactors. It can detect total cell density, pH, and DO and is operable in both batch and fed-batch culture modes.

1.19.3

Fed-Batch Culture: Dominator of Industrial-Scale Processes

In fed-batch cultures, on top of the basal medium a concentrated feed medium is provided intermittently to meet nutritional needs of the culture for an extended production period, achieving higher production (up to 10 g/L) than batch cultures. Fig. 4 illustrates the typical fed-batch culture kinetics, as characterized by the profile of glucose concentration in the culture medium that bounces depending on the balance of cellular consumption and the feeding schedule. In general, cell growth in a fed-batch culture is characterized with a higher VCD, a longer stationary phase and, thus, a larger IVCD than those in batch cultures. With the increased production capacity, product titer is also higher. This is why fed-batch culture is now most widely used in the biopharmaceutical industry. Increasing cell density and culture time can achieve a high productivity that is up to 10 times of the corresponding batch culture. Nevertheless, like batch cultures, a fed-batch culture must be terminated when it enters into the death or decline phase caused by the increasingly detrimental culture environment attributed to high osmolality and toxic waste levels. The success of a fed-batch culture, thus, depends on the feed medium formulation and feeding strategy, which require a good understanding of the culture kinetics and stoichiometry of nutrient utilization in support of cell growth and product formation. The development of a fed-batch process typically includes the optimization of feed medium composition, feeding strategy, and control of bioreactor operation parameters at different growth stages.

1.4

20 Fed-batch culture

1.2 VCD Glucose

1.0

IgG

12

0.8 0.6

8

IgG (g l−1)

VCD (106 cells/ml) Glucose (g l−1)

16

0.4 4 0.2 0

0.0 0

Figure 4

1

2

3

4

5

6

7 8 Day

9 10 11 12 13

Typical fed-batch culture kinetics with viable cell density (VCD), glucose concentration, and IgG titer profiles.

Modes of Culture/Animal Cells 1.19.3.1

281

Feed Medium Design

Medium development for a fed-batch process typically includes three aspects, the formulations of basal medium and feed medium, and the optimization of feeding strategy. For the detailed selection of an appropriate basal medium for a specific cell line, the reader is referred to Design of Culture Media (00028). Fortification of basal medium focuses on the prevention of nutrient depletion by enriching medium components, but excessive levels of enrichment for many nutrients impede cell growth directly or contribute to potential toxic waste accumulation. These growth-inhibitory byproducts including lactate and ammonia are regarded as the main root causes of cellular stress and the associated apoptosis in fed-batch culture, affecting product quantity and quality adversely. Feed supplementation, by feeding nutrients gradually to the culture and maintaining nutrient concentration at a low level during the course of the culture, is believed to be one of the main driving forces for the rapid advance of biopharmaceutical production. Feeding cannot only maintain sufficient nutrient supply but also prevent a high initial nutrient concentration that can lead to excessive toxic metabolite accumulation. Thus, a leaner basal medium plus nutrient feeding generally provides a more beneficial environment to cell growth and biologics production. In consideration of timeline, labor, and potential improvement, more optimization efforts are commonly spent in feed design, whereas basal medium is often chosen from off-the-shelf media in the market. One important factor to be taken into consideration in the feed design is the limitation of main energy sources, glucose and glutamine, and the accumulation of their relatively detrimental catabolites, lactate and ammonium. Fine tuning of glucose and glutamine addition to increase nutrient utilization efficiency has been reported to improve culture performance substantially.5 In general, lactate and ammonium are two main inhibitory byproducts and their buildup can be minimized by applying optimized time/rate of feed administration to sustain low glucose/glutamine levels. Identifying additional toxic waste byproducts and understanding intracellular factors regulating product synthesis and secretion are also important to the feed design. In addition to glucose and glutamine, other medium components such as lipid and phosphate (key substrates to form cell membrane) may also have to be replenished throughout a fed-batch culture.6 A multicomponent feed medium may be simply the concentrated basal medium excluding most of the salts, which can be implemented reliably if the limited medium components are not identified within a complex composition such as hydrolysate. Many commercial generic feed media belong to this category and they have been widely used in early-stage production, because they can quickly improve culture performance. However, the drawbacks are also obvious. The addition of nonrate-limiting components to excessive levels causes the risk of growth and production inhibition resulting from high-concentration suppression and byproduct buildup. It can also increase osmolality to a stress-inducing level, and it is not cost efficient.5 The development of cell line-specific feed formulation through iterative nutrition analysis and design of experiment (DOE) can potentially overcome these issues. Besides supplementation to prevent nutrient depletion and byproduct accumulation, feeding some specific agents can inhibit apoptosis or induce the desired metabolic pathway for protein production after reaching a high cell density. Apoptosis is generally the main cause of cell death and culture termination in production. Apoptosis inhibitors such as z-VAD-fmk, suramin, polysulfated compounds, or protein inducers can be applied directly to cell culture to increase culture longevity and to achieve a high cell density.5,7 As a popular productivity enhancer in industrial bioprocess, butyrate increases the accessibility of DNAase to chromosome DNA by hyperacetylating histone and, thus, enhances protein-specific productivity. However, it induces cellcycle arrest and inhibits cell growth. In this case, butyrate is often introduced at the late exponential growth phase to allow rapid growth at early stage and benefit production at late stage. The overall influence of these supplements on production performance varies depending on cell line and particular process strategy, and the dose/timing of their introduction need to be fine-tuned in order to achieve the best benefits. To attain the highest productivity, most of the large-scale fed-batch cultures use feed medium designed through iterative spent medium analysis and nutrient resupplementation. The principle in customizing a feed medium for a particular cell line is to supplement limiting nutrients by spent medium analysis in order to sustain all nutrients at a relatively constant level, which should be low but sufficient to support cell growth and protein production. Despite various strategies developed for feed medium optimization, a general procedure is suggested. A feed medium is formulated with an appropriate composition of depleted nutrients based on the analysis of collected spent medium and consumption rate calculation. This medium is used in the next culture followed by reiterative analysis and resupplementation as further refinement. Generally, a few repeats are enough to achieve most of benefits from feed medium optimization.7 In this approach, spent medium analysis requires the detection of residual nutrients to determine nutrient utilization and then the feed medium can be formulated with the desired ratio of individual components. Although most components in the feed medium can be detected by analytical technologies such as mass spectrometry, precise measurement of some medium components such as vitamins, lipids, and trace elements remains a challenge.5 In this case, parallel experiments in multiwell plates or shake flasks can be implemented with expertize reasoning and DOE principles to identify limiting nutrients. Another approach to simplify feed medium development is to balance feed components based on the stoichiometric ratio of biomass and protein product while maintaining low levels of glucose and glutamine throughout the culturing period.4 This approach is based on the theory that there is an ideal scenario that maximizes nutrient utilization efficiency through essential biological pathways and minimizes byproduct generation by bypassing nonessential pathways. Under this scenario, the consumption rates for various nutrients (other than glucose and glutamine) can be estimated from their contents in cell biomass and protein product. Formulation by this approach may be different from the final optimized medium, but it serves well as the baseline for further refinement. Xie and Wang8 largely reduced the generation of lactate and ammonia by designing a feed medium with a stoichiometric equation to control the hybridoma cell culture

282

Modes of Culture/Animal Cells

nutritional environment, and at the same time achieved 5 times higher cell density and 10 times higher monoclonal antibody production compared with conventional batch culture. Pfizer recently published their development of a single chemically defined nutrient feed (CDF) to replace a hydrolysatecontaining nutrient feed (HCF).9 In their work, they took a top-down approach by three major development steps: (1) starting with an overrich nutrient feed; (2) removing unnecessary components and reducing overfeed components; and (3) polishing and formulating. The developed CDF increased recombinant IgG titers by 115% for a fed-batch NS0 process and by 80% for a fed-batch CHO process, respectively, compared with their original fed-batch process using HCF. In another example, Genentech employed a systematic approach to develop a chemically defined platform feed that can support IgG production up to 4 g/L. Through iterative feed optimization and sophisticated feeding strategy development, an enhanced chemically defined feed was developed to boost IgG titers to around 8 g/L.10 Feed components must be dissolved in one or multiple small-volume solutions because substantial expansion of culture volume is not desirable. Concerns of feed medium preparation and storage are related to the solubility and stability, respectively. Adjustments of pH and temperature might be necessary to dissolve some components, and the combination of several concentrated solutions instead of adding all components into a bulk solution can be helpful. Because the feed medium is highly concentrated, some components may fall out of solution at low temperatures and, thus, sometimes the feed medium has to be kept at room temperature. This can result in an extremely short shelf life of a few days instead of months, making manufacturing schedule difficult. However, dry powder-based media with one solution and auto-pH are currently commercially available, such as Advanced Granulation Technology from Invitrogen (Carlsbad, CA), largely prolonging feed medium shelf life and reducing storage space.

1.19.3.2

Feeding Strategy

Once feed medium composition is optimized, proper timing/rate of feed administration ensures the delicate control on the nutrient concentration to prevent byproduct buildup and high osmolality while avoiding adverse consequences of lacking essential nutrients. Although there are many sophisticated systematic process algorithms to optimize feeding strategy, most of them can be classified into two categories: open- and closed-loop models, depending on whether nutrient consumption is either trajectory from accurate mathematical simulations or actually measured. In open-loop systems, dynamic programming, applying optimal control theory based on the understanding of cell growth and protein production kinetics, predicts the optimal control strategy without any feedback control.5 To eliminate the requirement of building models based on the very complicated cellular metabolism, closed-loop systems rely on either direct or indirect online measurements providing guidance for a series of feeding actions. Different PATs as discussed earlier are well suitable for this purpose. In-process real-time monitoring coupled with simple feedback algorithm controls the concentrations of crucial nutrients and metabolites by orchestrated nutrient supplementation. For example, the accumulation of inhibitory byproducts, lactate/ammonium, can be minimized through dynamic feeding of glucose/glutamine to keep these substrates at low levels according to the feedback from online monitoring.11 At the same time, the supply of other nutrient components from the feed medium with stoichiometric constitution is well controlled and their concentrations in culture are maintained at relatively constant levels. In case that reliable direct online sensors for these medium components are not available, indirect online measurements of some process parameters, such as pH, DO, CO2, turbidity (cell mass), and even base addition, can be used for the estimation of cell growth and nutrient consumption; and in conjunction with stoichiometric ratios they can be effectively used to develop a good feeding time/rate strategy.4 Zhou et al.12 used online oxygen uptake rate measurement to estimate nutrition consumption rate in order to determine the feeding rate of a stoichiometrically balanced feed medium, which resulted in a low accumulation of metabolites, including lactate, ammonia, and alanine, and an improved sustention of a high cell density over an extended period. However, due to its high complexity, dynamic and frequent feeding is not preferred in large-scale production. Simple offline sampling and empirical addition of daily bolus are a widely used feeding strategy in industry. Sauer et al.13 developed a generic process for Sp2/0 and NS0 antibody-producing cell lines using protein-free feed medium and a metabolically responsive feeding strategy based on offline glucose measurement. The integral of viable cell count (IVCD) was 4.3 times higher and qp was 1.7 times higher compared with the corresponding batch cultures.13 To improve culture performance and reduce developmental time, statistical DOE using small shake flasks is commonly implemented to study the dose and timing of nutrient introduction for feeding process optimization, followed by small-scale bioreactor culture for refinement and proof of concept.

1.19.3.3

Process and Bioreactor Control

Although feed medium design and feeding strategy can be optimized with less-defined systems such as multiwell plates and shake flasks in cell culture incubators, optimization of environmental parameters such as pH, DO, and temperature has to be conducted in fully controlled bioreactors. The key here is the balance between cell proliferation and product formation. For example, a low-temperature shift after the first few days of rapid cell growth at 37 C typically inhibits cell amplification by cell-cycle arrest but potentially can increase cell-specific productivity and extend culture longevity.14 Initial cell growth prefers a higher pH despite the association of lactate buildup, whereas a lower pH can potentially improve protein production at the

Modes of Culture/Animal Cells

283

late culture stage. High osmolality impedes cell replication but can facilitate cell-specific productivity. Thus, stepwise osmolality increase by feeding can be beneficial to culture performance. Orchestrated control of essential culture parameters to manipulate cell cycle and cell metabolism has proved to be valuable. Their main effects and interactions are typically studied in small/microscale bioreactors with statistical design approach such as partial fractional factorial design.14 SimCell, as a pioneer HTP technology allowing the control of key culture conditions and serving as an initial step of streamline process development, accelerates the establishment of reliable and high-yield production processes. Fed-batch process optimization is often considered in a cell line-dependent manner because of specific clone metabolism. Although some well-developed biopharmaceutical companies have established their own in-house platform technologies by generating high producer clones for different products from the same parental line using similar cell line generation procedures, and thus a generic feed and feeding strategy can promote production performance of most derived cell lines, further optimizations on a clone-specific base usually are beneficial.7 All discussions above focus on fed-batch process optimization to improve product titer, but from a regulatory viewpoint, consistent product quality is even more important than productivity. Thus, product quality is often analyzed starting from the very early development stage in many aspects, including glycosylation and other protein properties such as fragmentation, aggregation, and charge variance. Glycosylation can significantly influence clinical safety and efficacy such as circulatory halflife and/or immune responses of many enzymes, hormones, and mAbs.14 Both host cell line and culture conditions such as medium, glucose, DO, CO2, ammonia, and osmolality can alter glycosylation. In addition, high cell-specific productivity is a potential reason for immature forms of glycosylation, and extended culture duration may elicit heterogeneous and truncated oligosaccharides resulting from glycosidase released from dead cells. In certain situations, the highest productivity has to be sacrificed in favor of improving quality by shortening culture period and adopting processes with lower specific productivity. With a fed-batch process, Mab production from an NS0 cell line was improved by 10-fold for an extended period, but product charge variations and higher percentages of truncated and high mannose glycoforms appeared as the culture was extended.15

1.19.4

Perfusion Culture: The Most Sophisticated Process

A continuous culture provides fresh medium in a nonstop fashion to the culture vessel as cells continuously utilize nutrients and there is also a continuous harvest of medium to maintain the same culture volume in the bioreactor. Continuous culture is characterized with steady states for VCD, nutrient and metabolite levels, and product titer, respectively (Fig. 5). Depending on whether there is cell retention as medium flows through, continuous culture can be categorized as chemostat culture, where cells are washed away with medium, and perfusion culture, where cells are retained or recycled through a certain device inside or outside of the culture vessel. Chemostat culture can only be operated with a dilution rate less than cell growth rate in order to avoid washout. Chemostat is a valuable tool to study the fundamentals of animal cell culture by allowing precise metabolic-flux analysis, but it is not a commercial process because of low product titer and poor process stability.1,4 Therefore, continuous culture with cell retention/recycle, which is usually called perfusion culture, is developed to achieve high cell density and high volumetric productivity.

0.7

30 Perfusion culture

0.6

Glucose

0.4

IgG

15

0.3 10

IgG (g l−1)

0.5

VCD

20

0.2

26

28

24

22

18

20

16

14

10

12

6

0 8

0 4

0.1

2

5

0

VCD (106 cells/ml) Glucose (g l−1)

25

Day Figure 5

Typical perfusion culture kinetics with viable cell density (VCD), glucose concentration, and IgG titer profiles.

284

Modes of Culture/Animal Cells

1.19.4.1

Kinetics

The dilution rate is an important operation parameter for the continuous culture, which is defined as D¼

Q V

where Q is the volumetric flow rate and V the bioreactor working volume. For well-controlled perfusion culture, on the one hand, there is cell growth with the continuous feed, and on the other hand, there is cell bleeding for cell density control through harvest line and cell discard line. Therefore, cell density change can be expressed as follows16: dX CDR ¼ m$X  D$XH  $X dt V where CDR is the cell discard rate and XH the cell density in harvested medium, which is close to zero with cell retention or recycle. As a perfusion culture is usually operated at pseudosteady-state, the target cell density can be determined with the following equation16: X¼

D XH m  CDR=V

A useful variable developed for high-density perfusion culture is cell-specific perfusion rate (CSPR) defined as CSPR ¼

D X

CSPR is usually measured in the unit of nL/cell day, representing the volume of feed medium per cell per day. CSPR is a useful linker for important perfusion process variables, such as titer or product concentration (p), specific productivity (qp), and volumetric productivity (PV)16: P¼

qp CSPR

PV ¼ P$D ¼ X$qp Due to nutritional limit with a certain amount of feed medium per day, there is a minimum value for CSPR, indicating the minimal medium needed for 1 cell per day. A closely related concept with CSPR is the maximum number of cells that can be supported for 1 day with a unit amount of medium, which is termed as medium depth (MD) in the unit of cells day/mL16: MD ¼

1.19.4.2

1 Xmax ¼ CSPR min D

Perfusion Culture Processes

A perfusion culture system typically includes a bioreactor, a medium tank for continuously supplying feed medium to the bioreactor, a harvest tank for continuously receiving spent medium from the bioreactor, and a cell retention system.4 A crucial and unique component in a perfusion culture process is the cell retention device, which should operate efficiently with a high cell retention rate for the entire culturing period. It should not cause cell damage or affect any cell function such as protein productivity. Also, it is preferred that the system can selectively retain viable cells and remove dead cells and debris. Various cell retention systems have been employed and are discussed separately for adherent cells and suspension cell cultures. For adherent cells, adhesion to a solid support is commonly used to retain cells inside the bioreactors. For nonadherent cells, cell retention is achieved through either entrapment in a polymeric material or by using a membrane. Some commonly used cell immobilization and retention methods in perfusion cultures are discussed below.

1.19.4.2.1

Cell Retention Through Immobilization

For cells with difficulties to be adapted to suspension culture or with requirement of attachable surfaces to maintain cellular physiological properties such as embryonic stem cells and many cell lines for vaccine production, cell immobilization on or within a support matrix is necessary. Generally speaking, immobilization of cells is to restrict cells into a defined compartment inside of the bioreactor through cell physical entrapment or adhesion. With immobilized cells retained inside the bioreactor, spent medium can be easily drawn out. Microcarrier suspension culture in a stirred-tank is probably the most commonly used technology to scale-up adherent culture and it can retain cells easily. Microcarriers are typically small particles with biocompatible surface favorable for anchorage-dependent cell attachment and growth. Microcarriers provide a high surface-to-volume ratio for anchorage-dependent cells and ease the precise process control in large-scale bioreactors. Attached to microcarriers, anchorage-dependent cells can, in most cases, be treated as suspended cells in bioreactors except that cells are more sensitive to external collision and shear stress.

Modes of Culture/Animal Cells

285

Macroporous microcarrier beads are suitable for shear-sensitive cells by housing cells inside beads, but this advantage is countered by the inefficiency of intraparticle diffusion. Hollow fiber bioreactor (HFB), as both a culture vessel and a retention device, allows the permeation of nutrients and metabolites through ultrafiltration capillary fibers and retains cells and proteins and can reach a high cell density and high protein productivity and titer.17 However, due to its practical defects including pressure drop, nonhomogeneous nutrient/waste/ oxygen concentrations, and the absence of supportive techniques to monitor cell growth, its application is limited to small-scale preclinical production. Similar to HFB but with a better scale-up capability, the fibrous bed bioreactor (FBB) consisting of a 3D fibrous-packed bed inside a bioreactor was developed for culturing mammalian cells and tissues, including CHO, human trophoblast, osteosarcoma, colon cancer cells, breast cancer cells, and embryonic stem cells.18 Because the fibrous bed bioreactor has the advantages of high porosity, high surface-to-volume ratio, high permeability, low-pressure drop, easy downstream processing, and low material cost, the perfusion culture system based on it can support massive cell growth and sustain longterm cell viability and functional activities.18 However, direct quantification of immobilized cell density is still impractical during the course of a culture, making process characterization very difficult, although a novel online fluorescence probe was recently developed and validated in a lab-scale perfusion fibrous bed system with cells engineered to express green fluorescent protein, which quantified cell growth and protein production noninvasively.19 Polymerization and encapsulation are two cell immobilization methods that can be used for both adherent and nonadherent cells.4,20 Cells suspended in a polymeric solution can be fixed upon polymerization, and polymeric beads with entrapped cells are then formed by applying a physical pressure or emulsion. The major drawback of polymerization is low diffusion efficiency resulting in a suboptimal microenvironment for high-density cell growth. In encapsulation systems, cells are restricted into the hollow capsule surrounded by a semipermeable membrane. Nutrients and metabolites easily exchange across the membrane, while cells and large molecules such as proteins are retained. Compared with polymerization, liquid environment inside capsules has better diffusion efficiency and protein product is concentrated within capsules. However, high proteolysis is possible due to protease accumulation. Although these two methods have been widely used in small-scale cell culture research, especially in the field of tissue engineering, they are not used in cell culture manufacturing process because of the complexity and potentially high cost involved. Cell immobilization through adhesion on a solid surface or entrapment within a membrane or polymeric matrix creates a heterogeneous culture environment, which is desirable in some emerging cell culture applications, especially in the fields of stem cell, tissue engineering, and regenerative medicines.21,22 Proper cellular properties often require the interaction between cells and solid surface, so mass production of immobilized cells and functional tissues to support cell therapy and transplantation are indispensable. However, cell immobilization in perfusion culture is falling out of favor in industrial-scale biologics production due to its spatial variation in nutrient/waste distribution, poor oxygen transfer, difficulties of microenvironment control, and uneven inoculation. With increasing successful adaptation to suspension, most of current industrial cell lines can be cultured in suspension without a solid support.

1.19.4.2.2

Cell Retention Systems for Suspension Culture

In perfusion cultures with cells in suspension, cells (particles) can be separated based on differences in particle size, density, or both. Cell retention devices in homogeneous suspension culture generally can be classified into three categories – filtration, sedimentation, and centrifugation – which are discussed on their separation principles and factors affecting their retention performance. The placement of the cell retention device can be either external or internal to the bioreactor. Internal devices are usually preferred considering mechanical robustness and simplicity, but the replacement of the device is impractical when maintenance is required in the middle of cultivation, which can be up to a few months. External retention devices with cells pumped through a recirculation loop can be replaced when a fouling problem or mechanical failure happens, but the replacement requires extra care, increasing contamination risk and raising regulatory concerns. Also, as cells pass through these physiological suboptimal environments, nutrient depletion and oxygen starvation are likely to happen and cause cell damage. In this case, when using external devices, gentle cell handing and short external residence time should be considered to reduce cell damage and minimize the risk of failure.23 The merits and limitations, especially in scale-up potential for industrial applications, of different cell retention systems are discussed below. Cross-flow membrane filters, including hollow fiber cartridge and flat plate, are used as external retention filter modules. To reduce fouling of filters, cell suspension is pumped to flow tangentially to a membrane and cells are concentrated, leaving the permeate flow across the membrane (Fig. 6A). The concentrated cell suspension returns to the bioreactor, while the cell-free permeate is collected. Factors affecting performance include membrane pore size, permeate flux, and cell suspension flow rate. In general, increasing the membrane pore size also increases the perfusion capacity but decreases cell retention efficiency. A fast permeate flux demonstrating a high perfusion capacity can lead to a high rate of fouling and shorten the filter life. A high suspension flow rate with a high shear rate reduces the accumulation of solid particles on the filter surface but can increase pressure drop, elicit shear-induced damages, and affect viability adversely. It is important to control these three parameters in order to achieve a reliable long-term operation without filter change, maintain a benign culture environment, and at the same time increase perfusion capacity. Cross-flow microfiltration can supply cell-free permeate, which eliminates the step of primary recovery and allows the direct integration with downstream purification. However, despite all the efforts to optimize its performance, fouling problem leading to early culture termination happens frequently. In addition, nonhomogeneous filtration conditions are inevitable due to pressure drop. All reported perfusion processes with cross-flow filters are at small scale mostly due to

286

Modes of Culture/Animal Cells

A

B

C

D

Feed stream Cell suspension

Effluent stream

Effluent stream

Diaphragm pump

Membrane

Inlet stream

Effluent stream

Feed stream

Harvest

Inlet stream

E

Spinfilter

Hollow fiber module

Recycle stream

F

G Effluent stream

Effluent stream

Inlet stream

Feed stream

Recycle stream

Feed

Overflow=harvest

Ultrasonic resonator

Inclined setter Feed stream

Vertical setter

Inlet stream

Recycle stream

Primary vortex

Secondary vortex

Underflow

Figure 6 Cell retention devices for suspension culture: (A) cross-flow filtration; (B) ATF; (C) spin filter; (D) vertical sedimentation; (E) inclined sedimentation; (F) ultrasonic resonator; (G) hydrocyclone. (A, G) Reproduced from Voisard, D.; Meuwly, F.; Ruffieux, P. A.; et al. Potential of Cell Retention Techniques for Large-Scale High-Density Perfusion Culture of Suspended Mammalian Cells. Biotechnol. Bioeng. 2003, 82, 751–765. (C–F) Reproduced from Woodside, S. M.; Bowen, B. D.; Piret, J. M.; et al. Mammalian Cell Retention Devices for Stirred Perfusion Bioreactors. Cytotechnology 1998, 28, 163–175.

the difficulty of scale-up, because perfusion capacity, which increases linearly with filter surface area, has to be scaled up with bioreactor size proportionally. The high area required to maintain a constant area-to-volume ratio in large scale implies a physical constraint in providing the necessary filtration surfaces and pumping capacity in the context of industrial manufacturing processes.20,23 One exception is alternating tangential flow (ATF) from Refine Technology (Edison, NJ) with a fast diaphragm pump, which is mounted to one end of a hollow fiber module, providing a repeated reversible flow back and forth to reduce fouling of the module (Fig. 6B). The theoretical scale-up capacity is up to 1200 L/day. In a typical perfusion culture, the product is not kept inside the bioreactor during perfusion. However, with ultrafiltration, Percivia (Cambridge, MA) cultured human PER. C6 cells with XD process, which used Refine‘s ATF to retain both cells and product upon perfusion. A product concentration of 40 g/L with over 108 cells/mL was achieved. The productivity is probably the highest today, and the packed cell volume is close to the level that microbial culture can achieve. Spin filter was the retention device reported to support the first successful perfusion mammalian cell culture (Fig. 6C). A perfusion bioreactor with a spin filter as the retention device normally consists of a typical stirred-tank bioreactor and a cylindrical stainless-steel mesh or membrane. Spent medium crosses through the filter into the cylinder and is pumped out of the bioreactor. Perfusion flux, cell density, filter material, rotational rate, and pore size mainly determine the filter performance. It was reported that there were two stages of fouling with the first appearing right after the permeate flux became larger than perfusion capacity and the second occurring after a few weeks.24 Stainless-steel mesh was considered more fouling-inducing owing to their high surface charge density in comparison with some polymer materials. However, stainless steel is still preferred because it is nondegradable, autoclavable, and easy to clean.20 Perfusion capacity was reported proportional to the square of tangential speed, because high shear rate generated by rotation reduces fouling, thus increasing perfusion capacity.24 Pore size of spin filters can be larger than cell diameter, but the balance between retention efficiency and perfusion capacity requires careful manipulation. Perfusion capacity is normally higher with spin filter compared with cross-flow filters, and transmembrane pressure is also more uniform on the filter surface, providing a more controlled environment. It is usually an internal device and, thus, more cell growth friendly. However, fouling is still the main concern and scale-up becomes difficult when requiring very large filter surface of spin filters in large scale. Avgerinos et al.25 used an ethylene-tetraflouroethylene spin filter for CHO cell retention in a 20 L stirred-tank bioreactor for recombinant urinary type plasminogen activator production and maintained a high cell density of 60–74 million cells/mL at a perfusion rate of 3–4 bioreactor volume per day. Fifty-one grams of the product were produced in 1000 L medium through a culture period of 31 days without filter fouling.25 Commercially available devices include ESF

Modes of Culture/Animal Cells

287

100 G/200 G external spin filters using stainless-steel mesh for 2–50 L bioreactors and disposable spin filter p using polyethylene terephthalate polyester (PETP) fabric mesh, both of which are from Sartorius-BBI-Systems Cooperation. Vertical sedimentation exploits gravity to settle and separate cells that have a higher density than medium in a vertical counterflow, while drawing out cell-free spent medium on the top, in which a relatively quiescent liquid environment is required for cell settling (Fig. 6D). The maximum perfusion capacity is theoretically the product of the conical top area and the average settling velocity determined by the viable cell size and density. As there is no screen or filter for cell/medium separation, fouling and shear damage are no longer an issue in the sedimentation device. Also, because larger live cells settle faster than smaller dead cells and debris, it improves the selective recycle of live cells back to bioreactors and forms a steady state with higher cell density and longer cultivation time by avoiding dead cells and debris accumulation. However, because the single cell settling velocity is low, a large diameter of the cone top is required, which becomes a problem in large-scale bioreactors. The low recycle rate in a relative large volume retention chamber also causes a prolonged residence time (up to hours) under detrimental culture conditions without oxygen/nutrition supply and well mixing.4,20 Inclined sedimentation uses Boycotts effect that sedimentation can be enhanced by inclined tubes to improve settling efficiency and scalability (Fig. 6E). Cell suspension flows up between inclined parallel plates and cells settle on the sediment plate through a short distance. Clarified flow is removed from the top outlet and cells on the sediment plate slide down back to the bioreactor. Increasing the width and the number of plates improves the maximum perfusion capacity, which is theoretically the product of cell settling velocity and horizontal device area. The more compact the plate stack is, the sooner cells reach the plate surface and the shorter the length of plates can be. Therefore, a compact multiplate device formed by short and wide parallel plates is adopted during scale-up to reach a much higher perfusion flow rate compared with vertical settlers. However, although cells can reach sediment plates very soon, cells take a long period to move back to the bioreactor in a countercurrent flow, which affects cells adversely. In addition, it is susceptible to mechanical failure associated with fouling because debris and cells attach to plates and form sediment layers. Several approaches including intermittent settler vibration and cooling have been applied to increase the fallback of settled cells, reduce cell attachment, and protect cells from damages.20 Commercially available incline settlers from Biotechnology Solutions are up to 500 L. An ultrasonic resonator can increase settling efficiency by forming loose aggregates in an acoustic standing wave field based on the differences in density and compressibility between suspended cells and medium, and disaggregating during the recirculation to the bioreactor (Fig. 6F). Both high cell density and ultrasonic wave input power can allow high perfusion flow rate and increase separation efficiency. Similar to vertical settlers, failure associated with fouling during long-term culture is not likely, and the cells residence time in the device is much shorter than regular vertical devices. Potential problems are related to cell cavitation in ultrasonic field when using a high power.20,23 Also, the optimization to avoid heterogeneous field distribution and to remove heat produced is required in scale-up. The BioSep Perfusion from Applikon (Schiedam, Holland) is currently available with a capacity up to 200 L/day. An acoustic cell recycle system with the ability to selectively retain viable cells was used to support a perfusion culture over 700 h, maintaining a cell density higher than 50 million cells/mL with a 5 times higher product titer and a 70 times higher volumetric productivity compared with batch culture.26 Continuous centrifuges with rotating mechanical seals between the stream lines and the rotor overcome the problem of low settling velocities in gravity settlers.20,23 Like in gravity settlers, live cells can be separated from dead cells and recycled back to the bioreactor based on density difference. Centrifuges are very effective cell retention devices for large-scale bioreactor with a dynamic retention efficiency controlled by the centrifugal force, a compact size, and an inlet stream rate close to the perfusion rate. Thus, they are easy to scale up in comparison with other retention devices. The main concern is their complexity, reliability, durability, and their detrimental effects due to high shear stress. Today, there are several centrifuges in the market for industrial applications, including CentriTech Cell II Centrifuge using disposable sterile insert with a manufacturer-rated capacity of 120 L/h. Hydrocyclone is another centrifugal sedimentation device separating solid particles from liquid medium based on density differences (Fig. 6G). A hydrocyclone usually consists of an upper cylindrical part and a lower cone section. After cell suspension is injected to the cylinder tangentially, concentrated cells are drawn from the bottom while clarified liquid exits from the top. Performance and efficiency are mainly determined by inlet pressure, device configuration, and cell density. The compact size and high perfusion capacity make hydrocyclone very promising for scale up, but high pressure drop and shear stress generated due to high flow rate lead to viability decrease.4,20,23 Further optimization of designs is necessary before its application in industrialscale cell culture production. There is no universal optimum cell retention device for the homogeneous cell culture system. Depending on particular aspects of each application, simplicity, long-term operation, perfusion capacity, robustness, and cultivation environment are generally considered in determining a practical choice. Size-based filtration devices such as cross-flow filters and spin filters easily achieve a high separation efficiency. However, although long-term culture is possible, membrane fouling and clogging limit their culture duration or cause the replacement of external parts. The rate of fouling depends on cell density, cell biological properties, membrane material, port size, and elution rate. By contrast, there are no reports showing termination of culture due to clogging for density-based open devices such as vertical or inclined settler, centrifuge, ultrasonic resonator, and hydrocyclone, but separation efficiency can be problematic, limiting industrial applications of these devices. The elution flow rate influences the loss of cells as a partial breeding stream and the achievable maximal cell density. Currently, only spin filter, ultrasonic resonator, centrifuge, and hydrocyclone have a perfusion capacity higher than 250 L with centrifuge standing out in scale higher than 1000 L. These open separators have a certain level of scale-up superiority, because they can proportionally increase the perfusion capacity by expanding 3D structure, whereas the capacity of all filters relies on 2D surfaces.

288 1.19.4.3

Modes of Culture/Animal Cells Scale-Up and Optimization

Although scale-up of a perfusion culture mainly relies on a highly efficient, robust, and scalable cell retention system, bioreactor vessel itself also needs to be taken into account for maximizing cell growth. The volumetric productivity of a perfusion bioreactor is generally much higher than fed-batch cultures, so the size of vessel is more compact, which can alleviate scale-up difficulties. Similar to large-scale fed-batch culture, mixing, shear stress, and aeration can pose challenges, while there are still some differences due to the unique properties of perfusion. Using aeration as an example, a higher kLa is necessary because a higher cell density also requires an increased oxygen supply, while the CO2 removal problem is mitigated with continuous medium replacement. Compared with fed-batch cultures, perfusion cultures using the same basal medium as perfusion medium require a larger volume of medium usage and have a lower product titer. If volumetric productivity is limited by nutrition depletion instead of toxic metabolite accumulation, and if the product is stable, a low perfusion rate with fortified medium can be used to solve these problems. In addition to fine-tuning the perfusion rate, the development of a fortified medium with enriched nutrients can also optimize perfusion culture performance because the product titer is usually linear with 1/CSPR. An unfortified medium at a high perfusion rate can be replaced by a fortified medium at a low rate. The replacement reduces medium cost and storage space, increases product titer in the harvest stream, and most importantly, it largely lessens the burden of the cell retention device. The development of an enriched medium is similar to the feed medium development in the fed-batch culture, except that some nonconsumable components such as inorganic salts that provide optimal osmolality should be maintained at a similar level as the basal medium, instead of eliminating them as in the feed medium for a fed-batch culture.4,27 Compared with batch and fed-batch cultures, long-term perfusion culture with a high cell density requires closer monitoring and more timely control because small parameter offsets or incidents can accumulate over time and lead to process deviations. In this case, new PATs are highly recommended to monitor online parameters and to control the process in real time with a feedback loop, allowing elaborate and timely controls such as dynamic perfusion rate adjustment. The controls are based on cell density, nutrition and metabolite concentrations, oxygen consumption rate, and product titer, as previously discussed in fed-batch culture. Highly automated process control is preferred due to the dynamic nature of perfusion culture. However, one thing to be pointed out here is that almost all published research is focused on controlling the perfusion rate by monitoring nutrient consumption to ensure cell growth without nutrient depletion, whereas few implementations of control strategy take full advantages of the merit of perfusion culture by adjusting flow rate to both deliver nutrients and maintain toxic metabolite concentrations lower than detrimental levels. This is mainly because it is difficult to control multiple concentrations simultaneously by adjusting the flow rate of a single perfusion medium. Theoretically, the overall productivity of a 2000-L disposable bioreactor operated in a perfusion culture can easily match or be even higher than a 20,000-L fed-batch bioreactor. Therefore, there are growing interests in using disposable bioreactors in large-scale perfusion cultures.

1.19.5

Concluding Remarks on the Selection of Culture Mode

The performance of various culture modes is illustrated in Fig. 7. The choice among different culture modes is generally made based on their respective advantages/disadvantages and practical considerations about product stability, cell line stability, capital expense, manufacturing capacity, and company expertize (Table 3).1,7 Since the 1980s, improvements in mammalian cell expression system, medium formulation, reactor design, and process operation have increased the cell culture productivity to more than 100-fold. Current large-scale production processes can achieve a peak cell density higher than 107 cell/mL and a cell-specific productivity of tens of pg cell/day (pcd). On the other hand, compared with the first generation of biotherapeutics, current products, especially monoclonal antibodies, often require a more than 10 times higher dosing.14 This, plus the increasing number of products approved and in clinical trials, leads to a large demand of animal cell culture capacity. Another aspect about biotherapeutics in clinical trial and market is that regulatory agents such as FDA tightly monitor and regulate their manufacturing with detailed guidelines of current good manufacturing practice, and specific inspection and validation of facilities. Batch culture is the most operable and reliable process with the lowest risk of contamination and mechanical failure due to its simplicity. It is also with the shortest cultivation time and a relatively short product residence time, which is helpful for product quality controls. The kinetics of cell growth, nutrition consumption, and metabolites accumulation are simple and well characterized. This age-old technique is still very popular in many industrial productions and the most frequently used approach to study cell growth, product formation, medium components, culture environments, and cell line stability, even if fed-batch or perfusion has been selected as the final production mode. However, culture efficiency is always low because of its low productivity. Also, products generated by different batches can vary in quality and concentration, which causes potential regulatory concerns and downstream difficulties. Therefore, fed-batch and perfusion as more sophisticated culture modes become popular in mammalian cell culture processes, whereas batch culture fades out in large-scale production runs.4 Fed-batch culture in a stirred-tank bioreactor is the most attractive choice in commercial therapeutic protein production owing to its high product titer and relatively high volumetric productivity while keeping the similar level of easy operation, reliability, scalability, and flexibility of facility design to batch culture. The main driver is the substantially increased IVCD by nutrition supplementation, reduction of byproduct accumulation, and control of cell physiological status to extend culture longevity. Besides

Modes of Culture/Animal Cells

289

A

30

VCD (1e6 cells/ml)

25 20 15 Batch Fed-batch Perfusion Subculture

10 5 0 0

5

10

15

20

25

30

Day B

8 7 Glucose (g l−1)

6 5 Batch Fed-batch Perfusion

4 3 2 1 0

0

5

10

15

20

25

30

Day

C

1400 1200

IgG (mg l−1)

1000 800

Batch Fed-batch Perfusion

600 400 200 0

0

5

10

15

20

25

30

Day

Figure 7 Comparison of cell culture kinetics under different culture modes: (A) viable cell densities (VCD); (B) glucose concentrations; and (C) IgG titers in batch, fed-batch, and perfusion modes.

IVCD, qp can also be improved after changing culture environment from growth-beneficial conditions to production-beneficial conditions at late stage. Considering the high titer and the supplement of only essential components, medium is used more efficiently with less waste produced. Similar to batch culture, cells stay in the production reactor for about 2 weeks after being scaled up in seed train, and the next production normally uses a new vial thaw instead of cells from the previous run. Therefore, generation number from the master bank is relatively low and well controlled, so the risk of failure due to cell line instability can be minimized. Fed-batch culture is easy to implement and scale up and requires less capital cost, developmental effort, and technical expertize in comparison with perfusion culture, so process to reach market is faster and cheaper with less process development, validation, and regulatory concerns. In addition, the flexibility of fed-batch culture allows a given facility to change the products manufactured easily and saves capital cost. However, the improved culture efficiency compared with batch culture is accompanied with extra labor on development and operation, and increased risks of contamination and product quality deterioration associated with longer processes. Overall, considering what it can generate and the time/effort/cost paid for it, fed-batch is currently prevalent over other modes in therapeutic protein manufacturing.4,28

290 Table 3

Modes of Culture/Animal Cells Comparison of batch, fed-batch, and perfusion cultures

Volumetric Productivity (g/L h) Cell density Product quality consistency Cell line stability constraint Labor and energy cost Process development and validation (time/cost) Process control complexity Bioreactor volume requirement Facility flexibility Culture cycle time Regulatory approval of process Scalability (scale-up and scale-down) Tech transfer Risk of technical failure Risk of contamination Requirement of automation Desire of PAT Public availability of relative expertize

Batch

Fed-batch

Perfusion

Low Low Low Low High Low Low High High Short Easy Easy Easy Low Low Low Medium High

Medium Medium Medium Low Medium Medium Medium High High Medium Easy Easy Medium Medium Medium Medium Medium High

High High High High Low High High Low Low Long Hard Hard Hard High High High High Low

A perfusion culture can achieve the highest cell density, the longest culture duration, and the greatest IVCD. The demands on production capacity are largely increased in recent years, as monoclonal antibodies with high-dose requirement attract the most attention in the biopharmaceutical industry. High-yield processes with a superior operation strategy to maximize protein production are highly desired to fulfill this call. The first driver of perfusion applications is their high production rate to reduce manufacturing costs. With a great productivity, which is up to 10 times higher than its fed-batch counterpart, perfusion culture also allows the use of bioreactors more compact in size. The second driver and probably the most essential key advantage is the minimal exposure of product to harsh production conditions to achieve high quality, especially for labile products. This is attributed to the short residence time of secreted products and the removal of glycosidases and proteases, which are released from dead cells and often accumulated to an adverse level in fed-batch culture causing deglycosylation and deamination. Perfusion culture also minimizes labor and time on seed-train preparation and CIP/SIP and provides a more consistent environment for product formation, resulting in better reproducibility. The labile antihemophilic factor VIII (Bayer, Berkeley, CA), probably the largest molecule among all commercial biopharmaceuticals, is one of the successful cases using perfusion technology. However, perfusion culture is not as common as fed-batch culture due to prolonged time of development and validation, high risk, low plant flexibility, and low titer.4,5 Long course of bioprocesses and the addition of retention devices result in a substantially longer developmental time period, and at the same time the genetic stability of MCB and WCB should be evaluated to ensure constant productivity and quality within the production process period. The risk of contaminations is high because of the increased production cycle time and extra complexity, and the risk of equipment failure is also high due to fouling and clogging of the retention device. For economic consideration, many facilities are designed for multiproduct production for different periods, which has to be compromised if perfusion culture is the choice of production. Product changeover and turnaround in an existing facility is difficult due to complicated and long duration of perfusion culture. In most cases, due to low perfusion medium utilization efficiency, the product titer in the perfusion outlet stream is low and waste treatment cost is high. The tradeoff of high volumetric productivity and quality with extra risks and cost makes manufacturers reluctant to accept perfusion technologies if not necessary.

References 1. Rose, S.; Black, T.; Ramakrishnan, D. Mammalian Cell Culture: Process Development Considerations. In Handbook of Industrial Cell Culture: Mammalian, Microbial, and Plant Cells, Vol. 4, Vinci, V., Parekh, S., Eds.; Humana Press: Totowa, NJ, 2003; pp 69–103. 2. Birch, J. R.; Racher, A. J. Antibody Production. Adv. Drug Deliv. Rev. 2006, 58, 671–685. 3. Wurm, F. M. Production of Recombinant Protein Therapeutics in Cultivated Mammalian Cells. Nat. Biotechnol. 2004, 22, 1393–1398. 4. Ozturk, S. S.; Hu, W.-S. Cell Culture Technology for Pharmaceutical and Cell-based Therapies, Dekker: New York, NY, 2006. 5. Bibila, T. A.; Robinson, D. K. In Pursuit of the Optimal Fed-batch Process for Monoclonal Antibody Production. Biotechnol. Prog. 1995, 11, 1–13. 6. de Zengotita, V. M.; Miller, W. M.; Aunins, J. G.; Zhou, W. Phosphate Feeding Improves High-cell-concentration NS0 Myeloma Culture Performance for Monoclonal Antibody Production. Biotechnol. Bioeng. 2000, 69, 566–576. 7. Whitford, W. G. Fed-batch Mammalian Cell Culture in Bioproduction. BioProcess Int. 2006, 4, 30–40. 8. Xie, L.; Wang, D. I. Fed-batch Cultivation of Animal Cells Using Different Medium Design Concepts and Feeding Strategies. Biotechnol. Bioeng. 1994, 43, 1175–1189. 9. Ma, N.; Ellet, J.; Okediadi, C.; et al. A Single Nutrient Feed Supports Both Chemically Defined NS0 and CHO Fed-batch Process: Improved Productivity and Lactate Metabolism. Biotechnol. Prog. 2009, 25, 1353–1363. 10. Li, F. A Systematic Approach to Develop Chemically-defined Cell Culture Media. BioProcess Int. Conf. Exhib. October 2009. Raleigh, NC, USA.

Modes of Culture/Animal Cells

291

11. Chee Furng Wong, D.; Tin Kam Wong, K.; Tang Goh, L.; et al. Impact of Dynamic Online Fed-batch Strategies on Metabolism, Productivity and N-glycosylation Quality in CHO Cell Cultures. Biotechnol. Bioeng. 2005, 89, 164–177. 12. Zhou, W.; Rehm, J.; Europa, A.; Hu, W.-S. Alteration of Mammalian Cell Metabolism by Dynamic Nutrient Feeding. Cytotechnology 1997, 24, 99–108. 13. Sauer, P. W.; Burky, J. E.; Wesson, M. C.; et al. A High-yielding, Generic Fed-batch Cell Culture Process for Production of Recombinant Antibodies. Biotechnol. Bioeng. 2000, 67, 585–597. 14. Li, F.; Zhou, J. X.; Yang, X.; et al. Current Therapeutic Antibody Production and Process Optimization. BioProcessing J. 2005, 4, 23–30. 15. Robinson, D. K.; Chan, C. P.; Yu Lp, C.; et al. Characterization of a Recombinant Antibody Produced in the Course of a High Yield Fed-batch Process. Biotechnol. Bioeng. 1994, 44, 727–735. 16. Konstantinov, K.; Goudar, C.; Ng, M.; et al. The “push-to-low” Approach for Optimization of High-density Perfusion Cultures of Animal Cells. Adv. Biochem. Eng. Biotechnol. 2006, 101, 75–98. 17. Piret, J. M.; Cooney, C. L. Immobilized Mammalian Cell Cultivation in Hollow Fiber Bioreactors. Biotechnol. Adv. 1990, 8, 763–783. 18. Yang, S.-T.; Luo, J.; Chen, C. A Fibrous-bed Bioreactor for Continuous Production of Monoclonal Antibody by Hybridoma. Adv. Biochem. Eng. Biotechnol. 2004, 87, 61–96. 19. Zhang, X. 3-D Cell-based High-throughput Screening for Drug Discovery and Cell Culture Process Development. PhD thesis, Ohio State University, 2008. 20. Woodside, S. M.; Bowen, B. D.; Piret, J. M.; et al. Mammalian Cell Retention Devices for Stirred Perfusion Bioreactors. Cytotechnology 1998, 28, 163–175. 21. Griffith, L. G.; Swartz, M. A. Capturing Complex 3D Tissue Physiology In Vitro. Nat. Rev. Mol. Cell Biol. 2006, 7, 211–224. 22. Muschler, G. F.; Nakamoto, C.; Griffith, L. G. Engineering Principles of Clinical Cell-based Tissue Engineering. J. Bone Jt. Surg. 2004, 86, 1541–1558. 23. Voisard, D.; Meuwly, F.; Ruffieux, P. A.; et al. Potential of Cell Retention Techniques for Large-scale High-density Perfusion Culture of Suspended Mammalian Cells. Biotechnol. Bioeng. 2003, 82, 751–765. 24. Deo, Y. M.; Mahadevan, M. D.; Fuchs, R. Practical Considerations in Operation and Scale-up of Spin-filter Based Bioreactors for Monoclonal Antibody Production. Biotechnol. Prog. 1996, 12, 57–64. 25. Avgerinos, G. C.; Drapeau, D.; Socolow, J. S.; et al. Spin Filter Perfusion System for High Density Cell Culture: Production of Recombinant Urinary Type Plasminogen Activator in CHO Cells. Nat. Biotechnol. 1990, 8, 54–58. 26. Trampler, F.; Sonderhoff, S. A.; Pui, P. W.; et al. Acoustic Cell Filter for High Density Perfusion Culture of Hybridoma Cells. Nat. Biotechnol. 1994, 12, 281–284. 27. Ozturk, S. S. Engineering Challenges in High Density Cell Culture Systems. Cytotechnology 1996, 22, 3–16. 28. Shuler, M. L.; Kargi, F. Bioprocess Engineering: Basic Concepts, 2nd ed.; Prentice Hall: Upper Saddle River, NJ, 2002.

Relevant Websites http://www.applikon-bio.com – Applikon Biotechnology. http://www.bio-pro.de – Das Biotechnologie und Life Sciences Portal Baden-Württemberg. http://www.percivia.com – PER.C6 A Human Cell Lines Biotherapeutics of the Future. http://www.phrma.org – PhRMA. New Medicine. New Hope. http://www.refinetech.com – Refine Technology; ATF System Improve Your Bioprocess. http://www.sartorius-stedim.com – Sartorius Stedim Biotech. http://www.seahorsebio.com – Seahorse Bioscience. http://www.thermo.com – Thermo Scientific. http://www.wavebiotech.com – WAVE Bioreactor Systems. http://www.xcellerex.com – Xcellerex.

Modes of Microbial Cultureq

1.20

IK Blaby, Brookhaven National Laboratory, Upton, NY, United States V de Cre´cy-Lagard, University of Florida, Gainesville, FL, United States TJ Lyons, Evolugate, Gainesville, FL, United States © 2017 Elsevier B.V. All rights reserved. This is a reprint of I.K. Blaby, V. de Crécy-Lagard, T.J. Lyons, Modes of Culture/Microbial, Reference Module in Life Sciences, Elsevier, 2017.

1.20.1 1.20.2 1.20.2.1 1.20.2.1.1 1.20.2.1.2 1.20.2.2 1.20.2.2.1 1.20.2.2.2 1.20.3 1.20.3.1 1.20.3.2 1.20.3.2.1 1.20.3.2.2 1.20.3.2.3 1.20.4 1.20.5 References

Introduction Modes of Microbial Culture Batch Cultures Batch culture variations Limitations of batch cultures Continuous Culture and the Chemostat Chemostat variations Limitations of continuous culture When the Microbe Itself Is the End Product Harvesting Microbes and Wall Growth Experimental Evolution Serial transfer Continuous cultures New variations on long-term culture Using Microbial Communities for Industrial Purposes Concluding Remarks

292 293 293 294 295 295 296 296 297 297 297 297 299 300 302 303 303

Glossary Batch culture The process of growing cells in a bioreactor with limited nutrients, such that the culture reaches saturation and further growth is prevented. Biocatalysis, biocatalytic process A process through which a chemical product is made using enzymes derived from living organisms or by live cellular factories. Chemostat, auxostats, turbidostat, and retentostat Any of a variety of bioreactors designed to facilitate the continuous culture of cells. Continuous culture The process of growing cells in a bioreactor under conditions where spent medium is continuously replaced with fresh medium enabling a continuous growth rate. Experimental evolution The process of selecting organisms by culturing under specific conditions, or under a selection regime, with the goal of identifying a particular phenotype. Fed-batch culture A variation of batch culture in which a particular nutrient is slowly provided to the culture at a rate that is limiting to biomass increase.

1.20.1

Introduction

For many industrial processes that depend upon microbial biocatalysts, there are two central issues that must be addressed in order for the process to become economically viable. First, one must find the right culture conditions for efficient biocatalysis. Coaxing microbes to produce specific end products quickly and in high yield is often more complicated than simply giving them a food source and letting them grow. Second, one must find the “right microbe for the job” – an organism that has the appropriate genotype to perform the particular biocatalytic task of interest. These two considerations – culture conditions and microbial genotype – are intimately related, and often, one must be altered to accommodate the other in order to get a biocatalyst to grow vigorously under the conditions that are required for process optimization.

q

Change History: April 2016. I.K. Blaby, V. de Crécy-Lagard, and T.J. Lyons made edits throughout the chapter; added Table 1; replaced Section "Algal Biodiesel: A Case Study in Contemporary Challenges for Microbial Culture" by the Section 4 "Using Microbial Communities for Industrial Purposes"; and updated the biography section.

292

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.09021-X

Modes of Microbial Culture

293

In this chapter, we discuss different approaches to the culture of microorganisms for industrial applications along with the benefits and problems associated with each approach. We will include examples from the recent literature that highlight major advances in methodology or address crucial contemporary problems such as the production of biofuels. We also address the use of different modes of microbial culture for experimental evolution, with the ultimate goal of producing novel microbes with useful phenotypes that are not accessible by other methods of genetic engineering.

1.20.2

Modes of Microbial Culture

A quick perusal of the literature reveals that there are myriad ways to culture microorganisms; however, all methods can be characterized as variations of one of two basic methodologies – batch culture and continuous culture. These are discussed in detail in the following sections.

1.20.2.1

Batch Cultures

Batch cultures are closed systems that are essentially characterized by inoculating a microbe into a culture vessel, or bioreactor, containing a finite amount of nutrients and letting the culture grow to saturation over time. Although most batch cultures are agitated by means of mechanical mixing or by bubbling gas through the solution from beneath (called airlift reactors), some cell types perform better with no agitation. An archetypal growth curve of a batch culture is shown in Fig. 1 and is based on the growth phases identified by Monod (1949). Typically, a small inoculum of the microbe is added to a large volume of medium. Upon resuspension in the growth medium, the microbe remains quiescent for a period of time that is cell type and condition dependent. During this time (lag phase; phase A in Fig. 1), the cells are not dividing but are nevertheless metabolically active due to the need to physiologically and metabolically adapt to the new medium (Pin et al., 2009). Once the cells start to divide, they enter logarithmic phase (also called log phase or exponential phase, phase B in Fig. 1). In this phase, the cell density doubles at regular intervals that correspond to the duration of the cell division of the specific microorganism. Although many cultures display classic exponential growth curves, deviations from this dynamic occur for a variety of reasons. A common cause of this is a phenomenon known as diauxie, in which cells display one growth rate on a primary nutrient, yet display a different growth rate on a secondary nutrient after the primary nutrient is depleted (Monod, 1945; Fig. 2b). A classic example of this is the diauxic shift during ethanol fermentation by yeast when grown anaerobically. Although glucose might be abundant initially, the cells divide and they produce ethanol. However, as glucose is depleted, the cells begin to consume the ethanol as a secondary carbon source, which sustains a slower growth rate (Lewis et al., 1993). The result is two distinct stages of logarithmic growth. Of course, since ethanol is often the desired end product of yeast fermentations, its consumption represents a complication for biotechnological applications. As the nutrients in the medium are depleted, the growth rate slows down until the rate of cell division eventually equals the rate of cell death due to senescence or due to the accumulation of toxic metabolic waste products (ethanol, acetic acid, protons, etc.). At this point, there is no net change in cell density over time and the cells can be said to have entered stationary phase (Kolter et al., 1993; phase C in Fig. 1). Often, microbes remain metabolically active in stationary phase for extended periods of time by cannibalizing dead and dying cells. Indeed, it has been shown that cells continue to evolve in stationary phase (Finkel, 2006). Cells can also enter a quiescent state in which they undergo profound physiological changes to form nondividing spores or spore-like cells with minimal metabolic activity and increased stress resistance. Eventually, the rate of cell death can exceed the rate of cell division and the culture will enter the death phase, the onset and rate of progression of which are highly dependent upon the microbe in question.

A

B

C

D

Products (% maximal output) Nutrients (% input) Population size (% maximal yield)

100

Time Figure 1 Schematic of growth dynamics of a batch culture. The solid black line shows the increase in population density as the cells consume feedstock and grow. The dotted red line shows the decrease in nutrients in the culture medium as the culture progresses. The dotted/dashed green line indicates the amount of end product produced by the cells in culture. Letters refer to distinct phases of the growth curve. A, lag phase; B, logarithmic phase; C, stationary phase; D, death phase.

Modes of Microbial Culture

100

Population size (% maximal yield)

Population size (% maximal yield)

Nitrogen source (% input)

100

Preferred carbon source (% input)

Product (% maximum output)

B

Carbon source (% input)

Triglycerides (% maximum output)

A

Secondary carbon source (% input)

294

Time

Time

Figure 2 Variations on batch culture. (A) End products produced during stationary phase after cells have stopped growing due to nitrogen depletion. Depletion of carbon sources occurs in two stages. In the first, carbon is used to produce microbial biomass. In the second, it is diverted into generating product. (B) Overcoming catabolite repression by using a multistage culture with a primary, but repressing, carbon source to produce microbial biomass and a secondary carbon source to produce the desired end product.

From a biotechnological standpoint, cells can produce desired end products during any phase of growth, depending upon when the biochemical pathways producing the product are activated. Some products may be optimally produced during logarithmic phase, whereas others are optimally produced during stationary phase. Other products may result from heterologous expression using constitutive promoters that are continuously active regardless of growth phase.

1.20.2.1.1

Batch culture variations

Often, a biocatalytic process must take full advantage of different growth phases to enhance product yield and utilize variations of the batch culture method to achieve the best results. For example, some end products such as triglyceride oils are only produced en masse when the cells enter stationary phase. In such cases, it is often important to limit one essential nutrient, such as nitrogen, to induce the transition into stationary phase and add a carbon source from which the product will be made in excess (Meng et al., 2009; Fig. 2a). However, as the amount of product is ultimately proportional to the number of cells, a productive logarithmic phase is required to produce the maximal number of cells. These processes can be thought of as having two stages. In the first stage, the goal is to achieve the maximum possible cell density, and in the second stage, the goal is to produce the maximum amount of product. In one such example, glucose is used to produce biomass and ATP while a second carbon source, xylose, is used to make product (Park and Seo, 2004). Sometimes, a particular biocatalytic process works well on the laboratory scale but encounters problems upon scale-up where the dynamics of gas and heat exchange are markedly different. For example, if a hypothetical process requires O2 and there is poor gas exchange in the bioreactor, then microbial growth and oxygen consumption may exceed the replenishment of oxygen from the atmosphere or from air bubbles that are parsed through the medium. The result is the depletion of an essential nutrient, in this case O2, before the culture is able to achieve maximal cell density (Vanags et al., 2007). Fewer cells also means less end product (Fig. 3a). The solution to this problem is a variation on the batch process known as fed-batch, in which a concentrated solution of another nutrient, most often the carbon source, is slowly fed to the cells at a consistent but growth-limiting rate. By slowing down microbial growth, one allows gas exchange to catch up with consumption (Fig. 3b). Variations on the fed-batch methodology can also be used to alleviate problems with heat exchange in large bioreactors, to overcome toxicity of contaminants or to prevent the formation of undesirable side products (Vanags et al., 2007). B Batch culture with poor oxygen exchange 100

Fed-batch culture 100

Population size (% maximal yield)

Nutrients (% input)

pO2 (% maximum solubility)

Products (% maximum output)

A

Limiting nutrient feed

Time

Time

Figure 3 Batch versus fed-batch cultures. (A) A hypothetical batch process of obligate aerobes, in which oxygen consumption is faster than exchange with air. The result is oxygen limitation, poor biomass and product yields. (B) The same process in which the feedstock is fed slowly and consistently to the culture so that growth rate is limited. Oxygen consumption no longer exceeds exchange with air and the culture can reach higher biomass yields.

Modes of Microbial Culture

295

Fed-batch processes allow batch cultures to achieve much higher densities than batch cultures. The reason is that many essential nutrients (eg, iron) are toxic at high levels. As a result, the amount of said nutrient in the initial batch limits the maximum achievable cell density – yet, more nutrient cannot be added in batch culture due to inherent toxicity (Shiloach and Fass, 2005). By slowly feeding such nutrients, much greater biomass yields can be attained. As more biomass means greater product yields, it is not surprising that fed-batch is often preferred over batch cultures. However, it is important to note that fed-batch methods add a layer of complexity to a process that may reduce economic viability on the industrial scale. Conversely, batch culture is established, reliable, inexpensive to set up and simple to maintain, making it an attractive option despite limitations (discussed in Section 1.20.2.1.2).

1.20.2.1.2

Limitations of batch cultures

Simplicity is the true advantage of batch cultures; however, they also have important limitations. First, in batch, microbes are exposed to a constantly changing environment due to the consumption of nutrients and the buildup of waste products. Therefore, there is usually only a small optimal window in which the environmental conditions are ideal for the efficient biosynthesis of the desired end product. Second, batch cultures eventually reach an endpoint and must be restarted. For large bioreactors, there is a significant turnaround time required to empty, clean, sterilize, and reestablish the reactor for the next batch.

1.20.2.2

Continuous Culture and the Chemostat

In theory, the ideal situation for biotechnology would be to determine the best culture conditions for the production of the desired end product and maintain the cells under these conditions in a steady state, so that the product can be made continuously – a process known as continuous culture. Credit for continuous culture theory and methodology is generally given to Monod (1950) and Novick and Szilard (1950), who independently published their work in 1950. It is important to note, however, that the concept and practice of these ideas is much older and various apparatuses that maintained continuous cultures were published well before 1950 (Bilford et al., 1942; Myers and Clark, 1944). Ultimately, the name chemostat was coined by Novick and Szilard and came to be generally applied to the method. A chemostat is a single automated bioreactor in which spent medium is continuously replaced with fresh medium, where one nutrient is found in limiting quantities (Fig. 4). If the rate of medium replacement (the dilution rate) is lower than the growth rate of the microorganisms inside, then the cell density will increase. If the dilution rate is higher than the growth rate, then the cell density will decrease, and eventually, the cells will wash out. Conversely, if the dilution rate can be calibrated to equal the growth rate, a steady state is achieved. In the standard chemostat, the dilution rate is calculated based on the known growth rate of the microbe within and is fixed at the beginning of the experiment. Depending on the needs of the experiment, the cell density can be maintained at any level up to saturation as determined by the amount of a limiting nutrient.

Population size (% maximal yield)

Products (% maximum output) Nutrients (% input)

A

B pH Glucose O2 Turbidity

100

Flowin = Flowout Detector(s)

Nutrient feed Flowin

Equilibrium achieved

Saturated

Mixer

Flowout medium

Gasout Break to prevent chemotactic backgrowth into nutrient reservoir

Gasin Heat exchange

Time Figure 4 Continuous culture via chemostats and related devices. (A) The concept behind continuous culture is similar to fed-batch, in which fresh medium is continuously added to the bioreactor. The difference lies in the fact that saturated medium is removed from the bioreactor at a rate that is equivalent to that of medium addition. The dilution rate is ideally matched to the growth rate of the microorganisms to maintain constant cell density. (B) In the chemostat, fresh medium is added to the bioreactor at a rate that is equivalent to the removal of saturated medium. The medium is thoroughly mixed and the culture needs to be maintained at a constant temperature. Gas exchange is generally achieved by bubbling the optimal gas mixture through the solution. A physical barrier is needed to prevent microbial movement into the reservoir containing fresh medium. In variations of the chemostat, dilution rate is not calculated based on the growth rate of the cells, but rather on continuously modified based on physical measurements of the culture, such as pH, dissolved oxygen, glucose levels, or direct measurement of cell density.

296

Modes of Microbial Culture

1.20.2.2.1

Chemostat variations

As we saw with batch cultures, it is often important to grow microbial biocatalysts to high cell density in order to maximize product yield. In the chemostat, cell density is constrained by the concentration of the limiting nutrient, which often cannot be increased beyond a certain concentration due to toxicity effects. Cell densities that exceed saturation can be achieved in chemostats by recycling the cells that are removed (chemostat, with cell recycle, or retentostat) (Chesbro et al., 1979). In addition to allowing higher cell densities, and subsequently greater product yields, such bioreactors also allow for dilution rates that are higher than the growth rate without causing the cells to wash out. This is particularly useful when the product is toxic and needs to be rapidly removed. Another useful feature of this method is the fact that the growth rate decreases to near zero due to severe nutrient depletion. In essence, this method maintains cells in stationary phase for long periods of time. This is useful if a particular product is synthesized in stationary phase. One of the central limitations of the chemostat is that the culture is essentially on its own once the experiment has begun; the user has no opportunity to alter parameters once inoculated. This led to the development of chemostat variations that allow for the modification of dilution rate depending on real-time monitoring of changes in the culture conditions (Gostomski et al., 1994). For example, the turbidostat continuously measures cell density using a turbidimeter, which directly controls the dilution rate through measuring light scattering. However, turbidostats operate under the assumption that light diffraction correlates linearly with cell density, yet this is only true for transmission values that lie in the dynamic range of the particular tubidimeter. To complicate matters, significant changes in cell size also increase light diffraction, and consequently, turbidimeters may not accurately report the status of a particular culture. Turbidostats are also not particularly useful for measuring cell density with microbes that do not grow evenly in suspension, such as filamentous fungi, or for cultures in which the substrate is particulate, such as biomass. Finally, many microbes have the unfortunate property of adhering to reactor surfaces, including the optics of the turbidimeter. This can result in falsely assuming planktonic cell density is higher than it really is (see discussion on wall growth below). In these cases, it is often only possible to track microbial growth using a secondary reporter that is closely tied to growth, such as changes in pH or the consumption of essential nutrients such as oxygen or glucose. Devices that take this approach are called auxostats and their detectors directly control dilution rate. As with batch cultures, it is often useful to set up chemostats or related devices with multiple stages when the production of biomass must be separated from the biosynthesis of the desired end product, or to alleviate problems associated with catabolite repression. This can be achieved by altering the nutrient feed after the culture density has reached equilibrium so that one nutrient mix is used to produce biomass and a second is used to synthesize product. Alternatively, several chemostats could be set up in series, where the effluent from one, which is saturated with cells, serves as the inoculum for a downstream chemostat. One of the benefits of this approach is that the increased residence time in downstream bioreactors gives the biocatalysts more time to consume excess substrate. Otherwise, these nutrients would be lost in the effluent or possibly as secondary substrates that can only be utilized once a primary substrate is consumed (Govindaswamy and Vane, 2010). Multistage continuous culture is also helpful when the feed or environmental conditions, such as temperature or gas composition, for the downstream chemostat are significantly different from the upstream chemostat.

1.20.2.2.2

Limitations of continuous culture

Despite its advantages, continuous culture has several limitations that may restrict or prevent its use for industrial applications. One problem is encountered with cells that do not grow evenly in suspension such as filamentous fungi. Such cells grow as hyphal masses that are difficult to homogenize and remove as part of the effluent. The same problem is encountered if the microbial feedstock is particulate, as would be the case for biomass particles. In addition, although continuous cultures can theoretically be maintained indefinitely, they are particularly susceptible to contaminants that might infiltrate the nutrient feed. Thus, they must be periodically stopped, sterilized, and restarted. Other limitations of continuous culture are only revealed upon scale-up. As it takes up to five volumes of the bioreactor to achieve maximal cell density, the nutrient feed reservoir needs to be large. Vigorous mixing is also required to ensure homogenization of the culture and the feed. On the lab scale, these are not serious problems; however, on an industrial scale they may be considerable limitations. However, the most crucial limitation of continuous culture is the phenomenon called “wall growth,” in which cells adhere to and form biofilms on the inner surfaces of the bioreactor (Baltzis and Fredrickson, 1983). For microbial cells that require a solid surface to mediate growth, chemostats cannot be effectively used because maximal biomass yield is limited by the inner surface area of the bioreactor, and not by the limiting nutrient feed. Even when growing microbes that prefer to grow as planktonic populations, variants will rapidly appear that stick to the bioreactor walls and avoid being washed out during dilution, essentially turning the continuous culture into a selection scenario for adherent populations. Indeed, wall growth can occur within hours of establishing the culture and chemostats will quickly produce a heterogeneous mixture of planktonic and adherent populations that experience vastly different environmental conditions (Dykhuizen and Hartl, 1983). This can complicate attempts to maintain the right conditions for end product biosynthesis. Consequently, chemostats cannot be maintained for long periods of time and must be periodically emptied and cleaned to remove wall growth. Not surprisingly, chemostats are expensive to set up and maintain and despite their theoretical advantages, their practical limitations make them less desirable from an industrial perspective. Because batch cultures are simple, they are easy to establish on the commercial scale and still represent the most economical way of facilitating biocatalysis.

Modes of Microbial Culture

1.20.3

When the Microbe Itself Is the End Product

1.20.3.1

Harvesting Microbes and Wall Growth

297

Often, the desired end product of microbial culture is not a soluble chemical product but the microbial biomass itself. Examples of this include the harvesting of algae for biofuels or alimentary purposes. When the ultimate goal is to harvest the microbial biomass, the process is complicated by the aforementioned problem of wall growth. The fact that microbes have a tendency to adhere strongly to any surface makes harvesting highly problematic. Indeed, by some estimates, more than 30% of the biomass in a traditional bioreactor is due to wall growth (Larsen et al., 2003). Over the years, a variety of methods have been developed for reducing wall growth. These include coating the bioreactor walls with hydrophobic substances or the development of novel mixing systems, such as variomixing (Larsen et al., 2003), which increase turbulence in the solution. Although these methods work in the short term, they fail in the long term because of the tremendous selective pressure favoring wall growth. However, the main aim of this section is to highlight another powerful application of microbial culture in which the microbe itself is the end product: experimental evolution.

1.20.3.2

Experimental Evolution

In this case, the goal is to optimize the microbe for the required task when an optimized one does not already exist. In experimental evolution, microbes are maintained in stable long-term cultures with a particular selection regime. Over time, natural variants occur via low frequency errors in DNA replication machinery. A small fraction of these may be better adapted for the culture conditions, and these variants will out compete the rest of the population. Eventually, the product of this type of culture is a microbe with new phenotypic capabilities, preferably ones that are more conducive for efficient biocatalysis in an industrial setting. Experimental evolution has various advantages that make it preferable to other methods for altering microbial phenotypes such as genetic engineering. Targeted genetic engineering is impractical when the phenotype of interest is complex, poorly understood on the molecular level or requires many simultaneous mutations to confer a particular phenotype. Often, it is impossible to predict what type of mutation might achieve the desired results. On the other hand, experimental evolution does not require a priori knowledge about which type or how many mutations are needed to alter a particular phenotype. Experimental evolution circumvents this problem by facilitating the blind selection of as many favorable mutations as are required to adapt to a particular selection regime. Also, genetically engineered alterations in one phenotype often result in unintended effects on other cellular phenotypes, such as growth rate. Thus, genetically engineered strains are often less robust than their wild-type counterparts and, consequently, may be less useful in an industrial setting. Inherent to experimental evolution is the selection of only the most robust strains with a particular phenotype. Finally, when microbes are genetically engineered, via the deliberate removal of particular genes or the addition of foreign genes, they become classified as genetically modified organisms (GMOs). For microbes that are intended for release into the environment, this adds a layer of complexity in the form of regulatory hurdles and possible consumer backlash. Indeed, in some markets it may be illegal to release GMOs. Experimental evolution relies on the selection of naturally occurring genetic variants from within a particular population rather than the targeted addition or removal of DNA. Although some may not see the distinction (since both methods result in modification of DNA) experimental evolution is more akin to plant breeding than genetic engineering and therefore does not produce GMOs. Although experimental evolution has the advantage of being able to produce many synergistic mutations without the need for a priori knowledge, it is limited by the lack of reproducibility. Indeed, there is no guarantee that a particular evolutionary pathway will be repeated, especially when attempting to achieve complex traits. Thus, if a particular evolved strain is lost or suddenly changes phenotype, then the nature of the adaptation is lost forever. Although this may be vexing for academic researchers, it is more restrictive for commercial endeavors where the time required to regenerate lost strains is crucial. Moreover, for the purposes of protecting intellectual property it is important to know the exact genetic nature of the evolutionary adaptation. In a growing trend brought on by decreasing sequencing costs, experimental evolution is being combined with whole genome sequencing to identify adaptive mutations. Not only does this allow one to better understand the nature of adaptation, but it also allows one to reproduce a particular adaptive variant if the original strain is lost or changes phenotype. The power of combining experimental evolution with genome resequencing for improvement of microbes for industrial applications is exemplified by the explosion of published studies in the last few years (Table 1). The key to successfully altering microbial phenotypes using experimental evolution is maintaining stable long-term cultures. There are several different modes of culture that have been employed for experimental evolution for decades, including serial transfer of batch and continuous cultures. These methods have been successfully applied but possess important limitations. Newly reported methods have attempted to address these limitations.

1.20.3.2.1

Serial transfer

In serial transfer, a batch culture is serially propagated by repeatedly transferring a portion of an actively growing culture to a new, sterile, culture vessel. Periodically, samples are taken for long-term storage and to assess fitness relative to the ancestral strain. The selection imposed upon the culture is defined by the conditions in which it is grown; by exposing the population to stresses, such as varying the environmental conditions, or by growth in particular media, arising mutations with an increased fitness will overtake the population and ultimately become fixed. Critically, the severity of the selection is increased very gradually. In principle, if repeated over a sufficiently long timescale, the gap between the original and the desired phenotype should diminish. An early evolutionary

298 Table 1

Modes of Microbial Culture Strain improvement studies by adaptive laboratory evolution for applied purposes between January 2015 and March 2016

Species

Culture method

Selective pressure

Reference

Corynebacterium glutamicum (B) Lactococcus lactis (B) Thermotoga maritima (B) Thermococcus onnurineus (A) Escherichia coli (B) Sporomusa ovata (B) Saccharomyces cerevisiae (E) E. coli (B) Wine yeast strains (E) S. cerevisiae (ethanol red) (E)

Manual serial batch Manual serial batch Manual serial batch Manual serial batch Manual serial batch Manual serial batch Turbidostat Manual serial batch Manual serial batch Manual serial batch

Scheffersomyces stipitis (E) S. cerevisiae (E)

Continuous reactor Manual serial batch

High temperature High temperature Carbon source Carbon source Carbon source Toxic carbon source Toxic product Phosphate limitation Low temperature Carbon source and high temperature Toxic carbon source High temperature

E. coli (B) E. coli (B) Rhodococcus opacus (B)

Manual serial batch Manual serial batch Manual serial batch

Low pH Toxic compound Toxic compounds

Sulfolobus solfataricus (A) C. glutamicum (B) Phaeodactylum tricornutum (E)

Manual serial batch Manual serial batch Semi-continuous culture

Low pH and high temperature High temperature Light induced stress

Oide et al. (2015) Chen et al. (2015) Latif et al. (2015) Lee et al. (2016b) Lacroix et al. (2015) Tremblay et al. (2015) Voordeckers et al. (2015) Moreau and Loiseau (2016) Lopez-Malo et al. (2015) Wallace-Salinas and Gorwa-Grauslund (2013); Wallace-Salinas et al. (2015) Pereira et al. (2015) Caspeta et al. (2014); Caspeta and Nielsen (2015) Harden et al. (2015) Graves et al. (2015) Yoneda et al. (2016); Kurosawa et al. (2015) McCarthy et al. (2015) Lee et al. (2016a) Yi et al. (2015)

(A) Archaea; (B) Bacteria; (E) Eukarya.

investigator by the name of Dallinger pioneered the method of experimental evolution by serial transfer in the 1880s. Dallinger (1878) began a microbial culture with a starting temperature of 15
C, and over the course of 7 years, adapted the microbe for growth at 70
C by gradually increasing the temperature. Subsequent to Dallinger’s pioneering endeavors, the basic idea of continually subculturing while gradually exerting more severe external stimuli has been employed for academic evolution studies and with applied biotechnological intentions. Indeed, the literature abounds with examples of experiments conducted in this manner (Table 1). More recently, evolutionary biologists have employed microorganisms for long-term evolutionary experiments, which, due to their short generation time, large population size, simple genetic tractability, and the advent of relatively cheap sequencing, have been demonstrated to be useful models for the study of evolution (Elena and Lenski, 2003). The work of Lenski’s laboratory at Michigan State University with the long-term Escherichia coli experiment has, among others, been instrumental in illustrating the use of serial dilution-based techniques for demonstrating change in fitness as a result of evolution. Similar experiments have been conducted using eukaryotic and multicellular organisms and on solid media, indicating that the technique of serial transfer is not limited to bacteria or liquid culture. Serial transfer has the attraction that once the initial culture has begun growing and the experimental design (ie, selection criteria) decided on, little in the way of fundamental skill set is required. However, experiments designed to adapt organisms to novel conditions tend to be long, as there is no clear experiment ending or conclusion, and it is impossible to predict when the favorable mutations will appear. Rather, the experiment continues indefinitely as the number of required generations is unknown. The frequency of subculturing depends on the growth rate, and therefore the organism being used and the growth conditions, but is often performed on a daily basis, making for a labor-intensive process. Additionally, the subculturing must be performed with due care, as on each occasion during the transfer both the original (donor) culture and the new (recipient) culture must be exposed to the outside environment, providing the opportunity for contamination at each subculture. Academic evolution experiments have made clear several phenomena of which the applied investigator should be aware. Rather than accruing genetic diversity, it has been experimentally demonstrated that adaptive mutations become rapidly fixed in the population in a process known as periodic selection, and as beneficial, fitness-increasing mutations occur, selective sweeps of the culture result in outcompetition by the novel genotype. However, the large population size and short generation time experienced in bacterial cultures often result in multiple beneficial mutations co-occurring, in practice, resulting in a heterogeneous population. In asexual populations (and the associated lack of intrapopulation recombination), the inability to combine beneficial mutations causes competition between phenotypes, known as clonal interference, which can disrupt the predicted progression toward mutation fixation. In the absence of horizontal gene transfer, this is likely to result in the loss of one of the mutations from the population. An often-overlooked aspect of serial dilution experiments is the point at which the dilution should be made. Key to deciding when to dilute is having an accurate measure of the density of cells and the phase of growth. It is crucial that dilution occurs in late log phase, when the most robust, rapidly growing cells have the greatest numerical advantage (Fig. 5). This requires accurate, real-time monitoring of cell density, which is not practical with batch cultures. Also, high dilution ratios are essential in order to avoid severe evolutionary bottlenecks. Indeed, modeling studies have suggested that dilution ratios ranging from 1:10 (D¼0.1) to 1:5 (D¼0.2) are optimal for minimizing the impact of genetic bottlenecks (Wahl et al., 2002). However, high dilution ratios

Modes of Microbial Culture

299

Relative cycle times T = 6.2

10

00

)

00

)

T = 9.3

01 0 .0 D=

0 .0

1(

(1:

1 :1

) D=

(1: 2) (1: 0 .1 5 ) (1: 10 D=

D=

D=

0 .5

Log cell density (arbitary units)

100

0 .2

Dilution

T = 3.4 T = 2.3 T=1

10

1

0.1 Time (arbitary units) Figure 5 Dilution in serial transfer. A hypothetical growth curve of a culture undergoing serial transfer. The gray shading indicates the ideal part of the growth curve in which to initiate dilution. The dotted lines indicate growth curves at various dilution ratios (D). T, relative time for dilutions at each dilution ratio to achieve the optimal density for the next dilution.

require more frequent transfer, particularly with rapidly dividing cells. For example, E. coli doubles every 20 min and at a dilution ratio of 0.2, one would need to initiate dilution hourly, which is time and labor intensive. Otherwise, one runs the risk of allowing cells to oscillate between different growth phases with each dilution, resulting in a more complicated selection scenario. The solution is to use low dilution ratios (1:100, D¼0.01), so that cells remain in logarithmic phase for longer periods of time (Fong and Palsson, 2004). However, the transfer of fewer cells increases the severity of the genetic bottleneck and can potentially result in loss of beneficial mutations and lengthens the amount of time it takes to achieve desirable results. Serial transfer is also limited to cells that can be easily transferred. This is problematic for some cell types, such as filamentous fungi, that do not grow evenly in suspension. The same is true when the substrate is insoluble, as is the case with biomass. Serial transfer is also difficult when the cultured cells require a solid surface upon which to grow. In this case, experimental evolution requires either periodic removal of the cells from the vessel walls and homogenization (as is routinely done during passaging mammalian cells) or providing a continuous surface upon which to grow. The latter can be achieved if the surface is agar and the cells are allowed to grow directionally along a selection gradient. Despite its limitations, serial transfer has been successfully used to select for microbes with industrially important phenotypes, including growth at extreme temperature or pH. A particularly elegant application of serial dilution is to increase the yield of a valuable end product by coupling yield to an aspect of growth that can be selected for. For example, strains can be engineered in such a way that production of the desired product is the only route for reduced nicotinamide adenine dinucleotide (NADH) oxidation (to NADþ) under anaerobic conditions. This selection was employed by coupling to lactate production to growth rate, using E. coli strain SZ110, which contains mutations in the alcohol dehydrogenase and acetate kinase genes (adhE and ackA). The result was a strain that produced 1 mol of lactate per liter in complex media in 48 h, compared with 840 mmol lactate after 72 h for the ancestral strain. In a similar manner, the technique has been exploited for the production of succinate. By using an E. coli strain containing mutations in ldhA (D-lactate dehydrogenase), adhE, and ackA, growth rate was improved by serial dilution for approximately 2000 generations. Subsequent to the evolutionary regime, the strain produced 0.73 mol of succinate per mole of metabolized glucose, compared with 0.20 mol for the starter strain. The authors succeeded in performing further genetic manipulations of this optimized strain to further improve succinate yield, ultimately producing a strain capable of producing up to 1.6 mol succinate per mole metabolized glucose (Yomano et al., 2008). These and other examples have been reviewed by Jarboe et al. (2007). More generally, the ease of the experimental set-up makes experimental evolution by serial transfer the most commonly used for a variety of adaptation scenarios such as high temperature, low pH, toxic compounds or carbon source limitation and in all kingdoms: Bacteria, Archaea and Eukarya (Table 1).

1.20.3.2.2

Continuous cultures

The widespread use of serial transfer in the development of biological products has been limited by the long-term and laborintensive nature of the protocols and the slow pace of progress, which conflict with the industrial incentives of achieving goals

300

Modes of Microbial Culture

rapidly. Indeed, these inherent caveats led to the mechanization/automation of continuous culture and the development of the chemostat and its derivatives. The benefit of using continuous culture for experimental evolution is that the environment is stable, and the dilution rate can be precisely controlled to maintain maximal growth rate and reduce the severity of genetic bottlenecks. This means that, in theory, the selection of favorable variants occurs much faster using a chemostat than serial transfer. Indeed, chemostats are routinely used to study evolutionary phenomena (Dykhuizen and Hartl, 1983). There are many examples in the literature of the use of continuous culture for producing industrially important strains with altered phenotypes. An example of contemporary importance pertains to the use of continuous culture to improve the production of bioethanol from xylose by a strain of the yeast, Saccharomyces cerevisiae (Sonderegger and Sauer, 2003). Xylose is the second most abundant sugar in lignocellulosic biomass and there are few microbes that can efficiently ferment this sugar into a usable biofuel such as ethanol. One of the problems is that the metabolic pathways for channeling xylose into central metabolism do not exist in widely used ethanologens such as S. cerevisiae. Yeast can be genetically engineered to contain the appropriate enzymes for fermenting xylose; however, yeast maintains exquisitely tight control over metabolic pathways at the transcriptional and posttranslational level. This means that simply having the appropriate genotype does not mean that the desired phenotype will be expressed. This is particularly true for xylose metabolism in yeast, which, even with the appropriate enzymes, does not proceed under the anaerobic conditions that prevail in industrial fermentation. In an innovative series of experiments, Sonderegger and Sauer (2003) genetically engineered S. cerevisiae for xylose fermentation and slowly adapted the microbe to anaerobic conditions using a chemostat. The result was a strain of yeast capable of anaerobic fermentation of xylose. In addition, this strain could ferment xylose in the presence of glucose, indicating that catabolite repression had been alleviated. Most importantly, the resultant strain produced up to 19% more ethanol than the ancestral strain under the complex conditions that prevail in the fermentation industry. In another elegant example of the use of chemostats for evolutionary selection, Mondragón-Parada et al. (2008) isolated a community of microbes capable of degrading simazine, a triazinic herbicide that may have deleterious effects on the reproduction of aquatic animals. In this experiment, a chemostat inoculated with a community of soil bacteria known to be able to degrade simazine was fed medium containing simazine and other chemicals that are routinely added to simazine as adjuvants. The goal was to isolate strains that could degrade simazine in the presence of these adjuvants, which inhibit the growth of the original microbial community. After 42 days of continuous culture, a community of eight microbial species was isolated that could degrade 96% of the simazine present in the medium. Other recent examples of the use of continuous culture for strain improvement are given in Table 1. Despite the advantages of continuous culture, there are a variety of limitations that hinder its use for experimental evolution. First, although chemostats are less susceptible to contamination than serial transfer, continuous cultures are not closed systems, and contamination can enter the chemostat through the nutrient feed. The longer a continuous culture is run, the greater the possibility of contamination. Another possible complication that may arise is antagonistic pleiotropy, which is the inadvertent alteration of a secondary phenotype during the adaptation process. For example, adapting cells to grow in the presence of low levels of a particular sugar may reduce their ability to grow in the presence of high levels of the same sugar (Jansen et al., 2004). To a degree, antagonistic pleiotropy is a problem for all methods of experimental evolution; however, it may be more pronounced during selection in the steady-state conditions of continuous culture than during the fluctuating conditions of serial transfer. Such hyperspecialization would be problematic if one then wishes to use the resultant strain in a more dynamic industrial setting that may not be as uniform or stable as the selection scenario used for continuous culture. Above all, the central limitation of continuous culture is the fact that the desirable mutants are retained in the culture vessel rather than being transferred, as is the case in serial transfer. This means that continuous culture devices cannot be used to adapt microbes that require insoluble substrates or a solid surface upon which to grow because the cells or nutrients cannot easily be homogenized and removed through dilution. This also means that the fastest way to escape selective pressure is to adhere to the bioreactor walls and therefore, avoid being removed by dilution. Indeed, among the first mutants that will arise in populations maintained under continuous culture are ones that allow the microbes to form thick biofilms on every surface of the bioreactor. What results is a complex ecosystem of adherent and planktonic subpopulations that each experience different selective pressures. In essence, continuous culture exacerbates the problem of clonal interference that is only a mild nuisance for serial transfer techniques. Moreover, as was stated before, adherent microbes have the habit of fouling turbidimeters and other detectors and tricking the device into increasing dilution rate and washing out planktonic populations.

1.20.3.2.3

New variations on long-term culture

Bioreactor wall growth is the major reason that serial transfer, despite its many limitations, is more widely used than continuous culture for experimental evolution. One innovative solution to this problem was provided through the development of a proprietary technique called the Genetic Engine. In this technique, two chemostats are set up in series. Once the culture is established in the first chemostat, it is transferred to a clean chemostat for continued culturing. Although the culture continues in the second chemostat, the first chemostat is cleaned with NaOH to remove adherent cells. The result is a continuous culture/serial transfer hybrid process that counter-selects against wall growth in continuous cultures. In a proof-of-principle experiment, the Genetic Engine was used to select for strains of E. coli that could grow robustly under conditions of thymine starvation (de CrécyLagard et al., 2001). Although mainly an academic study, the lessons learned have practical applications because thymine starvation is known to cause cell lysis and is a major cause of clogging in continuous cultures. As in the case of serial transfer

Modes of Microbial Culture

301

Inoculum a 1

b

d

c

2

3

T

6

5

4

Growth chamber

Inoculation a 1

c

b 2

3

T

4

3

T

4

d 6

5

Growth round 1 a 1

c

b 2

z

a

Fresh medium

T

3

4

a 1

T

3

T

3

Sample 1

c

b 2

5

Sampling chamber

Mixing sampling

z 0

c

b 2

1

6

Saturated medium

Dilution

0

d 5

4

5

Growth round 2 z 0

b

a 1

2

c 4

5

T = Turbidimeter Successive rounds = Gates Figure 6 The Evolugator. Numbers indicate static positions on the mobile tubing. Letters indicate gates that are in constant contact with the tubing. An inoculum of cells is injected into the central growth chamber between gates (b) and (c) into tubing marked by the numbers 3 and 4. The cells are allowed to grow and cell density is measured through the translucent tubing using a turbidimeter (T). At the appropriate cell density, dilution is initiated. At this time, the gates and tubing are simultaneously moved to the right, pushing the culture and other contents of the tubing via peristaltic action. In the process, fresh medium is mixed with half of the saturated culture in the new growth chamber, which is marked by the numbers 2 and 3 on the tubing. The half of the saturated culture that is retained between gates (b) and (c) is now the sampling chamber and sample can be removed using a syringe without the risk of contaminating the culture chamber.

and chemostats, the Genetic Engine cannot be used for the continuous culture of microbes that require solid surfaces for growth or when the substrate is not soluble. A more recent development in long-term microbial culture is another proprietary device known as the Evolugator (Patent WO/ 2005/083052). In this method, the traditional bioreactor is replaced with a single length of flexible translucent tubing that is partitioned using gates, which are essentially clamps that pinch the tubing to prevent flow of culture and cells from one partition to another (Fig. 6). The tubing is filled with medium, partitioned with clamps and sterilized before loading on the Evolugator, a fully automated machine that initiates dilution. Cells to be evolved are inoculated into the growth chamber and mixed with culture medium using a sterile syringe that can penetrate the tubing. Multiple turbidimeters monitor the growth of cells by reading cell density through the translucent tubing. The tubing surrounding the growth chamber is enclosed in an environmentally controlled box to maintain constant culture conditions. Upon dilution, the tubing and the gates that partition it are moved in unison, resulting in the movement of the contents inside the tubing by peristaltic action. In this manner, fresh medium from

302

Modes of Microbial Culture

upstream of the growth chamber is mixed with half of the saturated culture in the growth chamber and allowed to grow again. The other half of the culture in the growth chamber is removed and is now in a chamber called the sampling chamber. Samples of the adapted microbes can be removed from this chamber using a syringe. The entire process is repeated, essentially making a continuous serial batch process in a single mobile vessel. Gas bubbles can be maintained between partitions so that fresh gas can be provided with each dilution. In addition, gas permeable tubing can be used to increase exchange with gases in the environmentally controlled box. Agitation is achieved by rocking the environmentally controlled box, causing the bubble to move back and forth through the culture chamber, making agitation akin to an airlift reactor. In theory, the time of dilution can be matched to the growth rate of the microbe, converting the method from a serial batch process to a continuous culture process. The Evolugator method provides a variety of benefits. First, dilution is controlled using specialized software linked to real-time measurement of turbidity; and dilution can be achieved at any phase of the growth curve and at any dilution ratio (a dilution ratio of 0.5 is shown in Fig. 6). This limits the severity of genetic bottlenecks and accelerates the appearance of adaptive mutations. Second, the vessel is a single length of tubing, and once the experiment has been started, there is no need to expose the culture to the outside environment. Moreover, the tubing can be made any length to accommodate experiments of varying timescales, creating a closed environment for the entire experiment. In addition, the physical separation of the sampling chamber from the growth chamber nearly eliminates the problem of contamination that plagues serial transfer. Third, as a portion of the growth chamber wall is removed with every dilution, an effective counter-selection against wall growth is provided. Fourth, as the Evolugator culture chamber is a continuous length of tubing, it can be used to culture cells that require a continuous surface upon which to grow. Finally, half of the contents of the culture chamber is retained and half is removed. This includes clumps of cells and substrate particles, which are evenly distributed between the effluent and what is retained in for the next round of growth. Consequently, the Evolugator can be effectively used for cells that do not grow evenly in suspension and for experiments in which the substrate is not soluble. The Evolugator is ideally suited for improving the growth rate of microorganisms (de Crécy et al., 2007). In an elegant example of the combination of genetic engineering and experimental evolution, the Evolugator was used to improve succinate production by a strain of E. coli. Genomatica, Inc. (Patent WO/2007/030830) used metabolic engineering to produce a strain of E. coli that was genetically crippled so that the only way to regenerate NADþ from NADH anaerobically was through the production of succinate. This was achieved by deleting the pfl (encoding pyruvate formate lyase), ldh (encoding lactate dehydrogenase), and adhE (encoding alcohol dehydrogenase) genes. The resultant strain produced approximately threefold more succinate than the ancestral wild type. However, it was growth attenuated. As the anaerobic growth rate of this strain was tied to succinate production, the selection for faster growing variants using the Evolugator was used to generate strains that produced more succinate. Although the unevolved input strain produced succinate with an approximate mass yield on glucose of 0.15, the evolved ouput strain was capable of a mass yield of 0.70. The Evolugator has also been applied to increase the maximal growth temperature of Metarhizium anisopliae, a filamentous entomopathogenic fungus that is commercially used as a biocontrol agent (de Crécy et al., 2009). The widespread use of M. anisopliae to control insects is limited by its poor resistance to high temperatures, a trait that insects take advantage of by generating body heat (fever) or by sunbathing (behavioral fever) in response to infection. A thermotolerant strain of M. anisopliae could potentially be a better entomopathogen by circumventing these febrile responses. Indeed, the thermotolerant strain was shown to more rapidly kill insects, although it is not yet clear if this trait is linked to the thermotolerant phenotype. Finally, the Evolugator was used to convert E. coli from a mesophile, preferring 37
C, to a facultative mesophile that actually grows better at 46
C (Blaby et al., 2012).

1.20.4

Using Microbial Communities for Industrial Purposes

It has long been recognized that microbial communities play a critical role in many industrial processes, from cheese making and beer brewing to wastewater treatment and anaerobic biogas production (Hays et al., 2015). Not surprisingly, we are seeing a paradigm shift in industrial microbiology away from microbial mono-cultures and toward the use of microbial co-cultures, which are defined as cultures of more than one strain of microorganism either cultured together or in a way they can share metabolites or enzymes (Goers et al., 2014; Bader et al., 2010). For example, some strains of antibiotic-producing Actinomycetes have been shown to produce more or different antibiotics when grown in co-culture, thereby improving antibiotic production or generating novel antibiotics (Moody, 2014; Antoraz et al., 2015). Alternatively, the use of multiple strains could greatly expand the metabolic capabilities of co-culture cellular factory resulting in more efficient feedstock utilization and improved product titers. In one interesting example, multiple engineered strains of E. coli were used to produce the valuable renewable chemical, muconic acid, with one strain producing an essential intermediate required by the other to make product (Zhang et al., 2015). Recent discoveries have also placed new emphasis on the importance of the microbiome in human health, with a flood of multi-strain and multi-species probiotic microbial products hitting the market. In this case, the thought is that different strains and species occupy different niches in the intestinal tract and having more than one microbe would augment the beneficial effects. Indeed, there is some evidence that co-culture probiotics perform better than monocultures for certain goals (Timmerman et al., 2004). Of course, these products claim to help with everything from weight loss to boosting the immune system and preventing Alzheimer’s disease (Hill et al., 2014), but since the industry is largely unregulated, it is unclear yet what the real benefits are.

Modes of Microbial Culture

1.20.5

303

Concluding Remarks

The exploitation of biological systems to reduce the costs of producing many goods has undergone a revolution in the last century. In addition, there are a variety of products, including some antibiotics and other pharmaceuticals that can only be made using biocatalysts. The importance of biocatalysis will only grow as we strive to replace petroleum as a feedstock for the production of fuels, plastics, and other staple commodities. This will require the continual development and implementation of innovative modes of microbial culture to solve the problems that are unique to each biocatalytic process. In addition, in the use of microbial culture to make tangible products, there is a growing resurgence in the use of long-term culture for experimental evolution as either a complement to or a replacement for traditional genetic engineering. A clear advantage obtained with experimental evolution is that evolution itself is harnessed to alter the organism at a global level, enabling complex phenotypes to be altered on a global scale. Numerous recent experimental evolution investigations have suggested that even in wellstudied systems adaptations evolve that may not be intuitive to those taking a solely reverse-genetics approach. Although experimental evolution experiments are currently being primarily used for strain enhancement, the property of being able to freeze and subsequently revisit a series of evolved strains (and thus directly compare evolved and ancestral strains), coupled with advances in whole genome sequencing, is making the technique an important tool for the elucidation for the molecular basis of the novel phenotype. Thus, experimental evolution should be seen as a complementary tool in the biologists’ toolkit for the development of industrially useful organisms.

References Antoraz, S., Santamaria, R.I., Diaz, M., et al., 2015. Toward a new focus in antibiotic and drug discovery from the Streptomyces arsenal. Frontiers in Microbiology 6, 461. Bader, J., Mast-Gerlach, E., Popovic, M.K., et al., 2010. Relevance of microbial coculture fermentations in biotechnology. Journal of Applied Microbiology 109, 371–387. Baltzis, B.C., Fredrickson, A.G., 1983. Competition of two microbial populations for a single resource in a chemostat when one of them exhibits wall attachment. Biotechnology and Bioengineering 25, 2419–2439. Bilford, H.R., Scalf, R.E., Stark, W.H., Kolachov, P.J., 1942. Alcoholic fermentation of molasses – Rapid continuous fermentation process. Industrial and Engineering Chemistry 34, 1406–1410. Blaby, I.K., Lyons, B.J., Wroclawska-Hughes, E., et al., 2012. Experimental evolution of a facultative thermophile from a mesophilic ancestor. Applied and Environmental Microbiology 78, 144–155. Caspeta, L., Chen, Y., Ghiaci, P., et al., 2014. Biofuels. Altered sterol composition renders yeast thermotolerant. Science 346, 75–78. Caspeta, L., Nielsen, J., 2015. Thermotolerant yeast strains adapted by laboratory evolution show trade-off at ancestral temperatures and preadaptation to other stresses. mBio 6, e00431. Chen, J., Shen, J., Ingvar Hellgren, L., et al., 2015. Adaptation of Lactococcus lactis to high growth temperature leads to a dramatic increase in acidification rate. Scientific Reports 5, 14199. Chesbro, W., Evans, T., Eifert, R., 1979. Very slow growth of Escherichia coli. Journal of Bacteriology 139, 625–638. de Crécy, E., Jaronski, S., Lyons, B., et al., 2009. Directed evolution of a filamentous fungus for thermotolerance. BMC Biotechnology 9, 74. de Crécy, E., Metzgar, D., Allen, C., et al., 2007. Development of a novel continuous culture device for experimental evolution of bacterial populations. Applied Microbiology and Biotechnology 77, 489–496. de Crécy-Lagard, V.A., Bellalou, J., Mutzel, R., Marlière, P., 2001. Long term adaptation of a microbial population to a permanent metabolic constraint: Overcoming thymineless death by experimental evolution of Escherichia coli. BMC Biotechnology 1, 1–7. Dallinger, W.H., 1878. On the life history of a minute septic organism: With an account of experiments made to determine its thermal death point. Proceedings of the Royal Society of London 27, 332–350. Dykhuizen, D.E., Hartl, D.L., 1983. Selection in chemostats. Microbiological Reviews 47, 150–168. Elena, S., Lenski, R., 2003. Evolution experiments with microorganisms: The dynamics and genetic bases of adaptation. Nature Reviews Genetics 4, 457–469. Finkel, S.E., 2006. Long-term survival during stationary phase: Evolution and the GASP phenotype. Nature Reviews. Microbiology 4, 113–120. Fong, S., Palsson, B., 2004. Metabolic gene-deletion strains of Escherichia coli evolve to computationally predicted growth phenotypes. Nature Genetics 36, 1056–1058. Goers, L., Freemont, P., Polizzi, K.M., 2014. Co-culture systems and technologies: Taking synthetic biology to the next level. Journal of the Royal Society, Interface 11. Gostomski, P., Mühlemann, M., Lin, Y.-H., et al., 1994. Auxostats for continuous culture research. Journal of Biotechnology 37, 167–177. Govindaswamy, S., Vane, L.M., 2010. Multi-stage continuous culture fermentation of glucose–xylose mixtures to fuel ethanol using genetically engineered Saccharomyces cerevisiae 424A. Bioresource Technology 101, 1277–1284. Graves Jr., J.L., Tajkarimi, M., Cunningham, Q., et al., 2015. Rapid evolution of silver nanoparticle resistance in Escherichia coli. Frontiers in Genetics 6, 42. Harden, M.M., He, A., Creamer, K., et al., 2015. Acid-adapted strains of Escherichia coli K-12 obtained by experimental evolution. Applied and Environmental Microbiology 81, 1932–1941. Hays, S.G., Patrick, W.G., Ziesack, M., et al., 2015. Better together: Engineering and application of microbial symbioses. Current Opinion in Biotechnology 36, 40–49. Hill, J.M., Bhattacharjee, S., Pogue, A.I., Lukiw, W.J., 2014. The gastrointestinal (GI) tract microbiome and potential link to Alzheimer’s disease (AD). Frontiers in Neurology 5. Jansen, M.L.A., Daran-Lapujade, P., De Winde, J.H., et al., 2004. Prolonged maltose-limited cultivation of Saccharomyces cerevisiae selects for cells with improved maltose affinity and hypersensitivity. Applied and Environmental Microbiology 70, 1956–1963. Jarboe, L.R., Grabar, T.B., Yomano, L.P., et al., 2007. Development of ethanologenic bacteria. Advances in Biochemical Engineering/Biotechnology 108, 237–261. Kolter, R., Siegele, D.A., Tormo, A., 1993. The stationary phase of the bacterial life cycle. Annual Review of Microbiology 47, 855–874. Kurosawa, K., Laser, J., Sinskey, A.J., 2015. Tolerance and adaptive evolution of triacylglycerol-producing Rhodococcus opacus to lignocellulose-derived inhibitors. Biotechnology for Biofuels 8, 76. Lacroix, R.A., Sandberg, T.E., O‫׳‬brien, E.J., et al., 2015. Use of adaptive laboratory evolution to discover key mutations enabling rapid growth of Escherichia coli K-12 MG1655 on glucose minimal medium. Applied and Environmental Microbiology 81, 17–30. Larsen, B., Rask Poulsen, B., Eriksen, N.T., Lønsmann Iversen, J.J., 2003. Homogeneous batch cultures of Aspergillus oryzae by elimination of wall growth in the variomixing bioreactor. Applied Microbiology and Biotechnology 64, 192–198. Latif, H., Sahin, M., Tarasova, J., et al., 2015. Adaptive evolution of Thermotoga maritima reveals plasticity of the ABC transporter network. Applied and Environmental Microbiology 81, 5477–5485.

304

Modes of Microbial Culture

Lee, J., Saddler, J.N., Um, Y., Woo, H.M., 2016a. Adaptive evolution and metabolic engineering of a cellobiose- and xylose- negative Corynebacterium glutamicum that co-utilizes cellobiose and xylose. Microbial Cell Factories 15, 20. Lee, S.H., Kim, M.S., Lee, J.H., et al., 2016b. Adaptive engineering of a hyperthermophilic archaeon on CO and discovering the underlying mechanism by multi-omics analysis. Scientific Reports 6, 22896. Lewis, J.G., Northcott, C.J., Learmonth, R.P., et al., 1993. The need for consistent nomenclature and assessment of growth phases in diauxic cultures of Saccharomyces cerevisiae. Journal of General Microbiology 139, 835–839. Lopez-Malo, M., Garcia-Rios, E., Melgar, B., et al., 2015. Evolutionary engineering of a wine yeast strain revealed a key role of inositol and mannoprotein metabolism during lowtemperature fermentation. BMC Genomics 16, 537. McCarthy, S., Johnson, T., Pavlik, B.J., et al., 2015. Expanding the limits of thermoacidophily in the archaeon Sulfolobus solfataricus by adaptive evolution. Applied and Environmental Microbiology 82, 857–867. Meng, X., Yang, J., Xu, X., et al., 2009. Biodiesel production from oleaginous microorganisms. Renewable Energy 34, 1–5. Mondragón-Parada, M.E., Ruiz-Ordaz, N., Tafoya-Garnica, A., et al., 2008. Chemostat selection of a bacterial community able to degrade s-triazinic compounds: Continuous simazine biodegradation in a multi-stage packed bed biofilm reactor. Journal of Industrial Microbiology & Biotechnology 35, 767–776. Monod, J., 1945. Sur la nature du phénomène de diauxie. Annales de L’Institut Pasteur 71, 37–38. Monod, J., 1949. The growth of microbial cultures. Annual Reviews of Microbiology 3, 371–394. Monod, J., 1950. La technique de culture continue. Théorie et applications. Ann Inst Pasteur 19. Moody, S.C., 2014. Microbial co-culture: Harnessing intermicrobial signaling for the production of novel antimicrobials. Future Microbiology 9, 575–578. Moreau, P.L., Loiseau, L., 2016. Characterization of acetic acid-detoxifying Escherichia coli evolved under phosphate starvation conditions. Microbial Cell Factories 15, 42. Myers, J., Clark, L.B., 1944. Culture conditions and the development of the photosynthetic mechanism: II. An apparatus for the continuous culture of Chlorella. The Journal of General Physiology 28, 103–112. Novick, A., Szilard, L., 1950. Description of the chemostat. Science 112, 715–716. Oide, S., Gunji, W., Moteki, Y., et al., 2015. Thermal and solvent stress cross-tolerance conferred to Corynebacterium glutamicum by adaptive laboratory evolution. Applied and Environmental Microbiology 81, 2284–2298. Park, Y.-C., Seo, J.-H., 2004. Optimization of culture conditions for D-ribose production by transketolase-deficient Bacillus subtilis JY1. Journal of Microbiology and Biotechnology 14, 8. Pereira, S.R., Sanchez, I.N.V., Frazao, C.J., et al., 2015. Adaptation of Scheffersomyces stipitis to hardwood spent sulfite liquor by evolutionary engineering. Biotechnology for Biofuels 8, 50. Pin, C., Rolfe, M.D., Munoz-Cuevas, M., et al., 2009. Network analysis of the transcriptional pattern of young and old cells of Escherichia coli during lag phase. BMC Systems Biology 3, 108. Shiloach, J., Fass, R., 2005. Growing E. coli to high cell density – A historical perspective on method development. Biotechnology Advances 23, 345–357. Sonderegger, M., Sauer, U., 2003. Evolutionary engineering of Saccharomyces cerevisiae for anaerobic growth on xylose. Applied and Environmental Microbiology 69, 1990–1998. Timmerman, H.M., Koning, C.J.M., Mulder, L., et al., 2004. Monostrain, multistrain and multispecies probiotics – A comparison of functionality and efficacy. International Journal of Food Microbiology 96, 219–233. Tremblay, P.L., Hoglund, D., Koza, A., et al., 2015. Adaptation of the autotrophic acetogen Sporomusa ovata to methanol accelerates the conversion of CO2 to organic products. Scientific Reports 5, 16168. Vanags, J., Rychtera, M., Ferzik, S., et al., 2007. Oxygen and temperature control during the cultivation of microorganisms using substrate feeding. Engineering in Life Sciences 7, 247–252. Voordeckers, K., Kominek, J., Das, A., et al., 2015. Adaptation to high ethanol reveals complex evolutionary pathways. PLOS Genetics 11, e1005635. Wahl, L., Gerrish, P., Saika-Voivod, I., 2002. Evaluating the impact of population bottlenecks in experimental evolution. Genetics 162, 961–971. Wallace-Salinas, V., Brink, D.P., Ahren, D., Gorwa-Grauslund, M.F., 2015. Cell periphery-related proteins as major genomic targets behind the adaptive evolution of an industrial Saccharomyces cerevisiae strain to combined heat and hydrolysate stress. BMC Genomics 16, 514. Wallace-Salinas, V., Gorwa-Grauslund, M.F., 2013. Adaptive evolution of an industrial strain of Saccharomyces cerevisiae for combined tolerance to inhibitors and temperature. Biotechnology for Biofuels 6, 151. Yi, Z., Xu, M., Magnusdottir, M., et al., 2015. Photo-oxidative stress-driven mutagenesis and adaptive evolution on the marine diatom Phaeodactylum tricornutum for enhanced carotenoid accumulation. Marine Drugs 13, 6138–6151. Yomano, L., York, S., Zhou, S., et al., 2008. Re-engineering Escherichia coli for ethanol production. Biotechnology Letters 30, 2097–2103. Yoneda, A., Henson, W.R., Goldner, N.K., et al., 2016. Comparative transcriptomics elucidates adaptive phenol tolerance and utilization in lipid-accumulating Rhodococcus opacus PD630. Nucleic Acids Research 44, 2240–2254. Zhang, H., Pereira, B., Li, Z., Stephanopoulos, G., 2015. Engineering Escherichia coli coculture systems for the production of biochemical products. Proceedings of the National Academy of Sciences 112, 8266–8271.

1.21

Photosynthesis and Photoautotrophy

NPA Hu¨ner, University of Western Ontario, London, ON, Canada B Grodzinski, University of Guelph, Guelph, ON, Canada © 2011 Elsevier B.V. All rights reserved. This is a reprint of N.P.A. Hüner, B. Grodzinski, 1.23 - Photosynthesis and Photoautotrophy, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 315-322.

1.21.1 1.21.2 1.21.3 1.21.4 1.21.5 1.21.6 References

Introduction Energy Absorption, Trapping, Conversion, and Storage Photostasis and Cellular Energy Imbalance Photoacclimation Tailors the Photosynthetic Apparatus Acclimation to Low Temperature Mimics Photoacclimation Conclusions

305 307 308 309 310 311 312

Glossary Light-harvesting complex Major chlorophyll a/b containing pigment–protein complexes involved in the initial absorption of light energy for photosynthesis and its subsequent transfer to reaction centers of plants and green algae. Nonphotochemical quenching The dissipation of excess absorbed light energy as heat. P680 Reaction center chlorophyll a associated with photosystem II. P700 Reaction center chlorophyll a associated with photosystem I. Phenotypic plasticity The ability of an organism to alter its form and function in response to environmental cues. Photoacclimation Adjustments in the structure and function of the photosynthetic apparatus to long-term changes in light intensity. Photoautotrophy The use of sunlight to convert CO2 into organic materials to be utilized in various cellular functions. Photoinhibition The light-dependent inhibition of photosynthesis. Photoprotection Cellular mechanisms by which photosynthetic organisms prevent damage to the photosynthetic apparatus when exposed to excess light. Photostasis The maintenance of energy balance or energy homeostasis in photosynthetic organisms. Photosystems Protein–pigment complexes involved in the conversion of absorbed light into electrochemical energy. Reaction center A specialized protein complex present within each photosystem that binds a special chlorophyll a and initiates light-dependent electron transport. Retrograde regulation Control of nuclear gene expression by organelles such as chloroplasts and mitochondria. Rubisco The enzyme that catalyzes the initial photosynthetic fixation of CO2 by the Calvin cycle.

1.21.1

Introduction

Life is an endergonic process. Energy needed to sustain life is required, by and large, to maintain structural and functional order. Such order transcends all spatial scales. It can be observed and quantified at the levels of atoms, molecules, cells, tissues, organs, whole organisms, communities, and ecosystems. Thermodynamically, energy is required to counteract the seemingly inevitable increase in entropy that is governed by the second law of thermodynamics. Thus, Schrödinger explained life as negative entropy.1 The ultimate source of the energy for almost all organisms on this planet is sunlight. One may reconcile that evolution has harnessed sunlight as the energy source for life because it is cheap, energy rich, abundant, readily available, and present in seemingly inexhaustible quantities when measured on a biological timescale. Only after the advent of photoautotrophy do we witness the creation of an oxygen-rich environment and an explosion in the diversity of both aquatic and terrestrial organisms.2 Photoautotrophic organisms are those which can utilize sunlight as their source of energy to synthesize organic complex compounds from CO2. The general mechanism by which sunlight is harvested and harnessed is called photosynthesis, the development of which required millions of years of evolution, and is first observed in the prokaryotesdphotosynthetic bacteria and cyanobacteria. With the advent of the eukaryotic cell, chloroplasts evolved through endosymbiosis to produce the green algae and land plants. However, with only some minor modifications, the mechanism by which sunlight is harnessed by photoautotrophs remains unchanged even after 4 billion years of evolution! Through biochemical and biotechnological advances, scientists have attempted and continue to try to simulate the biological process of photosynthesis through artificial photosynthesis, but none have been successful.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00051-3

305

306

Photosynthesis and Photoautotrophy

Thus, photoautotrophs are crucial in linking all other living organisms to the Sun through their ability to absorb, trap, and convert this light energy into useable forms of electrochemical potential energy (proton motive force, PMF) primarily in the form of a trans-thylakoid DpH and redox potential energy in the form of reduced nicotinamide adenine diphosphate (NADPH).3 Once the light energy has been converted to a DpH, it is enzymatically converted to ATP. The process of light absorption, energy trapping, and conversion takes place in specialized thylakoid membranes localized within chloroplasts.3 The light-dependent biosynthesis of ATP and NADPH may be regarded as end of the primary energy conversion process in all photoautotrophs. Subsequently, this cellular energy is stored in a much more stable form, reduced C intermediates or primary photoassimilates. In order to have a net uptake of inorganic C, there must be a reduction of CO2 first to triose-P via the Calvin cycle localized in the chloroplast stroma and then to the stable metabolic end products, either starch, within the chloroplast, or sucrose in the cytosol. Starch is synthesized in the light whenever the rate of CO2 assimilation exceeds either the rate at which sucrose is respired in the light or the rate at which sucrose is exported from the cell to other cells within a leaf or transported long distance through the vascular system from the leaf (source tissue) to sink tissues such as roots that are net consumers of carbon. In addition to these fates for fixed carbon, cereals can convert sucrose to a polymeric form called fructans for storage within the cell vacuole. The capacity for the biosynthesis and storage of fructans provides an additional carbon sink in many cereals.4 Although the light-dependent reduction of CO2 is considered the primary photosynthetic process, the photochemical generation of electrons is also consumed in the process of N and S reduction. Thus, not only C assimilation but also N and S assimilation should be considered photosynthetic.5 Evolution has integrated extremely fast biophysical processes such as light absorption and energy transfer that occur on a femtosecond (fs ¼ 1015 s) to picosecond (ps ¼ 1012 s) timescale with photochemical reactions that occur on a microsecond timescale (ms ¼ 106 s) (Table 1). Not only are these processes the fastest in biology, but they are also insensitive to temperature within the biologically relevant range of 0–30  C. These extremely fast, temperature-independent processes are, in turn, integrated with much slower, temperature-dependent biochemical redox reactions involved in photosynthetic electron transfer, which occur on a millisecond (ms ¼ 103 s) timescale, CO2 reduction, and carbohydrate biosynthesis and export, which occur on the timescale of seconds (s) to minutes (Table 2). The ultimate sink for metabolism is growth and development that occur at rates that are generally measured on a timescale of hours to days to months (Table 2). This means that the maintenance of photoautotrophic life requires the integration of processes that occur on drastically different spatial scales that vary from the atomic and molecular levels to the whole organism as well as temporal scales that differ by more than 10 orders of magnitude (Tables 1 and 2). Due to these disparate temporal and spatial scales, there must always be a potential for an imbalance in cellular energy budget due to the inability of the slowest biochemical steps to keep pace with the faster photochemical reactions. Unless this potential imbalance is overcome, the uncontrolled photochemistry will destroy the organism. The solution to this apparent conundrum is part of the magic of photosynthesis and is reflected in the remarkable plasticity of photoautotrophs. The capacity of terrestrial and aquatic photosynthetic organisms to adjust to or to acclimate to an environment that is changing with respect to temperature, light, CO2, and nutrient status on a daily as well as on a seasonal basis is dependent upon two important factors. First, the actual genetic makeup or genotype of the plant determines the potential of any species to acclimate. Second, the capacity to regulate the expression of this genome in response to environmental cues such as light, temperature, photoperiod, and water status is indicative of the remarkable flexibility or plasticity that a single species can exhibit with respect to form and function, that is, phenotypic plasticity. Although both factors are inextricably linked, it is the capacity to alter form and function in response to a changing environment that governs plant productivity and geographical distribution. As we will discuss below, photoautotrophs are constantly attempting to balance energy input as absorbed light energy with energy utilization through metabolism and growth. The attempts to maintain an energy balance can be detected not only at the level of photochemical and biochemical regulation but also at the level of gene regulation, which results in observable changes in phenotype.

Table 1

Temporal scales for energy trapping at the photosystem level

PSII (t1/2) (1) RC Excitation s1 can be estimated in vivo by the pulse amplitude modulated chlorophyll a fluorescence quenching parameter 1 – qP.17,18 An increase in 1 – qP induced by various environmental conditions has been called excitation pressure.17,18 Excitation pressure thus reflects the relative reduction state of QA, that is, [QA]/[QA þ QA], providing a nondestructive means to explore changes in energy balance as a result of changing environmental conditions. Indeed, chlorophyll a fluorescence has been used extensively to examine the effect of numerous environmental stresses on photosynthetic function. Excitation pressure may be induced by changes in several different environmental parameters. For example, increasing growth irradiance at a constant temperature would cause an over-reduction of QA due to an increase in irradiance and thus an increase in sPSII∙Ek. Assuming no changes in the capacity to utilize the absorbed energy, that is, no change in s1, energy balance would be disrupted. Theoretically, a similar over-reduction of QA could be created by maintaining the same irradiance but decreasing the growth temperature. The lower temperature would decrease the rate of the biochemical reactions that utilize the absorbed energy, decreasing s1, with no change in sPSII∙Ek. Similarly, drought or the lack of specific essential nutrients would also cause a decrease in

Photosynthesis and Photoautotrophy

309

s1 due to limitations in the availability of electron acceptors such as CO2, NO3  , or SO4 2 . The assimilation of NO3  alone is estimated to account for about 25% of total energy expenditures.24 What is/are the redox sensor(s) for excitation pressure? Early research with green algae indicated that a key sensor was the redox state of PQ, a mobile electron carrier that shuttles electrons from PSII to the cytochrome b6/f complex.17,18,25 This was based on experiments where the characteristic, yellow-green, high-light phenotype brought about by acclimation to high irradiance could be mimicked by chemically modulating the redox status of the intersystem PQ pool with the electron transport inhibitor, 2,5dibromo-3-methyl-6-isopropylbenzoquinone (DBMIB), in Dunaliella tertiolecta25 and Chlorella vulgaris.17,26 Since DBMIB inhibits the oxidation of PQH2 by the Cyt b6/f complex, PSII keeps the PQ pool reduced in the light. This induces the high-light phenotype that is characterized by low chlorophyll content per cell, high chlorophyll a/b ratio (>10), accumulation of the carotenoid-binding protein (CBR) but suppression of both Lhcb2 accumulation and Lhcb2 expression, the nuclear gene that encodes the major PSII light-harvesting antenna polypeptide.26 While low temperature does not affect the rate of light absorption, it severely restricts the rate of downstream, enzyme-catalyzed reactions. This restricts the capacity to utilize NADPH and ATP, the products of the PET, thus causing an over-reduction of the PQ pool due to negative feedback. As a consequence, the yellow, low-temperature phenotype is indistinguishable from the phenotype observed in the presence of DBMIB.17,26 In contrast, since 3-(30 ,40 -dichlorophenyl)-1,1-dimethylurea (DCMU) prevents the exit of electrons from PSII into the PQ pool, PSI is able to keep the PQ pool oxidized in the light. Under these conditions, cells exhibit a normal green phenotype that is associated with high chlorophyll content per cell, low chlorophyll a/b ratio (3.0–3.5), and high levels of Lhcb2 expression and Lhcb2 accumulation.17,25,26 This phenotype can also be generated by growth at either low irradiance or high temperature in C. vulgaris.17,26 In contrast to green algae, in Plectonema boryanum it was reported that the redox sensor for excitation pressure is not the PQ pool but rather must reside in the PET chain downstream of the PQ pool.27 Furthermore, recent research in Arabidopsis suggests that redox factors on the acceptor side of PSI may be important in redox signaling.28 These as well as additional signals including the precursor of chlorophyll synthesis, magnesium protoporphyrin,29 and ROS generated by the PET30 may constitute a complex network of signals involved in the retrograde pathway of communication from the chloroplast to the nucleus.31 Genetic analyses in Arabidopsis has identified STN732 as a chloroplast protein kinase involved in redox signaling essential for state transitions and photosynthetic acclimation.33 Yet the exact nature of the mechanisms by which the redox state of the chloroplast is signaled to the nucleus resulting in altered gene expression remains largely unknown. In nature, photosynthetic organisms are exposed to daily and seasonal fluctuations in irradiance, temperature, water, and nutrient availability. These can result in the induction of high excitation pressure and potential photodamage to PSII reaction centers.34 How do photoautotrophs maintain photostasis under such unpredictable changes in their environment? This question is addressed below with respect to changes in irradiance, light quality, and low temperature.

1.21.4

Photoacclimation Tailors the Photosynthetic Apparatus

The ability of photosynthetic organisms to adjust the structure and function of their photosynthetic apparatus in response to changes in growth irradiance is called photoacclimation. One mechanism of photoacclimation involves changes in sPSII through the modulation of the size and composition of the LHCII and LHCI of PSII and PSI, respectively. The green algae, D. tertiolecta and C. vulgaris, reduce the size of LHCII in response to modulation of the redox state of the PQ pool by high irradiance.25,26 This tailoring of the photosynthetic apparatus requires precise spatial and temporal coordination between the chloroplast, the nucleus, and the cytosol through retrograde regulation35 to ensure the establishment homeostasis and cellular energy balance. This response is mimicked by chemically modulating the redox state of the PQ pool. When the PQ pool is reduced by exposure to light in the presence of DBMIB, simulating sPSII∙Ek > s1, the transcription of the Lhcb genes is downregulated, decreasing the size of the LHCII and producing a yellow, high-light phenotype. In contrast, the PQ pool remains oxidized in the presence of DCMU, mimicking sPSII∙Ek < s1, and cells maintain a green, low-light phenotype.25,26 This acclimation mechanism is consistent with the notion that energy balance in response to high light may be attained through modulation of sPSII. It is now established that the xanthophyll cycle, championed by Demmig-Adams and Adams, is an important regulator of nonphotochemical dissipation of excess light.36 Xanthophyll-cycle-dependent antenna quenching is due to the light-dependent conversion of the light-harvesting xanthophyll, violaxanthin, to the energy-quenching xanthophylls, antheraxanthin and zeaxanthin. There is now a consensus that a close relationship exists between the increase in the capacity for nonphotochemical quenching (NPQ), the extent of the thylakoid DpH, and the increase in xanthophyll-cycle activity.36 The capacity for NPQ is also closely related to the expression of PsbS, a gene required for NPQ in Arabidopsis thaliana.37 Acclimation to prolonged exposure to high light appears to result, first, in an increase in xanthophyll-cycle pigments and, second, in a persistent engagement of the xanthophyll cycle and sustained antenna quenching of excess energy through NPQ.36 This aids in the maintenance of energy balance via a functional decrease in sPSII. However, the molecular mechanism underlying NPQ remains equivocal and controversial. One mechanism proposes that zeaxanthin itself acts directly to quench excess energy nonphotochemically within PSII antenna,37 whereas, alternatively, zeaxanthin may regulate NPQ indirectly by altering the organization and aggregation state of LHCII.22 Irrespective of the molecular mechanism, the capacity to regulate NPQ to maintain energy balance has a dramatic impact on the fitness of A. thaliana measured as net seed production under natural field conditions.38

310

1.21.5

Photosynthesis and Photoautotrophy

Acclimation to Low Temperature Mimics Photoacclimation

Based on the hypothesis that photosynthetic organisms respond to energy imbalance rather than high light or low temperature per se, low temperature should induce a high-light phenotype as the organism adjusts to decreased sink capacity (s1). This was demonstrated to be the case during cold acclimation of the unicellular green algae C. vulgaris and Dunaliella salina, where the regulation of photosynthesis by growth at low temperature and moderate irradiance 5  C/150 mmol m2 s1 (5/150) mimics photoacclimation at high light and moderate temperatures (27/2200).17,26 Cells grown at 5/150 are indistinguishable from those grown at 27/2200 with respect to photosynthetic efficiency, photosynthetic capacity, pigmentation, Lhcb content, and sensitivity to photoinhibition. These results are explained on the basis that cultures grown at either 5/150 or 27/2200 are exposed to comparable excitation pressure measured as 1 – qP.17,26 Similar conclusions regarding the role of excitation pressure have been reported for thermal and photoacclimation of Laminaria saccharina39 and the filamentous cyanobacterium, P. boryanum.27 These results are consistent with the thesis that exposure to low temperature creates a similar imbalance in energy budget as exposure to high light, and that similar protective mechanisms are utilized to defend the organism. Neither C. vulgaris nor D. salina is able to upregulate carbon metabolism and thus to adjust the capacity of electron-consuming sinks during growth and development at low temperature.17,18 As a consequence, these organisms exhibit a minimal capacity to adjust s1. This was observed for C. vulgaris through its inability to adjust exponential growth rates as a function of growth irradiance during growth at either 5 or 27  C.17 As a result, C. vulgaris and D. salina appear to primarily adjust sPSII through a reduction in the size of the PSII LHC coupled with an increased capacity for NPQ through the xanthophyll cycle to dissipate excess energy to maintain an energy balance under changing growth conditions.17,26 Cold temperate conifers such as lodgepole pine (Pinus contorta L.) and herbaceous cereals such as winter wheat (Triticum aestivum L.) and winter rye (Secale cereale L.) are representative of some of the most cold-tolerant plants.40 The capacity to cold acclimate is an essential requirement for surviving subzero temperatures during winter. However, these two groups of plants exhibit different strategies for the utilization of light energy during growth and cold acclimation.40 Cold acclimation of conifers induces the cessation of primary growth in contrast to winter cereals that require continued growth and development during the cold acclimation period to attain maximum freezing tolerance.26 In the context of these different growth strategies, the requirement for photosynthetic assimilates also differs considerably. Conifers exhibit a decreased requirement for photosynthetic assimilates upon the induction of dormancy and cold acclimation, representing a decrease in s1. In contrast, over-wintering cereals maintain a high demand for photoassimilates due to continued growth and development during cold acclimation. This keeps sink demand (s1) relatively constant. As a consequence of the decreased sink demand for photoassimilates, that is, a decrease in s1, conifers exhibit feedback inhibition of CO2 assimilation.40 To maintain energy balance under these conditions, conifers decrease their capacity and efficiency to absorb light by reducing PSII and LHC protein levels. In addition, conifers increase their capacity for NPQ through the upregulation of PsbS, accumulation of xanthophyll-cycle pigments, and aggregation of the major light-harvesting pigment proteins into energyquenching complexes.40 Energetically, this allows the plant to dissipate the majority of absorbed light as heat, effectively decreasing sPSII. This is one major reason that conifer needles stay green even during the coldest winter. Conifers recover fully from winterinduced quenched state with the onset of spring, suggesting that the capacity to downregulate photosynthesis during cold acclimation is an important mechanism for the successful establishment of evergreen conifers in cold-temperate and subarctic climates.40 In contrast to conifers, winter cereals such as wheat and rye grown at low temperatures maintain both high efficiency and capacity for light absorption with a minimum investment in NPQ.26,40 However, excitation pressure, measured as 1 – qP, is moderate due to the fact that a high flux of absorbed light energy (sPSII ∙ Ek) is matched by an increased capacity for CO2 assimilation through the upregulation of transcription and translation of genes coding for Rubisco and the regulatory enzymes of cytosolic sucrose and vacuolar fructan biosynthesis.41 Thus, the capacity for cold-acclimated wheat and rye to maintain energy balance upon exposure to low temperature appears to be primarily due to an enhanced capacity to utilize the absorbed light energy through an upregulation of carbon metabolism and growth at low temperature (s1). The reprogramming of carbon metabolism to match the continued absorption of light energy at low temperature has a dual function: It maximizes the chemical energy and carbon pool available for the renewed growth in the spring, and the accumulation of photosynthetic end products such as sucrose provides cryoprotectants to stabilize the cells during freezing events during the winter.41 Spring wheat cultivars exhibit a significantly lower capacity to maintain energy balance following cold acclimation, because they are unable to adjust carbon metabolism to as great an extent as winter cultivars.26 For maximal low temperature survival, upregulation of photosynthesis is absolutely critical for protection from both freezing and low-temperature-induced photodamage.34 Not surprisingly, the differential capacity to maintain cellular energy balance between winter and spring wheat cultivars is correlated with differential freezing tolerance and field survival.26 The herbaceous dicot, A. thaliana, appears to have an intermediate acclimation mechanism. In contrast to the extremes observed for conifers and winter cereals, cold-acclimated Arabidopsis exhibits an incomplete ability to adjust photosynthetic capacity relative to nonacclimated controls.17 As a consequence, the decreased sensitivity of Arabidopsis to photoinhibition appears to be the result of an upregulation of carbon metabolism (increased s1) combined with enhanced NPQ via the xanthophyll cycle to reduce sPSII. Thus, rather than only altering sPSII (green algae and conifers) or only adjusting s1 (winter cereals), Arabidopsis maintains energy balance through a combination of both processes. This is probably indicative of a more generalist approach than most plant species have to cold acclimation.

Photosynthesis and Photoautotrophy

1.21.6

311

Conclusions

A defining characteristic of all life is the ability to harness energy for the maintenance of its homeostasis, that is, maintenance of an energetic balance between complex, interdependent metabolic processes as well as between cellular compartments, tissues, and organs. They are all interconnected at varying spatial and temporal scales to form a complex, integrated living network that exhibits remarkable flexibility or phenotypic plasticity with respect to form and function. How is phenotypic plasticity regulated? Genotype determines the potential of any species to adjust or acclimate to a changing environment. However, the capacity to acclimate or adjust phenotype reflects the capacity to regulate the expression of this genome in response to environmental cues. Thus, this alters one’s view of the chloroplast, photosynthesis, and the role of the photosynthetic apparatus. Not only is the chloroplast a primary cellular energy transformer, but this organelle also acts as a global energy sensor whose impact extends beyond photosynthesis to plant and cell form and function by governing nuclear gene expression through retrograde regulation.17,19,26 Thus, sensing cellular energy imbalances is a major determinant in plant survival and productivity in an environment that constantly changes on a varying timescale of hours to days to weeks to years. Two fundamental principles have emerged from the research on excitation pressure and photostasis. First, although oxygenic photoautotrophs as varied as cyanobacteria, green algae, herbaceous plants, and conifers sense energy imbalances through changes in excitation pressure, the molecular response to this redox sensing/signaling pathway is species dependent.17,26 Green algae and cyanobacteria, overwintering evergreens, and spring cereals attain photostasis primarily by downregulation of light-harvesting efficiency coupled with an increase in NPQ to dissipate excess absorbed energy as heat. This occurs as a consequence of a lack of plasticity with respect to modulating sink capacity. In contrast, overwintering cereals, A. thaliana, Brassica napus, and Antarctic angiosperms minimize excitation pressure by stimulating sink capacity by upregulation of photosynthetic carbon metabolism, source–sink export capacity, and adjustment of leaf anatomy with minimal changes in sPSII.17 Second, the specific mechanism(s) used to attain photostasis is/are time dependent. On a short-term basis, transient energy partitioning mechanisms such as antenna quenching through the xanthophyll cycle and state transitions, both regulated by the redox state of the PQ pool and the transthylakoid DpH, are used to attain photostasis. In contrast, mechanisms during long-term steady-state growth and acclimation require regulation of gene expression and translation involved in either the structure and of the PSII–LHCII complex17,26,40 (Lhcb1, Lhcb2, psbA) or the increased levels of enzymes involved in CO2 assimilation (Rubisco, cFBPase, NADP-GPDH, PRK), sucrose synthesis (cyto-FBPase, SPS), and fructan biosynthesis.17,18,26,41 These molecular changes are combined with changes in plant phenotype. Thus, photoacclimation to attain photostasis is a ‘time-nested‘ phenomenon.21 Much of the recent research on the control of plant and crop productivity continues to be focused on the genomic and proteomic approach with little or no consideration for the contributions of plant architecture on plant biomass and fruit production. Plant growth and biomass production are the result of a systems-wide integration of light capture, energy sensing, and photosynthetic CO2 (C) and nitrate (N) assimilation.42 These processes are regulated from the level of the gene, to the cell, to the leaf, and ultimately to the whole-plant canopy level. Thus, in contrast to a purely genomics or proteomics approach, a broader approach that integrates plant morphology, physiology, biochemistry, and genetics/molecular biology that are indicative of altered patterns of energy flow will be important to understand the complexity of enhancing crop productivity. For example, some of our most potent agricultural herbicides such as DCMU were developed based on our understanding of their ability to inhibit primary electron transport processes during photosynthesis. Understanding the nuances of crop productivity will remain a critical challenge given the requirements to feed an ever-increasing human population under climate change conditions that will surely be suboptimal with respect to maximum crop growth and productivity. Although the ability of science to create artificial photosynthesis has not been successful to date, plants and microorganisms are being exploited as photosynthetic bioreactors. Biotechnology has allowed us to manipulate the genetic makeup and biosynthetic capacities of photoautotrophs to not only produce tradition products such as food but also create specific pharmaceuticals and nutraceuticals important for medicine and human health. This is called molecular farming. However, regardless of what the desired end product is, photosynthesis is at the heart of agriculture, forestry, environmental management, molecular farming, and the maintenance of life on the Earth. Of the total biomass of a typical crop plant, 96% is comprised of the three elements C, H, and O. This biomass is derived from photosynthesis and this is driven primarily by solar energy. We may use artificial light sources to supplement photoautotrophy in specialized situations such as greenhouses or algal chemostats, but, in the end, the simple biological fact is that we are using the unique photosynthetic capacity of plants, algae, and photosynthetic prokaryotes to harness one form of energy (light) and convert it into a wide range of usable forms of chemical energy based on reduced C and N as the core. The more complex the organism (e.g., a tree vs. a single-cell alga), the more we can appreciate the diverse spatial and temporal integration governed by the genotype that influences phenotypic plasticity. Of primary importance is an understanding of C partitioning and allocation of reduced C, N, and S compounds to different organs. Our exploitation of the unique nature of photoautotrophy and the continued use of complex photosynthetic bioreactors such as vascular plants, algae, and cyanobacteria, either in their natural form or as modified organisms, requires us to understand that light trapping provides a source of chemical energy that is consumed in the production of reduced C as triose-P, the first chemical sink for this absorbed light energy (Tables 1 and 2). Thus, a better understanding of the molecular basis by which photosynthesis and photoacclimation are coupled to phenotypic plasticity, crop productivity, and plant and algal survival mechanisms will be essential in addressing the challenge of at least maintaining and even perhaps enhancing sustainable production systems under the suboptimal growing conditions due to climate change events.

312

Photosynthesis and Photoautotrophy

See Also: 4.00 Introduction; 4.16 Increasing Photosynthesis/RuBisCO and CO2-Concentrating Mechanisms; 4.17 Improving Plant Nitrogen-Use Efficiency; 4.18 Sulfur Metabolism in Plants and Related Biotechnologies; 4.25 Microalgae as Bioreactors for Production of Pharmaceutical Proteins; 4.26 Algal Chemostats.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.

25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42.

Schrödinger, E. What Is Life?, 90p., Cambridge University Press: Cambridge, 1947. Dawkins, R. The Ancestor‘s Tale: A Pilgrimage to the Dawn of Life, 528p., Weidenfeld and Nicholson: London, 2004. Hopkins, W. G.; Huner, N. P. A. Introduction to Plant Physiology, 512p., John Wiley and Sons Inc: Hoboken, 2008. Pollock, C. J.; Farrar, J. F. Source-sink Relations: The Role of Sucrose. In Advances in Photosynthesis. Photosynthesis and the Environment, Vol. 5, Baker, N. R., Ed.; Kluwer Academic: Dordrecht, 1996; pp 261–279. Foyer, C. H.; Noctor, G. Redox Regulation in Photosynthetic Organisms: Signaling, Acclimation, and Practical Implications. Antioxidants Redox Signal. 2009, 11, 861–905. London: Mary Ann Liebert Inc. Ort, D. R.; Yocum, C. F. Electron Transfer and Energy Transduction in Photosynthesis: An Overview. In Advances in Photosynthesis. Oxygenic Photosynthesis: The Light Reactions, Vol. 4, Ort, D. R., Yocum, C. F., Eds.; Kluwer Academic Publishers: Dordrecht, 1996; pp 1–9. Bacon, K. E. The Stable Primary Electron Acceptor QA and the Secondary Electron Acceptor QB. In Advances in Photosynthesis. Photobiochemistry and Photobiophysics, Vol. 10; Kluwer Academic Publishers: Dordrecht, 2001; pp 290–304. Bacon, K. E. Photosystem I – Introduction. In Advances in Photosynthesis. Photobiochemistry and Photobiophysics, Vol. 10; Kluwer Academic Publishers: Dordrecht, 2001; pp 419–430. Bacon, K. E. Proton Translocation and ATP Synthesis. In Advances in Photosynthesis. Photobiochemistry and Photobiophysics, Vol. 10; Kluwer Academic Publishers: Dordrecht, 2001; pp 666–737. Merchant, S.; Sawaya, M. R. The Light Reactions: A Guide to Recent Acquisitions for the Picture Gallery. Plant Cell 2005, 17, 648–663. Nelson, N.; Ben-Shem, A. The Complex Architecture of Oxygenic Photosynthesis. Nat. Rev. Mol. Cell Biol. 2004, 5, 971–982. Haehnel, W. Photosynthetic Electron Transport in Higher Plants. Annu. Rev. Plant Physiol. 1984, 35, 659–693. Amunts, A.; Drory, O.; Nelson, N. The Structure of a Plant Photosystem I Supercomplex at 3.4A Resolution. Nature 2007, 447, 58–63. Ferreira, K. N.; Iverson, T. M.; Maghlaoui, K.; et al. Architecture of the Photosynthetic Oxygen-evolving Center. Science 2004, 303, 1831–1838. Kurisu, G.; Zhang, H.; Smith, J. L.; Cramer, W. A. Structure of the Cytochrome B6f Complex of Oxygenic Photosynthesis: Tuning the Cavity. Science 2003, 302, 1009–1014. Mittler, R. Abiotic Stress, the Field Environment and Stress Combination. Trends Plant Sci. 2006, 11, 15–19. Wilson, K. E.; Ivanov, A. G.; Öquist, G.; et al. Energy Balance, Organellar Redox Status and Acclimation to Environmental Stress. Can. J. Bot. 2006, 84, 1355–1370. Ensminger, I.; Busch, F.; Hüner, N. P. A. Photostasis and Cold Acclimation: Sensing Low Temperature through Photosynthesis. Physiol. Plantarum 2006, 126, 28–44. Murchie, E. H.; Pinto, M.; Horton, P. Agriculture and the New Challenges for Photosynthesis Research. New Phytol. 2009, 181, 532–552. Hüner, N. P. A.; Öquist, G.; Melis, A. Photostasis in Plants, Green Algae and Cyanobacteria: The Role of Light Harvesting Antenna Complexes. In Advances in Photosynthesis and Respiration; Green, B. R., Parson, W. W., Eds.; Light Harvesting Antennas in Photosynthesis, Vol. 13; Kluwer Academic Publishers: Dordrecht, 2003; pp 401–421. Falkowski, P. G.; Chen, Y.-B. Photoacclimation of Light Harvesting Systems in Eukaryotic Algae. In Advances in Photosynthesis and Respiration; Green, B. R., Parson, W. W., Eds.; Light Harvesting Antennas in Photosynthesis, Vol. 13; Kluwer Academic Publishers: Dordrecht, 2003; pp 423–447. Horton, P.; Johnson, M. P.; Perez-Bueno, M. L.; et al. Photosynthetic Acclimation: Does the Dynamic Structure and Macro-organisation of Photosystem II in Higher Plant Grana Membranes Regulate Light Harvesting States? FEBS J. 2008, 275, 1069–1079. Baker, N. R. Chlorophyll Fluorescence: A Probe of Photosynthesis In Vivo. Annu. Rev. Plant Biol. 2008, 59, 89–113. Raven, J. A.; Handley, L. L.; Andrews, M. Optimizing Carbon-nitrogen Budgets: Perspectives for Crop Improvement. In Advances in Photosynthesis and Respiration; Foyer, C. H., Noctor, G., Eds.; Photosynthetic Nitrogen Assimilation and Associated Carbon Respiratory Metabolism, Vol. 12; Kluwer Academic Publishers: Dordrecht, 2002; pp 265–274. Escoubas, J.-M.; Lomas, M.; LaRoche, J.; Falkowski, P. G. Light Intensity Regulates Cab Gene Transcription via the Redox State of the Plastoquinone Pool in the Green Alga, Dunaliella tertiolecta. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 10237–10241. Hüner, N. P. A.; Öquist, G.; Sarhan, F. Energy Balance and Acclimation to Light and Cold. Trends Plant Sci. 1998, 3, 224–230. Miskiewicz, E.; Ivanov, A. G.; Williams, J. P.; et al. Photosynthetic Acclimation of the Filamentous Cyanobacterium, Plectonema boryanum UTEX 485, to Temperature and Light. Plant Cell Physiol. 2000, 41, 767–775. Dietz, K.-J. Redox Signal Integration: From Stimulus to Networks and Genes. Physiol. Plantarum 2008, 133, 459–468. Fernandez, A. P.; Strand, A. Retrograde Signaling and Plant Stress: Plastid Signals Initiate Cellular Stress Responses. Curr. Opin. Plant Biol. 2008, 11, 509–513. Apel, K.; Hirt, H. Reactive Oxygen Species: Metabolism, Oxidative Stress, and Signal Transduction. Annu. Rev. Plant Biol. 2004, 55, 373–399. Woodson, J. D.; Chory, J. Coordination of Gene Expression between Organellar and Nuclear Genomes. Nat. Rev. Genet. 2008, 9, 383–395. Bonardi, V.; Pesaresi, P.; Becker, T.; et al. Photosystem II Core Phosphorylation and Photosynthetic Acclimation Require Two Different Protein Kinases. Nature 2005, 437, 1179–1182. Pesaresi, P.; Hertle, A.; Pribil, M.; et al. Arabidopsis STN7 Kinase Provides a Link between Short- and Long-term Photosynthetic Acclimation. Plant Cell 2009, 21, 2402–2423. Melis, A. Photosystem-II Damage and Repair Cycle in Chloroplasts: What Modulates the Rate of Photodamage in Vivo? Trends Plant Sci. 1999, 4, 130–135. Koussevitzky, S.; Nott, A.; Mockler, T. C.; et al. Signals from Chloroplasts Converge to Regulate Nuclear Gene Expression. Science 2007, 316, 715–719. Demmig-Adams, B.; Adams, W. W.; Ebbert, V.; Logan, B. A. Eco-physiology of the Xanthophyll Cycle. In Advances in Photosynthesis; Frank, H. A., Young, A. J., Britton, G., Cogdell, R. J., Eds.; The Photochemistry of Carotenoids, Vol. 8; Kluwer Academic Publishers: Dordrecht, 1999; pp 245–269. Niyogi, K. K.; Li, X.-P.; Rosenberg, V.; Jung, H.-S. Is PsbS the Site of Non-photochemical Quenching in Photosynthesis? J. Exp. Bot. 2005, 56, 375–382. Kulheim, C.; Agren, J.; Jansson, S. Rapid Regulation of Light Harvesting and Plant Fitness in the Field. Science 2002, 297, 91–93. Machalek, K. M.; Davison, I. R.; Falkowski, P. G. Thermal acclimation and Photoacclimation of Photosynthesis in the Brown Alga Laminaria saccharina. Plant Cell Environ. 1996, 19, 1005–1016. Öquist, G.; Hüner, N. P. A. Photosynthesis of Overwintering Evergreen Plants. Annu. Rev. Plant Biol. 2003, 54, 329–355. Stitt, M.; Hurry, V. M. A Plant for All Seasons: Alterations in Photosynthetic Carbon Metabolism during Cold Acclimation in Arabidopsis. Curr. Opin. Plant Biol. 2002, 5, 199–206. Zhu, X.-G.; Long, S. P.; Ort, D. R. Improving Photosynthetic Efficiency for Greater Yield. Annu. Rev. Plant Biol. 2010, 61, 235–261.

1.22

Protein Expression in Insect Cells

RB Hitchman, Oxford Expression Technologies Ltd, Oxford, United Kingdom RD Possee, NERC Centre for Hydrology & Ecology (CEH) Oxford, Oxford, United Kingdom LA King, Oxford Brookes University, Oxford, United Kingdom © 2011 Elsevier B.V. All rights reserved. This is a reprint of R.B. Hitchman, R.D. Possee, L.A. King, 1.24 - Protein Expression in Insect Cells, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 323-340.

1.22.1 Historical Background and General Introduction 1.22.2 Baculovirus Biology 1.22.2.1 Genomics and Phylogeny 1.22.2.2 Host Infection and Virus Trafficking to the Nucleus 1.22.2.3 Gene Expression and Virus Egress 1.22.2.3.1 Early Gene Expression 1.22.2.3.2 Late Gene Expression 1.22.2.3.3 Very Late Gene Expression 1.22.2.4 Host Death and Liquifaction 1.22.3 The Origins of the BEVS 1.22.3.1 Modification of Circular Virus DNA 1.22.3.2 Linear Virus DNA and Recombinant Virus Production 1.22.3.3 Replication Defective Triple-Cut DNA and lacZ 1.22.3.4 BacPAK6 1.22.3.4.1 BacPAK6 Transfer Plasmids 1.22.3.5 Bac-N-Blue 1.22.3.5.1 Bac-N-Blue Transfer Plasmids 1.22.4 Baculovirus Recombination in Bacteria: The Development of Bacmids 1.22.4.1 Bac-to-Bac 1.22.4.1.1 pFastBac Transfer Plasmids 1.22.4.2 MultiBac – Protein Production From Multiple Genes 1.22.5 Hybrid Systems: Bacmid Recombination in Insect Cells 1.22.6 Baculovirus Recombination In Vitro 1.22.6.1 BaculoDirect 1.22.6.1.1 BaculoDirect Transfer Plasmids 1.22.7 Nonlytic Systems for Protein Expression in Insect Cells 1.22.7.1 InsectDirect 1.22.7.2 Drosophila melanogaster S2 Cells 1.22.7.3 Insect Cell-Free Expression 1.22.8 Insect Cells 1.22.8.1 Spodoptera frugiperda Cell Lines 1.22.8.2 Trichoplusia ni Cell Lines 1.22.8.3 Modified Insect Cell Lines for Improved Recombinant Protein Expression 1.22.8.3.1 Vankyrin-Enhanced Cell Lines 1.22.8.3.2 Mimic Cells 1.22.9 Insect Cell Culture 1.22.9.1 Insect Cell Culture Scale-Up 1.22.10 Removing Bottlenecks in the BEVS 1.22.10.1 Cloning Genes Into Baculovirus Transfer Plasmids 1.22.10.1.1 Ligation-Independent Cloning 1.22.10.1.2 Creator 1.22.10.1.3 StarGate 1.22.10.2 Genetic Modification of the Baculovirus Genome 1.22.10.3 Baculovirus Titration: Ways to Avoid a Plaque Assay 1.22.10.4 High-Throughput Baculovirus Expression 1.22.11 Concluding Summary References Relevant Websites

314 315 315 316 316 316 316 316 318 318 318 318 319 319 319 320 320 321 321 321 321 322 322 322 323 323 323 323 323 324 324 324 324 324 324 325 325 325 325 326 326 326 326 328 328 329 329 330

Comprehensive Biotechnology, 3rd edition, Volume 1

313

https://doi.org/10.1016/B978-0-444-64046-8.00047-1

314

Protein Expression in Insect Cells

Glossary BAC (bacterial artificial chromosome) A cloning vector constructed from bacterial fertility (F) factors that accept large inserts (>100 kb). Bacmid Baculovirus genomes that contain a bacterial origin of replication so that they can replicate in bacteria as a plasmid. Codon A set of three nucleotides in mRNA, functioning as a unit of genetic coding by specifying a particular amino acid during the synthesis of polypeptides in a cell. A codon specifies a transfer RNA carrying a specific amino acid, which is incorporated into a polypeptide chain during protein synthesis. The specificity for translating genetic information from DNA into messenger RNA (mRNA), then to protein, is provided by codon–anticodon pairing. Codon optimization An experimental strategy in which codons within a cloned gene – ones not generally used by the host cell translation system – are changed by in vitro mutagenesis to the preferred codons, without changing the amino acids of the synthesized protein. Expression system Combination of host and vector, which provides a genetic context for making a cloned gene produce a recombinant protein in the host cell. Expression vector A cloning vector that has been constructed in such a way that, after insertion of a DNA molecule, its coding sequence is properly transcribed and the RNA is translated. The cloned gene is put under the control of a promoter sequence for the initiation of transcription and often also has a transcription termination sequence at its end. Gene expression The mechanism whereby the genetic directions in any particular cell are decoded and processed into the final functioning product, usually a protein. Transfer plasmid An extrachromosomal, self-replicating, circular segment of DNA that can be propagated in bacteria. They contain a baculovirus promoter and recombination sequences (as well as other sequences such as signal peptides and tags) to facilitate transfer of a foreign gene into a baculovirus or bacmid genome, for expression in insect cells. Protein A polypeptide consisting of amino acids. Each polypeptide consists of a chain of amino acids linked together by covalent (peptide) bonds. They are naturally occurring complex organic substances (egg albumen and meat) composed essentially of carbon, hydrogen, oxygen, and nitrogen, plus sulfur or phosphorus, which are so associated as to form submicroscopic chains, spirals, or plates and to which are attached other atoms and groups of atoms in a variety of ways. In their biologically active states, proteins function as catalysts in metabolism and, to some extent, as structural elements of cells and tissues. Promoter A nucleotide sequence of DNA to which RNA polymerase binds and initiates transcription. It usually lies upstream of (50 to) a coding sequence. A promoter sequence aligns the RNA polymerase so that transcription will initiate at a specific site. Proteolysis Enzymatic degradation of a protein. Recombination Formation of a new association of genes or DNA sequences from different parental origins, by crossing over, which occurs during meiosis I. It involves breakage in the same position of each of a pair of nonsister chromatids from homologous chromosomes, followed by joining of nonsister fragments, resulting in a reciprocal exchange of DNA between nonsister chromatids within a homologous pair of chromosomes. Recombinant baculovirus A hybrid virus produced by inserting pieces of foreign coding DNA from different organisms to produce recombinant protein. Recombinant protein A protein whose amino acid sequence is encoded by a cloned gene. Restriction nuclease A bacterial enzyme that cuts DNA at a specific restriction site, such as a nucleotide sequence in DNA that is recognized by a type II restriction endonuclease and makes a double-stranded cut within it.

1.22.1

Historical Background and General Introduction

Baculoviruses likely originated between 400 and 450 million years ago and are now ubiquitous in the environment. They produce large occlusion or polyhedral bodies containing rod-shaped virus particles. These occlusions have a refractive nature, which meant that they were visible using early optical microscopes to study extracts from infected insects. They were positively associated with the wilt or polyhedrosis disease of silkworms as described by both Cornalia and Maestri in 1856. Such diseases had been noted much earlier in ancient Chinese literature and also in 16th-century Western texts. In the 1920s, it was suggested that the polyhedra may contain infectious virus, a fact confirmed by Bergold in 1947. This soon led to baculoviruses being tested and proven effective for biocontrol of insect pest populations in agriculture. With the development of insect cell cultures from 1950 onward, a number of continuously cultured insect cell lines became available. These provided systems for propagating baculoviruses outside of the original host species. This approach was advanced considerably in the 1970s after the isolation of what has become the prototype baculovirus, Autographa californica multinucleopolyhedrovirus (AcMNPV). This virus replicates very efficiently in several insect cell lines and has been a model system of study for 40 years.1 Together with several other baculovirus–host combinations, there is now considerable information available on virus ultrastructure, particle assembly, and occlusion body (OB) formation. The virus genome has proved to be a double-stranded circular molecule with both relaxed and supercoiled DNA. In the 1970s, the AcMNPV genome was extensively mapped using restriction enzymes, which facilitated more detailed studies of baculovirus genetic diversity. These maps also paved

Protein Expression in Insect Cells

315

the way for the localization and sequencing of virus genes such as polyhedrin, which encodes the major OB protein. Subsequent analysis of other baculovirus polyhedrin genes provided the means to establish phylogenies of different isolates. The ability to genetically modify AcMNPV also meant that virus gene deletions could be done, which determined that polyhedrin was not essential for virus replication, only for the formation of the OBs. Studies of gene transcription and protein synthesis revealed that polyhedrin was expressed to very high levels, which made it an obvious first choice as the basis for a recombinant protein expression system. Its gene promoter is very active and results in the production of high levels of messenger RNA (mRNA) and polyhedrin protein thereafter. In 1983, Smith and Summers published the first report describing the use of AcMNPV for foreign gene expression2 and the baculovirus expression vector system (BEVS) was established. A US patent on this process was granted in 19883 and since this time a large number of BEVS-associated patents have been filed.4 The complete AcMNPV genome was published in 1994.5 Research on baculoviruses has largely been driven by their use in practical applications, such as for insect pest biocontrol and more recently in the BEVS. Both of these applications have been made possible through two unusual features in baculovirus biology. First, as shown in Figure 1, baculoviruses produce two structural forms of enveloped viruses (budded virus (BV) and occlusion-derived virus (ODV)) and an occluded form (polyhedra or OBs) that packages ODV (but not BV). In cell culture, only the BV phenotype is required for the virus replication cycle. In insects, BV can also initiate and complete an infection cycle providing it is introduced directly into the hemocoel and, thus, bypasses the midgut, the normal route of virus entry. Second, baculoviruses produce large quantities of protein during a very late phase of gene expression, which is unique in insect viruses. This stage in virus replication can be exploited for recombinant protein production, without affecting synthesis of infectious virus, which occurs in a preceding phase. By deleting the very late polyhedrin or p10 gene coding regions and replacing them with heterologous sequences, it is possible to produce recombinant proteins in virus-infected cells. Therefore, the evolution of the BEVS has paralleled basic research into baculovirus molecular and structural biology and as such, an overview of baculovirology is provided, to give the reader an insight into both the biology of these fascinating viruses and to illustrate some of the rationale behind their development as an expression system.

1.22.2

Baculovirus Biology

1.22.2.1

Genomics and Phylogeny

Baculovirus genomes consist of a covalently closed circle of double-stranded DNA, with examples ranging in size from 100–180 kilobase pairs (kbp) and a molecular weight of approximately 8  107 Da. The family Baculoviridae is divided into the Alphabaculovirus (genus Nucleopolyhedrovirus (NPV) (type species AcMNPV)) and the Betabaculovirus (genus Granulovirus (type species Plodia interpunctella GV)), although only NPVs have been extensively developed as expression vectors. NPV species can be further divided into two virion phenotypes: those occluded within a crystalline proteinaceous matrix comprising largely of a 29 kDa polyhedrin protein (OBs) and nonoccluded, budded virions (BV). OBs may contain either single- or multinucleocapsids (SNPVs and MNPVs, respectively), within a single viral envelope.

Baculovirus multicapsid nucleopolyhedrovirus Budded virus

Occluded virus

Occlusion body

Peplomers (gp64) DNA

Capsids

p74

Biological membrane Polyhedrin 50 nm

5 μm

Approximate scale

Approximate scale

Figure 1 Diagram of nucleopolyhedrovirus virions (originally drawn by Dr. D Lynn, USDA, Agricultural Research Service, US, and reproduced by F Murguia-Meca, Oxford Brookes University).

316

Protein Expression in Insect Cells

1.22.2.2

Host Infection and Virus Trafficking to the Nucleus

The prototype member of the family, AcMNPV, has a genome size of 134 kbp and is the virus most commonly used as an expression vector. The virus DNA is packaged into a rod-shaped nucleocapsid approximately 36  200–400 nm. Unlike other virus particles with icosahedral capsids, the length of the baculovirus capsids can adapt to the amount of DNA required to be packaged. The nucleocapsid is enclosed in a lipoprotein envelope to form virus particles, which may contain more than one nucleocapsid. Numerous virus particles are then occluded within the OBs. These OBs are visible under a light microscope and range in size from 1 to 15 mm in diameter and have an outer polysaccharide or polyhedron envelope that may give additional strength and protection. OBs are very resistant to degradation and are thought to serve as a survival mechanism for the virus in the environment, allowing it to persist when susceptible hosts are not available. In temperate regions, insects may only have one or two generations per year. Consequently, there are long periods when the host is not available. Figure 2 illustrates the baculovirus life cycle in more detail.

1.22.2.3

Gene Expression and Virus Egress

Baculovirus gene expression can be divided into four phases; immediate-early (IE, a), delayed-early (DE, b) – which are often considered together as the early phase – late (l), and very late (d). Following infection of an insect cell, baculovirus gene expression occurs in a temporally regulated cascade in a manner similar to herpes viruses.

1.22.2.3.1

Early Gene Expression

Early genes are transcribed by host RNA polymerase II and encode 19 late expression factors, such as a virus-encoded alpha amanitin-resistant RNA polymerase, that are required for late gene expression from about 6 h postinfection (h.p.i.). Unlike gene transcription in the three other temporal phases, transcription of the IE genes does not depend on production of other viral proteins because these genes are transcribed by host factors. Their products regulate the DE genes, and IE gene products have been found to transactivate the expression of DE genes after transfection of uninfected insect cells. Inhibition of cell protein synthesis has shown that after the inhibitors are removed some proteins are synthesized immediately by the cells, while others are produced after a delay. Prior to the late phase, however, virus DNA replication must be initiated and seems to be essential for subsequent high-level virus gene expression.

1.22.2.3.2

Late Gene Expression

Late gene expression occurs concurrently with the onset of DNA replication (approximately 6 h.p.i.) and includes the basic protein, P6.9, the capsid protein, P39, and the virus membrane glycoprotein, GP64. After virus particles enter the cell, the P6.9 protein may be phosphorylated causing the decondensation of the packaged viral DNA. The viral DNA then forms a chromatin-like structure within the virogenic stroma in the nucleus. PP31, a virally encoded phosphoprotein, binds to the viral DNA and virogenic stroma, providing a framework for the subsequent packaging of the viral DNA. Synthesis of the 6.9-kDa basic DNA-binding protein occurs between 10 and 24 h.p.i. and is believed to be involved in the condensation of viral DNA prior to packaging. Late and very late genes are transcribed by an a-amanitin-resistant RNA polymerase, and during the late phase, viral structural proteins are made. These package virus DNA into capsids, which start budding through the nuclear membrane at 12 h.p.i. The nuclear envelope acquired during this process is lost as the capsid transits the cytoplasm to the plasma membrane. Here, the capsid buds through the cell membranes, acquiring GP64 as spike-like structures at one end of a plasma membrane-derived envelope. The GP64 serves as an attachment protein for the virus to bind uninfected cells and allows dissemination of virus particles between cells. BV is 1000-fold more infectious in cell culture than OBs, which lack GP64. In contrast to the enveloped virus from the OBs, the BV does not fuse immediately with the plasma membrane but is taken up by receptor-mediated endocytosis before this vesicle fuses with an endosome. The acidic environment within the resulting structure causes fusion of the virus envelope with that of the endosome to release the capsid into the cytoplasm, where it travels to the nucleus and replicates. The BVs are then released into the hemolymph, and subsequent infection of different larval tissues occurs in a sequential manner.

1.22.2.3.3

Very Late Gene Expression

Within the original virus-infected cell, BV synthesis is replaced very late in infection (18–24 h.p.i.) by production of an enveloped virus form (ODV) that remains within the nucleus. Progeny viruses become occluded by polyhedrin within the nuclei of the infected insect cells. OBs, genetically identical with BV, obtain their lipid envelope de novo within the nucleus and lack GP64. Trilamellar membranes are synthesized de novo in the cell nucleus and envelope single (SNPV) or multiple (MNPV) nucleocapsids to form the virion. This membrane is antigenically distinct from that surrounding the BV and confers on the virus, the ability to infect cells of the midgut epithelium. Their envelope contains several ODV-specific proteins, including glycoproteins that appear to be essential for attachment and entry of the virus to the insect gut. From about 15 h.p.i., the virus-infected cell begins to produce two very late proteins, polyhedrin and P10, which are synthesized to very high levels. The polyhedrin protein occludes a number of ODV within a crystalline matrix to form the OB. Each cell nucleus may contain up to 100 OBs. The mature OB is surrounded by a protease-sensitive calyx or polyhedrin envelope (PE) containing a phosphoprotein (PP34). The role of P10 is unclear, but if it is deleted from the genome, then polyhedra lack the PE that normally surrounds them and appear more fragile.

Protein Expression in Insect Cells

317

A Midgut

Foregut

Hindgut

B

Ingested OBs

Peritrophic membrane Midgut epithelial layer

Midgut C

1

Occlusion-derived virus (ODV) released from polyhedra

Peritrophic membrane

2

Midgut cells

c

6

3 a b 5a

5b

4 GP64-enriched membrane Basal lamina

Hemolymph

Budded virus (BV) mediates systemic infection

Figure 2 Diagrammatic view of the baculovirus life cycle. (A) Infection of host begins at the larval stage of the insect life cycle where the host ingests the occlusion body. (B) Occlusion bodies dissolve in the alkaline midgut releasing the occlusion-derived virus (ODV). These penetrate the peritrophic membrane of the midgut releasing nucleocapsids. (C) (1) Nucleocapsids enter the midgut epithelial cells and migrate to the nucleus (2) where viral DNA is uncoated. (3) The first uncoated nucleocapsids to enter the midgut epithelia cells instigate viral DNA replication. (3a) Expression of GP64 and other essential genes occurs. (3b) At later time points, during infection nucleocapsids are packaged into a crystalline matrix forming occlusion bodies. (3c) The expression of GP64 leads to its accumulation at the plasma membrane. (4) Replicated virus genomes can be packaged into progeny nucleocapsids. (5) Progeny nucleocapsids form budded virus (BV) as they bud through the GP64-enriched basal lamina into the hemolymph promoting systemic infection (5a) or into adjacent cells. (5b) Alternatively, infecting nucleocapsids, which have not gone through a primary round of replication. (6) can then exit the cell through this GP64-enriched membrane into the hemolymph (F Murguia-Meca, Oxford Expression Technologies Ltd).

318 1.22.2.4

Protein Expression in Insect Cells Host Death and Liquifaction

In an overt baculovirus infection of insects, majority of the host’s cells are infected and it dies leaving a limp sac, filled with OBs. In the case of AcMNPV, the cadaver turns a creamy color and then ruptures or liquefies, reducing the larvae to what has been described as an amorphous puddle. This process appears to result from the action of virus-encoded chitinase (CHIA) and cathepsin (V-CATH) proteins that act in concert to reduce the structural integrity of the insect cuticle. The liquefaction process is important for the efficient release of OBs from the host and is assumed to increase the chances of the virus encountering a new host. A single liquefying larva may release as many as 109 OBs with each OB containing up to 100 virions. These virions are protected by the OB against desiccation and ultraviolet (UV) light while outside the host.6

1.22.3

The Origins of the BEVS

1.22.3.1

Modification of Circular Virus DNA

Manipulation of baculovirus DNA is complex because of its size, which makes direct insertion of foreign genes in the manner of bacterial plasmids difficult. Indeed, two of the most challenging bottlenecks in the BEVS are the processes of inserting the gene to be expressed into a suitable expression plasmid and then subsequently isolating recombinant viruses away from nonrecombinant parental viruses. Despite both polyhedrin and P10 proteins being synthesized to high levels in virus-infected cells, the polyhedrin locus has been most commonly used as the site for insertion of foreign genes (Figure 3). This was originally as a consequence of its easily visible phenotype of OBs in virus-infected cells, which is removed after insertion of a foreign-coding region to render the virus polyhedrin negative. Originally, the production of recombinant baculoviruses required the co-transfection of insect cells with circular AcMNPV DNA and a transfer plasmid, based on the polyhedrin gene region, which contained a foreign-coding sequence under the control of the polyhedrin gene promoter and flanked by DNA from the virus. Homologous recombination between virus DNA and the plasmid resulted in the replacement of the polyhedrin-coding region with that of the foreign gene. This process only yielded 0.1% recombinant viruses, which required careful separation from parental stock with the use of plaque purification, a technically challenging and time-consuming method. An advance to the system was achieved by linearizing the AcMNPV genome prior to co-transfection.

1.22.3.2

Linear Virus DNA and Recombinant Virus Production

The proportion of recombinant viruses produced was increased to 30% by using AcMNPV DNA modified to contain a unique restriction enzyme site in lieu of the polyhedrin gene-coding region, which permitted linearization of the virus genome prior to co-transfection with a transfer vector. The linear virus DNA was restored to an infectious circular form in the insect cell after recombination with the homologous sequences in the plasmid transfer vector. It is still unclear if virus gene expression is required for this process. As only circular baculovirus DNA can initiate an infection, it made isolation of recombinant viruses by plaque purification an easier procedure. Although linearization of the virus DNA was estimated to be nearly 100%, a relatively high background (70%) of parental viruses suggested the presence of nondigested DNA and some religation of digested DNA, within transfected cells.7 Clearly, there was room for improvement.

kDa 140 70 40 35

P35 Polh

25

15

1 2

3

4

5

6

Figure 3 Coomassie-stained SDS-PAGE analysis of baculovirus expression. Lane 1 shows a protein marker, lane 2 shows noninfected insect cells, lane 3 shows cells infected 24 h postinfection (h.p.i.) with wild-type baculovirus expressing polyhedrin (polh-positive), and lane 4 shows a recombinant baculovirus (polh-negative) expressing P35 under the control of the polh promoter 24 h.p.i. Lane 5 shows cells infected 48 h.p.i. with wild-type baculovirus expressing polyhedrin (polh-positive) and lane 6 shows a recombinant baculovirus (polh-negative) expressing P35 under the control of the polh promoter 48 h.p.i. (A. Chambers, Oxford Brookes University).

Protein Expression in Insect Cells

ORF 603

lacZ

319

ORF1629

A Bsu361

Bsu361

ORF 603

10  106 cells/ml (for Sf9) before viability starts to decline. Fetal bovine serum (FBS) is required for the maintenance of certain cell lines, but many have now been adapted to serumfree conditions. FBS replaced insect hemolymph as a growth supplement, as it was much more readily available, and although it makes the final medium less defined, it does offer protection to the cells from shear stress and proteases that can degrade recombinant proteins. However, even minor traces of serum components are undesirable if the purified protein is to be used to analyze its function or tested in clinical trials and because of fears of contamination with agents such as bovine spongiform encephalopies. Most cell lines are now available in serum-free medium that allows higher cell densities (>107 cells/ml) to be attained in suspension cultures, hence improving the productivity of the baculovirus expression system. Infected insect cells also have a high demand for oxygen, and the bubbling of gas through the cultures can result in foaming and shearing of the insect cells. However, serum-free media contain surfactants (Pluronic F-68) and other components to reduce this problem. The media is supplemented with yeast extract and L-glutamine for improved growth. Growth limitation is usually due to nutrient (glucose and glutamine) and amino acid (i.e., tyrosine and methionine) depletion and the corresponding production of significant amounts of toxic byproducts (i.e., ammonia, alanine, and lactate). One of the few problems associated with the use of serum-free medium is that Sf9 cells grown as a monolayer are susceptible to shear damage when scraped from a plastic flask. Instead, cells must be removed from the support by vigorous tapping. This is most successful when the cells are confluent.

1.22.9.1

Insect Cell Culture Scale-Up

The baculovirus insect cell culture system can be operated at a wide variety of scales for production of recombinant proteins from 3 l Erlenmeyer flasks to large (>10 l) agitated bioreactors. Most laboratories with facilities for bacterial or mammalian cell culture can usually adapt their facilities to grow insect cells. Stainless-steel and glass bioreactors have a variety of vessels, impellers, and control modules for pH, temperature, nutrients, and oxygenation to ensure optimal conditions for growing the cells. More recently, single-use disposable systems have become popular. These systems are not generally compatible with thermal sterilization methods and so are usually supplied presterilized by gamma irradiation. They have the capacity for single-use or integrated traditional controls for sensing critical growth parameters, as described above for stainless-steel/glass bioreactors. Such technologies offer considerable operational flexibility in adjusting to process changes and scale, remove the costs and time associated with cleaning, assembly, and sterilization, and reduce the risk of cross-contamination. They are also generally easier to use, that is, to install, to move when empty and for making aseptic connections. These factors all shorten the lead time and increase the turn-around time between process runs. However, on the downside, there may be issues with compatibility, leachable components, potential for puncture, pressure and temperature sensitivity, and disposal costs. Certain cell lines may also be difficult to grow in disposable vessels. One commonly used system for insect cells is the WAVE Bioreactor (GE Healthcare) where presterile Cellbags containing medium (0.1–500 l) are rocked on a platform providing mixing and oxygen transfer during growth. However, it is argued that agitated (stirred) bioreactors may provide better mixing and improved mass transfer than WAVE-type bioreactors. Such disposable bioreactors are now available at a range of scales from stainless steel outer support containers, which incorporate flexible >1000 l capacity disposable plastic containers to smaller benchtop stirred-tank bioreactors that are completely disposable, such as the 3 l MobiusÒ CellReady Bioreactor (Millipore).

1.22.10 Removing Bottlenecks in the BEVS 1.22.10.1 Cloning Genes Into Baculovirus Transfer Plasmids One of the major bottlenecks in the production of recombinant viruses is an efficient method for transferring the target gene into a suitable transfer plasmid for insertion into the baculovirus genome. Several companies such as GENEART offer gene synthesis and subsequent subcloning into transfer plasmids, which is particularly useful for difficult-to-clone genes. This approach also gives the option of codon optimization and allows the user to determine the cloning sites to be used, providing flexibility in the cloning design. However, the use of gene synthesis for routine cloning may prove costly, when compared to standard cloning methods. Many systems (e.g., Bac-to-BacÔ, BaculoDirectÔ, Bac-N-BlueÔ, and MultiBac) have system-specific transfer plasmids for inserting the target gene into the baculovirus genome. The Bac-to-Bac system benefits from the highly successful Gateway cloning process, which until recently placed the traditional BacPAK6-based systems at a disadvantage. The original transfer plasmids for use with BacPAK6 required the plasmid to be digested with unique restriction enzyme sequences within the MCS and for the foreign

326

Protein Expression in Insect Cells

gene to be inserted using T4 ligase, after polymerase chain reaction (PCR) amplification or subcloning from another plasmid with compatible restriction enzymes. These transfer plasmids are still in common use, but this process of cloning is cumbersome and not particularly amenable to high-throughput platforms. However, several alternative methods for inserting genes into traditional transfer plasmids for recombination into the baculovirus genome in insect cells have now been developed, and these will be discussed below.

1.22.10.1.1

Ligation-Independent Cloning

Ligation-independent cloning (LIC) is a method that utilizes short homology arms for the directional cloning of PCR products negating the need for restriction enzyme digestion or ligation reactions. The In-FusionÔ PCR Cloning System (BD Biosciences) is based on LIC and utilizes several transfer plasmids based on pBacPAK8 with C- and N-terminal 6 His tags. The foreign gene is amplified by PCR and its single-stranded ends are fused to the homologous ends of a linearized vector using a poxvirus enzyme. The 30 and 50 regions of homology are generated by adding 15 bp extensions to both PCR primers that precisely match the ends of the linearized vector. When the vector is combined with the insert, the poxvirus DNA polymerase 30 –50 exonuclease activity converts the double-stranded extensions into short single-stranded sequences and fuses these regions to the corresponding ends of the linearized vector. The noncovalently joined molecules are then repaired in E. coli to produce a seamless fusion. LIC versions of a range of transfer plasmids are now also available (i.e., pBAC-2cp Ek/LIC, Novagen), for rapid, directional cloning of PCR products.

1.22.10.1.2

Creator

CreatorÔ (Clontech Laboratories, Inc.) is a Cre-lox-mediated cloning system for transferring a target gene from a donor plasmid into one of two baculovirus acceptor plasmids, pLP-BacPAK9 and pLP-BacPAK9-6xHN. Donor plasmids contain two LoxP sites, which flank the 50 end of the MCS and the 50 end of the ORF for the chloramphenicol-resistance gene. Acceptor plasmids contain a single LoxP site, followed by a bacterial promoter, which drives expression of the chloramphenicol marker after Cre-lox-mediated recombination.

1.22.10.1.3

StarGate

Perhaps the most powerful of these systems in terms of ease of cloning and range of available combinations of plasmids and sequences (tags, signals, etc.) is the StarGateÔ (IBA) vector series (Figure 7). Initially, the target gene has combinatorial sites (four bases) added at both ends by PCR and is inserted into an entry vector by a simple one-tube reaction to produce a donor vector. The extremely short combinatorial sites ensure that the protein is not significantly modified by unwanted extra amino acids and its expression is not hampered by ribosomal frameshifting, which may, in other systems, employ longer recombination sequences (i.e., Gateway adds eight amino acids). The entry vector contributes the recognition sites and provides an operative linkage with the combinatorial sites for the next transfer step into an acceptor vector. This is again achieved by another one-tube reaction, mixing the donor vector with an acceptor vector. Recombinant clones can be identified through blue/white selection. There are at least 27 acceptor vectors, each providing a different genetic surrounding like host-specific promoters and different purification tags. Fusion vectors are also available for incorporating multiple genes into the same virus. The resulting destination vectors are then mixed with baculovirus DNA in a co-transfection reaction and added to insect cells. This system is analogous to the Gateway cloning system and offers a powerful alternative for those users who wish to take advantage of the modified baculovirus vectors available, which are based on homologous recombination in insect cells (i.e., flashBAC, BacMagic, and BaculoGold), which are incompatible with pFastBac-based plasmids.

1.22.10.2 Genetic Modification of the Baculovirus Genome Most of the commercially available parental baculovirus genomes used for foreign gene expression are wild-type genotypes only modified at the polyhedrin gene locus with removal of the coding sequence for polyhedrin. The discovery that baculoviruses encode chiA and v-cath, which result in liquefaction of the insect larval host, suggested that removal of these genes might improve recombinant protein production. Deletion of the v-cath gene from Bombyx mori NPV demonstrated that recombinant protein degradation in virus-infected insects was reduced. ChiA is expressed late in the infection cycle and has a signal peptide sequence that directs the protein into the secretory pathway where in cell culture it likely competes with recombinant proteins for the same cellular machinery. It also contains a C-terminal ER retention motif (KDEL), which results in its accumulation in the ER, forming a dense paracrystalline matrix and obstructing secretion of recombinant proteins. Deletion of both of these genes has been shown to improve both protein stability and yield. Another gene deletion shown to improve protein expression was p10. The function of P10 within the wild-type virus is currently unknown although phylogenetic studies show a high degree of conservation between baculovirus species. Deletion of p10 has been suggested to increase infected cell viability, resulting in an extended protein production window, although this may depend on host cell, the protein being expressed and any other deletions that may have occurred within the genome. The genes encoding P26 and P74 are also nonessential for baculovirus replication, and deletion vectors of all of these genes (chiA, v-cath, p10, p26, and p74) have been shown to greatly improve recombinant protein expression in cell culture (Figure 8).13 Besides deleting genes from the virus genome, addition of heterologous sequences to improve protein expression has also been reported. The most commonly described of these are chaperones, proteins that assist the noncovalent folding/unfolding and the assembly/disassembly of other protein structures. Chaperones are required to be produced with the newly synthesized protein in order to direct its folding, although an unmodified host cell often has a shortage, compared to the amount of foreign protein being synthesized. This can result in misfolding of the overexpressed proteins. To avoid

Protein Expression in Insect Cells

327

A

GOI

PCR GOI

Opened entry vector

StarCombinase™

One-tube reaction

B

GOI

Donor vector

StarCombinase™

Acceptor vector

One-tube reaction

C

GOI

Destination vector

Figure 7 Diagram showing StarGate cloning procedure. (A) The gene of interest (GOI) is polymerase chain reaction (PCR) amplified with combinatorial sites (4 bases) at both ends and inserted into an entry vector by a simple one-tube reaction. The opened entry vector contributes the recognition sites and provides an operative linkage with the combinatorial sites for the StarGate® gene transfer process. (B) After sequence confirmation, the resulting donor vector is the source for subcloning of the GOI by a second simple one-tube reaction into an acceptor vector. (C) The resulting destination vector is then ready for use (i.e., in a co-transfection reaction with baculovirus DNA). Reproduced with permission from IBA GmbH StarGate Instructional Manual, 2008 Version PR26-0020 by F Murguia-Meca, Oxford Brookes University).

this, insect cells can be co-infected with separate baculoviruses expressing chaperones and the gene of interest or by single bicistronic baculovirus vectors. One potential problem with this approach is that the chaperones may be found associated with the proteins of interest in pull-down assays, which may be a result of incomplete folding and can make downstream structural characterization more difficult.

328

Protein Expression in Insect Cells

100

Densitometry units (%)

90 80 70 60 50 40 30 20 10 0

AcMNPVΔchiA

AcMNPVΔchiA-p26-p10-p74

Figure 8 Densitometry analysis of comparative Western blot of a chiA deletion virus vector (AcMNPVDchiA) and a chiA, p26, p10, andp74 deletion virus vector (AcMNPVDchiA-p26-p10-p74), expressing a his-tagged glycoprotein. The additional gene deletions from the AcMNPV genome clearly have an effect on recombinant protein expression, boosting the average densitometry value of AcMNPVDchiA-p26-p10-p74 by approximately 20%. (Dr. RB Hitchman, Oxford Expression Technologies Ltd.).

1.22.10.3 Baculovirus Titration: Ways to Avoid a Plaque Assay It is a good practice to obtain an accurate estimate of your recombinant virus in order to maximize and ensure reproducibility of virus amplification and protein production. For example, amplification of virus using an excessively high multiplicity of infection (m.o.i.) may result in the generation of defective particles, and wasting valuable virus stocks. Conversely, infection of insect cells with too few virus particles may result in a nonsynchronous infection within the culture resulting in reduced yields of protein. Traditionally, baculoviruses have been titrated by plaque assay or endpoint dilution. Both methods are time consuming, taking 3–4 days and requiring a certain degree of technical skill in cell culture and virology. To avoid these problems and speed up the process of quantifying baculovirus stocks, a range of methods have been developed and commercialized such as immunological and ELISA-based assays (BacPAK Baculovirus Rapid Titer Kit, BaculoELISATiter Kit, Clontech, FastPlaxÔ, Novagen), high-performance capillary electrophoresis (deltaTITRE, deltaDOT Ltd), and quantitative PCR (baculoQUANT, OET Ltd) as well as many in-house technologies including the use of reporter genes such as b-galactosidase and GFP, flow cytometry, cell viability assays, magnetic cell-sorting technologies, and measurement of cell-diameter changes after virus infection. All of these are rapid and comparatively easy alternatives to the more traditional methods of assaying virus particles in cell culture. However, it is important to remember that unlike plaque and end-dilution assays, none of the methods mentioned will directly measure infectious virus particles and, therefore, are likely to be less accurate when assessing older virus stocks, which may have degraded or aggregated. Additionally, the use of different titration methods between users may result in considerable variability in the accuracy of the final titer, although using systems that incorporate automated data collection (i.e., quantitative PCR) will help reduce user-to-user variability, often one of the biggest sources of error.14

1.22.10.4 High-Throughput Baculovirus Expression HTP sequencing of eukaryotic, bacterial, and viral genomes has resulted in a wealth of database information for structure–function analysis. In response to this, a number of structural genomics projects have been initiated, such as the Structural Proteomics in Europe (SPINE) program. Many of these projects have developed and implemented parallel processes for HTP cloning, expression, and purification of recombinant proteins. Generally, these systems fall into two categories, either bespoke modular workcells that can be rapidly installed, configured, brought into operation, and reconfigured when the need arises or very large and expensive off-the-shelf turnkey machines such as Piccolo (The Automation Partnership), QPExpression (Genetix) and the Expression Factory (NextGen Sciences). One of the main advantages of using smaller workstations is flexibility, where several different instruments can be integrated into a work flow and a single instrument can also be used for different stages of the process, that is, using a CAS-1200 (Corbett Research) for making recombinant viruses in multiwell plates and then titrating them by quantitative PCR. Importantly, these robots are also small enough to be located within a class II hood and so can be maintained in a sterile environment relatively cheaply. Making recombinant baculoviruses in both 24- and 96-well plates using liquid handlers has been described for several bacmid-based systems in both bacteria and insect cells (Figure 9).15 The development of improved recombinant technologies

Protein Expression in Insect Cells

329

A

B A B C D E F G H Figure 9 Production of recombinant viruses in 96-well plate format. (A) Cells were seeded with Sf9 cells (3  104 cells per well in 100 ml Sf900II) using a liquid handler located in a Class II safety cabinet. flashBAC™ DNA was then mixed with transfer plasmid DNA containing lacZ and co-transfected onto four rows of cells (B, D, F, and H) with a transfection reagent, robotically. The intermediate rows (A, C, E, and G) were mock transfected with just the transfection reagent (minus DNA). (B) The cells were stained with X-gal 5 days postinfection and majority of wells transfected with DNA showed the production of b-galactosidase. Mock-transfected wells showed no blue coloration, suggesting there was no cross-contamination between wells. Plaque assays of the DNA co-transfections confirmed the production of recombinant virus (Dr. K Graumann, Oxford Brookes University, and Dr. RB Hitchman, Oxford Expression Technologies Ltd.).

such as those described in Section 1.22.10.1 has facilitated the HTP cloning of genes into baculovirus transfer plasmids, and the development of systems such as flashBAC has streamlined the process of inserting these genes into the baculovirus genome. One advantage of making the viruses in insect cells, without an intermediate bacterial stage (i.e., as required for Bac-to-Bac), is that there are less liquid-handling steps and so reduced risk of cross-contamination, particularly when using the same robotic platform. Automated transfections in 24-well plates yield reasonable virus titers that can then be used to infect either a fresh 24-well plate of cells or a deep-well block, for virus amplification or protein expression analysis.10 Insect cell growth and virus infection kinetics have been shown to be similar in 24 deep-well blocks to those of shake flasks when using the correct media volumes and agitation speeds in shakers such as the HiGroÒ (GeneMachines). These small-scale matrix experiments allow the user to rapidly optimize and screen multiple constructs (i.e., different tags, C and N termini, fusion partners, and signal sequences), protein production conditions (i.e., m.o.i., time of harvest, and cell lines) and identify any weak expressers, prior to scale-up. Plate-based protein purification technologies are now available that allow the rapid filtration and capture of 24–96 different proteins by affinity resins (GST, His, etc.). Magnetic bead systems are also available for such small-scale purifications. Larger parallel purification strategies can be carried out using HTP chromatography systems such as the AKTAexpress (GE Healthcare). Generally, laboratories that already have E. coli or mammalian HTP expression systems in place will be able to transfer the technology to a baculovirus-based HTP system relatively easily. However, although E. coli expression has been almost completely automated on a small scale, because of its complexity there are still several hurdles to overcome before the same can be achieved for the BEVS.

1.22.11 Concluding Summary In recent years, the BEVS has evolved to become one of the primary expression platforms for the production of recombinant proteins for structural biology, drug discovery, HTP screening, and biomedical research. It has also been demonstrated to be a viable candidate for large-scale flu vaccine production. This success has been largely driven by advances in molecular cloning technologies and the genetic engineering of improved virus vectors. The new era of BEVS is one of simplicity and speed, where foreign genes can be PCR amplified and inserted into expression cassettes relatively easily via recombination in bacteria. Recombinant viruses can then be made by simply mixing this cassette with the virus-vector DNA, as provided in a BEVS kit, and transfecting either bacteria or insect cells. These procedures can be done either manually or by robotic workstations. Depending on the system used, there is no requirement for plaque purification and the resulting recombinant viruses can be harvested directly for scale-up. Various kit-based virus titration methods are now widely available, which utilize equipment common to most molecular biology laboratories. The recombinant viruses can then be scaled up depending on requirements and used to produce protein, in shake flasks or bioreactors. Thus, the pipeline from gene cloning to protein production in insect cells can now be carried out with relative ease. This continued development toward improved expression levels, reduced gene to protein timelines, and user-friendliness will ensure that the BEVS will dominate the protein expression landscape for many years to come.

References 1. Summers, M. D. Milestones Leading to the Genetic Engineering of Baculoviruses as Expression Vector Systems and Viral Pesticides. Adv. Virus Res. 2006, 68, 3–73. 2. Smith, G. E.; Summers, M. D.; Fraser, M. J. Production of Human Beta Interferon in Insect Cells Infected with a Baculovirus Expression Vector. Mol. Cell Biol. 1983, 3 (12), 2156–2165. 3. Smith, G. E.; Summers, M. D. Method for Producing a Recombinant Baculovirus Expression Vector. US Patent 4745051, 1988. http://www.freepatentsonline.com/ 4745051.html.

330

Protein Expression in Insect Cells

4. Hitchman, R. B. Genetic Modification of a Baculovirus Vector for Increased Expression in Insect Cells. Cell Biol. toxicol. 2009, 26 (1), 57–68. Special Issue - Genetic Manipulation of Cells. http://www.springerlink.com/content/0742_2091/26/1/. 5. Ayres, M. D.; Howard, S. C.; Kuzio, J.; et al. The Complete DNA Sequence of Autographa Californica Nuclear Polyhedrosis Virus. Virology 1994, 202 (2), 586–605. 6. Fraser, J. M. J.; Vol. 8; Bentham Science: Bussum, 2007. 7. Kitts, P. A.; Ayres, M. D.; Possee, R. D. Linearization of Baculovirus DNA Enhances the Recovery of Recombinant Virus Expression Vectors. Nucleic Acids Res. 1990, 18 (19), 5667–5672. 8. Kitts, P. A.; Possee, R. D. A Method for Producing Recombinant Baculovirus Expression Vectors at High Frequency. Biotechniques 1993, 14 (15), 810–817. 9. Luckow, V. A.; et al. Efficient Generation of Infectious Recombinant Baculoviruses by Site-specific Transposon-mediated Insertion of Foreign Genes into a Baculovirus Genome Propagated in Escherichia coli. J. Virol. 1993, 67 (8), 4566–4579. 10. Possee, R. D.; Hitchman, R. B.; Richards, K. S.; et al. Generation of Baculovirus Vectors for the High-throughput Production of Proteins in Insect Cells. Biotechnol. Bioeng. 2008, 101 (6), 1115–1122. 11. Douris, V.; Swevers, L.; Labropoulou, V.; et al. Stably Transformed Insect Cell Lines: Tools for Expression of Secreted and Membrane-anchored Proteins and High-throughput Screening Platforms for Drug and Insecticide Discovery. Adv. Virus Res. 2006, 68, 113–156. 12. Shi, X.; Jarvis, D. L. Protein N-glycosylation in the Baculovirus-insect Cell System. Curr. Drug Targets 2007, 8 (10), 1116–1125. 13. Hitchman, R. B.; et al. Genetic Modification of a Baculovirus Vector for Increased Expression in Insect Cells. Cell Biol. Toxicol. 2009. 14. Hitchman, R. B.; et al. Quantitative Real-time PCR for Rapid and Accurate Titration of Recombinant Baculovirus Particles. Biotechnol. Bioeng. 2007, 96 (4), 810–814. 15. Hunt, I. From Gene to Protein: A Review of New and Enabling Technologies for Multi-parallel Protein Expression. Protein Expr. Purif. 2005, 40 (1), 1–22.

,

Relevant Websites http://www.baculovirus.com – Baculovirus Techniques. http://www.healthtech.com – Conferences: Healthtech.com; Cambridge Healthtech Institutes Annual ‘Baculovirus Technology‘ Conference. http://www.hpacultures.org.uk – Health Protection Agency; European Collection of Cell Cultures (ECACC). http://www.brookes.ac.uk – Oxford Brookes University; Insect Virology and Protein Expression. http://www.oetltd.com – Oxford Expression Technologies Ltd. http://www.virology.net – Virology Journal.

1.23

Stem Cells

SKW Oh and ABH Choo, Bioprocessing Technology Institute, Singapore © 2011 Elsevier B.V. All rights reserved. This is a reprint of S.K.W. Oh, A.B.H. Choo, 1.25 - Stem Cells, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 341-365.

1.23.1 1.23.2 1.23.2.1 1.23.2.2 1.23.2.3 1.23.2.4 1.23.2.5 1.23.2.5.1 1.23.2.5.2 1.23.2.6 1.23.3 1.23.3.1 1.23.3.2 1.23.3.3 1.23.3.4 1.23.3.4.1 1.23.3.4.2 1.23.3.4.3 1.23.3.5 1.23.4 1.23.4.1 1.23.4.2 1.23.4.3 1.23.4.4 1.23.4.5 1.23.4.5.1 1.23.4.5.2 1.23.4.5.3 1.23.4.6 1.23.5 1.23.5.1 1.23.5.2 1.23.5.3 1.23.5.4 1.23.5.5 1.23.5.5.1 1.23.5.5.2 1.23.5.5.3 1.23.5.5.4 1.23.5.5.5 1.23.5.6 1.23.6 1.23.6.1 1.23.6.2 1.23.6.2.1 1.23.6.3 1.23.6.3.1 1.23.6.4 1.23.6.4.1 1.23.6.4.2 1.23.6.5 1.23.6.5.1 1.23.6.5.2

Introduction Human Embryonic Stem Cells Initial Discovery Sources and Methods of Deriving hESCs Characteristics Differentiation Capabilities Clinical Trials Spinal Cord Injury Retinal Pigment Epithelial Cells Transplantation Conclusion Human-Induced Pluripotent Stem Cells Discovery of Reprogramming Characterization to Authenticate hiPSCs Reprograming Methods Study of Diseases Sources of Cells Disease Models hESC Versus hiPSC Conclusion Neural Stem Cells Initial Discovery Sources and Niche of NSCs Characteristics Differentiation Capabilities Clinical Trials Batten’s Disease Stroke Cancer Conclusion Mesenchymal Stem Cells Initial Discovery Sources and Niches of MSCs Characteristics Differentiation Capabilities Clinical Trials Immune-Modulatory Therapy Bone Regeneration Cartilage Regeneration Myocardium Regeneration Skeletal and Neurological Disorders Conclusion Hematopoietic Stem Cells Initial Discovery Sources Peripheral Circulating Blood Umbilical Cord Blood hESCs and hiPSCs Niches Osteoblastic Niche Vascular Niche Characteristics Self-Renewing HSC Population Common Lymphoid Progenitors

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00021-5

332 333 333 333 334 334 335 335 335 335 335 335 336 337 337 337 338 338 338 338 338 338 340 340 341 341 341 341 341 341 341 342 342 342 343 343 345 345 345 345 346 346 346 346 346 346 346 346 347 347 347 348 348

331

332

Stem Cells

1.23.6.5.3 1.23.6.6 1.23.6.6.1 1.23.6.6.2 1.23.6.7 1.23.6.7.1 1.23.6.7.2 1.23.6.7.3 1.23.6.8 References

Common Myeloid Progenitors Differentiation Capability Lymphoid Lineage Myeloid Lineage Clinical Applications HLA-Matching of HSC Sources BM, Peripheral Blood, and Cord Blood for HSCT Allogeneic Versus Autologous HSCT Conclusion

349 349 349 349 349 349 350 350 350 351

Glossary Ectoderm The outer layer of cells in a developing embryo, which comprises mainly of the neural and epithelial lineages. Endoderm The inner layer of cells that comprises the endocrine organs such as liver, pancreas, kidney, and lung. Mesoderm The middle layer of cells in the developing embryo that comprises muscle, bone, cartilage, heart, and blood lineages. Multipotent Cells capable of differentiation to a more limited range of cell types (e.g., either the ectoderm, mesoderm, or endoderm layers). Niche The anatomical location in adult tissues where stem cells reside. Pluripotent Cells capable of differentiation into all three germ layers and all tissue types of the body. Stem cells Cells capable of both unlimited self-renewal and differentiation to other cell types of the body.

1.23.1

Introduction

Stem cells have garnered the imagination of the scientific community in the late twentieth and early twenty-first centuries because of their perceived potential to self-renew and differentiate to specialized cell types of the body. With this promise comes the prospect that stem cells may become the vehicles for regenerative medicine to restore functions, lost through disease or aging, which cannot otherwise be circumvented by traditional medical therapies such as small molecules and biologics. We chose to review five types of stem cells: human embryonic stem cells (hESCs), induced pluripotent stem cells (iPSCs), neural stem cells (NSCs), mesenchymal stem cells (MSCs), and hematopoietic stem cells (HSCs). Figure 1 shows the relationship between each of these stem cells and their differentiation lineages. hESCs, derived from the inner cell mass (ICM) of the blastocyst (developing embryo), have the greatest regenerative capacity to form the three germ layers (i.e., endoderm, mesoderm, and ectoderm). From the mesoderm, two stem cell types are derived, MSCs and HSCs, which can form other progeny, while from the ectoderm, NSCs can be derived. Differentiated cells from these three germ layers, in turn, have been reprogrammed back to iPSCs that have similar pluripotent capability as hESCs. Each of these stem cells have significant regenerative capacity besides showing great promise in the study and treatment of diseases as shown by the number of early clinical trials being conducted with them. The exception is HSCs that have had a special status in blood transplantation for over 50 years and continues to produce fascinating science.

Box 1 When human-induced pluripotent stem cell (hiPSC) research was gaining momentum rapidly, Vierbuchen et al.62 surprised the stem cell community by demonstrating the possibility of converting mouse fibroblast into functional neurons without having to go through the hiPSC state. Through this research, they have added another dimension to the art of reprogramming somatic cells. They have shown that expressing three neural-lineage-specific transcription factors (Ascl1, Brn2, and Myt1l) were sufficient to induce differentiation of fibroblast into functional neurons. These neurons were shown to express neuron-specific proteins, generate action potential, and form functional synapses.

hESCs and human iPSCs (hiPSCs) are the most amazing discoveries that occurred within 10 years of each other. They share many similar characteristics and are able to differentiate into cells comprising the three germ layers. Astonishingly, differentiated cells such as fibroblasts, cord blood, neural cells, and hepatocytes, representatives from all three germ layers, have been reprogrammed into hiPSCs (Figure 1). The ability to create hiPSCs especially from diseased cell lines now enables the study of disease in vitro, which was not previously possible. Most recently, another amazing feat has shown that fibroblasts can be reprogrammed directly to motor neurons without forming iPSCs. NSCs are derived from the ectodermal layer and participate in the generation of many nervous system-related cells (Figure 1). They too are showing promise for the treatment of neurological

Stem Cells

Blastocyst

333

Mesenchymal stem cell Mesoderm

hESC

Hematopoietic stem cell

Ectoderm hiPSC Neural stem cell Endoderm

Hepatocyte

Neural stem cell

Cord blood cells

Fibroblast

Motor neuron Figure 1 The relationship between human embryonic stem cell (hESC) and human-induced pluripotent stem cell (hiPSC) and their differentiation into three classical germ layers. Subsequently, the differentiated cells from these different lineages can be reprogrammed back into hiPSC. In a recent study, fibroblasts have been shown to be able to be directly reprogrammed into motor neurons (refer to Box 1).

diseases. MSCs and HSCs are derived from the mesodermal layer (Figure 1). MSCs have the capability of differentiating into bone and cartilage as well as have immunomodulatory properties. In recent years, it has been studied extensively for its clinical potential. HSCs were the earliest stem cells to be isolated and have been shown to be able to differentiate to all the blood lineages. It is also widely used in HSC transplantations.

1.23.2

Human Embryonic Stem Cells

1.23.2.1

Initial Discovery

Embryonic stem cells from mice were derived in 1981,36 while the first hESCs were derived by James Thomson and his co-workers only in 1998. They used frozen or fresh blastocysts produced by in vitro fertilization (IVF), remaining from infertility treatment of couples. The blastocysts were cultured from the initial cleavage stage embryos and 14 ICMs were isolated from the blastocysts. From these, five different cell lines were derived.59

1.23.2.2

Sources and Methods of Deriving hESCs

Besides Thomson’s method of obtaining hESCs, as mentioned above, there are three other sources to obtain hESCs. First, hESCs can be derived from dead embryos. Landry and Zucker proposed that these embryos that are produced by IVF have undergone “irreversible cessation of cell division in the embryo observed in vitro” and are no longer suitable for the purpose of reproduction. However, these embryos still retain some normal blastomeres for harvesting and the derivation of stem cells.36,49 Another possible source of hESCs is the blastomeres obtained through the biopsy of young live embryosE.49 Recently, it has been demonstrated that hESCs can be generated from single blastomeres and this procedure is not harmful to the embryo.11,80

334

Stem Cells

The technique used in the biopsy of embryos is very similar to that of preimplantation genetic diagnosis, which is usually done prior to uterine transfer concurrently with assisted reproductive technologies to test for chromosomal and genetic abnormalities in IVF embryos.11,49,80 In this method, the extracted blastomeres are cultured using a tailored approach that aimed to recreate the ICM niche. This approach improved the efficiency of the rate of hESC derivation as compared to the derivation from whole embryos.11 The last procedure to produce embryos for the derivation of hESCs is called parthenogenesis. Through this procedure, the oocyte is treated biochemically to trick it into believing that it has been fertilized.49 One method developed by Rogers and his team was the microinjection of complementary ribonucleic acid for human phospholipase Cz, a sperm-specific protein, into human oocytes that have aged and failed to fertilize during IVF or through intracytoplasmic sperm injection. The study showed that cleavage division was observed and blastocysts were formed in some of the embryos.86 Although the treated oocyte underwent cell division, it is said to lack the ability to develop into a human being.49 After fertilization of the embryo from the above methods, the blastocyst is formed at roughly 4–5 days. The blastocyst has an outer shell called the trophectoderm and an aggregation of polarized inner cells called the ICM. To obtain hESCs, immunosurgery is done to remove the trophectoderm, and disaggregation of the ICM is done. The disaggregated ICMs are then plated on feeder cells. The resulting cell colonies are then isolated through mechanical means and replated until the colonies have achieved homogeneity. Each of the colonies is then screened for the presence of stem cells expressing the appropriate markers and capable of symmetrical division for unlimited self-renewal. Once the cell lines have been established, they can persist stably for many years.36

1.23.2.3

Characteristics

A study was performed by the International Stem Cell Initiative (ISCI), whereby 59 hESC lines from various laboratories worldwide were characterized. The study showed that each hESC line has similar expression patterns for several hESC markers. The reader is referred to Table 1 for the list of markers taken from the ISCI study.2 Common markers expressed by the hESC lines were TRA-1-60, TRA-1-81, SSEA-3, SSEA-4, NANOG, and OCT4. These markers represent the minimal criteria for the characterization of hESCs.77 However, not all hESC lines have the same expression pattern for each marker. For example, it has been shown that SSEA-4, TRA-1-60,74 OCT4,85,87 and NANOG87 were expressed with significant differences in several groups of hESC lines.

1.23.2.4

Differentiation Capabilities

Typically, in vitro differentiation of hESCs into the three germ layers first requires the formation of embryoid bodies (EBs), which are cellular aggregates in suspension.78 From there, the EBs are able to differentiate into several cell lineages such as definitive endoderm, mesoderm, ectoderm, primitive endoderm, and trophectoderm.75 Until now, germ cells, neurons, glia, endothelial cells, cardiomyocytes, keratinocytes, hepatocyte-like cells, hematopoietic precursors, osteogenic cells, insulin-producing cells, prostate tissue, adipocytes, and melanocytes have been differentiated from hESCs.72 An example of the formation of the ectoderm lineage is seen in neural differentiation. Large numbers of neural tube-like structures were formed from differentiating hESCs in the presence of fibroblast growth factor-2 (FGF-2). The neural precursors were then isolated from these structures. With the withdrawal of FGF-2, these precursors differentiated into neurons, astrocytes, and oligodendrocytes. When the precursors were transplanted into the brain of a neonatal mouse, they were incorporated into various parts of the brain where they differentiated into both astrocytes and neurons.88 Table 1

Summary of the different levels of characterizations of mouse and human embryonic stem cells and induced pluripotent stem cells Mouse

Characteristics

Embryonic stem cells

Induced pluripotent stem cells

Gene/protein markers

SSEA-1, Oct4, Nanog, Sox2, ERas, Fgf4, Cripto, Dax1, Zfp296, and alkaline phosphatase57

Telomerase activity In vitro differentiation to three germ layers

High57 Blood vessels,123 Cardiovascular cells,116 retinal dopaminergic, and 109 serotonergic neurons, cells,103 and hepatocytes112 pancreatic islet114 Forms cells from all three germ layers57 þ125 þ115 125 þ115 þ

Teratoma formation Chimera contribution Germline transmission

Human Embryonic stem cells

Induced pluripotent stem cells

SSEA-3, SSEA-4, TRA-1-60, TRA-1-81, Nanog, alkaline phosphatase, Sox2, Oct4, GDF3, REX1, FGF4, ESG1, DPPA2, DPPA4, CD9, TDGF1, DNMT3B, GABRB3, GCTM2, GCT343, Thy1, and Class 1 HLA2,58 High59,58 Motor Neurons,121 insulinMotor neurons,95 insulinproducing cells,127 and producing cells,127 and 106 cardiomyocytes cardiomyocytes128 Forms cells from all three germ layers59,58  

Stringencies increase from gene/protein markers to germ-line transmission. Chimera and germ-line transmission are not for human models due to ethical issues.

Stem Cells

335

Cardiomyocyte differentiation toward the mesoderm lineage has also been established. Four to twenty days after plating EBs onto culture dishes coated with gelatin, 8.1% of the EBs contained areas that were contracting rhythmically. Cardiac-specific genes and transcription factors were expressed by the cells isolated from these areas. In addition, the cells displayed a functional phenotype of immature human cardiomyocytes.78 hESCs have also been differentiated into a definitive endoderm lineage in low serum and activin A, which resulted in cultures consisting of up to 80% definitive endoderm cells. More mature cells of the endodermal organs are produced from the differentiation of the definitive endoderm cells after transplantation under the kidney capsule.75 The specific endodermal cell aggregates are defined through the expression of various marker genes. For example, the co-expression of HNF6, FOXA2, and PDX1 defines the cell aggregates of the pancreatic endoderm.82

1.23.2.5

Clinical Trials

At present, there are only a few ongoing clinical trials in humans involving hESCs. To date, other human cell-grafting therapies have mainly made use of adult stem cells derived from matched or allogeneic donors.64

1.23.2.5.1

Spinal Cord Injury

The first human trial of hESCs is for the treatment of spinal cord injury, approved by the Food and Drug Administration (FDA). Eight to ten patients with severe spinal cord injuries will be involved in the trial by Geron, a company based in California. The phase I clinical trial will primarily test the safety of the treatment. In this trial, various growth factors are used to induce hESCs differentiation into oligodendrocyte precursors before they are injected into the injured spinal cord. Oligodendrocytes play a part in supporting neural cells. Previous preclinical trials done in adult rats showed that the transplanted hESC-derived oligodendrocyte progenitor cells enhanced remyelination and promoted the improvement of motor function.79 The aim of this clinical trial is to enhance the repair of the myelin insulation around the nerve cells and subsequently reestablish the nerve cells’ ability to transmit signals.79

1.23.2.5.2

Retinal Pigment Epithelial Cells Transplantation

Retinal pigment epithelial (RPE) cells are derivatives of the neuroectoderm, which is crucial for the survival of photoreceptors. In age-related macular degeneration (AMD), RPE cells degenerate and cannot be replaced. Animal studies have shown that degenerated RPE cells can be replaced by transplanting donor RPE cells, saving the host photoreceptors and attenuating the loss of visual function.81 hESC has recently been recognized as a candidate for the transplantation of RPE cells.81 In culture, these hESC-derived RPE cells showed gene expression profiles that more closely resemble that of primary human RPE compared to ARPE19, an immortalized RPE cell line. Three previous studies used different neural differentiating protocols on mouse embryonic stem cells (mESCs) and hESCs to form RPE cells prior to transplantation and have found no evidence of any tumor formation.73,83,84 The finding shows the clinical potential of hESC-derived RPE cells in treating AMD.76 Recently, FDA has approved the testing of hESC for the treatment of Stargardt’s macular dystrophy (SMD) by advanced cell technology (ACT). This rare disease destroys the retinal cells of approximately 30 000 people worldwide. Currently, ACT has finished its studies in animals and is preparing for a phase I trial to establish the safety and tolerability of the RPE cells after their transplantation into the subretina of the patients with SMD. The preclinical trial on rat models showed a 100% improvement of visual performance without any undesirable effects in the treated rats when compared with their untreated counterparts.

1.23.2.6

Conclusion

Since the first derivation of hESCs in 1998, there have been several other methods of derivation that avoids the use of live embryos. The differentiation capabilities of hESCs into the various types of cells in the body promise to revolutionize cell therapy. hESCs have been studied extensively in recent years to apply them for therapeutic uses worldwide. The recent approval for phase I clinical trials of hESC-derived cells for spinal cord injury and SMD patients to prove their safety in human therapy is a significant step forward.

1.23.3

Human-Induced Pluripotent Stem Cells

1.23.3.1

Discovery of Reprogramming

Although the field of hESC research has progressed, it has always been surrounded by technical difficulties and ethical issues. In the midst of these, a breakthrough discovery by Takahashi and Yamanaka gave stem cell research a new dimension. They managed to show that pluripotent stem cells could be induced from mouse embryonic or adult fibroblasts, which revolutionized stem cell field.57 In their study, different combinations of 24 candidate reprogramming genes that were thought to play a role in maintaining embryonic stem cell identity were transduced by retrovirus into mouse fibroblasts. They found that the collective combination of four factors Oct4, Sox2, Klf4, and c-Myc was sufficient to reprogram somatic cell to become, what is known today as, iPSC. Soon after, three groups, including the pioneers, reported that they had successfully induced pluripotency in human somatic cells. Two groups used the same four factors (Oct4, Sox2, Klf4, and c-Myc), while another group used Oct4, Sox2, Nanog, and

336

Stem Cells Table 2

Different methods of reprogramming somatic cells into hiPSC

Method

Advantages

Disadvantages

References

Retrovirus carrying four reprogramming factors Retrovirus carrying four reprogramming factors þ siRNA against p53 and Utf1 cDNA Piggyback transposon system using four reprogramming factors “Minicircle” vector carrying four reprogramming factors Arginine peptide-tagged four protein factors

• Has a reprogramming efficiency

• Although silenced in pluripotent

58

of 0.1%

• 100 times increase in

reprogramming efficiency as compared to four factors alone

• Precise excision of reprogramming factor possible • Free of viral origins • Free of integration • No genetic modification

cells, there is a risk of genomic integration and insertional mutagenesis • Inhibition of p53 (a potent tumor suppressor)

129

• Excision might be inefficient

65

• Has a low reprogramming

105

efficiency of 0.005% • Has a low reprogramming efficiency of 0.001%

107

Lin28 to reprogram human somatic cells into iPSCs.47,58,70 Oct4, Sox2, and Nanog are transcription factors that have been shown to play a part in maintaining pluripotency of early embryos and embryonic stem cells. They synergistically upregulate pluripotentspecific genes while repressing differentiation-specific genes. In one study, human dermal fibroblasts were transduced with retrovirus carrying Oct4, Sox2, Klf4, and c-Myc genes.58 The transduced cells formed hESC-like colonies and were morphologically indistinguishable from hESCs. Yamanaka also proved these cells expressed hESC-specific markers, and called them hiPSCs. These hiPSCs grew exponentially for at least 4 months and its population doubling time was similar to hESCs. In another independent study, human embryonic fibroblasts were reprogrammed using retroviral constructs into hiPSCs using the same four factors.47 Separately, hiPSCs were also made using a different combination of reprogramming factors – Oct4, Sox2, Nanog, and Lin28.70 In this study, it was shown that Oct4 and Sox2 were vital to the reprogramming mixture and removal of any one the factors prevents cell reprogramming, while Nanog and Lin28 improved the efficiency of reprogramming. Like in the two previous studies, the reprogrammed cells resembled hESCs in many ways. Their gene expression profile closely resembled hESCs and also they were able to differentiate into all three germ layers in vitro and gave rise to teratomas upon injection into nude mice. Since the discovery of reprogramming, iPSC has generated a huge interest among the stem cell community. Many independent studies have confirmed the reproducibility of reprogramming capabilities in human cells (see Table 2). There are many studies done on characterizing hiPSCs as well as to improve the efficiency of reprogramming. In short, hiPSCs provide an alternative source of pluripotent stem cells that could be potentially used for disease modeling and clinical purposes.

1.23.3.2

Characterization to Authenticate hiPSCs

Characterization is a vital aspect of any research. Being a fast-growing arena, hiPSC research needs to have criteria and standards for characterization that would allow cross-lab comparison of data. Several criteria have been proposed to authenticate if a fully reprogrammed state of a somatic cell has been achieved, such as genetic and protein markers, telomerase activity, in vitro differentiation, teratoma formation, chimera contribution, and germline transmission. Table 1 summarizes the different levels of characterization of hESCs and hiPSCs. First and foremost, hiPSC colonies are identified by morphological criteria.117 If somatic cells have been reprogrammed to hiPSC state, cells would form compact colonies having high nucleus-to-cytoplasm ratios and prominent nucleoli, such as hESC colonies.58 Identified and selected colonies are put through an array of tests to further demonstrate its hESC-like properties. At the molecular level, gene and protein expression profiles of hiPSCs are screened to capture its resemblance to hESCs. Expression of key pluripotent factors, such as Oct4, Sox2, SSEA-3, SSEA-4, TRA-1-60, TRA-1-81, alkaline phosphatase, and Nanog, are normally profiled to show hESC-like properties in hiPSCs.58 Telomerase reverse transcriptase (hTERT) expression is another important property of hiPSCs. Telomerase100 is involved in adding DNA repeat sequences to the 30 telomere region of the chromosomes. Shortening of telomere region101 is linked with short replicative life span of diploid somatic cells. Hence, high expression levels of hTERT indicate high replicative life span of hESCs and hiPSCs. An authentic iPSC must be independent of exogenous genes expression used to reprogram these cells. To show that the iPSCs are indeed independent of exogenous gene expression, expression level of genes involved in retroviral silencing such as de novo methyltransferases110 can be analyzed to prove exogenous gene repression. Reverse transcription–polymerase chain reaction using primers specific for exogenous transcript are used to show exogenous gene silencing. Exogenous expression of reprogramming factors is only essential in the initial stages of reprogramming. Their expression reactivates endogenous pluripotent genes such as Oct4 and Sox2, which contribute to the resetting of somatic cells back into hESC-state. Bisulfite genomic sequencing is used to show the methylation state of endogenous hESC-specific gene promoters in hiPSCs.58 In their study, Takahashi showed that the promoters of hESC-specific genes such as Oct4, Rex1, and Nanog were highly unmethylated as compared to their parent cells. Demethylation state of promoters indicates active state of the genes.

Stem Cells

337

Histone modification pattern varies between hESCs and differentiated cells. As iPSCs are reprogrammed differentiated cells, their histone modification pattern must correspond more to hESC and not that of the differentiated cells. Trimethylation occurs on specific locations of histone, which either activates or represses a particular gene. Trimethylation at histone H3 lysine 4 activates a gene while trimethylation at H3 lysine 27 represses.3 Takahashi showed that the promoter region of pluripotent-specific genes was methylated at H3 lysine 4 and demethylated at H3 lysine 27 in their study.58 They also showed bivalent patterns,91 which are characteristic of hESCs, exist at developmental-associated genes.3 In such modifications, a gene is activated by the initial trimethylation but repressed by the trimethylation on lysine 27. Like hESCs, the pluripotency of hiPSCs is shown by teratoma formation.98 Demonstrating that cells form three germ layers, through histological and immunohistochemical analysis, proves pluripotency of hiPSCs. Directed differentiation has also been used to demonstrate pluripotency of hiPSCs in many studies.58 This is considered one of the most stringent test researchers used to demonstrate pluripotency of hiPSCs. hiPSCs can potentially be used to make patient-specific cells. Research has shown that hiPSCs can be differentiated into many functionally relevant cells. Human insulin-secreting cells,124,127 functional cardiomyocytes,128 and hematopoietic cells94 have all been differentiated from hiPSCs.

1.23.3.3

Reprograming Methods

Before hiPSC can realize its promises in therapeutics, there are fundamental questions to answer. The use of viruses in introducing reprogramming factors is a major hurdle to overcome. Many researchers have recently demonstrated better ways of reprogramming. Reprogramming of human fibroblasts using an adenoviral delivery system130 without any transgene integration was demonstrated. In another study, plasmid transfection118 has been used to reprogram mouse cells without any transgene integration. Table 2 shows different reprogramming methods and their advantages and disadvantages. Although the above research has offered safer means of making iPSCs, there have been other studies that have suggested better alternatives. Woltjen and his team in their study used a piggyBac (PB) transposon system to induce pluripotency in human embryonic fibroblast cells.65 In making this PB, the flanks of the transgenic genes were attached with an inverted repeats.97 The PB plasmid carrying the genes Oct4, Sox2, Klf4, and c-Myc was driven by deoxycycline inducible promoter. When this plasmid was transfected together with PB and transposase-expressing vector, expression of transposase enzyme catalyzes insertion and excision of PB-flanked genes, hESC-like cells were produced only when cultured with deoxycycline. Once the induced cells became independent of deoxycycline, additional transient transposase expression was used to excise the transgenes. The endogenous pluripotent genes remained active even after the removal of transgene. Recently, Jia reported the use of “minicircle”105 DNA to reprogram adult human adipose stem cells. Minicircle vectors are supercoiled DNA molecules that lack bacterial origin of replication and antibiotic-resistance gene. Their lower activation level of exogenous silencing mechanism allows them to express ectopic factors for longer time; hence, higher transfection efficiencies93 are achieved compared to plasmids. hiPSCs derived using this method had no transgene integration while having higher efficiency compared to other plasmid transfection-based reprogramming methods. Another recent advance in reprogramming method is the use of recombinant protein in reprogramming human somatic cells. Kim fused a polyarginine transduction domain to the four reprogramming proteins (Oct4, Sox3, Klf4, and c-Myc).107 These factors with the transduction domain were expressed by stable human embryonic kidney (HEK)-293 cell line. When fibroblasts were treated for a few rounds with HEK cell extracts, they formed hESC-like colonies that expressed hESC-specific markers and were capable of differentiating to all three germ layers, confirming the formation of hiPSC. Another hurdle in the use of hiPSCs for clinical practices is the use of oncogenes as reprogramming factors. C-Myc119 is a known oncogene. Oct4,126 Sox2,92 and Klf4120 have all been linked to cancer. Overexpression or transgene reactivation can be potentially hazardous if hiPSCs were to be used in therapies. The use of recombinant proteins to induce pluripotency could potentially solve these problems. Alternatively, many researchers have succeeded in reprogramming using fewer genes albeit with lower efficiency. For example, Oct4108 alone was used to induce human NSCs (hNSCs) to hiPSCs. In another study, Oct4 was shown to be dispensable in reprogramming and could be replaced by orphan nuclear receptor, Nr5a2.102 Mouse embryonic fibroblasts induced with Nr5a2, Sox2, Klf4, and c-Myc produced iPSCs that showed all characteristics of mESCs.

1.23.3.4 1.23.3.4.1

Study of Diseases Sources of Cells

Initially, dermal fibroblasts were used because they were easy to obtain. Alternative sources of cells that can be accessed easily and reprogrammed with fewer factors have been sought. So far, many different cell types such as keratinocytes,89 adipose stem cells,122 cord blood cells,99 NSCs,108 amniotic fluid-derived cells,111 gut mesentery-derived cells,113 and hepatocytes28 have been used as the parental cells, showing the versatility of reprogramming. The retroviral transduced keratinocytes89 were reprogrammed at a much higher (100-fold higher) efficiency as compared to retroviral-transduced fibroblasts. Keratinocytes have higher expression levels of c-Myc and Klf4 than fibroblasts, which might explain the higher efficiency of keratinocyte reprogramming. Some stem/progenitor cells have higher expression levels of some of the endogeneous reprogramming factors. For example, neural stem cells108 (NSCs) have a higher expression of endogeneous Sox2, Klf4, and c-Myc and have been reprogrammed using

338

Stem Cells

Oct4 alone. Although the efficiency was very low, this study demonstrates that endogeneous expression can complement exogenously added factors. Endoderm-derived hiPSCs were not reported until Hua demonstrated the prospect of using hepatocytes in reprogramming.28 Hepatocyte-derived hiPSCs would also serve as a more amenable system for studying liver-specific diseases. These cell lines would allow comprehensive comparative analysis of the quality of hiPSCs from different origins.

1.23.3.4.2

Disease Models

Many disease-specific hiPSCs have been produced that could hold key to understand many diseases better and to develop new drugs. A study by Park et al.46 generated hiPSCs from patients with a variety of genetic diseases such as adenosine deaminase deficiency-related severe combined immunodeficiency, Shwachman–Bodian–Diamond syndrome, Gaucher disease type III, Duchenne and Becker muscular dystrophy, Parkinson’s disease, Huntington disease, juvenile-onset type I diabetes mellitus, Down syndrome, and the carrier state of Lesch–Nyhan syndrome. Motor neurons of amyotrophic lateral sclerosis (ALS)95 of a patient have also been derived by reprogramming fibroblasts taken from an 82-year-old lady carrying ALS. In their study, no characterization of the resulting motor neurons was performed to show disease-related phenotypes. Ebert went one step further and reported disease-related phenotypes in motor neurons derived from fibroblasts taken from a patient carrying spinal muscular atrophy (SMA), a motor neuron disorder.96 The hiPSC from SMA patient expressed relatively low amounts of survival motor neuron protein which is characteristic of SMA. When these cells were grown in neural induction medium, cells positive for mature motor neurons were produced. The number of motor neurons decreased overtime due to degeneration. In another study, dyskeratosis congenital (DC)90 positive cells restored telomere elongation when the cells were reprogrammed. DC is a rare disorder that causes premature aging. Genetic lesions affecting telomerase maturation and stability have been associated with DC. Telomerase activity and telomere elongation were restored in hiPSCs but DC state-phenotype was re-identified in mature cells. The SMA and DC studies show the possibility of recreating pathogenesis in laboratories using hiPSCs. Recreating and tracking the progress of diseases would potentially reveal underlying mechanisms.

1.23.3.4.3

hESC Versus hiPSC

It is important to realize that hESC is the “gold standard” to which hiPSC should be compared to in terms of their differentiation capability. Two recent papers comparing differentiated cells derived from hiPSC and hESC show disparities in their characteristics, although both show similar developmental principles. Feng showed that hiPSCs are able to differentiate into hemangioblasts, endothelial cells, and hematopoietic cells, which share morphological and phenotypic characteristics to that of hESCs-derived progenitors but with much lower efficiency.20 In addition, hiPSC-derived hemangioblasts showed significant increase in apoptosis and limited growth and expansion capabilities. Also, the endothelial and retinal pigmented epithelium cells derived from hiPSCs showed early cellular senescing as compared to their hESC-derived counterparts. In another independent study, inconsistencies in neurons derived from hiPSCs and hESCs were shown.104 Although both hiPSC and hESC used the same transcriptional network to develop into functional neurons over a similar developmental period of time, hiPSCs do this with much lesser efficiency and increased variability. These variable results were discovered across the hiPSC lines and were independent of reprogramming factors and methods used in inducing pluripotency.

1.23.3.5

Conclusion

The findings by Feng and Hua have definitely dampened the early excitement surrounding hiPSCs.20,104 A big question, as pointed out by the above two research groups that remains to be answered, is the exact nature of hiPSCs and their likeness to hESCs. Are they true equivalents of hESC? At the moment, it seems that the difficulties and doubts surrounding hiPSCs are technically related. With better understanding and methodology of reprogramming, hiPSCs could serve as disease models and have the potential to be used in therapeutics.

1.23.4

Neural Stem Cells

1.23.4.1

Initial Discovery

Research into hematopoiesis and the mammalian nervous system sparked off the search for the existence of NSCs. Initially, it was hypothesized that the mechanisms governing proliferation and migration of the cells in the central nervous system (CNS) are inactive; therefore, self-renewal of the cells in the CNS is greatly reduced. However, 3H-thymidine labeling provided evidence of cellular proliferation in the adult CNS, and NSCs were isolated from the striatal tissue of adult mammals.51 These neural precursors are regional, multipotent cells that are capable of self-renewal and differentiation into a vast diversity of neural phenotypes of the CNS. Since then, hNSCs have also been isolated in fetal and adult human CNS.22

1.23.4.2

Sources and Niche of NSCs

Adult NSCs are obtained mainly from the CNS, within the niches of the lateral ventricles and the hippocampus that provide the most abundant supply of these multipotent cells (Figure 2(a)).53 These niches are regions where NSCs reside where they maintain

Stem Cells

339

A

Lateral ventricles Subependymal zone

Hippocampus

Subgranular zone

Dentate gyrus

B

Neural stem cells Oligodendrocyte

Bipolar

Unipolar

Multipolar Pyramidal cell

Astrocyte

Neurons Figure 2 Sources and differentiation capabilities of neural stem cells. (A) The diagram shows the sites within the central nervous system where neural stem cells (NSCs) are located. (B) NSCs harvested from the lateral ventricles and hippocampus are cultured as neurospheres and are capable of proliferation and differentiating into oligodendrocytes, neurons, and astrocytes.

their properties of indefinite self-renewal and multipotency. The ability to expand and generate NSC lines has enabled the study and the development of cell-based therapies for neurological diseases.35 The subependymal zone (SEZ) is a germinal cell layer bordering the lateral ventricles of the brain (Figure 2(a)). It was reported that neurospheres can be cultured from cells residing within the SEZ and the resulting neural precursors were found to express tenascin-C, an extracellular matrix (ECM) protein with anti-adhesive properties that cause the rounding of the cells. In addition, Ki67 positive cells were also discovered in the SEZ, indicating proliferative activity. The necessary factors for neurogenesis have also been detected within the SEZ. Thus, it was proposed that NSC undergoes cell division in the SEZ to replace the NSC that underwent neurogenesis.35 The next major source of NSCs in humans is the dentate gyrus (DG) of the hippocampus.35 Neurogenesis was observed to occur within the subgranular zone (SGZ)131 of the DG (Figure 2(a)). Furthermore, drugs and physiological factors that affect the control of neurogenesis were also found to have an effect on the proliferation activity of the SGZ. Antidepressants upregulated neurogenesis while the downregulation of neurogenesis was effected by the increase in physiological stress133,134 on the cells.63 Several NSC lines have been generated since its discovery. One of the more popular cell lines used in research is the immortalized hNSC line, HB1.F3.42 Over the years, newer sources of NSCs have also been investigated. Nestin expression in the peripheral nervous system (PNS)140 has led to the identification of the spinal cord as another potential source of NSCs. Recently, a novel type of NSC has been described which is known as neural rosette cells (R-NSCs)132 derived from the neural differentiation of hESCs. R-NSCs are able to generate neural precursors expressing NSC markers, nestin and Sox2, and also have the ability to differentiate into full neuronal diversity.

340 Table 3

Stem Cells Neural stem cell markers Antigenic markers defining neural stem cells Nestin

Sox2

Nature of antigen

Type VI Transcription intermediate factor filament protein Neural cell types Neural stem U U cells Other neural Neural Neural cell types progenitors precursors Other cell Follicle stem cells, Human types endothelial cells embryonic stem cells

References

30

137

Musashi-1

PSA-NCAM

Vimentin

Nucleostemin

RNAbinding protein

Glycoprotein

Intermediate filament

Nucleolar protein

U

U

U Neural precursors

135

43

CD90

U

CD133

p75(NTR)

Glycoprotein

Type I transmembrane protein

U

U

Neurons

21

Embryonic stem Hematopoietic Glioblastoma cells, cancer stem cells, stem cells, cells, primitive mesenchymal hematopoietic cells in the stem cells stem cells, bone marrow endothelial progenitor cells 35 38 69

It is also crucial to note that the endothelial cells within the NSC niche are influential in the self-renewal and differentiation of NSCs. By stimulating neuroendothelial cell contact, the Notch-Hes-1 signaling pathway is triggered to assist self-renewal. This indicates that endothelial cells are essential constituents of the NSC niche.54 Furthermore, the endothelial cells are separated from the brain by the basal lamina, and the presence of heparan sulfate glycosaminoglycans on the basal lamina increases the binding affinity136 of the ECM to factors affecting the development of the NSC. This allows the basal lamina to control the fate of NSCs.

1.23.4.3

Characteristics

The field of NSC is in a state of rapid growth. Although there is still a deficiency in specific markers for NSC, the more widely used markers (including nestin, Sox2, and Musashi-1) are listed in Table 3. Nestin is an intracellular marker belonging to the category of class VI intermediate filament (IF) protein and is used to identify NSCs. Johansson et al.30 conducted an investigation to determine whether the ependymal cells of the lateral ventricles are NSCs. In the study, significant nestin expression was observed in these ependymal cells, which decreased in the cellular proliferating region of the subventricular zone (SVZ). The ependymal cells also displayed self-renewal and multipotency, verifying the identity of NSCs.30 Studies also showed that cultured cells from the neurogenic areas demonstrated self-renewing and multipotency capability and expressed the marker, Sox2. The expression of the marker was also observed in nestin-positive cells of the SVZ and DG.137 Thus, this supports the use of Sox2 as an NSC marker. In the maturing CNS, Musashi-1 protein is highly enriched in NSC. Research has shown that this high Musashi-1 expression is associated with NSCs that are capable of differentiating into neurons and glia cells, and its expression is downregulated after differentiation.135 Polysialic acid–neural cell adhesion molecule (PSA–NCAM) is expressed in cells undergoing neurogenesis during the development of the nervous system. Unlike nestin, PSA–NCAM is used to identify immature neural cells. Upon lysolecithininduced demyelination, expression of PSA–NCAM by progenitors of the rostral SVZ and the rostral migratory pathway (RMS) was increased. An increase in the proliferation activity of the neural progenitors located in the SVZ and RMS was also observed. Furthermore, 3H-thymidine labeling revealed the migration of the PSA–NCAM-immunoreactive cells of the SVZ toward the lesion and their differentiation into oligodendrocytes and astrocytes. These events show that NSCs differentiated into neural precursors in the SVZ, thereby providing an additional source of oligodendrocytes along with the localized neural progenitors for myelin repair.43 These results also indicate the ability of PSA–NCAM to identify NSCs.

1.23.4.4

Differentiation Capabilities

Since the discovery of NSCs by Reynolds and Weiss, the neurosphere assay has been identified as the standard assay for NSCs.51 This neurosphere assay has been constantly revised and improved by Deleyrolle and Reynolds to obtain reproducible cultures. Neurospheres can either generate secondary neurospheres or differentiate into the different neural phenotypes.16 In the neurosphere assay, spheres of diameters between 100 and 150 mm are made when passaging to avoid unintended differentiation of the NSCs and difficulty in dissociating the clusters. During passaging, the presence of epidermal growth factor (EGF) and/or basic FGF (bFGF) will induce the proliferation of the NSCs into neurospheres. By removing these growth factors, the NSCs will differentiate into the neural phenotypes. This differentiation process may be carried out at either high or low neurosphere densities.16 High-density differentiation ascertains the variety of neural phenotypes while low-density differentiation is used to validate NSC pluripotency.

Stem Cells

341

Besides the ability to undergo continued proliferation, the multipotency of NSC enables its differentiation into the key neural cell types, namely neurons, astrocytes, and oligodendrocytes (Figure 2(b)). As reported by Shen, the endothelial cells of the NSCs niche are capable of secreting factors that restrict the differentiation of the NSC and enhance neuronal production at the expense of astrocytes and oligodendrocytes.54 In a separate study, it was proposed that peroxisome proliferator-activated receptor g (PPARg) mediates the proliferation of NSCs via upregulation of the EGF receptor and activation of the extracellular signal-regulated kinase (ERK) pathway.139 Furthermore, the activation of PPARg pathway also led to the inhibition of neurogenesis. This was shown through the loss of microtubule-associated protein 2 (MAP2, a marker for neurons) expression in NSCs cultured with PPARg agonist. Furthermore, nestin expression increased while MAP2 decreased in both concentration- and time-dependent manner. Thus, providing that the activation of the PPARg pathway stops neurogenesis while preserving the NSCs as undifferentiated neurospheres. The detailed mechanisms behind the pathway are, however, obscure.

1.23.4.5

Clinical Trials

Much attention has been given to the application of NSCs in both cell-based therapy and tissue engineering. This has led to the investigation of NSCs for therapeutic use in various neurological disorders, such as lysosomal storage diseases, stroke, and cancer.

1.23.4.5.1

Batten’s Disease

StemCells Inc. is transplanting fetal NSCs for children with neuronal ceroid lipofuscinosis (NCL), which is also known as Batten’s disease, and is currently seeking approval for the clinical trial of its human CNS-stem cells (HuCNS-SCs). The disorder is a result of genetic mutations, such that enzymes processing cellular waste substances are insufficiently produced. This leads to the accumulation of the waste and impairs neural cell functioning. Eventually, the patients will begin to lose motor skills, sight, and mental capacity. The proposed trial aims to further assess the safety of HuCNS-SC in NCL and to study the ability of the cells to reduce the advancement of the disease. The trial will require the patients to be transplanted with HuCNS-SC and immunosuppressed for 9 months, followed by a 12-month evaluation at regular intervals and a separate 4-year observational study upon the conclusion of this trial.

1.23.4.5.2

Stroke

Current advancements in NSC therapy have also reached the stage of the first-in-man clinical trial. This trial was put forward by ReNeuron Group plc to test its ReN001 stem cell therapy for stroke patients. The commercial-grade therapy will employ the NSC line, CTX, and has been proven capable of reversing the functional deficits linked to preclinical models of stroke.

1.23.4.5.3

Cancer

It is also worth noting that NSCs have an exceptional tropism for malignant gliomas. This property allows the engineered NSCs to be used as possible drug delivery agents138 to the gliomas. An investigation conducted by Aboody described the ability of NSC to track tumor cells and deliver the therapeutic cytosine deaminase.1 This enables the application of these NSCs as potential drugdelivery agents in the treatment of cancer. Several mechanisms govern the tumor homing of stem cells, and SDF1-CXCR4 chemokine axis-guided migration is one of the most significant mechanisms. NSCs that express CXCR4 move along the concentration gradient of SDF1a released by the tumorogenic cells to the site of the tumor.18

1.23.4.6

Conclusion

It is intriguing that cells isolated from what has long been viewed as the most quiescent among the bodily tissues can display such startling degree of plasticity and astonishing growth capacity. With new discoveries constantly adding to unraveling the enigma of the neurogenic process in adult CNS, the progress of the relatively new field of NSCs has greatly accelerated. The availability of hNSCs lines also promises to open several therapeutic venues for the treatment of neurological disorders. A rise in clinical trial approvals will be expected in the near future, which may eventually provide solutions to current long-term therapeutic hurdles.

1.23.5

Mesenchymal Stem Cells

1.23.5.1

Initial Discovery

MSCs are non-HSCs that are multipotent, being able to differentiate into bone, fat, muscle, and cartilage but not blood.5 They were first discovered by Alexander Friedenstein in the 1960s when he first observed bone marrow (BM)-stored stem cells that form mesenchymal tissues. In his pioneering experiment, Friedenstein immersed BM in plastic culture dishes for hours before removing nonadherent cells to eliminate most of the hematopoietic cells. The remaining cells appeared heterogeneous initially. However, the most adherent cells took on spindle-like shapes and remained dormant cells for about 3 days before dividing quickly. These spindle-shaped cells grew more homogeneously in morphology, after passaging. More importantly, the cells could differentiate into colonies that seemed like sediments of cartilage or bone.50 Such cells were defined as “colony forming units fibroblasts”

342

Stem Cells Table 4

Sources of MSCs and cell types that can be formed by MSCs Cell types differentiated from mesenchymal stem cells (MSCs)

Sources of MSC

Adipocyte

Chondrocyte

Osteoblast

Myoblast

Neuron

Cardiomyocyte

References

Bone marrow Muscle Trabecular bone Adipose tissue Periosteum Fetal MSCs Synovial membrane Umbilical cord blood hESC

U U U U U U U U U

U U U U U U U U U

U U U U U U U U U

U U

U

U

U U

U

17 29 60 71 44 8 15 19 4

U

(CFU-F). Subsequently, MSCs were renamed as marrow stromal cells in hematological literature, followed by MSCs, and most recently also known as multipotential mesenchymal stromal cells.146

1.23.5.2

Sources and Niches of MSCs

MSCs are normally obtained from BM aspirates of the superior iliac crest of human pelvis.17 Although the BM serves as the primary reservoir of MSCs, studies have shown the possibility of isolating MSCs from skeletal muscle connective tissue,29 human trabecular bones,60 adipose tissue,71 periosteum,44 fetal blood, fetal liver,8 synovial membrane,15 and umbilical cord blood (UCB).19 Table 4 summarizes the sources of MSCs and cell types that can be formed by MSCs. Homogeneous fibroblastic MSCs can also be derived from hESCs by culturing EBs for a week with hESC growth medium in culture dishes. The EBs are then plated onto gelatin-coated plates together with human BM stromal cell growth medium. Upon reaching confluency, cells can be passaged repeatedly until fibroblastic MSCs are observed.4 The niche where MSCs remain in their naïve state consists of nearby nonstem cells, extracellular matrix, and proximal soluble molecules.52 It was proposed that MSC niches possess a perivascular nature. Experiments on perivascular cells revealed the presence of MSC markers at the surface of native, noncultured perivascular cells. In addition, long-term cultured human perivascular cells from various organs (e.g., pancreas) expressed all established markers of MSCs. Intriguingly, they also retained myogenicity and displayed trilineage potential at the clonal level, irregardless of their tissue of origin.13 These data support the hypothesis that the origin of MSCs traces back to the walls of blood vessels. More specifically, MSCs belong to a subset of perivascular cells14 and this may be a useful indication for the elusive niche location of MSCs. Being easily accessible to the circulatory system may allow MSCs to reach other tissues and this suggests that MSCs do play an intrinsic role in the tissue-healing processes.

1.23.5.3

Characteristics

Currently, MSCs do not have a set of unique protein markers to allow for their reproducible isolation but they can be recognized by several cell markers and associated antibodies. Markers include certain cell adhesion molecules (e.g., CD106), receptors (e.g., CD44), and co-stimulatory molecules (e.g., CD73), while Stro-1 antibody can be used to identify nonhematopoietic progenitor BM stroma cells.55 It is useful to note that MSCs lack most hematopoietic markers (e.g., CD34), some immune cell markers (e.g., CD11b), and receptor molecules (e.g., CD14).48 Table 5 summarizes some of the commonly associated MSCs markers. MSCs do not have the ability for infinite self-renewal. The total numbers of MSCs decline with age and adult MSCs have a limited passage number.9 For example, the frequency of MSCs is about 1 in 10 000 nucleated marrow cells in a newborn baby. This drops to levels 10–100-fold lower in older individuals.

1.23.5.4

Differentiation Capabilities

MSCs have been shown to have trilineage differentiation capability into osteoblasts, adipocytes, and chondrocytes by many laboratories. Osteogenic differentiation requires a monolayer of MSCs to be incubated with b-glycerol-phosphate, ascorbic acid-2-phosphate, dexamethasone, and fetal bovine serum. The incubated MSCs should present osteoblastic morphology together with relatively high level of alkaline phosphatase and calcium. A suitable assay for osteoblast will be Von Kossa staining, a technique that subjects cell cultures to silver nitrate solution and strong light in order to quantify calcium deposition.5 For adipogenic differentiation, MSC monolayer cultures are incubated with isobutylmethylxanthine to form adipocytes with lipid vacuoles. This process is induced by receptors such as nuclear receptor, PPARg, transcription factors, and fatty acid synthetase. The lipid vacuoles within adipocytes can be visualized by oil red O staining, a technique that uses fat-soluble oil to stain lipid and fat on frozen sections.5

Stem Cells Table 5

Surface antigens commonly associated with MSCs

Marker type

Surface antigen

Marker description

References

Positive

CD29 CD44 CD71 CD73 (SH 3 and 4) CD90 CD105 (SH2)

Common beta subunit of integrins Hyaluronic acid receptor Transferrin receptor Costimulatory molecule Thy-1 Regulatory component of TGF-beta receptor complex Vascular cell adhesion molecule Receptor for tumor necrosis factor Interleukin-4 receptor subunit Cell-surface protein Immune cell marker Lipopolysaccharide (LPS) receptor Endothelial and hematopoietic cell marker Primitive hematopoietic stem cell marker Hematopoietic cell marker Hematopoietic stem/progenitor cell marker

48

Negative

343

CD106 CD120a CD124 Stro-1 CD11b CD14 CD31 CD34 CD45 CD117

45 48 45 48

Chondrogenic differentiation is initiated in a three-dimensional culture format with serum-free medium supplemented with transforming growth factor-b. Under such conditions, the morphology of MSCs changes from a fibroblastic appearance to developing cartilage-specific matrix layers filled with glycosaminoglycans. Toluidine blue indicator, a polychromatic dye, can be used to visualize glycosaminoglycan-containing components.5 In addition, such differentiated MSCs were found to generate type II collagen, a typical cartilage property.48 Other than these three lineages, MSCs have also been shown to be able to differentiate into myoblasts, cardiomyocytes, and even neurons. However, it was hypothesized that these cells with nonmesodermal origin may be a result of a phenomenon known as “stem cell plasticity,” a transdifferentiation process in which organ-specific stem cells are no longer restricted to forming the differentiated cell types of the tissue where they reside. To date, this phenomenon has been observed based on deductions or incomplete evidences as the specific mechanism behind transdifferentiation remains undefined.147 Figure 3 summarizes the multipotent traits of MSCs.

1.23.5.5 1.23.5.5.1

Clinical Trials Immune-Modulatory Therapy

A summary of the various ongoing or completed clinical trials involving MSCs is presented in Table 6. Currently, MSCs offer the most promising clinical candidate for immune-modulatory cell-based therapy. Osiris Therapeutics’s product Prochymal, a preparation of MSCs obtained from the BM of healthy adults, is presently being evaluated in two phase-III clinical trials for steroid refractory graft-versus-host disease (GVHD) and acute GVHD, after very successful phase II trials. Acute GVHD. Previously, Osiris Therapeutics performed a successful phase II clinical trial whereby human MSCs were used to treat de novo acute GVHD. Patients suffering from grades II–IV GVHD were chosen at random for two Prochymal treatments at dosages of 2 or 8 million MSCs per kg, together with infusion of corticosteroids. Out of 32 enrolled patients, Prochymal achieved 94% in overall response rate with an outstanding complete remission rate of 77%. Figure 4 depicts a pair of feet suffering from grade IV GVHD prior to treatment. Typically, the skin would breakdown at many locations together with severe blisters. However, most of these GVHD symptoms disappeared by day 18 after Prochymal treatment. In addition, these treatments induced neither administrative harm nor ectopic tissue development. Regardless of the variation in the dosage, safety, and efficacy levels remained the same. This study presents the remarkable suitability of MSCs in the treatment of acute GVHD.31 Crohn’s disease. Prochymal is also used in clinical trials for the treatment of moderate to severe Crohn’s disease. According to past phase II studies, patients’ inflamed intestines showed diminishing inflammatory conditions and crypt regeneration following Prochymal treatment. In one such open-label trial, 10 patients suffering from Crohn’s disease were randomly selected. Having been treated with steroids, methotrexate, and remicade previously, these patients showed little improvements in their illness. By day 9 of Prochymal treatment, there were signs of recovery in terms of intestinal inflammation and ulceration reduction as well as crypt formation144 in some patients. By day 28, every participating patient saw a drastic drop in Crohn’s disease severity and, at the same time, tolerated Prochymal relatively well.144 Figure 5 shows the large intestine before and after Prochymal treatment. These results suggest that MSCs may also be suitable to treat Crohn’s disease. Osiris Therapeutics’s phase III trial involving evaluation of Prochymal for Crohn’s disease recently halted enrolment of patients because of a flaw in the trial design, which resulted in significantly higher than expected placebo response rates. The company held an extended evaluation following the event and has successfully completed the study but is yet to publish the results.

344

Stem Cells

Mesenchymal stem cell

Osteogenesis

Adipogenesis

Chrondrogenesis

Osteocyte

Adipocyte

Chrondrocyte

Colony forming units-fibroblasts (CFU-F)

Myogenesis

Muscle myoblast

Others

Neuron

Cardiomyocyte

Figure 3 Mesenchymal stem cell (MSC) multilineage differentiation potential. MSCs are found to be able to differentiate into osteocytes, adipocytes, chondrocytes, myoblasts, neurons, and cardiomyocytes.

Table 6

MSCs and their related clinical trials

Condition

Intervention

Organization

Phase

Status

Crohn’s disease Graft vs. host disease Cartilage injury/osteoarthritis Myocardial infarction Myocardial infarction Type II/III

Drug: Prochymal Drug: Prochymal Drug: Cartistem Drug: Prochymal Drug: Provacel Mesenchymal stromal cells

Osiris Therapeutics Osiris Therapeutics Medipost Co Ltd. Osiris Therapeutics Osiris Therapeutics Children’s Hospital of Philadelphia

Phase Ill Phase Ill Phase III Phase II Phase I Phase l

Completed Ongoing Recruiting Recruiting Ongoing Recruiting

Figure 4 Prochymal treatment for acute graft-versus-host disease, Osiris Therapeutics Inc. (2010), prochymal graft-versus-host disease. Available at http://www.osiristx.com.

Stem Cells

Crohn’s disease before treatment

345

Crohn’s disease posttreatment at day 9

Figure 5 Prochymal treatment for Crohn’s disease (Source: Osiris Therapeutics Inc. (2010), prochymal Crohn’s disease. Available at http://www. osiristx.com).

1.23.5.5.2

Bone Regeneration

A defect known as osteogenesis imperfecta (OI) can be treated using MSC transplantation. This illness is associated with the generation of abnormal type I collagen in bones, causing slow bone development, frequent fractures, and bone distortion. In a demonstration, Horwitz et al. transplanted BM cells from human leukocyte antigen (HLA)-identical siblings to patients suffering from OI. Results showed about 2% of the osteoblasts in recipient’s BM came from the donor. These MSCs can develop into normal osteoblasts, leading to fast bone development and reduced fracture frequencies.142 Subsequent similar trials resulted in children with significant gain in total body length with a median of 7.5 cm, 6 months after transplantation. Bone mineral content improved by 45–77% of baseline values and frequency of fractures dropped from 10 to 2.141 Follow-up investigations proved that the introduction of purified allogeneic MSCs may reap better therapeutic results for allogeneic BM transplantation in the treatment of OI.41 A more recent clinical trial makes use of in utero MSC transplantation in patients with severe OI. Findings showed that allogeneic fetal MSCs can engraft and differentiate into bone in a human fetus even when the recipient is immunocompetent and HLA-incompatible.37 As such, MSCs can possibly be a worthy consideration in bone regeneration. The Children’s Hospital of Philadelphia in the United States is currently recruiting study subjects for their phase I study to assess the safety and feasibility of repeated infusions of MSCs in children with OI.

1.23.5.5.3

Cartilage Regeneration

The application of autologous BM-derived MSCs into patients with osteoarthritis has been reported recently. Twelve patients were treated with injected MSCs into an articular cartilage defect in their knee joints and results were compared with a control group with no MSCs treatment. Comparatively, MSC application resulted in defects being gradually covered with white soft tissue and exhibiting more desirable arthroscopic and histological grading results than the control group.149 Despite the insignificant difference between the two groups, this study demonstrated the possibility of using MSCs for cartilage repair. More studies will be required to demonstrate the usefulness of MSCs in cartilage regeneration even though results in animal models seem to be very promising. Articular cartilage regeneration is being investigated in a clinical study that showed the possibility of using UCB-derived MSCs to treat old patients who possess large lesions. This study is conducted by Medipost Co Ltd, which is currently recruiting participants to compare the efficiency and safety between cartistem and microfracture treatment in patients with knee articular cartilage defect.

1.23.5.5.4

Myocardium Regeneration

A new discovery indicates signs of cardiomyocyte regeneration of MSCs following myocardial infarction (MI). In a recent human study, patients were chosen at random for intracoronary introduction of autologous BM MSCs following MI. After 3 months, it was observed that there were less-damaged regions in the heart coupled with more significant improvements in terms of contraction and heart function after MSC treatment, when compared with patients who underwent standard therapeutic procedures.145 Although the exact mechanism behind such myocardium regeneration remains unknown, the study serves as a useful insight for future tissue-engineering and regeneration purposes.148 At the moment, Osiris Therapeutics is starting phase II clinical trials for the treatment of acute MI using Prochymal. Concurrently, Osiris Therapeutics is also carrying out a phase I randomized, double-blind, placebo-controlled, dose escalation, and multicenter study to determine the safety of intravenous ex vivo-cultured adult human MSCs (Provacel) following acute MI.

1.23.5.5.5

Skeletal and Neurological Disorders

Metachromatic leukodystrophy (MLD) or Hurler’s syndrome are diseases that result in severe skeletal and neurological disorders. A phase I clinical trial was conducted with Hurler syndrome patients, who previously underwent successful BM transplantation from

346

Stem Cells

an HLA-identical sibling. MSCs from a BM aspirate of the original donor were infused into these patients. Interestingly, four out of six patients with MLD present huge improvements in nerve conduction velocities after MSCs infusion.143 In addition, the bone mineral density either remained unchanged or improved mildly in all patients. More importantly, MSC infusion is considered safe as there were no apparent adverse effects in all patients. This study suggests that donor allogenic MSC infusion may be a method to improve the conditions among patients down with MLD. Nonetheless, the use of MSCs in such diseases needs further evaluation in order to establish it as a suitable cure.

1.23.5.6

Conclusion

The unique capabilities of MSC that are immunomodulatory, multipotential, and fast proliferating make them the most promising stem cell candidate for regenerative medicine. Although recent late-stage clinical data had been relatively impressive, the mechanism of action behind MSCs therapy remains largely unknown. It will be important to understand the properties of MSCs so as to develop more strategic medical solutions.

1.23.6

Hematopoietic Stem Cells

1.23.6.1

Initial Discovery

The existence of HSCs in mammalian BM was demonstrated between 1949 and mid-1950s. It was subsequently proven that in lethally irradiated mice, the protection of BM after intravenous infusion of marrow166 was due to transplantable HSCs.167 Till and McCulloch demonstrated the formation of clonogenic colony of hematopoietic lineages in the spleen termed CFU-spleen (CFU-S) of irradiated mice following transplantation of healthy BM donor cells as well as the self-renewal capacity of these colony-forming cells.7,172

1.23.6.2

Sources

Besides the adult BM being the established source of HSCs,177 there have been documentations of alternative sources of HSCs.

1.23.6.2.1

Peripheral Circulating Blood

HSCs can be obtained from peripheral circulating blood.164 To induce more cells to migrate from the marrow to the peripheral blood, the donor is injected with granulocyte-colony-stimulating factor (GCSF). The harvested blood is then filtered by apheresis to collect CD43þ white blood cells, while the remaining blood containing red blood cells is returned to the donor. The large infusion of T cells residing in the peripheral blood could result in complications of donor toxicities as well as greater risk of GVHD. However, a greater number of HSCs can be harvested in each session of apheresis and also faster hematopoietic recovery compared to a BM harvest, which requires spinal anesthesia and has its associated risks.33

1.23.6.3

Umbilical Cord Blood

UCB has also been documented in 1974 to be enriched with HSCs.34 The first cord blood transplantation involved the transfusion of cord blood cells (HLA identical) into a patient with Fanconi’s anemia.25 Fetal cord blood from a newborn’s placenta or umbilical cord can be stored in the cord blood bank for future transplants. HSC transplantation (HSCT) using UCB is believed to result in less GVHD and posttransplantation problems. Cord blood from the patient’s own body reduces the complications arising from immune rejections upon engraftment. UCB also requires a lower dose for infusion compared to adult BM due to the better engraftment of UCB-derived HSC forming larger in vitro colonies.26

1.23.6.3.1

hESCs and hiPSCs

Recently, researchers have derived multipotential common myeloid progenitors (CMPs) by differentiating hESCs and hiPSCs. CMPs are cells differentiated two stages after long-term HSC (LT-HSC) and can form myeloid lineages (Figure 6). Using growth factors combined with various cytokines in serum-free media, two independent hESC lines and one hiPSC line were differentiated into hematopoietic progenitors directed toward the monocyte–macrophage lineage.27 Due to the pluripotency and superior self-renewing capability of hESCs and hiPSCs, it is postulated that these could generate an abundant supply of hematopoietic progenitors for transplantation/engraftment studies or in vitro studies.

1.23.6.4

Niches

The anatomical location in adult human tissues where HSCs reside is termed the HSC niche. Schofield introduced the “niche” hypothesis for HSC BM microenvironment where HSCs are able to proliferate while maintaining their “stemness”.52 The HSC niche includes supporting cells that form the microenvironment for the HSCs as well as signals and extrinsic factors associated with these supporting cells.39 Thus, the niche plays a significant role in the regulation of HSC homeostasis through the balance of its self-renewal and differentiation potential.68 Figure 7(a) illustrates the location where HSCs reside within the human BM.

Stem Cells

347

LT-HSC

Basophil

ST-HSC CMP

Eosinophil

Dendritic cell CLP GMP

MEP

Pro-B-cell CFU-GM Neutrophil

CFU-Meg

CFU-E

Pro-T-cell

Natural killer cell

B lymphocyte Monocyte

T lymphocyte Erythrocytes Megakaryocyte

Macrophage

Platelets Myeloid pathway

Lymphoid pathway * Dendritic cells can be derived from CMPs (myeloid pathway) from monocytes

Figure 6 Flowchart depicting differential pathway of hematopoietic stem cells (HSCs) into individual hematopoietic cells. Long-term LT-HSCs that have migrated out of their niches, hence losing their self-renewal capability, are termed short-term ST-HSCs. ST-HSCs then differentiate into two different pathways: the myeloid and the lymphoid pathways. The myeloid pathway begins when ST-HSCs differentiate into common myeloid progenitors (CMPs), which further divide into granulocyte/macrophage progenitors (GMPs) and megakaryocyte/erythrocyte progenitors (MEPs). These progenitors then give rise to colony-forming units (CFUs). GMPs differentiate into CFU-granulocyte/macrophage while MEPs differentiates into CFU-megakaryocytes and CFU-erythrocytes. CFUs mature into neutrophils, macrophages, platelets, and erythrocytes. The lymphoid pathway begins when ST-HSCs differentiate into common lymphoid progenitors (CLPs) that differentiate into pro-T, pro-B, and natural killer (NK) as well as dendritic cells (DCs). Pro-T and Pro-B cells then mature into T and B lymphocytes of the immune system, respectively.

1.23.6.4.1

Osteoblastic Niche

Two extensively studied HSC niches are the endosteal region termed “osteoblastic niche” and the perivascular area termed “vascular niche,” as shown in Figure 7(b); however, most papers published defining these HSC niches are based on murine models. The barrier to the experimental studies of hHSC niches is the lack of suitable model systems. In the osteoblastic niche, the bonelining osteoblastic cells have been identified as part of the hematopoietic microenvironment.56 This hypothesis is based on the evidence that human endosteal osteoblasts constitutively produce GCSF that support hematopoiesis.175 Researchers have also presented evidence on the successful identification of human LT-HSCs that localize to this BM niche in their steady state, where they maintained their quiescent state by interacting with key components of the niche.66 In Figure 7(c), the signaling molecules and pathways that form a dynamic interaction with the osteoblastic niche are membrane-bound SCF/c-KIT and Sonic hedgehog that promote HSC self-renewal150,171; Notch ligand Jagged-1 for self-renewal and/or maintenance of multipotentiality61,161,162; and controlled regulation of Wingless (Wnt) signaling that promotes HSC quiescence.180,181

1.23.6.4.2

Vascular Niche

In the vascular niche, studies have shown that in the human BM, hematopoiesis develops specifically in structures organized by BM sinusoidal endothelial cells (BMECs).10 Functionally, the human BMECs are associated with HSC homing and the mobilization of HSC to the vascular niche under hematological stress to undergo self-renewal or differentiation of more committed progenitors into myeloid and megakaryocyte lineages.169

1.23.6.5

Characteristics

Characterization of protein markers for each differentiation stage of HSC is summarized in Table 7. LT-HSC is capable of self-renewal and maintenance of their multipotent state. As LT-HSCs become committed to differentiation, they lose CD133 expression and gain CD38 expression. Hence, together with CD34þ and Lin, it is possible to distinguish between the differentiation or self-renewal fates of HSCs.

348

Stem Cells

B Osteoblastic and vascular niches

A Location of HSCs within the human

bone marrow

Osteoblasts

Bonemarrow sinusoidal endothelial cells (BMESs)

Self-renewal

Self-renewal

Under Hematological Stress LT-HSC

Vascular niche Differentiation

ST-HSC

Trabecular zone

LT-HSC

LT-HSC

Notch Jagged-1 Nucleus

Bone-marrow sinusoidal endothelial cells (BMECs)

Long bone region

Bone

Membranebound SCF

c-KIT

Endosteum (osteoblastic niche)

LT-HSC Osteoblastic niche GCSF

Frizzled LRP 5/6

Nucleus

Niche osteoblast

Wnt

Smo Ptc

Shh

C Interactions in the osteoblastic niche governing HSC maintenance

Figure 7 Model of the adult human bone marrow, hematopoietic stem cell (HSC) niches, and interactions within the osteoblastic niche. (A) Schematic diagram showing the anatomical location where HSCs reside within the bone marrow. The endosteum separates bone from the bone marrow and it comprises different cell types including osteoblasts. (B) Quiescent HSCs undergo long-term self-renewal in the osteoblastic niche but migrate to the bone-marrow sinusoidal endothelial cells (BMECs) in the vascular niche in response to injury. Mobilized HSCs in the vascular niche will undergo either self-renewal or differentiation along the myeloid and megakaryocyte lineages. (C) Several signalling molecules and pathways identified at the osteoblastic niche have been implicated for long-term HSC maintenance. Protein markers expressed by LT-HSCs and their hematopoietic progenitors

Table 7

CD34 CD38 CD133 Lin HLA-DR Thy1/CD90 Flt3/CD135 CD 45RA CD 10 IL-3Ra/CD123 TpoR/CD110 References LT-HSC CLP CMP GMP MEP

þ þ þ þ þ

 þ þ þ þ

þ    

    

þ

þ 

lo/þ þ lo/þ hi/þ 

þ  þ 

þ   

lo/ lo lo/

  þ

6,23,32,67,155,157 23,32,155,157 23,32,40,153,155,157 23,32,40,153,155,157 23,32,40,153,155,157

Lin, negative for lineage specific markers (CD2, CD4, CD8, CD14, CD19, CD20, CD16, CD56, and glycophorin A).

1.23.6.5.1

Self-Renewing HSC Population

LT-HSCs can be identified by the expression of the LinCD34þThy-1þ protein markers6 where lineage-specific cell-surface markers (Lin) include all of the following markers: CD2, CD4, CD8, CD14, CD19, CD20, CD16, CD56, and glycophorin A.23 This HSC-enriched population further expresses Flt3, which has been found to maintain LT-HSC as well as CMPs and common lymphoid progenitors (CLPs).32 CMPs and CLPs are two different stages arising from LT-HSC, as shown in Figure 6. A novel marker CD133 has been discovered to characterize LT-HSC but it is not known whether the use of CD133 is more advantageous than CD34 as a HSC marker.67 The primitive LT-HSC population has also been reported to exhibit CD38 and HLA-DRþ phenotypes within the CD34þsubset.157 These characteristic markers are presented in Table 7.

1.23.6.5.2

Common Lymphoid Progenitors

CLPs are characterized phenotypically by the LinCD34þCD38þCD45RAþCD10þpopulation (see Table 7). This population has been seen to possess lymphoid differentiation potential (T-cell and B-cell) while being devoid of any myeloid/erythroid

Stem Cells

349

differentiation potential.23 Flt3 expression is downregulated in myeloid progenitors, but upregulated in the lymphoid pathway.32 Unlike LT-HSC, HSC that reaches the differentiation stage of CLPs have very low or undetectable levels of Thy-1.23 Differential expression of CD45RA can be used to distinguish between myeloid and B-lymphoid progenitors. Studies have shown that the CD34þCD45RA population gives rise to early myeloid progenitors while the CD34þCD45RAþ population gives rise to B-lymphoid cells.155

1.23.6.5.3

Common Myeloid Progenitors

CMPs are characterized phenotypically by the LinCD34þCD38þCD45RACD10population (Table 7). To distinguish between the different myeloid progenitors, Manz et al.40 proposed the use of IL-3Ra and CD45RA. They proposed three phenotypic definitions of early myeloid progenitors. CMPs can be represented as IL-3RaloCD45RA, granulocyte/macrophage progenitors (GMPs) represented as IL-3RaloCD45RAþ and megakaryocyte/erythrocyte progenitors (MEPs) represented as IL-3RaCD45RA.40 GMPs and MEPs are myeloid progeny differentiated from CMPs (Figure 6). However, due to the difficulty in achieving good isolation of the respective submyeloid populations using IL-3Ra, TpoR was employed as an additional characterization marker to distinguish GMPs from MEPs.153 TpoRþ cells were shown to limit differentiation to solely erythroid lineages while TpoR cells still maintained GM potential.153 Furthermore, Flt3 expression is upregulated in GMPs but downregulated in MEPs.32 Hence, distinction between GMPs and MEPs can be identified by the expression levels of Flt3 and TpoR in these progenitors. GMPs are high in Flt3 but negative for TpoR while MEPs are low in Flt3 and positive for TpoR. Table 7 illustrates the expression of these early myeloid characteristic markers.

1.23.6.6

Differentiation Capability

HSCs possess the ability to self-renew and differentiate into all types of blood cells, especially those involved in the human immune system. The differentiation fates of these blood-lineage committed cells may be divided into two classes: the lymphoid and myeloid lineages. Lymphoid cells consist of natural killer (NK), T- and B-lymphocytes, while myeloid cells consist of granulocytes (neutrophils, eosinophils, and basophils), monocytes, macrophages, erythrocytes, megakaryocytes, and mast cells. Though lymphoid and myeloid cells originate from the same LT-HSCs, they have distinct differential pathways.160 The reader is referred to Figure 6 for details of the HSC differentiation lineage pathways.

1.23.6.6.1

Lymphoid Lineage

LT-HSCs differentiate into short-term HSCs (ST-HSCs) that have limited self-renewal capacity. ST-HSCs then give rise to CLP and CMPs. CLPs (LinCD34þCD10þThy-1from human BM) are capable of differentiating into B cells, T cells, NK cells as well as lymphoid dendritic cells.163 CLPs differentiate into pro-B cells that are committed to B-cell lineages. In the thymus, CLP cells differentiate into pro-T cells. It was shown that human adult HSC expressing LinCD34þCD10þ markers are CLPs with complete lymphoid differentiation potential capable of forming B, NK as well as T cells.23 In the same study, after microinjection of CLPs into allogeneic fetal thymic grafts, cells derived were tested to express characteristic T-cell surface markers of CD4 and CD8. NK cells derived from adult BM cultures of CLPs expressed cell-surface phenotypes of CD56 and CD3. While CD19þ B cells were also derived from adult BM cultures directed to B lineage differentiation, hence proving the lymphoid differentiation potential of CLPs.23

1.23.6.6.2

Myeloid Lineage

ST-HSCs also differentiate into myeloid lineages. CMP cells possess two differentiation fates: CD45RAIL-3Ra MEPs and CD45RAþIL-3Ralo GMPs.40 GMPs subsequently differentiate into CFU-granulocyte/monocytes (CFU-GM), while MEPs give rise to CFU-megakaryocytes (CFU-Meg) and CFU-erythrocytes (CFU-E). Each CFU then further differentiates into their respective committed lineages as shown in Figure 6. Myeloid-committed lineage cells such as MEPs and GMPs has been shown to co-express CD34 and CD45RA surface cell markers allowing a two color flow cytometer analysis to quantify and characterize human early myeloid progenitors.155 Further, the expression levels of IL-3R on CD34þ cells determine the lineage fate of myeloid progenitor cells. As mentioned above, IL-3R cells differentiate into erythrocytes while IL-3Rlo cells differentiate into granulocytes and macrophages.40

1.23.6.7

Clinical Applications

HSC transplantation sprang to life in 1957 when the first paper on human marrow grafting (intravenous infusion) was published, in which only one out of six patients obtained a transient graft.178 Later, syngeneic marrow infusion in two patients with refractory leukemia was successfully carried out.179 The modern era of human allogeneic marrow grafting began in 1968 when the first successful BM transplantation was performed by Good and co-workers.24

1.23.6.7.1

HLA-Matching of HSC Sources

HSCT can render its cure by reconstituting damaged blood-forming cells and readjustment of the immunological rheostat after high-dose chemotherapy to eliminate disease.158 There are three types of HSCT: syngeneic, autologous, and allogeneic transplants. Syngeneic transplantations occur between identical twins. Autologous transplantations use the HSCs obtained directly from the patient and hence do not cause any complications of tissue incompatibility; whereas allogeneic transplantations involve the use

350

Stem Cells Table 8

Comparison of the sources of HSCs from bone marrow, peripheral blood, and cord blood used for HSCT

Aspect of comparison

Bone marrow

Cord blood

Peripheral blood stem cell

Stringency of donor–recipient HLA matching Speed of engraftmenta Risk of acute and chronic graft-versus-host diseases

Close matching needed

Less stringent

Close matching needed

þþ þþ

þ þ

þþþ þþþ

Duration to absolute neutrophil count (ANC) >500 and platelets >20 000 without growth factor/transfusion support.

a

Table 9

Commonly treated malignant and nonmalignant diseases with HSCT

Malignant diseases

Nonmalignant diseases

Allogeneic HSCT

Autologous HSCT

Acute myelogenous leukemia Chronic myelogenous leukemia Acute lymphoblastic leukemia Chronic lymphoblastic leukemia Multiple myeloma Non-Hodgkin’s lymphoma Hodgkin’s disease Sickle cell anemia Aplastic anemia Fanconi’s anemia Wiskott–Aldrich syndrome Thalassemia major

Acute myelogenous leukemia Multiple myeloma Non-Hodgkin’s lymphoma Hodgkin’s disease

Autoimmune diseases

of donor HSCs, either genetically related or unrelated to the recipient. Due to donor source, allogeneic HSCT must satisfy compatibility at the HLA loci to lower the risks of transplant, which include graft rejection and GVHD. Related donor–recipient HLA best matched at 6/6 alleles (A, B, DR), while unrelated donor transplants typically requires at least 8/8 alleles (A, B, C, DRB1) for optimal HLA matching.165 Any single mismatch of these HLA loci would render worse survival, treatment-related mortality and higher risk of acute GVHD.159

1.23.6.7.2

BM, Peripheral Blood, and Cord Blood for HSCT

The source of HSCs from UCB grants greater flexibility of donor–recipient HLA-matching.156 In comparison with HSCs from BM, the use of HSCs from UCB and peripheral blood stem cells (PBSC) for HSCT have been increasing worldwide between 1998 and 2007.168 In a large study of BM and PBSC recipients afflicted with myeloid malignancies, the Canadian Bone Marrow Transplant Group reported a faster engraftment and better overall survival in the PBSC group.151 However, this was at the expense of increased rates of acute and chronic GVHD as shown in separate randomized trials.151,152 Cohort studies have shown that although HSCT using UCB as the source requires less stringent HLA loci matching, it has slower time of engraftment and higher graft failure compared to the BM group.170,176 The advantages and disadvantages of these sources for HSCT are summarized in Table 8.26,156

1.23.6.7.3

Allogeneic Versus Autologous HSCT

Autologous HSCT offers the advantage of not causing the potentially lethal GVHD. The major complication with allogeneic HSCT is the possible of occurrence of GVHD; however, they contain cells that may initiate graft-versus-tumor (GVT) or graft-versus-leukemia (GVL) response.154 In contrast, the absence of GVT/GVL response in autologous HSCT increases the likelihood of relapse of the disease due to presence of tumor cells in the graft.174 Due to these complications, various strategies have been employed to create a better HSC graft including eradication of GVHD-causing cells, removal of relapse-causing cells, and addition of cells to increase the efficacy of graft function and expansion of donor cells when there is insufficient cell dose.158 With regards to preparative regimens used in HSCT, clinical advances in HSCT have allowed the use of nonmyeloablative preparative regimens that are safer than myeloablative regimens. The former does not compromise patient’s safety and necessitate a prolonged hospital stay under sterile conditions.173 The diseases commonly treated with HSCT are listed in Table 9.12

1.23.6.8

Conclusion

Huge advances in transplantation have been made since the discovery of HSCs. Over the past decade, HSCT has become an increasingly favorable treatment modality as it offers a good chance for a cure for many diseases of the hematopoietic and immune system. In spite of this, more work is required to improve the efficacy of grafts and preparative regimens, and to lower the risk of transplant complications including GVHD. The discovery of hESC and hiPSC as sources of HSCs opens up an interesting and attractive avenue for research work and can potentially be a viable source of HSCs in the near future. Manipulation and

Stem Cells

351

bioengineering of patient-specific hESC and hiPSC could potentially advance HSCT without the need for acquisition of HSCs, HLAmatching and preparative regimes for cell therapy.

References 1. Aboody, K. S.; Brown, A.; Rainov, N. G.; et al. Neural Stem Cells Display Extensive Tropism for Pathology in Adult Brain: Evidence From Intracranial Gliomas. Proc. Natl. Acad. Sci. 2000, 97, 12846–12857. 2. Adewumi, O.; Aflatoonian, B.; Ahrlund-Richter, L.; et al. Characterization of Human Embryonic Stem Cell Lines by the International Stem Cell Initiative. Nat. Biotechnol. 2007, 25, 803–816. 3. Azuara, V.; Perry, P.; Sauer, S.; et al. Chromatin Signatures of Pluripotent Cell Lines. Nat. Cell Biol. 2006, 8, 532–538. 4. Barberi, T.; Willis, L. M.; Socci, N. D.; Studer, L. Derivation of Multipotent Mesenchymal Precursors From Human Embryonic Stem Cells. PLoS Med. 2005, 2, 161. 5. Barry, F. P.; Murphy, J. M. Mesenchymal Stem Cells: Clinical Applications and Biological Characterization. Int. J. Biochem. Cell Biol. 2004, 36, 568–584. 6. Baum, C. M.; Weissman, I. L.; Tsukamoto, A. S.; et al. Isolation of a Candidate Human Hematopoietic Stem Cell Population. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 2804–2808. 7. Becker, A. J.; McCulloch, E. A.; Till, J. E. Cytological Demonstration of the Clonal Nature of Spleen Colonies Derived From Transplanted Mouse Marrow Cells. Nature 1963, 197, 452–454. 8. Campagnoli, C.; Roberts, I. A.; Kumar, S.; et al. Identification of Mesenchymal Stem/progenitor Cells in Human First-Trimester Fetal Blood, Liver, and Bone Marrow. Blood 2001, 98, 2396–2402. 9. Caplan, A. L. The Mesengenic Process. Clin. Plast. Surg. 1994, 21, 429–435. 10. Charbord, P.; Tavian, M.; Humeau, L.; Peault, B. Early Ontogeny of the Human Marrow From Long Bones: An Immunohistochemical Study of Hematopoiesis and Its Microenvironment. Blood 1996, 87, 4109–4119. 11. Chung, Y.; Klimanskaya, I.; Becker, S.; et al. Human Embryonic Stem Cell Lines Generated Without Embryo Destruction. Cell Stem Cell 2007, 2, 113–117. 12. Copelan, E. A. Hematopoietic Stem-Cell Transplantation. N. Engl. J. Med. 2006, 354, 1813–1826. 13. Crisan, M.; Yap, S.; Casteilla, L.; et al. A Perivascular Origin for Mesenchymal Stem Cells in Multiple Human Organs. Cell Stem Cell 2008, 3, 301–313. 14. Da Silva, M. L.; Caplan, A. L.; Nardi, N. B. In Search of the In Vivo Identity of Mesenchymal Stem Cells. Stem Cells 2008, 26, 2287–2299. 15. De Bari, C.; Dell’ Accio, F.; Tylzanowski, P.; Luyten, F. P. Multipotent Mesenchymal Stem Cells From Adult Human Synovial Membrane. Arthritis Rheum. 2001, 44, 1928–1942. 16. Deleyrolle, L.; Reynolds, B. A. Isolation, Expansion, and Differentiation of Adult Mammalian Neural Stem and Progenitor Cells Using the Neurosphere Assay. Methods Mol. Biol. 2009, 549, 91–101. 17. Digirolamo, C. M.; Stokes, D.; Colter, D.; et al. Propagation and Senescence of Human Marrow Stromal Cells in Culture: A Simple Colony-Forming Assay Identifies Samples With the Greatest Potential to Propagate and Differentiate. Br. J. Haematol. 1999, 107, 275–281. 18. Ehtesham, M.; Yuan, X.; Kabos, P.; et al. Glioma Tropic Neural Stem Cells Consist of Astrocytic Precursors and Their Migratory Capacity Is Mediated by CXCR4. Neoplasia 2004, 6, 287–293. 19. Erices, A.; Conget, P.; Minguell, J. J. Mesenchymal Progenitor Cells in Human Umbilical Cord Blood. Br. J. Haematol. 2000, 109, 235–242. 20. Feng, Q.; Lu, S. J.; Klimanskaya, I.; et al. Hemangioblastic Derivatives From Human Induced Pluripotent Stem Cells Exhibit Limited Expansion and Early Senescence. Stem Cells 2010, 28, 704–712. 21. Flax, J. D.; Aurora, S.; Yang, C.; et al. Engraftable Human Neural Stem Cells Respond to Developmental Cues, Replace Neurons, and Express Foreign Genes. Nat. Biotechnol. 1998, 16, 1033–1039. 22. Gage, F. H.; Ray, J.; Fisher, L. J. Isolation, Characterization, and Use of Stem Cells From the CNS. Annu. Rev. Neurosci. 1995, 18, 159–192. 23. Galy, A.; Travis, M.; Cen, D.; Chen, B. Human T, B, Natural Killer, and Dendritic Cells Arise From a Common Bone Marrow Progenitor Cell Subset. Immunity 1995, 3, 459–473. 24. Gatti, R. A.; Meuwissen, H. J.; Allen, H. D.; et al. Immunological Reconstitution of Sex-Linked Lymphopenic Immunological Deficiency. Lancet 1968, 2, 1366–1369. 25. Gluckman, E.; Broxmeyer, H. A.; Auerbach, A. D. Hematopoietic Reconstitution in a Patient With Fanconi’s Anemia by Means of Umbilical-Cord Blood From an HLA-Identical Sibling. N. Engl. J. Med. 1989, 321, 1174–1178. 26. Gluckman, E. Current Status of Umbilical Cord Blood Hematopoietic Stem Cell Transplantation. Exp. Hematol. 2000, 28, 1197–1205. 27. Grigoriadis, A. E.; Kennedy, M.; Bozec, A.; et al. Directed Differentiation of Hematopoietic Precursors and Functional Osteoclasts From Human ES and IPS Cells. Blood 2010, 115 (14), 2769–2776. Epub 2010 Jan 11. 28. Liu, H.; Ye, Z.; Kim, Y.; Sharkis, S.; Jang, Y. Y. Generation of Endoderm-Derived Human Induced Pluripotent Stem Cells From Primary Hepatocytes. Hepatology 2010, 51, 1810–1819. 29. Jankowski, R. J.; Deasy, B. M.; Huard, J. Muscle-Derived Stem Cells. Gene Ther. 2002, 9, 642–647. 30. Johansson, C. B.; Momma, S.; Clarke, D. L.; et al. Identification of a Neural Stem Cell in the Adult Mammalian Central Nervous System. Cell 1999, 96, 25–34. 31. Kebriaei, P.; Isola, L.; Bahceci, E.; et al. Adult Human Mesenchymal Stem Cells Added to Corticosteroids Therapy for the Treatment of Acute Graft-Versus-Host Disease. Biol. Blood Marrow Transplant. 2009, 15, 804–811. 32. Kikushige, Y.; Yoshimoto, G.; Miyamoto, T.; et al. Human Flt3 Is Expressed at the Hematopoietic Stem Cell and the Granulocyte/macrophage Progenitor Stages to Maintain Cell Survival. J. Immunol. 2008, 180, 7358–7367. 33. Körbling, M.; Anderlini, P. Peripheral Blood Stem Cell Versus Bone Marrow Allotransplantation: Does the Source of Hematopoietic Stem Cells Matter? Blood 2001, 98, 2900–2908. 34. Knudtzon, S. In Vitro Growth of Granulocytic Colonies From Circulating Cells in Human Cord Blood. Blood 1974, 43, 357–361. 35. Kukekov, V. G.; Laywell, E. D.; Suslov, O.; et al. Multipotent Stem/progenitor Cells With Similar Properties Arise From Two Neurogenic Regions of Adult Human Brain. Exp Neurol 1999, 156, 333–344. 36. Landry, D. W.; Zucker, H. A. Embryonic Death and the Creation of Human Embryonic Stem Cells. J. Clin. Investig. 2004, 114, 1184–1186. 37. Le Blanc, K.; Götherström, C.; Ringdén, O.; et al. Fetal Mesenchymal Stem-Cell Engraftment in Bone After in Utero Transplantation in a Patient With Severe Osteogenesis Imperfecta. Transplantation 2005, 79, 1607–1614. 38. Lee, A.; Kessler, J. D.; Read, T. A.; et al. Isolation of Neural Stem Cells From the Postnatal Cerebellum. Nat. Neurosci. 2005, 8, 723–729. 39. Li, L.; Xie, T. Stem Cell Niche: Structure and Function. Annu. Rev. Cell Dev. Biol. 2005, 21, 605–631. 40. Manz, M. G.; Miyamoto, T.; Akashi, K. Prospective Isolation of Human Clonogenic Common Myeloid Progenitors. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 11872–11877. 41. Marx, J. C.; Gordon, P. L.; Horwitz, E. M.; et al. Isolated Allogeneic Bone Marrow-Derived Mesenchymal Cells Engraft and Stimulate Growth in Children With Osteogenesis Imperfecta: Implications for Cell Therapy of Bone. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 8932–8937. 42. Müller, F. J.; Snyder, E. Y.; Loring, J. F. Gene Therapy: Can Neural Stem Cells Deliver? Nat. Rev. Neurosci. 2006, 7, 75–84. 43. Nait-Oumesmar, B.; Decker, L.; Lachapelle, F.; et al. Progenitor Cells of the Adult Mouse Subventricular Zone Proliferate, Migrate and Differentiate into Oligodendrocytes after Demyelination. Eur. J. Neurosci. 2008, 11, 4357–4366.

352 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96.

Stem Cells Nakahara, H.; Dennis, J. E.; Bruder, S. P.; et al. In Vitro Differentiation of Bone and Hypertrophic Cartilage From Periosteal-Derived Cells. Exp. Cell Res. 1991, 195, 492–503. Nauta, A. J.; Fibbe, W. E. Immunomodulatory Properties of Mesenchymal Stromal Cells. Blood 2007, 110, 3499–3506. Park, I. H.; Arora, N.; Huo, H.; et al. Disease-Specific Induced Pluripotent Stem Cells. Cell 2008, 134, 877–886. Park, I. H.; Zhao, R.; West, J. A.; et al. Reprogramming of Human Somatic Cells to Pluripotency by Defined Factors. Nature 2008, 451, 141–146. Pittenger, M. F.; Mackay, A. M.; Beck, S. C.; et al. Multilineage Potential of Adult Human Mesenchymal Stem Cells. Science 1999, 284, 143–147. President’s Council on Bioethics. White Paper: Alternative Sources of Pluripotent Stem Cells, Presidernt’s Council on Bioethics: Washington, DC, 2005. Prockop, D. J. Marrow Stromal Cells as Stem Cells for Nonhematopoietic Tissues. Science 1997, 276, 71–74. Reynolds, B. A.; Weiss, S. Generation of Neurons and Astrocytes From Isolated Cells of the Adult Mammalian Central Nervous System. Science 1992, 255, 1707–1710. Schofield, R. The Relationship Between the Spleen Colony-Forming Cell and the Haematopoietic Stem Cell. Blood Cells 1978, 4, 7–25. Schwartz, P. H.; Bryant, P. J.; Fuja, T. J.; et al. Isolation and Characterization of Neural Progenitor Cells From Post-mortem Human Cortex. J. Neurosci. Res. 2003, 74, 838–851. Shen, Q.; Goderie, S. K.; Jin, L.; et al. Endothelial Cells Stimulate Self-Renewal and Expand Neurogenesis of Neural Stem Cells. Science 2004, 304, 1338–1340. Simmons, P. J.; Torok-Storb, B. Identification of Stromal Cell Precursors in Human Bone Marrow by a Novel Monoclonal Antibody, STRO-1. Blood 1991, 78, 55–62. Taichman, R. S.; Emerson, S. G. The Role of Osteoblasts in the Hematopoietic Microenvironment. Stem Cells 1998, 16, 7–15. Takahashi, K.; Yamanaka, S. Induction of Pluripotent Stem Cells From Mouse Embryonic and Adult Fibroblast Cultures by Defined Factors. Cell 2006, 126, 663–676. Takahashi, K.; Tanabe, K.; Ohnuki, M.; et al. Induction of Pluripotent Stem Cells From Adult Human Fibroblasts by Defined Factors. Cell 2007, 131, 861–872. Thomson, J. A.; Itskovitz-Eldor, J.; Shapiro, S. S.; et al. Embryonic Stem Cell Lines Derived From Human Blastocysts. Science 1998, 282, 1145–1147. Tuli, R.; Seghatoleslami, M. R.; Tuli, S.; et al. A Simple, High-Yield Method for Obtaining Multipotential Mesenchymal Progenitor Cells From Trabecular Bone. Mol. Biotechnol. 2003, 23, 37–49. Varnum-Finney, B.; Xu, L.; Brashem-Stein, C.; et al. Pluripotent, Cytokine-Dependent, Hematopoietic Stem Cells Are Immortalized by Constitutive Notch1 Signaling. Nat. Med. 2000, 6, 1278–1281. Vierbuchen, T.; Ostermeier, A.; Pang, Z. P.; et al. Direct Conversion of Fibroblasts to Functional Neurons by Defined Factors. Nature 2010, 463, 1035–1041. Warner-Schmidt, J. L.; Duman, R. S. Hippocampal Neurogenesis: Opposing Effects of Stress and Antidepressant Treatment. Hippocampus 2006, 16, 239–249. Wobus, A. M.; Boheler, K. R. Embryonic Stem Cells: Prospects for Developmental Biology and Cell Therapy. Physiol. Rev. 2005, 85, 635–678. Woltjen, K.; Michael, I. P.; Mohseni, P.; et al. piggyBac Transposition Reprograms Fibroblasts to Induced Pluripotent Stem Cells. Nature 2009, 458, 766–770. Yahata, T.; Muguruma, Y.; Yumino, S.; et al. Quiescent Human Hematopoietic Stem Cells in the Bone Marrow Niches Organize the Hierarchical Structure of Hematopoiesis. Stem Cells 2008, 28, 3228–3236. Yin, A. H.; Miraglia, S.; Zanjani, E. D.; et al. AC133, a Novel Marker for Human Hematopoietic Stem and Progenitor Cells. Blood 1997, 90, 5002–5012. Yin, T.; Li, L. The Stem Cell Niches in Bone. J. Clin. Investig. 2006, 116, 1195–1201. Young, K. M.; Merson, T. D.; Sotthibundhu, A.; et al. p75 Neurotrophin Receptor Expression Defines a Population of BDNF-Responsive Neurogenic Precursor Cells. J. Neurosci. 2007, 27, 5146–5155. Yu, J.; Vodyanik, M. A.; Smuga-Otto, K.; et al. Induced Pluripotent Stem Cell Lines Derived From Human Somatic Cells. Science 2007, 318, 1917–1920. Zuk, P. A.; Zhu, M.; Ashjian, P.; et al. Human Adipose Tissue Is a Source of Multipotent Stem Cells. Mol. Biol. Cell 2002, 13, 4279–4295. Allegrucci, C.; Young, L. E. Differences between Human Embryonic Stem Cell Lines. Hum. Reprod. Update 2007, 13, 103–120. Banin, E.; Obolensky, A.; Idelson, M.; et al. Retinal Incorporation and Differentiation of Neural Precursors Derived From Human Embryonic Stem Cells. Stem Cells 2006, 24, 246–257. Carpenter, M. K.; Rosler, E. S.; Fisk, G. J.; et al. Properties of Four Human Embryonic Stem Cell Lines Maintained in a Feeder-Free Culture System. Dev. Dyn. 2004, 229, 243–258. D’Amour, K. A.; Agulnick, A. D.; Eliazer, S.; et al. Efficient Differentiation of Human Embryonic Stem Cells to Definitive Endoderm. Nat. Biotechnol. 2005, 23, 1534–1541. Da Cruz, L.; Chen, F. K.; Ahmado, A.; Greenwood, J.; Coffey, P. RPE Transplantation and its Role in Retinal Disease. Prog. Retin. Eye Res. 2007, 26, 598–635. Hoffman, L. M.; Carpenter, M. K. Characterization and Culture of Human Embryonic Stem Cells. Nat. Biotechnol. 2005, 23, 699–708. Kehat, I.; Gepstein, L. Derivation and Properties of Human Embryonic Stem Cell-Derived Cardiomyocytes. Gene Ther. Regul. 2002, 1, 387–398. Keirstead, H. S.; Nistor, G.; Bernal, G.; et al. Human Embryonic Stem Cell-Derived Oligodendrocyte Progenitor Cell Transplants Remyelinate and Restore Locomotion After Spinal Cord Injury. J. Neurosci. 2005, 25, 4694–4705. Klimanskaya, I.; Chung, Y.; Becker, S.; et al. Human Embryonic Stem Cell Lines Derived From Single Blastomeres. Nature 2006, 444, 481–485. Klimanskaya, I.; Hipp, J.; Rezai, K. A.; et al. Derivation and Comparative Assessment of Retinal Pigment Epithelium From Human Embryonic Stem Cells Using Transcriptomics. Cloning Stem Cells 2004, 6, 217–245. Kroon, E.; Martinson, L. A.; Ksdoya, K.; et al. Pancreatic Endoderm Derived From Human Embryonic Stem Cells Generated Glucose-Responsive Insulin-Secreting Cells In Vivo. Nat. Biotechnol. 2008, 26, 443–452. Meyer, J. S.; Katz, M. L.; Maruniak, J. A.; Kirk, M. D. Neural Differentiation of Mouse Embryonic Stem Cells In Vitro and After Transplantation into Eyes of Mutant Mice With Rapid Retinal Degeneration. Brain Res. 2004, 1014, 131–144. Meyer, J. S.; Katz, M. L.; Maruniak, J. A.; Kirk, M. D. Embryonic Stem Cell-Derived Neural Progenitors Incorporate into Degenerating Retina and Enhance Survival of Host Photoreceptors. Stem Cells 2006, 24, 274–283. Rho, J. Y.; Yu, K.; Han, J. S.; et al. Transcriptional Profiling of the Developmentally Important Signalling Pathways in Human Embryonic Stem Cells. Hum. Reprod. 2006, 21, 405–412. Rogers, N. T.; Hobson, E.; Pickering, S.; et al. Phospholipase Czeta Causes Ca2þ Oscillations and Parthenogenetic Activation of Human Oocytes. Reproduction 2004, 128, 697–702. Skottman, H.; Mikkola, M.; Lundin, K.; et al. Gene Expression Signatures of Seven Individual Human Embryonic Stem Cell Lines. Stem Cells 2005, 23, 1343–1356. Zhang, S.-C.; Wernig, M.; Duncan, I. D.; et al. In Vitro Differentiation of Transplantable Neural Precursors From Human Embryonic Stem Cells. Nat. Biotechnol. 2001, 19, 1129–1133. Aasen, T.; Raya, A.; Barrero, M. J.; et al. Efficient and Rapid Generation of Induced Pluripotent Stem Cells From Human Keratinocytes. Nat. Biotechnol. 2008, 26, 1276–1284. Agarwal, S.; Yuin-Han, L.; McLoughlin, E. M.; et al. Telomere Elongation in Induced Pluripotent Stem Cells From Dyskeratosis Congenita Patients. Nature 2010, 464, 292–296. Bernstein, B. E.; Mikkelsen, T. S.; Xie, X.; et al. A Bivalent Chromatin Structure Marks Key Developmental Genes in Embryonic Stem Cells. Cell 2006, 125, 315–326. Chen, Y.; Shi, L.; Zhang, L.; et al. The Molecular Mechanism Governing the Oncogenic Potential of Sox2 in Breast Cancer. J. Biol. Chem. 2008, 283, 17969–17978. Chen, Z. Y.; He, C. Y.; Kay, M. A. Improved Production and Purification of Minicircle DNA Vector Free of Plasmid Bacterial Sequences and Capable of Persistent Transgene Expression In Vivo. Hum. Gene Ther. 2005, 16, 126–131. Choi, K. D.; Yu, J.; Smuga-Otto, K.; et al. Hematopoietic and Endothelial Differentiation of Human Induced Pluripotent Stem Cells. Stem Cells 2009, 27, 559–567. Dimos, J. T.; Rodolfa, K. T.; Niakan, K. K.; et al. Induced Pluripotent Stem Cells Generated From Patients With ALS Can Be Differentiated into Motor Neurons. Science 2008, 321, 1218–1221. Ebert, A. D.; Yu, J.; Rose, F. F., Jr.; et al. Induced Pluripotent Stem Cells From a Spinal Muscular Atrophy Patient. Nature 2009, 457, 277–280.

Stem Cells

353

97. Fraser, M. J.; Ciszczon, T.; Elick, T.; Bauser, C. Precise Excision of TTAA-Specific Lepidopteran Transposons PiggyBac (IFP2) and Tagalong (TFP3) From the Baculovirus Genome in Cell Lines From Two Species of Lepidoptera. Insect Mol. Biol. 1996, 5, 141–151. 98. Gertow, K.; Przyborski, S.; Loring, J.; et al. Isolation of Human Embryonic Stem Cell-Derived Teratomas for the Assessment of Pluripotency. Curr. Protocol Stem Cell Biol. 2007. U1.B4. https://doi.org/10.1002/9780470151808.sc01b04s3. 99. Giorgetti, A.; Montserrat, N.; Rodriguez-Piza, I.; et al. Generation of Induced Pluripotent Stem Cells From Human Cord Blood Cells With Only Two Factors: Oct4 and Sox2. Nat. Protoc. 2010, 5, 811–820. 100. Greider, C. W.; Blackburn, E. H. Identification of a Specific Telomere Terminal Transferase Activity in Tetrahymena Extracts. Cell 1985, 43, 405–413. 101. Harley, C. B.; Futcher, A. B.; Greider, C. W. Telomeres Shorten during Ageing of Human Fibroblasts. Nature 1990, 345, 458–460. 102. Heng, J. D.; Feng, B.; Han, J.; et al. The Nuclear Receptor Nr5a2 Can Replace Oct4 in the Reprogramming of Murine Somatic Cells to Pluripotent Cells. Cell 2009, 6, 167–174. 103. Hirami, Y.; Osakada, F.; Takahashi, K.; et al. Generation of Retinal Cells From Mouse and Human Induced Pluripotent Stem Cells. Neurosci. Lett. 2009, 458, 126–131. 104. Hu, B. Y.; Weick, J. P.; Yu, J.; et al. Neural Differentiation of Human Induced Pluripotent Stem Cells Follows Developmental Principles But With Variable Potency. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 4335–4340. 105. Jia, F.; Wilson, K. D.; Sun, N.; et al. A Nonviral Minicircle Vector for Deriving Human IPS Cells. Nat. Methods 2010, 7, 197–199. 106. Kehat, I.; Kenyagin-Karsenti, D.; Snir, M.; et al. Human Embryonic Stem Cells Can Differentiate into Myocytes With Structural and Functional Properties of Cardiomyocytes. J. Clin. Investig. 2001, 108, 407–414. 107. Kim, D.; Kim, C. H.; Moon, J. I.; et al. Generation of Human Induced Pluripotent Stem Cells by Direct Delivery of Reprogramming Proteins. Cell Stem Cell 2009, 4, 472–476. 108. Kim, J. B.; Greber, B.; Araúzo-Bravo, M. J.; et al. Direct Reprogramming of Human Neural Stem Cells by Oct4. Nature 2009, 461, 649–653. 109. Lee, S. H.; Lumelsky, N.; Studer, L.; et al. Efficient Generation of Midbrain and Hindbrain Neurons From Mouse Embryonic Stem Cells. Nat. Biotechnol. 2000, 18, 675–679. 110. Lei, H.; Oh, S. P.; Okano, M.; et al. De Novo DNA Cytosine Methyltransferase Activities in Mouse Embryonic Stem Cells. Development 1996, 122, 3195–3205. 111. Li, C.; Zhou, J.; Shi, G.; et al. Pluripotency Can Be Rapidly and Efficiently Induced in Human Amniotic Fluid-Derived Cells. Hum. Mol. Genet. 2009, 18, 4340–4349. 112. Li, W.; Wang, D.; Qin, J.; et al. Generation of Functional Hepatocytes From Mouse Induced Pluripotent Stem Cells. J. Cell. Physiol. 2010, 222, 492–501. 113. Li, Y.; Zhao, H.; Lan, F.; et al. Generation of Human-Induced Pluripotent Stem Cells From Gut Mesentery-Derived Cells by Ectopic Expression of Oct4/Sox2/Nanog. Cell. Reprogram. 2010, 12 (3), 237–247. 114. Lumelsky, N.; Blondel, O.; Laeng, P.; et al. Differentiation of Embryonic Stem Cells to Insulin-Secreting Structures Similar to Pancreatic Islets. Science 2001, 292, 1389–1394. 115. Maherali, N.; Sridharan, R.; Xie, W.; et al. Directly Reprogrammed Fibroblasts Show Global Epigenetic Remodeling and Widespread Tissue Contribution. Cell Stem Cell 2007, 1, 55–70. 116. Narazaki, G.; Uosaki, H.; Teranishi, M.; et al. Directed and Systematic Differentiation of Cardiovascular Cells From Mouse Induced Pluripotent Stem Cells. Circulation 2008, 118, 498–506. 117. Meissner, A.; Wernig, M.; Jaenisch, R. Direct Reprogramming of Genetically Unmodified Fibroblasts into Pluripotent Stem Cells. Nat. Biotechnol. 2007, 25, 1177–1181. 118. Okita, K.; Nakagawa, M.; Hyenjong, H.; et al. Generation of Mouse Induced Pluripotent Stem Cells Without Viral Vectors. Science 2008, 322, 948–953. 119. Ponzielli, R.; Katz, S.; Barsyte-Lovejoy, D.; Penn, L. Z. Cancer Therapeutics: Targeting the Dark Side of Myc. Eur. J. Cancer 2005, 41, 2485–2501. 120. Rowland, B. D.; Peeper, D. S. KLF4, P21 and Context-Dependent Opposing Forces in Cancer. Nat. Rev. Cancer 2006, 6, 11–23. 121. Shin, S.; Dalton, S.; Stice, S. Human Motor Neuron Differentiation From Human Embryonic Stem Cell. Stem Cells Dev. 2005, 14, 1–4. 122. Sun, N.; Panetta, N. J.; Gupta, D. M.; et al. Feeder-Free Derivation of Induced Pluripotent Stem Cells From Adult Human Adipose Stem Cells. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 15720–15725. 123. Svetlana, N. R.; Paramjeet, K. R.; Victoria, L. B. In Vitro Differentiation of Mouse Embryonic Stem Cells into Primitive Blood Vessels. Methods Enzymol. 2008, 443, 103–117. 124. Tateishi, K.; He, J.; Taranova, O.; et al. Generation of Insulin-Secreting Islet-Like Clusters From Human Skin Fibroblasts. J. Biol. Chem. 2008, 283, 31601–31607. 125. Thompson, S.; Clarke, A. R.; Pow, A. M.; et al. Germ Line Transmission and Expression of a Corrected HPRT Gene Produced by Gene Targeting in Embryoni Stem Cells. Cell 1989, 56, 313–321. 126. Trosko, J. E. From Adult Stem Cells to Cancer Stem Cells: Oct-4 Gene, Cell–cell Communication, and Hormones During Tumor Promotion. Ann. N.Y. Acad. Sci. 2006, 1089, 36–58. 127. Zhang, D.; Jiang, W.; Liu, M.; et al. Highly Efficient Differentiation of Human ES Cells and IPS Cells into Mature Pancreatic Insulin Producing Cells. Cell Res. 2009, 19, 429–438. 128. Zhang, J.; Wilson, G. F.; Soerens, A. G.; et al. Functional Cardiomyocytes Derived From Human Induced Pluripotent Stem Cells. Circ Res. 2009, 104, 30–41. 129. Zhao, Y.; Yin, X.; Qin, H.; et al. Two Supporting Factors Greatly Improve the Efficiency of Human IPSC Generation. Cell Stem Cell 2008, 3, 475–479. 130. Zhou, W.; Freed, C. R. Adenoviral Gene Delivery Can Reprogram Human Fibroblasts to Induced Pluripotent Stem Cells. Stem Cells 2009, 27, 2667–2674. 131. Encinas, J. M.; Vaahtokari, A.; Enikolopov, G. Fluoxetine Targets Early Progenitor Cells in the Adult Brain. Proc. Natl. Acad. Sci. 2006, 103, 8233–8238. 132. Elkabetz, Y.; Panagiotakos, G.; Shamy, G. A.; et al. Human ES Cell-Derived Neural Rosettes Reveal a Functionally Distinct Early Neural Stem Cell Stage. Genes Dev. 2007, 22, 152–165. 133. Malberg, J. E.; Duman, R. S. Cell Proliferation in Adult hippocampus Is Decreased by Inescapable Stress: Reversal by Fluoxetine Treatment. Neuropsychopharmacology 2003, 28, 1562–1571. 134. Malberg, J. E.; Eisch, A. J.; Nestler, E. J.; Duman, R. S. Chronic Antidepressant Treatment Increases Neurogenesis in Adult Rat hippocampus. J. Neurosci. 2000, 20, 9104–9110. 135. Maslov, A. Y.; Barone, T. A.; Plunkett, R. J.; Pruitt, S. C. Neural Stem Cell Detection, Characterization, and Age-Related Changes in the Subventricular Zone of Mice. J. Neurosci. 2004, 24, 1726–1733. 136. Mercier, F.; Kitasako, J. T.; Hatton, G. I. Anatomy of the Brain Neurogenic Zones Revisited: Fractones and the Fibroblast/macrophage Network. J. Comp. Neurol. 2002, 451, 170–188. 137. Pevny, L. H.; Nicolis, S. K. Sox2 Roles in Neural Stem Cells. Int. J. Biochem. Cell Biol. 2010, 42, 421–424. 138. Trounson, A. New Perspectives in Human Stem Cell Therapeutic Research. BMC Med. 2009, 7, 29. 139. Wada, K.; Nakajima, A.; Katayama, K.; et al. Peroxisome Proliferator-Activated Receptor g-Mediated Regulation of Neural Stem Cell Proliferation and Differentiation. J. Biol. Chem. 2006, 281, 12673–12681. 140. Xu, R.; Wu, C.; Tao, Y.; et al. Nestin-Positive Cells in the Spinal Cord: A Potential Source of Neural Stem Cells. Neuroscience 2008, 26, 813–820. 141. Gordon, P. L.; Horwitz, E. M.; Prockop, D. J.; et al. Clinical Responses to Bone Marrow Transplantation in Children With Severe Osteogenesis Imperfecta. Blood 2001, 97, 1227–1231. 142. Horwitz, E. M.; Prockop, D. J.; Gordon, P. L.; et al. Transplantability and Therapeutic Effects of Bone Marrow-Derived Mesenchymal Cells in Children With Osteogenesis Imperfecta. Nat. Med. 1999, 5, 309–313. 143. Koç, O. N.; Day, J.; Nieder, M.; et al. Allogeneic Mesenchymal Stem Cell Infusion for Treatment of Metachromatic Leukodystrophy (MLD) and Hurler Syndrome (MPS-IH). Bone Marrow Transplant. 2002, 30, 215–222. 144. Onken, J.; Gallup, D.; Hanson, J.; et al. Successful Outpatient Treatment of Refractory Crohn’s Disease Using Adult Mesenchymal Stem Cells. In Paper Presented at the American College of Gastroenterology Conference; 2006. 2006. 145. Orlic, D.; Kajstura, J.; Chimenti, S.; et al. Bone Marrow Cells Regenerate Infracted Myocardium. Nature 2001, 410, 701–705.

354

Stem Cells

146. Pochampally, R. Colony forming Unit Assays for MSCs. Methods Mol. Biol. 2008, 449, 83–91. 147. Sato, Y.; Araki, H.; Kato, J.; et al. Human Mesenchymal Stem Cells Xenografted Directly to Rat Liver Are Differentiated into Human Hepatocytes Without Fusion. Blood 2005, 106, 756–763. 148. Strauer, B. E.; Brehm, M.; Zeus, T.; et al. Repair of Infarcted Myocardium by Autologous Intracoronary Mononuclear Bone Marrow Cell Transplantation in Humans. Circulation 2002, 106, 1913–1918. 149. Wakitani, S.; Imoto, K.; Yamamoto, T.; et al. Human Autologous Culture Expanded Bone Marrow Mesenchymal Cell Transplantation for Repair of Cartilage Defects in Osteoarthritic Knees. Osteoarthritis Cartilage 2002, 10, 199–206. 150. Bhardwaj, G.; Murdoch, B.; Wu, D.; et al. Sonic Hedgehog Induces the Proliferation of Primitive Human Hematopoietic Cells via BMP Regulation. Nat. Immunol. 2001, 2, 172–180. 151. Couban, S.; Simpson, D. R.; Barnett, M. J.; et al. A Randomized Multicenter Comparison of Bone Marrow and Peripheral Blood in Recipients of Matched Sibling Allogeneic Transplants for Myeloid Malignancies. Blood 2002, 100, 1525–1531. 152. Cutler, C.; Antin, J. H. Peripheral Blood Stem Cells for Allogeneic Transplantation: A Review. Stem Cells 2001, 19, 108–117. 153. Edvardsson, L.; Dykes, J.; Olofsson, T. Isolation and Characterisation of Human Myeloid Progenitor Populations-TpoR as Discriminator Between Common Myeloid and Egakaryocyte/Erythroid Progenitors. Exp. Hematol. 2006, 34, 599–609. 154. Ferrara, J. L.; Levine, J. E.; Reddy, P.; Holler, E. Graft-Versus-Host Disease. Lancet 2009, 373, 1550–1561. 155. Fritsch, G.; Buchinger, P.; Printz, D.; et al. Rapid Discrimination of Early CD34þ Myeloid Progenitors Using CD45-RA Analysis. Blood 1993, 81, 2301–2309. 156. Haspel, R. L.; Miller, K. B. Hematopoietic Stem Cells: Sources Matter. Curr. Stem Cell Res. Ther. 2008, 3, 229–236. 157. Huang, S.; Terstappen, L. W. M. M. Lymphoid and Myeloid Differentiation of Single Human CD34þ, HLA-DRþ, CD38 Hematopoietic Stem Cells. Blood 1994, 83, 1515–1526. 158. Hwang, W. Y. K. Haematopoietic Graft Engineering. Ann. Acad. Med. Singapore 2004, 33, 551–558. 159. Hwang, W. Y. K.; Ong, S. Y. Allogeneic Haematopoietic Stem Cell Transplantation Without a Matched Sibling Donor: Current Options and Future Potential. Ann. Acad. Med. Singapore 2009, 38, 340–345. 160. Iwasaki, H.; Akashi, K. Hematopoietic Developmental Pathways: On Cellular Basis. Nat. Oncogene 2007, 26, 6687–6696. 161. Karanu, F. N.; Murdoch, B.; Gallacher, L.; et al. The Notch Ligand Jagged-1 Represents a Novel Growth Factor of Human Hematopoietic Stem Cells. J. Exp. Med. 2000, 192, 1365–1372. 162. Kondo, M.; Weissman, I. L.; Akashi, K. Identification of Clonogenic Common Lymphoid Progenitors in Mouse Bone Marrow. Cell 1997, 91, 661–672. 163. Körbling, M.; Burke, P.; Braine, H.; et al. Successful Engraftment of Blood-Derived Normal Hematopoietic Stem Cells in Chronic Myelogenous Leukemia. Exp. Hematol. 1981, 9, 684–690. 164. Lee, S. J.; Klein, J.; Haagenson, M.; et al. High-Resolution Donor–recipient HLA Matching Contributes to the Success of Unrelated Donor Marrow Transplantation. Blood 2007, 110, 4576–4583. 165. Lorenz, E.; Uphoff, E. D.; Reid, T. R.; Shelton, E. Modification of Acute Irradiation Injury in Mice and Guinea Pigs by Bone Marrow Injection. Radiology 1951, 58, 863–877. 166. Main, J. M.; Prehn, R. T. Successful Skin Homografts after the Administration of High Dosage X Radiation and Homologous Bone Marrow. J. Natl. Cancer Inst. 1955, 15, 1023–1029. 167. Pasquini, M. C.; Wang, Z. Current Use and Outcome of Hematopoietic Stem Cell Transplantation: Part I – CIBMTR Summary Slides. Center for International Blood and Marrow Transplant Research Newsletter 2009, 15 (1), 7–11 [serial online]. 168. Rafii, S.; Shapiro, F.; Pettengell, R.; et al. Human Bone Marrow Microvascular Endothelial Cells Support Long-Term Proliferation and Differentiation of Myeloid and Megakaryocytic Progenitors. Blood 1995, 86, 3353–3363. 169. Rocha, V.; Labopin, M.; Sanz, G.; et al. Transplants of Umbilical-Cord Blood or Bone Marrow From Unrelated Donors in Adults With Acute Leukemia. N. Engl. J. Med. 2004, 351, 2276–2285. 170. Sharma, S.; Gurudutta, G. U.; Satija, N. K.; et al. Stem Cell C-KIT and HOXB4 Genes: Critical Roles and Mechanisms in Self-Renewal, Proliferation and Differentiation. Stem Cells Dev. 2006, 15, 755–778. 171. Siminovitch, L.; McCulloch, E. A.; Till, J. E. The Distribution of Colony-Forming Cells Among Spleen Colonies. J. Cell Comp. Physiol. 1963, 62, 327–336. 172. Spitzer, T. R. The Expanding Applications of Non-Myeloablative Stem Cell Transplantation. Pediatr. Transplant. 2003, 7, 95–100. 173. Tabbara, I. A.; Zimmerman, K.; Morgan, C.; Nahleh, Z. Allogeneic Hematopoietic Stem Cell Transplantation: Complications and Results. Arch. Intern. Med. 2002, 162, 1558–1566. 174. Taichman, R. S.; Emerson, S. G. Human Osteoblasts Support Hematopoiesis Through the Production of Granulocyte-Colony Stimulating Factor. J. Exp. Med. 1994, 179, 1677–1682. 175. Takahashi, S.; Ooi, J.; Tomonari, A.; et al. Comparative Single-Institute Analysis of Cord Blood Transplantation From Unrelated Donors With Bone Marrow or Peripheral Blood Stem-Cell Transplants From Related Donors in Adults Patients With Hematologic Malignancies After Myeloablative Conditioning Regimen. Blood 2007, 109, 1322–1330. 176. Thomas, E. D.; Storb, R. Technique for Human Marrow Grafting. Blood 1970, 36, 507–515. 177. Thomas, E. D.; Lochte, H. L., Jr.; Lu, W. C.; Ferrebee, J. W. Intravenous Infusion of Bone Marrow in Patients Receiving Radiation and Chemotherapy. N. Engl. J. Med. 1957, 257, 491–496. 178. Thomas, E. D.; Lochte, H. L., Jr.; Cannon, J. H.; et al. Supralethal Whole Body Irradiation and Isologous Marrow Transplantation in Man. J. Clin. Investig. 1959, 38, 1709–1716. 179. Van Der Berg, D. J.; Sharma, A. K.; Bruno, E.; Hoffman, R. Role of Members of the Wnt Gene Family in Human Hematopoiesis. Blood 1998, 92, 3189–3202. 180. Wilson, A.; Trumpp, A. Bone-Marrow Haematopoietic Stem-Cell Niches. Nat. Immunol. 2006, 6, 93–106. 181. Wright, D. E.; Cheshier, S. H.; Wagners, A. J.; et al. Cyclophosphamide/granulocyte Colony-Stimulating Factor Causes Selective Mobilization of Bone Marrow Hematopoietic Stem Cells into the Blood After Mphase of the Cell Cycle. Blood 2001, 97, 2278–2285. 182. International Society for Stem Cell Research. Scientists See Progress in FDA Stem Cell Trial Approval, Sciencedaily, 2009. http://www.sciencedaily.com. 183. Vergano, D. Stem Cells Approved to Treat ‘orphan’ Disease, Advanced Cell Technology, 2010. http://www.advancedcell.com. 184. Douglass, B. Advanced Cell Technology’s Studies to Support Phase 1 Multicenter Trial of Patients with Stargardt’s Macular Dystrophy Demonstrate Excellent Safety Profile, Advanced Cell Technology, 2010. http://www.advancedcell.com. 185. ReNeuron. ReNeuron Receives Final Regulatory Approval to Commence Landmark Stroke Clinical Trial in UK, 2010. http://www.reneuron.com. 186. StemCells Inc. StemCells Inc. Plans to Advance to Second Clinical Trial in Batten Disease, 2010. http://www.stemcellsinc.com. 187. Medipost, C. L. Study to Compare the Efficacy and Safety of Cartistem® and Microfracture in Patients with Knee Articular Cartilage Injury or Defect, 2009. http:// clinicaltrials.gov. 188. Osiris, T. Extended Evaluation of Prochymal[tm] Adult Human Stem Cells for Treatment-resistant Moderate-to-severe Crohn’s Disease, 2007. http://clinicaltrials.gov. 189. Children’ss Hospital of Philadelphia. Repeated Infusions of Mesenchymal Stromal Cells in Children with Osteogenesis Imperfecta (STOD3), 2010. http://clinicaltrials.gov. 190. Medicine Encyclopedia. Physiological Changes: Stem Cells – Mesenchymal Stem Cell Numbers – Age, Aging, Differences, 2010. http://medicine.jrank.org.

1.24

Structural Organization of CellsdThe Cytoskeletonq

E Frixione and M Herna´ndez, Department of Cell Biology, Center for Research and Advanced Studies IPN (Cinvestav), Mexico City, Mexico © 2019 Elsevier B.V. All rights reserved. This is an update of E. Frixione, M. Hernández, 1.26 - Structural Organization of Cells – The Cytoskeleton, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 367-381.

1.24.1 Introduction 1.24.2 Molecular and Supramolecular Components 1.24.2.1 Evolution 1.24.2.2 Monomers and Polymers 1.24.2.2.1 Microfilaments 1.24.2.2.2 Microtubules 1.24.2.2.3 Intermediate Filaments 1.24.2.3 Polymerization, Polarity, and Treadmilling 1.24.2.4 Associated Proteins 1.24.3 Cytoskeletal Arrays and Their Structural Functions 1.24.3.1 Arrays of Actin Microfilaments or Actin-like Polymers 1.24.3.2 Arrays of Microtubules or Tubulin-like Polymers 1.24.3.3 Arrays of Intermediate Filaments or Homologue-Protein Polymers 1.24.4 Motility 1.24.4.1 Cytomotive Filaments 1.24.4.2 Molecular Motors 1.24.4.2.1 Actin Microfilament-Based Molecular Motors: Myosins 1.24.4.2.2 Microtubule-Based Molecular Motors: Kinesins and Dyneins 1.24.4.3 Motile Systems 1.24.4.3.1 Intracellular Transport 1.24.4.3.2 Cell Locomotion and Movement of the External Medium 1.24.4.3.3 Contractility 1.24.4.4 Control of Motility 1.24.5 Diseases and the Cytoskeleton 1.24.5.1 Actin Microfilament-Related Diseases 1.24.5.2 Microtubule-Related Diseases 1.24.5.3 Intermediate Filament-Related Diseases Acknowledgments References Relevant Websites

356 356 356 357 357 357 358 359 360 361 361 362 362 362 363 363 363 364 365 365 365 367 369 369 370 370 370 370 370 371

Glossary Actin (so-named because it “activates” myosin) Globular protein abundant in all eukaryotic cells that polymerizes as a couple of intertwined rows known as a “microfilament”, thousands of which form diverse types of networks and bundles that constitute one of the three major components of the cytoskeleton. ATP (adenosine triphosphate) Organic molecule capable of storing chemical energy which can be rapidly released at specific points within cells, thus serving as a main fuel for powering most metabolic reactions, including conformational changes of proteins that function as molecular motors. Cytoskeleton (CSK) Fibrous intracellular framework composed of linear polymers of specific proteins –i.e., microfilaments, microtubules, and intermediate filaments in eukaryotic cells, and their homologues in prokaryotic cells–, plus all their respective associated proteins, which supports both cell structure and directional motion mechanisms driven by polymerization dynamics and/or molecular motors.

q

Change History: October 15, 2017. E. Frixione and J. M. Hernández, the original authors of this article, completed an updated and expanded revision of the full text, with an updated reference list to recently published research, plus an improved re-elaboration of several figures created especially for this article.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00022-7

355

356

Structural Organization of CellsdThe Cytoskeleton

Dynein (from Gr. dynamis, force power, and -in, generic suffix for protein, i.e., “powering protein”) Motor protein capable of “reverse walking” (usually from plus-end to minus-end) along microtubules by means of ATP-fueled conformational changes, thus functioning as an intracellular carrier or tension-generating device in eukaryotic cells. Intermediate filament Highly stable linear polymer (
10 nm ¼ 1  108 m diam.) of any of six protein types, which builds sturdy structural networks that constitute a major part of cytoskeletons in eukaryotic cells. Kinesin (from Gr. kinein, to move, and -in, generic suffix for protein, i.e., “moving protein”) Motor protein capable of “forward walking” (usually from minus-end to plus-end) along microtubules by means of ATP-fueled conformational changes, thus functioning as an intracellular carrier in eukaryotic cells. Microfilament Linear polymer (7 nm ¼ 7  109 m diam.) constituted by two intertwined rows of actin monomers, which forms bundles and networks supporting the plasma membrane, and serves as a track for the active movement of myosins in eukaryotic cells. Microtubule Hollow cylindrical structure (25 nm ¼ 2.5  108 m ext. diam./14 nm int. diam.) composed of longitudinal “protofilaments”, i.e., rows of dimers of tubulins. Microtubules form bundles and networks supporting internal cytoplasmic arrangements, serving also as tracks for the movement of kinesins and dyneins in eukaryotic cells. Myosin (from Gr. myo or mys, muscle, and -in, generic suffix for protein, i.e., “muscle protein”) Motor protein capable of “walking” (toward to minus-end) along microfilaments by means of ATP-fueled conformational changes, thus functioning as both a tension-generating device in muscle fibers and as an intracellular carrier in other eukaryotic cells. Tubulin Any of two closely similar and related proteins (a and b) abundant in all eukaryotic cells, which coupled as heterodimers polymerize to form linear “protofilaments”, which in turn assemble into hollow cylindrical structures called microtubules which constitute one of the three major components of the cytoskeleton.

1.24.1

Introduction

In order to survive and execute their many vital functions, living cells need to maintain peculiar overall shapes as well as distinctly organized distributions of parts in their internal space. On the other hand, to both reproduce themselves and comply with their physiological roles in multicellular organisms, they often require modifying those very morphological characteristics either transiently or permanently. In addition, many cells have the ability to move through their surroundings in search of nutrients or more favorable environments, as well as to flee away from dangers or to hunt for potentially harmful agents that somehow have got access into the multi-cellular organisms where they belong. All these activities are possible because every cell has a cytoskeleton with mechanical attributes and operative features suited for such purposes. Moreover, the cytoskeleton also participates in cell communication through its association with signaling molecules. In multicellular organisms, the many different functions carried out by cytoskeletons in diverse tissues account for a remarkable variety of critical roles, from growing to hearing to muscle contraction and movement.1 The name cytoskeleton, for a concept with a surprisingly long history,2 is now given to an intracellular framework composed of various thin, linear, individually nonbranched polymers of specific proteins that provide various types of structural support and selfmovement in both eukaryotic and prokaryotic cells. This system of fibers sustains in due places the cell membrane and a variety of internal structured components like plasmids and organelles, all of which in turn determine cell form and polarity; therefore, both normal and abnormal changes in the spatial arrangement of the cytoskeleton are usually accompanied by alterations of cell shape and/or distribution of cytoplasmic constituents. Furthermore, some cytoskeletal polymers serve also as supporting tracks for selfpropelled molecular motors that mediate displacements of the fibers in relation to one another, or directional transportation of particles and macromolecules through the cytoplasm. In eukaryotic cells dthe systems best understood at present in this respect because they have extensive, comparatively complex and more thoroughly studied cytoskeletonsd there are three major cytoskeletal polymers: actin microfilaments, microtubules, and intermediate filaments, all of which are constituted by highly conserved fundamental proteins usually accompanied by a constellation of other associated macromolecules. Prokaryotic cells may contain primitive smaller homologues of these fundamental proteins, apart from others with equivalent characteristics. The present article covers (1) the molecular building blocks of cytoskeletal polymers and their main supramolecular constructions, (2) their most important structural and dynamic functions, and (3) selected diseases related to the cytoskeletons. Excellent book chapters1,3 and several reliable websites are recommended below for a broad background on the general subject, so numbered references in superscript point only at sources about recent developments on the particular topics or to some special issues.

1.24.2

Molecular and Supramolecular Components

1.24.2.1

Evolution

Cytoskeletal structures were for some time supposed to occur exclusively in eukaryotic cells, but work over the last 25 years have revealed otherwise.4,5 Current living prokaryotes contain not only counterparts of actin, tubulins, and intermediate-filament proteins but, in addition, they express also a larger inventory of their homologues.4–6 Such similitude is further emphasized by

Structural Organization of CellsdThe Cytoskeleton

357

the presence in eukaryotic DNAs of genes with both archaeal and bacterial ancestry that encode for at least some of those cytoskeletal proteins. Moreover, a viral homologue of bacterial tubulin (TubZ) has been found in bacteriophages.7 Consequently, it is now believed that eukaryotic cells conserve only those filament-forming protein lineages most favored by natural selection, among an assortment of variants produced over evolution. Despite such broad genetic and functional analogies, however, the cytoskeleton proteins of eukaryotes differ greatly in primary structures (i.e., amino acid sequences) from those in prokaryotes, and at present not all such proteins in the latter have been proved to constitute indeed functional cytoskeletal structures in vivo. Their similarity is based mainly on the common significant feature of having the ability to polymerize, at least in vitro, into some sort of filament through progressive linear linking of monomers while binding and next hydrolyzing some nucleotide like guanosine triphosphate (GTP) or adenosine triphosphate (ATP). Two families of cytoskeleton proteins showing this property have been identified. One of them includes actin, which is universal in eukaryotic cells, along with MreB, ParM and other relatives in prokaryotes. The second family comprises the also ubiquitous tubulins of eukaryotes and their homologues in prokaryotes, such as FtsZ and TubZ. Although the members in each family have little homology among them in sequences of amino acids dexcept, in a few cases, for the series of residues directly involved in nucleotide binding and hydrolysis for polymerizationd, the proteins in each of these two families share remarkably similar three-dimensional configurations in relation to their respective counterparts. Other cytoskeleton-like proteins apparently exclusive to prokaryotes, such as the intermediate-filament-homologue crescentin and the Walker A Cytoskeletal ATPases (WACAs), must also be considered in the spectrum.8,9

1.24.2.2 1.24.2.2.1

Monomers and Polymers Microfilaments

The simplest example known of a cytoskeletal polymer is a straight filament formed by two parallel rows of subunits or monomers, such as those the MreB or the FtsZ globular proteins build up in bacterial cells (Fig. 1A and C). In a variant of this model, also found in bacteria, the ParM protein polymerizes likewise as a two-row filament but which, instead of straight, is helically twisted with lefthanded torsion throughout its length (Fig. 1B). A helically twisted twin-chain configuration closely similar to that of ParM filaments, except for the opposite handedness in torsion, occurs in the actin microfilaments of eukaryotic cells (1D and 2B). Each actin microfilament, measuring only 7 nm (7  109 m) in width, is constituted by two intertwined chains of actin monomers. The two states of actin di.e., free soluble subunits and a polymerized helical twin-chaind are commonly known as globular (Gactin) and filamentous (F-actin), respectively. Depending upon the type of cell, one of various actin isoforms (a, b, or g) may predominate1,9: muscle cells deither in skeletal, cardiac or smooth muscled express mostly the a isoform, whereas the remaining tissues express the b and g isoforms.

1.24.2.2.2

Microtubules

A totally different supramolecular architecture, still producing an ultimately linear structure, is found in the microtubules of eukaryotic cells (Figs. 1E and 2A). These hollow cylinders measuring 25 nm (2.5  108m) in total diameter, of which about half is the width of the lumen at the center, count among the most remarkable polymers in biology.1,3 They are typically constituted by 13

Figure 1

Cytoskeletal polymers and monomers (see text for descriptions).

358

Structural Organization of CellsdThe Cytoskeleton

Figure 2

Assembly of cytoskeletal proteins in eukaryotic cells (see text for descriptions).

parallel protofilaments arranged in a circle, although microtubules with higher or lower numbers of protofilaments also exist; an example of this is the 12 protofilament-microtubules in the parasitic protozoan Trypanosoma cruzi. In turn, each protofilament is a succession of globular heterodimers of two closely related proteins known as tubulin-a and tubulin-b, ordered so that each of these connects with a complementary one both in front and behind along the protofilament. Because the tubulin heterodimers along each protofilament are set slightly off-register with regard to those in the immediately adjoining protofilaments, the microtubule wall is actually a helical lattice of heterodimers, as well as a cylinder made up of longitudinal rows of heterodimers arranged in a circle as described above. A third type of tubulin dtubulin-gd is found at the end of microtubules where they attach to an intracellular site known as microtubule organizing center or MTOC (see Section 1.24.3.2). Alternative assemblies of tubulin protofilaments are found in certain organelles of eukaryotic cells. Thus the axoneme or motile apparatus that extends all along the core of each individual cilium or flagellum (see Section 1.24.4.3.2, and Fig. 9) includes doublet microtubules, each doublet consisting of a typical single microtubule plus a 10-protofilament partial one attached all over its length (Fig. 1F). Short triplet microtubules constructed in the same way, except that an additional second partial tubule is added on the back of the first partial one (Fig. 1G), constitute the centrioles or basal bodies involved in the organization of arrays of microtubules within cells (see Section 1.24.3.2). At least two bacterial proteins dBtubA and BtubBd seem to be closely related to tubulins a and b and also form heterodimers capable of polymerizing in a GTP-dependent manner. This remarkable resemblance between the two couples of proteins suggests that BtubA and BtubB may actually have eukaryotic origin through horizontal gene transfer. FtsZ, another bacterial protein, can also polymerize into tubular and sheet-like structures in vitro, although it is still not certain whether FtsZ filaments are indeed tubular in vivo. Other tubulin-like proteins in prokaryotic cells are TubZ and RepX.10

1.24.2.2.3

Intermediate Filaments

A third major class of cytoskeletal structures comprises the intermediate filaments (Fig. 1H), so called because their 10-nm (1  108 m) diameter lies in between those of microtubules and actin microfilaments. Intermediate filaments are found almost exclusively in cells of animals with relatively soft bodies such as vertebrates, mollusks and worms, i.e., those having no outer shells or rigid integuments, so their internal tissues may be exposed to varying mechanical stresses through bodily motions. These sturdy filaments can be constituted by varying combinations from an assortment of over 70 proteins with very dissimilar molecular weights, which are differentially expressed in various types of cells. They have been identified and classified into six groups based on similarities between their amino acid sequences, and depending on the tissues being considered.11 In epithelia, they are represented by the keratins, while in nervous tissue they include a protein triplet that composes the neurofilaments in neurons, plus a fibrillary acidic protein abundant in glial cells. Desmin is characteristic of striated muscle. Vimentin, found in diverse kinds of

Structural Organization of CellsdThe Cytoskeleton

359

tissues, and the lamins, associated to the nuclear envelope of all eukaryotic cells, are the most widespread proteins that polymerize as intermediate filaments. The assemblage of intermediate-filament proteins is far more complex than that of microfilaments and microtubules (Fig. 2C). Two slender, helical monomers intertwine to form a coiled-coil dimeric subunit, which then couples in staggered fashion with another similar but antiparallel subunit to constitute a tetramer. Series of tetramers aggregated sidewise then make up protofibrils, in which a cylinder of eight tetramers amounts to a “unit-length filament” (ULF). Next, end-to-end serial linking of ULFs builds up short protofilaments, which connected by their ends finally constitute an intermediate filament. In addition, several of these filaments often cross-link in parallel by means of associated proteins (see Section 1.24.2.4), the robust bundle being known as a tonofilament. Further, spectroplakin, another associated protein, allows the interaction of intermediate filaments with microtubules and microfilament for the construction of a whole cytoplasmic framework. Crescentin, a bacterial peptide that assembles in vitro into 10-nm thick filaments, is presently the only representative of intermediate filament-like proteins found in prokaryotic cells.8

1.24.2.3

Polymerization, Polarity, and Treadmilling

Actin and tubulin dand their homologues in prokaryotic cellsd can both rapidly polymerize and depolymerize into or from their corresponding cytoskeletal structures according to physiological needs or experimental circumstances. Polymerization starts as soluble actin monomers or tubulin heterodimers first assemble into a small ‘polymerization nucleus’, from which the polymer then grows by subsequent addition of subunits in strict head-to-tail orientation during an elongation phase (Fig. 2A and B).1,3,9 This process occurs as a sequential allosteric reaction, in which incorporation of a subunit at one end of the already-assembled structure facilitates the attachment of another free subunit to the last one, and so on. A stationary phase is eventually reached in which free and assembled subunits attain equilibrium, which is in turn dependent upon the concentration of free subunits and sensitive to the local physicochemical environment. Under appropriate laboratory conditions dbasically a critical concentration of subunits in a medium of certain ionic composition and pH at the right temperature, in the presence of ATP or GTPd MreB, ParM, or actin monomers readily polymerize into protofilaments or microfilaments. A similar process occurs with TubZ, RepX, and BtubA/BtubB or with a, b-tubulin dimers, which assemble as microtubules.6,10 Sheets or filament bundles are also found in these preparations as a result of lateral associations of the basic linear polymers. Conversely, cytoskeletal polymers break down into free subunits as conditions change. One of these variations in the levels of free Ca2þ or Mg2þ ions in the medium, which are conducive to depolymerization of microtubules or actin microfilaments, constitutes a major factor in the physiological management of many functions of the cytoskeleton in eukaryotic cells (see Section 1.24.4.4). In vivo, the polymerization of cytoskeletal proteins is further controlled by subtler mechanisms, notably the binding of specific accessory proteins (see Section 1.24.2.4). Combinations of factors influencing the polymer-to-subunit equilibrium can lead to sudden bursts of assembly or, on the contrary, abrupt episodes of depolymerization. This property, known as dynamic instability, which is well characterized for actin and tubulin in eukaryotic cells, has also been demonstrated in vitro for the actin-homologue ParM of prokaryotes. Certain drugs that affect dynamic instability through binding to subunits either assembled or in the free state, thereby stabilizing or destabilizing the respective polymers, are useful tools for experimental research. Thus, actin microfilaments are disassembled by cytochalasins and strengthened by phalloidin, whereas microtubules become depolymerized by colchicine or nocodazole and stabilized by taxol. Since polymerization involves a precisely oriented linking of each added subunit, as determined by the specific molecular 3D shapes of the respective proteins, cytoskeletal polymers have an intrinsic structural polarity. In other words, their two ends are not equivalent, and this asymmetry constitutes a key property because it determines polymer direction at every point of its length. A consequence of this structural polarity is, for example, the differential rate of subunit incorporation and loss at opposite ends of some cytoskeletal polymers, depending on the local concentration of free subunits (Fig. 2A and B). Assembly and therefore elongation occur by progressive addition of subunits preferentially at one end (the plus or þ end, especially if others have been recently added at this extremity), while disassembly and thus shortening result from removal of subunits, also mainly at the þ end of the polymer. The opposite (minus or –) end is comparatively less active in both the assembly and disassembly processes, at least in microtubules and microfilaments of eukaryotic cells.1,3,9 Generally, polymerization and depolymerization occur in vivo simultaneously at opposite ends of a cytoskeletal polymer, thus producing a net flow of subunits along its whole structure in a dynamic steady state known as treadmilling (Fig. 2B).12 A mechanism of this type exists also in prokaryotic cells, where labeled monomers of the bacterial tubulin TubZ, as well as of the actin-homologue MreB, have been reported to move down throughout the corresponding polymers for the full length of the cells.6 A notable exception to the highly dynamic behavior of most cytoskeletal polymers are the intermediate filaments which, given their complex construction, are much more stable and even capable of withstanding harsh chemical or mechanical treatments. This toughness not only greatly contributes to the structural integrity of the whole cytoplasm, but also presents a difficult challenge for experimental analysis. Hence, the dynamics of the intermediate filaments is currently less understood than that of microfilaments or microtubules. In fact, apart from phosphorylation that interferes with their polymerization, it was originally thought that they were almost inalterable physiologically. This view has been corrected in recent years, as evidence was obtained that monomers can be incorporated into, as well as released from, existing intermediate filaments.11

360 1.24.2.4

Structural Organization of CellsdThe Cytoskeleton Associated Proteins

The three major cytoskeletal filaments in eukaryotic cells have each a collection of associated proteins that regulate changes in polymer length, contribute to its structural stability, and mediate passive linking with other elements of the cytoskeleton or anchoring to membranous cellular components (Fig. 3). To some extent the same is true for cytoskeletal structures in prokaryotic cells. Additionally, eukaryotes possess molecular motors that generate mechanical forces over actin microfilaments or microtubules, and thus either develop tension or “walk” along the respective polymer carrying various kinds of cargoes with them across the cytoplasm. In this section, we survey the cytoskeleton-associated proteins that serve purely or primarily structural roles in general, leaving the molecular motors for the section that deals on motility (1.24.4). The function of microfilaments is regulated in various ways by specific actin-binding proteins (ABPs), some of which simply join to the free monomers and regulate polymerization by either hindering (thymosin) or promoting (profilin) their assembly.9 On the other hand, the assembled filaments can be protected against depolymerization by actin-capping proteins such as CapZ, which binds to the plus (þ) end of the polymer and thus prevents both addition and loss of monomers, or tropomodulin, which blocks monomer release by capping the minus () end. Severing ABPs, such as gelsolin and cofilin, break filaments into shorter fragments and further control their length by promoting either assembly or disassembly. Other ABPs cross-link the filaments, either in parallel to form bundles (fascin and fimbrin) or transversely at various angles, thus building networks (filamin, spectrin, and dystrophin). Some of the latter also attach the filaments to integral proteins in the plasma membrane, so the networks become anchored at the periphery of the cell. Since most cells in tissues are covered on the outside by a fibrous coat called the extracellular matrix, which is in turn attached to similar coats of adjacent cells, the intracellular microfilament network becomes greatly stabilized. A peculiar type of ABP is tropomyosin, a thin double-fiber extended within the groove left between the paired rows of actin monomers along the middle of microfilaments in muscle fibers. Under the control of troponin, itself regulated by local Ca2þ levels, tropomyosin can change slightly its position so that actin becomes either exposed or hidden for binding with the motor protein myosin II. As with actin filaments, microtubule length is controlled by capping, though with GTP or guanosine diphosphate (GDP) instead of proteins. GTP bound to the b-tubulin of free heterodimers promotes their assembly but becomes hydrolyzed in the process. The resulting GDP is adverse to the incorporation of more subunits and can therefore temporarily block further polymerization, or even facilitate polymer breakdown. Microtubule dynamics is also regulated by cytoplasmic linker proteins (CLIPs) and their associating proteins (CLAPs) that, acting together, foster elongation and stabilization of microtubules selectively at certain cytoplasmic regions, and therefore control overall cell structure.13

Figure 3

Cytoskeleton-associated proteins (see text for descriptions).

Structural Organization of CellsdThe Cytoskeleton

361

Microtubule structural stabilization and cross-linking with each other, or with filaments and membranes, are mediated by two types of microtubule-associated proteins (MAPs) particularly abundant in neuronal processes.13 Type I MAPs (MAP 1A and MAP 1B) have a stabilizing role, contributing to microtubule firmness through shielding electrical repulsion between the negatively charged tubulin subunits, whereas type II MAPs (MAP 2, MAP 4, and tau) serve a cross-bridging function with other microtubules and adjacent organelles. Intermediate filament-associated proteins (IFAPs), such as plectin and filaggrin, cross-link these filaments into bundles (i.e., tonofilaments; see Section 1.24.2.2.3) or with other cytoplasmic structures. In contrast to ABPs and MAPs, however, the IFAPs seem to have little regulatory influence on the intrinsic high stability of the intermediate filaments themselves.3

1.24.3

Cytoskeletal Arrays and Their Structural Functions

All three major cytoskeletal polymers and their respective associated proteins build up extensive networks in animal cells and protozoans, each with a distinct spatial distribution and several specialized functions (Fig. 4). In contrast, only a limited deployment of intracellular networks is found in plants and prokaryotes; nevertheless, cytoskeletal polymers control cell morphology and physiology also in many of these organisms. Thus, in virtually all instances the cytoskeleton is crucial for the spatial organization of cells, from determining their overall shape to securing the precise internal localization and activity of organelles.

1.24.3.1

Arrays of Actin Microfilaments or Actin-like Polymers

The cross-linking and attachment of actin filaments by various ABPs in particular ways allows for the formation of different kinds of arrays, which in turn determine the local or whole shape of the cell. Planar webs of microfilaments underlying and ultimately anchored to integral proteins at the plasma membrane constitute a major component of the cell cortex (Fig. 4A and D), from which bundles of microfilaments extend inward in diverse directions and cooperatively contribute to determine cell configuration and cytoplasm consistency in animal cells.

Figure 4 Microscopic views of the major cytoskeletal networks. Pictures correspond to cells of the human glioblastoma-astrocytoma epithelial line U373MG.

362

Structural Organization of CellsdThe Cytoskeleton

Small signaling proteins generically known as Rho-GTPases control the specific kind and amount of filament cross-linking by the various ABPs in order to produce particular structures.1,14 Thus, for example, the Rho-GTPases Rac1 and Cdc42 induce the generation of local networks capable of pushing the plasma membrane outward from the cell body, either as flat, shallow and wide lamellipodia or as fingerlike filopodia and microvilli, respectively.9,15 Another Rho-GTPase called RhoA signals for the formation of bundles of filaments that, through interaction with the motor protein myosin II (see Section 1.24.4.2.1 below), constitute contractile stress fibers in nonmuscle cells. The most developed and regular arrays of actin microfilaments are found in muscle fibers, i.e., cells specialized in movement (see Section 1.24.4.3.3). Polymers of actin-homologues such as MreB also play both structural and dynamic roles in prokaryotes, from forming a spiral pattern (Fig. 10) that maintains a specific cell profile through regulation of peptidoglycan distribution over the cell wall, to directly segregating plasmids or chromosomes.6

1.24.3.2

Arrays of Microtubules or Tubulin-like Polymers

In contrast to actin microfilaments, which are characteristically tethered or otherwise related to the plasma membrane at the cell periphery, microtubules typically radiate from dense intracellular regions called ‘microtubule-organizing centers’ (MTOCs) that commonly include a couple of centrioles (see Section 1.24.2.2.2). MTOCs contain the protein pericentrin and a third type of tubulin (tubulin-g; Fig. 2A), both of which are involved in nucleating the origin of new microtubules. Most animal cells at interphase have a single MTOC known as centrosome situated near the nucleus1,12 (Fig. 4C and D). Since the centrosome has a special affinity for the minus or () end of the microtubules (see above), these become preferentially oriented with their more actively polymerizing plus or (þ) end pointing toward the periphery. On the other hand plant cells, as well as animal embryos and many epithelia, contain multiple MTOCs. Some of these are positioned near the cell cortex, so that microtubules growing from them lie parallel to the plasma membrane. In addition, the dynamics of assembly and association properties intrinsic to the microtubules, which are governed at least in part by MAPs, seem to permit self-organization processes and stabilization independently from MTOCs in some plant cells. While changes in the positioning of actin microfilament arrays are generally expressed as surface phenomena over limited regions of a cell, like ruffling or endocytosis, transformations of microtubule networks are commonly reflected in whole-cell shape alteration or distribution of internal organelles. The building up of large microtubule frameworks is behind such spectacular processes like the growth of a new branch from a nerve cell. Conversely breakdown of microtubules, through treatment with substances such as colchicine, for example, leads to large-scale disorganization of the normal microanatomy in the cytoplasm. Both of these cytoskeletal systems dmicrofilaments and microtubulesd are centrally involved and collaborate with each other in animal cell division, when the mother-cell elongation driven by the construction and operation of a microtubule-built mitotic spindle is followed by the progressive narrowing of a perpendicular, membrane-attached contractile ring made up of actin microfilaments that completes the separation of the original cytoplasm into two equal parts for the daughter cells (see Section 1.24.4.3.3). Microtubules and motor proteins participate also in intermediate-filament polymerization from ULFs in nervous tissue.

1.24.3.3

Arrays of Intermediate Filaments or Homologue-Protein Polymers

Intermediate filaments act as mechanical integrators of the cytoplasm and are distributed accordingly from the nucleus to the plasma membrane (Fig. 4B).11,16 One type of intermediate filament forms a network layer underlying the inner side of the nuclear envelope, while a layer of a different type of these filaments lines the cytoplasmic surface of the plasma membrane.17,29 In addition radial intermediate filaments connect both layers or fasten at their distal ends to aggregates of membrane proteins called desmosomes, so that a flexible framework traverses the whole cell.18 Most cytoplasmic organelles, including actin filaments and microtubules, establish relatively firm associations with intermediate filaments through associated linking proteins like spectroplakin, as mentioned above (see Section 1.24.2.2.3). The cross-linking of both actin filaments and microtubules to intermediate filaments helps in achieving an overall integration of mechanical tensions, either internally generated or in response to externally applied stress, throughout the entire cytoskeleton in a phenomenon called ‘tensegrity’.17,29 Permanent active work of myosin motors (see Section 1.24.4.2) over actin microfilaments underlying the plasma membrane, for example, permits cells to stand up due to an intermediate filament-assisted peripheral distribution of tensions against a supporting inner scaffold of microtubules. In bacteria, the intermediate-filament protein homologue crescentin,8 by localizing at just one side of the rod-shaped cell, restricts local elongation of the wall. Asymmetric wall elongation then results in the characteristic crescent-like form of the cell.

1.24.4

Motility

Apart from their structure-sustaining and reinforcing functions, cytoskeletons account for most forms of biological self-motion or motility. This term includes a number of apparently diverse dynamic phenomena, from plasmid segregation in prokaryotes and chromosome separation in eukaryotes to the displacement of whole cells, and from the relatively slow and primitive amoeboid

Structural Organization of CellsdThe Cytoskeleton

363

movement to the highly specialized and fast activities mediated by contractions of striated muscles in higher animals. Yet, as discussed in this section, similar mechanical principles apply in the various classes of motility.

1.24.4.1

Cytomotive Filaments

The mechanisms producing internal movements or changes of cell shape in prokaryotes are now being better understood. It is already clear that proteins of the cytoskeleton play a key role in distributing other intracellular components such as DNA, and in constricting the plasma membrane during cell division. No polymer-riding molecular motors such as those present in eukaryotic cells (see Section 1.24.4.2) have been found so far in prokaryotes, nor are they apparently necessary for their motile processes. Instead, the polymers themselves, consuming the energy provided by nucleotide hydrolysis coupled to polymerization, seem capable of developing linear force with enough power as for acting mechanically upon other structures.4,5 The same capacity is likely present in actin and tubulin polymers of eukaryotes, even though, having to move much greater loads, they also serve as supporting guides for molecular motors.

1.24.4.2

Molecular Motors

Molecular motors are mechano-enzymes capable of undergoing a forceful and reversible conformational change using up energy derived from the hydrolysis of ATP. When acting upon cytoskeletal polymers, such forces allow these molecules to move in a steplike fashion, in effect “walking” over and along the linear structures. Actin microfilaments and microtubules dthough not intermediate filamentsd support, transmit, and/or apply mechanical forces generated by molecular motors in eukaryote cells. The fact that both actin microfilament-based and microtubule-based molecular motors share the same mechanical principle strongly suggests that they may have evolved from a common ancestor. In the following paragraphs we first discuss the main properties of the most-studied molecular motors and then focus on examples of the various roles these devices play in the most important motile systems.

1.24.4.2.1

Actin Microfilament-Based Molecular Motors: Myosins

Myosins, mechano-enzymes that work upon actin microfilaments to produce cell contraction and other functions, are representative examples of molecular motors. These large proteins consist of a heavy chain with three well-defined domains (Fig. 5A): (1) a globular head dthe actual motord with specific binding sites for actin and ATP; (2) a neck where two or more regulatory light chains govern the mechanical process depending on certain conditions, such as the local level of free Ca2þ ions (see Section 1.22.4.4); and (3) a long coiled tail that can mediate either the pairing with another myosin molecule to integrate a dimer, or the attachment to other cytoplasmic structures. The myosin family consists of 18 elements differing in the carboxyl terminal, where each member possesses specific functional domains.19 Myosins II, V, and XVIII function paired as parallel dimers, but the rest of them dtypically myosin Id work as monomers.3,30

Figure 5

Myosin I as an example of molecular motor in successive steps of interaction with actin microfilament (see text for detailed description).

364

Structural Organization of CellsdThe Cytoskeleton

In the absence of ATP, a myosin head can attach strongly at discrete periodical sites along actin microfilaments (Fig. 5B). Binding of ATP at the proper place on the myosin head permits two successive steps (Fig. 5C): (1) release from the microfilament and (2) an ATP hydrolysis-coupled conformational change of the myosin head, such that it can reach the next attachment site on the microfilament (toward the þ end). Since this site must be available (given that the immediately following myosin head was also released by ATP), and because ATP is now locally gone after being converted to adenosine diphosphate (ADP), the myosin head attaches again firmly at the corresponding advanced position on the microfilament (Fig. 5D). This new attachment then causes dissociation of the ADP and a vigorous return of the head to its original conformational state, in effect exerting on the filament a thrusting jerk parallel to its axis and toward the () end of the filament (Fig. 5E). The force involved and the movements it can generate under experimental conditions of virtually zero load have been both measured in single discrete events, revealing values of a few piconewtons and a few micrometers, respectively. The ratchet-like mechanical cycle, including myosin head release, conformational change, reattachment, and power stroke, can be repeated continuously with a net advancement along the actin microfilament every time, just like footsteps while walking. Which member of the pair dmyosin or actind moves more as a result of this interaction depends on their relative mobilities (see examples in Section 1.24.4.3 below). The mechanical cycle goes on for as long as adequate levels of ATP are present and nothing obstructs the myosin–actin interaction. As discussed in Section 1.24.4.4, this second condition is variable and under physiological control, so that motor action can be increased or decreased and even turned on and off.

1.24.4.2.2

Microtubule-Based Molecular Motors: Kinesins and Dyneins

Just like actin microfilaments serve as tracks for the myosins’ self-propelled displacement, microtubules support and guide the movement of two kinds of molecular motors dkinesins and dyneinsdwith general features resembling those of the myosins. Thus, the kinesins are also dimers in which each member of the pair consists of one heavy and one light chain configuring three domains: a large globular head, a long slender coil that in turn coils together with the parallel member of the dimer, and a small globular end that includes the light chain.1,3 The large globular heads of kinesin attach to microtubules, along which they execute the ATP-dependent cyclic mechanical action, whereas the small globular domains at the end may bind to another microtubule, a vesicle, or other intracellular organelle (Fig. 6A–C). The driving force developed by the kinesin motor heads, transmitted through the coiled-coil central domain that pairs together the two monomers, pulls the attached organelle along or makes it to roll over the supporting microtubules. There are 14 kinesins, most of them involved in transport on microtubules or helping microtubular structures to slide over each other, but kinesin XIII also named catastrofin promotes microtubule depolymerization.1 Different amino acid sequences in the end domains of kinesin molecules specify the particular types of cargo dmicrotubules, mitochondria, cytoplasmic granules, lysosomes, or chromosomesd that each of them is capable of driving.20 Most kinesins propel themselves and hence their cargoes toward the (þ) end of microtubules, that is, away from the MTOC near the cell nucleus. Yet, kinesins involved in the activities of the mitotic apparatus (see Section 1.24.4.3.3) exist as both (þ) end and () end types. Thus, in contrast to actin microfilaments, which only support myosins walking toward their (þ) end, microtubules can accept molecular motors moving toward both ends. The most abundant () end-directed molecular motors are the dyneins, which drive the beating of cilia and flagella as well as the centripetal hauling of organelles within cells (see Section 1.24.4.3 below). They are large proteins constituted of two or three heavy chains plus a variable number of lighter chains. Still, morphologically, they conform to the design of myosins and kinesins, that is,

Figure 6 Kinesins functions. (A) A single kinesin, with a specific molecule or other particle (not shown here) attached to its distal (here superior) end, walks carrying its cargo along a microtubule. (B) A number of kinesins attached to a microtubule (or mitochondrion) by their distal (here superior) ends slide their common cargo over another microtubule. (C) A number of kinesins attached to a vesicle by their distal (here outer) ends allow their common cargo to roll over a microtubule.

Structural Organization of CellsdThe Cytoskeleton

365

large globular headsdthe actual motorsdconnected by flexible stalks to a common axis terminated by small globular domains involved in attaching the structure that is being acted upon. A singular feature of some dyneins, however, is that attachment to the cargo requires the participation of accessory linking proteins such as dynactin.1,3 The dynein family has expanded also to 14 members, including those from ciliated protozoans.

1.24.4.3

Motile Systems

The actual result of the force exerted by molecular motors upon cytoskeletal tracks depends on the number and positions of simultaneously active motor heads and on the relative mobility of their associated structures. Here, we review just the most important examples.

1.24.4.3.1

Intracellular Transport

Myosins I and V, interacting with actin microfilaments associated to the plasma membrane, are known to participate in some examples of intracellular transport of vesicles and other organelles within cells. A large-scale instance of this process is the cytoplasmic streaming or cyclosis observed in some green algae, where the whole cytoplasm is kept going on continuously around a large central vacuole for as long as the cells are illuminated. Bundles of actin microfilaments, all with the same polar orientation and attached to a layer of chloroplasts fixed to the plasma membrane, provide a peripheral bed of unidirectional tracks over which myosin molecules can collectively drive the cyclic motion of membrane-bound organelles, which through viscous drag move the fluid cell contents as well. For the most part, however, it is the kinesins and cytoplasmic dyneins that carry things around within cells, over short or long distances and often along bidirectional routes. Transport within nerve fibers of large animals, for example, occurs through distances that may be in the order of tens of centimeters, so macromolecules and organelles produced in the cell body at the spinal cord level can be supplied to the remote nerve terminals in the periphery, and these can in turn report back to the cell body about local conditions and needs (Fig. 7). A precisely choreographed instance of intracellular transport, involving a variety of molecular motors, is the separation of chromosomes by the mitotic apparatus during the division of eukaryotic cells (see Section 1.24.4.3.3, and Fig. 10A). Two or more bundles of numerous microtubules, transiently taking up virtually all of the tubulin available in the cytoplasm, form a spindleshaped scaffold that successively (1) forces the dividing cell to elongate in a specific direction, (2) captures the chromosomes and organizes them into two equivalent sets at its equator, and (3) pulls the two sets apart from each other toward the spindle poles, thus driving each set to the cytoplasmic region that will soon become a daughter cell.21 Cell division in prokaryotes is mediated also by polymers of the tubulin homologues TubZ, RepX and FtsZ, which assist plasmid DNA replication and segregation into two nucleoids, apart from leading the development of a septum for binary fission of the original cytoplasm.5,10 Hence this system may constitute a primitive precedent for the highly elaborate mitotic apparatus present today in nearly all eukaryotic cells.

1.24.4.3.2

Cell Locomotion and Movement of the External Medium

A directional flow of cytoplasm driven by myosin I motors moving over actin filaments can become a pushing force that forms a transient cell projection called pseudopod (Fig. 8) and can lead to whole cell displacement when accompanied by new attachments of the pseudopod membrane to the substratum, simultaneous with detachment at other points, as observed in amoebas and similar cells. Ciliary and flagellar beating driven by dynein (Fig. 9) provides a wholly different and much faster class of relative displacement, either of a cell or of its immediate surroundings. In contrast to the prokaryotic flagellum, which is a simple filament of a single protein (flagellin) that passively follows the spinning of a rotary motor at its base, cilia and flagella in eukaryotes are internally supported and actuated by a full-length axial apparatus called axoneme (Fig. 9B). This little marvel of natural bioengineering is built

Figure 7 The cytoskeleton of a nerve fiber supports axonal transport mediated by kinesins in anterograde direction (away from the cell body), and by dyneins in retrograde direction (toward the cell body).

366

Structural Organization of CellsdThe Cytoskeleton

A

B

C

D

Figure 8 Amoeboid movement. (A) Cytoplasmic flow driven by actin-myosin I interaction starts to push the plasma membrane out as a pseudopod at the leading front (here upper) of the amoeboid cell. (B) and (C), the pseudopod grows progressively in the same direction. (D) A new pseudopod starts to form from the left side of the previous one.

Figure 9 Axoneme beating motion in cilia. (A) Lateral view of successive positions of an active cilium, similar to the arm movement of a swimmer. (B) Axoneme structure in cross-section (above) and side-view (below). (C) and (D) Cilium bending occurs by longitudinal displacement of some microtubule doublets relative to others. Whip-like flagellar movements result from the same mechanism occurring throughout a much longer axoneme.

with two types of tubulin polymers in parallel: 1) a sheathed central pair of single microtubules and b) a surrounding palisade of several doublet microtubules, usually nine in most cases (see Section 1.24.2.2.2 and Fig. 1F). Accessory proteins provide radial spokes that bridge at regular intervals from the central pair of single microtubules to each of the peripheral doublets, and each doublet is in turn strapped at multiple levels to each of its immediate neighbor doublets, so the whole axoneme is a thoroughly cross-linked construction. When active, the axoneme is forced to respond mechanically to longitudinal stresses imposed by hundreds or thousands of dynein motors distributed periodically along each microtubule doublet, each dynein pulling downward upon the immediately neighboring doublet. This longitudinal pull on each doublet toward the cell body, working against the opposition presented by

Structural Organization of CellsdThe Cytoskeleton

367

the internally braced scaffold of the whole structure, bends the axoneme in a specific direction (Fig. 9C). Sequential pulling in all the doublets one after the other translates into rotational bending of the cilium or flagellum (Fig. 9A). Depending on the length of the entire structure, the resulting movement can be undulating whip-like dfound mainly in flagellad or swimmer arm-like as in cilia. Viscous drag caused by the motions of these appendages within the external medium permits displacement of free-living cells, or sweeping of the medium over the cells if these are fixed as part of an epithelium.

1.24.4.3.3

Contractility

Cell cortex tension. As explained above (Section 1.24.3.1), most eukaryotic cells have extensive webs of actin microfilaments underlying the plasma membrane. Myosin II motors pulling on these filaments generate tension that is continuously maintained around the cell cortex and can be accentuated at some regions depending on the local distribution and numbers of filaments.17,22 One instance of this arrangement is the circumferential belt of actin microfilaments near the apical surface of epithelial cells, at the level where each cell establishes a tight-junction with its immediate surrounding neighbors. The motor action of myosin II on these microfilaments exerts a constricting tension at the top of each cell and, because contiguous cells are connected with one another by desmosomes (see Section 1.24.3.3), the simultaneous collective tension tightens up the whole epithelium right at the border of the tissue and keeps it effectively sealed against free permeation of most solutes. The lateral force pulling the cells together, combined with filament-driven protrusions from healthy cells at the periphery of a wound, also helps healing by gradually obliterating gaps opened in epithelia. A temporary though more general example of a circumferential band of actin microfilaments is the contractile ring that fully separates the cytoplasm of a dividing animal cell into two portions that become daughter cells (Fig. 10A; see also Section 1.24.4.3.1). In this process called cytokinesis the myosin II-powered contractile ring, situated at the equator of the original cell and at right angles to the spindle of microtubules that has just moved apart the two sets of chromosomes, constricts the cell body with a cleavage furrow that finally separates the cytoplasm in two cells. A similar pattern is observed for cell partition in some bacteria (Fig. 10B). Septum construction between the two daughter nucleoids starts with the formation of a cortical Z-ring constituted by filaments of the tubulin-like protein FtsZ. This ring then acts as a recruiting band for the aggregation of the actin-like FtsA and other proteins necessary for complete partition into two daughter cells.4,10 The same machinery, but located close to the poles of rod-shaped bacteria, is used for asymmetrical partition during sporulation. Muscle contraction. Biological motion is mainly manifest in animals because they are provided with large amounts of specialized equipment for the purpose, in the form of two classes of muscle tissue dsmooth and striatedd distinguished by the internal architecture of their constituent cells and their different properties. Both types of muscle are packed with actin microfilaments and myosin II in a more highly organized arrangement than in other cells.1,3 Myosin dimers occur here associated by their tails in both parallel and antiparallel orientations, thus constituting bipolar bundles that appear as thick filaments (Fig. 11A). The staggered bundling of myosin dimers in a thick filament results in a symmetrical distribution of motor heads along the two outer thirds of the filament length. Because these heads protrude radially in various directions around the thick filament, each of these rods is capable of making contact at multiple levels with several adjacent actin microfilaments at once.

Figure 10 Cell division. (A) In eukaryotic cells a spindle of microtubules and dyneins carries equal sets of chromosomes to opposite poles of the mother cell (mitosis), while a contractile ring of myosin II-actuated microfilaments constricts the middle of the cell (cytokinesis). (B) In prokaryotic cells a helicoidal filament of MreB extends the original cell, while a filament of the microtubule-like FtsZ marks and possibly helps to constrict the cell by the middle to create a partition septum.

368

Structural Organization of CellsdThe Cytoskeleton

Figure 11 Myosin-actin force-generating interaction. (A) Myosin II thick filament. (B) Mechanical interaction of myosin II motor heads with adjacent actin filaments anchored to a Z-plate in striated muscle. For illustration purposes, actin filaments are here represented at about the same thickness as myosin filaments, though they are in fact comparatively thinner (see Fig. 12B).

Force-generating interaction will occur only when the myosin heads attach to properly oriented actin microfilaments, so that their respective polarities permit functional coupling (Fig. 11B). Whenever this occurs, numerous myosin heads pull coherently on parallel microfilaments, so the latter slide over the thick filaments. Because these are bipolar bundles, each of them can interact at the same time with oppositely oriented microfilaments. Thus, upon activation, the microfilaments move in opposite directions, that is, toward the middle section of the thick filament, the net result being a diminution in length of the whole assembly. Since the actin microfilaments drag with them any linked membranes or other structures, collective activation of this mechanism produces contraction of a whole muscle cell. The overall arrangement greatly increases the efficiency of a system evidently evolved to optimize contractile power, and its advantages are exploited in two different versions. Smooth muscle, found mainly in arterial walls and abdominal organs of vertebrates, takes its name from the fact that it lacks the periodical transverse stripes characteristic of striated muscle fibers, such as those in the heart and in skeletal musculature (see below). This is explained by the loose internal distribution of cytoskeleton components in smooth muscle cells. These are fusiform fibers akin to other cells in that they contain bundles of actin microfilaments fastened by their ends to integral proteins of the plasma membrane, usually at discrete patches or attachment plates. Smooth muscle cells are peculiar, however, in having also actin microfilament bundles arising from dense bodies distributed within the cytoplasm, and a number of interspersed thick filaments of myosin. The two populations of complementary filaments have thus many possibilities for interacting nearly everywhere inside the cells. Hence, if the muscle is distended while at rest, the myosin heads will always find nearby actin microfilaments to interact with, although probably other than those met in the previous contraction. This plasticity of smooth muscle, a property absent in the striated type, permits the bladder or the uterus to contract with similar force, irrespective of their initial extent of distension. Because myosin thick filaments and microfilaments are repositioned every time, however, such an advantage comes at the price of a slow speed of response. This last property becomes maximized, on the other hand, in the precisely positioned and comparatively constant arrangement of the cytoskeleton in striated muscle fibers (Fig. 12). The cooperative action of myosin heads is here multiplied by a regular overlapping of the two types of filaments, so that each thick filament of myosin is surrounded by several actin microfilaments, and each of these is in turn cross-bridged by myosin heads protruding from two or more adjacent thick filaments (Fig. 12B). This pattern, consisting of a set of parallel myosin thick filaments partly interpenetrated by two oppositely oriented sets of actin microfilaments, is the basic building plan of the sarcomere or functional contractile unit of all striated muscles. A long series of consecutive sarcomeres constitutes a myofibril, and bundles of myofibrils make up a muscle cell or fiber. The relatively constant transverse alignment of sarcomeres across the myofibrils in a fiber is the reason for the characteristic striations observed all along skeletal and cardiac muscle under the microscope.1 Simpler instances of contractile cytoplasmic fibrils that cause transient changes in cell shape by mechanisms unrelated to actin and myosin are found in some unicellular organisms. Ciliated protozoans contain myonemes and spasmonemes that are capable of shortening faster than any muscle fiber. Thus in Spiroplasma, a small bacterium without a cell wall, contractile fibrils participate in producing a corkscrew shape as well as locomotion dfibril contractions cause variations in the pitch of coiling of the whole helical body, thus propelling the organism through the surrounding liquid medium.

Structural Organization of CellsdThe Cytoskeleton

369

Figure 12 Contraction of striated muscle. (A) Symmetrical arrangement of a contractile unit (sarcomere) in striated muscle, where simultaneous traction of myosin II motor heads in bipolar thick filaments over antiparallel actin filaments result in shortening. (B) Schematic illustration of filament arrangement in the resting and active conditions, as seen in cross-section (left) and side-view (right).

1.24.4.4

Control of Motility

As already mentioned (see Sections 1.24.2.3 and 1.24.4.2.1), assembly and stability of both actin microfilaments and microtubules depend upon fluctuations in the concentration of free Ca2þ ions in the cytoplasm. This factor is also a virtually universal stimulus-response coupling link for mechanisms of biological motion. Amoeboid movement, for example, follows chemical gradients in the environment because the local concentration of stimulant molecules modifies the internal level of Ca2þ in that region of the cell, thus creating an intracellular Ca2þ gradient that in turn affects the motile machinery in two complementary ways: (1) Ca2þ increase activates myosin-actin coupling to produce contraction of the cell cortex and (2) at the same time Ca2þ destabilizes inner actin microfilaments, so the cytoplasm becomes more fluid. Since opposite changes occur at cytoplasmic regions with low levels of Ca2þ, the cell becomes mechanically polarized and responds accordingly. Such Ca2þ influences on motility are mediated by regulatory proteins such as calmodulin, which can bind four Ca2þ ions and, in this state, it is able to associate with myosin light-chain kinase and thus activate myosin II through phosphorylation. Still subtler regulation of motility involves signal-transduction mechanisms composed of sequential steps performed by various intracellular messengers acting in cascade-like fashion. The control of motility is best understood in the contraction of skeletal muscle. In response to the physiological stimulus (a transient electrical depolarization of the plasma membrane), all the sarcomeres in a myofibril and all the myofibrils in a fiber shorten almost synchronously, thus producing a twitch contraction. Activation of sarcomere shortening is mediated by an equally transient rise of intracellular Ca2þ, which through a regulatory complex of proteins (troponin) induces a change in the shape of tropomyosin (see Section 1.24.2.4), ultimately permitting and/or facilitating the traction of myosin heads over actin microfilaments. The Ca2þ levels in the cytoplasm, in turn, are themselves regulated by intracellular sequestering and releasing mechanisms.

1.24.5

Diseases and the Cytoskeleton

Cytoskeletal polymers and their associated proteins are involved in diverse health problems of considerable relevance. Some bacterial infections interfere with normal cytoskeletal function by producing toxins that bind to the polymers or their associated proteins, or by secreting peptides similar to mammalian actin-binding proteins so they compete with the natural ones. In a more sophisticated mechanism, invading bacteria express effector proteins that promote abnormal actin polymerization into a specialized apparatus that propels the microbes across the membrane and through the cytoplasm of infected cells.23 In other cases an overly active cytoskeleton participates in rapid proliferation of cancerous cells and in moving them to establish metastases; hence, targeting cytoskeletal mechanisms is a promising option for complementary therapy in some patients with the disease.28 The same is true against viruses that use normal intracellular transport as invading pathway. Yet, as discussed below (Sections 1.24.5.1–1.24.5.3), the most prevalent pathological conditions related to cytoskeletal systems are due to alterations of the constituent proteins themselves deither by mutation, misfolding, or anomalous aggregationd, leading to faulty operation of the respective mechanisms. Furthermore, autoimmune responses also target the cytoskeleton, like in systemic lupus erythematosus, coeliac disease, and the WiskottAldrich syndrome.24

370 1.24.5.1

Structural Organization of CellsdThe Cytoskeleton Actin Microfilament-Related Diseases

Muscular dystrophy, a progressive weakness of skeletal muscle that eventually causes premature death, is due to a mutation in the gene that encodes dystrophin, a protein involved in the attachment of actin microfilaments to the cell membrane25 (see also Section 1.24.2.4). Mutations in proteins closely associated to dystrophin are linked to milder cases of muscular dystrophy. In addition, mutations in some myosins have been related to human deafness, for they impair muscle-mediated regulation of stimulus transmission across the middle ear. Mutations in tropomyosin cause loss of function and are associated with skeletal muscle myopathy. Some of them affect the site of actin-tropomyosin interaction and thus interfere with myosin-actin coupling. Some neurodegenerative diseases share the feature of being accompanied by the presence of anomalous cofilin-rich, rod-shaped bundles of actin microfilaments in the cytoplasm of affected cells. Other ill-defined cytoplasmic inclusions dsuch as irregular aggregates called aggresomes, or paracrystalline lattices known as Hirano bodies, typical of the Alzheimer’s and Creutzfeldt–Jakob diseasesd may also be actin–cofilin complexes at various stages of defective formation.24

1.24.5.2

Microtubule-Related Diseases

Most outstanding in this class are Alzheimer’s disease and other taupathies26 that is, severe and progressive neurodegenerative diseases in which the microtubule-associated protein tau appears forming aberrant filamentous aggregates, usually called neurofibrillary tangles, within neurons. Such structures are produced through a sequence of modifications of tau, including phosphorylation at various sites, fragmentation by proteolysis, and the formation of intracellular and extracellular aggregates of the protein fragments. Brain tissue of patients with Alzheimer’s disease also shows structures containing actin and its associated protein cofilin. Other taupathies include progressive supranuclear palsy, Pick’s disease, cortico-basal ganglionic degeneration, fronto-temporal dementia, and Parkinsonism linked to chromosome 17. On the other hand, there are numerous diseases involving brain development, memory, learning and metabolic disorders, where kinesins are affected.20

1.24.5.3

Intermediate Filament-Related Diseases

Given the various families of intermediate filaments, their pathologies are likewise diverse, so only the most significant will be mentioned here. Many disorders are attributed to mutations in keratin genes.27 Thus, for example, palmoplantar keratoderma has been traced to mutations in keratins 1, 9, 10, and 16, while hypertrophic nail dystrophy is caused by mutations in keratins 6a, 6b, and 17. Some types of cataracts derive also from mutations of intermediate filament genes. Some neurodegenerative afflictions, such as fatal amyotrophic lateral sclerosis and Alexander disease, or the Charcot-Marie-Tooth condition and certain forms of Parkinsonism, result from mutations to genes encoding neurofilament or glial fibrillary acidic proteins28. In turn, a few myopathies are due to mutated desmin.

Acknowledgments The authors are especially grateful to Armando Pérez-Rangel for the quality artwork, and to Lourdes Ruiz for the micrographs in Fig. 8 and help in reviewing the manuscript.

See Also: 1.14 Design of Culture Media.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

Alberts, B.; Johnson, A.; Lewis, J.; et al. Molecular Biology of the Cell, 6th ed.; Garland Science: New York, NY, 2015. Frixione, E. Recurring Views on the Structure and Function of the Cytoskeleton: a 300-year Epic. Cell Motil Cytoskeleton 2000, 46, 73–94. Lodish, H.; Berk, A.; Kaiser, C. A.; et al. Molecular Cell Biology, 8th ed.; W.H. Freeman: New York, NY, 2016. Erickson, H. P. The Discovery of the Prokaryotic Cytoskeleton: 25th Anniversary. Mol. Biol. Cell 2017, 28, 357–358. Wickstead, B.; Gull, K. The Evolution of the Cytoskeleton. J. Cell Biol. 2011, 194, 513–525. Carballido-Lopez, R. The Actin-like MreB ’Cytoskeleton’. In Bacillus: Cellular and Molecular Biology; Graumann, P. L., Ed., 3rd ed.; 2017; pp 223–261. Oliva, M. A.; Martin-Galiano, A. J.; Sakaguchi, Y.; Andreu, J. M. Tubulin Homolog TubZ in a Phage-encoded Partition System. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 7711–7716. Charbon, G.; Cabeen, M. T.; Jacobs-Wagner, C. Bacterial Intermediate Filaments: In Vivo Assembly, Organization, and Dynamics of Crescentin. Genes Dev. 2009, 23, 1131–1144. Pollard, T. D. Actin and Actin-binding Proteins. Cold Spring Harbor Perspect. Biol. 2016, 8, a018226. Coltharp, C.; Xiao, J. Beyond Force Generation: Why Is a Dynamic Ring of FtsZ Polymers Essential for Bacterial Cytokinesis. Bioassays 2017, 39, 1–11. Herrmann, H.; Aebi, U. Intermediate Filaments: Structure and Assembly. Cold Spring Harbor Perspect. Biol. 2016, 8, a018242. Kueh, H. Y.; Mitchison, T. Structural Plasticity in Actin and Tubulin Polymer Dynamics. Science 2009, 325, 960–963. Wieczorek, M.; Bechstedt, S.; Chaaban, S.; Brouhard, G. J. Microtubule-associated Proteins Control the Kinetics of Microtubule Nucleation. Nat. Cell Biol. 2015, 7, 907–916. Blanchoin, L.; Boujemaa-Paterski, R.; Sykes, C.; Plastino, J. Actin Dynamics Architecture and Mechanics in Cell Motility. Physiol. Rev. 2014, 94, 235–263. Hodge, R. G.; Ridley, A. J. Regulating Rho GTPases and Their Regulators. Nat. Rev. Mol. Cell Biol. 2016, 17, 496–510.

Structural Organization of CellsdThe Cytoskeleton 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30.

371

Chang, L.; Goldman, R. D. Intermediate Filaments Mediate Cytoskeletal Crosstalk. Nat. Rev. Mol. Cell Biol. 2004, 5, 601–613. Ingber, D.; Wang, N.; Stamenovi, D. Tensegrity, Cellular Biophysics, and the Mechanics of Living Systems. Rep. Prog. Phys. 2014, 77, 046603. Holthöfer, B.; Windoffer, R.; Troyanovsky, S.; Leube, R. E. Structure and Function of Desmosomes. Int. Rev. Cytol. 2007, 264, 65–116. Hartman, M. A.; Spudich, J. A. The Myosin Superfamily at a Glance. J. Cell Sci. 2012, 125, 1627–1632. Hirokawa, N.; Tanaka, Y. Kinesin Superfamily Proteins (KIFs): Various Functions and Their Relevance for Important Phenomena in Life and Diseases. Exp. Cell Res. 2015, 334, 16–25. Garzon-Coral, C.; Fantana, H. A.; Howard, J. A Force-generating Machinery Maintains the Spindle at the Cell Center during Mitosis. Science 2016, 352, 1124–1127. Chugh, P.; Clark, A. G.; Smith, M. B.; et al. Actin Cortex Architecture Regulates Cell Surface Tension. Nat. Cell Biol. 2017, 19, 689–697. Frankel, G.; Phillips, A. D. Attaching Effacing Escherichia coli and Paradigms of Tir-triggered Actin Polymerization: Getting off the Pedestal. Cell Microbiol. 2008, 10, 549–556. Todoric, K.; Koontz, J. B.; Mattox, D.; Tarrant, T. K. Autoimmunity in Immunodeficiency. Curr. Allergy Asthma Rep. 2013, 13, 361–370. Wallace, G. Q.; McNally, E. M. Mechanisms of Muscle Degeneration, Regeneration, and Repair in the Muscular Dystrophies. Annu. Rev. Physiol. 2009, 71, 37–57. Bamburg, J. R.; Bloom, G. S. Cytoskeletal Pathologies of Alzheimer Disease. Cell Motil Cytoskeleton 2009, 66, 635–649. Toivola, D. M.; Boor, P.; Alam, C.; Strnad, P. Keratins in Health and Disease. Curr. Opin. Cell Biol. 2015, 32, 73–81. Eira, J.; Silva, C. S.; Sousa, M. M.; Liz, M. A. The Cytoskeleton as a Novel Therapeutic Target for Old Neurodegenerative Disorders. Prog. Neurobiol. 2016, 141, 61–62. Jones, J. C.; Kam, C. Y.; Harmon, R. M.; et al. Intermediate Filaments and the Plasma Membrane. Cold Spring Harb. Perspect. Biol. 2017, 9, a025866. Maravillas-Montero, J. L.; Santos-Argumedo, L. The Myosin Family: Unconventional Roles of Actin-dependent Molecular Motors in Immune Cells. J. Leukoc. Biol. 2012, 91, 35–46.

Relevant Websites https://www.youtube.com/watch?v¼uwnw4vg9I5Q. http://cellix.imba.oeaw.ac.at/. http://biochemweb.net/cytoskeleton.shtml. http://www.cellmigration.org/index.shtml. http://home.uni-leipzig.de/pwm/web/?section¼introduction&page¼cytoskeleton. http://pfam.sanger.ac.uk – Wellcome Trust Sanger Institute; Family: Actin (PF00022).

Viruses Produced From Cellsq

1.25

KM Coombs, University of Manitoba, Winnipeg, MB, Canada; Manitoba Centre for Proteomics and Systems Biology, Winnipeg, MB, Canada; and Manitoba Institute of Child Health, Winnipeg, MB, Canada © 2017 Elsevier B.V. All rights reserved. This is a reprint of K.M. Coombs, Viruses Produced From Cells, Reference Module in Life Sciences, Elsevier, 2017.

1.25.1 Introduction 1.25.2 Cell Culture 1.25.3 Types of Growth Flasks 1.25.4 Parameters of Virus Growth 1.25.4.1 Cell Type 1.25.4.2 Initial MOI 1.25.4.3 Cell Density 1.25.4.4 Media pH 1.25.4.5 Incubation Temperature 1.25.4.6 Media Replenishment During Infection 1.25.4.7 Flask Size 1.25.5 Virus Purification 1.25.5.1 Gradient Ultracentrifugation 1.25.5.2 Ultrafiltration 1.25.5.3 Chromatography 1.25.6 Future Perspectives References Relevant Website

372 373 373 377 377 377 378 378 379 379 379 379 379 380 381 381 382 382

Glossary Cytopathic effect Detectable morphological changes that take place in cells as a result of virus infection and often involves rounding up, detachment from solid substrates, and eventual death. Multiplicity of infection (MOI) A number that represents the average number of viruses infecting each cell; determined by dividing total numbers of viruses by total numbers of cells. Particle to plaque forming unit (PFU) ratio The number of total particles per individual infectious virus. Most animal viruses are defective and not infectious. For example, a virus with a particle to PFU ratio of 100 has one infectious virion per 100 particles, or specific infectivity of 1%. Plaque forming unit (PFU) A measure of the number of infectious virus particles in a sample, determined by incubating virus samples with susceptible cells and counting numbers of lesions in the cell monolayer. Virion The extracellular form of a virus with a defined shape and size, often referred to simply as virus.

1.25.1

Introduction

Although the concept of an infectious virus is just over 100 years old, diseases caused by some of these agents (ie, rabies and polio) appear to have been known in Egypt about 4000 years ago. Anton van Leeuwenhoek’s work about 350 years ago allowed visualization of bacteria, and Louis Pasteur established his disease-causing properties about 200 years later. However, it became appreciated toward the end of the 19th century that a few disease-causing agents were small enough to pass through filters known to block bacteria. Thus, the term “virus” (latin for poison) was coined to describe these filterable toxins. However, viruses were soon shown to differ from poisons and toxins, which can be diluted when serially passaged from one host to another, whereas viruses multiply. Indeed, this capacity of viruses to multiply, often rapidly and exponentially, can be exploited by scientific researchers. Viruses are among the simplest and smallest of currently known living organisms. There is some debate as to whether they should be considered living. Viruses generally exist in two forms. The actively replicating virus inside an infected cell is the form that is usually considered alive. The extracellular form of the virus, known as the “virion,” is analogous to a seed or spore. The virion

q

Change History: March 2016. K.M. Coombs updated the Section “Future Perspectives,” made minor changes to Sections “Introduction” and “Parameters of Virus Growth” and added references to the “References” section.

372

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.09026-9

Viruses Produced From Cells

373

is a stable crystalline structure of defined shape and size, whose primary function is to protect the viral genetic material until the viral genome enters a suitable host cell. There is considerable variability in the size and complexity of virions. Parvovirus virions are among the simplest of animal viruses, being composed of a few copies of a few proteins and a single piece of DNA nucleic acid. They are also among the smallest, with virion sizes in the range of 25 nm diameter. Most other animal viruses have larger and/ or more complex virion structures. For example, poxviruses, with a virion size of about 200
300 nm, are among the largest of currently known animal viruses, being composed of about 100 proteins and also containing, in addition to the DNA genome, a host-derived lipid envelope. Some recently discovered fungal viruses, such as Mimivirus and Pandoravirus, are much larger. No virus is capable of growing by itself. All must make use of macromolecular building blocks (amino acids, nucleotides, and in some cases, lipids) and employ enzymes found within living cells. Thus, all viruses are obligate intracellular parasites. Therefore, deliberate production of virions from cells for scientific research or for therapeutic purposes also requires considerable expertise in appropriate cell culture methods and optimization. Where a suitable in vitro cell culture system exists, infection is generally most easily detected by the capacity of the virus to induce cytopathic effect in the cells. Such conditions are also exploited to measure the amounts of infectious virus, which is most easily carried out by plaque assays. However, these types of assays suggest that most preparations of animal viruses contain noninfectious particles. The relationship between numbers of infectious (as measured by plaque assay) and total particles (as measured by direct electron microscopic counts), also known as the particle to plaque forming unit (PFU) ratio, can vary from 10:1 to >10,000:1. This may have implications for the amounts of virus or virions one wishes to produce for any particular experimental protocol. Recent work and technical innovations aimed at optimizing cell and virus growth are the subject of this chapter.

1.25.2

Cell Culture

Theoretically, almost any type of nucleated cell should be capable of some growth in culture, provided the appropriate nutrients and conditions are provided. Unlike many bacteria, which can grow on simple agar plates, animal cells require a complex mix of components in order to survive in culture. Many of these requirements were determined empirically in the mid-20th century. Cells in culture require balanced salts, a variety of nutrients, including most amino acids, several vitamins and coenzymes, and an energy source such as glucose. Each of approximately 20–100s of components must be present in a precise concentration. In addition to the above, many animal cells require other factors we do not know, which are provided by inclusion of 5–20% serum, and many cells also require incubation in an atmosphere of 5–10% CO2 to maintain correct pH in bicarbonate buffered media. The precise formulations and required supplements vary among the several hundred media recipes that have been designed for specific purposes, and there are a growing number that are serum free, especially in the interest of reducing animal-derived components for human therapy. Most cells of animal origin will assume one of two distinct morphologies (epithelial or fibroblastic, see Fig. 1; a few cells, such as nerve cells, are exceptions). These morphologies become less obvious as cells become confluent and occupy the entire surface area (Fig. 1, bottom). Most cells are also anchorage dependent; they require calcium ions and a flat surface on which to attach and grow. Thus, most animal cells will not grow on untreated plastic bacterial plates and will soon die (hybridoma and many Chinese hamster ovary cells are an exception). Even under optimum conditions, animal cells tend to grow relatively slowly. Many divide approximately once per day. Animal cell culture media is considerably richer than most bacterial culture media. Thus, under unfortunate conditions where bacteria are inadvertently introduced into an animal cell culture, the bacteria will quickly outgrow the animal cells. This might be prevented by including antibiotics in animal cell media, but most investigators purposefully do not include antibiotics because inclusion can lead to development of resistant organisms and hide mycoplasma contamination, the cell phenotype may change after long-term exposure, and it is simply human nature to be less careful if it is assumed that cultures cannot be contaminated. Thus, scrupulous aseptic techniques must be followed. Most primary cells will divide a very limited number of times before they cease to replicate and (generally) die. Thus, many investigators attempt to make use of continuous cell lines for long-term work. Continuous cell lines are immortalized and theoretically have unlimited division potential. However, most such cell lines are immortalized by an oncogenic event. For example, HeLa cells contain papillomavirus oncogenes and an activated telomerase enzyme. Presence of oncogenes in cell culture may need to be considered in some applications, such as human therapy, where contamination of the final product for human usage needs to be avoided.

1.25.3

Types of Growth Flasks

Adherent cells generally grow well on glass surfaces. Because of the problems of bacterial contamination (described in earlier section), reusable glassware needs to be carefully cleaned and sterilized. Many laboratories have switched to using disposable plastic culture vessels to avoid processing problems. Native plastic is not conducive to animal cell growth so most manufacturers treat the internal surfaces the cells will be grown on to promote cell growth. Most manufacturers produce a variety of plastic tissue-culture dishes and flasks. Culture dishes are manufactured in a variety of sizes, including 35, 60, 100, and 150 mm diameters. These types of vessels are generally used for terminal experiments because it is more difficult to maintain long-term sterility in such containers. Thus, long-term cell maintenance is usually most conveniently carried out in flasks with sealable caps, usually left loose to allow gas exchange. Some newer flasks have gas-permeable filters built into the cap so caps may be fully closed. Most common tissue culture flask sizes include 25, 75, 150, and 225 cm2 (often referred to as T25, T75, T150, and T225 flasks, respectively). The precise flask shape and configuration, and removable cap format and color, vary from manufacturer to manufacturer (Fig. 2). Most flasks we have worked with behave similarly.

374

Viruses Produced From Cells

A

B

C

D

E

F

Figure 1 (A, C, and E) Epithelial and (B, D, and F) fibroblast cells at low (10%, top), medium (40%, middle), and high (100%, bottom) levels of confluency. Micrographs at same magnification, with indicated scale bar.

The numbers of flasks needed for a specific experiment, that is, to grow sufficient cells for various needs, are determined by what the virus will be used for (some examples are listed in Table 1). For example, to amplify a small amount of virus for molecular biologic purposes, and depending upon how well the virus grows in culture, one may only require the number of cells that can be grown in a 35-mm dish, or a single T25 flask. Similarly, enough virions may be produced in these small containers to allow electron microscopic visualization, particularly if the produced virus is concentrated before application to the electron microscope grid. Finally, if the virus grows to the range of 108 PFU/mL (z 1010 particles/mL if the particle to PFU ratio is 100:1), it may also be possible to purify sufficient viral nucleic acid from these sized flasks to detect it by gel electrophoresis. Other needs may require more and/or bigger flasks and are determined by knowledge of how well the virus grows. For example, if it is decided that 1014 viral particles are required to initiate crystallization trials, and if the virus of interest is known to routinely be produced at 1010 particles/ mL, it can be calculated that a minimum of 10,000 mL of viral-infected culture is needed, which corresponds to approximately 400 T150 flasks, assuming that losses during purification are negligible. If the virus grows more poorly, as many do, or it is anticipated that a substantial portion will be lost during purification, as routinely happens, then more flasks will be required. Considerations of exa-scale (1018) production of viruses, such as for human therapy, have been discussed in [1] and necessitate means to conveniently and easily grow very large numbers of cells. Working with large numbers of flasks, as suggested in the previous paragraph, is a laborious process, and the risk of contamination increases dramatically. Thus, a variety of strategies and innovative technologies have been recently developed in order to allow investigators to work with the large numbers of cells required for some applications but in a manner that reduces the

Viruses Produced From Cells

375

A

B

C

D

~0.2 mm Figure 2 (A) Various common formats of tissue culture vessels. Different sized disposable plastic formats (dishes and T-flasks) are above the ruler scale and a variety of re-usable glass vessels are to the left and right of the ruler. From middle left, clockwise: a Wheaton 250 mL stirring flask for suspension cultures; T25 flasks (Corning/Costar, orange caps; Falcon, blue caps; Greiner, red caps); P60 dishes; T75 flasks; P100 dishes; T150 flasks; and small (11 cm diameter
32 cm high) roller bottle. (B) The Corning HYPERFlask, which contains 10 layers for cell growth (image provided by Corning, Inc., with permission). (C) A Corning CellStack, consisting of multiple layers of 636 cm2 trays (image provided by Corning, Inc., with permission). (D) A 1 L Bellco stirring flask; backdrop is a photomicrograph of Solohill Hillex polystyrene microcarriers, covered with Vero cells. Image provided by Solohill Engineering, Ann Arbor, MI, with permission. Table 1

Virion quantities needed for various purposes

Purpose

Amount a

Genome amplification by PCRb Visualization by electron microscopy Detection of nucleic acid by electrophoresis Determination of protein copy numbers per virion Crystallization for structure determination Therapeutics (mass vaccination)

100–102 102–106 107–1010 1010–1012 1013–1016 1015–1018

a

Approximate range only; dependent upon virion, nucleic acid and protein size, copy number, and detection method. Polymerase chain reaction if DNA genome; reverse transcriptase-PCR if RNA.

b

numbers of vessels and therefore, manipulations needed for larger scale processes. The simplest strategy to produce more cells and virus might be to increase the size of the flasks used. However, this is limited by the sizes of available flasks and whether virus grows as efficiently in different sized flasks (see Section 1.25.4.7). Roller bottles (Fig. 2) have a larger useable surface and are one strategy to reduce the numbers of vessels. A medium-sized roller bottle (11 cm diameter
47.5 cm height¼1330 cm2 surface area) provides the same surface area as nine T150 flasks. However, use also requires additional specialized equipment such as a means of slowly and continuously rotating the bottle. In addition, depending upon whether the roller can fit in a CO2 incubator or not, appropriate media that either is compatible with CO2 or that does not require CO2 if the roller apparatus does not fit in a standard CO2

376

Viruses Produced From Cells

incubator are required. If the roller apparatus and bottles will be maintained in an incubator, then there are limits as to how many bottles will fit because they have a relatively large volume-to-surface ratio. Roller bottles have been used for decades in many applications. In efforts to maintain relatively simple static cultures that require no more than an appropriate incubator, and in attempts to dramatically increase the surface area to volume ratios of various formats to allow more efficient incubator space usage, some manufacturers have recently created multilayered flasks or cell factories. For example, Corning, Inc. makes a HYPERFlask, which features 10 thin separated layers of gas-permeable membrane to allow efficient gas exchange, all in a T175 footprint, providing 1750 cm2 of total surface area (Fig. 2B). Nunc, Inc. makes a cell factory, which are narrow stackable trays, each with a surface area of about 630 cm2, so stacks of 10 such trays provide >6000 cm2 area for cell growth. Corning, Inc. makes a similar product, called CellStack (Fig. 2C). Similarly, products such as the Corning CellCube consist of multiple layers of parallel plates to increase available surface area in a small volume. These formats have been used successfully to grow cells to high concentrations. Reports on virus production are inconsistent. Some reports indicate excellent production, comparable to production from regular single-layer flasks. However, other users report that less virus is often produced per unit area in these configurations than in standard flasks, which is attributed to poorer gas exchange and, therefore, suboptimal cell growth. This suboptimal virus growth has been overcome in some instances by forced culture gassing. Efficient gas exchange is most easily carried out in cultures that are agitated, rather than in static cultures. Thus, in the case of large-scale bacterial growth for protein expression, the bacteria are cultured in shaker flasks that can be up to 5–10 L or in bioreactors than can be up to 100,000 L. A similar strategy will work for animal cells and viruses, if the particular cells grow in suspension. For example, reovirus may be grown in mouse L929 cells, which can be grown in suspension cultures. A 1 L suspension culture contains as many cells as 40 T150 flasks. This technique has been used to amplify some viruses, such as reovirus, in up to 10 L volumes. In addition, a number of other viruses have been grown in suspension cultured cells, including adenoviruses in HEK-293 and PER.C6 cells [2], and retroviruses in HEK-293 cells. Unfortunately, most animal cells are anchorage dependent and thus, cannot be grown in suspension this way. In addition, one study, using a direct comparison of infectious adenovirus yields in HEK-293 cells grown under various conditions, showed 2000–7000 PFU/cell produced from suspension growth in serum-free media, 4000–7000 PFU/cell from microcarriers in serum-supplemented media, and up to 10,000 PFU/cell in standard T-flasks in serum-supplemented media [3]. Very large scale cell growth, and virus amplification, such as for mass therapeutic use, is now most often performed in bulk microcarrier cultures (Fig. 2D). Such cultures offer multiple benefits. For recent review, the reader is referred to Ref. [4]. Cells are grown on small beads, allowing growth of anchorage-dependent cells, and the beads may be kept in suspension by low-speed stirring. In some cases, a few hundred cells may occupy the surface of each bead, and beads may be used in culture at up to 2–5
104 beads per mL (z4–10
106 cells/mL); thus, a 5-L microcarrier culture contains approximately the same number of cells as 600–1500 T150 flasks. This type of cell-culturing format also requires additional specialized equipment – a slow-speed stirrer, so may not be economically reasonable for routine small- or medium-scale virus preparations. Microcarrier beads are produced by several manufacturers and are produced in various sizes and chemical configurations, including native glass beads, glass-coated plastic beads, solid beads, macroporous beads, and beads whose surfaces are modified in any of several ways. Choosing the best bead is often empiric. A large, and growing, number of viruses have been grown to high titers in various microcarrier formats, including foot and mouth disease virus in BHK-21 cells on glass beads, poliovirus in Vero cells, measles virus in Vero cells on dextran-based Cytodex-1 beads, HIV from HT-29 cells, hepatitis A virus, various herpes viruses in several different types of cells on glass and Cytodex-1 beads, hepatitis B virus in HepG2 cells on Cytodex-3 beads, dengue virus in COS-1, or in mosquito C6/36 and Vero cells on a variety of Cytodex beads, reovirus in Vero cells on Cytodex-1, rabies virus in BHK-21 and Vero cells on Cytodex beads, vaccinia virus in HeLa cells on Cytodex-3 beads, and influenza virus in Madin Darby Canine Kidney (MDCK) cells. Although most cells tested on solid and macroporous microcarriers appear to grow similarly, and often to higher cell densities on the macroporous beads, virus production is often higher from cells grown on the solid substrates. Microcarriers in stirred culture are scaleable to >10,000 L and often used in the pharmaceutical industry. In addition to the need for additional specialized equipment (described above), other problems with this process include cell damage and bead fracturing from bead collisions and shear stresses. Mammalian cells are susceptible to such stresses. Different cells appear to be differentially susceptible to shear stress. A variety of new novel technologies have been recently developed to attempt to reduce these stresses and generate higher cell concentrations. Packed bed reactors allow cell growth approaching 100
than that found in typical cell suspension and microcarrier cultures (for review, see Ref. [5]). Packed bed reactors generally consist of small inert and nontoxic glass, ceramic, polyester, polyurethane, or polyvinyl beads, strips, fibers, or chips that provide very high surface-to-volume ratios and on which adherent cells are allowed to grow. Cell densities of 1
108–5
108 mL1, approaching about one-half the density of a solid organ, have been reported [5]. Despite these impressive cell concentration numbers possible with packed bed reactors, virus growth is generally lower, or only marginally better on a per volume basis, and generally is lower on a per cell basis. Similarly, cells such as HEK-293, which will grow in suspension, can be allowed to aggregate, thereby increasing cell numbers. Some reports indicate that virus production is not improved in such clump cultures (reviewed in Ref. [6]), but others, which employ media perfusion, report dramatic virus production. The capacity to replace media, for example, by perfusion, generally seems to help maintain cell health and increase virus production (see below). Additional techniques have been developed to reduce contamination risks, including plastic disposable Wave bags [7, 8] and other presterilized single-use disposable configurations. Such formats may also be constructed to reduce crucial component absorption (ie, fluorinated ethylene propylene to reduce cholesterol absorption). Wave bags are presterilized plastic bags that are provided

Viruses Produced From Cells

377

empty or prepacked with any of various cell-growing matrices (beads, disks, etc.). Empty bags may be filled with suspension cells; matrix-containing bags may be filled with adherent cells. The bags are then placed on a rocking platform in an incubator and incubated, with rocking, to generate waves that both agitate and aerate the cells. A variety of manufacturers also make articulated bottles, which, with an appropriate mechanism, will perfuse cultures. Examples include Dunn Labortechnik GmbH’s BelloCell, a disposable plastic bottle that, when used with a bellows compressor platform, induces media and air flow across porous matrices on which cells grow, allowing up to 15,600 cm2 surface area in a 500-mL size format. This system has been reported to produce higher yields of Japanese encephalitis virus from Vero cells than microcarrier spinner flasks. New Brunswick Scientific makes a similar product, called FibraStage, which has 500 mL bottles filled with the popular Fibra-Cel polyester fiber/polypropylene disk system used in packed-bed reactors [5].

1.25.4

Parameters of Virus Growth

In addition to absolute numbers of available cells and cell concentration (above), there are a large number of other parameters that may contribute to optimal virus growth in animal cell culture, including cell type, initial multiplicity of infection (MOI), cell density, media pH, incubation temperature, whether media is replenished during infection, and even flask size.

1.25.4.1

Cell Type

As indicated earlier, most animal cells assume either an epithelial or fibroblast morphology in cell culture. As a general rule, any given virus will grow better in one cell morphology than in another. There also is great variability in how well viruses will grow in different cells of similar morphology. For example, retroviral vectors grow better in human HEK-293 cells than in mouse NIH 3T3 cells [6], BHK-21 and Vero cells produce Peste de petits ruminants virus more quickly than HEK-293 cells, and CHO K1 and MRC-5 are not suitable for growth of this virus, and reovirus grows better in mouse L929 cells than in many other cells, including NIH 3T3, which is generally attributed to the presence or absence of activated cellular oncogenes. Thus, it often is a good idea for investigators wishing to optimize their virus’ growth to empirically determine what cell types support optimal growth of each virus of interest.

1.25.4.2

Initial MOI

The relative numbers of viruses and cells is referred to as MOI. Statistical calculations (Poisson’s distribution) suggest that if cells and virus are mixed at a 1:1 ratio (MOI¼1), many cells are actually infected with multiple viruses and, therefore, only 63.2% of cells are initially infected. An MOI of 5 PFU/cell is needed to obtain an initial infection efficiency of 99.3%. This suggests more is better. However, there also are diminishing returns. For example, using 5
as much virus only results in a theoretical infection increase of 57% (99.3–63.2/63.2%). As indicated earlier, one of the remarkable features of viruses is that their numbers can dramatically increase through infection in a relatively short period of time. Therefore, it often is advantageous to infect cultures with a small amount of virus, conserving inoculum stocks, and allow the virus to grow through two or more rounds of replication, which results in the entire cell culture being eventually infected. An example of reovirus production from cultures of mouse L929 cells initially infected at various MOIs is shown in Fig. 3. As is seen, cultures infected at higher MOIs produce more virus at early time points;

Figure 3 Growth profiles of reovirus, serotype 1 Lang (T1L), in adherent L929 cells. Cells were infected at various multiplicities of infection (indicated), harvested at various time points, and amount of infectious virus produced measured by plaque assay.

378

Viruses Produced From Cells

however, cultures infected at substantially lower MOI still produce large amounts of virus, albeit at slightly later times. Similar results have been reported by others for many viruses, including adenovirus [9], influenza virus [7], vesicular stomatitis virus and Sindbis viruses. Thus, stock viruses are usually generated by set procedures, and wherever possible, low MOIs are chosen and used because repeated high MOI passages lead to accumulation of mutations, including defective interfering particles. For example, herpes viruses are usually amplified by infecting cells at an initial MOI of 2 million cells/mL) should be considered. This would include isolating better clones that could support high cell density culture, medium development to support high productivity after transfection, and feeding strategies to replenish the nutrient components consumed during production. To make the process cost-effective and with increased transfection efficiency, the research work shifted to find novel reagents and diverted to modify the polycations by the addition of chemicals, by changing the chain lengths, or by changing the head groups by attaching multiple positive charges. It was shown that polyamine-based cation lipids had much higher efficiency than monocationic lipids. New approaches to develop cost-effective, optimized transfection systems are showing promising results.

1.26.4

Conclusion

In conclusion, stable cell line generation for r-protein for drug screening and preclinical testing is time-consuming. Transfection at any scale is a faster and economical alternative for the production of a desired material. Since the demonstration of transfection in a serum-free medium using suspension cell cultures by Schlaeger and Christensen,2 the work continued using CHO and HEK293 cells29 but also insect cells.30 Operation of suspension cultures and the removal of serum from the medium for transfection were the major improvements in transient r-protein production and facilitated the downstream processing. Polycation–DNA condensation methods, such as PEI–DNA condensation method, are applicable for the production of r-proteins at large scale in a serum-free medium using suspension cell cultures. Therefore, PEI has been preferred among the researchers to further the advancement of transfection technology. For cell lines highly refractory to transfection such as primary cell lines, viral transduction using recombinant nonreplicating adenoviruses or lentiviruses remains the most effective approach for transgene delivery and expression. Delivering of cost-effective and rapid production in serum-free suspension culture conditions has been one of the challenges faced in biopharmaceutical industries. Transient cell transfection technology met the challenges of delivering sufficient quantities of recombinant human therapeutic proteins in a cost-effective manner within a couple of weeks. In addition, progress in large-scale transfection technology has been successfully implemented in efficient manufacturing of viral vectors for cell and gene therapy.

390

Cell Transfection

See Also: 1.25 Viruses Produced from Cells.

References 1. Alexander, H. E.; Koch, G.; Mountain, I. M.; Van Damme, O. Infectivity of Ribonucleic Acid from Poliovirus in Human Cell Monolayers. J. Exp. Med. 1958, 108, 493–506. 2. Schlaeger, E. J.; Christensen, K. Transient Gene Expression in Mammalian Cells Grown in Serum-free Suspension Culture. Cytotechnology 1999, 30, 71–83. 3. Steger, K.; Brady, J.; Wang, W.; et al. CHO-S Antibody Titers >1 Gram/Liter Using Flow Electroporation-mediated Transient Gene Expression Followed by Rapid Migration to High-yield Stable Cell Lines. J. Biomol. Screen 2015, 20, 545–551. 4. Yamashita, Y.; Shimada, M.; Tachibana, K.; et al. In Vivo Gene Transfer into Muscle via Electro-sonoporation. Hum. Gene Ther. 2002, 13, 2079–2084. 5. Yang, N. S.; Burkholder, J.; Roberts, B.; et al. In Vivo and in Vitro Gene Transfer to Mammalian Somatic Cells by Particle Bombardment. Proc. Natl. Acad. Sci. U. S. A. 1990, 87, 9568–9572. 6. Antkowiak, M.; Torres-Mapa, M. L.; Stevenson, D. J.; et al. Femtosecond Optical Transfection of Individual Mammalian Cells. Nat. Protoc. 2013, 8, 1216–1233. 7. Yudin, M. A.; Bykov, V. N.; Nikiforov, A. S.; et al. Study of the Efficiency of the Hydroporation for Delivery of Plasmid DNA to the Cells on the Model of Toxic Neuropathy. Bull. Exp. Biol. Med. 2018, 164, 798–802. 8. Smolders, S.; Kessels, S.; Smolders, S. M.; et al. Magnetofection Is Superior to Other Chemical Transfection Methods in a Microglial Cell Line. J. Neurosci. Meth. 2018, 293, 169–173. 9. Novickij, V.; Grainys, A.; Novickij, J.; Markovskaja, S. Irreversible Magnetoporation of Micro-organisms in High Pulsed Magnetic Fields. IET Nanobiotechnol. 2014, 8, 157–162. 10. Sharei, A.; Cho, N.; Mao, S.; et al. Cell Squeezing as a Robust, Microfluidic Intracellular Delivery Platform. JoVE 2013, e50980. 11. Kim, T. K.; Eberwine, J. H. Mammalian Cell Transfection: the Present and the Future. Anal. Bioanal. Chem. 2010, 397, 3173–3178. 12. Jordan, M.; Schallhorn, A.; Wurm, F. M. Transfecting Mammalian Cells: Optimization of Critical Parameters Affecting Calcium-phosphate Precipitate Formation. Nucleic Acids Res. 1996, 24, 596–601. 13. Pandey, A. P.; Sawant, K. K. Polyethylenimine: A Versatile, Multifunctional Non-viral Vector for Nucleic Acid Delivery. Mater. Sci. Eng. C Mater. Biol. Appl. 2016, 68, 904–918. 14. Koynova, R.; Tenchov, B. Enhancing Nucleic Acid Delivery, Insights from the Cationic Phospholipid Carriers. Curr. Pharmaceut. Biotechnol. 2014, 15, 806–813. 15. Zhi, D.; Bai, Y.; Yang, J.; et al. A Review on Cationic Lipids with Different Linkers for Gene Delivery. Adv. Colloid Interface Sci. 2018, 253, 117–140. 16. Sharon, D.; Kamen, A. Advancements in the Design and Scalable Production of Viral Gene Transfer Vectors. Biotechnol. Bioeng. 2018, 115, 25–40. 17. Lee, C. S.; Bishop, E. S.; Zhang, R.; et al. Adenovirus-mediated Gene Delivery: Potential Applications for Gene and Cell-based Therapies in the New Era of Personalized Medicine. Genes Dis. 2017, 4, 43–63. 18. Kamen, A.; Henry, O. Development and Optimization of an Adenovirus Production Process. J. Gene Med. 2004, 6 (Suppl. 1), S184–S192. 19. Schiedner, G.; Morral, N.; Parks, R. J.; et al. Genomic DNA Transfer with a High-capacity Adenovirus Vector Results in Improved in Vivo Gene Expression and Decreased Toxicity. Nat. Genet. 1998, 18, 180–183. 20. Dormond, E.; Chahal, P.; Bernier, A.; et al. An Efficient Process for the Purification of Helper-dependent Adenoviral Vector and Removal of Helper Virus by Iodixanol Ultracentrifugation. J. Virol. Methods 2010, 165, 83–89. 21. Berggren, W. T.; Lutz, M.; Modesto, V. General Spinfection Protocol. In StemBook; 2008. 2008. Cambridge (MA). 22. Ansorge, S.; Lanthier, S.; Transfiguracion, J.; et al. Development of a Scalable Process for High-yield Lentiviral Vector Production by Transient Transfection of HEK293 Suspension Cultures. J. Gene Med. 2009, 11, 868–876. 23. Aiuti, A.; Cossu, G.; de Felipe, P.; et al. The Committee for Advanced Therapies’ of the European Medicines Agency Reflection Paper on Management of Clinical Risks Deriving from Insertional Mutagenesis. Hum. Gene Ther. Clin. Dev. 2013, 24, 47–54. 24. Tyzack, E.; Pettman, G.; Daramola, L. An ‘Industry First’ 500L Bioreactor CHO Transient Culture: Development of Large Scale Transient Expression Capabilities. In 25th European Society for Animal Cell Technology Meeting: Cell Technologies for Innovative Therapies, 3. Lausanne, Switzerland, 14–17 May 2017. BioMed Central; 2017. 2017. 25. Geisse, S.; Henke, M. Large-scale Transient Transfection of Mammalian Cells: a Newly Emerging Attractive Option for Recombinant Protein Production. J. Struct. Funct. Genom. 2005, 6, 165–170. 26. Baldi, L.; Muller, N.; Picasso, S.; et al. Transient Gene Expression in Suspension HEK-293 Cells: Application to Large-scale Protein Production. Biotechnol. Prog. 2005, 21, 148–153. 27. Backliwal, G.; Hildinger, M.; Chenuet, S.; et al. Rational Vector Design and Multi-pathway Modulation of HEK 293E Cells Yield Recombinant Antibody Titers Exceeding 1 G/l by Transient Transfection under Serum-free Conditions. Nucleic Acids Res. 2008, 36, e96. 28. Jäger, V.; Büssow, K.; Schirrmann, T. Transient Recombinant Protein Expression in Mammalian Cells. In Animal Cell Culture; Al-Rubeai, M., Ed., Springer International Publishing, 2015; p 1. 29. Gutierrez-Granados, S.; Cervera, L.; Kamen, A. A.; Godia, F. Advancements in Mammalian Cell Transient Gene Expression (TGE) Technology for Accelerated Production of Biologics. Crit. Rev. Biotechnol. 2018, 38, 918–940. 30. Drugmand, J. C.; Schneider, Y. J.; Agathos, S. N. Insect Cells as Factories for Biomanufacturing. Biotechnol. Adv. 2012, 30, 1140–1157.

1.27 mRNA Translation and Recombinant Gene Expression From Mammalian Cell Expression Systems EJ Mead and CM Smales, University of Kent, Canterbury, United Kingdom © 2011 Elsevier B.V. All rights reserved. This is a reprint of E.J. Mead, C.M. Smales, 1.29 - mRNA Translation and Recombinant Gene Expression from Mammalian Cell Expression Systems, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 403-409.

1.27.1 1.27.2 1.27.3 1.27.4 1.27.5 1.27.6 1.27.7 1.27.8 References

Introduction Translational Machinery Manipulation of mRNA for Optimal Translational Efficiency Importance of 50 -UTR and Secondary Structure in 50 -UTR Region of mRNA mRNA Translation Shutdown MicroRNAs and Translational Control In Vitro mRNA Translation Systems Conclusions and Future Prospects

391 391 394 395 395 396 396 396 396

Glossary (de)Phosphorylation A regulatory mechanism whereby addition or removal of a phosphate group from a protein molecule, often resulting in a conformational change in the protein, leads to either activation or deactivation. Eukaryotic initiation factor (eIF) A group of proteins which, by promoting the correct association of ribosomes with messenger RNA (mRNA), facilitate the initiation of protein synthesis. Messenger RNA (mRNA) An RNA template molecule produced by transcription of DNA by RNA polymerase II. Specifies the amino acid sequence of a protein to the translational machinery. MicroRNA (miRNA) Short nucleotide sequences of 18–22 base pairs (bp) which can exhibit either exact or partial base paring to target mRNAs and modify their stability and/or translation. Translation A ribosome-mediated process during which mRNA is decoded to a specific amino acid sequence.

1.27.1

Introduction

Significant advances have been made in recent years with regard to increasing the maximal cell concentration achievable while in vitro culturing mammalian cells in the bioreactor and, to a lesser extent, in improving specific productivity from such cells, ultimately resulting in enhanced recombinant protein (rP) yields from such expression systems. To date, these advances have been achieved through two primary areas: (1) with regard to increased viable cell concentration improved media design, feeding, and bioprocesses and (2) with regard to specific productivity via the redesign of vectors and promoter systems for high-level messenger RNA (mRNA) expression. However, increased mRNA levels have not always resulted in improved rP yield and, despite improvement, industrial cell culture systems currently have a maximum specific productivity of around 20–50 pg/cell/day for immunoglobulin Gs (IgGs) from mammalian expression systems. This compares to a production rate for B-cells which has been calculated at approximately 100 pg/cell/day5 and from hybridomas at 2300–8000 monoclonal molecules/cell/s12 which equates to approximately 47–170 pg/cell/day. There are numerous points throughout the gene expression pathway of rPs which are potential bottlenecks (e.g., see Fig. 1). The advent of systems biology is beginning to enable the complex mathematical modeling required to elucidate precise limiting points in the rP production pathway of individual cell lines and proteins.9 In the interim, the discrepancy between high-level mRNA expression and productivity suggests that constraints on rP production are no longer at the DNA level but are at the translational or posttranslational levels. Here, we describe the manipulation of mRNA translation in mammalian cells to improve rP production.

1.27.2

Translational Machinery

In the early stages of culture, cells focus energy into division and growth, resulting in a lag between growth curves and rP production. As cells reach the mid-exponential phase of growth, much of the energy resource of the cell is diverted into protein synthesis via mRNA translation and global protein synthesis. When a culture approaches maximum cell concentration, a number of

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00024-0

391

392

mRNA Translation and Recombinant Gene Expression From Mammalian Cell Expression Systems

HC Polypeptide

Half antibody

VH HC mRNA HC DNA Transcription

CH2 C H3 Translation

LC DNA LC mRNA Cell growth Cell death

CH1 Secretion Folding intermediates Assembled antibody

VL CL

LC dimer

LC Polypeptide

Figure 1 Potential limiting points in the gene expression pathway of recombinant antibody production. Folding intermediates and assembly are dependent on antibody subtype. Adapted from Ho, Y., Varley, J., Mantalaris, A (2006). Development and analysis of a mathematical model for antibody producing GS-NS0 cells under normal and hyperosmotic culture conditions. Biotechnol. Prog. 22: 1560–1569.

environmental stresses, such as nutrient deprivation and endoplasmic reticulum (ER) stress, are encountered and perceived. One mechanism by which mammalian cells respond to such environmental, ER, and other stresses encountered as a result of the demands of high-level expression of a rP is to modulate mRNA translation rates and hence global protein synthesis rates. Much of this control is imparted by (de)phosphorylation of key factors of the mRNA translational machinery, with a potential outcome being a reduction in recombinant mRNA translation and hence rP synthesis. Therefore, in order to maximize rP synthesis, differences in the rate of mRNA translation throughout the entire culture must be determined and optimized through each culture phase, although this is rather difficult to achieve. mRNA translation comprises three main steps – initiation, elongation, and termination – each of which requires ribosomes, mRNA, energy, and a number of additional factors. Translation initiation is the most complex of the steps and depends upon the assembly and coordinated action of multiple initiation factors, many of which consist of multiple subunits (Fig. 2). Consequently, the initiation step is rate limiting for translation and provides the cell with a key control point in the gene expression pathway. In yeast cells, it has been shown that the copy numbers per cell of the eukaryotic initiation factors (eIFs) and their subunits are spread over an approximately 30-fold range. Knowledge of the stoichiometry of initiation factors not only gives a better insight into translation initiation as a whole process but also provides key information on the impact of individual eIFs in translational regulation.15 In mammalian cells, the stoichiometry of the eIFs remains to be determined, although a number of studies have shown that changes in levels of individual eIFs or their subunits are implicated in malignant transformation and changes in cell growth in cancer. Despite the stoichiometry of the mammalian eIFs being as-yet unknown, a number of key control points of translation initiation have been well studied in mammalian cells. The phosphorylation status of several eIFs tightly regulates translation initiation. The best characterized of these are eIF2a-8 and eIF4E-binding protein (4E-BP)10 phosphorylation (see Fig. 2). eIF2 is a trimeric protein which joins with the initiator methionyltRNA (Met-tRNA(i)) and guanosine triphosphate (GTP) to form the ternary complex and bring Met-tRNA(i) onto the 40S ribosomal subunits (tRNA, transfer RNA). During each cycle of translation initiation, the GTP associated with eIF2 is hydrolyzed to guanosine diphosphate (GDP) and subsequently the eIF2/GDP complex is released. To participate in another round of translation, the GDP must be exchanged for a GTP molecule. This exchange is accomplished with the help of the guanine nucleotide exchange factor (GEF) eIF2B. In times of cellular stress, the eIF2a subunit is hyperphosphorylated at Ser51 and forms an inhibitory complex with eIF2B, thus preventing the assembly of the ternary complex and translation initiation. As eIF2 normally occurs in excess of eIF2B, and has a higher affinity for eIF2B in its phosphorylated form, only a small proportion of eIF2 needs to be phosphorylated to inhibit nucleotide exchange and thus cap-dependent initiation. The phosphorylation event can be carried out by one of four protein kinases (PKR-like endoplasmic reticulum kinase (PERK), heme-regulated inhibitory kinase (HRI), general control nonderepression-2-kinase (GCN2), and interferon-induced protein kinase (PKR)), each of which responds to different cellular stresses. The level of phosphorylated eIF2a has been shown to increase toward the end of mammalian cell culture, at the same point at which there was a marked increase in secreted rP. Phosphorylated eIF2a levels may therefore act as an indicator of cellular stress perception, which could be monitored during rP production and used to inform feeding and engineering strategies.14 eIF4E is a member of the cap-binding complex (eIF4F) which binds the 7-methyl GTP-cap structure of mRNA for cap-dependent translation. Inhibition of cap binding of eIF4E is a potential control point for translation initiation both through eIF4E phosphorylation and through its interaction with the 4E-BPs. The members of the 4E-BP family inhibit translation by competing

elF4F M

elF4H

GTP Met-tRNA

elF2B GTP

GDP

elF4B

ATP

elF4G

elF2

elF4E

ADP + Pi

Poly A

elF4A

PABP

GTP

M

Ternary Complex

m7G

α β γ Ser51P

AUG

mRNA GDP 4E-BP

P elF1A

M GTP

GTP M

M GTP

Circularized mRNA

Multifactor complex

AUG

43S pre-initiation complex Scanning

40S subunit

ATP

elF4A, 4E, 4G,4B, 4H

ADP+Pi

M

GTP

elF1

elF3

elF5

m7G

AUG

48S complex

GTP

GDP +Pi

ElF5B

GTP

elF2(GDP),1,3,5

M GTP

m7G

AUG

GDP +Pi GTP

60S subunit

elF1A,5B

M

m7G

AUG

80S assembled ribosome

Figure 2 Schematic representation of key control points and interacting factors in mammalian translation initiation, based on current understanding. Arrows at eIF2 and eIF4E control points represent reversible reactions as opposed to equilibrium reactions.

mRNA Translation and Recombinant Gene Expression From Mammalian Cell Expression Systems

AUG

393

394

mRNA Translation and Recombinant Gene Expression From Mammalian Cell Expression Systems

for the eIF4G-binding site of eIF4E. Affinity of the 4E-BP for eIF4E is reduced by its hyperphosphorylation. The hyperphosphorylation can be induced by a variety of stimuli, including hormones, mitogens, and growth factors, via a number of signaling pathways such as the mammalian target of rapamycin (mTOR) pathway. The converse of this is hypophosphorylation of 4E-BP when there is nutrient or growth factor deprivation.7 Hypophosphorylation of 4E-BPs increases their affinity for eIF4E, inhibiting cap-dependent translation in times of stress. In a recent study investigating the impact of cellular energy on rP production capacity, it was shown that upon addition of adenosine to a culture, the level of cellular adenosine triphosphate (ATP) increased, cell growth was arrested and average specific productivity was increased 2.5-fold. Under these conditions, hypophosphorylation of 4E-BP1 is expected due to the activation of the adenosine monophosphate-activated protein kinase (AMPK) and subsequent inhibition of the mTOR pathway of 4E-BP1 phosphorylation. However, the high levels of ATP kept 4E-BP1 hyperphosphorylated, probably by a direct interaction of ATP with the mTOR pathway. The hyperphosphorylation allowed translation to continue and increased average specific productivity to be achieved, suggesting that the manipulation of the repression of translation by pathways involving 4E-BP may improve rP yield under stress conditions.3 The eIFs are not the only cellular machinery critical to the rate of translation. The number of ribosomes present in the cell can also limit translational capacity. Ribosome biogenesis in mammalian cells is a complex process during which RNA polymerases I (45S processed to 18S, 28S, and 5.8S) and III (5S) transcribe rRNA species in different cellular compartments. Additionally, RNA polymerase II transcribes a large number of mRNAs, approximately 80 of which are translated to ribosomal proteins. At any one time, a significant proportion of ribosomal RNA (rRNA) genes in a cell is transcriptionally silent. Ribosome assembly begins in the nucleolus before transport out through the nucleus into the cytoplasm where the final activation of the large subunit occurs (for review, see Ref. 4). In yeast, it has been shown that engineering of a gene BMS1, whose product is required for 18S ribosome biogenesis, altered the ratio of 60S:40S subunits from 1:1 to 2:1 and correspondingly 25S:18S subunits from 2:1 to 3:1, resulting in a high-yielding phenotype. These data suggest that the balance of cellular ribosomal components is critical to efficient rP production.1 Furthermore, a study showed that, in mammalian cells producing the model rPs, luciferase and human placental secreted alkaline phosphatase (SEAP), protein yield was increased two- to fourfold when ribosome synthesis was enhanced. The increase in ribosome synthesis was achieved using an epigenetic engineering approach to limit the silencing of rRNA genes by knockdown of a subunit of the nucleolar remodeling complex (NoRC) which normally represses rDNA transcription via histone modification and DNA methylation.11

1.27.3

Manipulation of mRNA for Optimal Translational Efficiency

In addition to modifications in translational machinery, translational efficiency can be significantly up- or downregulated according to the specific mRNA sequence being translated. As a result, a number of approaches have been taken to manipulate mRNAs encoding rPs for improved yield. The genetic code encodes for 20 amino acids using 61 different codons, allowing for significant degeneracy in the code, from those amino acids encoded by only one codon (Met and Trp) to those encoded for by six codons (Arg, Leu, and Ser). Synonymous codons are not used with equal frequencies; correspondingly, concentrations of the respective tRNAs differ. This bias in the usage of codons and tRNAs is extremely organism dependent. One approach to enhance mRNAs for expression in mammalian cells is to engineer the sequence to optimal codons for individual cell lines with the aim of increasing the speed of mRNA translation and hence protein expression. Optimized codon usage has also been found to have a significant effect on rP production in prokaryotes such as E. coli. Recently in E. coli, it has been shown that at protein domain boundaries it may be beneficial to slow translation down and allow time for protein folding or chaperone activity to occur. In this context, it is not only codon usage that must be considered but also the use of adjacent codons or codon pairs which determine the positioning of tRNAs next to one another on the ribosome. Codon optimization of mRNAs for production of complex multidomain proteins in mammalian cells also requires careful consideration of the fine balance between translation rates and the requirement for folding. Splicing of mRNA uses time and energy and as such introns are frequently removed from engineered mRNAs. The first intron is retained as there is a requirement for the splice machinery to be present for optimal translation, to the extent that adding an intron into a normally intronless gene significantly enhances its production. Considerable attention has also been paid to enhancing translation by modification of the 50 -untranslated region (UTR); however, the possibility of engineering of the 30 -UTR should not be ignored. Critically, translation termination efficiency in mammalian cells can be significantly altered by changing the base following the stop codon. This modification can prevent read-through and hence improve protein authenticity. 30 -UTR structure is also crucial for mRNA stability. One key predictor of mRNA stability is the presence of AU-rich elements (AREs) in the 30 -UTR, and removal of these elements from unstable mRNAs increases mRNA half-life. A second predictor of mRNA stability is the addition and length of a poly(A) tail. The poly(A) tail can interact with the cap structure via a bridging association between the poly(A)-binding protein (PABP), PABP-interacting protein (PAIP-1) and eIF-4G, resulting in circularization of mRNA (see Fig. 2). This phenomenon enables re-initiation of ribosomes onto the mRNA, increasing translation rates. All of these modifications in the 30 -UTR can be coordinated not only to improve rP authenticity via the stop codon but also to increase yield through higher mRNA levels and re-initiation of translation.

mRNA Translation and Recombinant Gene Expression From Mammalian Cell Expression Systems

1.27.4

395

Importance of 50 -UTR and Secondary Structure in 50 -UTR Region of mRNA

The ribosome is usually recruited to the 50 -UTR of the mammalian mRNA in a cap-dependent manner, and the beginning of protein synthesis by the assembled ribosome requires the presence of specific consensus regions surrounding the AUG start codon. Elements within the 50 -UTR such as stable hairpins, upstream open reading frames, and internal ribosome entry sites (IRESes) can be engineered to influence translation initiation. As with the sequence surrounding the stop codon, the sequence surrounding the AUG start codon is critical for fidelity of rP production. In mammalian cells, the majority of mRNAs contains a Kozak consensus sequence (GCCGCCACC ATG G). This sequence slows ribosome scanning at the initiation codon, can significantly enhance the probability of efficient initiation, and helps prevent leaky scanning which leads to nonauthentic protein production. For this reason, a Kozak sequence is commonly engineered into the 50 -UTR of mRNAs for production in mammalian cells. During times of cellular stress, cap-dependent translation of proteins is downregulated via a number of pathways. However, under times of stress not all translation ceases, and one mechanism for overcoming the translational block is to utilize a capindependent mechanism of ribosome recruitment via an IRES. The IRESes which were first discovered as part of the mechanism to hijack host, cell translation in picornaviruses, are present in the 50 -UTR of a number of mRNAs encoding cellular proteins with roles in critical control processes such as cell growth and cell death. Specific IRESes respond to specific stress stimuli, including hypoxia, ER stress, and nutrient deprivation. This avoidance of translation inhibition allows proteins, which are key to the cellular response to the particular stress the cell is encountering, to continue to be produced and the cell to maintain its integrity through a variety of environmental changes. The viral IRESes have been shown to have different canonical initiation factor requirements for their activity, and this differential requirement has been found to extend to the cellular IRESes. Within the cell are a group of auxiliary proteins called IRES trans-activating factors (ITAFs), some of which are generally required for IRES activity (such as polypyrimidine-tract-binding protein), and others which are more specific. It has been proposed that the ability of different IRESes to respond to different stress stimuli depends on changes in the expression and localization of ITAFs.13 At the end of mammalian cell culture when cells are stressed, cap-dependent translation, including that of any recombinant mRNA, will be significantly reduced. IRESes have already been utilized along with alternative splicing for the development of dicistronic expression systems which couple selectable markers to expression of the recombinant gene of interest, such as the glutamine synthase selection marker in NS0 cells. An engineering strategy could be envisaged whereby recombinant mRNAs are designed incorporating specific IRESes and expressed in cells with a tailored ITAF repertoire, allowing the recombinant mRNA to exploit cap-independent translation when cap-dependent translation is inhibited. In contrast to the potential benefits of an IRES, the presence of stable secondary structures, such as hairpins, in the 50 -UTR can significantly decrease efficiency of cap-dependent translation initiation; as a result, these structures are normally removed from the 50 -UTR.

1.27.5

mRNA Translation Shutdown

Nutrient deprivation, including amino acid deprivation, is severely limiting on translation, resulting in the accumulation of uncharged tRNAs and activation of the general amino acid control (GAAC) pathway. Upon activation of the GAAC pathway, eIF2a becomes phosphorylated via the GCN2 kinase, and global protein synthesis is inhibited until the nutrient stress is alleviated. Improvements in culture media and optimization of processes in the bioreactor, such as fed-batch strategies, help to minimize translational limitations arising from nutrient deprivation. Complex metabolomics analysis during rP production may help to further inform feeding strategies to avoid uncharged tRNA accumulation and hence relieve translational inhibition at the end of culture. Temperature downshift is frequently used in the production of rPs, and changes in yield and cell viability following downshift are the culmination of complex cellular responses to cold. It has been shown that in general as temperature decreases, overall protein synthesis rates gradually decline. As in other cellular stresses during the response to more extreme cold shock, there are changes in eIF phosphorylation status decreasing global translation. Transient transfection of mutated eIF2aSer51Ala- or eIF4ESer209Glu-modified expression of a model rP in both a cell-line- and temperature-dependent manner. There is a delicate balance in temperature reduction, which must be enough for improved yield, while avoiding shutdown of cellular processes necessary for production. Mild temperature downshift of CHO cells from 37 to 32  C does not significantly reduce global protein synthesis and allows translation to proceed sufficiently to produce higher rP yields, partly as a result of increased stability and levels of mRNA. The UPR is activated in times of ER stress by the accumulation of unfolded or misfolded proteins. It is a complex response which ultimately leads to transcriptional and translational repression of gene expression and a reduction of the stress placed upon the ER. The primary mediators of UPR-activated translational repression are PERK via eIF2a phosphorylation and activating transcription factor 4 (ATF4) via induction of 4E-BP1. Notably, ATF4 mRNA is specifically translated when eIF2a is phosphorylated, owing to four short open reading frames in the 50 -UTR. Under normal cellular conditions, ribosomes translate these short sequences and fall off the mRNA, never reaching the authentic start codon. When eIF2a is hyperphosphorylated, and cap-dependent translation is inhibited, the process of leaky scanning means that a proportion of 40S subunits does not form an intact ribosomal complex until they reach the true start codon, allowing ATF4 to be translated. The IRESes provide one option for manipulating translation during the UPR. An alternative to this is to engineer UPR intermediates to prevent translational repression. A clear problem with this

396

mRNA Translation and Recombinant Gene Expression From Mammalian Cell Expression Systems

approach is the highly interlinked nature of the UPR transduction pathway and the crosstalk between each of the elements making it difficult to manipulate one part of the pathway without having serious consequences for other parts. Although complex to engineer, natural plasma cells have achieved such control of the UPR. During plasma cell differentiation, p54XBP-1 and p50ATF6a are produced, increasing ER protein synthesis. In contrast, the translational repression pathway mediated by PERK has not been identified in these cells. Understanding how plasma cells regulate the UPR response may offer insights into engineering strategies for manipulation of the UPR-signaling pathways in industrially relevant cell lines.

1.27.6

MicroRNAs and Translational Control

MicroRNAs (miRNA) are short nucleotide sequences of 18–22 base pairs (bp) originally identified in Caenorhabditis elegans and subsequently found in a wide range of eukaryotic species including those most relevant to industrial rP production, mouse, and CHO. In classical miRNA regulation, RNA polymerase II transcribes a pri-miRNA which is both capped and polyadenylated. This pri-miRNA is subsequently processed in the nucleus to a stem-loop structure of approximately 33 nucleotides (nt) termed a pre-miRNA. The pre-miRNA is exported to the cytoplasm and further processed by the enzyme Dicer to form a mature miRNA duplex. The duplex is short-lived, being quickly unwound; one strand is discarded and the other is retained in complex with an Ago effector protein. Ago protein and miRNA form the core of the miRISC complex, which finds its specific target using the miRNA sequence and is then able to regulate particular mRNAs (for review, see Ref. 2). The exact mechanism of repression by miRISC remains undetermined but it certainly involves the repression of translation. Any one miRNA has the potential to repress many mRNAs, and it is estimated that in the human genome up to 30% of genes are targets. Included in these targets are a number of proteins proved to either enhance or reduce rP yield. A study looking at the changes in miRNA expression during batch culture and biphasic culture of CHO cells has identified 26 differentially expressed miRNAs between cells growing exponentially at 37  C and those in stationary phase at 31  C.6 As knowledge of miRNA expression, mechanisms of action, and targets rapidly increases, the potential scope for utilization of this relatively new area of translational control in cell engineering is vast.

1.27.7

In Vitro mRNA Translation Systems

Cell-free translation systems have primarily been used to study mechanisms of translational control, incorporate unnatural amino acids, or produce small amounts of protein for structural and functional studies. They also have the advantage of being able to express proteins which are normally toxic to cells. E. coli extract is established for cell-free prokaryotic protein production but has limitations for expressing eukaryotic proteins. Rabbit reticulocyte lysate is the most authentic commonly used lysate for the study of mammalian cell translation but is difficult to produce and therefore costly. Wheat germ lysates have provided some advancement on E. coli lysates producing eukaryotic proteins in a relatively cost-effective way; however, a major limitation in each of these systems for the production of complex mammalian proteins is the lack of complex posttranslational modifications. Work has been undertaken to investigate the possibility of producing mammalian cell lysates from cultured cells and to characterize the capacity of these lysates not only to translate recombinant mRNAs but also to glycosylate them. Inhibition of translation initiation through phosphorylation of eIF2a has been overcome in these lysates by using nonphosphorylatable mutants, and some simple glycosylation patterns have been achieved, but significant advances will have to be made before cell lysates are a realistic prospect for the authentic production of complex rPs such as monoclonal antibodies.

1.27.8

Conclusions and Future Prospects

The control and manipulation of mRNA translation is a complex process involving multiple pathways, proteins, and mechanisms. Despite this, there are clear opportunities to improve rP synthesis via the careful control and manipulation of key components of the mRNA translational machinery and recombinant mRNAs. A failure to consider such issues certainly will impinge on the ability to achieve high-level expression of recombinant genes from in vitro-cultured mammalian cells. Ongoing investigations toward defining the cellular limitations on mRNA translation during culture of mammalian cells are likely to open up further opportunities for the manipulation of translational elements and machinery with a view to increasing rP yields from these important gene expression systems.

References 1. 2. 3. 4. 5.

Bonander, N.; Darby, R. A.; Grgic, L.; et al. Altering the Ribosomal Subunit Ratio in Yeast Maximizes Recombinant Protein Yield. Microb. Cell Factories 2009, 8, 10. Carthew, R. W.; Sontheimer, E. J. Origins and Mechanisms of miRNAs and siRNAs. Cell 2009, 136, 642–655. Chong, W. P.; Sim, L. C.; Wong, K. T.; Yap, M. G. Enhanced IFNgamma Production in Adenosine-treated CHO Cells: A Mechanistic Study. Biotechnol. Prog. 2009, 25, 866–873. Dinman, J. D. The Eukaryotic Ribosome: Current Status and Challenges. J. Biol. Chem. 2009, 284, 11761–11765. Dinnis, D. M.; James, D. C. Engineering Mammalian Cell Factories for Improved Recombinant Monoclonal Antibody Production: Lessons from Nature? Biotechnol. Bioeng. 2005, 91, 180–189.

mRNA Translation and Recombinant Gene Expression From Mammalian Cell Expression Systems

397

6. Gammell, P.; Barron, N.; Kumar, N.; Clynes, M. Initial Identification of Low Temperature and Culture Stage Induction of miRNA Expression in Suspension CHO-K1 Cells. J. Biotechnol. 2007, 130, 213–218. 7. Gingras, A. C.; Gygi, S. P.; Raught, B.; et al. Regulation of 4E-BP1 Phosphorylation: A Novel Two-step Mechanism. Gene Dev. 1999, 13, 1422–1437. 8. Kimball, S. R.; Fabian, J. R.; Pavitt, G. D.; et al. Regulation of Guanine Nucleotide Exchange through Phosphorylation of Eukaryotic Initiation Factor eIF2alpha. Role of the Alphaand Delta-subunits of eiF2b. J. Biol. Chem. 1998, 273, 12841–12845. 9. Mead, E. J.; Chiverton, L. M.; Smales, C. M.; von der Haar, T. Identification of the Limitations on Recombinant Gene Expression in CHO Cell Lines with Varying Luciferase Production Rates. Biotechnol. Bioeng. 2009, 102, 1593–1602. 10. Pause, A.; Belsham, G. J.; Gingras, A. C.; et al. Insulin-dependent Stimulation of Protein Synthesis by Phosphorylation of a Regulator of 5’-cap Function. Nature 1994, 371, 762–767. 11. Santoro, R.; Lienemann, P.; Fussenegger, M. Epigenetic Engineering of Ribosomal RNA Genes Enhances Protein Production. PLoS One 2009, 4, e6653. 12. Savinell, J. M.; Lee, G. M.; Palsson, B. O. On the Orders of Magnitude of Epigenic Dynamics and Monoclonal-antibody Production. Bioprocess Eng. 1989, 4, 231–234. 13. Spriggs, K. A.; Stoneley, M.; Bushell, M.; Willis, A. E. Re-programming of Translation Following Cell Stress Allows IRES-mediated Translation to Predominate. Biol. Cell 2008, 100, 27–38. 14. Underhill, M. F.; Birch, J. R.; Smales, C. M.; Naylor, L. H. EIF2alpha Phosphorylation, Stress Perception, and the Shutdown of Global Protein Synthesis in Cultured CHO Cells. Biotechnol. Bioeng. 2005, 89, 805–814. 15. von der Haar, T.; McCarthy, J. E. Intracellular Translation Initiation Factor Levels in Saccharomyces cerevisiae and Their Role in Cap-complex Function. Mol. Microbiol. 2002, 46, 531–544.

1.28

Posttranslation Modifications Other Than Glycosylation

N Jenkins, University College Dublin, Dublin, Ireland © 2011 Elsevier B.V. All rights reserved. This is a reprint of N. Jenkins, 1.30 - Posttranslation Modifications Other Than Glycosylation, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 411-417.

1.28.1 Introduction 1.28.2 Cell Influences on Protein Expression 1.28.3 Induction of Protein Expression 1.28.4 Improving the Protein Folding and Secretory Pathways 1.28.5 Role of Chaperones 1.28.6 Multiple Gene Activators 1.28.7 Cell Clearance of Misfolded Proteins 1.28.8 Protein Aggregation 1.28.9 Analytical Techniques for Protein Aggregate Detection 1.28.10 Asparagine Deamidation 1.28.11 Methionine Oxidation 1.28.12 Surface Plasmon Resonance 1.28.13 Conclusions Acknowledgments References Relevant Websites

398 398 399 399 400 400 400 401 401 403 403 403 403 403 403 404

Glossary Aggregation The binding of two or more identical protein molecules. This can be covalent or noncovalent. Asparagine deamidation Involves the non enzymic conversion of asparagine residues to form aspartic acid and iso-aspartic acid and is a cause of protein degradation, particularly during long-term storage. Misfolding Incorrect folding pattern of a protein molecule. This can be covalent or noncovalent. Methionine oxidation Some of the methionine residues in a protein can be oxidized to methionine sulfoxide or even methionine sulfone. Posttranslational modifications Any modifications made to a protein once it has been synthesized on the ribosome of a cell.

1.28.1

Introduction

Recombinant proteins, especially monoclonal antibodies (MAbs) and their derivatives, constitute a growing sector of the biotechnology and pharmaceutical industries. The ability to perform complex posttranslational modifications (PTMs) is one of the major reasons that most biotherapeutics are manufactured in mammalian cells.1 However, proteins are prone to several modifications that can affect their efficacy and limit shelf life. Common protein isoforms and degradation pathways include variable glycosylation, misfolding, aggregation, methionine oxidation, asparagine deamidation, and proteolysis.2 Protein N- and O-linked glycosylation is covered in several recent reviews including Ref. 3. This article focuses on how the production and purification of biopharmaceuticals can be improved without compromising the quality attributes of recombinant proteins. Potential bottlenecks in protein expression, folding, and secretion are described in the first sections of this article, together with the strategies that may be used to alleviate these constraints. In addition to genetic manipulation of cell lines, many environmental conditions that occur during bioprocessing can contribute to protein secretion and stability, and these are listed in the main text and associated tables. Robust and accurate assays for PTMs are essential for the understanding of protein stability at all level stages of bioprocessing, and recent improvements in protein analysis are described at the end of this article.

1.28.2

Cell Influences on Protein Expression

In order to better understand the potential constraints in protein expression and secretion, studies have compared cell lines producing different levels of recombinant protein (from low to high producers) and examined the effects of supplements (e.g., sodium butyrate (NaBu) or dimethyl sulfoxide (DMSO) that result in increased productivity4). In a study comparing recombinant

398

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00025-2

Posttranslation Modifications Other Than Glycosylation

399

Chinese hamster Ovary (CHO) cells expressing different amounts of green fluorescent protein (GFP), 20 proteins were identified as being differentially regulated and were involved in protein biosynthesis and folding, including eIF2S3 and Hspd1. The challenge is to translate these findings into functional cell responses to bioprocess conditions in order to improve productivity and product quality attributes. Four mouse myeloma (NS0) cells that had approximately equal growth rates but varying monoclonal antibody (MAb)-specific production rates have also been compared. Proteins that were found to be upregulated with increasing levels of productivity included endoplasmic reticulum (ER) luminal chaperones involved in folding (endoplasmin, ENPL) and the chaperone heavy-chain binding protein (BiP). Other upregulated proteins included cytosolic and mitochondrial chaperones (HSC7 and HSP60), proteins involved in calcium binding and microtubule stabilization (TCTP), nucleoside metabolism (NDKA), and oxidative stress (PDX1). Downregulated proteins included two glycolytic enzymes (ALFA and KPYc). In a related study, the microsomal component was isolated from the same set of NS0 cells; protein disulfide isomerase (PDI) and BiP were again found to be increased with increased levels of MAb productivity. A set of 79 proteins was identified in NS0 cells and analyzed across all four NS0 lines with varying levels of specific MAb productivity. The relative abundance of several ER chaperones, non-ER chaperones, cytoskeletal, cell-signaling, metabolic, and mitochondrial proteins were significantly increased with increased productivity of MAb. The authors suggest that individual cells within parental populations are more equipped for high-level recombinant protein production than others and that this difference could be used to select cells that are likely to achieve high productivity. A study profiling NS0 cells with high-productivity levels using both cDNA microarray and proteomic technologies found that a large number of genes and proteins related to protein synthesis, cell growth, and cell death pathways were altered in high-producing cell lines. While most proteolytic events are damaging to the protein product, there are several products where a defined cleavage event is necessary for efficient secretion. For example, differences in the proteome of CHO-DUKX cells expressing recombinant human bone morphogenetic protein-2 (rhBMP-2) were compared to cells co-expressing the cleavage enzyme PACE-sol, which improves posttranslational processing of the rhBMP-2 dimer.4 PACE-sol co-expression was associated with a significant increase in cellular productivity of rhBMP-2. Sixty proteins were also found to be differentially expressed, and a substantial number of these proteins were found to have chaperone or folding activity (e.g., BiP and PDI). Several proteomics technologies are used for protein expression profiling, including two-dimensional polyacrylamide gel electrophoresis (2D-PAGE), surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-ToF MS), protein arrays, and stable isotope labeling technologies such as Isotope Coded Affinity Tags (ICAT), Isobaric Tagging for Relative and Absolute Protein Quantitation (iTRAQÔ), and Stable Isotope-Labeled Amino Acids in Culture (SILAC). These techniques are described in more detail in a recent paper from our lab.1

1.28.3

Induction of Protein Expression

NaBu is well known to enhance the productivity of recombinant proteins and appears to cause a complex series of changes in CHO cells related to longevity of culture and specific productivity. A study investigated proteins altered following elevated levels of human growth hormone (hGH) expression in recombinant CHO cells.5 Cells were exposed to zinc and NaBu, and the expression of a number of cellular proteins besides hGH was increased in response, including GRP75, enolase, and thioredoxin. NaBu induces changes in expression of a number of proteins involved in either protein folding (GRP78), cell metabolism (GAPDH), or cytoskeletal architecture (tropomyosin 4, b tubulin), among others. DMSO has previously been shown to increase specific productivity of a fusion protein in CHO cells, but it is not clear how DMSO exerts this effect. In addition to recombinant hepatitis B surface antigen, seven proteins were altered following exposure to DMSO, including four glycolytic enzymes (triose–phosphate isomerase, GAPDH, aldolase, and phosphoglycerate kinase). A switch to low-temperature cultivation has also been shown to significantly improve the specific productivity of recombinant cells,6 but the molecular mechanisms underlying this response are poorly understood. It is common practice in industry to use a temperature switch to induce a static culture (which is more productive). The effect of lowering cultivation temperature from 37 to 30
C on the productivity of CHO cells producing secreted alkaline phosphatase (SEAP) has been monitored using 2D-PAGE and Western blotting. Ten 2D-PAGE spots had significantly altered intensities from cells incubated at 30
C compared to the 37
C controls, and Western blots showed altered levels of two tyrosine-phosphorylated proteins. The transcriptome and proteome of CHO cells producing recombinant erythropoietin (EPO) grown at different temperatures have also been investigated. The expression levels of several proteins (PDI, vimentin, NDK B, ERp57, phosphoglycerate kinase, and 71 kDa heat shock protein) were increased over twofold at 33
C, and two proteins (HSP90-beta and EF2) were decreased over twofold, compared to the 37
C control culture. In summary, a number of studies have been performed to date profiling mammalian cells used for biopharmaceutical production to gain a better understanding of the biology of these cells, with the ultimate aim to generate a cell line capable of high productivity. The data generated so far from these studies suggest that expression of a wide range of genes and proteins are altered, especially from functional classes such as protein folding and secretion, protein synthesis, cellular architecture (cytoskeleton), and cellular metabolism and cell growth.

1.28.4

Improving the Protein Folding and Secretory Pathways

The lumen of the ER provides an oxidizing environment and access to several enzymes and chaperones that assist in the folding and secretion of proteins. PDI is a 58-kDa ER-resident enzyme that catalyzes the formation and breakage of disulfide bonds between

400

Posttranslation Modifications Other Than Glycosylation

thiol groups of cysteine residues using the substrate glutathione. It allows proteins to find the correct alignment of disulfide bonds7 and operates as a chaperone to inhibit the aggregation of misfolded proteins. This reaction normally allows proteins to fold quickly, but a mismatch of thiol pairing can also be the source of covalent aggregate formation and protein misfolding. Results from transfecting extra copies of the gene coding for PDI alone has yielded mixed results. For example, a 2.5-fold increase in the level of PDI resulted in a 15–27% rise in antibody productivity in one CHO line but failed to increase thrombopoietin secretion in another CHO line. Besides the overall level of PDI, glutathione availability and the activity of Ero1p (the enzyme that oxidizes PDI) can affect PDI activity. Increased expression of Ero1 results in the acceleration of disulfide bond formation and correct protein folding. However, reducing the levels of glutathione in the cell can lead to an increase in the rate of disulfide bond formation without leading to correct protein folding, by compromising the reductive pathway. It is likely that multiple intervention points in the reductive pathway and formation of disulphide bonds will be required for maximum secretion and product quality.

1.28.5

Role of Chaperones

Chaperones such as heavy chain BiP (a member of the hsp70 family) have been linked to many beneficial ER functions such as the protein translocation, folding, and oligomerization.8 However, insufficient adenosine triphosphate (ATP) levels or the lack of co-chaperones such as Lhs1p may become rate limiting to BiP functions, and increased BiP activity may stall other chaperone functions (e.g., GRP94, calnexin, or calreticulin). Engineering chaperone systems by overexpressing a single component of the ER secretion machinery has yielded mixed results in improving productivity. In one case, the overexpression of BiP actually decreased the secretion rate of recombinant antibody in a CHO line. The calreticulin and calnexin chaperones also provide a quality-control mechanism to ensure proper protein folding. They transiently bind to newly synthesized glycoprotein intermediates and ensure that only correctly folded proteins are transported to the Golgi apparatus. Calnexin and calreticulin overexpression were found to nearly double the specific productivity of thrombopoietin in recombinant CHO cultures. However, as seen with PDI, multiple chaperone intervention points may need to be engineered to consistently improve secretion rates.

1.28.6

Multiple Gene Activators

Because of the inconsistency observed when using the one-gene-at-a-time approach, several groups have considered multiple gene activation pathway for improving expression and secretion. The truncated form of X-box-BiP (XBP-1S) may be a more effective target for improving productivity, since it regulates multiple genes in the secretory pathway. XBP-1S is essential for the generation of plasma cells, a cell type optimized for high-level production and secretion of antibodies.9 The precursor (XBP-1) activation is triggered by the accumulation of unfolded or misfolded proteins in the ER, and this, in turn, activates the formation of its active form, XBP-1S. XBP-1S induces the expression of many ER, Golgi, and mitochondrial chaperones. Overexpression of XBP-1S has overall yielded mixed results in improving productivity. Heterologous expression of XBP-1S led to an increase in ER content and specific MAb productivity of CHO-DG44 cells resulting in a 40% increase in antibody titers. The most common interventions in the secretory pathway for recombinant proteins are shown in Table 1. If cells are selected based on high expression, there is a risk that protein quality may be compromised, since the posttranslational machinery of the cells may become saturated.

1.28.7

Cell Clearance of Misfolded Proteins

In theory, misfolded proteins undergo proteolysis in the endosome and the resultant amino acids are recycled in the unfolded protein response (UPR)9 and the ER overload response. However in practice, cells used for bioprocessing can become overloaded Table 1

Genetic manipulation studies on industrial cell lines

Protein

Mechanism of action

Effect on productivity

Protein disulfide isomerase Formation and breakage of disulfide Increased MAb production in CHO cells (PDI) bonds Heavy-chain-binding protein ER-based protein chaperone Overexpression delayed thyroglobulin secretion (BiP) in CHO cells Increased MAb production in insect cells Calnexin and calreticulin Chaperones that detect glycosylation Increased thrombopoietin production in CHO status cells Human XBP-1S Gene activator of multiple ER Increased MAb production in CHO cells proteins

Notes Did not increase thrombopoietin production Not useful for production Beneficial effect Beneficial effect Comparable glycosylation patterns

Posttranslation Modifications Other Than Glycosylation

401

with misfolded recombinant protein, leading to release of misfolded and aggregated proteins particularly at high levels of protein expression. This can be a particular problem with multimeric proteins such as recombinant IgG and blood-clotting proteins such as Factor VIII, where protein aggregates trigger an immune response in the patient resulting in blocking antibodies to the therapeutic protein.10 Various strategies are being evaluated to combat aggregation and misfolding that include altering expression levels of chaperone proteins or modulating the redox potential of the cell. Genetic engineering of the secretion pathway in cells used for bioprocessing (e.g., CHO cells) holds the promise of increasing protein yields without compromising product quality, that is, avoiding the formation of aggregates or misfolded proteins.

1.28.8

Protein Aggregation

Aggregation during the manufacture of recombinant proteins can occur at all stages of bioprocessing: during cell culture, protein purification, formulation, and filling.11 Protein aggregates can arise by different mechanisms, including reversible and irreversible reactions, noncovalent interactions between hydrophobic domains, or the formation of intermolecular disulfide bonds. Some aggregates are insoluble while others remain in solution. For nonvaccine biotherapeutics, all types of aggregates are considered undesirable, since small soluble aggregates may become immunogenic10 resulting in inhibitory antibodies to the therapeutic protein. Larger particulates may cause problems at the site of administration. While there are US Pharmacopeia guidelines for the number of particles of size 10 mm and 25 mm that are acceptable in pharmaceutical preparations, the permissible levels of soluble aggregates (e.g., dimers and high-molecular-weight (HMW) soluble aggregates) are ill defined. The first steps in protein aggregation typically arise from weak noncovalent protein interactions. Exposure of hydrophobic surfaces in partially denatured proteins can lead to noncovalent aggregates, and these can form precursors to covalent aggregates. There is an equilibrium between the monomers and higher-order forms (e.g., dimers and tetramers) that may shift as a result of a change in conditions such as protein concentration or pH.11 A further step in aggregate formation is disulfide bonding between unpaired thiols. These covalently bonded aggregates are very difficult to disrupt and, once formed, are often discarded during the protein purification process through chromatography or filtration. One way of reducing the free thiol content in recombinant CHO cells is to add low amounts (up to 100 mM) of the oxidizing agent, copper sulfate. This results in a 10-fold reduction of percentage-free thiols in IgG, and significant (threefold) reductions were observed with as little as 5 mM copper sulfate. Other cell culture media components can also affect the level of total aggregates and the distribution between noncovalent and disulfide-bonded aggregates.1 Evidence has been found that protein aggregates can continue to form in the supernatant after cells producing recombinant IgG have been harvested. This necessitates cooling or other controlled hold steps between bioreactor harvest and the first chromatography capture step. The aggregation problem is not confined to upstream bioprocessing, for example, the techniques used to inactivate viruses during downstream processing such as exposure to detergents or extremes of pH can inadvertently damage and aggregate the protein product.12 Low pH conditions (pH 2–4) are also commonly used to elute antibodies from Protein A capture columns during purification. Downstream intermediate and polishing steps typically include ion-exchange chromatography, which elutes the protein with high-ionic-strength solutions, or in high pH conditions which can damage the product. Multiple filtration steps are also used in protein purification for concentration, buffer exchange, and virus removal. Large protein aggregates can cause membrane fouling, and the high pressures employed may also increase aggregation during these process steps. Stresses to proteins such as freezing, thawing, lyophilization, prolonged exposure to air, light, or interactions with metal surfaces may also result in surface denaturation, which then leads to the formation of aggregates. Excipients such as sugars and arginine are often used to suppress aggregate formation during protein purification and formulation. However, there are examples where the drug product forms aggregates with its excipients, for example, a formulation of interferon-alpha became oxidized at room temperature and formed aggregates with the excipient human serum albumin. This induced an immune response to interferon-alpha, and changing to a liquid albumin-free formulation stored at 4
C reduced the immune reaction. Table 2 summarizes the environmental effects on protein modifications.

1.28.9

Analytical Techniques for Protein Aggregate Detection

Several analytical techniques exist to detect protein aggregates (see Table 3), but all have their strength and weaknesses.1 The standard (FDA-approved) method has been size exclusion chromatography (SEC) performed by high-performance liquid chromatography (HPLC) where aggregated species are separated by size from monomers in a porous gel matrix. SEC-HPLC permits relatively rapid and cost-effective analysis of aggregate content but is hampered by nonspecific interactions with the column matrix. Adding excipients such as arginine to the SEC elution buffer reduces nonspecific interactions and permit more accurate measurement of HMW species. Arginine suppresses protein aggregation through interactions between the guanidinium group of arginine and tryptophan side chains. Including arginine in the elution buffer from SEC can provide a more accurate estimation of total aggregate content. However, the largest soluble HMW species may not interact with the column material at all and be of sufficiently low concentration to escape detection. A tandem combination of SEC-HPLC and a multi-angle laser light scattering (MALLS) facilitates the detection of dimers and larger HMW species.

402 Table 2

Posttranslation Modifications Other Than Glycosylation Effects of environmental conditions on protein modifications

Environmental condition

Main effect

Context and other effects

Temperature, protein concentration, ionic strength, pH Electrostatic interactions

All influence aggregate levels

Can occur at all phases of the production cycle

Increase viscosity and aggregation

Mechanical stress

Increases aggregation

Oxidative intracellular environment Media components

Decreases aggregate formation Influence distribution between monomers, dimers, and HMW species Decreases aggregation Increases aggregation Increases aggregation

Can occur between monomers and with surfaces of containers Impeller speed in bioreactor, rate of pumping and filtration Bioreactor conditions Bioreactor conditions

Copper sulfate Time in bioreactor Freeze–thaw cycles and lyophilization Viral inactivation and low pH elution from Protein A Excipients, e.g., arginine and sugars

Table 3

Increases aggregation Decrease aggregation

Toxic to cells in mM doses Affects duration of bioreactor run Can occur at process hold steps and in formulation Minimize time the protein is exposed Used in formulation and some purification steps

Analytical methods used to detect protein quality changes

Method Size exclusion chromatography (SEC) Hydrophobic interaction chromatography (HIC) Analytical ultracentrifugation (AUC) Field flow fractionation (FFF)

Principle

Separations based on hydrodynamic radius Adsorption of hydrophobic protein residues under high salt conditions Separates molecules based on sedimentation coefficients Separates species based on diffusion coefficients Uses scattered light to measure Dynamic light scattering the rate of diffusion or geometric (DLS) and multi-angle laser light scattering (MALLS) size of protein particles Liquid chromatography–mass Mass spectroscopy used on spectrometry (LC-MS) proteolytic digests

Advantages

Limitations

FDA-approved, simple and robust

Can underestimate high-molecularweight aggregates, sample is diluted High salt conditions, can be corrosive Time-consuming, complex analysis Requires extensive optimization and low throughput Cannot resolve low-molecular-weight species

Concentrates sample more than SEC Minimal dilution effects. Good orthogonal method to SEC Broad dynamic range, column-free Avoids dilution effects, good detection of HMW aggregates Comprehensive characterization of protein primary structure and PTMs

Low-throughput, expensive equipment

Other modes of chromatography such as hydrophobic interaction chromatography (HIC) and ion-exchange chromatography (IEX) may be more useful in downstream processing (DSP) at separating aggregates from monomers. However, both HIC and IEX typically use high salt concentrations in the mobile phase, and this may alter the nature or concentration of aggregates. Other column-free methods have been used to detect aggregates such as analytical ultracentrifugation (AUC) and field flow fractionation (FFF). AUC separates molecules based on their varying sedimentation coefficients, which derive from their size and weight. AUC is arguably the optimal means of analyzing HMW distribution and is an excellent choice for orthogonal confirmation of results obtained using SEC. Unfortunately, AUC analysis requires a specialized centrifuge, complex data analysis, and is low throughput. An emerging alternative is FFF, which offers similar benefits to AUC with faster run times, simpler postrun analysis, and can be coupled to MALLS. However, FFF needs considerable optimization to ensure reproducible results and involves sample dilution. Polyacrylamide gel electrophoresis performed under native conditions can also demonstrate aggregation. This is inexpensive and can deliver nanogram detection limits with sensitive staining methods such as silver staining (0.2 ng protein/lane) and coomassie brilliant blue G-250, but it is seriously constrained by size limits, which restrict its usefulness with many biopharmaceuticals. An asset of optics-based detection technologies such as dynamic light scattering (DLS) is the ability to resolve larger HMW species that may escape detection by conventional SEC. Because the signal increases with increasing molecular size, DLS is ideal for detection of HMW species at the upper scale of >1000 kDa. However, resolution of DLS suffers below this threshold, and it is unable to resolve monomer species from dimers. There is a requirement for more rapid assessment methods for aggregation that could be used in process development (e.g., in clone selection, setting bioreactor parameters, chromatography column, and filter selection). Dye-based assays are already used in process analytics (e.g., Ellman’s reagent): 5, 50 -dithiobis-2-nitrobenzoic acid is routinely used to determine total free thiol content in product batches. Hydrophobic dye-based analysis is also compatible with the 96- or 384-well microtiter platform that lends itself to liquid handling automation and spectrophotometric readouts; however, their use in routinely detecting biopharmaceutical aggregates has yet to be established.

Posttranslation Modifications Other Than Glycosylation

403

1.28.10 Asparagine Deamidation Deamidation is reported to be the most common PTM found in proteins and involves the nonenzymic conversion of asparagine residues to a cyclic imide intermediate, which rapidly hydrolyzes to form a mixture of iso-aspartic and aspartic acid at ratios of 3:1.13 This PTM can arise during every stage of biopharmaceutical production from cell culture through to formulation and can occur in vivo post-administration. Asparagines at the protein surface appear more vulnerable, and the micro-environment is known to influence the rate of deamidation. Also, the presence of a carboxyl glycine greatly enhances the rate of deamidation, and computer modeling has been used to predict deamidation. In one case, deamidation caused a 70% loss of potency of a marketed MAb. This is known to promote aggregation in certain proteins and enhance the auto-immune response. Deamidated proteins have been identified via a variety of chromatographic techniques including HIC and IEX but are most comprehensively characterized through peptide mapping of proteolytic digests where deamidation can be tracked to a specific asparagine residue. Promega‘s ISOQUANT kit exploits the ability of the enzyme L-isoaspartyl methyltransferase (PIMT) to transfer the active methyl group of S-adenosyl-L-methionine (SAM) onto the free alpha-carboxyl of iso-aspartate to form an O-methyl ester. The isoaspartyl methyl ester rapidly breaks down at neutral pH to form the cyclic imide with concomitant release of methanol.

1.28.11 Methionine Oxidation Methionine oxidation is another common PTM which is minimized within the cell by the methionine sulphoxide reductase pathway. Once secreted, vulnerable methionine residues can be oxidized (particularly in protein-free or low protein culture media), but formulation excipients can help protect the residues. Methionine oxidation has been reported in MAbs during bioprocessing and after long-term storage.14 Methionine oxidation in the Fc portion of IgG1 adversely affects both MAb structure and stability and can decrease its binding to protein A and protein G resins that are commonly used in capture DSP. No rapid method for methionine oxidation detection exists, but it can be detected by peptide mapping with MS detection, rpHPLC, HIC, and weak cation-exchange chromatography. Currently, only MS can determine which methionine residue within the protein is oxidized.

1.28.12 Surface Plasmon Resonance Many companies are now replacing traditional enzyme-linked immunosorbant assay (ELISA) methods for measuring product concentrations with surface plasmon resonance (SPR) methods. Concentration is determined by monitoring the interaction of a molecule with a prepared sensor surface in the presence of a target molecule in solution (solution inhibition) or excess analyte (surface competition). Concentrations down to the nanomolar range can be measured (for IgG this equates to 0.15 mg l1). Not only does this provide more accurate concentrations but it also provides kinetic (ka and kd) information on the ligand–protein interaction that can reveal any damage to the binding site.15 In our lab, we are designing an interface from the SPR instrument to the bioreactor for at-line measurements of product concentrations and binding kinetics. Not only will this powerful technology detect product damage under different bioreactor conditions, it is also capable of predicting several antibody functions such as half-life, complement activation, and free radical or antibodydependent cell cytotoxicity (ADCC) mechanisms of cancer cell killing.

1.28.13 Conclusions Substantial progress has been made in recent years in understanding the bottlenecks to protein secretion and modifications through a combination of cellular and analytical techniques. However, many challenges still exist to achieve consistently high yields for biopharmaceutical production while avoiding posttranslational modifications that may damage protein stability, efficacy, shelf-life, and minimize immunogenicity.

Acknowledgments The author wishes to thank Raymond Tyther, Lisa Murphy, Virginie Terraube, Paula Meleady and also thanks the support from the Irish Industrial Development Association (IDA).

See Also: 1.05 Structure and Biosynthesis of Glycoprotein Carbohydrates; 1.30 Glycomics.

References 1. Jenkins, N.; Murphy, L.; Tyther, R. Post-translational Modifications of Recombinant Proteins: Significance for Biopharmaceuticals. Mol. Biotechnol. 2008, 39, 113–118. 2. Chirino, A. J.; Mire-Sluis, A. Characterizing Biological Products and Assessing Comparability Following Manufacturing Changes. Nat. Biotechnol. 2004, 22, 1383–1391.

404

Posttranslation Modifications Other Than Glycosylation

3. Arnold, J. N.; Wormald, M. R.; Sim, R. B.; et al. The Impact of Glycosylation on the Biological Function and Structure of Human Immunoglobulins. Annu. Rev. Immunol. 2007, 25, 21–50. 4. Meleady, P.; Henry, M.; Gammell, P.; et al. Proteomic Profiling of CHO Cells With Enhanced RhBMP-2 Productivity Following Co-expression of PACEsol. Proteomics 2008, 8, 2611–2624. 5. Alete, D. E.; Racher, A. J.; Birch, J. R.; et al. Proteomic Analysis of Enriched Microsomal Fractions from GS-ns0 Murine Myeloma Cells With Varying Secreted Recombinant Monoclonal Antibody Productivities. Proteomics 2005, 5, 4689–4704. 6. Fox, S. R.; Patel, U. A.; Yap, M. G.; et al. Biotechnology and Bioengineering Maximizing Interferon-gamma Production by Chinese Hamster Ovary Cells Through Temperature Shift Optimization: Experimental and Modeling. Biotechnol. Bioeng. 2004, 85, 177–184. 7. Mohan, C.; Park, S. H.; Chung, J. Y.; et al. Effect of Doxycycline-regulated Protein Disulfide Isomerase Expression on the Specific Productivity of Recombinant CHO Cells: Thrombopoietin and Antibody. Biotechnol. Bioeng. 2007, 98, 611–615. 8. Schroder, M. The Unfolded Protein Response. Mol. Biotechnol. 2006, 34, 279–290. 9. Brewer, J. W.; Hendershot, L. M. Building an Antibody Factory: A Job for the Unfolded Protein Response. Nat. Immunol. 2005, 6, 23–29. 10. Hermeling, S.; Crommelin, D. J.; Schellekens, H. W.; et al. Structure–immunogenicity Relationships of Therapeutic Proteins. Pharmaceut. Res. 2004, 21, 897–903. 11. Cromwell, M. E.; Hilario, E.; Jacobson, F. Protein Aggregation and Bioprocessing. AAPS J. 2006, 8, E572–E579. 12. Lin, J. J.; Meyer, J. D.; Carpenter, J. F.; et al. Stability of Human Serum Albumin during Bioprocessing: Denaturation and Aggregation During Processing of Albumin Paste. Pharmaceut. Res. 2000, 17, 391–396. 13. Chelius, D.; Rehder, D. S.; Bondarenko, P. V. Identification and Characterization of Deamidation Sites in the Conserved Regions of Human Immunoglobulin Gamma Antibodies. Anal. Chem. 2005, 77, 6004–6011. 14. Liu, D.; Ren, D.; Huang, H.; et al. Structure and Stability Changes of Human IgG1 Fc as a Consequence of Methionine Oxidation. Biochemistry 2008, 47, 5088–5100. 15. Gurbaxani, B. M.; Morrison, S. L. Development of New Models for the Analysis of Fc–fcrn Interactions. Mol. Immunol. 2006, 43, 1379–1389. 16. Chung, J. Y.; Lim, S. W.; Hong, Y. J.; Hwang, S. O.; Lee, G. M. Effect of Doxycycline-regulated Calnexin and Calreticulin Expression on Specific Thrombopoietin Productivity of Recombinant Chinese Hamster Ovary Cells. Biotechnol. Bioeng. 2004, 89, 539–546.

Relevant Websites http://books.google.co.uk – Google Books; Protein Misfolding, Aggregation and Conformational Diseases. http://ecb_icbm.med.uchile.cl – Hetz Lab; Laboratory of Cellular Stress and Biomedicine. http://www.impactanalytical.com – Impact Analytical. http://media.wiley.com – Methionine oxidation. http://www.nature.com – Nature.com. http://www.rsc.org – RSC Advancing the Chemical Sciences; Handbook of Surface Plasmon Resonance. http://themedicalbiochemistrypage.org – The Medical Biochemistry Page; Glycoproteins. http://www.proteinchemist.com – Welcome to mach7; Dynamic Light Scattering.

1.29

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories

J McLeod and DC James, University of Sheffield, Sheffield, United Kingdom © 2011 Elsevier B.V. All rights reserved. This is a reprint of J. McLeod, D.C. James, 1.31 - Engineering Protein Folding and Secretion in Eukaryotic Cell Factories, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 419-426.

1.29.1 1.29.2 1.29.3 1.29.4 1.29.5 1.29.6 References

Introduction Direct Engineering of Recombinant Protein Folding and Assembly Engineering the Regulation of Protein Folding and Assembly: The Unfolded Protein Response Glycosylation Engineering for Improved Protein Processing Engineering of the Secretory Apparatus Mathematical Modeling of Recombinant Protein Synthesis and Secretion

405 406 407 407 408 409 410

Glossary Binding immunoglobulin protein (BiP) An endoplasmic reticulum (ER) chaperone protein which binds native proteins as they enter the ER lumen and maintains them in a state competent for subsequent folding and oligomerization. Chaperones A class of proteins which facilitate the proper folding of native proteins by binding to or stabilizing the unfolded or partially folded protein state. N-glycosylation Co-translational modification necessary for glycoprotein folding and secretion. Protein disulfide isomerase (PDI) An enzyme which catalyzes the rearrangement of disulfide bonds during protein folding. Soluble N-ethylmaleimidine-sensitive-factor attachment receptor (SNARE) Protein superfamily involved with membrane fusion, intracellular protein trafficking, and secretory processes. Unfolded protein response (UPR) An adaptive response triggered by ER stress resulting in inhibition of global protein synthesis, and the selective transcription and translation of specific proteins, which helps the cell to deal with ER stress.

1.29.1

Introduction

Mammalian expression systems are utilized for the production of the majority of recombinant proteins for therapeutic and research purposes. The sustained rapid growth seen in this sector over the past two decades has been driven by the success of therapeutic recombinant proteins in clinical trials, with monoclonal antibody (mAb) therapeutics representing the second largest class of biopharmaceutical products in development,1 of which
18–29% in development succeed to market.2 With the annual growth rate of the recombinant mAb sector alone forecast to be 21%,3 it is anticipated that biopharmaceuticals will comprise 30% of the total pharmaceutical market by 2015. Chinese hamster ovary (CHO) cells are by far the most widely utilized cell type used to produce correctly folded proteins with the necessary posttranslational modifications, such as glycosylation,4 although other mammalian systems are in use, such as the NS0 and Sp2/0 murine myeloma cell lines. Nonmammalian production systems, such as bacteria, yeast, and transgenic plants, are currently used for less complex biomolecules and are being developed further to enable increased protein folding and mammalian-like N-glycosylation.5–7 Typically, mammalian cell-based production systems capable of generating multikilogram quantities of product are required to support administration of relatively high doses (>100 mg) of therapeutic proteins in the clinic. This has placed significant demands on the biopharmaceutical industry to develop high-yielding production systems employing mammalian host cells. Accordingly, over the last 15 years extensive empirical optimization of mammalian cell-based production systems has substantially increased both volumetric concentration of recombinant product and shortened cell-line development time.8 In the case of recombinant mAb’s, volumetric productivities exceeding 5 g l1 are now achievable in a drastically reduced development time.9 This has largely been a consequence of the systematic optimization of viable cell lifetime10 and the advent of rapid screening technologies able to identify and isolate productive transfectants.11 However, targeted cell-engineering approaches to improve specific limiting functions within the host cell background are anticipated to provide step changes in cell-specific production rates. Here, we review the current status of engineering approaches targeted to improve the folding and secretion of many recombinant proteins within different cell expression systems.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00026-4

405

406

1.29.2

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories

Direct Engineering of Recombinant Protein Folding and Assembly

Consequently, the process of protein folding within the ER has been the target of many cellular engineering strategies. ER molecular chaperones, such as BiP (GRP78), protein disculfide isomerase (PDI), and calnexin/calreticulin, have been most intensively studied. BiP facilitates the folding of secreted and transmembrane proteins within the ER lumen by binding the nascent polypeptide chain to prevent misfolding during translation permitting further folding and assembly reactions. PDI catalyzes intra- and intermolecular disulfide bond formation and isomerization as proteins fold, stabilizing native protein structure. Both BiP and PDI have been shown to be present at higher levels in mammalian cell factories with an increased specific production rate, qP.13 Calnexin and calreticulin are ER lectins and molecular chaperones that transiently bind newly synthesized glycoproteins, together functioning in a folding and assembly quality control mechanism prior to ER exit.14 Engineering strategies to increase the BiP within the ER have therefore aimed to increase protein folding to the native state while preventing aggregation of misfolded proteins within the ER which can lead to the unfolded protein response (UPR). Engineering levels of BiP to increase protein folding have given mixed results across numerous expression platforms. Single overexpression of this protein has shown improvements in protein secretion for approximately half of all heterologous proteins studied12,15–20 with studies in yeast having reported both an increase in protein productivity21 and no effect on productivity22 with elevated levels of BiP; whereas studies in mammalian cells demonstrated that the downregulation of BiP expression was fundamental to the increased release of protein from the ER.17,23 It has been postulated that insufficient levels of adenosine triphosphate (ATP) or a lack of co-chaperones such as Lhs1p within the ER environment may contribute to the limiting of BiP functions within some of the overexpression models. Additionally, increased levels of BiP within this cellular organelle could stall GRP94, the calnexin–calreticulin chaperone cycle, and UPR machinery, due to the hierarchy of ER luminal chaperone systems.12 As with BiP, engineering the overexpression of PDI has not consistently led to an improvement in protein secretion. Studies using yeast expression systems have shown much promise, for example, significant increases in protein productivity with PDI overexpression were observed in Pichia pastoris16,24,25 and Saccharomyces cerevisiae.21,26 However, expression of PDI within the more complex mammalian cell background has yielded variable results on product yield. An inducible PDI overexpression system generated by Kitchin and Flickinger within a hybridoma cell line showed no increase in antibody productivity. Similarly, both PDI knockdown and overexpression approaches failed to show significant increases in antibody productivity in both NS0 and CHO cells.27 Moreover, thrombopoietin levels were unaffected by inducible PDI overexpression within CHO cells, whereas a mild increase in productivity was observed with antibodies within the same system.28 Indeed, a recent study involving both knockdown and overexpression of PDI in NS0 and CHO cell lines stably expressing an IgG4 mAb concluded that PDI did not contribute to increasing productivity, as this was not the limiting reaction for protein synthesis within these cells.29 In contrast, Borth et al. demonstrated a 40% increase in antibody productivity with PDI overexpression in CHO cells. An alternative method of engineering disulfide bond formation and reducing the free thiol content of proteins (which may lead to protein aggregation and misfolding), within recombinant IgG molecules produced in CHO cells, is to add small amounts of the oxidizing reagent copper sulfate. A 10-fold reduction in the percentage of free thiol was observed with additions of up to 100 mM.30 The rate of PDI-catalyzed disulfide bond formation depends upon the regeneration of oxidized PDI. Upstream factors in this process such as glutathione availability and Ero1p31 activity, which is responsible for PDI oxidation, may become rate limiting in PDI overexpressing cells, hindering its function within the cell factory. Indeed, increasing Ero1p levels in mammalian cells has been shown to increase the level of antibody oxidation, while expression of a mutant inactive Ero1p decreased protein oxidation.32 Overexpression of a catalytically inactive PDI isoform also increased protein secretion within a yeast model system producing human lysozyme by increasing the “holdase” capacity of the ER, preventing aggregation of unfolded polypeptides.33 Similarly, PDI overexpression has been shown in a number of studies to increase secretion of protein products which do not contain any disulfide bonds, such as b-glucosidase and human parathyroid hormone.34,35 It is thought that substrate specificity may be responsible for the lack of effect with PDI overexpression in certain systems and suggested that an alternative holdase such as GRP94, which is responsible for acting as a molecular chaperone to partially folded polypeptides, may work as a substitute in such cases.12 Studies examining the effects of calnexin and calreticulin overexpression have shown improved production of heterologous proteins in both fungal and mammalian systems.15,36,37 However, it is unlikely that these lectins are limiting within the ER; therefore, overexpression should not increase the calnexin–calreticulin cycle capacity for folding and thus productivity. It is thought instead that the increased levels of calnexin and calreticulin may inactivate other lectins such as ER degradation enhancing mannosidase-like protein (EDEM), which target proteins for degradation, thus reducing heterologous protein turnover and increasing productivity.38,39 Typically, the effects of molecular chaperone overexpression to increase protein folding and secretion in production systems depend on the expression system used, the target proteins, and the chaperones concerned. These results reflect the idiosyncratic nature of protein expression systems, where all cell lines present different functional limitations in terms of overexpression of heterologous proteins, and indeed all proteins have different folding requirements, meaning each engineering strategy must be tailored to the individual host cell and recombinant protein under investigation. Additionally, in order to effectively improve protein production through the use of molecular chaperones, the targeting of single components in isolation of such a complex system as the secretory pathway may not be the best approach. The co-overexpression of a number of chaperones, co-chaperones, and secretory apparatus in a functionally meaningful ratio to modulate the secretion machinery in a global fashion may be necessary.

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories

1.29.3

407

Engineering the Regulation of Protein Folding and Assembly: The Unfolded Protein Response

The unfolded protein response (UPR) is found to function in all eukaryotic cells as a cellular stress response to a buildup of unfolded or misfolded proteins in the ER of eukaryotic organisms. Induction of this response causes the upregulation of numerous genes involved in protein folding, secretion, degradation, and expansion of the ER, to re-dress the balance within the ER of protein load with folding and secretory capacity. Hence, the possibility of engineering functional effectors of the UPR to promote protein folding and secretion within host cell factories has shown some promising results. Inositol requiring kinase 1 (IRE1) was first identified in S. cerevisiae and shown to be essential in the cellular response to an accumulation of unfolded proteins.40,41 Cox et al.40 showed IRE1 to be a type I transmembrane protein with serine/threonine kinase activity, where IRE1/ mutants were unable to activate the transcription of ER molecular chaperones. It was later shown that IRE1 operates upstream of the transcription factor HAC1 which is responsible for the expression of genes involved in the UPR or ER stress response.42 Upon activation of IRE1, its RNase activity is initiated which gives rise to the splicing of HAC1 messenger RNA (mRNA).42,43 Without this splicing, which removes a 30 intron, HAC1 mRNA cannot be translated.43,44 Following the splicing and translation of HAC1 mRNA, the transcriptional activator translocates to the nucleus where it is active in globally upregulating genes which contain an unfolded protein response element (UPRE). These include factors involved in protein folding and modification, ER-Golgi transport, and ER-associated protein degradation (ERAD).45 The mammalian UPR has a high level of complexity with three ER-localized UPR-signaling components: protein kinase-like ER kinase (PERK), IRE1, and activating transcription factor 6 (ATF6). PERK is the UPR signaling component that causes the most immediate response to ER stress. It causes a reversible and transient attenuation of mRNA translation.46 When PERK becomes activated, it homodimerizes and undergoes trans-autophosphorylation.47 This active form of PERK phosphorylates the downstream eukaryotic initiation factor 2a (eIF2a) to inactivate it, preventing translation initiation and causing a general attenuation of protein synthesis within the cell.48–52 However, eIF2a still retains an ability to initiate the selective translation of a few mRNAs, one of the major ones being ATF4.52 ATF4 activation leads to the transcription of genes which play a role in anti-oxidative stress, amino acid biosynthesis, ER chaperones, and apoptotic factors.52,53 Translational attenuation is stopped when eIF2a becomes dephosphorylated by the stress-induced phosphatase growth arrest and DNA damage-inducible gene 34 (GADD34) whose expression is induced by ATF4.54 In mammals, two IRE1 homologs, IRE1a and IRE1b, exist, which have been found to be expressed in different tissues.55,56 Similar to the yeast homolog of IRE1, they promote the essential splice event of the mRNA of the mammalian homolog of Hac1, X-box-binding protein 1 (XBP1), allowing translation.55,57,58 Following translation, XBP1 translocates to the nucleus where it can activate a wide variety of target genes containing ER stress-response element (ERSE) promoters.59 Additionally, XBP1 is able to upregulate ERAD by activating genes such as EDEM and removing the excess misfolded protein from the ER.45,60 ATF6 is the third mammalian UPR-signaling component. Like IRE1, it is also found in the membrane of the ER and has two mammalian forms, ATF6a and ATF6b.61 Under normal physiological conditions, ATF6 is bound to BiP within the ER; however, following induction of the UPR, BiP is released allowing ATF6 to be transported to the Golgi where it is cleaved to its active form.62 ATF6 is then free to translocate to the nucleus where it induces expression of target genes containing ATF6/cAMP response elements or ERSE with its cofactor NF-Y.58 ATF6 target genes include the ER molecular chaperones such as GRP78, GRP94, and PDI.63 The transactivators of the UPR are targets of whole organelle cell-engineering studies to increase protein production. The XBP1 yeast homolog HAC1 has been used to improve production of an heterologous protein in S. cerevisiae.64 Inactivation of HAC1 caused a decrease (70–75% reduction) in a-amylase production and a reduction of the UPR, whereas overexpression of HAC1 caused a 70% increase in the production of the heterologous protein. In a strain of P. pastoris, overexpression of S. cerevisiae HAC1 gave rise to a 1.3-fold increase in full antibody production and a 1.5-fold increase in production of a human antibody fragment.24 Studies carried out overexpressing XBP-1 in mammalian cells have demonstrated increases in the production of secreted alkaline phosphatase (SEAP) by sixfold and human vascular endothelial growth factor by up to fivefold, by expanding the secretory capacity of the cell.65 When overexpressed in CHO-DG44 cells stably producing a therapeutic antibody, XBP1(s) increased production by 40%.66 However, in one such study, Becker et al.66 reported difficulties in generating stable XBP-1(s) expressing antibody cell lines, indicating a negative selection pressure imposed upon cell lines expressing high XBP-1(s). Conversely, stable production of mAbs, interferon-g, and antithrombin-III was not enhanced by overexpression of active XBP-1.67,68 This demonstrates the cell line-specific effects of ER expansion on protein production, where global expansion of the secretory pathway, by such means as XBP-1 overexpression, may prove to be beneficial in host cells where the secretory pathway is forming a bottleneck to protein production, but is not effective in cell lines where production is limited further upstream in the production process. The use of ATF6 has not been well studied thus far, though its overexpression has recently been shown to result in ER expansion via a different mechanism to that of XBP-1,69 making it a possible new target for the engineering of cell factories. PERK is not considered as such a potential target for modulation in recombinant protein producing cells due to its effects in attenuating translation. However, some positive effects have been seen with overexpression of its downstream activators ATF4 and GADD34, where overexpression in CHO-K1-derived cell lines resulted in increases in the cell-specific productivity of antithrombin III.68,70

1.29.4

Glycosylation Engineering for Improved Protein Processing

Glycosylation is one of the most common posttranslational modifications to occur in protein biosynthesis. Initial glycosylation takes place within the ER and is carried out simultaneously with protein translation into the ER lumen. Protein glycosylation plays

408

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories

many key roles in protein folding, oligomer assembly, and secretion processes of the relevant protein, as well as in clearance of the glycoprotein from the bloodstream. As nascent polypeptides enter the ER through the Sec61 translocon complex, they are scanned by oligosaccharyltransferase (OST) for glycosylation sites. Carbohydrate groups can be added to proteins either via an N-glycosidic bond to the R-group of an asparagines residue within the consensus sequence Asn-X-Ser/Thr, or via an O-glycosidic bond to the R-group of serine or threonine; these account for both N- and O-glycosylation, respectively. Additionally, carbohydrate groups form part of the glycophosphatidylinositol anchor which serve to bind proteins to the cell membrane. The presence of such consensus sequences within the primary protein sequence does not ensure site glycosylation. Glycosylation of any consensus sequence within a protein is dependent upon the position within the protein and its conformation and the host cell type used for protein expression. Further to this, the type of sugar residues from which the glycan group is composed can vary again depending on the protein conformation and position of the glycosylation site within the structure and the cell type used for protein expression.71 Simple organisms such as bacteria are not capable of posttranslational modifications such as glycosylation, and so for production of more complex proteins eukaryotic host cells are required. The majority of yeast strains is known to add large numbers of additional mannose residues to the core oligosaccharide,72 which can affect therapeutic protein efficacy, such as with the hepatitis B vaccine.73 This can be overcome by the use of mutant yeast strains, such as mnn-9 and ngd-29, restricted in their N-glycosylation capabilities to limited mannose content within the core oligosaccharide.73,74 However, the O-glycosylation sites used by mammalian cells and yeast are also different.75 Mammalian rodent cell lines are most commonly used for the production of large complex therapeutic proteins and have glycosylation patterns more similar to that of humans. However, within humans the enzyme a1–3 galactosyltransferase is silenced,76 and so viable production cell lines for protein therapeutics must also have inactivated this enzyme to allow comparable glycan profiles and prevent an immune response in the patient. Fortunately, this enzyme is not active within the mouse NS0, rat Y0, and CHO cell lines (the industrial production vehicle of choice),77–79 and these cell lines can be further genetically modified to resemble the human glycoprofile by transfection of the appropriate glycosyltransferases.71 Immediately following transfer of the initial glycan group to the protein, it is susceptible to processing via a number of endogenous enzymes before it reaches its final form. These processing steps take place both within the ER and later within the Golgi apparatus prior to secretion. Correct processing of the glycan group by the glycosylation machinery is necessary for correct folding and secretion capability of the protein. Within the ER lumen, an N-glycan group of a newly synthesized protein having been processed by glucosidase-I and glucosidase-II to Glc1Man5-9GlcNAc is bound by the lectin-like molecular chaperones calnexin and calreticulin. If the glycoprotein is correctly folded within the ER, the terminal glucose is removed by glucosidase-II and the protein released from the calnexin/calreticulin cycle to leave the ER into the later secretory system. However, if the protein structure is not folded correctly, the terminal glucose is once again attached by the action of UDP-glucose:glycoprotein glucosyltransferase (UGGT), which discriminates between folded and unfolded substrates.80 These unfolded glycoproteins then re-enter the calnexin/calreticulin cycle in an attempt to refold. Ultimately if the protein fails to fold correctly, it is targeted for degradation and removed from the ER.80,81 As the addition of N-glycan groups to proteins aids their folding within the ER, it has been the target of engineering studies to improve folding and secretion of recombinant proteins. There has been some evidence to suggest that increasing the number of glycosylation sites within production proteins serves to increase their secretion and productivity. Within S. cerevisiae and P. pastoris secretion of cutinase and llama, VHH antibody fragments were increased by up to fivefold through the introduction of an additional glycosylation site at either the N- or C-terminal, with greater increases in secretion noted at when the glycosylation site was incorporated at the N-terminal of the proteins.82 This effect was also shown in mammalian cells where N-glycosylation sequons were introduced into seven different model recombinant proteins both with and without endogenous N-glycosylation sites with the result that six showed increases in secretion from HEK293 cells, again with increased efficacy noted in N-glycosylation sites introduced at the N-terminal of the protein compared to the C-terminal. This was considered to be due to the possibility of the nascent polypeptide having already folded prior to attachment of the glycan group at the C-terminal end of the protein.83 Overall, protein secretion increases in proportion to the number of N-glycan oligosaccharides, which reportedly increases thermal stabilization of the proteins by the polysaccharide chains, and increases in proportion with the number of attached chains, though the size of the attached glycan has little effect. This enhanced thermodynamic stabilization is due to an increase in the destabilization of the unfolded state by glycan attachment to the polypeptide.84 Despite the increased secretion of proteins when engineered to contain more N-glycosylation sites, there is evidence to suggest that the glycosylation of these proteins is not as complete when turned over more quickly. For example, SEAP produced by CHO cell cultures was found to contain a higher relative glycan content and increased extent of sialylation when produced at a slower rate in a semicontinuous perfusion cultures compared to more rapid repeated fed-batch cultures.85

1.29.5

Engineering of the Secretory Apparatus

The rate at which folded and assembled proteins are trafficked between cellular compartments can be considered a substantial bottleneck in the mammalian cell synthetic process. However, new strategies to increase the rate of these energy-intensive, complex processes have arisen from an understanding of the cellular machinery involved in the control of vesicle formation and trafficking along microtubules. The use of bacterial cell factories has been hindered by their inability to secrete the heterologous product, making downstream purification from host cell protein difficult. Escherichia coli mutants with defects in outer membrane structures have been exploited for the extracellular production of recombinant proteins. This approach involves first, fusing an N-terminal sequence to the recombinant gene so that the recombinant fusion protein translocates to the periplasmic space. Second, inducing specific mutations

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories

409

in the outer membrane structures that confer increased permeability to the structure, this allows recombinant protein in the periplasm to escape into the culture medium.86 An advantage of this approach is that it is nonspecific, applies to all proteins expressed in such systems, and therefore no specific engineering is required based only on the protein product in question. However, a potential disadvantage is the effect of membrane mutations on cell growth. For example, the absence of a cell wall in L-form E. coli strains makes this type of mutant extremely fragile and sensitive to the environmental stresses typically encountered during large-scale fermentations.87 It has been recently demonstrated that this negative effect on cell growth is not always apparent when the cell wall mutation is carefully selected. For example, a single lpp gene deletion mutant has been shown to have a near identical growth rate when compared with the wild-type strain of E. coli.88 This mutation has been utilized for the high levels of secretion of maltose-binding protein (MalE), xylanase, and cellulose.89 Genome-wide screening of a yeast cDNA library in S. cerevisiae identified the cell wall proteins CCW12, SED1, and CWP2 as beneficial to the secretion of various antibody fragments.90 These proteins have generally been implicated in providing cell wall stability and resistance to stresses. For example, CCW12 deletion or overexpression increases the sensitivity to known cell wall perturbants calcofluor white and Congo red,91 deletion of SED1 made stationary-phase cells more sensitive to Zymolase treatment,92 and deletion of CWP2, like CCW12, increased sensitivity to calcofluor white and Congo red while also increasing the sensitivity of exponential growing cells to Zymolase treatment.93 Thus, it may be concluded that the stresses imposed by heterologous protein display and secretion are diminished by overexpression of cell wall proteins, although further study will be required to elucidate the mechanism whereby the cell wall proteins that are identified assist secretion and display. The eukaryotic secretory apparatus consists of a complex network of vesicles and secretory organelles, to manage the precise distribution of molecular cargo between cellular organelles as well as the secretion of recombinant proteins from cell factories. Fusion of secretory vesicles and their target organelle membranes along the secretory pathway is driven by different soluble Nethylmaleimidine-sensitive-factor attachment receptor (SNARE) and Sec1/Munc18 (SM)-like proteins. The assembly of t-SNAREs on the target membrane and v-SNAREs on the transport vesicle induces the formation of trans-SNARE complexes or SNAREpins, which bring the membranes together and catalyze membrane fusion.94,95 Overexpression of the plasma membrane t-SNAREs Sso1 and Sso2 has been reported to increase secretion levels of laccase,96 a-amylase,97,98 and invertase98 by up to sixfold in S. cerevisiae. However, Gasser et al.99 reported only a moderate improvement (20%) in secretion of an antibody fragment when SSO2 was overexpressed in P. pastoris. This difference in response to overexpression of SSO may be due to the direct correlation between the amount of secreted recombinant product and Sso protein levels,98 which were up to 20 times overproduced (due to multiple copies of the SSO gene) in S. cerevisiae,97,98 compared to an additional one copy into the P. pastoris genome.99 Increasing evidence suggests that the specificity of membrane-fusion events is regulated by Sec1/Munc18 (SM) proteins.100,101 A number of SM proteins have been shown to promote membrane fusion including Sly1,102,103 Sec1,104 Munc18a,105 and Munc18c.106 Accordingly, transgenic expression of the SM proteins Sly1 and Munc18c has been shown to significantly increase SNARE-based fusion of ER-to-Golgi and Golgi-to-plasma membrane vesicle trafficking, increasing the productivity of CHO cells expressing an IgG protein over eightfold, up to 40 pg/cell/day, and showing up to a fivefold increase in secretion of SEAP, secreted a-amylase and VEGF121.107 Furthermore, transgenic expression of Munc18b enhanced heterologous protein production in HeLa, HEK-293, and HT-1080 mammalian cell lines.108 The mechanism by which vesicles for TGN to plasma membrane transport are generated is only starting to be understood.109 Central to this mechanism is the recruitment of members of the protein kinase D (PKD) family to the TGN. PKD is first recruited to the TGN through binding to diacylglycerol (DAG)110 and is activated by PKC-h-mediated phosphorylation.111 PKD, in turn, phosphorylates and activates phosphatidylinositol 4-kinase IIIb (PI4KIIIb) which appears to regulate TGN-to-plasma membrane vesicle trafficking.112 Another key player in TGN-to-plasma membrane transport is the ceramide transfer protein, CERT, which mediates ATP-dependent ceramide transport from the ER to the Golgi complex.113 CERT has been identified as a PKD substrate and it has been shown that this PKD–CERT interaction regulates the secretory activity of the Golgi complex.114 Accordingly, heterologous expression of CERT has been shown to increase production of both the plasma protein HSA and two different IgG1 mAbs in CHO cells,115 demonstrating that engineering of the secretory system is a potentially rewarding target for the increase of recombinant protein yields in systems where the bottleneck is in the final secretory stage of production.

1.29.6

Mathematical Modeling of Recombinant Protein Synthesis and Secretion

Many mathematical models of protein production have highlighted the folding and secretory systems as bottlenecks to increased productivity116–120; however, targeted engineering of these functions within the host cell has yielded variable results in different systems and with the differing approaches used. It seems that the protein folding and secretory capacity of cell factories is under the control of a complex global regulation within the cell, and greater understanding of the interactions of such systems may be required to give consistent functional enhancements. More recently, the results of a mathematical model of mAb synthesis, examining a panel of CHO cell lines with varying rates of productivity of the same recombinant IgG4 protein, have demonstrated the importance of the host cell background used in recombinant protein expression.121 This study showed that a large degree of functional heterogeneity exists between cell lines with respect to protein processing reactions, from transcription to protein folding and secretion, within a host cell population, meaning that the cells within such a population will vary with respect to the engineering targets that will be successful in increasing productivity.13,121 This level of heterogeneity can be expected to exist in all cell populations, with different constraints imposed by different recombinant proteins. It can be concluded from this and the

410

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories

variable results observed with engineering folding and secretion in different cell lines that though engineering of cellular folding and secretory mechanisms may be required to increase productivity in cells with ultimately high transcription and translation rates, these approaches must be targeted in a cell-specific and product-specific manner to result in consistent outcomes. Moreover, it may be necessary to target numerous cell processes, such as translation and protein folding in order to allow efficient folding of polypeptides at a specified translation rate within a single production vehicle, to overcome the functional redundancy of the host cell background.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43.

Walsh, G. Biopharmaceutical Benchmarks – 2003. Nat. Biotechnol. 2003, 21 (8), 865–870. Reichert, J. M.; et al. Monoclonal Antibody Successes in the Clinic. Nat. Biotechnol. 2005, 23 (9), 1073–1078. Pavlou, A. K.; Belsey, M. J. The Therapeutic Antibodies Market to 2008. Eur. J. Pharm. Biopharm. 2005, 59 (3), 389–396. Walsh, G. Biopharmaceutical Benchmarks 2006. Nat. Biotechnol. 2006, 24 (7), 769–776. Gomord, V.; et al. Production and Glycosylation of Plant-made Pharmaceuticals: The Antibodies as a Challenge. Plant Biotechnol. J. 2004, 2 (2), 83–100. Li, H.; et al. Optimization of Humanized IgGs in Glycoengineered Pichia pastoris. Nat. Biotechnol. 2006, 24 (2), 210–215. Simmons, L. C.; et al. Expression of Full-length Immunoglobulins in Escherichia coli: Rapid and Efficient Production of Aglycosylated Antibodies. J. Immunol. Meth. 2002, 263 (1–2), 133–147. Wurm, F. M. Production of Recombinant Protein Therapeutics in Cultivated Mammalian Cells. Nat. Biotechnol. 2004, 22 (11), 1393–1398. Birch, J. R.; Racher, A. J. Antibody Production. Adv. Drug Deliv. Rev. 2006, 58 (5–6), 671–685. Sauer, P. W.; et al. A High-yielding, Generic Fed-batch Cell Culture Process for Production of Recombinant Antibodies. Biotechnol. Bioeng. 2000, 67 (5), 585–597. Browne, S. M.; Al-Rubeai, M. Selection Methods for High-producing Mammalian Cell Lines. Trends Biotechnol. 2007, 25 (9), 425–432. Schroder, M. Endoplasmic Reticulum Stress Responses. Cell. Mol. Life Sci. 2008, 65 (6), 862–894. Smales, C. M.; et al. Comparative Proteomic Analysis of GS-ns0 Murine Myeloma Cell Lines With Varying Recombinant Monoclonal Antibody Production Rate. Biotechnol. Bioeng. 2004, 88 (4), 474–488. Hammond, C.; Helenius, A. Quality Control in the Secretory Pathway. Curr. Opin. Cell Biol. 1995, 7 (4), 523–529. Chung, J. Y.; et al. Effect of Doxycycline-regulated Calnexin and Calreticulin Expression on Specific Thrombopoietin Productivity of Recombinant Chinese Hamster Ovary Cells. Biotechnol. Bioeng. 2004, 85 (5), 539–546. Damasceno, L. M.; et al. Cooverexpression of Chaperones for Enhanced Secretion of a Single-chain Antibody Fragment in Pichia pastoris. Appl. Microbiol. Biotechnol. 2007, 74 (2), 381–389. Dorner, A. J.; Wasley, L. C.; Kaufman, R. J. Overexpression of GRP78 Mitigates Stress Induction of Glucose Regulated Proteins and Blocks Secretion of Selective Proteins in Chinese Hamster Ovary Cells. EMBO J. 1992, 11 (4), 1563–1571. Hsu, T. A.; Betenbaugh, M. J. Coexpression of Molecular Chaperone BiP Improves Immunoglobulin Solubility and IgG Secretion From Trichoplusia Ni Insect Cells. Biotechnol. Prog. 1997, 13 (1), 96–104. Kim, M. D.; et al. Coexpression of BiP Increased Antithrombotic Hirudin Production in Recombinant Saccharomyces cerevisiae. J. Biotechnol. 2003, 101 (1), 81–87. Smith, J. D.; Robinson, A. S. Overexpression of an Archaeal Protein in Yeast: Secretion Bottleneck at the ER. Biotechnol. Bioeng. 2002, 79 (7), 713–723. Shusta, E. V.; et al. Increasing the Secretory Capacity of Saccharomyces cerevisiae for Production of Single-chain Antibody Fragments. Nat. Biotechnol. 1998, 16 (8), 773–777. Robinson, A. S.; et al. Reduction of BiP Levels Decreases Heterologous Protein Secretion in Saccharomyces cerevisiae. J. Biol. Chem. 1996, 271 (17), 10017–10022. Dorner, A. J.; Kaufman, R. J. The Levels of Endoplasmic Reticulum Proteins and ATP Affect Folding and Secretion of Selective Proteins. Biologicals 1994, 22 (2), 103–112. Gasser, B.; et al. Engineering of Pichia pastoris for Improved Production of Antibody Fragments. Biotechnol. Bioeng. 2006, 94 (2), 353–361. Inan, M.; et al. Enhancement of Protein Secretion in Pichia pastoris by Overexpression of Protein Disulfide Isomerase. Biotechnol. Bioeng. 2006, 93 (4), 771–778. Butz, J. A.; Niebauer, R. T.; Robinson, A. S. Co-expression of Molecular Chaperones Does Not Improve the Heterologous Expression of Mammalian G-protein Coupled Receptor Expression in Yeast. Biotechnol. Bioeng. 2003, 84 (3), 292–304. Hayes, N. V.; Smales, C. M.; Klappa, P. Protein Disulfide Isomerase Does Not Control Recombinant IgG4 Productivity in Mammalian Cell Lines. Biotechnol. Bioeng. 2010, 105 (4), 770–779. Mohan, C.; et al. Effect of Doxycycline-regulated Protein Disulfide Isomerase Expression on the Specific Productivity of Recombinant CHO Cells: Thrombopoietin and Antibody. Biotechnol. Bioeng. 2007, 98 (3), 611–615. Hayes, N. V.; Smales, C. M.; Klappa, P. Protein Disulfide Isomerase Does Not Control Recombinant IgG4 Productivity in Mammalian Cell Lines. Biotechnol. Bioeng. 2010, 105 (4), 770–779. Chaderjian, W. B.; et al. Effect of Copper Sulfate on Performance of a Serum-free CHO Cell Culture Process and the Level of Free Thiol in the Recombinant Antibody Expressed. Biotechnol. Prog. 2005, 21 (2), 550–553. Tu, B. P.; Weissman, J. S. Oxidative Protein Folding in Eukaryotes: Mechanisms and Consequences. J. Cell Biol. 2004, 164 (3), 341–346. Mezghrani, A.; et al. Manipulation of Oxidative Protein Folding and PDI Redox State in Mammalian Cells. EMBO J. 2001, 20 (22), 6288–6296. Hayano, T.; Hirose, M.; Kikuchi, M. Protein Disulfide Isomerase Mutant Lacking its Isomerase Activity Accelerates Protein Folding in the Cell. FEBS Lett. 1995, 377 (3), 505–511. Powers, S. L.; Robinson, A. S. PDI Improves Secretion of Redox-inactive Beta-glucosidase. Biotechnol. Prog. 2007, 23 (2), 364–369. Vad, R.; et al. Engineering of a Pichia pastoris Expression System for Secretion of High Amounts of Intact Human Parathyroid Hormone. J. Biotechnol. 2005, 116 (3), 251–260. Conesa, A.; et al. Calnexin Overexpression Increases Manganese Peroxidase Production in Aspergillus niger. Appl. Environ. Microbiol. 2002, 68 (2), 846–851. Kato, T.; et al. Improvement of the Production of GFPuv-b1,3-n-acetylglucosaminyltransferase 2 Fusion Protein Using a Molecular Chaperone-assisted Insect-cell-based Expression System. Biotechnol. Bioeng. 2005, 89 (4), 424–433. Molinari, M.; et al. Role of EDEM in the Release of Misfolded Glycoproteins From the Calnexin Cycle. Science 2003, 299 (5611), 1397–1400. Oda, Y.; et al. EDEM as an Acceptor of Terminally Misfolded Glycoproteins Released From Calnexin. Science 2003, 299 (5611), 1394–1397. Cox, J. S.; Shamu, C. E.; Walter, P. Transcriptional Induction of Genes Encoding Endoplasmic Reticulum Resident Proteins Requires a Transmembrane Protein Kinase. Cell 1993, 73 (6), 1197–1206. Mori, K.; et al. A Transmembrane Protein With a Cdc2þ/CDC28-related Kinase Activity Is Required for Signaling From the ER to the Nucleus. Cell 1993, 74 (4), 743–756. Cox, J. S.; Walter, P. A Novel Mechanism for Regulating Activity of a Transcription Factor That Controls the Unfolded Protein Response. Cell 1996, 87 (3), 391–404. Sidrauski, C.; Walter, P. The Transmembrane Kinase Ire1p Is a Site-specific Endonuclease That Initiates mRNA Splicing in the Unfolded Protein Response. Cell 1997, 90 (6), 1031–1039.

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories

411

44. Ruegsegger, U.; Leber, J. H.; Walter, P. Block of HAC1 mRNA Translation by Long-range Base Pairing Is Released by Cytoplasmic Splicing upon Induction of the Unfolded Protein Response. Cell 2001, 107 (1), 103–114. 45. Malhotra, J. D.; Kaufman, R. J. The Endoplasmic Reticulum and the Unfolded Protein Response. Semin. Cell Dev. Biol. 2007, 18 (6), 716–731. 46. Kaufman, R. J. Regulation of mRNA Translation by Protein Folding in the Endoplasmic Reticulum. Trends Biochem. Sci. 2004, 29 (3), 152–158. 47. Liu, C. Y.; Schroder, M.; Kaufman, R. J. Ligand-independent Dimerization Activates the Stress Response Kinases IRE1 and PERK in the Lumen of the Endoplasmic Reticulum. J. Biol. Chem. 2000, 275 (32), 24881–24885. 48. Harding, H. P.; Zhang, Y.; Ron, D. Protein Translation and Folding Are Coupled by an Endoplasmic-reticulum-resident Kinase. Nature 1999, 397 (6716), 271–274. 49. Harding, H. P.; et al. Perk Is Essential for Translational Regulation and Cell Survival during the Unfolded Protein Response. Mol. Cell 2000, 5 (5), 897–904. 50. Harding, H. P.; et al. Regulated Translation Initiation Controls Stress-induced Gene Expression in Mammalian Cells. Mol. Cell 2000, 6 (5), 1099–1108. 51. Prostko, C. R.; Brostrom, M. A.; Brostrom, C. O. Reversible Phosphorylation of Eukaryotic Initiation Factor 2 Alpha in Response to Endoplasmic Reticular Signaling. Mol. Cell. Biochem. 1993, 127 and 128, 255–265. 52. Ron, D. Translational Control in the Endoplasmic Reticulum Stress Response. J. Clin. Invest. 2002, 110 (10), 1383–1388. 53. Harding, H. P.; et al. An Integrated Stress Response Regulates Amino Acid Metabolism and Resistance to Oxidative Stress. Mol. Cell 2003, 11 (3), 619–633. 54. Novoa, I.; et al. Feedback Inhibition of the Unfolded Protein Response by GADD34-mediated Dephosphorylation of EIF2alpha. J. Cell Biol. 2001, 153 (5), 1011–1022. 55. Tirasophon, W.; Welihinda, A. A.; Kaufman, R. J. A Stress Response Pathway From the Endoplasmic Reticulum to the Nucleus Requires a Novel Bifunctional Protein Kinase/ endoribonuclease (Ire1p) in Mammalian Cells. Genes Dev. 1998, 12 (12), 1812–1824. 56. Wang, X. Z.; et al. Cloning of Mammalian Ire1 Reveals Diversity in the ER Stress Responses. EMBO J. 1998, 17 (19), 5708–5717. 57. Calfon, M.; et al. IRE1 Couples Endoplasmic Reticulum Load to Secretory Capacity by Processing the XBP-1 mRNA. Nature 2002, 415 (6867), 92–96. 58. Yoshida, H.; et al. XBP1 mRNA Is Induced by ATF6 and Spliced by IRE1 in Response to ER Stress to Produce a Highly Active Transcription Factor. Cell 2001, 107 (7), 881–891. 59. Yoshida, H.; et al. Identification of the cis-acting Endoplasmic Reticulum Stress Response Element Responsible for Transcriptional Induction of Mammalian Glucose-regulated Proteins – Involvement of Basic Leucine Zipper Transcription Factors. J. Biol. Chem. 1998, 273 (50), 33741–33749. 60. Yoshida, H.; et al. A Time-dependent Phase Shift in the Mammalian Unfolded Protein Response. Dev. Cell 2003, 4 (2), 265–271. 61. Haze, K.; et al. Identification of the G13 (cAMP-response-element-binding Protein-related Protein) Gene Product Related to Activating Transcription Factor 6 as a Transcriptional Activator of the Mammalian Unfolded Protein Response. Biochem. J. 2001, 355, 19–28. 62. Haze, K.; et al. Mammalian Transcription Factor ATF6 Is Synthesized as a Transmembrane Protein and Activated by Proteolysis in Response to Endoplasmic Reticulum Stress. Mol. Biol. Cell 1999, 10 (11), 3787–3799. 63. Okada, T.; et al. Distinct Roles of Activating Transcription Factor 6 (ATF6) and Double-stranded RNA-activated Protein Kinase-like Endoplasmic Reticulum Kinase (PERK) in Transcription during the Mammalian Unfolded Protein Response. Biochem. J. 2002, 366, 585–594. 64. Valkonen, M.; Penttila, M.; Saloheimo, M. Effects of Inactivation and Constitutive Expression of the Unfolded-protein Response Pathway on Protein Production in the Yeast Saccharomyces cerevisiae. Appl. Environ. Microbiol. 2003, 69 (4), 2065–2072. 65. Tigges, M.; Fussenegger, M. Xbp1-based Engineering of Secretory Capacity Enhances the Productivity of Chinese Hamster Ovary Cells. Metab. Eng. 2006, 8 (3), 264–272. 66. Becker, E.; et al. An XBP-1 Dependent Bottle-neck in Production of IgG Subtype Antibodies in Chemically Defined Serum-free Chinese Hamster Ovary (CHO) Fed-batch Processes. J. Biotechnol. 2008, 135 (2), 217–223. 67. Ku, S. C.; et al. Effects of Overexpression of X-box Binding Protein 1 on Recombinant Protein Production in Chinese Hamster Ovary and NS0 Myeloma Cells. Biotechnol. Bioeng. 2008, 99 (1), 155–164. 68. Ohya, T.; et al. Improved Production of Recombinant Human Antithrombin III in Chinese Hamster Ovary Cells by ATF4 Overexpression. Biotechnol. Bioeng. 2008, 100 (2), 317–324. 69. Bommiasamy, H.; et al. ATF6alpha Induces XBP1-independent Expansion of the Endoplasmic Reticulum. J. Cell Sci. 2009, 122 (Pt 10), 1626–1636. 70. Omasa, T.; et al. Overexpression of GADD34 Enhances Production of Recombinant Human Antithrombin III in Chinese Hamster Ovary Cells. J. Biosci. Bioeng. 2008, 106 (6), 568–573. 71. Jenkins, N.; Parekh, R. B.; James, D. C. Getting the Glycosylation Right: Implications for the Biotechnology Industry. Nat. Biotechnol. 1996, 14 (8), 975–981. 72. Herscovics, A.; Orlean, P. Glycoprotein Biosynthesis in Yeast. FASEB J. 1993, 7 (6), 540–550. 73. Kniskern, P. J.; et al. Characterization and Evaluation of a Recombinant Hepatitis B Vaccine Expressed in Yeast Defective for N-linked Hyperglycosylation. Vaccine 1994, 12 (11), 1021–1025. 74. Lehle, L.; et al. Glycoprotein Biosynthesis in Saccharomyces cerevisiae: Ngd29, an N-glycosylation Mutant Allelic to Och1 Having a Defect in the Initiation of Outer Chain Formation. FEBS Lett. 1995, 370 (1–2), 41–45. 75. Kalsner, I.; et al. Comparison of the Carbohydrate Moieties of Recombinant Soluble Fc Epsilon Receptor (sFc epsilon RII/sCD23) Expressed in Saccharomyces cerevisiae and Chinese Hamster Ovary Cells. Different O-glycosylation Sites Are Used by Yeast and Mammalian Cells. Glycoconj. J. 1992, 9 (4), 209–216. 76. Larsen, R. D.; et al. Frameshift and Nonsense Mutations in a Human Genomic Sequence Homologous to a Murine UDP-Gal:beta-D-Gal(1,4)-D-GlcNAc Alpha(1,3)galactosyltransferase cDNA. J. Biol. Chem. 1990, 265 (12), 7055–7061. 77. Ip, C. C.; et al. Structural Characterization of the N-glycans of a Humanized Anti-CD18 Murine Immunoglobulin G. Arch. Biochem. Biophys. 1994, 308 (2), 387–399. 78. Lifely, M. R.; et al. Glycosylation and Biological Activity of CAMPATH-1H Expressed in Different Cell Lines and Grown under Different Culture Conditions. Glycobiology 1995, 5 (8), 813–822. 79. Smith, D. F.; et al. Transfer and Expression of a Murine UDP-Gal:beta-D-Gal-alpha 1,3-galactosyltransferase Gene in Transfected Chinese Hamster Ovary Cells. Competition Reactions Between the Alpha 1,3-galactosyltransferase and the Endogenous Alpha 2,3-sialyltransferase. J. Biol. Chem. 1990, 265 (11), 6225–6234. 80. Cabral, C. M.; Liu, Y.; Sifers, R. N. Dissecting Glycoprotein Quality Control in the Secretory Pathway. Trends Biochem. Sci. 2001, 26 (10), 619–624. 81. Ruddock, L. W.; Molinari, M. N-glycan Processing in ER Quality Control. J. Cell Sci. 2006, 119 (Pt 21), 4373–4380. 82. Sagt, C. M.; et al. Introduction of an N-glycosylation Site Increases Secretion of Heterologous Proteins in Yeasts. Appl. Environ. Microbiol. 2000, 66 (11), 4940–4944. 83. Liu, Y.; et al. Enhancing the Secretion of Recombinant Proteins by Engineering N-glycosylation Sites. Biotechnol. Prog. 2009, 25 (5), 1468–1475. 84. Shental-Bechor, D.; Levy, Y. Effect of Glycosylation on Protein Folding: A Close Look at Thermodynamic Stabilization. Proc. Natl. Acad. Sci. U.S.A. 2008, 105 (24), 8256–8261. 85. Lipscomb, M. L.; et al. Effect of Production Method and Gene Amplification on the Glycosylation Pattern of a Secreted Reporter Protein in CHO Cells. Biotechnol. Prog. 2005, 21 (1), 40–49. 86. Ni, Y.; Chen, R. Extracellular Recombinant Protein Production From Escherichia coli. Biotechnol. Lett. 2009, 31 (11), 1661–1670. 87. Gumpert, J.; Hoischen, C. Use of Cell Wall-less Bacteria (L-forms) for Efficient Expression and Secretion of Heterologous Gene Products. Curr. Opin. Biotechnol. 1998, 9 (5), 506–509. 88. Ni, Y.; Reye, J.; Chen, R. R. lpp Deletion as a Permeabilization Method. Biotechnol. Bioeng. 2007, 97 (6), 1347–1356. 89. Shin, H. D.; Chen, R. R. Extracellular Recombinant Protein Production From an Escherichia coli lpp Deletion Mutant. Biotechnol. Bioeng. 2008, 101 (6), 1288–1296. 90. Wentz, A. E.; Shusta, E. V. A Novel High-throughput Screen Reveals Yeast Genes That Increase Secretion of Heterologous Proteins. Appl. Environ. Microbiol. 2007, 73 (4), 1189–1198. 91. Mrsa, V.; et al. Deletion of New Covalently Linked Cell Wall Glycoproteins Alters the Electrophoretic Mobility of Phosphorylated Wall Components of Saccharomyces cerevisiae. J. Bacteriol. 1999, 181 (10), 3076–3086.

412

Engineering Protein Folding and Secretion in Eukaryotic Cell Factories

92. Shimoi, H.; et al. Sed1p Is a Major Cell Wall Protein of Saccharomyces cerevisiae in the Stationary Phase and Is Involved in Lytic Enzyme Resistance. J. Bacteriol. 1998, 180 (13), 3381–3387. 93. van der Vaart, J. M.; et al. Identification of Three Mannoproteins in the Cell Wall of Saccharomyces cerevisiae. J. Bacteriol. 1995, 177 (11), 3104–3110. 94. Chen, Y. A.; Scheller, R. H. SNARE-mediated Membrane Fusion. Nat. Rev. Mol. Cell Biol. 2001, 2 (2), 98–106. 95. Ungar, D.; Hughson, F. M. SNARE Protein Structure and Function. Annu. Rev. Cell Dev. Biol. 2003, 19, 493–517. 96. Larsson, S.; Cassland, P.; Jonsson, L. J. Development of a Saccharomyces cerevisiae Strain With Enhanced Resistance to Phenolic Fermentation Inhibitors in Lignocellulose Hydrolysates by Heterologous Expression of Laccase. Appl. Environ. Microbiol. 2001, 67 (3), 1163–1170. 97. Toikkanen, J. H.; Sundqvist, L.; Keranen, S. Kluyveromyces Lactis SSO1 and SEB1 Genes Are Functional in Saccharomyces cerevisiae and Enhance Production of Secreted Proteins when Overexpressed. Yeast 2004, 21 (12), 1045–1055. 98. Ruohonen, L.; et al. Enhancement of Protein Secretion in Saccharomyces cerevisiae by Overproduction of Sso Protein, a Late-acting Component of the Secretory Machinery. Yeast 1997, 13 (4), 337–351. 99. Gasser, B.; et al. Transcriptomics-based Identification of Novel Factors Enhancing Heterologous Protein Secretion in Yeasts. Appl. Environ. Microbiol. 2007, 73 (20), 6499–6507. 100. Whyte, J. R.; Munro, S. Vesicle Tethering Complexes in Membrane Traffic. J. Cell Sci. 2002, 115 (Pt 13), 2627–2637. 101. Koumandou, V. L.; et al. Control Systems for Membrane Fusion in the Ancestral Eukaryote; Evolution of Tethering Complexes and SM Proteins. BMC Evol. Biol. 2007, 7, 29. 102. Peng, R.; Gallwitz, D. Sly1 Protein Bound to Golgi Syntaxin Sed5p Allows Assembly and Contributes to Specificity of SNARE Fusion Complexes. J. Cell Biol. 2002, 157 (4), 645–655. 103. Peng, R.; Gallwitz, D. Multiple SNARE Interactions of an SM Protein: Sed5p/Sly1p Binding Is Dispensable for Transport. EMBO J. 2004, 23 (20), 3939–3949. 104. Scott, B. L.; et al. Sec1p Directly Stimulates SNARE-mediated Membrane Fusion In Vitro. J. Cell Biol. 2004, 167 (1), 75–85. 105. Shen, J.; et al. Selective Activation of Cognate SNAREpins by Sec1/Munc18 Proteins. Cell 2007, 128 (1), 183–195. 106. Latham, C. F.; et al. Molecular Dissection of the Munc18c/syntaxin4 Interaction: Implications for Regulation of Membrane Trafficking. Traffic 2006, 7 (10), 1408–1419. 107. Peng, R. W.; Fussenegger, M. Molecular Engineering of Exocytic Vesicle Traffic Enhances the Productivity of Chinese Hamster Ovary Cells. Biotechnol. Bioeng. 2009, 102 (4), 1170–1181. 108. Peng, R. W.; et al. The Vesicle-trafficking Protein Munc18b Increases the Secretory Capacity of Mammalian Cells. Metab. Eng. 2010, 12 (1), 18–25. 109. Ghanekar, Y.; Lowe, M. Protein Kinase D: Activation for Golgi Carrier Formation. Trends Cell Biol. 2005, 15 (10), 511–514. 110. Baron, C. L.; Malhotra, V. Role of Diacylglycerol in PKD Recruitment to the TGN and Protein Transport to the Plasma Membrane. Science 2002, 295 (5553), 325–328. 111. Diaz Anel, A. M.; Malhotra, V. PKCeta Is Required for Beta1gamma2/beta3gamma2- and PKD-mediated Transport to the Cell Surface and the Organization of the Golgi Apparatus. J. Cell Biol. 2005, 169 (1), 83–91. 112. Hausser, A.; et al. Protein Kinase D Regulates Vesicular Transport by Phosphorylating and Activating Phosphatidylinositol-4 Kinase IIIbeta at the Golgi Complex. Nat. Cell Biol. 2005, 7 (9), 880–886. 113. Hanada, K.; et al. Molecular Machinery for Non-vesicular Trafficking of Ceramide. Nature 2003, 426 (6968), 803–809. 114. Fugmann, T.; et al. Regulation of Secretory Transport by Protein Kinase D-mediated Phosphorylation of the Ceramide Transfer Protein. J. Cell Biol. 2007, 178 (1), 15–22. 115. Florin, L.; et al. Heterologous Expression of the Lipid Transfer Protein CERT Increases Therapeutic Protein Productivity of Mammalian Cells. J. Biotechnol. 2009, 141 (1–2), 84–90. 116. Gonzalez, R.; Andrews, B. A.; Asenjo, J. A. Metabolic Control Analysis of Monoclonal Antibody Synthesis. Biotechnol. Prog. 2001, 17 (2), 217–226. 117. Whiteley, E. M.; Hsu, T. A.; Betenbaugh, M. J. Modeling Assembly, Aggregation, and Chaperoning of Immunoglobulin G Production in Insect Cells. Biotechnol. Bioeng. 1997, 56 (1), 106–116. 118. Bibila, T. A.; Flickinger, M. C. A Model of Interorganelle Monoclonal Antibody Transport and Secretion in Mouse Hybridoma Cells. Biotechnol. Bioeng. 1991, 38 (7), 767–780. 119. Bibila, T. A.; Flickinger, M. C. Use of a Structured Kinetic Model of Antibody Synthesis and Secretion for Optimization of Antibody Production Systems: II. Transient Analysis. Biotechnol. Bioeng. 1992, 39 (3), 262–272. 120. Bibila, T. A.; Flickinger, M. C. Use of a Structured Kinetic Model of Antibody Synthesis and Secretion for Optimization of Antibody Production Systems: I. Steady-state Analysis. Biotechnol. Bioeng. 1992, 39 (3), 251–261. 121. O’Callaghan, P. M.; et al. Cell Line-specific Control of Recombinant Monoclonal Antibody Production by CHO Cells. Biotechnol. Bioeng. 2010, 106 (6), 938–951.

1.30

Glycomicsq

EFJ Cosgrave and JJ Kattla, National Institute for Bioprocessing Research and Training, Dublin, Ireland MP Campbell, Macquarie University, Sydney, NSW, Australia WB Struwe and MR Wormald, University of Oxford, Oxford, United Kingdom PM Rudd, National Institute for Bioprocessing Research and Training, Dublin, Ireland © 2017 Elsevier B.V. All rights reserved. This is a reprint of E.F.J. Cosgrave, J.J. Kattla, M.P. Campbell, W.B. Struwe, M.R. Wormald, P.M. Rudd, Glycomics, Reference Module in Life Sciences, Elsevier, 2017.

1.30.1 1.30.2 1.30.2.1

Introduction Methods for the Structural Analysis of Released Glycans U/HPLC-Based High-Throughput Analysis: This Technology Can be Used Alone or Coupled to MS and Exoglycosidase Arrays Mass Spectrometry Bioinformatics and Glycomics Glycomics in Bioproduction Glycoprotein Therapeutics Hormones: FSH Growth Factors: EPO Enzyme Replacement Therapies for Lysosomal Storage Diseases Recombinant a-galactosidase A for the treatment of Fabry disease Production of b-glucocerebrosidase for the treatment of Gaucher’s disease Cytokines: IFN-g Antibody-Derived Therapeutics Engineering Glycosylation for Improved Glycoprotein Function Sialylation in the Context of PK Behavior and Biological Activity Core a(1,6)-Fucosylation Affects the Biological Activity of IgG The Bisecting b(1,4)-N-Acetylglucosamine Influences ADCC Eukaryotic Expression Systems for Recombinant Glycoprotein Production Transgenic Plants Modified for “Humanized” N-Glycosylation Baculovirus-Mediated Insect Cell Expression Systems Restructuring of Yeast N-Linked Glycan Biosynthesis CHO: The Workhorse in Bioproduction Murine-Derived N-Linked Glycosylation The Changing Landscape of Regulatory Agencies Toward Glycosylation of Biopharmaceuticals Summary

1.30.2.2 1.30.2.3 1.30.3 1.30.4 1.30.4.1 1.30.4.2 1.30.4.3 1.30.4.3.1 1.30.4.3.2 1.30.4.4 1.30.4.5 1.30.5 1.30.6 1.30.7 1.30.8 1.30.9 1.30.10 1.30.11 1.30.12 1.30.13 1.30.14 1.30.15 1.30.16 References Relevant Websites

414 415 415 417 417 420 421 422 422 422 423 423 423 423 424 425 425 425 426 426 427 427 428 429 429 430 430 434

Glossary Biotherapeutic A large and typically complex biomolecule, often protein based, used for the treatment or management of disease. Glycoform An isoform of a protein defined by the specific glycan or collection of glycans attached, where the same protein with different glycans represents different glycoforms of the protein. Glycomics The study of how a collection of glycans contribute to or are associated with a particular biological event. Glycosylation The nontemplate-driven posttranslational modification of proteins whereby carbohydrate moieties are attached to either asparagine (N-linked) or serine/threonine (O-linked) residues in the peptide backbone.

q

Change History: August 2016. M.P. Campbell and P.M. Rudd updated the article to describe significant improvements that have occurred in analytical technologies including advances in ion mobility and the shift toward U/HPLC and their impact on high-throughput quantitative glycosylation analysis; updated Sections “Methods for the Structural Analysis of Released Glycans,” “U/HPLC-Based High-Throughput Analysis: This technology Can be Used Alone or Coupled to MS and Exoglycosidase Arrays,” “Mass Spectrometry,” and “Bioinformatics and Glycomics;” and extended Reference list for each modified section.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-12-809633-8.09031-2

413

414

1.30.1

Glycomics

Introduction

The discipline of glycomics is one within the large field of systems biology that aims to understand glycan structures and functions. Functional glycomics aims to tackle three general questions: (1) how glycans function in cellular communication; (2) what is the basis for specificity between proteins and glycans; and (3) how glycan diversity and microheterogeneity result as a function of biology in development and disease. In contrast to DNA, RNA, and proteins, carbohydrates form branching structures and, as a result, a relatively small set of monosaccharides can form considerably complex structures. Furthermore, the synthesis of glycans is a complex nontemplate-driven process that employs over 200 enzymes in a precise nonuniform steady-state distribution in the Golgi [1,2]. The biological significance of glycoconjugates is not yet fully understood, but the consequence of defective glycosylation is severe, and complete loss of glycosylation is lethal in all metazoans [3,4]. The most common types of protein glycosylation are large branched glycans covalently linked to asparagine residues in an N-X-S/T sequon (N-glycans), smaller linear and/or biantennary glycans attached to serine or threonine residues (O-glycans), and linear-sulfated glycans bound to serine residues (glycosaminoglycans (GAGs)). All N-glycans contain a conserved GlcNAc2Man3 core structure and are characterized as paucimannose, high mannose, hybrid, or complex, depending on further processing/ additions in the Golgi. Paucimannose structures contain 3–5 mannose residues attached to the chitobiose core (GlcNAc2) and are found in lower species. High-mannose structures contain between 5 and 9 mannose residues attached to the chitobiose core. Complex structures have all but three mannose residues removed and are modified with numerous residues and various degrees of branching. Hybrid structures have an antenna with unsubstituted mannose residues and branching typical of complex type structures. O-glycans have eight core structures and express considerable structural diversity, although the extent of branching and overall size is less than that of N-glycans. GAGs are long linear-sulfated structures that typically have disaccharide repeats. Hyaluronic acid, chondroitin sulfate, and heparin are all GAGs. The structural diversity of N- and O-glycans is determined through a complex biosynthetic pathway that begins on the surface of the endoplasmic reticulum (ER) or in the Golgi, respectively. The early stages of N-glycan biosynthesis is conserved among Saccharomyces cerevisiae, Caenorhabditis elegans, and vertebrates [5,6]. The biochemical pathway for N-linked glycan synthesis occurs in four distinct stages: (1) formation of a lipid-linked precursor oligosaccharide; (2) en bloc transfer of the oligosaccharide to a nascent polypeptide; (3) trimming of oligosaccharides in the ER and Golgi; and (4) addition of new sugars in the medial and trans-Golgi. All eukaryotes share the first three steps of N-glycosylation, but the greater glycan complexity seen in higher species is due to extensive processing that occurs in the Golgi. As the nascent glycoprotein begins transport from the ER, glucosidase I removes the outermost glucose residue. Glucosidase II then removes the second proximal glucose leaving a glycoprotein with a single terminal glucose. Subsequently, two quality-control lectins, calnexin and calreticulin, ensure proper folding by binding to the monoglucosylated glycopeptides. Glucosidase II acts again by cleaving the third glucose and the calnexin/calreticulin complex dissociates from the glycoprotein (Fig. 1). At this point, the glycoprotein can follow two distinct routes depending on proper protein folding. If the calnexin/calreticulin complex determines the glycoprotein to be properly folded, it will continue toward the Golgi in the intermediate compartment for further processing alone or with the aid of an additional ER lectin, ERGIC-53. If the glycoprotein is partially unfolded, it is retained in the ER and reglucosylated by a luminal glucosyltransferase and will once again interact with calnexin/calreticulin and ERp57 to achieve proper folding. If the protein is folded correctly, the ER (1,2)-mannosidase-I cleaves a specific terminal mannose residue from the Man9GlcNAc2 oligosaccharide to generate Man8GlcNAc2 before the glycoprotein exits the ER. However, if correct folding is not achieved, the glycoprotein will trigger the ER-associated protein degradation pathway and will be translocated to the cytoplasm and digested in the proteasome [9]. In addition to the inherent complexity of glycoconjugates, a particular oligosaccharide of a glycoprotein may occur in forms that differ in one or more of its oligosaccharide residues, a trend known as microheterogeneity. Many factors influence N-glycan

Figure 1 Calnexin (CNX) and calreticulin (CRT) function as molecular chaperones that assist the folding and assembly of glycoproteins that pass through the endoplasmic reticulum (ER). Removal of the outermost glucose residues by glucosidase I and glucosidase II exposes the Glc1Man9GlcNAc2 epitope. This structure is then recognized by CNX and CRT followed by the removal of the innermost glucose by glucosidase II, thereby releasing the glycoprotein from the lectin anchor. If the protein forms a non-native/disordered three-dimensional structure the N-glycan is glucosylated by the soluble ER enzyme glucosyltransferase, as this enzyme behaves as a conformational sensor. Adapted from Ref. [7,8].

Glycomics

415

microheterogeneity such as transport rates from the ER to the Golgi, the duration that glycoconjugates are in the Golgi, sugar nucleotide metabolism, and localization of glycosyltransferases in the Golgi [10]. By and large, the diversity of glycan structures is due to over 200 different glycosyltransferases located in the Golgi [1]. On the other hand, the precise localization of only a small number of these glycosyltransferase enzymes within the Golgi has been determined through subcellular fractionation and immunolabeling of thin sections followed by electron microscopy [11]. O-glycan biosynthesis occurs exclusively in the Golgi and unlike N-linked glycans, where a common lipid-linked precursor core is attached en bloc to proteins that contain the Asn-X-Ser/Thr sequon, O-glycans have a series of eight common core structures that can be linked to serine or threonine amino acid residues. The initiating step of O-linked assembly is catalyzed by the covalent linkage of an N-acetylgalactosamine by polypeptide GalNAc transferase. Through a stepwise process, the GalNAc residue on serine or threonine can be modified by a series of enzymes to generate the eight core structures. These include core 1 (b(1,3)-galactosyltransferase (core 1 GalT)), core 2 (b(1,6)-N-acetylglucosaminyltransferase (C2GnT)), core 3 (b(1,3)-Nacetylglucosaminyltransferase (C3GnT)), and core 4 (b(1,6)-N-acetylglucosaminyltransferase (C4GnT)).

1.30.2

Methods for the Structural Analysis of Released Glycans

Glycoprotein analysis can be approached at several levels: the analysis of the intact glycoprotein (to obtain information about glycoforms), the analysis of glycopeptides (to determine the glycans at individual sites), the profiling of released glycans (to give a snapshot of the complexity of the pool) and detailed glycan analysis (to obtain monosaccharide sequence and linkage information). Several techniques are employed in the structural elucidation of glycans, which can either be used alone or coupled. The most commonly used methods are ultra/high-performance liquid chromatography (U/HPLC), mass spectrometry (MS), nuclear magnetic resonance spectroscopy (NMR), capillary electrophoresis (CE), gas chromatography, lectin microarrays and exoglycosidase arrays, to name a few. There are advantages (and limitations) to each technique; therefore, most approaches for detailed structural assignments incorporate a combination of any of the above methods to provide orthogonal confirmation. This is due to the complex nature of glycans, where no single approach can fully characterize a given glycan structure and/or function (with perhaps NMR being the exception; however, it suffers from low sensitivity, necessitating a large amount of material for characterization). Analysis aims to answer the overall glycan topology, monomer identification, anomericity, and detection of isomers/isobars in a given sample. Here, we present two of the most popular techniques used to investigate the structure of glycans.

1.30.2.1 Arrays

U/HPLC-Based High-Throughput Analysis: This Technology Can be Used Alone or Coupled to MS and Exoglycosidase

The ability to detect and characterize protein glycosylation has become paramount interest due, in part, to the association of aberrant glycosylation with a multitude of diseases, including cancer, autoimmune disorders, and infections [12–20]. The subsequent characterization of disease-associated glycans stemming from serum glycoproteins has resulted in a more in-depth understanding of disease pathologies, additionally providing the potential for early disease detection through identification of new candidate glyco-biomarkers demonstrating altered glycan profiles. However, analysis of the serum glycome relies on the ability to detect glycans in small quantities due to the low abundance of the majority of serum glycoproteins. The lack of sensitivity for several glycan analytical approaches has, therefore, been a bottleneck for more extensive investigations of glycosylation in disease. The development of a sensitive, quantifiable, robust, and reproducible method for the detection and separation of femtomolar quantities of glycans is, therefore, of principal interest in the field of glycomics. Several techniques are currently available for glycan analysis including glycan arrays, MS, CE, NMR, and U/HPLC. The use of U/HPLC-based methods for glycan detection and characterization has emerged as a powerful method for glyco-profiling, with this particular method progressing rapidly due to the development of high-throughput glycan-release platforms [21]. The strength in this method is based on a number of significant advantages that accelerate the process of glycan analysis. First, the method of amide-based normal phase-U/HPLC (NP-U/HPLC) glycan separation is a well-established technique capable of providing highly reproducible profiles [22–24]. Second, both charged and neutral glycans can be analyzed in the same U/HPLC run. In the case of MS or CE, analysis often occurs only after sialic acids have been enzymatically removed. Third, the use of labels such as the fluorophores 2-aminobenzamide (2AB), 2-anthranilic acid (2AA), procainamide or Waters recently developed Rapifluor generates a 1:1 stoichiometry between glycan and label when coupled, permitting the quantitation of individual glycans within a glycan pool. In some cases, the label significantly increases the sensitivity of the MS, improving the quality of data from coupled LC/MS [25]. Fourth, the use of a glycan standard such as dextran presents an opportunity to calibrate glycan retention times occurring in an U/HPLC run into standardized values, often referred to as glucose unit (GU) values. This enables the normalization of all glycans identified by NP-U/HPLC, therein eliminating variation between U/HPLC equipment and providing a platform for comparative analysis and curation of data based on a standard reporting system. Fifth, exoglycosidase-assisted digestion of glycan pools allows for the characterization of complex carbohydrates, based on the knowledge that specific monosaccharides demonstrate GU shifts of known and predictable GU increments when compared in U/HPLC profiles, facilitating the identification of glycan structure. Moreover, many of the known exoglycosidases are specific for certain linkages, therein allowing for the determination of linkage information in complex carbohydrate structures (Fig. 2). This is an important aspect as several alternative techniques are not capable of providing both compositional and structural information. Finally, and perhaps of particular significance for industry-related glycan analysis, NP-U/HPLC does not require skilled personnel, making it ideal for general use in either academic or industrial settings (Fig. 3).

416

Glycomics

Symbol

Monosaccharide

Abbreviation

N-acetylglucosamine

GlcNAc

Linkage type Known α -linkage

Glucose

Glc

Known β -linkage

Galactose

Gal

Unknown α -linkage

N-acetylgalactosamine

GalNAc

Unknown β -linkage

Fucose

Fuc

Mannose

Man

N-acetylneuraminic acid

Neu5Ac (NANA)

Linkage position

N-glycolylneuraminic acid

Neu5Gc (NGNA)

Xylose

Xyl

8 6 4

Unknown pentose

3 2

Unknown hexose

Figure 2 The Oxford nomenclature for identifying and reporting glycan structural information, which depicts the monosaccharide linkages with embedded specificity and anomericity. This format is widely used in the community and adopted in this article, in part, due to its clear representation of linkage information. Recently, in conjunction with the third edition of Essentials in Glycobiology, the Essentials/Consortium for Functional Glycomics graphical notation has been extended and now supports the Oxford linkage option. This updated format is available for review at http://www.ncbi. nlm.nih.gov/books/NBK310273/ and Ref. [26]. Trans-golgi network

Transgolgi

ST6Gal-I

β (1,4) GalT-I Lewis A antigen

FucT-VIII Medialgolgi

Plant GlcNAcT-II

β (1,2) Xylose α(1,3) Fucose

α-Man II

Insect

Cisgolgi

GlcNAcT-I

α (1 ,2 IA )-M an IB

IC

Paucimannose

ER an

M

ER

I

α-Glu II α-Glu I

Yeast

High mannose

: Protein

Figure 3 Comparative assessment of N-glycan biosynthesis in eukaryotic systems. Normal human N-linked glycosylation occurs in the endoplasmic reticulum (ER) and Golgi apparatus, resulting in the production of fully processed complex oligosaccharide structures. While other eukaryotic systems perform N-linked glycosylation, they differ in the extent of processing and this has consequential effects. Yeast diverge early in the biosynthetic process and result in typically high mannose structures. Insects are capable of trimming high mannose but are unable to extend oligosaccharides beyond paucimannosidic structures. Plants are capable of capping oligosaccharides with N-acetylglucosamine, but plant glycosyltransferases produce structures that are highly allergenic in humans. Rodent cell lines such as CHO or NS0 are popular expression hosts for recombinant protein expression in part because of the conservation of N-linked glycosylation. Here, the only difference observed is the sialic linkage where humans cap with a(2,6) and CHO caps with a(2,3)-linked sialic acid.

Glycomics

417

HPLC techniques include the option of sialic acid quantitation and speciation, facilitated by the use of weak anion-exchange HPLC (WAX-HPLC) and 1,2-diamino-4,5-methylene-dioxybenzene$2HCl (DMB) labeling, respectively. In sialic acid speciation, terminal sialic acid residues are typically released from glycoproteins by acid hydrolysis and the resulting isolated sialic acid-based saccharides are labeled with DMB. Similar to 2AB, DMB couples with saccharides in a 1:1 stoichiometry, allowing for the relative quantitation of individual sialic acid species. Labeled glycans are separated by reversed phase-U/HPLC (RP-U/HPLC) using a C18 column, where sialic acid orthologs such as Neu5Gc (N-glycolylneuraminic acid; NGNA), Neu5,7Ac2, and Neu5,Gc9Ac are identified and quantified. Separation of glycans by WAX-U/HPLC assists in identifying the number of sialic acid moieties present on a given glycoprotein. Used in conjunction with sialidase digestions, the number of terminal sialic acids on a given glycan can be verified and whether the negative charge is provided by N-acetylneuraminic acid (NANA) or alternative residue modifications/ substitutions (eg, sulfation). Sialic acid content is of significant importance to the bio-industry as sialylation can greatly impact the safety and efficacy of a glycoprotein therapeutic. Moreover, inappropriate sialylation, such as incorporation of NGNA, can lead to premature clearance of the therapeutic, loss of efficacy and cause untoward reactions in some patients [27–29]. Therefore, the characterization of sialic acids represents a critical feature for potential therapeutics. From a bioproduction perspective, quantitation of individual sialic acid species provides essential information including whether the terminally charged glycan moieties are in fact NANA (Neu5Ac) or potentially NGNA, with the latter an obvious concern in terms of safety and efficacy.

1.30.2.2

Mass Spectrometry

MS is a key tool in defining glycoconjugate structure, which is generally applied to released glycan samples. Numerous MS-based methods incorporate LC or CE separation prior to MS, as well as glycan derivatization (commonly permethylation/peracetylation), exoglycosidase digestions similar to HPLC analyses, and isotopic labelling. Two principle ionization techniques, matrix-assisted laser desorption/ionization (MALDI) and electrospray ionization (ESI), are used, coupled to a variety of mass analyzers. MALDI–time of flight (TOF) MS is commonly used for the analysis of permethylated N- and O-glycans because of its low sample expenditure, capacity to produce singly charged ions, detection in the high mass range, tolerance of salt contaminants, and speed of sample preparation and analysis [30,31]. MALDI MS does have limitations, specifically in the degree of information it provides, whereby a single MS spectrum only defines the overall composition of a glycan. Tandem MS (MS/MS) and sequential MS (MSn) with low-energy collision-induced dissociation (CID) provides structural data relating to branching, linkage, reducing end modifications, as well as the detection of isomers and/or isobars. For any MS approach, sample preparation is crucial. Instruments are increasingly sensitive, permitting detection at remarkably low levels. A typical sample preparation workflow involves the removal of the peptide component enzymatically (PNGase F or Endo H) or chemically (hydrazinolysis, b-elimination). Further purification may include desalting via cation-exchange, normal phase and/or porous-graphitized carbon solid-phase extraction. Purified glycoconjugates may be analyzed without further derivatization; yet glycan composition greatly influences the type of ion formed during both MALDI and ESI ionization as well as the polarity used (positive or negative analysis). Neutral glycans (eg, high-mannose type) preferentially form metal adducts, generally sodium or potassium as positive ions and chloride or phosphate as negative ions, as these salts are present throughout sample preparation. Sialic acid-containing glycans form deprotonated ions from proton loss on the carboxylic acid, and sulfated/phosphorylated structures are commonly observed as deprotonated ions. Sialic acids can be problematic, particularly in MALDI analyses, due to their susceptibility to fragmentation during ionization. To revolve this, sialic acid-containing glycans can be derivatized to methylesters and analyzed in positive mode [32]. Likewise, released glycans can be permethylated, whereby the hydroxyl groups on each residue are modified to methyl ethers without altering the structure of the glycan [10]. Additionally, permethylation stabilizes sialic acid residues and increases the ionization efficiency compared to native analysis [33]. Sequencing of permethylated glycans by sequential MS in an ion-trap mass analyzer provides detailed structural information that can be overlooked by MS/MS and/or U/HPLC techniques or where exoglycosidases are ineffective [34]. However, analysis of negative glycan ions is advantageous in that collision-induced dissociation (CID) yields abundant cross-ring fragments that are more information-rich than similar MS/MS approaches in positive mode where most fragmentation occurs between residues at glycosidic bonds (ie, neutral loss) [35–37]. However, some structural features remain difficult to resolve from MS/MS of negative ions, namely complex core structures (extensive fucosylation), outer arm modifications (Lewis/blood group epitope isomers), and linkage isomers. However, the development and commercial availability of ion mobility (IM) MS instruments has offered a new tool for glycomics analysis that can separate isomeric glycans [38–42]. IM separates ions based on the size, shape and charge of gas-phase glycan structures; however, our understanding of how glycan structures are formed in the gas-phase is incomplete but improving instrumentation especially resolution will further our understanding. MS is a fundamental tool in glycomics, but the need for high-throughput, comprehensive methods remains critical. Glycan MS is continually evolving with technical advances in separation like IM, increased speed of analyses, miniaturization and automation [25,43]. A comprehensive MS analysis of glycans employs several established techniques, using different ion sources, mass analyzers, and sample derivatization procedures. There is no single bona fide technique to understanding and dissecting the complex nature of glycans by MS, but as more technologies are developed and software tools expand, the potential of MS will be increasingly realized.

1.30.2.3

Bioinformatics and Glycomics

In contrast to genomics and proteomics, glycosciences lacks a centralized, well-curated, validated structural and experimental database that collates the structure, biological origin, and potential function of glycans that have been reported in the literature. The lack of such a database can be attributed to the inherent structural complexity of carbohydrates, the difficulty in determining

418

Glycomics

their structures, and the nontemplate-driven biosynthesis process. The development and advancement of sensitive and high-throughput analytical methods and technologies for structure determination of carbohydrate structures including U/HPLC and mass spectrometric glycan profiling, array- and chip-based technologies, have resulted in large data collections. The growth in glycomics data over recent years now requires supporting bioinformatics tools and databases to (semi-) automate data analysis, data management, and data sharing. The development of bioinformatics resources has increased in recent years; however, the field is defined by diverse and incompatible experimental data, databases, and applications that implement different encoding standards, all of which act to hinder cross-referencing interoperability. The lack of standardization and data-sharing mechanisms has hampered the development of comprehensive bioinformatics tools for the interpretation of experimental data and large-scale glycomics studies. There is general agreement that the availability of software tools to handle and process data and access to robust carbohydrate databases will facilitate the development of glycomics analytical technologies, experiments to decipher the functions of glycans, and integration with other life sciences. The need to develop coordinated resources has been raised by international investigators and stated in a National Institutes of Health (NIH) white paper [44] and re-emphasized at the CFG/NIGMS Sponsored Workshop “Analytic and Bioinformatic Glycomics” (April 16–18, 2009, NIH, Bethesda, MD). The successful development and adoption of glyco-related databases depends on provision of interactive and well-curated data and services that can be integrated with other glycoinformatics resources and shared across a common framework. Achieving this objective requires the implementation and adoption of common data-sharing methods and standards for exchanging annotated data between international consortia and local research groups. The requirement to define standards and workflows for data exchange led to the formation of a “Working Group on Glycomics Database Standards” (WGGDS) at the Consortium for Functional Glycomics (CFG) workshop on Analytical and Bioinformatic Glycomics in April 2009. The working group brings together partners of the CFG [45], GlycomeDB [46], RINGS, [47] and EUROCarbDB [48] consortia with a focused goal of developing web services and controlled standards to enable data sharing between international glycobiology groups. This effort has continued with the GlycoRDF initiative (see below) established in 2013 at the 22nd International Symposium on Glycoconjugates (Dalian, China) [49]. The necessity for a centralized database of all carbohydrate structures published in refereed scientific journals was recognized in the mid-1980s. This led to international efforts to ensure high-quality curation of structure data and the creation of the Complex Carbohydrate Structure Database (CCSD) (commonly referred to as CarbBank). The CCSD was supported by the NIH and developed and maintained by the Complex Carbohydrate Research Center, University of Georgia. During the second half of the 1990s, funding stopped, and CarbBank became a static entity. However, with over 49,000 entries, corresponding to over 23,000 distinct glycan structures, the CCSD is the largest publicly available repository; recent open access glyco-databases have used CarbBank to underpin database designs and seed structure entries. Many large-scale international initiatives have been established since CarbBank such as EuroCarbDB, CFG, Complex Carbohydrate Research Center, Human Disease Glycomics/Proteome Initiative, and Kyoto Encyclopedia of Genes and Genomes (KEGG-Glycan) [50]. These collaborative efforts are developing novel resources including databases and bioinformatic platforms to integrate and share the growing data generated by this field, addressing the challenges of glycomics and the rapid evolution of analytical technologies. This effort was led, in part, by the formation of a group at the DKFZ, Heidelberg (German Cancer Research Centre), who realized the potential of carbohydrates and the need to continue the work of the CCSD. The work and ambition of the DKFZ lab under the guidance of Claus-Wilhelm von der Lieth led to the glycosciences.de [51] portal and subsequently the formation of the EUROCarbDB design studies. The CFG has been an important feature in the advancement of glycobioinformatics, addressing the need for informatics to handle and annotate vast amounts of experimental data generated by glycomics research. Moreover, the extension of KEGG to integrate glycosylation pathways with glycan structure databases has been pivotal in connecting structural evidence with enzyme information. CAZy is a database of Carbohydrate-Active enZYmes (CAZymes) that provide rich information on the proteins involved in the synthesis/degradation of glycans or in their recognition including glycoside hydrolases, glycosyltransferases, polysaccharide lyases, carbohydrate esterases and carbohydrate-binding families [52]. Several other database initiatives have followed, notably: (1) the curated Carbohydrate Structure Database platform for bacterial, archaeal, plant and fungal carbohydrates [53]; (2) GlycomeDB, an initiative to consolidate structures from a number of established glycan structure databases; (3) The Japan Consortium for Glycobiology and Glycotechnology DataBase (JCGGDB); and (4) UniCarbKB that emerged from the transformation of GlycoSuiteDB [54,55]. Most recently, an international repository of uncurated glycan structures was launched (GlyTouCan) that assigns unique accession numbers to any glycan, providing a foundation for effective data sharing beyond databases [56]. The nonlinear and complex nature of carbohydrate sequences complicates the representation of glycan structures. The notation issues are further complicated by the implementation of different formats for storing annotations, such as taxonomic and tissue information. Recent efforts to establish a common encoding format for storing carbohydrate sequence data have led to GlycoCT, Glyde and WURCS [57–59]. For example, GlycoCT provides translational tools that encompass all structure-encoding capabilities into a unified single sequence, providing the basis for data-sharing protocols and cross-database referencing querying. A number of tools are available for displaying graphical representations of glycan structures. These tools include GlycanBuilder [60,61], which can be used to build and visualize glycan structures from a predefined set of monosaccharide building blocks that can be inter-converted into different graphical formats, both the CFG color and black/white formats, both color and black/white University of Oxford formats, and a descriptive graphical text format. The ability to inter-convert between different notation formats enhances user accessibility and ensures that the development of centralized resources encompasses the major features.

Glycomics

419

110000

120000 130000 140000 150000

100000

90000

80000

70000

60000

40000

5.00

50000

EU

20000

30000

A major factor restricting progress in the glycomics/glycoproteomics field is the lack of (semi-) automated tools capable of analyzing and retrieving structural information from U/HPLC and MS data. Previous efforts by a European-funded design study (EUROCarbDB) established a framework for such a system that includes formats and standards for data exchange, tools to assist the interpretation of experimental data, workflows, and database infrastructures for storing and retrieving glycan data. The set of standards were subsequently adopted by the WGGDS to facilitate the extension of data-sharing between international databases. The development of tools for data annotation of U/HPLC profile and MS spectra data is an active area of research and is vital for high-throughput glycomics investigations. For example, tools and databases have been created to assist in high-throughput U/HPLC data interpretation, including GlycoBase and autoGU [62]. GlycoBase is a novel database solution containing the U/HPLC elution positions for 2AB-labeled N- and O-linked glycan structures, the predicted products of exoglycosidase digestions and supporting literature information. The interpretation of U/HPLC data including exoglycosidase digestions can be time consuming, and database-matching software (autoGU) is available to assist in the assignment of possible glycan structures to each U/HPLC peak. When used in combination with data from a series of exoglycosidases, autoGU will create a refined list of structures based on the digest footprint, that is, shifts in GU values due to cleavage of terminal monosaccharides dependent on enzyme specificity (Fig. 4). GlycoDigest is a tool that simulates exoglycosidase digestion, based on controlled rules acquired from expert knowledge

EU

20.00

EU

0.00 40.00

0.00 60.00 40.00 20.00 0.00

8.9 ABS 6.7 7.5 2x BTG

EU

5.9 1x

50.00

2x

0.00 EU

0.00 60.00 EU

5.9–5.5 = 0.4 0.4 = 1 × Fucose

1x

10.00

4.4

40.00

GUH 5.5–4.4 = 1.1 1.1 = 2 × GlcNAc

2x

20.00 0.00 60.00

JBM

2.6

40.00

6.7–5.9 = 0.8 0.8 = 1 × Galactose 7.5–5.9 = 1.6 1.6 = 2 × Galactose

BKF

5.5

EU

8.9–7.5 = 1.4 1.4 = 2 × Sialic acid

4.4–2.6 = 1.8 1.8 = 2 × Mannose

2x

20.00 0.00 30

40

50

60

70 80 90 100 Retention time (min)

110

120

130

140

Figure 4 Structural assignment of murine IgG1 using exoglycosidase digestions with normal-phase high-performance liquid chromatography (NPU/HPLC) analysis. Series of exoglycosidases (ABS, BTG, BKF, GUH, and JBM) are capable of sequentially cleaving complex oligosaccharides back to a common core structure from which all glycoforms in a pool are derived. The specificity of exoglycosidases with example are provided at https:// glycobase.nibrt.ie/glycobase/view_enzymes.action and https://github.com/alternativeTime/unicarb_static/wiki/GlycoDigest-Guide-(UniCarbKB). In this case, a dextran ladder is used to generate glucose unit (GU) values based on the chromatographic retention time, eliminating instrument variability. By correlating the enzyme specificity with the observed shift in GU value (due to the digestion), more complex structures can be sequentially reconstructed to identify all the structures in the glycan pool. In this example, a full panel of exoglycosidases is able to cleave all the glycan structures in the pool to the common Man1 structure as seen in the bottom chromatogram (GU value 2.6). Based on the enzyme used and the shift in retention time, the oligosaccharides can be reassembled based on the knowledge that individual monosaccharides have characteristic GU shifts in NP-U/HPLC. For example, mannose residues typically demonstrate a GU value of 0.9. In the above example, the bottom chromatogram shows a shift in the main peak from 4.4 (in the chromatogram immediately above) to 2.6 due to the addition of an a-mannosidase (JBM), which correlates to the loss of two mannose residues. Knowing the structure of the 2.6 peak, we can add two mannose residues to solve the structure of the peak at 4.4. By applying this logic to all the chromatograms, a stepwise restructuring process occurs whereby all structures can be identified in the total glycan pool.

420

Glycomics

and experimental evidence available in GlycoBase, which allows the targeted design of glycosidase enzyme mixtures by allowing researchers to model the action of exoglycosidases [63]. The latest release of GlycoBase (version 3.2) now provides data from RP-UPLC and CE for N- and O-linked glycans and milk sugars obtained from a variety of species, tissues and important biotherapeutic products with comprehensive information on the experimental procedures used to derive the glycan data. These applications have been developed to meet the requirements of analytical methodologies described above. The database is publicly available and was developed in partnership with the EUROCarbDB framework to ensure compatibility with proposed database standards and data-exchange protocols. A number of software solution and databases are available that assist the interpretation of MS data. One of the earliest tools developed was GlycoMod that can be used to putatively assign glycan compositions from experimental mass values of either free or derivatized glycans and/or glycopeptides. A widely used tool for the analysis of MS and MSn data is GlycoWorkBench, which matches a theoretical list of fragment masses against experimental peak list derived from the mass spectrum by using a comprehensive library of fragmentation rules and annotation options [64]. An emerging database for the deposition of annotated and curated MS data is UniCarb-DB that provides tools for spectral matching to identify unknown structures [65]. In addition, RINGS (Resource for INformatics of Glycomes at Soka) provides a number of web-based software, for example, Glycan Miner Tool to analyze glycan fragments from glycan profiling (mass spectrometry) data. As part of the Glycome Informatics Consortium (GLIC) a list of maintained databases/tool is provided at https://github.com/glycoinfo/glic/blob/gh-pages/database/index.md. Extensions to existing databases are being developed in parallel with new analytical strategies, which meet the growing demand and requirements including descriptions of glycan structures sequenced from a range of biotherapeutics. For example, the launch of an ion mobility mass spectrometry (IM-MS) database (GlycoMob), within the UniCarb framework, provides access to over 900 collision cross section values of biologically derived and synthetic N-glycans and their fragments [66]. The in-house availability of glycan structural information enables the implementation of novel high-throughput interpretation workflows to screen and fingerprint the glycosylation profiles of therapeutic antibodies and batch-to-batch variability. The combination of information from several databases is necessary to better understand the structure–function roles of carbohydrates. Developing this understanding requires access to the primary structures and associated data such as biological contexts, specification of glycan-binding proteins, supporting experimental data and publication details. However, it is widely accepted that the lack of standardization is the largest obstacle for automatic data exchange, hindering the development of bioinformatic tools for sharing information with applications in data mining and systems biology. It is clear that there is a need to develop solutions for storing structural data collections, analyzing experimental data, and providing interfaces for cross-referencing data collections. The glycoinformatics community have embraced Semantic Web technologies and created a Resource Description Framework (RDF) representation for glycomics data. This standard, defined by the GlycoRDF ontology, provides a solution for integration of these diverse data, ultimately allowing researchers to answer more complex questions than could be addressed with individual database queries [49,67]. The current coordinated trend toward RDF-based data integration is already shaping future developments of glycoscience databases and will help bridge the gap between glycomics and other -omics that have already adopted RDF ontologies. For this reason, the rapidly expanding glycoscience field is being increasingly recognized as an important component of life science research, as highlighted in a recent report commissioned by the European Science Foundation (ESF), which stated in its policy briefing that “to further develop diagnostic tools, preventive medicines (vaccines) and therapeutic drugs, a better understanding of glycosylated molecules is required.”

1.30.3

Glycomics in Bioproduction

Advances in the biopharmaceutical industry have led to a new generation of therapeutics, capable of providing novel strategies in disease detection and treatment. Principally, the industry has experienced a shift from small-molecule to large-molecule therapeutics, with the latter predominantly represented by protein-based structures. Isolation of therapeutic proteins initially relied on the ability to purify individual target proteins from their natural environment, such as insulin, raising issues with possible contamination during purification, and the inherent risks to the patient. However, with the advent of recombinant DNA technology, the same targets can now be expressed as heterologous proteins in specific cell lines engineered for maximal protein expression in defined media, eliminating numerous risk factors such as viral or prion contamination. Optimization of this process has subsequently led to the production of therapies used for treatment of numerous diseases, with perhaps the biggest impact in the treatment of various forms of cancer and autoimmune disease. As our knowledge of disease develops, the biopharmaceutical industry is ideally poised to translate this knowledge to therapeutic design, with the intent of developing therapies that help treat either new diseases with no current solutions or old diseases with novel approaches. Critical to the shift from small-molecule to large-molecule therapies has been the development of production systems that are capable of performing posttranslational modifications, as many new therapeutics require modifications such as glycosylation to improve their activity, solubility, pharmacokinetic (PK) behavior, potency, and stability. The pharmaceutical trend is toward increased functional specificity in treatment; thus, the ability of a given production system to achieve the various posttranslational modifications necessary for the safety and efficacy of the proposed therapeutic remains the caveat. In other words, the complexity of the therapy is limited by the ability of the expression system to generate the required biological features. There are 64 currently approved glycoprotein therapeutics and a further 500 more protein-based therapeutics, anticipated to reach the market, that are in preclinical or clinical development [68]. Nearly 70% of these recombinant proteins are

Glycomics

421

expected to be glycoproteins, emphasizing the necessity for the integration of glycomics and glycan analytical techniques into the current bioproduction model. The role of glycosylation in therapeutics is widely appreciated to modulate therapeutic safety and efficacy [69–71]. Appropriately, glycosylation of protein-based therapeutics is now the subject of intense interest as manipulation of the glycan biosynthetic pathway in well-established cell lines potentially yields therapeutics with improved PK behavior and biological activity. Manipulation of glycosylation machinery typically targets N-linked glycans. O-linked glycosylation continues to be a rapidly advancing area; however, techniques and methods for O-linked glycan analysis are not as well established as those for N-linked analysis. Accordingly, N-linked glycan modification has been the focus of bioproduction and the discussion of glycoprotein therapeutics will therefore focus on the role of N-linked glycosylation in protein function. Although the N-linked glycan biosynthetic pathway exhibits a certain degree of conservation among eukaryotes, important differences are evident, which lead to the generation of structures that not only influence the efficacy and activity of a therapeutic but also potentially introduce foreign epitopes that subsequently induce immunogenic responses by the host (Fig. 4). Several glycan moieties are now described that trigger immune responses and consequently, regulatory agencies are demanding a more detailed characterization of glycan structures on candidate glycoprotein therapeutics seeking approval. Thus, significant efforts have been made to engineer current eukaryotic industrial mammalian expression systems to both reduce the incorporation of immunogenic epitopes and “humanize” the glycosylation pathways, typically through elimination of native enzymes responsible for generating immunogenic epitopes and complementation with heterologous glycosyltransferases, respectively. The ultimate intention of this strategy is the improvement of the PK behavior and biological activity of glycoprotein therapeutics. Critically, a full appreciation of the glycans carried by a given therapeutic is necessary in terms of both quality control and therapeutic potency [71].

1.30.4

Glycoprotein Therapeutics

Carlson reported that global 2012 revenue from biologics reached at least $125 billion while McKinsey and Company (New York) estimated that revenue from global biopharmaceuticals was $168 billion [72]. This figure has amassed from the contribution of several classes of protein therapeutics, including (1) monoclonal antibodies, such as anti-tumor necrosis factor-alpha (TNFa) and anti-VEGF therapeutics (vascular endothelial growth factor, VEGF); (2) blood factors, including recombinant factor VIII; (3) anticoagulants; (4) recombinant enzymes, such as b-glucocerebrosidase; (5) hormones, including follicular stimulating hormone (FSH); (6) cytokines, including interferon-a (IFN-a), IFN-b, and IFN-g; (7) vaccines; (8) growth factors, most notably erythropoietin (EPO); and (9) fusion proteins, such as Fc-TNFa receptor fusions (Fig. 5). Importantly, each class of therapeutic

Interferon-gamma (IFN-γγ )

Erythropoietin

Human chorionic gonadotropin (HCG)

Immunoglobulin G1 (IgG1)

Figure 5 Elucidated structures of glycoprotein therapeutics. Glycosylation of recombinant protein therapeutics plays an important role in the function of the underlying protein. In these examples, the N- and O-linked glycosylation provides key attributes to the protein such as improved solubility, proteolytic protection, and changes in the protein tertiary structure that modulate interaction with other proteins. Manipulation of the glycosylation through recombinant expression has the potential to affect the biological activity of these proteins, both positively and negatively. For each glycoprotein, the modeled glycan structures are shown in the ball and stick representation.

422

Glycomics

listed contains glycoprotein representatives, with these glycoproteins often contributing the majority of the sales for its given category. This emerging trend indicates the importance of appropriate glycosylation during therapeutic bioprocessing and manufacture. The distribution of therapeutic sales identifies growth factors and monoclonal antibodies (mAbs) as the major contributors both in terms of the number of approved therapies and annual sales. The following selection of glycoprotein therapeutics demonstrates the importance of glycosylation and how changes to the glycosylation process affect biological activity and PK behavior.

1.30.4.1

Hormones: FSH

Hormone therapies are largely represented by recombinant insulin, glucagon-like peptide-1 (GLP-1), and human growth factor, which together accounted for a total of $6 billion (USD) in sales, representing the fastest growing category of biologics in 2007 [73]. To a lesser degree, the hormone category is also resident to hormones involved in fertility such as human chorionic gonadotropin, luteinizing hormone, and FSH, which together account for over $600 million (USD) in sales per annum [73]. The fertility hormones demonstrate a certain degree of similarity, each being heterodimeric structures of an a and a b subunit. Although conservation is observed in the a subunit for each fertility hormone, variability exists in each hormone b subunit. Interestingly, the b subunit contains sites of glycosylation that influence the biological activity of the hormone. One such a/b hormone is FSH, a heterodimeric glycoprotein regulator of gonadal function in both men [74,75] and women [76]. The use of gonadotropins for the treatment of infertility in women has been in practice for over four decades, principally for their role in stimulating the growth of multiple oocytes during in vitro fertilization and improving the likelihood of a live birth [77]. Prior to the development of recombinant FSH expression, all human FSH was purified from postmenopausal urine that typically exhibited reduced activity due to contamination by other urine-related proteins [78]. FSH expressed as a recombinant protein in a controlled environment subsequently eliminated the associated immunogenicity of naturally purified FSH, therein improving the therapeutic potency and PK behavior. Glycosylation of FSH is observed on both the a and b subunit, each encoding two sites for N-linked glycosylation; at positions Asn52 and Asn78 for the a subunit and positions Asn7 and Asn24 for the b subunit [79]. Selective manipulation of individual N-linked glycosylation sites on FSH does not affect secretion; however, complete removal of all glycans prevents its secretion [80]. Specifically, removal of FSH occurs through glomerular filtration and this in turn is directly influenced by sialylation. Sialylation affects the isoelectric point of FSH at biological pH, imposing a net negative charge that hinders the capacity of the glomeruli to filter such proteins. Increased sialylation of FSH has subsequently been associated with extended serum half-life and increased biological activity [81–83]. Integration of additional glycosylation sites both imbedded in FSH and at its peptidyl N-terminus result in a significant increase in the serum half-life when compared to its unmanipulated equivalent. Modification of recombinant FSH by modulating changes in N-linked glycosylation represents an opportunity to improve both the efficacy and biological activity of the product, ultimately improving patient care through relaxed dosing regimens and associated costs.

1.30.4.2

Growth Factors: EPO

Sales of EPO therapies in 2008 accumulated to over $7 billion (USD), emphasizing the importance of its use in the treatment of numerous disease pathologies. EPO is a 36-kDa glycoprotein belonging to the cytokine superfamily and exhibits both N- and O-linked glycosylation. Recombinant EPO is used for the treatment of anemia stemming from chronic kidney disease, cancer, and HIV infections and has also been shown to play an important role in response to injury. This growth factor not only functions in erythrocyte differentiation but also plays an important function in neuroprotection [84–87]. The contrasting role between EPO-associated erythropoiesis activation and neuroprotective functions is largely influenced by its half-life in vivo, where extended exposure results in erythropoietic activity and shortened exposure leads to neuroprotective features [88–91]. Not surprisingly, its in vivo half-life is largely determined by the glycosylation state of the protein. EPO contains three N-linked glycan sequence motifs and two O-linked motifs, all of which contribute to its efficacy. Modulation of its glycosylation state significantly affects both its biological activity and production, where mutagenesis of individual Asn residues leads to poor secretion during production, and mutagenesis of multiple Asn residues affects in vivo biological activity [88]. Further to this point, the sialylation of EPO, in particular, plays an important role in its efficacy, evidenced by a reduction in serum half-life from approximately 2 h to less than 10 min for desialylated EPO. Depending on the intended aim for its use, the sialylation state may be seen as an advantage or a disadvantage. Extended serum half-life of EPO stimulates red cell production, potentially causing platelet aggregation and subsequent thrombosis. Alternatively, removal of sialic acid residues from EPO provides neuroprotective functions in animal models of stroke and spinal cord injury, while avoiding the activation of erythropoiesis [92]. In cases such as anemia, erythropoiesis is a critical feature of EPO administration and extended serum half-life is advantageous for improved potency and reduced dosage frequency. With this in mind, EPO has been engineered to include additional N-linked sites for glycosylation, resulting in the production of therapies such as Darbopoetin Alpha, an EPO derivative with improved PK behavior and in vivo activity [89,90]. The glycosylation status of recombinant EPO, therefore, plays a significant role in determining the ultimate suitability of the therapeutic for a given condition.

1.30.4.3

Enzyme Replacement Therapies for Lysosomal Storage Diseases

Over 40 different diseases exist due to deficiencies in lysosomal proteins [93]. Lysosomal storage diseases (LSDs) develop as a result of deficiencies or defects in the enzymes or enzyme co-factors responsible for the catabolism of glycoconjugates in lysosomes.

Glycomics

423

Accumulation of intermediates in the catabolic pathway due to the inability for removal causes numerous symptoms of varying severity such as growth retardation and skeletal deformities [94,95]. Current methods for treatment of LSDs include chemical chaperone therapy, bone marrow transplantation, substrate deprivation therapy, and enzyme replacement therapy (ERT). Each has their own drawbacks in terms of safety and efficacy; however, developments in the area of ERT have shown promise, largely due to the engineering of recombinant glycoprotein therapeutics. Currently, ERT is used to treat Gaucher’s disease [96], Fabry disease [97], Pompe disease [98], and more recently, mucopolysaccharadosis I [99], II [100], and VI [101].

1.30.4.3.1

Recombinant a-galactosidase A for the treatment of Fabry disease

Fabry disease manifests from a deficiency of the lysosomal enzyme a-galactosidase A (a-GalA), resulting in the accumulation of the enzyme substrate globotriaosylceramide [102]. Recent advancements in ERT have resulted in the production of recombinant glyco-engineered a-GalA, expressed in S. cerevisiae modified by the deletion of och1 (a(1,6)-mannosyltransferase) and mnn1 (a(1,3)-mannosyltransferase) eliminating highly mannosylated structures [103]. Additionally, mnn4 was engineered for constitutive expression, resulting in the continued availability of phosphomannosyltransferase [104]. This results in the phosphorylation of mannose residues, effectively targeting the recombinant a-GalA to mannose-6-phosphate receptors. Directing recombinant enzymes to the mannose-6-phosphate receptor represents a novel strategy for the delivery of enzymes to the lysosome, whereupon enzymes can then begin the process of catabolic degradation of the accumulated substrates.

1.30.4.3.2

Production of b-glucocerebrosidase for the treatment of Gaucher’s disease

Although typically highly mannosylated structures are undesirable, some therapeutics have been engineered to interact with mannose receptors for targeted uptake. Cerezyme is a recombinant glucocerebrosidase used as an ERT for the treatment of Gaucher’s disease. Cerezyme has been engineered to contain glycan structures of the high-mannose type that consequently become the target of macrophages within the liver. Cerezyme interacts with the mannose receptor on macrophages where it is then endocytosed, effectively delivering the recombinant enzyme to the site of glucocerebroside accumulation within lysosomes. Glycan engineering of this description represent an emerging trend in tissue targeting of glycoprotein therapeutics.

1.30.4.4

Cytokines: IFN-g

IFN-g is a soluble cytokine that exists as a dimerized glycoprotein and is the only member of the type II class of interferons [105]. IFN-g is a key regulator in both innate and adaptive immunity and is secreted primarily by natural killer (NK) and Natural Killer T (NKT) cells [106]. Glycosylation of IFN-g can occur on Asn25 and/or Asn97 [107] where glycans identified are typically found to be of the biantennary type with variable degrees of fucosylation and sialylation in both natural human and Chinese hamster ovary (CHO)-expressed IFN-g [108–110]. Recombinant IFN-g is used for the treatment of chronic granulomatous disease; however, efforts to pursue its use as a therapeutic were subsequently abandoned following phase III trials showing no benefit for its use. Critical to the function of recombinant IFN-g is its glycosylation, where the N-linked glycans influence its dimerization and secretion [111], protease resistance [112], antigenicity [113], and protein stability [113]. As with all therapeutic glycoproteins, improved PK behavior and biological activity is desired and so efforts have been made to improve its glycosylation. Notably, IFN-g has been expressed in a CHO derivative transfected with the a(2,6)-sialyltransferase to humanize the sialylation [114].

1.30.4.5

Antibody-Derived Therapeutics

Aggarwal estimated that in 2012, mAbs US sales reached $24.6 billion, an 18.3% growth over their 2011 sales [115] with two therapies (Remicade and Humira) responsible for over $8.2 billion (USD) alone. Monoclonal antibodies are scaffolds of immunoglobulin G (IgG), a heterodimeric glycoprotein constructed of two heavy and two light chains, covalently linked by disulfide bridges of varying number depending on the IgG subclass (Fig. 6). Functionally, the assembled IgG contains two domains of importance: the Fab region and the Fc region. Recognition of antigens by IgG is the sole function of the Fab region while the Fc region acts to coordinate with specific receptors of the immune system to instigate immune effector functions. Within the IgG class are four subclasses, namely IgG1, IgG2, IgG3, and IgG4, which demonstrate varying physiological preferences for Fc receptors of the immune system [116]. Monoclonal antibodies have a wide distribution of biological function and are used for the treatments of several pathologies, including cancer, autoimmune diseases, viral infections, asthma, and transplantation rejection. Of significance to both the PKs and biological activity of these products is the glycosylation of the Fc region of the monoclonal antibody, where loss of the glycan moieties at the conserved Asn297 ablates immune effector functions [117]. A review by Reusch and Tejada focuses on the role of Fc glycans as Critical Quality Attributes (CQA) to ensure the desired product quality, safety and efficacy for mAbs [118]. The composition of the Asn297-linked (N-linked) carbohydrate of IgG plays a highly influential role in determining the affinity of IgG for the receptors of the immune system, which include three classes of Fcg receptors (FcgRIa, FcgRIIa/b, and FcgRIIIa/b), the complement receptors (C1q and mannose-binding lectin (MBL)), and the neonatal receptor (FcRn). These receptors are responsible for coordinating the IgG Fab-directed recognition of antigens with specific immune effector functions (mediated by the Fc region), which then act to eliminate the IgG Fab-targeted antigen. The neonatal receptor, FcRn, is largely responsible for the recycling and transport of IgG [119,120] and as such, it plays a significant role in IgG half-life. Interestingly, the Fc–FcRn does not appear to be influenced by Fc-glycosylation as aglycosylated IgGs demonstrate a similar clearance rate as glycosylated equivalents [121,122]. In mammalian cells, glycosylation of IgG is observed as a single N-linked glycan on each of the two CH2 domains in the Fc region is

424

Glycomics

IgG1 IgG2

IgG4 IgG3

Figure 6 Immunoglobulin G subclass structures. IgG contains four subclasses that all demonstrate structural differences. Most noticeably different is the IgG3 subclass, which exhibits an extended hinge region that makes it susceptible to proteolytic cleavage.

composed of a core biantennary heptapolysaccharide containing N-acetylglucosamine (GlcNAc) and mannose. Further modification of the core carbohydrate structure is species specific, with fucose, bisecting GlcNAc, galactose, and terminal sialic acids. Differential glycosylation of the Fab and Fc region has also been reported [123], although a full appreciation of Fab glycosylation function has yet to be elucidated. IgG oligosaccharides are sequestered within the interstitial space enclosed by the two CH2 domains and interact with each other and the protein–carbohydrate interface, contributing to the structural conformation of the two heavy chains, in turn, influencing the binding of IgG to the family of Fc receptors. The absence of the Fc glycan results in a misconformation that perturbs the binding regions for the Fcg receptors, resulting in impaired effector functions of IgGs. The affinity of such aglycosylated antibodies for their antigens remains essentially unchanged, highlighting the influence of the glycan specifically on the Fc–Fcg receptor interactions. Manipulation of Fc glycosylation significantly impacts IgG association with Fc receptors, potentially improving the ability to activate specific immune effector functions. This has profound effects on the potency of monoclonal antibodies, specifically in cancer therapies where targeted cell killing via antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) is the preferred mode of action. Activation of ADCC is largely facilitated by the recruitment of NK cells, which express FcgRIIIa of the Fc receptor family exclusively. Hence, an extensive amount of research has been directed toward improving ADCC activation through enhancing the Fc–FcgRIIIa interaction, which has led to the identification of the core a(1,6)-fucose as being largely responsible for the improved binding.

1.30.5

Engineering Glycosylation for Improved Glycoprotein Function

Safety and efficacy of glycoprotein therapeutics and, indeed, all therapeutics are of paramount concern for regulatory agencies. Manipulation of the glycan biosynthetic machinery within a mammalian host provides an opportunity to selectively glycosylate therapeutics for improved half-life, subsequently improving the efficacy and potentially reducing the likelihood of immunogenicity of the therapy. The half-life of a given therapeutic is governed by its interaction with specific glycan-recognizing receptors whose sole purpose is to eliminate foreign material from the host environment. Removal of glycoprotein therapeutics primarily occurs through aberrant glycosylation identified by the host immune system. The soluble receptors of the complement system, namely the MBL and

Glycomics

425

complement protein C1q, bind to glycans containing terminal N-acetylglucosamine (GlcNAc) and target the identified protein for removal by complement [124,125]. Glycoproteins that exhibit terminal mannosylation are recognized by the mannose receptor present on macrophages. Interaction with the mannose receptor results in phagocytic uptake and processing of the therapeutic. Terminal mannose (Man) or N-acetylglucosamine (GlcNAc) influence the PK of glycoprotein therapeutics through interaction with circulating mannose receptors expressed on NK cells and macrophages [126].

1.30.6

Sialylation in the Context of PK Behavior and Biological Activity

The role of sialylation in glycoprotein function can be considered in two ways: (1) its impact on the associated protein half-life or (2) its impact on biological activity. In the context of Fc-derived therapeutics, sialylation is associated with anti-inflammatory properties, which given the current intention of monoclonal antibody-based therapeutics is considered a disadvantage. Individual sugar residues can modulate specific interactions between the Fc of IgG and individual Fc receptors. Biantennary sialic acid-capped glycans have been demonstrated to provide the anti-inflammatory properties of intravenous IgG (IVIG) [127]. The use of IVIG for the treatment of several autoimmune and inflammatory diseases requires high doses (1 g kg 1), raising speculation as to its mode of action. However, Kaneko et al. [127] were able to demonstrate that the high-dose requirements were a reflection of the sialylated IgG content of IVIG. Sialic acid containing IgGs were purified from IVIG with this enriched fraction shown to provide equivalent anti-inflammatory properties. Desialylation of IVIG subsequently removed its protective quality, clearly demonstrating the function of IgG sialylation in IgG anti-inflammatory properties. The involvement of the C-type lectin SIGN-R1 has recently been proposed as the receptor for the sialic acid residues, in turn, providing the anti-inflammatory properties [128]. Although sialylation has now been shown to affect the IgG biological activity, there does not appear to be any deleterious effect on the IgG half-life due to loss of sialic acid [129], suggesting that clearance mechanisms of IgG may differ from other glycoproteins. Interestingly, sialylation has a markedly different outcome on the PK of non-Fc-based glycoprotein therapeutics, where desialylation significantly impacts the in vivo half-life of several therapeutics such as EPO, FSH, and IFN [82,90,130]. Hence, a challenge for the production of recombinant proteins with N-linked glycosylation involves the improved sialylation of the final product. For example, asialo-EPO exhibits a significantly reduced half-life of 2 min while sialylated EPO has an observed half-life of between 5 and 6 h [92], clearly defining a role for sialylation in PK behavior. To overcome the challenge of protein sialylation, popular eukaryotic expression systems such as Pichia pastoris have been engineered to hypersialylate recombinant proteins such as EPO in an effort to improve its half-life [131].

1.30.7

Core a(1,6)-Fucosylation Affects the Biological Activity of IgG

Of the current FDA-approved monoclonal antibody therapies, over 50% are engineered for use in either the treatment or diagnosis of cancer [132]. The mode of action for many of the antibodies is through targeting of specific cancer-associated markers and the subsequent recruitment of immune effector mechanisms, resulting in targeted elimination of the cancerous cells. Paramount to the recruitment of the appropriate immune cells is the association with Fcg receptors, specifically FcgRIIIa. NK cells express this receptor exclusively and engagement of this receptor with the Fc of IgG results in the activation of ADCC, a critical immune effector function in the destruction of targeted cancerous cells. Loss of the core fucose leads to enhanced ADCC activity [133]. Removal of the core fucose leads to an improved affinity for FcgRIIIa, in turn, inducing ADCC activity. Moreover, the glycosylation of FcgRIIIa itself affects the binding affinity. Two key glycans at positions Ans162 and Asn45 modulate the binding of IgG [134,135]. Appropriately, CHO cell lines deficient in the fucosyltransferase, FUT8 have been engineered for the production of monoclonal antibody therapeutics deficient in a(1,6) core fucose for the treatment of cancer [136].

1.30.8

The Bisecting b(1,4)-N-Acetylglucosamine Influences ADCC

Loss of the core a(1,6)-fucose residue significantly enhances the ADCC activity of therapeutic monoclonal antibodies and methods for its elimination have subsequently become a goal in bioproduction for improving the biological activity of therapeutic antibodies intended for use in cancer treatment. As mentioned previously, this has been facilitated in part by the disruption of FUT8, the glycosyltransferase responsible for the addition of a(1,6)-fucose. The inhibition of core fucosylation has also been achieved by an alternative approach that introduces a bisecting b(1,4)-GlcNAc residue to the core mannose-b(1,4)-GlcNAc carbohydrate that acts to eliminate the substrate required for a(1,6)-fucosyltransferase activity, namely GlcNAc2Man3GlcNAc2Asn [137]. The b(1,4)-N-acetylglucosaminyltransferase (GnT-III) enzyme is responsible for the addition of a bisecting GlnNAc to any glycan structure modified by GnT-I but not GalT, whether or not the glycan is fucosylated. Overexpression of GnT-III represents a novel strategy for engineering monoclonal antibodies with improved biological activity. However, popular production cell lines such as NS0 or CHO do not express the enzyme [138,139], introducing a minor impediment for the employment of this strategy in bioproduction. This point has been addressed by Umana et al. [140], where CHO was engineered to overexpress GnT-III in a cell line producing a low ADCC activity anti-neuroblastoma IgG1. Glycan analysis of the IgG isolated from the engineered cell line generated predominantly bisecting hybrid structures and this glyco-engineered IgG demonstrated enhanced ADCC activity. The isolation of the bisecting hybrid structures prompted further manipulation of the N-linked biosynthetic pathway. Overexpression of GnT-III

426

Glycomics

resulted in primarily high-mannose structures containing bisecting GlcNAc, most likely due to the inhibitory activity of GnT-III on a-mannosidase II (Man-II). To overcome this premature stop in glycan processing, further modifications were made to the heterologous GnT-III by generating chimeric GnT-III fused with localization domains from other Golgi-residing enzymes and co-expressing the fusion construct with Man-II [141]. The co-expression of Man-II provided exogenous enzyme to further drive the processing of the high-mannose structures through to complex biantennary glycans. Fusion of the GnT-I, a(1,6)-FucT, GnT-II, or Man-II to the GnT-III catalytic domain produced a fusion construct (GnT-IIIManII) with an enhanced ability to catalyze the addition of the b(1,4)-GlcNAc moiety, which resulted in the isolation of IgG glycans of two descriptions: (1) increased bisecting, nonfucosylated hybrid structures when the fusion construct was expressed exclusively and (2) complex, nonfucosylated glycans containing a bisecting GlcNAc when the fusion construct was co-expressed with Man-II. IgGs were isolated from each cell line and assayed for ADCC activity where each IgG glycoform demonstrated enhanced potency when measured in B-cell depletion assays. Manipulation of the N-linked biosynthetic pathway by providing increased GnT-III and Man-II has clearly demonstrated the advantages for eliminating core fucosylation by upregulating the addition of a bisecting GlcNAc. It should be noted that enhanced ADCC and CDC were only observed when Man-II was co-expressed with GnT-III. Inducible expression of exogenous glycosyltransferases offers a novel strategy for improving the biological activity of monoclonal therapeutics, but this technology must be regulated with caution. Disruption of the N-linked biosynthetic pathway potentially results in the formation of structures that may influence both efficacy and safety of the candidate therapy.

1.30.9

Eukaryotic Expression Systems for Recombinant Glycoprotein Production

Unquestionably, the production of glycoprotein therapeutics hinges upon the ability of the selected expression system to perform N-linked glycosylation in a manner similar to that observed in humans. Failure to do so faces the likelihood of generating a product that is rapidly cleared or potentially induces immune responses upon administration. With an obvious necessity for N-linked glycosylation in glycoprotein therapeutics, protein expression workhorses such as Escherichia coli are not an option due to the absence of the required glycosylation machinery. However, several eukaryotic cell lines are available that are capable of glycosylating recombinant proteins. This includes transgenic plants, yeast, insect cell lines, murine myeloma cell lines, and the industry-preferred CHO cell lines. Each of these systems boasts several advantages in terms of their ability to produce glycosylated proteins. With that said, there are also several disadvantages. Appropriately, each cell expression system has experienced significant attention, and several modifications to each N-linked biosynthetic pathway have been in an effort to improve its ability to “humanize” glycosylation, therein improving therapeutic efficacy, safety, and biological activity.

1.30.10 Transgenic Plants Modified for “Humanized” N-Glycosylation Plants present an attractive alternative to many of the complications experienced with mammalian cell expression systems. Generally, plants require less investment and operating costs and are typically devoid of human or animal pathogens and endotoxins. Subsequent protein purification is facilitated by directing recombinant proteins to the plant endosperm, where proteins can then be easily extracted. In light of these advantages, plants have been selected for the production of several recombinant proteins such as avidin [142], b-glucuronidase [143], and hirudin [144]. However, plants still exhibit a number of disadvantages such as poor expression levels and difficulties with downstream processing. More prevalent is the inconsistencies between plant and mammalian N-linked glycosylation. Plant N-linked glycans can be broadly categorized into three groups: (1) glycans of the high-mannose type (Man5GlcNAc2 to Man9GlcNAc2); (2) complex N-linked glycans; and (3) paucimannosidic-type glycans. Other glycans observed in plants include the presence of the Lewis a (Lea) antigen and the inclusion of known immunogenic epitopes b(1,2)-xylose and a(1,3)-fucose [145]. Further concern with the biosynthetic pathway in plants involves the lack of both galactosylation and sialylation of glycoproteins, which is a significant disadvantage when considering the impact of both carbohydrates on the PK behavior of current glycoprotein therapeutics. Hence, the use of transgenic plants is reluctantly accepted at the cost of potential side effects, which include the activation of immune responses and/or premature clearance of the glycoprotein therapeutic. In terms of premature clearance, IgE and IgG class antibodies specific for plant glycan moieties have been identified in allergic patients [146], raising concern for the use of so-called “plantibodies” and other plant-derived glycoproteins for therapeutic use in patients with underlying allergies. However, with the obvious advantages for plant-derived expression of recombinant glycoproteins, significant advances have been made to genetically manipulate plants for “humanized” glycosylation as a strategy to improve the biological activity and efficacy of candidate therapeutics. The humanization of N-linked glycan synthesis in plants has been principally achieved through two approaches: (1) the reduction or elimination of plant-specific glycosylation generated by the activity of a(1,3)fucosyltransferase (a(1,3)-FucT) and b(1,2)-xylosyltransferase (b(1,2)-XylT) and (2) the introduction of foreign enzymes to mimic human glycosylation, such as b(1,4)-galactosyltransferase (b4GalT-I) and b(1,4)-N-acetylglucosaminyltransferase (GnTIII). One strategy for eliminating a(1,3)-fucosylation includes the disruption of genes associated with the synthesis of L-fucose such as mur1 in Arabidopsis [147]. In this system, nearly 95% of the resulting glycans analyzed lacked fucose with no other carbohydrate providing a substitute. Rat GnTIII has also been expressed in Nicotiana tabacum, providing the bisecting b(1,4)-GlcNAc that prevents the addition of both a(1,2)-xylose and a(1,3)-fucose [148]. A primary candidate for biosynthetic manipulation of N-linked

Glycomics

427

glycosylation is the moss Physcomitrella patens, primarily due to its ease of use in genetic engineering [149]. Two critical manipulations to P. patens have been the elimination of both a(1,3)-FucT (fucT) and the b(1,2)-XylT (xylT), which when used to express recombinant EPO, resulted in the isolation of glycosylated protein completely devoid of either b(1,2)-xylose or a(1,3)-fucose [150]. The isolation of high-mannose plantibodies in tobacco was achieved by the incorporation of the ER retention signal to the C-terminus of the IgG heavy chain, resulting in the isolation of antibodies primarily containing Man6-9GlcNAc2 structures. The evolution of plants for the bioproduction of recombinant glycoprotein therapeutics is an exciting prospect, and efforts to adapt several plant species for humanized N-linked glycosylation are well underway. Key to this process is the elimination of potential allergenic epitopes and addition of enzymes capable of further synthesis of complex N-linked glycans. This technology still faces several challenges with protein yield and potential immunogenicity of resulting glycoproteins most prevalent. Should these challenges be met, the use of transgenic plants offers a highly attractive process for recombinant protein production.

1.30.11 Baculovirus-Mediated Insect Cell Expression Systems The use of insect cell lines for the production of recombinant glycoproteins offers a number of distinct advantages, principally the high protein yields obtained when transfected with baculovirus-derived expression vectors and the high degree of similarity in posttranslational modifications when compared to mammalian glycosylation. Growth of insect cell lines does not require serum, making their selection cost effective and biosafe, and insect cells are second only to mammalian cells in terms of posttranslational modifications. Insect cells are capable of both N- and O-linked glycosylation with a similar core structure to that observed in mammalian cells, namely the heptapolysaccharide structure containing terminal b(1,2)-N-acetylglucosamines observed during glycan processing in the medial Golgi. However, insect cell lines glycosylate proteins not unlike plants, with the majority of structures being paucimannosidic or modified glycans containing a(1,3)-fucose [151]. The selection of insect cell lines for recombinant glycoprotein expression faces many obstacles. For their use in bioproduction to be safe and efficacious, elimination of a(1,3)-fucosylation must be approached as a top priority. The activity of GlcNAcase must also be prevented if complex biantennary structures are to be assembled. GlcNAcase removes the b(1,2)-N-acetylglucosamine addition to mannose, preventing the extension of more complex glycans [152]. Extending this point, the activity of GlcNAc transferase II must then be elevated to improve the likelihood of antennary arm glycosylation. In light of several advantages for protein expression in insect cell lines, research has progressed toward modifying these cells for humanized glycosylation. Notably, the transfection of b(1,2)-N-acetylglucosaminyltransferase I (GnTI) in Spodoptera frugiperda (Sf9) cells was observed to progress the processing of insect-derived paucimannosidic type glycans toward hybrid structures, where a terminal GlnNAc saccharide was identified on the core Man3GlcNAc2 structure [153]. Although this structure (GlcNAcMan3GlcNAc2) provides the substrate for subsequent glycosyltransferase reactions, the authors did not report the presence of either galactose or sialic acid on glycans from this Sf9 transfectant. Perhaps a greater stumbling block when considering recombinant protein expression in insect cells is the lack of sialyltransferase activity, but more so the lack of activated sialic acid in the form of cytidine monophosphate (CMP)– NeuAc. Accordingly, Sf9 cells have benefited from extensive engineering that complements the glycosylation machinery with (b4GalT-I) and a(2,6)-sialyltransferase [154]. Salvage pathways for the metabolism of CMP–NeuAc have also been identified [155], providing all the constituents for the production of fully processed, complex biantennary glycan structures, albeit under strict conditions where culturing must be in the presence of fetal bovine serum. Alternative methods to side-step the requirement of serum in culture media have been addressed, where Sf9 cells have been engineered to express sialic acid synthase and CMP–sialic acid synthase [156]. The efforts to humanize N-linked glycosylation in insect cells has culminated in the development of the SfSWT-3 cell line, engineered to express five fundamental enzymes required for further processing of the typical paucimannosidic glycans observed in insect cells [157]. Compared with other transgenic insect cells, SfSWT-3 offers a distinct advantage in that growth can occur in the absence of serum provided the media is supplemented with N-acetylmannosamine. This aids in improving the safety and downstream purification of recombinant protein expression.

1.30.12 Restructuring of Yeast N-Linked Glycan Biosynthesis Yeast are a popular eukaryotic expression system for the production of recombinant proteins due to the ability to achieve high cell densities in culture, shorter fermentation times, and the utilization of chemically defined media that eliminate the possibility of viral or prion contamination. Accordingly, yeast together with E. coli – for similar reasons – have been employed for the synthesis of the majority of nonglycosylated proteins, with yeast specifically responsible for the production of over half of the world’s supply of insulin. It is believed that approximately one-sixth of therapeutic proteins currently approved are produced in yeast; however, none of these therapeutics are glycosylated, largely due to the differences observed between human and yeast in N-linked glycosylation pathways. Yeast typically synthesize glycans of the high-mannose type that are invariably recognized by the mannose receptor in humans and subsequently removed from circulation. N-linked glycosylation of proteins in yeast occurs similarly to that observed in humans during the initial assembly stages in the ER. Yeast catalyze the transfer of the Glc3Man9GlcNAc2 oligosaccharide to the nascent polypeptide at the Asn-X-Ser/Thr consensus sequence (where X is any amino acid excluding proline) whereupon glucosidase I and II remove the three terminal glucose residues,

428

Glycomics

and the a-1,2 mannosidase residing in the ER removes a terminal a-1,2 mannose. Glycan processing then continues to the Golgi apparatus, where the similarities between humans and yeast end. While humans proceed to cleave the Man8GlcNAc2 structure back to Man5GlcNAc2, yeast glycosylation is limited to mannosylation and the addition of mannosylphosphate groups by a range of mannosidases including a-1,2, a-1,3, and a-1,6 mannosyltransferases as well as mannosylphosphate transferases [158]. Thus, the resulting outcome of glycoprotein production in yeast ultimately leads to the generation of a high-mannose recombinant protein that, when administered to humans, faces the likelihood of rapid clearance through phagocytosis or the lectin pathway of CDC, mediated by mannose receptors (MR) and mannose binding lectin (MBL). However, with several advantages in terms of protein yield and media composition, production of recombinant glycoproteins in yeast has undergone a revolution in terms its capacity to humanize N-glycosylation of recombinant proteins. This has been achieved largely by the genetic manipulation of yeast to perform glycosylation consistent with humans and to eliminate enzymes that lead to the incorporation of undesirable sugar constructs. Several groups have focused on genetically manipulating yeast to simulate the N-linked glyco-biosynthetic pathway in humans. Two approaches were applied: (1) eliminate enzymes in yeast not consistent with human glycosylation and (2) complement yeast with enzymes that incorporate human carbohydrates. Elimination of genes encoding the a(1,6) and a(1,2) mannosyltransferases was an initial focus, as this strategy prevented the hypermannosylation commonly associated with yeast glycosylation. Chiba et al. [159] performed a double knockout of och1 and mnn1 and subsequently identified the presence of Man5GlcNAc2 glycans, reflective of a block in hypermannosylation. Further development of complex glycan structures in yeast proved more problematic, largely due to the lack of success in progressing Man8GlcNAc2 structure to the Man5GlcNAc2; an essential substrate for the initiation of complex glycan synthesis. Trimming of Man8GlcNAc2 is performed by a(1,2) mannosidase, an ER-specific glycosidase responsible for removal of three a(1,2)-linked mannose residues to yield Man5GlcNAc2 [160]. Several attempts have been made to provide this enzyme exogenously; however, issues have arisen with its localization to the ER, the activity once there, and competition between the exogenous enzyme and endogenous yeast ER-residing enzymes. However, through construction of an extensive yeast combinatorial library, yeast mutants were isolated that were capable of efficiently trimming Man8GlcNAc2 to Man5GlcNAc2 [161]. The library was constructed through fusion of N-terminal signal sequences (obtained from yeast enzymes known to localize to the ER and Golgi) with enzyme catalytic domains and subsequent expression of the chimeric enzymes in a yeast mutant lacking a(1,6)mannosyltransferase activity. By a similar approach, Choi et al. [161] simultaneously engineered the localization of GlcNAcT I – the N-acetylglucosaminyltransferase required for the addition of a b(1,2)-GlcNAc to the a(1,3)-mannose of Man5GlcNAc2 – to the Golgi in an effort to progress the development of complex N-glycan synthesis. Further engineering of yeast N-linked glycosylation has resulted in the introduction of a number of key enzymes responsible for human N-linked glycosylation, including GlcNAcT II and b(1,4)-GalT. To achieve efficient terminal galactosylation, several additional components were required for co-expression with b(1,4)-GalT. Lack of uridine diphosphate (UDP)–galactose was hypothesized as a contributor to the poor efficiency in yeast strains expressing endogenous b(1,4)-GalT based on attempts to improve galactosylation with limited success [162,163]. The co-expression of UDP–galactose-epimerase and the UDP–galactose transporter combined with b(1,4)-GalT subsequently generated glycans with extensive terminal galactose residues [164]. As sialylation is a critical feature for many glycoprotein therapeutics, yeast must be additionally capable of performing this last step in the development of complex N-linked glycans. Accordingly, P. pastoris was complemented with enzymes responsible for CMP–NeuAc biosynthesis, CMP–NeuAc transport, and transfer of sialic acid to the nascent protein [131]. The resulting construct generated in excess of 90% sialylation using recombinant EPO as a model glycoprotein. Taken together, the genetic manipulation of yeast to perform humanized glycosylation represents a significant step forward in the large-scale production of glycosylated protein therapeutics. With issues pertaining to high-mannose structures potentially eliminated, yeast can realistically be considered for the production of complex glycoprotein therapeutics.

1.30.13 CHO: The Workhorse in Bioproduction CHO cells continue to be the preferred cell line for recombinant glycoprotein production for a number of reasons. Primarily, CHO has been the focus of extensive manipulation all in an effort to provide a cell line capable of sustainable, high-yielding recombinant protein expression. These attributes have been engineered by the introduction of several components that increase both cell viability and productivity, which include: (1) the introduction of anti-apoptotic factors [165]; (2) the co-expression of chaperones such as Hsp70 and Hsp27, which assist in the formation and breakdown of disulfide bonds [166]; and (3) the introduction of proteins such as Sly1 and Munc18c [167] as well as the ceramide transfer protein [168], which are all involved with vesicle trafficking from the ER to the Golgi. Significant efforts have also been implemented to improve the N-linked glycan biosynthetic pathway in CHO. Nucleotide precursors such as CMP–sialic acid were added to the culture media, resulting in a slight improvement in sialylation [169]. However, to improve the efficiency of sialylation, the supporting components such as UDP–GlcNAc-2-epimerase (which synthesizes sialic acid), CMP–sialic acid transporter, and the sialyltransferases themselves must be present and active, which would ultimately encourage the synthesis, transport, and addition of sialic acids during N-linked glycan biosynthesis. Sialylation has been the target of many of these manipulations, where both a(2,3)-sialyltransferase [170] and a(2,6)-sialyltransferase [130] have been overexpressed in CHO. Each engineered CHO cell line produced recombinant proteins demonstrating enhanced sialylation, suggesting that expression of sialyltransferases is a suitable process for enhancing sialylation of recombinant proteins. Core

Glycomics

429

fucosylation also remains a keen interest of the biopharmaceutical industry, and several advances have been made with respect to eliminating core fucosylation as have been previously discussed (see Section Core a(1,6)-Fucosylation Affects the Biological Activity of IgG).

1.30.14 Murine-Derived N-Linked Glycosylation Bioproduction for the majority of glycoprotein therapeutics is reliant on two mammalian cell lines: CHO and mouse myeloma (NS0 and Sp2/0). These two cell lines offer a distinct advantage over other cell systems such as yeast, insect, or plant as they are well characterized, are adapted for suspension cell growth, can be easily transfected, and high producers can be isolated in a relatively straightforward manner [171]. Four FDA-approved monoclonal antibodies (Erbitux, Remicade, Synagis, and Zenapax) are produced in murine cells, while 20 FDA-approved glycoprotein therapeutics are expressed in CHO [68]. One notable reason for the obvious preference between CHO and murine myeloma cell lines is the incorporation of immunogenic epitopes on recombinant proteins glycosylated in murine cells. Specifically, murine cell lines express a(1,3)-galactosyltransferase and CMP–NeuAc-hydroxylase, which generate galactose-a(1,3)-galactose and NGNA, respectively. Both are known immunogens with approximately 1% of normal human serum IgG containing anti-a(1,3)-galactose antibodies. The full impact of immunogenic epitopes on commercial monoclonal antibodies was recently made known when approximately 20% of patients receiving cetuximab (an anti-epidermal growth factor receptor (EGFR) antibody expressed in Sp2/0) experienced severe anaphylaxis, mediated by IgE antibodies specific for galactose–a(1,3)-galactose epitopes [172]. NGNA is a derivative of NANA (sialic acid) not found in humans [173] but synthesized by both NS0 and CHO [174]. Moreover, it is described as an oncofetal antigen and human serum also contains anti-NGNA antibodies [175]. Low levels of NGNA are somewhat acceptable; however, higher levels are prone to immune responses as evidenced by the rapid clearance of CT4-IgG [176]. Thus, glycoprotein therapeutics expressed in murine cell lines run the risk of incorporating immunogenic epitopes, subsequently becoming the target of clearance through antibodymediated mechanisms.

1.30.15 The Changing Landscape of Regulatory Agencies Toward Glycosylation of Biopharmaceuticals The importance of glycosylation became significant after the discovery of monoclonal antibodies. In the 1970s, the technology of generating monoclonal antibodies specific for predetermined antigens signaled a revolution of new biologics for the treatment of diseases that were previously thought untreatable such as cancer, autoimmune diseases, infections, and respiratory diseases. More importantly, approval of these therapies was based on regulations dating to the public service Act of 1944 which itself was an update from the biological control act of 1902 – certainly a different time in the world of biology. The first therapeutic antibody that entered the market in 1986 was muromonab, murine anti-CD3 IgG; it was withdrawn due to lack of efficacy and rapid clearance due to patients’ production of human anti-mouse antibodies (HAMAs). Concerns of HAMA responses have led to the development of humanized and chimeric monoclonal antibody (mAbs). These issues, in due course, became driving forces for the evolution of regulatory guidelines toward monoclonal antibodies and biopharmaceuticals. The European Medicines Agency (EMEA) introduced the first guidelines on production and quality control of monoclonal antibodies of murine origin in June 1987, which was then followed by quality-control guidelines for human monoclonal antibody in July 1990, with the FDA introducing similar guidelines in 1994. Bilateral discussions between Europe, Japan, and the United States during the WHO-sponsored conference of Drug Regulatory Authorities (ICDRA) in 1989 gave birth to international conference of harmonization (ICH) in April 1990. The ICH was primarily designed to set common guidelines for medicines produced by pharmaceutical companies to protect the public health from an international perspective. In 1999, the ICH expert working group published guidelines for the quality of biotechnological products. The current ICHQ6B article delineates the guidelines for carbohydrate analysis on biopharmaceuticals. The regulatory authorities require the pharmaceutical industry to characterize and study the carbohydrate content in relation to the presence of neutral sugars, amino sugars, and sialic acid sugars. Moreover, it is mandatory to analyze the structure of carbohydrate chains, their antennary profile, and glycosylation sites of the polypeptide chains. In addition, ICH6QB necessitates detection and characterization of altered glycosylation by relevant analytical methods. However, the current ICH regulatory guidelines on glycosylation lack clarity as to what extent glycan analysis is to be carried out in relation to site occupancy and detailed glycoforms analysis. In the recent past, microhetrogeneity and macroheterogeneity in many glycosylated biopharmaceuticals including viraferon (interferon a-2B, Schering-Plough), Myozyme (alglucosidase alfa, Genentech), EPO, mAbs, and tissue plasminogen activator (tPA) raised diverse safety concerns to the regulatory authorities. More importantly, it is not possible to predict more precise functional consequences of altered glycosylation. For example, glycoproteins produced in CHO and murine cells could result in antibodies with immunogenic glycoforms such as NGNA and galactose–a(1,3)-galactose residues. In March 2006, an immunomodulatory drug CD-28 SuperMAB (also known as TGN1412) designed for the treatment of B-cell leukemia and rheumatoid arthritis was withdrawn from the clinical trials and development. The administration of the drug resulted in catastrophic effects with multiorgan failure and cancer. Following the incidence, an interim report from the Medicines and Healthcare products Regulatory Agency (MHRA) concluded that the above effect was due to unpredicted biological action of the drug. In addition to the existing biologics, biosimilars or follow-on biologics have generated concerns due to altered glycoforms [177]. The EMEA approved four biosimilars in 2008, despite noted differences from the original product. At the same time, an interferon

430

Glycomics

biosimilar was rejected due to safety concerns [177]. New developments in mAb technology and intense research on the role of glycan architecture in various aspects of protein function such as immunogenicity, safety, efficacy, bioavailability, biodistribution, biological activity, and immune effector functions have forced the regulatory authorities in the United States and Europe to develop more explicit guidelines on glycan characterization. In Europe, the EMEA has taken a worldwide lead in establishing revised guidelines for monoclonal antibodies that came into effect from June 2009 and the revised directive outlines more stringent guidelines for glycan characterization. Under Section 4.3 of the revised guidelines, the regulatory directive requires manufacturers to confirm the presence or absence of glycosylation on the additional glycosylation sites on heavy chains of mAbs. It is also vital to characterize glycan structures in relation to the degree of mannosylation, galactosylation, fucosylation, and sialylation. With respect to galactosylation, it is required to determine the percentage and distribution of main glycan structures present (often G0, G1, and G2). Section 4.3.2 of the revised guidelines requires detailed biochemical analysis on glycan structures corresponding to immunological properties. Furthermore, under Section 4.3.4, it is essential to assess the heterogeneity of N-linked oligosaccharide by a combination of orthogonal methods. The above directive clearly displays the importance of glycosylation and will facilitate biomanufacturers to produce improved biological medicines.

1.30.16 Summary Posttranslational modification through glycosylation plays a critical role in the ultimate function of biological macromolecules, influencing their solubility, activity, stability, and localization. Significant strides have been made to determine the details of the machinery responsible for glycosylation processing and the role of glycosylation in biology; however, carbohydrate complexity and heterogeneity remain cryptic and many aspects of glycobiology have yet to be elucidated. The importance of glycosylation is such that novel strategies, including high-throughput U/HPLC, MS, and both glycan and lectin arrays, have been designed to assess glycan structures. Together, these methods have resulted in the accumulation of critical data that relates glycan structures to biological function. These techniques are providing a comprehensive approach toward understanding not only the function of individual glycans in an isolated system but also the function of the total glycan population in multiple systems. The advancements in glycomics and glycan analysis have been of significant benefit to the health industry, where disorders in glycan-related processes are now known to manifest in diseases, such as congenital disorders in glycosylation. Appropriately, characterization of glycan structure and function has provided opportunities to the biopharmaceutical industry for the engineering of novel biotherapeutics, capable of highly specific and potent activity, ultimately improving patient care by improving the safety and efficacy of these treatments while maintaining biological activity. However, scientific discovery progresses with astonishing pace and advancements in biotherapeutic discovery often outpace the regulatory guidelines that govern the safety of these new therapies. In terms of bioproduction, glycan analysis is playing a critical role in identifying the quality of the engineered therapeutic through assessment of critical features such as sialylation, fucosylation, and incorporation of immunogenic epitopes – all of which are susceptible to change due to variation in bioproduction parameters. The current methods for glycan analysis have provided robust tools for glycoprotein characterization, but often the accuracy of the method is subject to a time-consuming process. Future considerations for glycan analysis in the bioproduction environment will inevitably require streamlining with the goal of obtaining critical glycosylation data that allows for adjustments to maintain or improve the glycan quality of the glycoprotein being manufactured. The trend in protein biotherapeutics is demonstrating an increased contribution from glycoproteins and it is now appreciated that the extent of glycosylation results in varied protein function. Understanding the glycosylation process provides an opportunity for industry to select the most appropriate expression system based on the intended role of the candidate therapy.

References [1] Opat, A.S., van Vliet, C., Gleeson, P.A., 2001. Trafficking and localisation of resident Golgi glycosylation enzymes. Biochimie 83 (8), 763–773. [2] Ungar, D., et al., 2002. Characterization of a mammalian Golgi-localized protein complex, COG, that is required for normal Golgi morphology and function. J. Cell Biol. 157 (3), 405–415. [3] Lehle, L., Strahl, S., Tanner, W., 2006. Protein glycosylation, conserved from yeast to man: A model organism helps elucidate congenital human diseases. Angew. Chem., Int. Ed. Engl. 45 (41), 6802–6818. [4] Freeze, H.H., Westphal, V., 2001. Balancing N-linked glycosylation to avoid disease. Biochimie 83 (8), 791–799. [5] Altmann, F., et al., 2001. Genetic model organisms in the study of N-glycans. Biochimie 83 (8), 703–712. [6] Huffaker, T.C., Robbins, P.W., 1982. Temperature-sensitive yeast mutants deficient in asparagine-linked glycosylation. J. Biol. Chem. 257 (6), 3203–3210. [7] Branza-Nichita, N., et al., 1999. Tyrosinase folding and copper loading in vivo: A crucial role for calnexin and alpha-glucosidase II. Biochem. Biophys. Res. Commun. 261 (3), 720–725. [8] Schrag, J.D., et al., 2001. The structure of calnexin, an ER chaperone involved in quality control of protein folding. Mol. Cell. 8 (3), 633–644. [9] Jones, J., Krag, S.S., Betenbaugh, M.J., 2005. Controlling N-linked glycan site occupancy. Biochim. Biophys. Acta. 1726 (2), 121–137. [10] Hossler, P., Mulukutla, B.C., Hu, W.S., 2007. Systems analysis of N-glycan processing in mammalian cells. PLOS ONE 2 (8), e713. [11] Elsner, M., Hashimoto, H., Nilsson, T., 2003. Cisternal maturation and vesicle transport: Join the band wagon! (Review). Mol. Membr. Biol. 20 (3), 221–229. [12] Abd Hamid, U.M., et al., 2008. A strategy to reveal potential glycan markers from serum glycoproteins associated with breast cancer progression. Glycobiology 18 (12), 1105–1118. [13] Opdenakker, G., et al., 2006. Remnant epitopes, autoimmunity and glycosylation. Biochim. Biophys. Acta. 1760 (4), 610–615. [14] Radcliffe, C.M., et al., 2007. Human follicular lymphoma cells contain oligomannose glycans in the antigen-binding site of the B-cell receptor. J. Biol. Chem. 282 (10), 7405–7415.

Glycomics [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68]

431

Ritchie, G.E., et al., 2002. Glycosylation and the complement system. Chem. Rev. 102 (2), 305–20–19. Rudd, P.M., et al., 2001. Glycosylation and the immune system. Science 291 (5512), 2370–2376. Saldova, R., et al., 2007. Ovarian cancer is associated with changes in glycosylation in both acute-phase proteins and IgG. Glycobiology 17 (12), 1344–1356. Saldova, R., et al., 2008. Glycosylation changes on serum glycoproteins in ovarian cancer may contribute to disease pathogenesis. Disease Markers 25 (4–5), 219–232. Van den Steen, P., et al., 1998. Cytokine and protease glycosylation as a regulatory mechanism in inflammation and autoimmunity. Adv. Exp. Med. Biol. 435, 133–143. Watson, M., et al., 1999. Sugar printing rheumatic diseases: A potential method for disease differentiation using immunoglobulin G oligosaccharides. Arthritis Rheum. 42 (8), 1682–1690. Royle, L., et al., 2008. HPLC-based analysis of serum N-glycans on a 96-well plate platform with dedicated database software. Analytical Biochemistry 376 (1), 1–12. Anumula, K.R., 2000. High sensitivity and high resolution methods for glycoprotein analysis. Glycobiology 10 (10), 1138. Anumula, K.R., 2006. Advances in fluorescence derivatization methods for high-performance liquid chromatographic analysis of glycoprotein carbohydrates. Analytical Biochemistry 350 (1), 1–23. Guile, G.R., et al., 1996. A rapid high-resolution high-performance liquid chromatographic method for separating glycan mixtures and analyzing oligosaccharide profiles. Analytical Biochemistry 240 (2), 210–226. Lauber, M.A., et al., 2015. Rapid preparation of released N-glycans for HILIC analysis using a labeling reagent that facilitates sensitive fluorescence and ESI-MS detection. Analytical Chemistry 87 (10), 5401–5409. Varki, A., et al., 2015. Symbol nomenclature for graphical representations of glycans. Glycobiology 25 (12), 1323–1324. Dalziel, M., et al., 2014. Emerging principles for the therapeutic exploitation of glycosylation. Science 343 (6166), 1235681. Tangvoranuntakul, P., et al., 2003. Human uptake and incorporation of an immunogenic nonhuman dietary sialic acid. Proceedings of the National Academy of Sciences of the United States of America 100 (21), 12045–12050. Ghaderi, D., et al., 2010. Implications of the presence of N-glycolylneuraminic acid in recombinant therapeutic glycoproteins. Nat. Biotechnol. 28 (8), 863–867. Beavis, R.C., Chait, B.T., 1990. High-accuracy molecular mass determination of proteins using matrix-assisted laser desorption mass-spectrometry. Analytical Chemistry 62 (17), 1836–1840. Beavis, R.C., Chait, B.T., 1990. Rapid, sensitive analysis of protein mixtures by mass-spectrometry. Proceedings of the National Academy of Sciences of the United States of America 87 (17), 6873–6877. Harvey, D.J., 2005. Structural determination of N-linked glycans by matrix-assisted laser desorption/ionization and electrospray ionization mass spectrometry. Proteomics 5 (7), 1774–1786. Kang, P., et al., 2005. Solid-phase permethylation of glycans for mass spectrometric analysis. Rapid Communications in Mass Spectrometry 19 (23), 3421–3428. Ashline, D.J., et al., 2007. Carbohydrate structural isomers analyzed by sequential mass spectrometry. Analytical Chemistry 79 (10), 3830–3842. Harvey, D.J., 2005. Fragmentation of negative ions from carbohydrates: Part 3. Fragmentation of hybrid and complex N-linked glycans. Journal of the American Society for Mass Spectrometry 16 (5), 647–659. Harvey, D.J., 2005. Fragmentation of negative ions from carbohydrates: Part 2. Fragmentation of high-mannose N-linked glycans. Journal of the American Society for Mass Spectrometry 16 (5), 631–646. Harvey, D.J., 2005. Fragmentation of negative ions from carbohydrates: Part 1. Use of nitrate and other anionic adducts for the production of negative ion electrospray spectra from N-linked carbohydrates. Journal of the American Society for Mass Spectrometry 16 (5), 622–630. Harvey, D.J., et al., 2016. Travelling-wave ion mobility and negative ion fragmentation of high-mannose N-glycans. J. Mass Spectrom. 51 (3), 219–235. Hinneburg, H., et al., 2016. Distinguishing N-acetylneuraminic acid linkage isomers on glycopeptides by ion mobility-mass spectrometry. Chem. Commun. (Camb) 52 (23), 4381–4384. Hofmann, J., et al., 2014. Estimating collision cross sections of negatively charged N-glycans using traveling wave ion mobility-mass spectrometry. Anal. Chem. 86 (21), 10789–10795. Pagel, K., Harvey, D.J., 2013. Ion mobility-mass spectrometry of complex carbohydrates: Collision cross sections of sodiated N-linked glycans. Anal. Chem. 85 (10), 5138–5145. Struwe, W.B., et al., 2015. Collision cross sections of high-mannose N-glycans in commonly observed adduct states – Identification of gas-phase conformers unique to [M– H] ions. Analyst 140 (20), 6799–6803. Stockmann, H., et al., 2015. Ultrahigh throughput, ultrafiltration-based n-glycomics platform for ultraperformance liquid chromatography (ULTRA(3)). Anal. Chem. 87 (16), 8316–8322. Packer, N.H., et al., 2008. Frontiers in glycomics: Bioinformatics and biomarkers in disease – An NIH White Paper prepared from discussions by the focus groups at a workshop on the NIH campus, Bethesda MD (September 11–13, 2006). Proteomics 8 (1), 8–20. Raman, R., et al., 2006. Advancing glycomics: Implementation strategies at the consortium for functional glycomics. Glycobiology 16 (5), 82R–90R. Ranzinger, R., et al., 2008. GlycomeDB – Integration of open-access carbohydrate structure databases. BMC Bioinformatics 9, 384. Akune, Y., et al., 2010. The RINGS resource for glycome informatics analysis and data mining on the Web. OMICS 14 (4), 475–486. von der Lieth, C.W., et al., 2011. EUROCarbDB: An open-access platform for glycoinformatics. Glycobiology 21 (4), 493–502. Ranzinger, R., et al., 2015. GlycoRDF: An ontology to standardize glycomics data in RDF. Bioinformatics 31 (6), 919–925. Hashimoto, K., et al., 2006. KEGG as a glycome informatics resource. Glycobiology 16 (5), 63R–70R. Lutteke, T., et al., 2006. GLYCOSCIENCES.de: An Internet portal to support glycomics and glycobiology research. Glycobiology 16 (5), 71R–81R. Lombard, V., et al., 2013. The carbohydrate-active enzymes database (CAZy) in 2013. Nucleic Acids Res. 42, 490–495. Egorova, K.S., Toukach, P.V., 2014. Expansion of coverage of Carbohydrate Structure Database (CSDB). Carbohydr. Res. 389, 112–114. Campbell, M.P., et al., 2014. UniCarbKB: Building a knowledge platform for glycoproteomics. Nucleic Acids Res. 42 (1), D215–D221. Cooper, C.A., et al., 2001. GlycoSuiteDB: A new curated relational database of glycoprotein glycan structures and their biological sources. Nucleic Acids Res. 29 (1), 332–335. Aoki-Kinoshita, K., et al., 2016. GlyTouCan 1.0 – The international glycan structure repository. Nucleic Acids Res. 44 (D1), D1237–D1242. Herget, S., et al., 2008. GlycoCT – A unifying sequence format for carbohydrates. Carbohydrate Research 343 (12), 2162–2171. Tanaka, K., et al., 2014. WURCS: The Web3 unique representation of carbohydrate structures. J. Chem. Inf. Model 54 (6), 1558–1566. Sahoo, S.S., et al., 2005. GLYDE – An expressive XML standard for the representation of glycan structure. Carbohydr. Res. 340 (18), 2802–2807. Ceroni, A., Dell, A., Haslam, S.M., 2007. The GlycanBuilder: A fast, intuitive and flexible software tool for building and displaying glycan structures. Source Code Biol. Med. 2, 3. Damerell, D., et al., 2012. The GlycanBuilder and GlycoWorkbench glycoinformatics tools: Updates and new developments. Biol. Chem. 393 (11), 1357–1362. Campbell, M.P., et al., 2008. GlycoBase and autoGU: Tools for HPLC-based glycan analysis. Bioinformatics 24 (9), 1214–1216. Gotz, L., et al., 2014. GlycoDigest: A tool for the targeted use of exoglycosidase digestions in glycan structure determination. Bioinformatics 30 (21), 3131–3133. Ceroni, A., et al., 2008. GlycoWorkbench: A tool for the computer-assisted annotation of mass spectra of Glycans. Journal of Proteome Research 7 (4), 1650–1659. Hayes, C.A., et al., 2011. UniCarb-DB: A database resource for glycomic discovery. Bioinformatics 27 (9), 1343–1344. Struwe, W.B., et al., 2015. GlycoMob: An ion mobility-mass spectrometry collision cross section database for glycomics. Glycoconj. J. Aoki-Kinoshita, K., et al., 2013. Introducing glycomics data into the Semantic Web. J. Biomed. Semantics 4 (1), 39. Hossler, P., Khattak, S.F., Li, Z.J., 2009. Optimal and consistent protein glycosylation in mammalian cell culture. Glycobiology 19 (9), 936–949.

432 [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] [107] [108] [109] [110] [111] [112] [113] [114] [115] [116] [117] [118] [119] [120] [121] [122] [123] [124]

Glycomics Jefferis, R., 2009. Recombinant antibody therapeutics: The impact of glycosylation on mechanisms of action. Trends in Pharmacological Sciences 30 (7), 356–362. Li, H.J., d’Anjou, M., 2009. Pharmacological significance of glycosylation in therapeutic proteins. Current Opinion in Biotechnology 20 (6), 678–684. Walsh, G., Jefferis, R., 2006. Post-translational modifications in the context of therapeutic proteins. Nature Biotechnology 24 (10), 1241–1252. Carlson, R., 2016. Estimating the biotech sector’s contribution to the US economy. Nat. Biotechnol. 34 (3), 247–255. Aggarwal, S., 2009. What’s fueling the biotech engine – 2007. Nat. Biotechnol. 27 (2), 205 (2008, vol. 26, p. 1227). Sairam, M.R., Krishnamurthy, H., 2001. The role of follicle-stimulating hormone in spermatogenesis: Lessons from knockout animal models. Archives of Medical Research 32 (6), 601–608. Simoni, M., et al., 1999. Role of FSH in male gonadal function. Annales D Endocrinologie 60 (2), 102–106. Robker, R.L., Richards, J.S., 1998. Hormone-induced proliferation and differentiation of granulosa cells: A coordinated balance of the cell cycle regulators cyclin D2 and p27(Kip1). Molecular Endocrinology 12 (7), 924–940. Waterstone, J., Parsons, J., Bolton, V., 1991. Elective transfer of two embryos. Lancet 337 (8747), 975–976. Giudice, E., et al., 1994. Composition of commercial gonadotrophin preparations extracted from human post-menopausal urine: Characterization of non-gonadotrophin proteins. Hum. Reprod. 9 (12), 2291–2299. Ulloa-Aguirre, A., Timossi, C., 1998. Structure–function relationship of follicle-stimulating hormone and its receptor. Hum. Reprod. Update 4 (3), 260–283. Bishop, L.A., et al., 1994. Specific roles for the asparagine-linked carbohydrate residues of recombinant human follicle stimulating hormone in receptor binding and signal transduction. Mol. Endocrinol. 8 (6), 722–731. Creus, S., et al., 2001. Human FSH isoforms: Carbohydrate complexity as determinant of in-vitro bioactivity. Molecular and Cellular Endocrinology 174 (1–2), 41–49. D’Antonio, M., et al., 1999. Biological characterization of recombinant human follicle stimulating hormone isoforms. Human Reproduction 14 (5), 1160–1167. Miller, C., et al., 1983. Pituitary follicle-stimulating-hormone heterogeneity – Assessment of biologic activities of each follicle-stimulating-hormone form. Fertility and Sterility 40 (2), 242–247. Brines, M., et al., 2008. Nonerythropoietic, tissue-protective peptides derived from the tertiary structure of erythropoietin. Proceedings of the National Academy of Sciences of the United States of America 105 (31), 10925–10930. Fantacci, M., et al., 2006. Carbamylated erythropoietin ameliorates the metabolic stress induced in vivo by severe chronic hypoxia. Proceedings of the National Academy of Sciences of the United States of America 103 (46), 17531–17536. Leist, M., et al., 2004. Derivatives of erythropoietin that are tissue protective but not erythropoietic. Science 305 (5681), 239–242. Mennini, T., et al., 2006. Nonhematopoietic erythropoietin derivatives prevent motoneuron degeneration in vitro and in vivo. Molecular Medicine 12 (7–8), 153–160. Delorme, E., et al., 1992. Role of glycosylation on the secretion and biological activity of erythropoietin. Biochemistry 31 (41), 9871–9876. Elliott, S., et al., 2004. Structural requirements for additional N-linked carbohydrate on recombinant human erythropoietin. J. Biol. Chem. 279 (16), 16854–16862. Elliott, S., et al., 2004. Control of rHuEPO biological activity: The role of carbohydrate. Exp. Hematol. 32 (12), 1146–1155. Fukuda, M.N., et al., 1989. Survival of recombinant erythropoietin in the circulation: The role of carbohydrates. Blood 73 (1), 84–89. Erbayraktar, S., et al., 2003. Asialoerythropoietin is a nonerythropoietic cytokine with broad neuroprotective activity in vivo. Proceedings of the National Academy of Sciences of the United States of America 100 (11), 6741–6746. Chiba, Y., Akeboshi, H., 2009. Glycan engineering and production of ‘humanized’ glycoprotein in yeast cells. Biological & Pharmaceutical Bulletin 32 (5), 786–795. Pastores, G.M., Meere, P.A., 2005. Musculoskeletal complications associated with lysosomal storage disorders: Gaucher disease and Hurler–Scheie syndrome (mucopolysaccharidosis type I). Current Opinion in Rheumatology 17 (1), 70–78. Pastores, G.M., et al., 2004. Therapeutic goals in the treatment of Gaucher disease. Seminars in Hematology 41 (4), 4–14. Elstein, D., Zimran, A., 2009. Review of the safety and efficacy of imiglucerase treatment of Gaucher disease. Biologics 3, 407–417. Schiffmann, R., et al., 2000. Infusion of alpha-galactosidase A reduces tissue globotriaosylceramide storage in patients with Fabry disease. Proc. Natl. Acad. Sci. USA 97 (1), 365–370. Van den Hout, J.M., et al., 2004. Long-term intravenous treatment of Pompe disease with recombinant human alpha-glucosidase from milk. Pediatrics 113 (5), e448–e457. Pastores, G.M., 2008. Laronidase (Aldurazyme): Enzyme replacement therapy for mucopolysaccharidosis type I. Expert Opin. Biol. Ther. 8 (7), 1003–1009. Clarke, L.A., 2008. Idursulfase for the treatment of mucopolysaccharidosis II. Expert Opin. Pharmacother. 9 (2), 311–317. Hopwood, J.J., Bate, G., Kirkpatrick, P., 2006. Galsulfase. Nat. Rev. Drug Discov. 5 (2), 101–102. Brady, R.O., et al., 1967. Enzymatic defect in Fabry’s disease. Ceramidetrihexosidase deficiency. N. Engl. J. Med. 276 (21), 1163–1167. Chiba, Y., et al., 2002. Production in yeast of alpha-galactosidase A, a lysosomal enzyme applicable to enzyme replacement therapy for Fabry disease. Glycobiology 12 (12), 821–828. Odani, T., et al., 1997. Mannosylphosphate transfer to cell wall mannan is regulated by the transcriptional level of the MNN4 gene in Saccharomyces cerevisiae. FEBS Letters 420 (2–3), 186–190. Gray, P.W., Goeddel, D.V., 1982. Structure of the human immune interferon gene. Nature 298 (5877), 859–863. Schoenborn, J.R., Wilson, C.B., 2007. Regulation of interferon-gamma during innate and adaptive immune responses. Adv. Immunol. 96, 41–101. Rinderknecht, E., O’Connor, B.H., Rodriguez, H., 1984. Natural human interferon-gamma. Complete amino acid sequence and determination of sites of glycosylation. J. Biol. Chem. 259 (11), 6790–6797. Mutsaers, J.H., et al., 1986. Structural studies of the carbohydrate chains of human gamma-interferon. Eur. J. Biochem. 156 (3), 651–654. Yamamoto, S., et al., 1989. Structures of the sugar chains of interferon-gamma produced by human myelomonocyte cell line HBL-38. J. Biochem. 105 (4), 547–555. Yamamoto, S., et al., 1989. Studies on the sugar chains of interferon-gamma from human peripheral-blood lymphocytes. J. Biochem. 105 (6), 1034–1039. Sareneva, T., et al., 1994. Role of N-glycosylation in the synthesis, dimerization and secretion of human interferon-gamma. Biochem. J. 303 (Pt 3), 831–840. Sareneva, T., et al., 1995. N-glycosylation of human interferon-gamma – Glycans at Asn-25 are critical for protease resistance. Biochemical Journal 308, 9–14. Sareneva, T., et al., 1993. Effect of carbohydrates on the pharmacokinetics of human interferon-gamma. J. Interferon Res. 13 (4), 267–269. Lamotte, D., et al., 1999. Na-butyrate increases the production and alpha2,6-sialylation of recombinant interferon-gamma expressed by alpha2,6-sialyltransferase engineered CHO cells. Cytotechnology 29 (1), 55–64. Aggarwal, R.S., 2014. What’s fueling the biotech engine – 2012 to 2013. Nat. Biotechnol. 32 (1), 32–39. Bruhns, P., et al., 2009. Specificity and affinity of human Fcgamma receptors and their polymorphic variants for human IgG subclasses. Blood 113 (16), 3716–3725. Nezlin, R., Ghetie, V., 2004. Interactions of immunoglobulins outside the antigen-combining site. Adv. Immunol. 82, 155–215. Reusch, D., Tejada, M.L., 2015. Fc glycans of therapeutic antibodies as critical quality attributes. Glycobiology 25 (12), 1325–1334. Roopenian, D.C., Akilesh, S., 2007. FcRn: The neonatal Fc receptor comes of age. Nat. Rev. Immunol. 7 (9), 715–725. Goebl, N.A., et al., 2008. Neonatal Fc receptor mediates internalization of Fc in transfected human endothelial cells. Mol. Biol. Cell 19 (12), 5490–5505. Hobbs, S.M., Jackson, L.E., Hoadley, J., 1992. Interaction of aglycosyl immunoglobulins with the IgG Fc transport receptor from neonatal rat gut: Comparison of deglycosylation by tunicamycin treatment and genetic engineering. Mol. Immunol. 29 (7–8), 949–956. Tao, M.H., Morrison, S.L., 1989. Studies of aglycosylated chimeric mouse–human IgG. Role of carbohydrate in the structure and effector functions mediated by the human IgG constant region. J. Immunol. 143 (8), 2595–2601. Mimura, Y., et al., 2007. Contrasting glycosylation profiles between Fab and Fc of a human IgG protein studied by electrospray ionization mass spectrometry. Journal of Immunological Methods 326 (1–2), 116–126. Arnold, J.N., et al., 2006. Mannan binding lectin and its interaction with immunoglobulins in health and in disease. Immunology Letters 106 (2), 103–110.

Glycomics

433

[125] Malhotra, R., et al., 1995. Glycosylation Changes of IgG associated with rheumatoid-arthritis can activate complement via the mannose-binding protein. Nat. Med. 1 (6), 599 (1995, vol. 1, p. 237). [126] Rademacher, T.W., 1993. Glycosylation as a factor affecting product consistency. Biologicals 21 (2), 103–104. [127] Kaneko, Y., Nimmerjahn, F., Ravetch, E.V., 2006. Anti-inflammatory activity of immunoglobulin G resulting from Fc sialylation. Science 313 (5787), 670–673. [128] Anthony, R.M., et al., 2008. Identification of a receptor required for the anti-inflammatory activity of IVIG. Proceedings of the National Academy of Sciences of the United States of America 105 (50), 19571–19578. [129] Jones, A.J.S., et al., 2007. Selective clearance of glycoforms of a complex glycoprotein pharmaceutical caused by terminal N-acetylglucosamine is similar in humans and cynomolgus monkeys. Glycobiology 17 (5), 529–540. [130] Bragonzi, A., et al., 2000. A new Chinese hamster ovary cell line expressing alpha2,6-sialyltransferase used as universal host for the production of human-like sialylated recombinant glycoproteins. Biochim. Biophys. Acta 1474 (3), 273–282. [131] Hamilton, S.R., et al., 2006. Humanization of yeast to produce complex terminally sialylated glycoproteins. Science 313 (5792), 1441–1443. [132] Walsh, G., 2006. Biopharmaceutical benchmarks 2006. Nat. Biotechnol. 24 (7), 769–776. [133] Okazaki, A., et al., 2004. Fucose depletion from human IgG1 oligosaccharide enhances binding enthalpy and association rate between IgG1 and Fc gamma RIIIa. Journal of Molecular Biology 336 (5), 1239–1249. [134] Ferrara, C., et al., 2006. The carbohydrate at Fc gamma RIIIa Asn-162 – An element required for high affinity binding to non-fucosylated IgG glycoforms. Journal of Biological Chemistry 281 (8), 5032–5036. [135] Shibata-Koyama, M., et al., 2009. The N-linked oligosaccharide at Fc gamma RIIIa Asn-45: An inhibitory element for high Fc gamma RIIIa binding affinity to IgG glycoforms lacking core fucosylation. Glycobiology 19 (2), 126–134. [136] Yamane-Ohnuki, N., et al., 2004. Establishment of FUT8 knockout Chinese hamster ovary cells: An ideal host cell line for producing completely defucosylated antibodies with enhanced antibody-dependent cellular cytotoxicity. Biotechnology and Bioengineering 87 (5), 614–622. [137] Schachter, H., 1986. Biosynthetic controls that determine the branching and microheterogeneity of protein-bound oligosaccharides. Biochemistry and Cell Biology – Biochimie et Biologie Cellulaire 64 (3), 163–181. [138] Bergwerff, A.A., et al., 1995. Variation in N-linked carbohydrate chains in different batches of two chimeric monoclonal IgG1 antibodies produced by different murine SP2/ 0 transfectoma cell subclones. Glycoconj. J. 12 (3), 318–330. [139] Lifely, M.R., et al., 1995. Glycosylation and biological activity of CAMPATH-1H expressed in different cell lines and grown under different culture conditions. Glycobiology 5 (8), 813–822. [140] Umana, P., et al., 1999. Engineered glycoforms of an antineuroblastoma IgG1 with optimized antibody-dependent cellular cytotoxic activity. Nat. Biotechnol. 17 (2), 176–180. [141] Ferrara, C., et al., 2006. Modulation of therapeutic antibody effector functions by glycosylation engineering: Influence of Golgi enzyme localization domain and co-expression of heterologous beta1, 4-N-acetylglucosaminyltransferase III and Golgi alpha-mannosidase II. Biotechnol. Bioeng. 93 (5), 851–861. [142] Masarik, M., et al., 2003. Application of avidin–biotin technology and adsorptive transfer stripping square-wave voltammetry for detection of DNA hybridization and avidin in transgenic avidin maize. Anal. Chem. 75 (11), 2663–2669. [143] Kusnadi, A.R., et al., 1998. Production and purification of two recombinant proteins from transgenic corn. Biotechnol. Prog. 14 (1), 149–155. [144] Parmenter, D.L., et al., 1995. Production of biologically active hirudin in plant seeds using oleosin partitioning. Plant Molecular Biology 29 (6), 1167–1180. [145] Bardor, M., et al., 2003. Immunoreactivity in mammals of two typical plant glyco-epitopes, core alpha(1,3)-fucose and core xylose. Glycobiology 13 (6), 427–434. [146] Bencurova, M., et al., 2004. Specificity of IgG and IgE antibodies against plant and insect glycoprotein glycans determined with artificial glycoforms of human transferrin. Glycobiology 14 (5), 457–466. [147] Rayon, C., et al., 1999. Characterization of N-glycans from Arabidopsis. Application to a fucose-deficient mutant. Plant Physiol. 119 (2), 725–734. [148] Frey, A.D., Karg, S.R., Kallio, P.T., 2009. Expression of rat beta(1,4)-N-acetylglucosaminyltransferase III in Nicotiana tabacum remodels the plant-specific N-glycosylation. Plant Biotechnol. J. 7 (1), 33–48. [149] Koprivova, A., et al., 2004. Targeted knockouts of Physcomitrella lacking plant-specific immunogenic N-glycans. Plant Biotechnol. J. 2 (6), 517–523. [150] Weise, A., et al., 2007. High-level expression of secreted complex glycosylated recombinant human erythropoietin in the Physcomitrella Delta-fuc-t Delta-xyl-t mutant. Plant Biotechnol. J. 5 (3), 389–401. [151] Lerouge, P., et al., 1998. N-glycoprotein biosynthesis in plants: Recent developments and future trends. Plant Mol. Biol. 38 (1–2), 31–48. [152] Altmann, F., et al., 1995. Insect cells contain an unusual, membrane-bound beta-N-acetylglucosaminidase probably involved in the processing of protein N-glycans. J. Biol. Chem. 270 (29), 17344–17349. [153] Wagner, R., et al., 1996. Elongation of the N-glycans of fowl plague virus hemagglutinin expressed in Spodoptera frugiperda (Sf9) cells by coexpression of human beta 1,2-N-acetylglucosaminyltransferase I. Glycobiology 6 (2), 165–175. [154] Hollister, J.R., Jarvis, D.L., 2001. Engineering lepidopteran insect cells for sialoglycoprotein production by genetic transformation with mammalian beta 1,4galactosyltransferase and alpha 2,6-sialyltransferase genes. Glycobiology 11 (1), 1–9. [155] Hollister, J., Conradt, H., Jarvis, D.L., 2003. Evidence for a sialic acid salvaging pathway in lepidopteran insect cells. Glycobiology 13 (6), 487–495. [156] Lawrence, S.M., et al., 2001. Cloning and expression of human sialic acid pathway genes to generate CMP-sialic acids in insect cells. Glycoconj. J. 18 (3), 205–213. [157] Aumiller, J.J., Hollister, J.R., Jarvis, D.L., 2003. A transgenic insect cell line engineered to produce CMP-sialic acid and sialylated glycoproteins. Glycobiology 13 (6), 497–507. [158] Herscovics, A., Orlean, P., 1993. Glycoprotein biosynthesis in yeast. FASEB J. 7 (6), 540–550. [159] Chiba, Y., et al., 1998. Production of human compatible high mannose-type (Man5GlcNAc2) sugar chains in Saccharomyces cerevisiae. J. Biol. Chem. 273 (41), 26298–26304. [160] Herscovics, A., 2001. Structure and function of Class I alpha 1,2-mannosidases involved in glycoprotein synthesis and endoplasmic reticulum quality control. Biochimie 83 (8), 757–762. [161] Choi, B.K., et al., 2003. Use of combinatorial genetic libraries to humanize N-linked glycosylation in the yeast Pichia pastoris. Proceedings of the National Academy of Sciences of the United States of America 100 (9), 5022–5027. [162] Schwientek, T., Narimatsu, H., Ernst, J.F., 1996. Golgi localization and in vivo activity of a mammalian glycosyltransferase (human beta 1,4-galactosyltransferase) in yeast. Journal of Biological Chemistry 271 (7), 3398–3405. [163] Vervecken, W., et al., 2004. In vivo synthesis of mammalian-like, hybrid-type N-glycans in Pichia pastoris. Appl. Environ. Microbiol. 70 (5), 2639–2646. [164] Bobrowicz, P., et al., 2004. Engineering of an artificial glycosylation pathway blocked in core oligosaccharide assembly in the yeast Pichia pastoris: Production of complex humanized glycoproteins with terminal galactose. Glycobiology 14 (9), 757–766. [165] Figueroa, B., et al., 2007. Enhanced cell culture performance using inducible anti-apoptotic genes E1B-19K and Aven in the production of a monoclonal antibody with chinese hamster ovary cells. Biotechnology and Bioengineering 97 (4), 877–892. [166] Lee, Y.Y., et al., 2009. Overexpression of heat shock proteins (HSPs) in CHO cells for extended culture viability and improved recombinant protein production. Journal of Biotechnology 143 (1), 34–43. [167] Peng, R.W., Fussenegger, M., 2009. Molecular engineering of exocytic vesicle traffic enhances the productivity of Chinese hamster ovary cells. Biotechnol. Bioeng. 102 (4), 1170–1181. [168] Florin, L., et al., 2009. Heterologous expression of the lipid transfer protein CERT increases therapeutic protein productivity of mammalian cells. J. Biotechnol. 141 (1–2), 84–90.

434

Glycomics

[169] Bork, K., Horstkorte, R., Weidemann, W., 2009. Increasing the sialylation of therapeutic glycoproteins: The potential of the sialic acid biosynthetic pathway. J. Pharm. Sci. 98 (10), 3499–3508. [170] Jeong, Y.T., et al., 2009. Enhanced sialylation of recombinant erythropoietin in genetically engineered Chinese-hamster ovary cells. Biotechnol. Appl. Biochem. 52 (Pt 4), 283–291. [171] Butler, M., 2005. Animal cell cultures: Recent achievements and perspectives in the production of biopharmaceuticals. Appl. Microbiol. Biotechnol. 68 (3), 283–291. [172] Chung, C.H., et al., 2008. Cetuximab-induced anaphylaxis and IgE specific for galactose-alpha-1,3-galactose. N. Engl. J. Med. 358 (11), 1109–1117. [173] Muchmore, E.A., et al., 1989. Biosynthesis of N-glycolyneuraminic acid. The primary site of hydroxylation of N-acetylneuraminic acid is the cytosolic sugar nucleotide pool. J. Biol. Chem. 264 (34), 20216–20223. [174] Baker, K.N., et al., 2001. Metabolic control of recombinant protein N-glycan processing in NS0 and CHO cells. Biotechnol. Bioeng. 73 (3), 188–202. [175] Zhu, A., Hurst, R., 2002. Anti-N-glycolylneuraminic acid antibodies identified in healthy human serum. Xenotransplantation 9 (6), 376–381. [176] Flesher, A.R., et al., 1995. Fluorophore-labeled carbohydrate analysis of immunoglobulin fusion proteins: Correlation of oligosaccharide content with in vivo clearance profile. Biotechnol. Bioeng. 46 (5), 399–407. [177] Covic, A., et al., 2008. Biosimilars and biopharmaceuticals: What the nephrologists need to know – A position paper by the ERA-EDTA Council. Nephrology Dialysis Transplantation 23 (12), 3731–3737.

Relevant Websites www.ccrc.uga.edu – CFG, Complex Carbohydrate Research Center. www.eurocarb.org – EuroCarbDB. www.emea.europa.eu – European Medicines Agency. http://glycobase.nibrt.ie – GlycoBase 3.2 – National Institute for Bioprocessing Research and Training. www.hgpi.jp – Human Disease Glycomics/Proteome Initiative. www.ich.org – International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use. www.fda.gov – U.S. Food and Drug Administration.

1.31 Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics U Roessner and A Nahid, The University of Melbourne, Melbourne, VIC, Australia B Chapman, A Hunter, and M Bellgard, Murdoch University, Murdoch, WA, Australia © 2011 Elsevier B.V. All rights reserved. This is a reprint of U. Roessner, A. Nahid, B. Chapman, A. Hunter, M. Bellgard, 1.33 - Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 447–459.

1.31.1 Introduction 1.31.2 Technologies Used to Measure Metabolites 1.31.2.1 GS Coupled to MS 1.31.2.2 Liquid Chromatography Coupled to Mass Spectrometry 1.31.2.3 Nuclear Magnetic Resonance 1.31.3 Metabolomics Approaches 1.31.4 Bioinformatics: What Can It Do 1.31.5 What Does the Informatician Need to Analyze the High-Density Data? 1.31.6 Data Preprocessing: From Raw to Sense 1.31.6.1 Normalization and Data Transformation 1.31.6.1.1 Z-Scores 1.31.6.1.2 Median 1.31.6.1.3 Sum of All or Some of the Metabolites 1.31.6.1.4 Internal and External Analytical Standards 1.31.7 Requirements and Problems of Statistical and Multivariant Analysis of Metabolomics Data 1.31.7.1 Univariate Analysis 1.31.7.2 Multivariate Data Analysis 1.31.7.2.1 Principal Component Analysis 1.31.7.2.2 Linear Discriminant Analysis 1.31.7.2.3 Hierarchical Cluster Analysis 1.31.7.2.4 Partial-Least Squares Discriminant Analysis 1.31.7.3 Treatment of Missing Values 1.31.8 Conclusions Acknowledgments References

436 436 436 437 438 438 438 439 440 441 443 443 444 444 444 444 445 445 445 445 445 446 446 447 447

Glossary Bioinformatics The application of information technology, computer sciences, software development, and statistical analysis to a particular biological problem. Especially omics-based sciences such as genomics, proteomics, and metabolomics require sophisticated bioinformatics tool to aid analysis and interpretation of the large data sets produced by the technologies in a biological context. Databases Software programs that store related information in an organized and regular structure. They allow the information to be retrieved and manipulated. Mass spectrometry An analysis tool that measures ionized forms of molecules and atoms. It records the mass to charge ratio of those ions that are related to the molecular weight of the ion. Mass spectrometry also allows breaking molecules into fragments, enabling the elucidation of the structure of the molecule under analyses. Metabolomics The science of a comprehensive analysis of small molecules, the metabolites, in a biological system. Metabolomics used techniques from analytical chemistry aiming to measure as many metabolites as possible linking the resulting data to biochemistry, biology, and physiology. Statistics A specialized area of applied mathematics that uses probability theories to estimate and predict parameters and patterns in populations of quantitative data. Depending on the problem in question and how what type of data are collected, it is important to understand which statistical methodologies to apply appropriately.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00027-6

435

436

1.31.1

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

Introduction

Metabolomics is a rapidly emerging field that combines the approach of analytical biochemistry, to identify and quantify metabolites in biological systems in a high-throughput manner, with the application of sophisticated informatics and bioinformatics tools for data extraction, analysis, and statistics to support biological interpretation. In today’s biological sciences, the utilization of “omics”-type analyses of biological molecules including genes (genomics), transcripts (transcriptomics), proteins (proteomics), and metabolites (metabolomics) are important for the comprehensive investigation of a particular biological problem and system. These approaches enhance our understanding of the biology of an organism and its response to environmental stimuli or genetic perturbation. Unification of the information generated from these “omics” studies via a systems biology approach requires the application of bioinformatics strategies in order to extract the information for a biological interpretation from the high-density data sets. Metabolites are required as precursors for the building blocks and energy carrier molecules that are needed for the growth and maintenance of living cells; therefore, each metabolite is synthesized in order to fulfill a finite and specific biological function. They undergo chemical reactions that are carried out by specialized proteins called enzymes. During these reactions, the chemical properties and often energy content of metabolites are changed. A series of these chemical reactions is called a pathway, and the connection of these pathways, metabolism. Metabolites have been described as the interface between genetic architecture and the environment,1 providing a direct description of the physiological state of an organism.2 The development of metabolomics as a scientific field has been proved to be most exciting and also challenging. The inherent challenges derive from the fact that each metabolite is characterized by a unique chemical structure, which determines its physical and chemical properties. These differences in chemical and physical features of a metabolite, including molecular weight, molecular size, polarity, stability, volatility, solubility, and many more, determine the means by which the compound is to be extracted, separated from other biomolecules, detected, and quantified (for more details see Ref. 3). If the aim of an accurate and comprehensive measurement of metabolites is desired, a number of different extraction procedures, as well analytical methodologies and instrumentation, must be used for a metabolomics approach. A number of analytical technologies have been utilized and successfully employed to analyze metabolites in many different organisms, tissues, and fluids (for more details see Ref. 4). Major tools for the simultaneous analysis of many metabolites include chromatographic separation techniques coupled to mass spectrometry (MS) and nuclear magnetic resonance (NMR) spectroscopy. A more detailed description of these technologies is provided below. Since instrumentation has become increasingly more sensitive and reliable, thereby allowing the analysis of a huge number of metabolites simultaneously, a greater current challenge is the extraction of information from the analytical analyses, the handling of the resultant raw data, and interpretation of this vast amount of information in a biological context. Here, a strong collaboration between analytics and bioinformatics is required to apply sophisticated statistical and multivariant analyses tools. The tools required may be adapted from other existing life science fields, such as astronomy or physics, where high-density data sets occur, or new algorithms will need to be developed and validated. Currently, a number of routine methods are applied within the metabolomics community including classification techniques, comparative overlays of data sets, network constructions based on correlation analysis, or simple representation of the data in heatmaps or pathway mappings. As already mentioned, analytical instrumentation has become more and more sensitive, correspondingly much more background noise is now detected from contamination, solvents, and such others. The question of how we are able to distinguish between noise and real sample-related information has developed into a great challenge and has proved to be very important to our data interpretations. Here, we offer a playground for bioinformaticians to develop new algorithms that may allow the identification of background noise in raw data; for instance, in mass spectrometry, the aim would be to extract only sample-related signals to be further investigated. In summary, metabolomics is a newly developed biological science field which has presented a great deal of excitement and challenges in the past decade. Together with other complimentary “omics” technologies, a way is now opened, allowing research to analyze a huge number of cell components simultaneously, highly improving our understanding of the organism of interest.

1.31.2

Technologies Used to Measure Metabolites

There are a number of analytical platforms used to analyze small biological molecules including MS and NMR. When MS is applied, often metabolites in complex mixtures need to be separated prior to analysis. This can be achieved by gas chromatography (GC), liquid chromatography (LC), or capillary electrophoresis (CE). A detailed description of these technologies and their advantages and disadvantages are presented in.4 Here, we only summarize the major tools currently used in the metabolomics community: GC–MS, liquid chromatography coupled to mass spectrometry (LC–MS), and NMR.

1.31.2.1

GS Coupled to MS

Gas chromatography coupled to mass spectrometry (GC–MS) has been the well-accepted working horse in the metabolomics field due to its great separation power and reproducibility. Additionally, the technology is well advanced with regard to routine applications to a wide range of tissues and fluids and also combines with well-established mass spectral libraries for compound identifications.5,6 GC– MS enables the analysis of up to 400 compounds including sugars, amino acids, organic acids, fatty acids, sterols, and amines in a single run of which about 100–150 are known with respect to their chemical nature.7 However, GC–MS is limited to only those compounds that are amenable for analysis and which are either volatile or can be made volatile using chemical derivatisation. Modification by

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

437

chemical derivatization alters the chemical nature of the molecule substantially and often makes identification based on mass spectra very difficult. Current GC–MS mass spectral libraries require the analysis of authentic chemical standards to record mass spectra as well as retention time in a standardized method for unambiguous identification of detected metabolites. Once a sample of volatile compounds (either naturally or through derivatisation) is injected into the instrument, the compounds are separated based on their volatility temperature and the interaction with the column stationary phase (e.g., polarity). Once eluted off the column into an ion source, most commonly, an electron impact ionization (EI) source, they are ionized to enter the MS. Different mass analyzers have been used including low-resolution quadrupole instruments; however, fast-scanning time-offlight (TOF) MS is becoming more accepted. EI is a relatively harsh ionization process in which electron energy is transferred to the molecule, resulting in highly reproducible fragmentations. Resultant fragments are then analyzed by their mass to charge ratio (m/z) and their abundance detected in the detector. The fragmentation of molecules is highly dependent on the chemical structure and therefore a particular fragmentation pattern, which is then called mass spectrum is observed and can be utilized for the construction of GC-MS-based mass spectral libraries (e.g., National Institute for Standard Technology (NIST) (http://www.nist.gov/srd/nist1a.htm), Golm Metabolome Database (GMD) (http://csbdb.mpimp-golm.mpg.de/csbdb/gmd/msri/gmd_msri.html)). If retention time is also recorded for each compound, unambiguous peak identification across instruments, biological sample, and laboratories can be achieved. By analysis of authentic standards, a huge number of mass spectra, sometimes together with a retention time, are now available. However, about 60% of metabolites detectable with GC-MS remain unknown with respect to chemical nature. The problem is that most natural compounds are not commercially available and therefore there is no possibility to obtain their authentic mass spectra and retention time for peak identifications. Another type of ionization available with GC-MS is chemical ionization (CI), which uses gases (methane or ammonia), providing a much softer fragmentation of the molecules. This may have the advantage of detecting the molecular ion and its isotopic pattern to determine molecular weight of the molecule; however, it does not provide a typical fragmentation pattern to aid peak identification. This technique is often applied to determine stable isotope distribution in molecules for flux analysis following a stable isotope labeled feeding experiment.8 Because GC-MS only allows the analysis of thermally stable volatile or compounds that can be made volatile through derivatization, only low-molecular-weight compounds of up to around 1000 Da can be detected. Therefore, comprehensive techniques such as LC-MS or NMR have to be used to cover a greater diversity of metabolites.

1.31.2.2

Liquid Chromatography Coupled to Mass Spectrometry

LC–MS technology for untargeted and quantitative analyses has been widely used in proteomics applications for a long time and has gained considerable attention within the metabolomics community as well. In contrast to GC–MS, for analysis by LC–MS compounds do not require derivatization to become amendable for measurement, so no chemical interference with the molecule is needed. Advantageously, higher molecular mass or lower thermostability metabolites can be analyzed with LC-MS. In order to couple the LC to a MS, the liquid elutents of the column need to be converted into the gas phase concurrent with simultaneous ionization of the compounds to be analyzed in the MS. The most commonly used ionization techniques include electrospray ionization (ESI) and atmospheric pressure ionization (API). Most important for any LC–MS-based metabolomics analysis is the right choice of the appropriate column matrix for separation (e.g., ion-exchange, reversed-phase, and hydrophobic interaction chromatography) and suitable elution procedures for the compound, compound classes, or metabolome of interest. No single methodology is able to separate, detect, and quantify the range of chemical diversity metabolites represented. This becomes especially important when the aim of the experiment is to analyze as many metabolites as possible. To be most comprehensive, several approaches, all based on LC–MS, have to be followed. The initial separation of polar and apolar compounds can be reached through extraction (bi-phasic, e.g., methanol, water, and chloroform). The lipophilic phase can then be analyzed using lipidomics approaches using Liquid Chromatography–Quadrupole Time of Flight–Mass Spectrometry (LC–QTOF–MS) technology for screening of all present lipid classes9 followed by absolute quantification of each lipid species using multiple reaction monitoring (MRM) using authentic standards on a Liquid Chromatography–Triple Quadrupole–Mass Spectrometry (LC–QQQ–MS) instrument.10 The polar phase of the extraction can then undergo different chromatographic separations based on C18 reversed phase for the more apolar components in this polar phase (e.g., nonpolar amino acids such as isoleucine, leucine, phenylalanine, tryptophan, or secondary metabolites such as flavonoids or certain plant hormones and small peptides), with further hydrophilic or aqueous normal phase-based separations for the polar component of the extract.11 In addition to the different separation chemistries, positive and negative ionization methods need to be applied to cover both positively and negatively charged molecules present in each separation. In each of these approaches there are between 500 and 2000 mass features detectable11 represented with their accurate mass, isotopic pattern, and retention time. Accurate mass (as obtained with a QTOF or FT-type instrument, see Ref. 4) and isotopic pattern as well as additional secondary and tertiary MS (MS2 or MS3) can aid structural elucidation of the compound of interest. However, there are a number of metabolites that are similar in structure, mass, and fragmentation pattern (e.g., isobaric compounds or sugars of similar mass), which will make identification of certain peaks almost impossible. There are a number of commercial and publically available libraries of accurate masses of compounds available (e.g., Kegg Ligand (http://www. genome.jp/ligand/), Human Metabolome Database (HMD; http://www.hmdb.ca/), Metlin (http://metlin.scripps.edu/), or Chemspider (http://www.chemspider.com/)); however, unambiguous identification is not always possible as a number of metabolites can have exactly the same molecular formula and mass. Different initiatives around the world have been attempting

438

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

to develop solutions which would require standardized LC–MS method setup including column stationary phase, elution solvents and gradient, ionization procedure (e.g., positive/negative ESI with fixed conditions), and MS scan setup including accuracy of mass detection. Once the community has settled on a selected number of methodologies, we will be able to create LC–MS-based mass spectral information including accurate mass, MS2 and MS3 fragmentation pattern and very importantly, retention time under very specific LC conditions.

1.31.2.3

Nuclear Magnetic Resonance

The advantages of using NMR for metabolomics analyses are numerous. Most importantly, NMR is a nondestructive measurement that allows the sample to be further analyzed using other analytical techniques. Additionally, NMR can quantitatively detect a large number of different compound classes independent of size, charge, volatility, or stability.12 NMR also represents a fast and relatively cheap analysis per sample, although a substantial capital investment has to be considered for the instrumentation. The disadvantages of NMR lie in the relatively low sensitivity, which results in larger sample volumes to be required for extraction. NMR-based analysis is based on radio-frequency pulses generating high-energy spin states in nuclei with odd atomic or mass numbers (e.g., 1H, 13 C, or 31P) in a strong magnetic field. The resultant radiation is then emitted and detected while the nuclei return to their lowerenergy spin state, where the resultant signal is further transformed using Fourier transformation calculations. The technology has been most commonly used in the analyses of biological fluids, such as blood, serum, or urine, and has demonstrated great success in medical applications.13 However, in plant metabolomics applications, NMR is becoming more and more important, as summarized by Ward et al.14

1.31.3

Metabolomics Approaches

Widely accepted definitions of the different approaches used in metabolomics have been given in Ref. 15. These include target analysis, metabolite profiling, metabolomics, and metabolic finger- or footprinting. Targeted analysis has been done for many decades now, well before the field of metabolomics was born. In target analysis, a small and well-defined set of known metabolites are measured and quantified using one particular analytical method such as Gas Chromatography–Flame Ionisation Detector (GC–FID) or high-performance liquid chromatography (HPLC) coupled to an ultraviolet visible (UV–VIS) or diode array detector (DAD). A huge number of methods are available for many kinds of targeted metabolites such as sugars, amino acids, nucleotides, flavonoids, alkaloids, or fatty acids (see Ref. 4). Newer, sophisticated technologies such as GC–MS or LC–MS have now allowed the analysis of many more compounds simultaneously in a more unbiased manner, a process which is called metabolite profiling. Often, the analyzed compounds may not be identified with respect to their chemical nature. As described above, GC-MS represents one of the technologies used for metabolite profiling, as it enables the detection of amino acids, organic acids, fatty acids, sugars, sugar alcohols, amines, and sterols. However, a typical GC–MS run detects about 400 compounds of which only about 100–120 are known, depending upon the biological matrix. Utilizing the GC–MS-based mass spectral databases mentioned above, GC–MS metabolite profiling is now routinely set up in many laboratories around the world. The combination of complementary analytical technologies with the aim of analyzing as many metabolites as possible is called metabolomics. The analyzed metabolites may be either identified or unidentified. As described above, a single analytical platform or methodology has its limitations to detect the huge chemical diversity that metabolites represent. Therefore, metabolomics, by definition, would require the application of multiple platforms such as GC–MS, LC–MS, and NMR. The advantage of this multifaceted approach is that not only is a lot more information obtained but also there will be some compounds which will be detected in two or more technologies giving greater confidence in the estimation of the analytical error inherent in each technology. Once it has been confirmed that similar results are obtained from each technology, validation of that combination of the data for any subsequent bioinformatics analyses is possible. The fourth approach is called metabolic fingerprinting (for cellular metabolites) or footprinting (for extracellular metabolites). A metabolic signature of a sample is generated in a very crude way. Often, the complex mixtures are not separated using chromatographic methodologies but rather all are injected directly into a mass spectrometer (direct infusion MS). For this approach, an instrument with high mass resolution is required (e.g., high-accuracy QTOF–MS, Orbit-Trap-based MS, or Fourier transform ion cyclotron resonance MS (FT-ICR-MS)), allowing the separation of metabolites in the resulting mass spectra of the sample on a mass basis. The resulting mass spectra are then compared between samples allowing the identification of differences between a huge numbers of samples in a very short time frame.

1.31.4

Bioinformatics: What Can It Do

Bioinformatics is the discipline that undertakes detailed computational analysis and interpretation of computer-readable biological data resulting from genetics, genomics, and high-throughput biochemical analyses of gene products such as transcripts, proteins, or metabolites. This data is generated, for instance, through a DNA sequencer or a mass spectrometer, along with associated metadata information such as experimental design, or it might be obtained through clinical or phenotype assessment. In this regard, bioinformatics is considered a very broad area that spans the individual diverse life science disciplines as well as disciplines including

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

439

information systems, software development, human computer interaction, high-performance computing, pure/applied mathematics, and statistics. Bioinformatics activities can include development of new algorithms for data analysis for metabolomics or proteomics, detailed data and molecular sequence analysis, comparative genomics and evolutionary analysis, statistical analysis of biological results, as well as software development of information systems to support experimental design, data storage, and data management. The field of bioinformatics benefits greatly from, and promotes the reuse of software libraries through, open source licensing. These benefits include the ability to leverage existing works in the field. In emerging fields such as metabolomics, the benefits of open source are in contrast to vendor-specific software systems and file formats which can limit the ability of the community to customize analysis to fully exploit the scientific value of the generated experimental data and advances in technology. The majority of available bioinformatics resources are Internet based, and now a diverse number of tools can also be accessed through the Internet. From a software development perspective, delivering robust and sustainable Internet-based resources require close interaction with analytical and biological scientists as well as bioinformaticians. This collaboration is critical in order to, for example, articulate system requirements that have been typically gathered by the scientists through pragmatic use of existing legacy bioinformatics tools and systems for conducting detailed data analysis. Internationally, a number of collaborative initiatives have been undertaken for the development of an Internet-based information management system to manage metabolomics experiments.16 For instance, at the University of Davis, an open source approach is being undertaken called SetupX (http://setupx. fiehnlab.ucdavis.edu:8080/m1/index.jsp). SetupX is essentially a laboratory information management system (LIMS) that enables scientists to detail and set up biological experiments.16 The various functionalities of SetupX allow a user to continually track and run an experiment smoothly and efficiently at all stages, from receiving the samples to generating of reports on the results. The program interacts with the instrument, in this case a GC-MS, where it creates a sequence of samples to be run. Once the samples have been run, the resulting data can be sent to any connected annotation software. The result files are then made available and can be downloaded. Viewing of the data can be limited to those who have been specifically granted access to the experiment, or can be made publically available. Integration of analysis tools and data involves the process of creating analysis pipelines or workflows to conduct a particular data analysis. Workflows can be implemented in various programming languages, for example, Perl (http://www.perl.org/), Python (http://www.python.org), and Java (http://java.sun.com/), utilizing extensive application programming interfaces such as BioPerl17 and BioJava.18 More recently, bioinformatics tools and data can be accessed via web services. Construction of workflows in this manner requires a level of programming proficiency that might present a barrier to life science researchers that do not possess these skills. A possible solution is to provide molecular biologists with a visual environment that abstracts the technical detail of scripting workflows. Well-known examples include the Taverna19 and Kepler projects that provide visual workflow programming environments. Further work has been built on these sophisticated workflow environments and abstracted the complexity to improve accessibility for nontechnical domain scientists. In addition, there are projects to enable web-based submission of workflows generated by desktop applications.20 Another analysis environment is KNIME, which is a modular data exploration platform that allows a user to plug-and-play with different data flows. It has a suite of modules, which can be used together to analyze large amounts of data. The program has the ability to handle data intelligently, has an intuitive interface, allows importing/exporting other workflows, parallel execution on multi-core systems, and is command line capable. The modules which KNIME has available include IO (inputs/outputs to files or databases), data manipulation (filtering, group-by, pivoting, binning, normalization, aggregation, joining, sampling, partitioning, etc.), views (several interactive views), highlighted (ensures that hilting data points are hilting in all views), statistics (regression, linear correlation, correlation filter, statistics, and value counter), data mining (clustering, rule induction, neural network, decision tree, association rules, naive Bayesian networks, support vector machines, etc.), chemistry (translators, connectivity, fingerprints, substructure search, two-dimensional (2D) coordinates, 3D viewer, etc.), distance matrix, image processing, loop support, and many other modules. The KNIME program runs on either Windows, Linux, or Mac setups (http://www.knime.org/).21 Alternately, Rich Internet Applications (RIAs) provide the ability to efficiently manipulate data sets using intuitive functionality traditionally associated with desktop applications from within a web browser (http://download.macromedia.com/pub/flash/whitepapers/richclient.pdf). There exists an opportunity for RIAs that facilitate a pipeline or workflow development and are accessible to a nontechnical audience. A system developed in this style is known as YABI (ccg.murdoch.edu.au/yabi). It is an RIA that is aimed at molecular biologists who wish to conduct a range of bioinformatics analysis. YABI integrates bioinformatics tools and data via an intuitive workflow creation and management environment. All analysis results and a comprehensive analysis audit trail22 are stored online within a user‘s YABI workspace. Currently, YABI incorporates tools and resources for molecular sequence analysis but can be easily extended to incorporate analysis of other “omics” technologies. In this regard, systems such as YABI are cross-platform analysis environments which have seamless integration with high-performance computing and peta-scale storage solutions.

1.31.5

What Does the Informatician Need to Analyze the High-Density Data?

Conducting ongoing metabolomics, proteomics, and genomics analysis in a scalable and reusable fashion is critical given the pace of molecular data generation via mass spectrometers and next-generation genome sequencers. For instance, a single metabolomics experiment using high-sophisticated analytical instrumentation can generate over potentially millions of data points that require careful analysis to distinguish between noise and signals from the sample of interest. In an analogous way, a single genomic

440

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

experiment, using next-generation sequence technologies, can generate over 4 million sequence reads that must be processed, mapped, assembled, and analyzed in context with other data both already available (e.g., whole genome) or complementary experiments such as transcriptome experiments or metabolic fingerprint analysis using untargeted fast LC with accurate mass detection. Devising strategies to manage, analyze, and integrate large amounts of data is a major focus of international bioinformatics activities. In a field such as metabolomics there are fewer initiatives underway compared to say a mature field such as genomics, where there are a number of groups that are working on developing strategies and analysis pipelines for their internal processes for delivering bioinformatics support and analyses to research collaborators and external clients. Some of these groups make their software tools and processes available for other labs to adopt. Other groups focus on developing suites of tools in the public domain. For instance, the Generic Model Organism Database (GMOD) makes available a range of tools for creating and managing genome-scale biological databases (http://www.gmod.org). GMOD can be used to create small laboratory database of genome annotations, or larger web-accessible community databases. Some examples of GMOD enabled tools include Apollo (annotation/visualization), Chado (database), Caryscope (gene visualization), BioMart (database tool) and Pathway Tools (molecular pathway visualization). Some biological data repositories have been moving to a newer more interpretive data format. One such format is the Minimum Information about a Microarray Experiment (MIAME), which is essentially a data format schema that lists the minimum requirements to enable the interpretation of the results of an experiment unambiguously and to potentially reproduce the experiment. Any one particular format is not specified, but data formats are in MAGE-TAB, based on a spreadsheet format or MAGE-ML (http://www. mged.org/Workgroups/MIAME/miame.html). Another data format similar to MIAME is a data model called Architecture for Metabolomics (ArMet). The model describes plant metabolomic experiments and their results. The data model represents the first step to standardize the data for the field. The data model is an object-oriented data model and may be implemented in a variety of ways to produce databases and data management tools to support plant metabolomics experiments. The implementation of this data model to produce the database would translate to more efficient use of time, a reduction in errors from manual input, and allow customization.23 Open source interfaces to data repositories for biological data are becoming more prevalent throughout the research community. One such database that accepts data in an XML format is the PRoteomics IDEntifications (PRIDE) database (http://www.ebi.ac.uk/ pride/). PRIDE is a centralized data repository for proteomics data, which provides protein and peptide identifications, as well as the evidence for the identification, and can capture the details of post-translational modifications. PRIDE was developed through a collaboration of the EMBL-EBI and Ghent University in Belgium. It was developed to provide a common data exchange format and repository to support proteomics literature publications. The PRIDE database features a PRIDE converter and data submission, where a stand-alone application exists for conversion of open source data format (mzData) to PRIDE XML format. Direct submission of pride XML files is possible once the user is logged into PRIDE. The PRIDE XML or mzData files can be validated through PRIDE before submission. Experiments can be browsed by project name, species, tissue types, cell types, gene ontology, or disease types. An advanced search feature is also available, where searches can be performed based on particular criteria such as accession number, author, sample, and/or peptide sequence. Experiments can also be searched by the mass of the protein‘s post-translational modifications. Experimental data can be viewed and an Ontology Lookup Service (OLS) is available, which is a web service interface that allows multiple queries on ontologies from a single location with a unified output format, and can integrate any ontology available in the Open Biomedical Ontology (OBO) format. Another feature called the PRIDE BioMart performs searches on the Pride or Reactome data sets. It allows retrieval of PRIDE data from a data warehouse. The interface allows selection of the attributes to be viewed, avoiding the need to search through a massive table of results. PRIDE is available to download as open source software and is also available on the EBI web site (Figures 1–4).

1.31.6

Data Preprocessing: From Raw to Sense

A typical pipeline for a metabolomic experiment from biological question to statistical analysis is given in Figure 5. Data processing is a two-stage process. The first stage, normally known as preprocessing, includes steps that reduce the raw data into an easy to understand format for further data analysis. The second stage includes normalization/transformations and statistical analysis (e.g., univariate, multivariate) to explore the answer to the biological questions under study. Data preprocessing starts from a set of raw data files acquired from the instruments. This step includes filtering, deconvolution, peak detection or feature finding, and alignment of the peaks found from different samples. The main purpose of these steps is to extract the characteristics of each observed ion (e.g., m/z ratio, retention time and ion intensity) from the raw data file into an easy to interpret form. The aim of filtering the raw data is to remove the measurement noise or baseline correction. After filtering, the next step is identification of signals caused by true ions. Once all of the signals are identified, it is important to align all the signals across different experiments in order to deal with retention time drifts between runs. There are a number of open source and vendor-based softwares available to perform these operations. A review on the preprocessing steps and softwares available for data preprocessing for MSbased metabolites is given by Katajamaa and Oresic.24 A review on deconvolution process for data acquired by GC-MS is given by Likic.25 Once the data matrix is generated in a table format with samples as rows and metabolites as columns, the next step is data normalization and subsequent transformation. We have found that this is still a gray area and there is a distinct need to develop a greater understanding of when, why, and how to normalize or transform the data. In the following section, we explain this step in detail.

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

Figure 1

YABI workflow design interface displaying the creation of a simple bioinformatics analysis workflow.

Figure 2

PRIDE example of experiments available to download and view.

1.31.6.1

Normalization and Data Transformation

441

One of the crucial steps in metabolomic data analysis is normalization and an appropriate transformation of the measured data. It should be understood that normalization and transformation serve different purposes. Normalization of data depends on the properties of data and attempts to remove the impact of nonbiological influence on biological data from an instrument. It is an operation that is applied to the data from each sample and its purpose is to make all samples to be directly comparable with each other.

442

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

Figure 3

PRIDE example view of experiment data.

Figure 4

PRIDE view of spectral data, which can assist with manual de novo sequencing.

Transformation is an operation that is applied to each variable (intensities of metabolite) and is dependent upon the statistical test that is needed to be performed to answer a biological question. For example, the t-test assumes that data are normally distributed so it is important to transform a right- or left-skewed data to normally distributed data prior to performing this test. It is also important to explore the data before choosing any normalization or transformation method, for example, heteroscedastic noise (where the standard deviation of each metabolite in replicate samples changes with the mean of the metabolite) may change the results from the normalized data profoundly26. Keller et al.27 have demonstrated that heteroscedastic noise in the data from LC with diode array detection (LC-DAD) gave rise to spurious components when principal component analysis (PCA) was used to decompose the data. Most biological data are characterized by heteroscedasticity, as shown for data obtained by transmittance Fourier Transform– Infrared Spectroscopy (FT-IR) analyses.28 The optimal transformation to change heteroscedastic noise into homoscedastic depends on the structure of the heteroscedasticity in the signals. If the standard deviation is proportional to the mean of the signal then log-transformation is optimal to treat heteroscedasticity in the data. If the standard deviation is proportional to the root of the mean, then square root transformation provides a homoscedastic noise pattern.

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

443

Biological question Experimental design Sample processing and Data acquisition

Filtering Peak finding or feature detection or deconvolution

Alignment

Normalization Transformation

Statistical analysis

Figure 5

Metabolomic data analysis pipeline.

Although data normalization is an optional step, if the assumptions of a normalization method are violated, then it could harm the data instead of being beneficial. It is important to choose an appropriate method for normalization by including prior knowledge about the experiment (e.g., additional technical or biological knowledge) and assumptions under which the data was generated. It is also important to understand that each data set is different and one normalization method may not be appropriate for all data sets. Some of the commonly used methods to normalize the data are given below:

1.31.6.1.1

Z-Scores

In Z-scores, normalization deviation of each point is taken from the mean and divided by the standard deviation and thus standard deviation is set to 1 and mean is set to zero. The underlying assumption is that neither mean nor variance contains any valuable information (which is often not the case). The main drawback of this normalization is that it gives equal weight to very small values (which could only be noise) and large values. Scaling up the small variables (may be nonrelevant also) could change the results profoundly. If heteroscedasticity is present in the data then it cannot be treated with this normalization process as data cannot be log-transformed after Z-scores normalization due to negative values in Z-scores can be calculated as Z
scores ¼

xij
xi si

(1)

where xij is the intensity of the sample i and for metabolite j. xi and si are the average and standard deviation, respectively, for all the metabolite in sample i. If the data are log-transformed before normalization then the normalized intensity exij from Eq. (1) is   ln xij
lnðxi Þ (2) SDðlnðxi ÞÞ Eq. (2) is equivalent to normalization by geometric mean, log-transformation and then dividing by the standard deviation of the log-transformed intensities (peak areas) which makes it hard to interpret the Z-scores.

1.31.6.1.2

Median

In this case, each data point (peak area) in the sample is divided by the median of all the metabolites (all the peak areas) in that sample. It is assumed that although all the samples are different, median is more or less the same across all the samples. If this assumption is not true, then the results from normalization by this method might be difficult to interpret. This method is particularly useful when one wants to retain the relative variance of the variables, as high variance would correspond to a high relative change, which is useful in analyses that consider variance (e.g., PCA). It is better to use the median rather than the mean, as it is more robust to outliers. If heteroscedasticity is present in the data, log transformation can be applied before or after the median normalization, but care should be taken in which way one performs transformation:

444

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics  exij ¼ log

   exij ¼ log xij
log½medianðxi Þ medianðxi Þ

(3)

Thus, if log transformation is taken first, then median of each sample‘s intensities should be subtracted instead of dividing by it. Median and log transformation (if necessary) will give an approximately evenly distributed data that is centered to zero. Another approach to normalize by median is to use the median of the control samples if available (e.g., in a study where one has wild type in mutant experiments or time point zero samples in a time series study). Ideally, the median of control samples should be the same and thus can be used to normalize the data. It is anticipated that results from both normalization methods (dividing by median of all the metabolites in a sample or dividing by the median of controls) will be quite similar but in the case of dividing by the median of control this will make the results easy to interpret.

1.31.6.1.3

Sum of All or Some of the Metabolites

A commonly used method to normalize the data is to divide each entry in the sample by the sum of all or few metabolites in the sample. The main assumption in this case is that total area under the chromatogram is the same in all the samples. So, dividing by the sum will normalize the data and make them comparable. If strong heteroscedasticity is present in the data (which is commonly found in the metabolomic data), then normalizing by sum introduces spurious correlation between the variables.28 This can be understood by the following explanation. The intensity of a signal j of a sample i can be expressed as xij ¼ exij þ eij ðfor all jÞ

(4)

where exij is the true signal and eij is the noise term in that signal. Constant sum of the sample i can be written as: X X X exij þ xij ¼ eij

(5)

j

j

j

Now if the noise is heteroscedastic (where noise is increasing with the signal intensity) then the error term

P eij will be j

dominated by the largest signals and the intensities of the smaller signals are suppressed. This will lead to false negative or positive correlation between the major variables and other small variables. In short, this normalization technique affects the covariance structure of the variables and special consideration is needed when using method for multivariate analysis.

1.31.6.1.4

Internal and External Analytical Standards

Another commonly used way of normalizing the data is to use internal and/or external standards where either one or multiple standards are used to normalize the data. The external standards (used after the extraction of sample) could be useful to remove any instrumental variation. Internal standards (used prior to extraction) are good to remove any nonbiological variations from the samples but due to the large number of metabolites and chemical diversity one standard is not enough for all the experiments and it is also not practical to use a separate standard for each metabolite. The above section addressing normalization and transformation is discussed from a multivariate or univariate analysis point of view, so special care should be taken when performing these procedures to answer any other question. For example, to look at the fold change between different metabolites under different experimental conditions, it is important to take care in how to calculate fold change depending on whether the log transformation has been applied or not. Technically, there is no need to transform the data for fold change and it should be calculated just on the normalized data. Once the data are normalized by removing the systematic errors and transformed to follow a normal distribution, it is ready for univariate and multivariate analysis in order to calculate statistical significance between different samples and metabolites.

1.31.7

Requirements and Problems of Statistical and Multivariant Analysis of Metabolomics Data

The type of statistical analysis on the normalized and transformed data set depends on the question we need to answer. For example, classification of two or more groups when groups are known or unknown, or finding the metabolites contributing to the difference between two or more groups, etc.

1.31.7.1

Univariate Analysis

After normalization, the next step is to look at the statistical properties of the data in order to explore whether the difference between the groups is statistically significant or not. A common test to test whether the mean of two groups are distinct or not is the t-test that gives a P-value showing whether the distinction between the groups mean is statistically significant or not. If the number of samples for t-test is very small, then it is quite possible to have many false negative and false positives. Sometimes, it is also possible that many metabolites do not show any statistical significance individually but a combination of them could be significant and observation of this requires multivariate data analysis.

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

445

Another common univariate analysis is to calculate the fold-change between the metabolites and plot the fold-change against the P-value from t-test. This plot is commonly known as a volcano plot and it compares the size of the fold-change to the statistical significance level.

1.31.7.2

Multivariate Data Analysis

Among commonly used multivariate analysis metabolomic data is PCA, linear discriminant analysis (LDA), hierarchical cluster analysis (HCA), partial least squares discriminant analysis (PLS-DA), and independent component analysis (ICA).

1.31.7.2.1

Principal Component Analysis

PCA is a data transformation technique that is used to reduce multidimensional data sets to a lower number of dimensions for further analysis (e.g., ICA). In PCA, a data set of interrelated variables is transformed to a new set of variables called principal components (PCs) in such a way that they are uncorrelated and the first few of these PCs retain most of the variation present in the entire data set. Thus, the first PC is a linear combination of all the actual variables in such a way that it has the greatest amount of variation. Second, PC is also a linear combination of the original variables in such a way that it has the most variation in the remaining PCs. PCA is an unsupervised technique where knowledge of prior groups is not required and thus sometimes it is useful to explore potential grouping of samples in an experiment. Two plots can be generated from PCA – a score plot (Figure 6a) and a loading plot. A score plot gives the relationship between the samples, and loading plot gives the relationship between the metabolites. A bi-plot is a combination of the score and loading plots and gives information on metabolites contributing to the differences between the groups. For more detail on PCA, the reader is recommended to see Ref. 30.

1.31.7.2.2

Linear Discriminant Analysis

LDA is a classical technique to predict groups of samples. This is a supervised technique and needs prior knowledge of groups. Therefore, LDA is well suited for nontargeted metabolic profiling data, which is usually grouped. LDA is very similar to PCA, except that this technique maximizes the ratio of between-class variance to the within-class variance in a set of data and thereby gives maximal separation between the classes. LDA and PCA are similar in the sense that both of them reduce the data dimensions but LDA provides better separation between groups of experimental data compared to PCA.29 This is because LDA models the differences between the classes of data, whereas PCA does not take account of these differences. The loading from LDA shows the significance of metabolite in differentiating the groups. The main objective of LDA in the analysis of metabolomic data is not only to reduce the dimensions of the data but also to clearly separate the sample classes, if possible.

1.31.7.2.3

Hierarchical Cluster Analysis

HCA is a statistical technique that identifies groups of samples that behave similarly or show similar characteristics and thus quantify the structural characteristics of the samples or variables. The procedure of the hierarchical clustering involves the construction of hierarchy of treelike structure. There are two kinds of procedures to construct a structure, namely agglomerative and divisive. In the agglomerative method, each observation starts in a cluster of its own and then continuously joins clusters together until there is only one cluster consisting of all the observations. The divisive method proceeds in the opposite direction to the agglomerative method. Different similarity measures can be used in HCA, including average linkage, complete linkage, single linkage, and Ward‘s linkage, and these may result in different clusters. The main objective of HCA for the analysis of metabolomic data is to classify the data into different groups by structuring it. This would then help in identifying the relationship among observations. HCA can be applied to samples as well as to the metabolites. In either case, when it is applied to samples or metabolites, it shows which are grouped together on similarity basis (unsupervised classification such as PCA). The results from HCA can be shown either as a clustering dendrogram or as a heat map (Figure 6(b)).

1.31.7.2.4

Partial-Least Squares Discriminant Analysis

PLS-DA is a powerful supervised classification method. This method has proved to be robust for high-dimensional data and is used for other “omics” data analysis. This technique is useful for studies aiming for diagnosis, prognosis, or treatment outcomes. PLS-DA uses multiple linear regression technique to find the direction of maximum covariance between a data matrix (X) and its class grouping (Y). Both X and Y are reduced to principal components, then the components of X are used to predict the scores on the Y components, the predicted Y components scores are then used to predict the actual values of grouping, Y. Like PCA, PLS-DA also gives classification (score plot) and feature selection (loading plot). Since it is a supervised technique, there is a danger of over fitting the model, but most of the software available for PLS-DA has different options for cross validation to check the validity of models. Apart from the commonly used methods mentioned above, there are number of other univariate and multivariate techniques. For instance, there are classification techniques and decision trees that are quite common in machine learning and that can be used in metabolomics data to classify the samples depending on the measured profiles of the metabolites. This is also a supervised method and like other supervised methods it has the danger of overfitting and it would be necessary to check the validity of the trees with cross-validation techniques.

446

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

Figure 6 (A) Principal component analysis of GC–MS resulted metabolite profiles of leaf tissue of four different species. Data have been produced and analyzed as described in Ref. 7. (B) Heatmap analysis combined with hierarchical cluster analysis of the same data set using the R program called made4 as described in Ref. 31 (Ref. 32, Copyright Biotechniques).

1.31.7.3

Treatment of Missing Values

Like other “omics” approaches, a metabolomics data matrix contains missing values and some metabolites can have no numerical value in some samples. One of the reasons for this could be that a particular metabolite is not found in the sample or may be lower than the threshold set in the algorithm for peak detection or, it could just be that algorithm has not picked that peak. There are two kinds of missing values: one where a particular metabolite is missing in all the samples of a group of replicated (biological) measurements and the other where a particular metabolite is missing in some of the samples in a group. In either case, it is recommended to confirm the missing values from the raw data files. Once it is confirmed that values are really missing then they should only be replaced when necessary. For example, univariate analysis like t-test or multivariate analysis such as HCA does not require a fully defined matrix. Some statistical tests (e.g., PCA) require a full data matrix, that is, without any missing values, which make the treatment of missing values an important step of data processing for these tests. One of the approaches to deal with this situation is to remove all the rows (metabolites) that have missing values but this approach comes with the heavy cost of losing important information that may be contained in those rows. There are several ways to estimate the missing values and some open source and commercial packages have algorithms to fill the missing values as a part of processing steps. The simplest methods include 1. if the values are missing in the entire group then those values can be replaced by a value lower than the lowest value in the data matrix and 2. if the values are missing in some of the samples in a group then replacing them either with the average of the remaining values of the group or with the lowest value of that metabolite across all the groups is quite common. For the purpose of PCA there are a number of missing value imputation algorithms available, for example, probabilistic PCA (PPCA), Bayesian PCA (BPCA), and singular value decomposition imputation (SVDImpute).30 It is also important to note that although replacing the missing values is necessary for the computation of some statistical tests, it will not significantly affect the outcome of that analysis provided that number of missing values is small. For statistical analysis, rates of less than 1% missing data are considered trivial and 1–5% is manageable, whereas 5–15% requires sophisticated handling methods and more than 15% may have a severe impact on any kind of interpretation.

1.31.8

Conclusions

Here, we summarize the latest state-of-the art technologies used in metabolomics, an emerging field in functional genomics and systems biology. We have emphasized the current bottleneck in any metabolomics approach; the extraction of information from raw data and its subsequent analyses using different bioinformatics tools. The development of algorithm and automated pipelines enabling to look holistically at the highly dimensional data obtained from a metabolomics experiment remains challenging. This includes peak detection, alignment of detected peaks, distinguishing between noise and real data, and most importantly to accurately integrate each peak allowing quantification. Here, the analysis is not finalized, as described in the article, a number of preprocessing steps of resulting data matrices are as equally important including filtering, normalization and transformation. The preprocessed data matrix can then undergo statistical, uni- or multivariant analyses aiding visualization of the data and

Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics

447

interpretation in a biological context. It is almost unnecessary to stress that a close collaboration and communication between biologists, analytical chemists, bioinformaticians, and statisticians is a prerequisite to gain best and validated information from a metabolomics experiment.

Acknowledgments The authors thank for support from Bioplatforms Australia Pty Ltd which is funded through the National Collaborative Research Infrastructure Strategy (NCRIS), 5.1 biomolecular platforms and informatics with co-investments from the Victoria and Western Australian State Government. U.R also thanks for funding from the Australian Centre for Plant Functional Genomics Pty Ltd.

References 1. Jewett, M. C.; Hansen, M. E.; Nielsen, J. Saccharomyces cerevisiae metabolomics: A Driver for Developing Integrative Analytical Tools for Discerning Metabolic Function. In Metabolomics; Jewett, M. C., Nielsen, J., Eds., Springer: Heidelberg, 2007. 2. Gieger, C.; Geistlinger, L.; Altmaier, E.; et al. Genetics Meets Metabolomics: A Genome-wide Association Study of Metabolite Profiles in Human Serum. PLoS Genet. 2008, 4, e1000282. 3. Villas-Bôas, S. G.; Roessner, U.; Hansen, M.; et al. Metabolome Analysis: An Introduction, Wiley: New Jersey, NJ, 2007. 4. Roessner, U.; Beckles, D. M. Metabolite Measurements. In Plant Metabolic Networks; Schwender, J., Ed., Springer: New York, NY, 2009. ISBN 978 0-387-78744-2. 5. Kopka, J.; Schauer, N.; Krueger, S.; et al. [email protected]: The Golm Metabolome Database. Bioinformatics 2005, 21, 1635–1638. 6. Kind, T.; Wohlgemuth, G.; Lee do, Y.; et al. FiehnLib: Mass Spectral and Retention Index Libraries for Metabolomics Based on Quadrupole and Time-of-flight Gas Chromatography/mass Spectrometry. Anal. Chem. 2009, 15, 10038–10048. 7. Jacobs, A.; Lunde, C.; Bacic, A.; et al. The Impact of Constitutive Expression of a Moss Naþ Transporter on the Metabolomes of Rice and Barley. Metabolomics 2007, 3, 307–317. 8. Roessner-Tunali, U.; Lui, J.; Leisse, A.; Perez-Melis, A. Flux Analysis of Organic and Amino Acid Metabolism in Potato Tubers by Gas Chromatography-mass Spectrometry Following Incubation in 13C Labelled Isotopes. Plant J. 2004, 39, 668–679. 9. Ivanova, P. T.; Milne, S. B.; Myers, D. S.; Brown, H. A. Lipidomics: A Mass Spectrometry Based Systems Level Analysis of Cellular Lipids. Curr. Opin. Chem. Biol. 2009, 13, 526–531. 10. Ståhlman, M.; Ejsing, C. S.; Tarasov, K.; et al. High-throughput Shotgun Lipidomics by Quadrupole Time-of-flight Mass Spectrometry. J. Chromatogr. B 2009, 15, 2664–2672. 11. Callahan, D. L.; De Souza, D.; Bacic, A.; Roessner, U. Profiling of Polar Metabolites in Biological Extracts Using Diamond Hydride-based Aqueous Normal Phase Chromatography. J. Separ. Sci. 2009, 32, 2273–2280. 12. Dunn, W. B.; Ellis, D. I. Metabolomics: Current Analytical Platforms and Methodologies. Trends Anal. Chem. 2005, 24, 285–294. 13. Bollard, M. E.; Stanley, E. G.; Lindon, J. C.; et al. NMR-based Metabonomic Approaches for Evaluating Physiological Influences on Biofluid Composition. NMR Biomed. 2005, 18, 143–162. 14. Ward, J. L.; Baker, J. M.; Beale, M. H. Recent Applications of NMR Spectroscopy in Plant Metabolomics. FEBS J. 2007, 274, 1126–1131. 15. Fiehn, O. Metabolomics – The Link Between Genotypes and Phenotypes. Plant Mol. Biol. 2002, 48, 155–171. 16. Scholz, M.; Fiehn, O. SetupX – A Public Study Design Database for Metabolomic Projects. Pac. Symp. Biocomput. 2007, 12, 169–180. 17. Stajich, J. E.; Block, D.; Boulez, K.; et al. The Bioperl Toolkit: Perl Modules for the Life Sciences. Genome Res. 2002, 12, 1611–1618. 18. Pocock, M.; Down, T.; Hubbard, T. BioJava: Open Source Components for Bioinformatics. ACM SIGBIO Newsletter 2000, 20, 10–12. 19. Oinn, T.; Addis, M.; Ferris, J.; et al. Taverna: A Tool for the Composition and Enactment of Bioinformatics Workflows. Bioinformatics 2004, 20, 3045–3054. 20. Romano, P.; Bartocci, E.; Bertolini, G.; et al. Biowep: A Workflow Enactment Portal for Bioinformatics Applications. BMC Bioinf. 2007, 8 (Suppl. 1), S19. 21. Berthold, M. R.; Cebron, N.; Dill, F.; et al. KNIME: The Konstanz Information Miner. In Data Analysis, Machine Learning and Applications; Preisach, C., Burkhardt, H., SchmidtThieme, L., Decker, R., Eds., Springer: Berlin, Heidelberg, 2007. 22. Bellgard, M. I. Bioinformatics From Comparative Genomic Analysis Through to Integrated Systems. In Mammalian Genomics; 2005; pp 393–409. 23. Jenkins, H.; Hardy, N.; Beckmann, M.; et al. A Proposed Framework for the Description of Plant Metabolomics Experiments and Their Results. Nat. Biotechnol. 2004, 22, 1601–1606. 24. Katajamaa, M.; Oresic, M. Data Processing for Mass Spectrometry-based Metabolomics. J. Chromatogr. A 2007, 27, 318–328. 25. Likic, V. A. Extraction of Pure Components From Overlapped Signals in Gas Chromatography-Mass Spectrometry (GC-MS). BioData Min. 2009, 12, 2–6. 26. Kvalheim, O. M.; Brakestad, F.; Liang, Y. Preprocessing of Analytical Profiles in the Presence of Homoscedastic or Heteroscedastic Noise. Anal. Chem. 1994, 66, 43–51. 27. Keller, H. R.; Massart, D. L.; Liang, Y.; Kvalheim, O. M. Evolving Factor Analysis in the Presence of Heteroscedastic Noise. Anal. Chem. Acta 1992, 263, 29–36. 28. Toft, J.; Kvalheim, O. M. Eigenstructure Tracking Analysis for Revealing Noise Pattern and Local Rank in Instrumental Profiles: Application to Transmittance and Absorbance IR Spectroscopy. Chemometr. Intell. Lab. Syst. 1993, 19, 65–73. 29. Jolliffe, I. T. Principal Component Analysis, Springer: New York, NY, 1986. 30. Stacklies, W.; Redestig, H.; Scholz, M.; et al. pcaMethods – A Biocondutcor Package Providing PCA Methods for Incomplete Data. Bioinformatics 2007, 23, 1164–1167. 31. Culhane, A. C.; Thioulouse, J.; Perriere, G.; Higgins, D. G. MADE4: An R Package for Multivariate Analysis of Gene Expression Data. Bioinformatics 2005, 21, 2789–2790. 32. Roessner, U.; Bowne, J. What Is Metabolomics All About? Biotechniques 2009, 46, 363–365.

1.32

Theory and Applications of Proteomics

BJ McConkey, University of Waterloo, Waterloo, ON, Canada © 2011 Elsevier B.V. All rights reserved. This is a reprint of B.J. McConkey, 1.34 - Theory and Applications of Proteomics, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 461–469.

1.32.1 1.32.2 1.32.2.1 1.32.2.2 1.32.3 1.32.3.1 1.32.3.2 1.32.3.3 1.32.4 1.32.4.1 1.32.4.1.1 1.32.4.1.2 1.32.4.2 1.32.4.3 1.32.5 1.32.6 References

Introduction Proteomics Technologies Peptide Ionization MS Instrumentation Separation Technologies Gel-Based Proteomics LC-Based Proteomics LC or Gel Based? Quantitative Proteomics Quantitative Proteomics of 2D Gels DIGE Experimental Design Considerations Statistical Analysis of DIGE Gels Quantitative Methods in LC–MS Proteomics Label-Free LC–MS Data Processing Applications in Biotechnology

448 449 449 450 450 450 451 451 451 452 452 452 452 453 453 454 455

Glossary 2D gel electrophoresis Separation of intact proteins in a gel matrix by charge in the first dimension and by molecular weight in the second dimension. Differential in-gel electrophoresis (DIGE) 2D gel electrophoresis technique for relative quantification of protein abundance using up to three fluorescent dye labels. Electrospray ionization (ESI) Method for ionization and transfer of peptides to gas phase from liquid phase by elution through a capillary column at high voltage. False discovery rate Statistical technique to control the rate of false positives among all detected positives, correcting for multiple comparisons in high-throughput experiments. Liquid chromatography mass spectrometry (LC–MS) Analytical chemical technique combining separation of complex mixtures by liquid chromatography with molecular mass analysis by mass spectrometry. Mass spectrometer An instrument that measures mass-to-charge ratios of ions, producing a mass spectrum. Matrix-assisted laser desorption ionization (MALDI) Method for ionization and transfer of peptides to gas phase from a solid matrix by absorption of laser energy. Proteomics The characterization of the protein complement of a cell, tissue type, or organism under given experimental conditions.

1.32.1

Introduction

Genomics, transcriptomics, and proteomics, respectively, have allowed the exploration of the cellular DNA, RNA, and protein content of entire cells or organelles, in contrast to traditional investigations of single target genes, proteins, or pathways. The development of these omics technologies has provided the means to investigate cellular systems at a much more comprehensive level than previously possible. While an organism’s genome is typically fixed, the proteome can vary considerably depending on cell type and internal and external cellular conditions. The proteome is effectively a catalog of the presence and/or abundance of a large set of proteins, expressed at a given time and in a given environment. As such, proteomics data can provide insight into the functioning of the cell and its response to a given environment and can provide further information on protein posttranslational modifications (PTMs), interactions, degradation rates, and changes in protein expression. Importantly, proteomics approaches are not

448

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00028-8

Theory and Applications of Proteomics

449

constrained by prior knowledge and can be used to discover novel gene and protein targets, which can be engineered for increased efficiency of bioproduction, improvement of strains, and development of bioprocesses.

1.32.2

Proteomics Technologies

A typical proteomics experiment consists of a number of steps: preparation of sample(s), separation of proteins or peptides, digestion of proteins into peptides (either before or after separation), analysis of peptides by mass spectrometry (MS), identification of peptides, and interpretation of the resulting data.1–3 Advances in MS, and also concomitant advances in genomics and computation, have been vital in the development of proteomics technologies. The analytical challenge of ionization and translation of peptides to gas phase has largely been solved through electrospray and laser-assisted desorption ionization strategies in MS (discussed below). The wealth of genome information now available has produced comprehensive databases of protein sequences, which can be matched to experimental mass spectral data, and finally, there exists the computing power and software to perform the somewhat daunting task of matching, scoring, and sorting thousands of experimental spectra, and translating that information into useful data on protein occurrence and abundance. More recently, quantitative proteomics has been emphasized,4 where either relative abundances of proteins between two samples or, in some cases, absolute quantification of peptides is the experimental objective. The comparison of two or more samples allows for the researcher to focus on the relative differences between the samples, which is often far more informative than abundance in a single sample. These data can highlight functional and regulatory changes in response to environmental conditions or differences between cellular states. From a biotechnology perspective, this information can be used to identify targets for optimization or to help select between strains for production. The development of effective methods for identification of proteins and peptides by MS underlies all proteomics techniques. The two dominant experimental approaches in proteomics have been gel-based proteomics, which separates intact proteins prior to digestion and peptide identification, and liquid chromatography/mass spectrometry (LC–MS) approaches, where proteins are first digested into peptides, then the peptides separated and identified. Several methods for identification and quantification of peptides have been developed for LC methods in particular. Both approaches rely on similar MS technologies, discussed below.

1.32.2.1

Peptide Ionization

Although it is possible to obtain molecular weight information from intact proteins using some systems, peptides tend to be a more efficient and informative target for analysis by MS. Using tandem MS (MS/MS), the amino acid sequence of peptides 5–25 amino acids in length can be determined for most sufficiently abundant peptides. By contrast, little sequence information is available for intact proteins using most analytical systems. Trypsin is a particularly effective enzyme for digestion, as it is stable, relatively inexpensive, and most importantly has a high activity and specificity. Trypsin cleaves polypeptide chains on the C-terminal side of the lysine and arginine, resulting in a set of peptides ending in the positively charged residues K or R. Other enzymes of different specificities, such as Lys-C, Glu-C, or Asp-N, may also be used on their own or to complement analysis of trypsin digested proteins. One of the key events in protein MS was the development of techniques for ionization of peptides and transfer to gas phase. In the 1990s, two techniques were developed to accomplish this: electrospray ionization (ESI) and matrix-assisted laser desorption ionization (MALDI). Both of these techniques provided the means to transfer polar peptides into gas phase and introduce the sample into the mass spectrometer. In ESI, peptides are very slowly eluted from a capillary through a spray needle, often at the outlet of a low-volume, high-pressure liquid chromatography (LC) column used for desalting and peptide separation. The spray needle is kept at a high positive electric potential of several kilovolts, effectively transferring a positive charge to the peptides in the solvent. Highly charged droplets elecrostatically disperse at the needle outlet and undergo further loss of solvent until the peptides are desorbed from the surface or are effective isolated to a droplet small enough to contain only a single analyte ion.5 The use of very low flow nano-LC columns has provided increased sensitivity. ESI will often generate multiply charged ions that can provide a stronger signal for MS/MS fragmentation, though spectra complexity may also be increased. MALDI is a second technique for peptide ionization. In contrast to ESI, MALDI does not interface to a chromatography system; instead, samples are prepared on a spot on a target plate. The analyte of interest (often representing a mixture of peptides from one or a small number of proteins) is mixed with an excess of strongly ultraviolet (UV) absorbing material, typically a lowmolecular-weight aromatic acid. This mixture is dried on the target plate and irradiated with a short laser burst of an appropriate wavelength. Energy from the laser is absorbed by the matrix, which evaporates from the plate surface and carries the analyte peptides with it. A portion of the analyte peptides will obtain a positive charge from the matrix and from intermolecular collisions in the vapor phase and can be accelerated by an electric potential and introduced into the mass spectrometer. The development of related techniques (surface assisted laser desorption ionization (SALDI); desorption/ionization on silicon(DIOS); and atmospheric pressure MALDI (AP-MALDI))1 has provided a greater versatility in this technique, such as increased tolerance to salts and detergents and atmospheric pressure MALDI for greater interchangeability with ESI systems. A single-stage mass spectrometer can effectively identify the molecular weight of a set of input ions, such as a set of peptides from a target protein. This can be sufficient for the identification of the protein based on the set of observed peptide masses (the peptide mass fingerprint). Here, the set of observed peptides is compared with the theoretical masses of peptides in a database of proteins, such as all proteins from the target organism‘s genome.

450 1.32.2.2

Theory and Applications of Proteomics MS Instrumentation

Mass spectrometers typically consist of an ion source, a mass analyzer, and detector, along with data-processing electronics.1–3 The central component of these systems is the mass analyzer, of which several types exist: time-of-flight (TOF) MS, quadrupole MS, ion-trap MS, Fourier transform ion cyclotron resonance (FTICR) MS, and orbitrap MS. These systems are also often combined in a series configuration, such as quadrupole time-of-flight (Q-ToF) or triple quadrupole (TQ or QQQ) mass spectrometers. Each system measures essentially the same molecular property, the spectrum of mass-to-charge (m/z) ratios for molecular analytes within the sample, where the systems differ is in their sensitivity, resolution, range of m/z ratios that can be measured, ease of operation, and analysis speed. Quadrupole mass filters consist of four conducting rods to which a combination of radio frequency alternating current and direct current is applied. Ions are introduced at one end of the quadrupole and have a stable trajectory for only a narrow mass-to-charge range that is dependent on the applied electric fields. Ions with a stable trajectory exit the far end of the quadrupole mass filter, where they can be detected or passed to a second stage of the mass spectrometer. The direct current and radio frequency potentials on the quadrupole can be varied so that m/z values can be quickly scanned and a spectrum of m/z values obtained. Quadrupole instruments are versatile and relatively inexpensive and are suitable for many high-throughput proteomics studies. They have a reduced mass resolution compared to other systems but are often used very effectively either in series or in combination with other MS systems such as ToF MS. ToF instruments use an alternate approach, where ions are accelerated through an electric field and then follow a set ion path. The transit time through the ion path is proportional to the square root of their atomic mass. The ToF can be precisely measured, providing excellent mass resolution and very low sensitivity. Newer systems often incorporate ion reflection systems with a resulting V- or W-shaped ion trajectory, leading to more accurate m/z measurements. ToF MS is often paired with one or more quadrupole mass filters in an MS/MS systems, as discussed above. Ion-trap systems work similarly to quadrupole mass filters, except that ions of the target m/z are retained and accumulated in the ion trap, then sequentially ejected. A variation on the quadrupole ion trap is the linear ion trap, which has greater ion-trapping capacity and greater dynamic range. These systems are well suited for high-throughput, robust analyses and have high sensitivity and reasonable mass resolution. The fast scanning capabilities and good sensitivity of ion traps make them suitable for analysis of complex proteomics samples.1 FTICR mass spectrometers operate on a different principle, where ions are confined using a strong magnetic field from a superconducting magnet. The ions cycle within the FTICR at frequencies inversely proportional to their m/z ratios and induce an alternating current within FTICR. This time-varying current represents a frequency spectrum of the trapped ions and can be converted into an m/z spectrum through a Fourier transform. As the frequency spectrum can be measured with very high accuracy, the mass resolution of the FTICR is correspondingly very high (>100 000). The cost of FTICR systems is quite high, however, primarily due to the cost of the superconducting magnet. The Orbitrap MS combines the features of an ion trap and an FTICR system. The Orbitrap uses a static electric field around a central electrode, around which ions oscillate in an axial direction. Similar to an FTICR, the ion oscillations can be detected as a time-dependent signal and converted into an m/z spectrum. The Orbitrap has a very high mass resolution, high mass accuracy, reasonable dynamic range, and lower cost compared to an FTICR.

1.32.3

Separation Technologies

The two dominant types of proteomics separations are gel-based systems, in which intact proteins are separated in a two-dimensional (2D) gel matrix, then excised, digested, and identified by MS, and LC systems, where complex mixtures of peptides are separated by their physicochemical properties such as charge and hydrophobicity. 1D gel-LC-MS approaches are also common, where a protein mixture is separated by molecular weight in a 1D gel, then bands excised, digested, and the resultant peptides are analyzed by LC–MS.

1.32.3.1

Gel-Based Proteomics

2D electrophoresis has been the workhorse of proteomics since its inception in the 1990s. 2D electrophoresis can simultaneously separate thousands of proteins and protein isoforms as spots within a 2D gel matrix, which can be subsequently identified and characterized by MS. Samples for 2D electrophoresis can be whole-cell extracts from a given cell type, a type of tissue (e.g., leaf tissue and liver tissue), or a subset of cellular components obtained from isolating an organelle or fractionation of cellular contents. Proteins are separated in the first dimension by charge using an immobilized pH gradient (IPG). Originally, this separation was done using tube gels, containing a mixture of ampholytes to establish the pH gradient. They suffered from a lack of reproducibility, however, and now have been largely replaced by pH gradient strips. The manufactured IPG strips are available in a variety of ranges of pH and can provide highly reproducible separations. After the sample is introduced to the IPG strip, an electric field is applied to the strip, causing the proteins to migrate to their respective isoelectric points (pI, the point at which a protein has zero net charge) on the pH gradient. Separation in the second dimension is by molecular weight, using the standard electrophoretic technique of sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE). The IPG strip containing proteins separated in the first dimension is placed on top of an SDS–PAGE gel, and proteins are drawn through the gel matrix using an electric field. In this case, the SDS provides a net negative charge on each protein, so proteins migrate vertically in the gel at a rate approximately inversely proportional to their molecular weight. Protein spots on resulting 2D gel can be imaged using stains or fluorescent dyes and spots excised for identification by MS.

Theory and Applications of Proteomics

451

One of the advantages of 2D gel-based approaches is that as intact proteins are separated, identification of target proteins can be easier in some cases. If the genome of target organism has not been fully sequenced, for example, it can be easier to identify the protein based on comparison with homologous proteins in other species.6 This task is more difficult using peptide-based approaches, as there is no a priori information to group peptides with a homologous parent protein. Another potential advantage is that different protein isoforms are separated, so it is possible to detect PTMs that result in a different charge state (e.g., phosporylation) or molecular weight of the protein. A downside of gel-based approaches is the bias toward highly abundant proteins. Protein spots need to be sufficiently abundant for visualization and quantification, so low-abundance proteins may be missed. Highly hydrophobic proteins or proteins with extreme molecular weights or pIs may also be more difficult to analyze using standard 2D gel techniques.

1.32.3.2

LC-Based Proteomics

LC-based proteomics separates complex mixtures of peptides prior to analysis by MS. In contrast to gel-based techniques, proteins are digested first; then components of the peptide mixture are separated. A whole-cell sample containing thousands of protein isoforms can result in hundreds of thousands of peptides, which should undergo sufficient separation to produce unambiguous protein identifications. Low-abundance peptides, in particular, need to be sufficiently resolved so the signal is not dominated by overlapping higher-abundance peptides. High-performance liquid chromatography (HPLC) systems are typically coupled with ESI to provide a continuous source of ions. Separation of peptides is done using a combination of chromatographic materials in columns, with ion exchange, affinity, and reverse phase (RP) being common packings. RP resins are essential to proteomic LC separations and separate peptides based on hydrophobicity. The mobile phases used with RP are compatible with ESI and provide high resolution and reproducibility. Greater resolutions are obtained with small particle size of the packing material and greater column length. These systems typically require very high pressures and low flow rates and have improved peak capacity, sensitivity, and resolution.1 A high-complexity proteome sample may contain thousands of proteins, with abundances covering several orders of magnitude. Proteolytic digests of these samples result in numerous peptide fragment per protein, producing tens or hundreds of thousands of unique peptides within a sample. Multidimensional separations have been applied to address this issue, where several separation techniques are used in series to improve resolving power. Ideally, the separation techniques are orthogonal, where different molecular properties are used in each dimension. One established technique couples a final RP separation with an initial separation based on strong cation exchange (SCX). A highly complex sample is first loaded onto an SCX column, then eluted in a series of steps using increasing salt concentrations. Sample fractions from the SCX column are further separated using an RP column and directly eluted into an ESI source and MS. Other separation materials used in the first dimension include size exclusion, anion exchange, or affinity methods. A combined gel–LC method also provides effective first dimension separation of proteins by molecular weight, followed by digestion and separation by RP in the second dimension. For more targeted studies, affinity chromatography can be used to isolate protein complexes or proteins with specific PTMs. Selective extraction of a given target protein and other proteins binding to the target protein can be accomplished using antibody-based interactions, or via peptide tags added to the target protein. PTMs often play a role in cell-signaling and regulatory pathways, and enrichment of target PTMs such as phosphorylated proteins can enrich low-abundance signaling proteins prior to MS analysis. A method for phosphoprotein enrichment is immobilized metal affinity chromatography (IMAC), where immobilized metal ions on a column are used to selectively bind phosphopetides. Specificity may be altered according to ion type, with iron and nickel as frequently used options. Other options include Zr4þ, Ga3þ, and oxides of Ti, Fe, and Zr.1

1.32.3.3

LC or Gel Based?

Comparison of gel and LC methods has typically concluded that the two broad methods tend to have somewhat different peptide and protein coverage, and each has their own advantages. The more recent LC methods have started to overtake the popularity of gel-based methods and have some analytical advantages including less bias in protein detection, more standardized sample preparation, and faster data collection time than gel-based methods. Gel-based methods, however, have their advantages as well. If the organism or cell line being investigated does not have corresponding genome sequence available, the separation of intact proteins permits stronger identifications, as the set of peptides from a given spot can be mapped to homologous proteins with greater accuracy. PTMs resulting in a change in charge or molecular weight relative to a reference sample are also more easily detected on a gel. It has been shown that the approach with the ability to detect the greatest number of proteins is the hybrid method of 1D gel electrophoresis followed by RP separation and LC–MS identification.7

1.32.4

Quantitative Proteomics

The goal of a proteomics experiment is often to provide a proteome-level comparison of differences in expression, whether the differences are between treatments, cell types, or strains, or to identify responses to a given experimental condition. Quantifying these differences is an important task, and several strategies have been developed to address this for both gel- and LC-based proteomics.

452

Theory and Applications of Proteomics

1.32.4.1

Quantitative Proteomics of 2D Gels

Initial efforts in quantitative gel-based proteomics involved comparison of two or more 2D gels, stained with fluorescent dyes, silver, or coomassie stains. This required accurate image alignment of the different gels and suffered from a lack of repeatability, poor gel-to-gel alignments, and reduced dynamic range. The introduction of differential in-gel electrophoresis (DIGE) has greatly reduced these issues, through the use of multiplexed samples. In DIGE, samples with three different fluorescent labels can be run on the same gel, typically as two sample channels plus an internal reference channel. The usual approach is to use a minimal labeling strategy, in which 1–3% of proteins are labeled with the amine-reactive CyDyes. This ensures that the majority of detected proteins have only one label, so multiply labeled proteins do not need to be considered. The DIGE CyDyes used are charge- and molecular weight-matched, so otherwise identical protein isoforms with different dye labels migrate to the same gel coordinates. The internal reference standard is a mixture of all samples run within a given experiment, so a reference sample of identical composition is present on each gel. The internal reference allows for accurate matching of samples across gels, greatly improving gel alignment. In addition to the three-dye system mentioned above, alternate designs such as one sample plus an internal reference per gel have been proposed.8

1.32.4.1.1

DIGE Experimental Design Considerations

A primary consideration in a DIGE experiment is the number of gels to run. A larger number of replicates will typically result in an increased number of statistically significant differences; however, the number of replicates may be limited by other factors such as available samples, cost, and time constraints, and the number of gels that may be run per day. It is often the case that running samples in triplicate has insufficient power to detect changes of interest; typical DIGE experiments often have four to six replicates per sample type. A more quantitative estimate of the degree of replication can be obtained through statistical power analysis, which may be used to estimate the number of samples required to detect an effect of a given magnitude. Estimates of variance are required for power analysis, however, and these may not be available prior to running the experiment. It is also possible to do a post hoc power analysis on a given data set, which provides an estimate of the magnitude of change that is likely to be detected.

1.32.4.1.2

Statistical Analysis of DIGE Gels

Multiple hypothesis testing is an important consideration for any high-throughput technique. For single hypothesis testing, a p-value of 0.05 is often used to determine whether a difference is statistically significant. For the hundreds or thousands of spots on a 2D gel, this translates to a false positive rate of approximately 5%, or 50 spots per 1000. As there may only be a small number of true positives present, steps need to be taken to reduce the rate of false positives. A very conservative procedure would be to use a family-wise error rate, where a value of 0.05 would indicate the probability of any false positives. This, however, results in many of the true positives (the spots of interest on a 2D gel) being rejected in a high-throughput study. A more practical means of correcting for multiple hypothesis testing is the application of false discovery rates (FDRs), which controls the expected proportion q of false positives within the set of identified positive cases. For example, in a proteomics experiment where 50 protein spots were identified as differentially expressed, a q-value of 0.10 would indicate that 5 of these 50 are expected to be false positives. The set of p-values calculated using single hypothesis testing may be used as the input for several FDR procedures, and as a result, the test can be applied to a variety of high-throughput data sets. The use of FDR has become standard in the analysis of DIGE data sets.6,8

1.32.4.2

Quantitative Methods in LC–MS Proteomics

In comparison to gel-based methods, LC–MS methods of quantification compare differences in peptide abundances directly using MS data. Quantification can be done using isotopic labels, or it can be label free, comparing MS signal strength in sequential analyses. Several methods for isotopic labeling of peptides for quantification have been developed. In all isotopic labeling methods, different stable isotopes are added to peptides so that peptides from different sources can be differentiated using small differences in mass. The chemical behavior of the peptides is largely unchanged, so in theory peptides of the same sequence but different isotopic masses will behave the same during fractionation and separation and will co-elute. Difference is mass can be detected during first-stage MS or during MS/MS, and the relative peak intensities can be used to determine the ratio of abundances of the isotopes and the corresponding peptides.9 The isotopic labels can be incorporated at several points: metabolic incorporation of stable isotopes in living cells; labeling of proteins or peptides with covalently bonded isotope tags; incorporation of H218O isotopes during proteolysis; or finally, comparing with isotopically labeled synthetic peptides added as internal standards. The first method for relative quantification of peptides was isotopically coded affinity tags (ICATs), which labels cysteine residues with an isotopically light or heavy reagent tag. The original ICAT reagents consisted of a cysteine reactive group (iodoacetimide), an isotopically labeled linker, and a biotin tag. The biotin tag is used for affinity purification of labeled peptides to selectively extract labeled peptides and reduce sample complexity. A more recent version of the ICAT reagents uses a linker region containing 13C atoms instead of 2H in the heavy tag, resulting in improved co-elution of labeled peptides during RP separations. The mass difference of the 13C-labeled reagent is 9 atomic mass units (amu), providing clear separation between the different isotopic labels. A second modification to the ICAT reagents is the inclusion of an acid cleavable link, so that the biotin group used for affinity

Theory and Applications of Proteomics

453

purification may be removed prior to MS analysis. The removal of the biotin tag decreases the size of the peptide and reduces fragmentation of the label, improving the quality of MS/MS spectra and subsequent peptide identifications. An advantage of ICAT is the specificity of the reactive group to cysteines, greatly reducing the complexity of peptide analysis. Proteins that do not contain a cysteine group, however, will not appear in the analysis. A second strategy for quantification, isobaric tags for relative and absolute quantitation (iTRAQ), incorporates labels that are distinguished within the fragmentation spectra in MS/MS analysis. The 4- and 8-plex iTRAQ reagents allow for simultaneous relative quantification of up to four or eight different samples, respectively. The iTRAQ tags consist of an amine-specific reactive group, isotopically labeled reporter ions, and a balance group. For the 4-plex reagents, the reporter groups have four different isotopic masses (114, 115, 116, 117 u). The balance groups have a complementary mass so the total mass of the balance group and reporter ions is equal for the set of tags. Each of the tags behaves the same during chromatographic separation, and as the total mass of each tag is equal, the observed mass of the intact peptide is not affected. Fragmentation of the peptide during MS/MS, however, separates the reporter and balance groups, so that the unique m/z value for each reporter group is observed within the MS/MS spectra. The mass ranges of the reporter groups have been selected in a region that does not overlap with amino acid fragmentation spectra, so the relative intensities of the reporter groups can be used to quantify the relative abundances of the target peptide in the original samples. In metabolic labeling, stable isotopes are introduced to the sample during cell growth. Metabolic labeling was originally demonstrated using simple model organisms such as bacteria or yeast, where the organism was fed a diet containing the heavy nitrogen isotope 15N. The label is incorporated into the organism, and the differences in mass can be detected at the peptide level. The isotopically labeled organisms can be fed to small model organisms in turn, such as Caenorhabditis elegans or Drosophila melanogaster. It is possible that higher organisms such as rats or plants could also be labeled with 15N. This strategy has some pitfalls, however, not least of which is the cost involved in labeling larger organisms. The mass increment for the heavy peptides depends on the sequence of the peptide, greatly complicating both the identification of the peptides and comparison of peptide abundance prior to accurate sequencing. The related technology of stable isotope labeling with amino acids in cell culture (SILAC) circumvents many of the above issues. In this procedure, cell cultures are grown in media supplemented with isotopically heavy or light amino acids such as lysine, arginine, leucine, and tyrosine. The target heavy amino acids contain atoms of 13C, 15N, or 2H such that the amino acid mass of the heavy amino acid is 4–9 Da heavier than the corresponding light version (depending on the target amino acid). The labeled amino acids are incorporated into cellular proteins, which can then be identified and relative abundances determined. This strategy has been expanded to label entire organisms (mouse and fruit fly) by providing a diet containing isotopically labeled lysine. Heavy isotopes can also be incorporated during proteolysis by endoproteinases Lys-N or Glu-C. If the reaction occurs in light (H216O) or heavy (H218O) water, one or two oxygen atoms will be added to each peptide. This technique has not been widely used, however, as the introduced mass difference is relatively small (2 or 4 Da) and the labeling is often incomplete.

1.32.4.3

Label-Free LC–MS

Although the intensity of ion peaks in MS spectra is highly dependent on peptide sequence, advances in the accuracy and repeatability of LC–MS have led to the development of label-free LC–MS. If peptides can be repeatably separated and quantified in multiple MS runs, it is possible to compare total peptide abundances. Two methods of label-free quantification have emerged – peak intensity comparisons and spectral counting. In the peak intensity method, the peak areas of peptides are measured at their respective m/z ratio, and the area of these extracted ion chromatograms (XICs) is compared across multiple runs. The abundance of the peptide should be linearly related to its abundance over a reasonable range (assuming consistent chromatographic separation), so the relative abundance between runs can be directly quantified. Peak intensity comparisons of MS data can be interleaved with peptide identification in MS/MS mode, or identifications can be done separately from quantitative MS runs. The spectral counting method provides a semiquantitative analysis of protein abundance based on MS/MS data. Based on the observation that the number of identified peptide fragment spectra is approximately proportional to the logarithm of protein abundance, differences in the number of identified peptides associated with a protein can be used to estimate relative amounts between samples. Peak intensity measurements tend to provide more accurate ratios for protein abundance; however, spectral counting methods can have a larger dynamic range and greater reproducibility. In general, label-free methods require experimental conditions to be highly similar, with only minimal system variability. Likewise, software used for peak detection, peptide identification, quantitative estimates, and statistical analysis also has to be accurate and consistent, with a low rate of false positives.

1.32.5

Data Processing

The data generated by a proteomics experiment can be very extensive and require considerable data processing and analysis, particularly for large MS/MS data sets. The simpler case of MS data processing can provide good-quality identifications of proteins based on peptide mass fingerprints. Here, a set of peptide masses generated for an individual protein (e.g., MALDI analysis of a spot excised from a 2D gel) can be matched to theoretical peptide masses. The input data set is the list of experimentally observed peptide masses from one protein or a simple mixture of a few proteins, which is then compared to a list of theoretical masses generated from a protein sequence database. The database search can be limited to the specific target organism or a taxonomic subset. This peptide mass fingerprinting procedure is fairly straightforward and has been utilized in a number of MS analysis programs such as Mascot (Matrix Science).

454

Theory and Applications of Proteomics

The data generated by an MS/MS for a specific peptide is the fragment ion spectrum, usually produced by collision-induced dissociation of the parent peptide. Tens to hundreds of thousands of these fragment ion spectra may be generated within a proteomics experiment, and accurate, statistically valid methods of assigning fragment ion spectra to peptide identifications are required. The fragmentation of a peptide is a complex process and depends on the peptide sequence, net charge, and experimental conditions. The process is predictable enough, however, that peptide sequences can be assigned to fragmentation spectra with high accuracy, given good-quality data. There are three general computational approaches for assigning peptide sequences: (1) database searching, where spectra are compared to theoretical spectra derived from a protein sequence database; (2) de novo sequencing, in which the peptide sequence is derived directly from the MS/MS data; and (3) hybrid approaches, combining database searching and short de novo sequence matches.11 All require an input set of peptide components, which at a minimum include the 20 amino acids. Numerous other modifications to this basic schema are possible, with common additions including biological modifications (phosphorylation, acetylation, and methylation), chemical modifications during analysis (methionine oxidation and carbamidomethyl blocking of cysteine), or addition of tags (Biotin, ICAT, and iTRAQ). Each additional modification increases the range of peptides than can be identified, but at the cost of lower statistical significance due to the increase in search space needed. Other parameters include the mass tolerance, dependent on the MS system used, and the proteolytic enzyme used in digestion. The most widely used tools for peptide identifications have been based on database search algorithms, such as Mascot, Sequest, and Protein Prospector.11 These algorithms compare fragment ion spectra with theoretical spectra derived from a target protein database. Common target databases include Uniprot, the International Protein Index (IPI) database, and two within the National Center for Biotechnology Information (NCBI): RefSeq (an annotated listing of best available sequences in NCBI) and the Entrez protein database (more comprehensive, but with greater redundancy, partial sequences, and less detailed annotations). Databases that restrict the search space to the target group such as human, vertabrates, or bacteria are also available and have the advantage of greater power due to the reduced search space. A search score is used to measure the degree of similarity between the theoretical and observed spectra, and the top matches in the database are retained. Several scoring schemes have been created, based on correlation functions, empirical rules, or statistical derivations. Many scores are converted to expectation values or E-values, which is an estimate of the number random matches of similar score expected in the target database. Calculated E-values should be largely independent of the underlying scoring system and provide a useful means to make comparisons between search tools. A high-quality match should have an E-value less than 0.05, similar to p-values; however, this is not a guarantee that the correct match has been found. An extension to the database search is spectral matching, where a library of experimentally observed spectra is used to augment or replace the theoretical spectra used in a database search. This has the potential to be faster and more reliable, but it also requires a database that can provide close to comprehensive coverage of all observable peptides in the proteome. In the absence of a high-quality database, spectral matching could augment current strategies by providing reliable first-pass identifications for a subset of proteins. Spectral matching algorithms have been integrated into existing tools, such as the X! P3 (Proteotypic Peptide Profiler) algorithm available within the X!Tandem software. Proteotypic peptides are commonly detected peptides representative of the presence of a unique protein. Once identified, a targeted search can be conducted for nonproteotypic peptides to identify mutations and PTMs that could not be directly detected in a database search. Reliable spectral libraries with good proteome coverage are available for human and yeast and are being constructed for a number of other organisms.11 In contrast to database searching, de novo sequencing does not rely on selecting from potential peptide sequences and identifying the best match. Instead, peptide matches are generated directly from the observed peaks in the MS/MS spectra. Absolute m/z values and differences between peak m/z values are used to identify the most likely sequence corresponding to the observed spectra. De novo sequencing is particularly useful for organisms that do not have corresponding genome sequence data, or for identifying sequence polymorphisms or modified peptides. Protein identifications are still required, however, necessitating a search of sequence databases with the de novo peptide sequences as queries. The spectra analyzed also need to be of sufficient quality to effectively generate reliable de novo sequences. This procedure can be more time consuming than direct database searches, but it has the additional flexibility of being able to identify peptides that are not exact matches to existing peptides, permitting identification via similar sequences in homologous proteins. Hybrid approaches combine elements of database searching and de novo sequencing, often through identification of short sequence tags from the MS/MS data, which can be used to restrict subsequent database searches. This is particularly important if PTMs are being considered in the search, as this can greatly increase the search space. The use of sequence tags and similar approaches provide sufficient restrictions so that the search is of a manageable size. In all of the above approaches, identifications are typically done for large numbers of peptides, each representing individual identifications. As such, the scores and statistical significance need to be corrected for multiple hypothesis testing to avoid a high rate of false positive identifications. As with 2D gels, FDR corrections can be applied to estimate the number of false positives within the set of identified peptides.11 E-value scores are particularly amenable to this approach, as they do not require a score-specific function. In response to the complexity of the methods of data collection and analysis within a proteomics experiment, efforts have been made to standardize methods and analysis of proteomics data through the Human Proteome Organization ‘minimum information about a proteomics experiment‘ (MIAPE).12 The development of common standards will facilitate the collection, integration, storage, and dissemination of data and aid in comparison across experiments.

1.32.6

Applications in Biotechnology

Proteomics can help provide a more comprehensive view of biological systems than traditional biochemical approaches. Information on changes in protein expression, PTMs, and protein interactions is available for large numbers of proteins, providing information on dynamic changes within the cell. This comprehensive, or discovery-based, approach has less of a bias toward

Theory and Applications of Proteomics

455

existing knowledge and may be a valuable source of new information for strain improvement or process development. Importantly, proteomics techniques can identify novel gene and protein targets, besides providing additional information about known proteins and pathways. From a biotechnology perspective, this information can be used to identify protein targets for improved expression, localization, stable protein folding, and optimization of production. Genetic manipulation of microorganisms such as Eecherichia coli and eukaryotic expression systems including yeast and mammalian cell lines has markedly increased the production of medically and industrially useful proteins. At the genetic level, increased productivity can be accomplished primarily through changes in promoter sequences, signal sequences, and plasmid origin. These changes are reflected at the protein level and result in increased expression or improved protein localization and reduced degradation. A key strategy in increasing production has been the identification and development of high-expression promoters, motifs, and plasmids.13 This has been accomplished through trial and error or through integration of exogenous DNA into the cell. The use of -omics techniques provides a more informative and streamlined approach to the development of promoter systems. Using comparative proteomics analyses, proteins that are highly expressed under given conditions can be used to identify corresponding upstream promoter regions, particularly in prokaryotic organisms. This approach has been used to develop novel expression systems with increased productivity. Stress response pathways offer another venue for improving production of recombinant proteins. In general, chemical or physical stress induces protective mechanisms to minimize damage or remove an agent of stress. Recombinant protein production at high levels is likely to induce some level of stress response in the cell, which can be either beneficial or detrimental to production. Chaperone proteins plan a key role in the stress response. Many chaperone proteins aid in the correct folding or refolding of proteins, and induction of chaperones can increase the proportion of active (correctly folded) recombinant proteins in the cell. The chaperones GroEL, GroES, and DnaK have all been used to improve yields of soluble active protein.13 Other stress-induced proteins can reduce the rate of protein degradation by cellular proteases, or increase secretion of recombinant proteins. The production of recombinant proteins adapts the native machinery of the cell to produce a foreign protein, or increased amounts of a native protein. However, production remains dependent on the metabolic pathways and precursors required for protein synthesis. Bottlenecks in these pathways will result in decreased potential yields of bioproducts. Proteomics approaches are particularly useful for acquiring information on changes to alterations in protein expression, and this can be effectively applied to monitor changes in regulation and metabolism. Using this approach, previous work has identified decreases in the levels of certain amino acid biosynthesis enzymes, which provides an explanation for the increased recombinant protein production on addition of amino acids and nitrogen compounds. Proteomics can also be used to identify rate-limiting steps in metabolic pathways and identify targets for manipulation. Results of genetic manipulations can also be assessed via proteomics, providing more information than simple changes in net production. In addition to manipulation of proteins and pathways, proteomics techniques can be used to identify fusion partners for increasing protein solubility, and motifs that may be used as membrane anchors for cell-surface display. Applications of cell-surface display are expanding, and include biosensors, live vaccines, antibody production, and screening of peptide libraries. A promising fusion technique is the use of fusion partners to target proteins to excretory systems, allowing for increased ease of purification and a high proportion of correctly folded proteins. Additional strategies for optimization of production via changes to protein processing and transport within the cell are possible through manipulation of ancillary proteins. For example, variation in the expression of chaperones and protein export systems can greatly increase the production of stable recombinant products, through more efficient folding, localization, and protein export. The use of proteomics in biotechnology has the potential to provide a more mechanistic understanding of the processes involved in protein production, which can be utilized to develop more efficient systems for production of target proteins.

See Also: 1.50 Mass Spectrometry.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

Yates, J. R.; Ruse, C. I.; Nakorchevsky, A. Proteomics by Mass Spectrometry: Approaches, Advances, and Applications. Annu. Rev. Biomed. Eng. 2009, 11, 49–79. Whitelegge, J. P. Protein Mass Spectrometry, 1st ed.; Elsevier: Amsterdam, London, 2009. Wilkins, M. R., Appel, R. D., Williams, K. L., Hochstrasser, D. F., Eds.; Proteome Research: Concepts, Technology and Applications, 2nd ed.; Springer: Berlin, 2007. Schulze, W. X.; Usadel, B. Quantitation in Mass-spectrometry-based Proteomics. Annu. Rev. Plant Biol. 2010, 61, 491–516. Steen, H.; Mann, M. The ABC’s (And XYZ’s) of Peptide Sequencing. Nat. Rev. Mol. Cell Biol. 2004, 5 (9), 699–711. Cheng, Z.; Woody, O. Z.; Glick, B. R.; McConkey, B. J. Characterization of Plant–bacterial Interactions Using Proteomic Approaches. Curr. Proteomics 2010, 7 (4), 244–257. Fang, Y.; Robinson, D. P.; Foster, L. J. Quantitative Analysis of Proteome Coverage and Recovery Rates for Upstream Fractionation Methods in Proteomics. J. Proteome Res. 2010, 9 (4), 1902–1912. Timms, J. F.; Cramer, R. Difference Gel Electrophoresis. Proteomics 2008, 8 (23–24), 4886–4897. Bantscheff, M.; Schirle, M.; Sweetman, G.; et al. Quantitative Mass Spectrometry in Proteomics: A Critical Review. Anal. Bioanal. Chem. 2007, 389 (4), 1017–1031. Mueller, L. N.; Brusniak, M. Y.; Mani, D. R.; Aebersold, R. An Assessment of Software Solutions for the Analysis of Mass Spectrometry Based Quantitative Proteomics Data. J. Proteome Res. 2008, 7 (1), 51–61. Nesvizhskii, A. I.; Vitek, O.; Aebersold, R. Analysis and Validation of Proteomic Data Generated by Tandem Mass Spectrometry. Nat. Methods 2007, 4 (10), 787–797. Taylor, C. F.; Paton, N. W.; Lilley, K. S.; et al. The Minimum Information about a Proteomics Experiment (MIAPE). Nat. Biotechnol. 2007, 25 (8), 887–893. Han, M. J.; Lee, S. Y.; Koh, S. T.; et al. Biotechnological Applications of Microbial Proteomes. J. Biotechnol. 2010, 145 (4), 341–349.

1.33 Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds D Na, MY Kim, JY Park, and SY Lee, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea © 2014 Elsevier B.V. All rights reserved. This is a reprint of D. Na, M.Y. Kim, J.Y. Park, S.Y. Lee, 1.35 - Systems Metabolic Engineering for the Production of Non-innate Chemical Compounds, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 471–482.

1.33.1 1.33.2 1.33.2.1 1.33.2.2 1.33.2.2.1 1.33.2.2.2 1.33.2.2.3 1.33.2.3 1.33.2.3.1 1.33.2.3.2 1.33.2.3.3 1.33.2.4 1.33.2.4.1 1.33.2.4.2 1.33.2.4.3 1.33.3 References

Introduction and Scope Systems Metabolic Engineering Strategy Reorganization of Existing Metabolic Networks Omics Analysis In Silico Approach Case Studies Synthetic Pathway Construction Synthetic Pathway Construction From Heterologous Enzymes Synthetic Pathway Optimization Case Studies Synthetic Organisms Genome Reduction Genome Synthesis Case Studies Summary

456 457 457 457 457 457 459 460 460 461 462 465 465 465 465 466 467

Glossary Metabolic engineering Engineering of cellular metabolism to reallocate cellular resources toward a specific goal, usually to optimize the production of a desired chemical substance. Synthetic biology Engineering of cellular systems in analogy with the system design principles of engineering. Synthetic biology aims at creating synthetic genetic or metabolic networks exhibiting novel biological functions and at creating synthetic organisms. Systems biology Analysis of cellular systems in an interdisciplinary manner, incorporating knowledge from different fields such as biology, computer science, and nanotechnology. Systems biology aims to unravel the underlying principles of biological systems in a systematic way and makes use of omics data, including gene expressions, protein concentrations, and metabolic fluxes. Systems metabolic engineering Systematic and rational design of cellular systems based on the comprehensive understanding of cellular systems for the production of a desired chemical substance.

1.33.1

Introduction and Scope

In biorefineies, microorganisms have attracted a great interest due to their ability to produce useful chemicals from renewable biomass. With the advances in genetic engineering and analytical technologies, the internal mechanisms of genetic regulations and metabolic reactions were better understood and have given hints for identifying target components within the cell to improve the production of a target compound from a specific feedstock (e.g., glucose). Early efforts in metabolic engineering of microorganisms have reported a number of success stories of improved production of target compounds. However, engineering of microorganisms into microbial factories to efficiently produce the target products is difficult because living organisms, even a single-cell organism, are composed of highly complex intracellular systems with sophisticated interactions between the different intracellular systems to perform important biological functions, such as generating energy and adapting to environmental changes. Therefore, without a comprehensive understanding of the biological systems, it is difficult to identify which genes are to be manipulated and to predict the outcomes from the modification of a single gene. Recent advances in genetic technology, computational biology, and nanotechnology have enabled us to unravel the sophisticated cellular mechanisms of genetic regulation and metabolic fluxes. These advances have provided clues and tools that allow

456

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00029-X

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

457

us to redesign the internal networks of microorganisms to transform them into microbial cell factories. Owing to these advances, metabolic engineering strategies are moving from the classical ad hoc manner toward a systematic manner, where the target genes are identified rationally, considering the cellular system as a whole. Recently, the shift in strategies has been further accelerated by synthetic biology, which is capable of creating and reorganizing cellular machineries, allowing microorganisms to produce non-innate chemicals that they inherently cannot produce. Our aim in this article is to describe the strategies adopted in systems metabolic engineering to harness the power of systems and synthetic biology for the production of value-added chemicals. We focus on how the systems metabolic engineering strategies are used to engineer microorganisms to produce novel and valuable chemical substances. We shall briefly introduce the overall strategy and describe the required methodologies to deal with the hurdles in each step of the process. In addition, successful metabolic engineering cases selected from the literature are described to show how these methodologies were applied.

1.33.2

Systems Metabolic Engineering

1.33.2.1

Strategy

The strategies adopted in engineering microorganisms to produce noninnate and valuable chemicals involve reorganization of the existing metabolic and regulatory networks to maximize the flux toward precursors of the desired target substance, and reconstruction of the heterologous metabolic pathways to convert the precursor to the desired chemical compound (Figure 1). The reorganization of metabolic networks is supported by systems biology, which aims at unraveling the hidden principles of cellular functions in a high-throughput manner in assistance with nanotechnology and computer science. As the scale of the cellular networks to consider for engineering is beyond our imagination, in silico approaches have been of great help in identifying the target genes to be engineered by simulating the global genome-scale metabolic network. The reconstruction of heterologous metabolic pathways is supported by the recently emerging field of synthetic biology. Heterologous metabolic pathways, commonly termed synthetic pathways, are intended to convert an intracellular metabolite into a noninnate chemical substance through artificially constructed chemical reaction cascades using enzymes originated from various living organisms.

1.33.2.2 1.33.2.2.1

Reorganization of Existing Metabolic Networks Omics Analysis

After deciding on a target product to be produced, a proper host organism is selected in consideration of the physical and chemical properties of the target product. The host organism can be a well-established engineering organism, such as Escherichia coli or Saccharomyces cerevisiae, or a less-studied organism with superior properties that make it more suitable as a host organism. If the physiology of the host organism is not well understood, then its physiological properties, including genetic regulatory networks or metabolic pathways, should be unraveled first. Then the genes or proteins involved in the same functional category could be identified and the genes to be engineered could be selected accordingly. Genome-wide analysis of intracellular biological systems is required to unravel the underlying principles of biological systems in a systematic way using omics data, including gene expressions, protein concentrations, and metabolic fluxes. The analysis of genomic-scale expression profiles allows for the understanding of genetic regulatory mechanisms, and the approach is applied to identifying genes operating in the same pathway for a specific physiological behavior and metabolic genes for producing useful metabolites.1,2 Proteomic analysis also provides system-wide changes of cellular proteins in response to a specific intracellular and extracellular change and assists in identifying suitable target proteins for productivity improvement. Proteomic analysis was applied during the fermentation of E. coli to discover a target gene enhancing the production of a recombinant antibody fragment.3 Fluxome, a collection of internal metabolite fluxes, could also be measured through the use of isotopomers. Though the target fluxes determined were limited to the relatively smaller-scale metabolic pathways, they provided information on the changes to the internal metabolite concentrations with the assistance of computational tools and gave a clear evidence of the activity of metabolic pathways. Recently, the reaction type and kinetics of whole-cell metabolic enzymes, namely, reactomes, was measured using metabolite-bound nanoparticles on an array. This information allows for the identification of metabolic phenotypes and networks of organisms, even for those whose genomic sequence is unknown.4 The reactome array was applied to confirm the global metabolic network of Pseudomonas putida and Streptomyces coelicolor and to discovering the networks of three microbial communities derived from acidic volcanic pool, deep-sea brine lake, and hydrocarbon-polluted seawater.

1.33.2.2.2

In Silico Approach

A single-cell organism is composed of highly complex intracellular systems with sophisticated interactions between the systems to generate energy, adapt to environmental changes, and perform a number of other important biological functions. It is difficult and beyond our capability to identify the genes or proteins to manipulate and to predict the outcomes from the modification of even a single gene in the same metabolic system. Therefore, target genes and proteins to be manipulated should be identified in consideration of genetic and metabolic networks in a global context. For such purposes, the in silico model, representing cellular interactions expressed in mathematical form, is constructed to simulate cellular behaviors under the condition of genetic modifications. Conversion of cellular systems into in silico models involves the transformation of topological interaction structures into a mathematical form by incorporating reaction parameters, environmental boundaries, and physicochemical conditions. The models, in short, describe what machineries are operating and

458

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

Microbial factory

Synthetic metabolic pathway reconstruction Pathway optimization Enhancement of catalytic activity through directed evolution Balancing fluxes by balancing the protein expression level of enzymes Enhancing enzyme reactions by proximity optimization

Cellular network reorganization In silico analysis S·ν = 0 dxi f (xi ,t) dt

Prediction of genetic engineering effects on productivity Identification of genes to be removed Identification of genes to be amplified

Pathway reconstruction A B A B

Y Y

Reconstruction of metabolic pathways from heterologous enzymes Building feasible pathways using kinetic/ chemical information of enzymes

Omics analysis Integration of omics data Construction of networks Identification of genes in a target pathway Understanding of cell’s physiology

Classical random engineering Evolution Repeated mutagenesis by treating with mutagen and selecting the ones with higher productivity

Microorganism in nature Figure 1 A systems metabolic engineering strategy for the production of noninnate but value-added chemical substances in microorganisms. A general strategy intended to produce a noninnate chemical substance in microorganisms is illustrated. An appropriate host organism should be selected in consideration of its physiological properties. Unless genotypic and phenotypic information of the organism is available for engineering, the organism is evolved through repeated processes of mutagenesis and selection using chemical mutagens. For advanced and efficient organism engineering, various high-throughput technologies are exploited to discover genetic and physiological properties and thereby to identify potential target genes. As cells are built with a great number of interacting components, it is hardly possible to predict the outcomes of a single gene modification. In silico approaches are taken to identify target genes by analyzing outcomes of genetic modifications in a global context. These approaches would help to enhance the flux to the precursor of a target product. The next step is to build a synthetic metabolic pathway driving the precursor to a target chemical substance by integrating enzymes from heterologous organisms. The first attempts using a synthetic pathway may fail to achieve high productivity because of flux imbalance caused by a bottleneck enzyme which might have a relatively low catalytic activity or be expressed at a low level. The expression of each enzyme in a synthetic pathway can be optimized in consideration of their expression rates. Protein engineering gathering the enzymes into a close proximity, mimicking substrate tunneling, would further enhance the productivity of noninnate chemicals by reducing the loss of intermediates by diffusion or by consumption by competing pathways. Through these processes of engineering, a microbial factory with the highest productivity can be constructed.

how fast they operate in response to internal or external changes. With an objective function that maximizes the production yield of a target product with several constraints, including biomass demand, computation of the models can find candidate genes to be manipulated so that the maximum production yield can be increased by the modification of the candidates. A widely used approach for modeling intracellular activities, specifically metabolic pathways, is flux balance analysis (FBA).5 In FBA, topological metabolic pathways and the stoichiometry of enzymatic reactions are modeled into a mathematical form and the flux distribution of the pathways is calculated in the assumption of a pseudo-steady state for the reactions. The FBA approach was used in the engineering of Clostridium acetobutylicum for selective butanol production and for investigating the effect of host–pathogen interaction (E. coli–MS2 phage) on metabolism.6,7 FBA with modifications has also been used to understand the regulatory circuitry of Pichia pastoris by measuring intracellular metabolic fluxes through the central carbon pathways in the production of recombinant human erythropoietin.8 Although FBA utilizes a pseudo-steady state assumption, the internal state of components in living organisms is in continuous fluctuation over time. To model the dynamic phenomena of living organisms, several kinetic models using the formulas of ordinary

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

459

differential equations have been developed. These models enable the simulation and tracing of the temporal changes in the metabolic fluxes regulated by genetic interactions.9 However, dynamic models rely heavily on a large number of precise parameter values, which are not available for every metabolic reaction, and therefore the current dynamic models focus on small-scale networks.10 To take advantage of both these approaches, significant genetic regulatory systems are integrated into metabolic models to simulate physiological states of the cells under diverse environments and to dynamically simulate metabolic activity regulations. This hybrid approach was applied to the E. coli in silico model to understand the dynamic regulations in response to diverse carbon limitations and to predict the gene knockout effects.11,12

1.33.2.2.3

Case Studies

1.33.2.2.3.1 Amino Acid Production The primary target metabolites in metabolic engineering are in general components of food additives, medical substances, fuels, and other materials. Among them, L-valine, an essential amino acid widely used as a component of pharmaceuticals and cosmetics, as well as an animal feed additive, has been reported to be overproduced in E. coli using systems metabolic engineering (Figure 2).13 Although E. coli is a widely used strain for amino acid production, the complex regulations in L-valine biosynthesis had obscured the engineering of E. coli as a host strain for L-valine production. According to the report, despite the difficulties, E. coli was successfully engineered with the assistance of the combined approach of transcriptome analysis and computational gene knockout simulations. First, the selected host strain was rationally engineered by manipulating the genes related to L-valine biosynthesis. Further, the target genes were identified by comparative transcriptome analysis and genome-wide in silico gene knockout simulations. This stepwise metabolic engineering process successfully achieved efficient production of L-valine in E. coli. The target genes responsible for the negative regulations mediated by L-valine were removed by site-specific genome engineering. The feedback inhibition in the small subunit of acetohydroxy acid synthase (AHAS) isoenzyme III (encoded by ilvH) and transcription attenuation regulations in the leader region of the ilvGMEDA and ilvBN operons were removed to eliminate a major bottleneck

Glucose Pentose phosphate pathway

pfkA

ilvA

Pyruvate

mdh aceF TCA

L-Threonine

2-Ketobutyrate

Acetate

ilvlH ilvBN

ilvGM

2-Acetolactate 41st G 50th C

A T

ilvC L-Isoleucine

2,3-Dihydroxyisovalerate ilvD 2-Ketoisovalerate leuA

panB

LRP ilvE Pantothenate L-Leucine L-Valine

Targets rationally selected yga ZH

Targets identified by transcriptome analysis Targets identified by in silico simulation

Figure 2 Systems metabolic engineering approach for L-valine production. Feedback inhibitions by L-valine were disrupted, genes in TCA and glycolysis were eliminated to increase the pyruvate pool, and genes in competing pathways were also eliminated to enhance the flux from pyruvate to L-valine. Identified target genes are indicated: rationally identified targets (bulb icons), targets identified by transcriptome analysis (DNA chip icons), and targets identified by in silico simulation (computer icons).

460

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

for L-valine biosynthesis. To direct more carbon flux toward L-valine production, the genes involved in competing pathways were disrupted. In addition, AHAS I (encoded by ilvBN) was selected for amplification in order to increase the first metabolite (2-acetolactate) for L-valine. All the genetic targets in the first step were related to L-valine biosynthesis. For further identification of additional targets and understanding of the alteration of global gene expression levels, comparative transcriptome analysis of the engineered host strain and wild-type strain was performed. The comparative analysis showed that the expression levels of the genes involved in L-valine biosynthesis were upregulated (ilvC, ilvE, and ilvD) in the recombinant host strain but the extents of increase of ilvCDE were less than those of ilvBN. These results indicate that ilvCDE should be amplified in order to balance the flux to L-valine. The transcriptome analysis also showed the downregulation of the lrp gene, which activates the expression of ilvIH (encoding AHAS III). To overcome the ilvIH expression problem, the lrp gene was amplified. In addition to the lrp gene, the expression of ygaZH, playing a role as an L-valine exporter, was also down regulated. As the accumulation of L-valine inside a cell might limit its production, the ygaZH genes were amplified as well. The additional amplification of the ilvCED, lrp, and ygaZH genes increased the production of L-valine by 113%. For further improvement, the authors performed in silico gene knockout simulations and as a result identified the target genes to be deleted: aceF, pfkA, and mdh. The deletion of aceF and mdh increased the pyruvate pool directly and indirectly, respectively. The deletion of pfkA increased the availability of NADPH, which is an important cofactor for L-valine production. As a result of the comprehensive engineering, the host strain was able to produce 7.55 g l 1 L-valine from 20 g l 1 glucose, an impressively high yield of 0.378 g L-valine per gram glucose.

1.33.2.2.3.2 Lycopene Production For rapid targeted metabolic engineering for strain improvement, it is necessary to adopt rational and systematic approaches. Genome-scale in silico models can serve as a platform for predicting physiological outcomes in response to perturbations. Owing to their prediction capability, in silico models are commonly used for the identification of promising target genes. In silico identification of target genes for enhanced yield is presented in this section (Figure 3).14 A genome-wide stoichiometric FBA was performed to identify single- or multiple-gene knockout targets in E. coli that improved lycopene yield, while maintaining an acceptable growth rate. Specifically, metabolic fluxes were computed to optimize cell growth rate using the genome-wide biochemical reaction network of E. coli in order to identify potential target genes for manipulation leading to improved lycopene production. A limitation of FBA is that the calculated fluxes are not necessarily the same as the in vivo fluxes, especially for genetically perturbed organisms, in which the resulting fluxes are often suboptimal in growth rate and target metabolite production rate. To overcome this problem, flux profiles were optimized using a different objective function: the minimization of metabolic adjustment required between the wild-type strain and the genetically modified strain. The resulting flux profiles are between the wild-type optimal and the gene knockout mutant optimal. The in silico simulation identified promising genetic targets expected to enhance lycopene production: gdhA, gpmA/gpmB, aceE, fdhE, and talB. The effects of single genetic knockouts of the identified genes were investigated, and as a result gdhA was predicted to show the best increase in lycopene production. Therefore, based on the gdhA-knocked-out strain model, a second target gene was also searched using the simulation. This process sequentially identified multiple genetic knockout targets, which were gdhA, aceE, and fdhF. This triple knockout resulted in the increase of lycopene production by 40% over that of its parental strain.

1.33.2.3 1.33.2.3.1

Synthetic Pathway Construction Synthetic Pathway Construction From Heterologous Enzymes

Genome-wide analysis and in silico modeling are intended to optimize global metabolic fluxes to increase productivity within the inherent production capability of the host organism. For novel chemical production, these approaches can be applied to enhance the production of a precursor. Then the precursor can be transformed into a noninnate chemical substance through an artificially constructed reaction cascade composed of enzymes from heterologous organisms. In recent years, synthetic biology has gained prominence in metabolic engineering. Synthetic biology is expected to be capable of enhancing inherent production capability by empowering host microorganisms to gain new abilities, which are not inherent in the organisms, for example, noninnate compounds production. Synthetic metabolic pathways, which are cascades of reactions mediated by enzymes originated from various organisms, can be rationally reconstructed. To date, synthetic metabolic pathways redirecting the cellular fluxes to the formation of noninnate target compounds have been developed and applied to the production of plastics, biofuels, and therapeutics from renewable sources. Implementation of a synthetic metabolic pathway carrying out cytochrome P450-based oxidation in vivo enables E. coli to produce intermediates of plant-derived antimalaria terpenoids.15 Synthetic pathways implemented in E. coli and composed of metabolic enzymes originated from several microorganisms produced noninnate alcohols or isopropanol with a higher titer than native producers.16,17 E. coli cells incorporated with a synthetic regulatory module, which controls the balance between lycopene production (target material) and acetate production (waste material) dynamically in response to acetate concentration, showed an increased production of lycopene as compared to uncontrolled cells.18 Production of a promising bioplastic polymer in E. coli was improved by a synthetic metabolic pathway for polylactic acid production and metabolism optimization with the aid of genome-scale in silico analysis.19

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

461

Glucose

Glyceraldehyde-3-phosphate talB 3-Phosphoglycerate gpmA

Pentose phosphate pathway

gpmB

2-Phosphoglycerate

Lycopene

Phosphoenolpyruvate

Pyruvate aceE

fdhF

PP c Acetyl-CoA

Formate

GLU

α-Ketoglutarate

Oxaloacetate TCA

CO2 + H2O

gdhA

Figure 3 Genetic knockout targets identified by in silico simulations for improved lycopene production. The metabolic pathway in E. coli and the genetic knockout targets identified by an in silico approach are illustrated. Among the target candidates, the triple knockout combination of aceE, gdhA, and fdhF resulted in up to 40% increase of lycopene production.

1.33.2.3.2

Synthetic Pathway Optimization

Reconstructed synthetic metabolic pathways might show relatively low production yield. The gene expression levels and kinetic activities of metabolic enzymes in nature have been evolutionarily optimized for the efficient conversion of metabolites, whereas the heterologous enzymes in synthetic pathways have not been optimized in nature and therefore might show flux imbalance because of the presence of a rate-limiting enzyme exhibiting a relatively low activity or being expressed at a lower level as compared to the other enzymes in the pathway. Synthetic pathway optimization aims at enhancing the rate of conversion of the precursor to the final target substance without any interference from rate-limiting enzymes. To avoid such flux imbalances, a reasonable reaction cascade should be reconstructed taking into consideration chemical structural changes, reaction mechanisms, and enzyme information. For such calculations, a bioinformatics approach can be used for finding the optimal combination of enzymes from databases.20 If the bottleneck in the production of the target product results from a relatively low expression of an enzyme, control of the expression of enzymes can be achieved computationally. In various genetic engineering reports, employing an alternative protein expression cassette with a different promoter strength is widely used but might fail to work due to the change of the secondary structure of the ribosome binding site by a downstream coding sequence, leading to a change in protein expression level. The recently reported computational approaches to design synthetic ribosome binding sites could be applied to control the expression levels of enzymes in synthetic pathways.21,22 If the bottleneck in the production of the target product results from an inherently low catalytic activity, the activity can be enhanced through evolutionary mutagenesis and selection. In mutagenesis, mutations are artificially introduced using chemical mutagens, error-prone PCR, or DNA shuffling. For selecting the mutant with enhanced performance, proteins produced from the mutated nucleotide sequences are screened by measuring their enzymatic activities. To date, several advanced mutationinducing technologies have been developed for enhancing the mutation rate, but the difficulties in measuring the catalytic activity from numerous mutants, which might differ enzyme to enzyme, limit the rapid evolutionary development of proteins. Recently,

462

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

a novel kinetic optimization method of the synthetic pathway has been reported based on the fact that enzymes in the same pathway are evolved to stick to each other to use the substrates quickly, a phenomenon commonly known as substrate tunneling.23 The tunneling phenomenon can be mimicked in such a way that enzymes fused with a protein-binding domain associate onto a scaffold protein. This association forces the enzymes into physical proximity and thereby dramatically enhances productivity of the synthetic pathways.

1.33.2.3.3

Case Studies

1.33.2.3.3.1 Algorithm for Synthetic Pathway Reconstruction The rapidly accumulating information on proteins of various sequenced organisms and enzymatic reaction information make it possible to reconstruct novel synthetic metabolic pathways, which mediate enzymatic reactions by redirecting intracellular metabolites to noninnate chemical substances. There have been several computational algorithms for the reconstruction of synthetic metabolic pathways leading to the production of a desired chemical substance. However, as the algorithms produce a huge number of possible synthetic metabolic pathways, it is difficult to validate and apply the predicted pathways in real experiments. A recently reported algorithm reduces the number of potential synthetic pathways by scoring the predicted pathways depending on chemical structural changes, enzyme characteristics, and reaction mechanisms (Figure 4).20 The authors proposed a pathway prediction system framework and applied the framework to the reconstruction of the isobutanol biosynthesis pathway. The framework applies predefined reaction rules to generate possible reaction paths from various substrates to the target compound. Among the possible substrates, 2-ketoisovalerate was selected as a starting compound of the synthetic metabolic pathway. Subsequently, two base routes from 2-ketoisovalerate to isobutanol were selected. The reactions in the base routes, converting 2-ketoisovalerate to intermediates and the intermediates to isobutanol, were assigned to real enzymes. The possible enzyme combinations were scored based on five priority factors: binding site covalence, chemical similarity, thermodynamic favorability, pathway distance, and organism specificity. Binding site covalence describes the substructural similarity of changed chemical structures. Chemical similarity represents the similarity of entire molecular structures. For two given reactions, as the structural differences get smaller, the two reactions are closer and thereby the enzyme–reaction combination is more preferred. Thermodynamic favorability describes the Gibbs free energy of formation representing the thermodynamic feasibility of reactions. An increase in the Gibbs free energy in a reaction indicates that more energy is required for activating the

A

B (1) Binding site covalence

2nd reactions

10 3

Route 1 (blue)

H C

O

O

OH

p

OH

OH

Reaction 1

OH

Unknown

1st reactions

0

O

OH

ΔG(kcal/mol)

3

H C O HO

1

−10

OH

HO

O

CH3 O

C01412 OH

H3C

CH3

CH3 OH

6

Route 2 (red)

Reaction steps

O

(4) Pathway distance

(2) Chemical similarity

O

B

A

Reaction 2 C

(5) Organism specificity

OH

H3C

Isobutanol

2-ketoisovalerate

5

−30

O C00141

OH

4

HO O

H H3C

3

CH3 CH3

CH3

H C00479

2

−20

−40

CH3

O

H3C

OH C05979

H3C C14710

(3) Thermodynamic favorability

O

Proteobacteria

Unknown

Epsilonproteobacteria Gammaproteobacteria Alphaproteobacteria

Figure 4 An algorithm for rational design of synthetic metabolic pathways. The algorithm searches for possible reaction routes from every starting compound to the target compound (isobutanol). Among the possible reaction routes, a known compound was set as a starting one (2-ketoisovalerate) and two base routes were found (the two paths in the dotted box). (A) Reactions in the base routes are then assigned to real enzymatic reactions, and among the possible combinations of enzymatic reactions, candidate reaction cascades (synthetic pathways) are ranked depending on the five prioritization factors (B): (1) Binding site covalence describes substructural similarity of changed chemical structures and (2) chemical similarity describes the similarity of entire molecular structures. As the structural differences get smaller, the two reactions are closer and thereby the enzyme–reaction combination is preferred. For (3) thermodynamic favorability, reaction cascades with continuously decreasing Gibbs free energy or those with less fluctuating free energy are favored (Route 2 (red) is preferred over Route 1 (blue)). For (4) pathway distance, enzymes found at a closer distance in a metabolic pathway are preferred (A–C enzyme pair is preferred to A–B enzyme pair). For (5) organism specificity, enzymes found in close organisms are preferred. The tree represents lineage relationships among organisms. Any given enzyme in Alphaproteobacteria is closer to an enzyme in Proteobacteria than to those in Gammaproteobacteria or in Epsilonproteobacteria.

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

463

reaction, and thereby the reaction might slow down the overall reaction cascade due to the energy barrier. Therefore, a route with a continuously decreasing Gibbs free energy or a low Gibbs free energy fluctuation is favored. Pathway distance represents the shortest distance between two given enzymes in a real pathway. The shorter the pathway distance of two enzymes is, the more closely the enzymes are related. As the pathway distance of two enzymes has been discovered to indicate genomic relation between their genes, enzymes with a shorter pathway distance are favored. Similarly, organism specificity represents the distance between the two organisms to which the enzymes belong. Enzymes from close organisms are preferred in a synthetic pathway. The framework suggested prominent synthetic pathway candidates in consideration of the above factors. The candidate synthetic pathways for isobutanol showed a high reliability: the experimentally verified synthetic pathways were listed within the top 0.089% of the pathway candidates. The developed framework would be useful for the in silico design of novel metabolic pathways to be employed for the efficient production of noninnate chemical compounds. 1.33.2.3.3.2 Synthetic Pathway Reconstruction Annually about 1 million people die of malaria, and the appearance of multidrug-resistant strains of the malaria parasite Plasmodium falciparum makes it difficult to control the disease. Artemisinin is highly effective against the multidrug-resistance of Plasmodium spp. but is in short supply because artemisinin is extracted from Artemisia annua L (sweet wormwood). A reported method to produce the precursor of artemisinin in S. cerevisiae by employing a synthetic metabolic pathway is presented in Figure 5.24 As S. cerevisiae is incapable of producing artemisinic acid, it was engineered to obtain the capability. A synthetic metabolic pathway converting an intracellular metabolite, farnesyl pyrophosphate (FPP), to artemisinic acid was employed by cloning the

Glycolysis

Genes

Acetyl-CoA ERG10

HMGR

Encodes 3-hydroxy-3-methyl-gluarylcoenzyme A reductase (tHMGR) Overexpressed with truncated form by chromosomal integration

ERG9

Encodes squalene synthase Downregulated by using a methionine repressible promoter

UPC2

Encodes global transcription factor regulating the biopsynthesis of sterols in S. cerevisiae Overexpressed by chromosomal integration

Amorpha-4,11-diene Acetoacetyl-CoA ERG13

H

ADS

HMG-CoA

H

tHMGR Mevalonate upc2-1

ERG12

CYP71AV1/CPR

Mevalonate-P Artemisinic acid

ERG8 Mevalonate-PP

ERG20

IPP

H

DMPP

ADS

HO O

Semisynthesis

ERG20 GPP

CYP71AV1

ERG20 Met

Encodes FPP synthase Overexpressed by chromosomal integration

H

ERG19 IDII

Description

Artemisinin

FPP erg9

Encodes amorphadiene synthase from A: annua Heterologously expressed under the control of the GAL1 promoter on the pRS425 plasmid Encodes cytochrome P450 monooxygenase from A. annua Heterologously expressed under the control of the galactoses-inducible promotor

H O

Squalene O H

O H

CPR

O H O

Ergosterol

Encodes NADPH: cytochrome P450 oxidoreductase from A. annua Heterologously expressed under the control of the galactose-inducible promotor

Figure 5 Schematic representation of the engineered artemisinic acid biosynthetic pathway in S. cerevisiae. Both the synthetic artemisinin biosynthesis pathway constructed in S. cerevisiae and the genetic manipulations performed in the study are shown. The farnesyl pyrophosphate (FPP) pool was increased through the engineering of the FPP biosynthetic pathway and then FPP was converted to the precursor of artemisinin, artemisinic acid, by the synthetic artemisinic acid biosynthetic pathway containing amorphadiene synthase, a novel cytochrome P450, and its redox partner from Artemisia annua. Truncated HMGR (tHMGR) and erg20 were amplified, while erg8, erg12, and erg13 were indirectly amplified by the incorporation of a semidominant mutant of upc2-1. The erg9 was downregulated by replacing its promoter. The pathway intermediates IPP, DMAPP, and GPP denote isopentenyl pyrophosphate, dimethyl allyl pyrophosphate, and geranyl pyrophosphate, respectively. The synthetic pathway containing ADS, CYP71AV1, and CPR was employed from A. annua.

464

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

amorphadiene synthase (ADS) and cytochrome P450 (CYP71AV1/CPR) genes. The ADS gene from A. annua was cloned to convert FPP to amorphadiene. The cytochrome P450 that performs a three-step oxidation of amorphadiene to artemisinic acid was cloned from A. annua. For the functional expression of CYP71AV1, its native redox partner, NADPH:cytochrome P450 oxidoreductase (CPR), was also cloned from A. annua. To improve the overall artemisinic acid production, the FPP biosynthetic pathway was engineered to accumulate a pool of FPP, a precursor for amorphadiene, while decreasing the flux from FPP to sterols. Several genes in the FPP biosynthetic pathway were amplified including a truncated form of 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (tHMGR), FPP synthase (ERG20), and a semi-dominant mutant of upc2-1 increasing the expression of erg13, erg12, and erg8. The expression of erg9 consuming FPP was downregulated. As a result of this metabolic engineering, the engineered yeast cell was able to produce artemisinic acid efficiently. The employment of a synthetic metabolic pathway empowered S. cerevisiae to produce artemisinic acid, and the metabolic engineering of intracellular fluxes toward the accumulation of the precursor for the synthetic pathway has successfully produced the value-added antimalaria drug with a high yield. 1.33.2.3.3.3 Synthetic Pathway Optimization Construction of synthetic pathways using enzymes originated from different organisms often lacks the regulatory systems to control the fluxes and thereby results in flux imbalance, leading to a reduced yield. The flux imbalance within the synthetic pathway might result from the presence of a bottleneck enzyme which inevitably slows down the overall reaction rate, because either the expression level of the bottleneck enzyme might be relatively lower or the enzyme might possess lower catalytic activity than the other enzymes in the pathway. Therefore, the optimization of enzyme expression and catalytic activity is key to employing synthetic metabolic pathways. There has been a report describing a strategy using a scaffold that optimizes the proximity between enzymes and the stoichiometry of the enzymes as well (Figure 6).23 A synthetic pathway composed of three enzymes was optimized, which converts acetyl-CoA to mevalonate via the enzymes acetoacetyl-CoA thiolase (AtoB) originated from E. coli, hydroxymethylglutaryl-CoA synthase (HMGS) from S. cerevisiae, and hydroxymethylglutaryl-CoA reductase (HMGR) from S. cerevisiae. Each enzyme was fused with a protein-binding domain to bind to the scaffold protein: AtoB binding to the GTPase-binding domain (GBD), HMGS to the Src homology 3 (SH3) domain, and HMGR to the PSD95/DlgA/Zo-1 (PDZ) domain. The three enzymes meet in the scaffold and mimic substrate tunneling effect in such a way that a product from one enzyme can be rapidly consumed as a substrate by the next enzyme without diffusing and being consumed by competing pathways. The stoichiometry of the synthetic pathway was also optimized by recruiting different number of enzymes to the scaffold. Among the various combinations of the stoichiometry of the enzymes, AtoBx1–HMGSx2–HMGRx2 showed the maximal increase in mevalonate production by about 77-fold. The same scaffold strategy was also applied to the synthetic pathways for glucaric acid, a potent building block for several polymers including nylons and hyperbranched polyesters. Scaffolding the bottleneck enzymes of the pathway (myo-inositol-1phosphate synthase and myo-inositol oxygenase) improved glucaric acid titer over a nonscaffold control. The scaffold-based optimization strategy can be applied to the optimization of diverse biosynthetic pathways. The strategy optimizes the intrinsic efficiency of the pathways but does not dynamically regulate the fluxes in a global context, which may lead to unfavorable perturbation and thereby suboptimal productivity in a host organism. Therefore, synthetic

Acetyl-CoA

AtoB

HMGS

HMGR

Mevalonate

Non-optimized pathway

Acetyl-CoA

AtoB

HMGS HMGS

HMGR HMGR

GBD

SH3

PDZ

Mevalonate

Optimized pathway

Figure 6 Spatial optimization of synthetic pathways. A synthetic protein scaffold provides a frame that clusters enzymes fused with a proteininteracting domain and thereby facilitates the conversion of metabolites by reducing the diffusion of intermediates. In the mevalonate synthetic pathway, the GBD, SH3, and PDZ domains recruit acetoacetyl-CoA thiolase (AtoB), hydroxyl-methylglutaryl-CoA synthase (HMGS), and hydroxymethylglutaryl-CoA reductase (HMGR), respectively.

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

465

regulatory systems, which dynamically control the fluxes and the expression of enzymes in response to global flux perturbations, would enable the operation of synthetic pathways in a desired way. The development of such synthetic regulatory systems is one of the great challenges in systems metabolic engineering.

1.33.2.4

Synthetic Organisms

For decades, scientists have attempted to build synthetic organisms, which have minimal but replicable genome. From the perspective of metabolic engineers, such synthetic organisms can provide a platform for building superior organisms that produce fewer byproducts, have increased genome stability, and tolerate the stress during metabolite production. There are two well-known approaches for synthetic organism construction: top-down (genome reduction) and bottom-up (genome synthesis).

1.33.2.4.1

Genome Reduction

Genome reduction is a process of eliminating nonessential genes. From the late 1950s, the genome reduction approach has been adopted to search for autonomous self-replicating minimal genomes of Mycoplasma genitalium, which has the smallest number of DNA sequences. Recently, genome comparison and transposon mutagenesis have revealed that 382 out of the 482 genes in M. genitalium are nonessential and thereby might be eliminated while sustaining its ability for replication.25 Of the nonessential genes 48% were hypothetical or encoding proteins of unknown function. The genome reduction approach has also been applied to the genome of E. coli. About 14.3% of the genes of the parental strain were successfully eliminated, and the resulting strain produced less unwanted byproducts and showed increased genome stability.26

1.33.2.4.2

Genome Synthesis

Genome synthesis is a bottom-up style construction of self-replicating organisms. Every single base is artificially synthesized and assembled into an intact genome. Recently, Graig et al. have succeeded in synthesizing an entire chromosome sequence and have successfully transplanted it into living cells to create Mycoplasma laboratorium. Such recent advances in synthetic biology enable us to design small-scale genetic systems exhibiting various biological behaviors. Synthetic biology is expected to make possible the engineering of cells to show more complex regulatory behaviors in response to intracellular and extracellular changes and thereby foster more dynamic biological behaviors of microorganisms to adapt to diverse physical or chemical changes. Now the wave of synthetic biology is accelerating toward creating higher levels of systems and synthetic biology, which, in conjunction with genome synthesis technology, open a new path to creating organisms for an enhanced productivity from the blueprint of DNA sequences.

1.33.2.4.3

Case Studies

1.33.2.4.3.4 Amino Acid Production on a Genome-Reduced Organism As microorganisms have been evolutionarily adapted to survive under various chemical and physical environments, microorganisms engineered as a microbial factory and intended to be cultivated in a laboratory might contain genes unnecessary for growth in the simplified culture environments and for the production of the desired chemicals. Therefore, elimination of the non-essential genes could improve the metabolic performance of strains by ways such as producing less unwanted byproducts, increasing genome stability, and streamlining metabolism without affecting the physiological compromise. E. coli is one of the widely used host strains in biotech industry to produce value-added chemical substances. A reduced-genome E. coli strain MDS42 whose genome was reduced by 14.3% as compared to its original strain MG1655 has been constructed by eliminating recombinogenic or mobile DNA and cryptic virulence genes (Figure 7).27 The genome-reduced MDS42 showed robust growth under normal laboratory conditions and showed unanticipated properties, such as high electroporation efficiency and robust propagation of unstable recombinant genes. The benefits of genome-reduced E. coli have been applied for the production of L-threonine.26 In order to compare the effects of genome reduction on the productivity of L-threonine, the genome-reduced E. coli MDS42 strain and the wild-type MG1655 strain were metabolically engineered in the same way. The feedback-resistant threonine operon (thrA*BC) was constitutively expressed. Threonine dehydrogenase (tdh) and threonine uptake proteins (tdcC and sstT), which consume L-threonine, were disrupted. A mutant threonine exporter (rhtA23) was overexpressed to pump out the accumulated L-threonine. The resulting L-threonine production strain (MDS-205) modified from the genome-reduced MDS42 strain showed twofold increase in production as compared with the engineered strain (MG-105) modified from MG1655 strain through the same modifications. The increased L-threonine productivity might have resulted from the reduced metabolic burden of maintaining unnecessary gene expressions, from enhanced carbon utilization and robustness, and from the redirection of the overflowed carbon metabolism in MDS42 strain into the production of L-threonine. Thus, microorganisms with a reduced genome can serve as a platform for microbial factory construction owing to their capability to improve metabolic performance by reducing byproducts and metabolic burden. 1.33.2.4.3.2 Synthetic Genome Since the completion of the human genome project in 2003, there has been an accelerated development of computational and experimental tools for understanding the genome of organisms. These advances have opened a new way for experimental paradigm shift from genetic modification of natural organisms toward chemical synthesis of designed genomes. Synthetic organisms have been created, including M. laboratorium and Mycoplasma mycoides JCVI-syn 1.0, that are able to replicate and propagate their genome to daughter cells.28,29 For the creation of M. mycoides JCVI-syn1.0, Craig Venter’s group had chemically

466

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

MDS42

Genome reduction

MG1655

Metabolic engineering

Metabolic engineering

Unnecessary genes Deleted genes Introduced plasmid

MG-105

MDS - 205

HO

HO HO

HO H N

O OH

H N

H N O OH

Threonine production

O HO OH H N

O OH

H N

O

H N

HO

OH O OH

Threonine production

Figure 7 Construction of threonine-producing E. coli strains from a wild-type strain or from a genome-reduced strain. E. coli strains (MG1655 and MDS42) were engineered to increase threonine yield through the same modifications. The feedback-resistant threonine operon (thrA*BC) was constitutively expressed. Threonine dehydrogenase (tdh) and threonine uptake proteins (tdcC and sstT) which consume L-threonine were disrupted. A mutant threonine exporter (rhtA23) was overexpressed to pump out the accumulated L-threonine. The resulting L-threonine production strain (MDS-205) modified from the genome-reduced MDS42 strain showed increased production by twofold as compared with the engineered strain (MG-105) modified from the MG1655 strain.

synthesized the 1.08 Mbp genome sequence using the natural genome of M. mycoides as a reference and incorporated designed watermark sequences, gene deletions, and polymorphisms into the genome. Briefly, several cassettes each with a length of 1 kb was chemically synthesized, and about 10 cassettes were assembled via homologous recombination in yeast to form a 10 kb synthetic intermediate. About 10 synthetic assembly intermediates with a length of 10 kb each were assembled again to produce 100 kb intermediates. The 11 resulting 100 kb assemblies were combined into a complete genome in yeast. The complete genome of M. mycoides JCVI-syn 1.0 was then transplanted into restriction-minus Mycoplasma capricolum in order to prevent the degradation of unmethylated donor DNA during transplantation by the restriction system of the recipient cell. The successful transplantation of the synthetic genome was confirmed using the primers specific to the watermark sequences. The Craig Venter group‘s approach to synthetic genome was quite different from that of conventional genome engineering: They created cells displaying wanted phenotypic properties. The creation of life from digital DNA information may foster the design of living organisms with all the capabilities required for a superior microbial factory: improved resistance to toxic intermediates, reduced metabolic burden, increased growth rates, and so on.

1.33.3

Summary

Microorganisms have become an attractive platform for the production of chemicals, fuels, and other materials from renewable resources. Despite such attractiveness, the difficulty in predicting the outcomes of genetic modification due to the complex networking of proteins within cells has forced researchers to adopt an ad hoc manner or random evolution to engineer microorganisms. Such approaches dampen the inspiration for further improvement of cellular performances. Consequently, a need for a rational engineering strategy to move toward constructing a microbial factory has surfaced. In this article, a systems metabolic engineering strategy for the production of noninnate compounds, from the systematic omics analysis to artificial synthetic pathway reconstruction, was presented with relevant examples. Systems metabolic engineering became possible owing to systems biology capable of systematic analysis and rational targeting of genetic modifications and synthetic biology capable of rational design of cellular systems empowering microorganisms to gain new physiological properties enhancing the overall productivity. The recent great leap in genome synthesis opens a new path to rational design of microorganisms using only DNA sequences, and the technology might even be able to resurrect non-existing organisms, whose chromosomal sequence is digitized, into a living organism. With more advanced systems and synthetic biology methodologies together with genome synthesis technology, systems metabolic engineering would evolve in such a way that host organisms are rationally designed and synthesized for microbial factories from nucleotide sequences rather than by modifying several genes of existing organisms.

See Also: 1.31 Metabolomics – The Combination of Analytical Biochemistry, Biology, and Informatics.

Systems Metabolic Engineering for the Production of Noninnate Chemical Compounds

467

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.

Alper, H.; Moxley, J.; Nevoigt, E.; et al. Engineering Yeast Transcription Machinery for Improved Ethanol Tolerance and Production. Science 2006, 314 (5805), 1565–1568. Alper, H.; Stephanopoulos, G. Global Transcription Machinery Engineering: A New Approach for Improving Cellular Phenotype. Metab. Eng. 2007, 9 (3), 258–267. Kuhner, S.; van Noort, V.; Betts, M. J.; et al. Proteome Organization in a Genome-reduced Bacterium. Science 2009, 326 (5957), 1235–1240. Beloqui, A.; Guazzaroni, M.-E.; Pazos, F.; et al. Reactome Array: Forging a Link between Metabolome and Genome. Science 2009, 326 (5950), 252–257. Kim, H. U.; Kim, T. Y.; Lee, S. Y. Metabolic Flux Analysis and Metabolic Engineering of Microorganisms. Mol. Biosystems 2008, 4, 113–120. Lee, J. Y.; Jang, Y.-S.; Lee, J.; et al. Metabolic Engineering of Clostridium Acetobutylicum M5 for Highly Selective Butanol Production. Biotechnol. J. 2009, 4 (10), 1432–1440. Jain, R.; Srivastava, R. Metabolic Investigation of Host/pathogen Interaction Using MS2-infected Escherichia coli. BMC Systems Biol. 2009, 3 (1), 121. Çelik, E.; Çalõk, P.; Oliver, S. G. Metabolic Flux Analysis for Recombinant Protein Production by Pichia pastoris Using Dual Carbon Sources: Effects of Methanol Feeding Rate. Biotechnol. Bioeng. 2010, 105 (2), 317–329. Wang, X.; Dalkic, E.; Wu, M.; Chan, C. Gene Module Level Analysis: Identification to Networks and Dynamics. Curr. Opin. Biotechnol. 2008, 19 (5), 482–491. Chassagnole, C.; Noisommit-Rizzi, N.; Schmid, J. W.; et al. Dynamic Modeling of the Central Carbon Metabolism of Escherichia coli. Biotechnol. Bioeng. 2002, 79 (1), 53–73. Lemuth, K.; Hardiman, T.; Winter, S.; et al. Global Transcription and Metabolic Flux Analysis of Escherichia coli in Glucose-limited Fed-batch Cultivations. Appl. Environ. Microbiol. 2008, 74 (22), 7002–7015. Covert, M. W.; Xiao, N.; Chen, T. J.; Karr, J. R. Integrating Metabolic, Transcriptional Regulatory and Signal Transduction Models in Escherichia coli. Bioinformatics 2008, 24 (18), 2044–2050. Park, J. H.; Lee, K. H.; Kim, T. Y.; Lee, S. Y. Metabolic Engineering of Escherichia coli for the Production of L-valine Based on Transcriptome Analysis and in Silico Gene Knockout Simulation. Proc. Natl. Acad. Sci. 2007, 104 (19), 7797–7802. Alper, H.; Jin, Y.-S.; Moxley, J. F.; Stephanopoulos, G. Identifying Gene Targets for the Metabolic Engineering of Lycopene Biosynthesis in Escherichia coli. Metab. Eng. 2005, 7 (3), 155–164. Chang, M. C. Y.; Eachus, R. A.; Trieu, W.; et al. Engineering Escherichia coli for Production of Functionalized Terpenoids Using Plant P450s. Nat. Chem. Biol. 2007, 3 (5), 274–277. Hanai, T.; Atsumi, S.; Liao, J. C. Engineered Synthetic Pathway for Isopropanol Production in Escherichia coli. Appl. Environ. Microbiol. 2007, 73 (24), 7814–7818. Zhang, K.; Sawaya, M. R.; Eisenberg, D. S.; Liao, J. C. Expanding Metabolism for Biosynthesis of Nonnatural Alcohols. Proc. Natl Acad. Sci. 2008, 105 (52), 20653–20658. Farmer, W. R.; Liao, J. C. Improving Lycopene Production in Escherichia coli by Engineering Metabolic Control. Nat. Biotechnol. 2000, 18 (5), 533–537. Jung, Y. K.; Kim, T. Y.; Park, S. J.; Lee, S. Y. Metabolic Engineering of Escherichia coli for the Production of Polylactic Acid and its Copolymers. Biotechnol. Bioeng. 2010, 105 (1), 161–171. Cho, A.; Yun, H.; Park, J.; et al. Prediction of Novel Synthetic Pathways for the Production of Desired Chemicals. BMC Systems Biol. 2010, 4 (1), 35. Salis, H. M.; Mirsky, E. A.; Voigt, C. A. Automated Design of Synthetic Ribosome Binding Sites to Control Protein Expression. Nat. Biotechnol. 2009, 27 (10), 946–950. Na, D.; Lee, S.; Lee, D. Mathematical Modeling of Translation Initiation for the Estimation of its Efficiency to Computationally Design mRNA Sequences with a Desired Expression Level in Prokaryotes. BMC Systems Biol. 2010, 4 (1), 71. Dueber, J. E.; Wu, G. C.; Malmirchegini, G. R.; et al. Synthetic Protein Scaffolds Provide Modular Control over Metabolic Flux. Nat. Biotechnol. 2009, 27 (8), 753–759. Ro, D.-K.; Paradise, E. M.; Ouellet, M.; et al. Production of the Antimalarial Drug Precursor Artemisinic Acid in Engineered Yeast. Nature 2006, 440 (7086), 940–943. Glass, J. I.; Assad-Garcia, N.; Alperovich, N.; et al. Essential Genes of a Minimal Bacterium. Proc. Natl. Acad. Sci. 2006, 103 (2), 425–430. Lee, J.; Sung, B.; Kim, M.; et al. Metabolic Engineering of a Reduced-genome Strain of Escherichia coli for L-threonine Production. Microbial Cell Factories 2009, 8 (1), 2. Posfai, G.; Plunkett, G. I. I. I.; Feher, T.; et al. Emergent Properties of Reduced-genome Escherichia coli. Science 2006, 312 (5776), 1044–1046. Gibson, D.; Glass, J.; Lartigue, C.; et al. Creation of a Bacterial Cell Controlled by a Chemically Synthesized Genome. Science 2010, 329 (5987), 52–56. Gibson, D. G.; Benders, G. A.; Andrews-Pfannkoch, C.; et al. Complete Chemical Synthesis, Assembly, and Cloning of a Mycoplasma Genitalium Genome. Science 2008, 319 (5867), 1215–1220.

1.34

Apoptosis: The Signaling Pathways and Their Control

TM Sauerwald, Centocor R&D, Inc., PA, United States A Lewis, IRP, NIDA, NIH, DHHS, Baltimore, MD, United States; and Johns Hopkins University, Baltimore, MD, United States H Dorai, Centocor R&D, Inc., PA, United States MJ Betenbaugh, Johns Hopkins University, Baltimore, MD, United States © 2011 Elsevier B.V. All rights reserved. This is a reprint of T.M. Sauerwald, A. Lewis, H. Dorai, M.J. Betenbaugh, 1.36 - Apoptosis: The Signaling Pathways and Their Control, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 483–494.

1.34.1 1.34.2 1.34.3 1.34.3.1 1.34.3.2 1.34.3.3 1.34.4 1.34.5 1.34.6 1.34.7 References

Introduction Apoptosis Regulators and Executioners Apoptotic Pathways Intrinsic Pathway ER Stress Pathway Extrinsic Pathway Apoptosis and Autophagy Inhibition of Apoptosis Apoptosis Affects Metabolic Pathways Conclusion

468 469 469 470 471 472 472 474 475 476 476

Glossary Knock-down Refers to techniques that reduce the expression of a particular gene(s). Knock-out Refers to techniques that remove a gene, or genes, from a cell’s genome. Overexpression Refers to techniques that excessively express a particular gene(s).

1.34.1

Introduction

Many biotherapeutics are produced in recombinant mammalian cell lines including Chinese hamster ovary (CHO), NS0, baby hamster kidney (BHK), PerC.6, and human embryonic kidney (HEK) 293 cells. These cell lines are often adapted from adherent to suspension and have been weaned to grow in animal component-free, chemically defined medium in order to achieve increasingly higher product titers, higher viable cell densities, and extended viabilities in bioreactors. One of the major goals of the bioprocessing industry is to maximize the integrated viable cell concentration, which considers both viable cell density and time of culture. Unfortunately, the drive for high viabilities and extended culture times is inevitably hindered by the activation of the cell death phase that occurs in all bioreactors. Cell cultures grown in bioreactors can experience two types of cell death: necrosis and programmed cell death (PCD; types I and II). Necrosis is a spontaneous form of cell death marked by swelling and lysing of the cell. It is usually due to a physical trauma such as shear forces from mechanical agitation, loss of cell integrity due to a chemical toxin, or injury due to a toxin. By contrast, PCD is a highly regulated, genetically conserved mode of death. Apoptosis, referred to as type-I PCD, is triggered by a multitude of insults such as nutrient depletion, the accumulation of metabolic byproducts, poor oxygenation, and pH fluctuations. Apoptosis displays distinct morphological features such as cell shrinkage, chromatin condensation and fragmentation, blebbing of the plasma membrane, and then the pinching off of apoptotic bodies containing tightly packed organelles. In cell culture, these apoptotic bodies then lyse, resulting in the spilling of their contents into the medium in what is referred to as secondary necrosis. The second type of PCD called type II PCD, or autophagy, causes the self-digestion of the cell‘s components. This catabolism is the result of the fusion of autophagosomes with lysosomes to form autophagolysosomes responsible for the degradation of the cell‘s internal components. This type of death is characterized by the accumulation of autophagic vacuoles and of processing of the LC3 protein into the 16-kDa LC3-II. Typically, the primary mode of cell death in bioreactors is PCD because bioreactors regulate the culture-feeding cells during fed-batch or perfusion runs. There are a number of signaling pathways that one needs to consider in order to effectively inhibit apoptosis in cell cultures in hopes of increasing productivity in an industrial setting. Current approaches for apoptosis inhibition include the supplementation of nutrients usually through a fed-batch, perfusion, or hybrid approach, the addition of chemical inhibitors, which can become costly at large-scale production mode, and genetic engineering of the cells with genes that can manipulate the cell death pathway to slow the progression of cell death. It is the latter approach, especially as it applies to apoptosis

468

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00030-6

Apoptosis: The Signaling Pathways and Their Control

469

and metabolism, which we focus our discussion on from herein. Furthermore, we discuss the links between apoptosis and autophagy, and how inhibiting autophagy in addition to apoptosis could be beneficial.

1.34.2

Apoptosis Regulators and Executioners

Several families of genes are responsible for maintaining the delicate balance between life and death within the cell. The largest family of apoptotic regulators is the Bcl-2 family. All members of the Bcl-2 family share regions of sequence homology, known as Bcl-2 homology (BH) domains numbered BH1–BH4; however, all members of the Bcl-2 family do not contain all four of the homology domains (Figure 1). The BH3 domain, found in all family members, regulates the homo- and heterodimerization between the anti- and pro-apoptotic family members as a way of neutralizing the competition between these proteins. The anti-apoptotic family members typically contain all four BH domains and include Bcl-2, Bcl-xL, Bcl-w, and Mcl-1. These proteins mainly reside in the mitochondria, although their presence has been reported in the cytosol and endoplasmic reticulum (ER) as well. The pro-apoptotic proteins are split into two groups: the Bcl-2 effector molecules (Bax and Bak) that contain the BH1–BH3 domains and BH3-only proteins such as Bik, Bad, and Bid, which directly activate Bax and Bak.52 Ultimately, it is the overall ratio of anti-apoptotic Bcl-2 members to proapoptotic members that determines the susceptibility of the cell to death stimuli. Caspases (cysteine-dependent aspartate-specific proteases) are the main family of proteins responsible for the execution of apoptosis and exist within cells as inert pro-caspases called zymogens. These zymogens contain a highly diverse structure on their N-terminal domain that is required for caspase activation. Induction of apoptosis usually leads to the activation of initiator or signaling caspases (caspases 2, 8, 9, 10, and 12), which cleave their target substrates after aspartic acid residues. These caspases can further activate themselves in an autocatalytic manner or other caspases in what is known as a caspase cascade. Eventually, the cascade leads to the activation of effector or executioner caspases (caspases 3, 6, and 7), resulting in the cleavage of essential cellular proteins, which trigger the morphological changes typically seen in cells undergoing apoptosis. For example, caspase-8 was shown to be responsible for the release of pro-apoptotic proteins from the mitochondria.45 Currently, 18 classes of caspases have been identified; however, some are species and/or function specific.24 Furthermore, anti-apoptotic Bcl-2 and Bcl-xL have the ability to inhibit the activation of caspases by sequestering pro-caspases.108 Inhibitors of apoptosis (IAP) family proteins represent another important family of anti-apoptotic regulators. Currently, six family members have been identified – NAIP, c-IAP1, c-IAP2, X-linked inhibitor of apoptosis (XIAP), Survivin, and BRUCE – and are believed to suppress apoptosis by inhibiting caspases. Members of the IAP family are characterized by up to three tandem repeats of baculoviral IAP repeats (BIRs), a highly conserved domain of approximately 70 amino acids; however, not all BIR-containing proteins are apoptosis regulators. Although still unclear, it is possible that the RING domain on the c-terminus of IAPs and the caspase recruitment domain are also necessary for the suppression of apoptosis (for a more detailed review, the reader is referred to Ref. 21). Studies have supported the notion that IAPs have the ability to bind and inhibit initiator caspases 8 and 9 as well as the executioner caspase-3.

1.34.3

Apoptotic Pathways

To understand how to manipulate apoptosis pathways to prolong culture viabilities, one must have an appreciation of the highly regulated pathways involved. The apoptotic cascade can be initiated by the intrinsic (or mitochondrial) pathway, the extrinsic (or Anti-apoptotic Bcl-2 Bcl-xL

BH4

BH3

BH1

BH2

Membrane anchor

Bcl-w Mcl-1

Pro-apoptotic Mutidomain Bax Bak BH3-only Bik Bid Bad Figure 1 Bcl-2 family members. Schematic representation of anti-apoptotic and pro-apoptotic members of the Bcl-2 family of proteins. Anti-apoptotic typically contain all four BH4 domains. Pro-apoptotic family members include multidomain members containing domains BH1–3 and BH3-only members.

470

Apoptosis: The Signaling Pathways and Their Control

Fas ligands Cell stress

Fas

BH3 proteins

Extrinsic pathway

FADD

tBid

Bax/Bak

Bid ER stress pathway

ER

Bcl-2/Bcl-xL

Bcl-2/Bcl-xL

Ca2+

Smac/ DIABLO

Caspase-8 (Caspase-10)

Intrinsic pathway Cyto c Mitochondria

Release of pro-apoptotic factors (including cytochrome c and Smac/DIABLO)

XIAP

Cyto c DISC Smac/ DIABLO

FLIP

Apoptosome

Caspase-9 Apaf-1 Executioner caspases (Caspases 3, 6, 7)

Apoptosis

Figure 2 Apoptosis pathways. The intrinsic pathway is initiated by many intracellular stresses resulting in the upregulation of Bax and Bak levels. Bax and Bak are both pro-apoptotic proteins containing multiple Bcl-2 homology domains that translocate to the mitochondrial membrane leading to its permeablility and the release of apoptotic proteins. The endoplasmic reticulum (ER) stress pathway is induced by misfolded and aggregated proteins and other stresses in the ER that leads to the release of Ca2þ and activation of the instrinsic pathway and independent ER-associated caspases. Bcl-2 and Bcl-xL are anti-apoptotic proteins that inhibit the translocation of pro-apoptotic proteins to the mitochondria. The extrinsic pathway includes the binding of a ligand to the TNF family of receptors and is generally activated by external stimuli. The mitocondria-dependent extrinsic pathway leads to the truncation of Bid, a BH3-only proapoptotic protein. Truncated Bid (tBid) translocates to the mitochondria and activates the intrinsic pathway. The mitochondriaindependent extrinsic pathway leads to the formation of DISC to activate executioner caspases. Smac, DIABLO, Apaf-1 and cyt c are among the apoptotic proteins housed in the mitochondria and are released upon loss of mitochondrial membrane integrity. XIAP is a member of the inhibitor of caspase-8. Apaf-1, apoptosis protein-activating factor-1; cyt c, cytochrome c; DIABLO, direct inhibition of apoptosis protein (IAP)-binding protein with low pl; DISC, death-inducing signaling complex; FLIP, Fas-associated death domain (FADD) interleukin-1b-converting enzymes (FLICE)-like inhibitory protein; Smac, second mitochondrial activator of caspases; XIAP, X-linked inhibitor of apoptosis.

death receptor (DR)) pathway, and the ER stress pathway (Figure 2). Although all three known pathways are distinct, they have some overlapping mitochondrial involvement and ultimately converge at the execution pathway, which activates effector caspases and results in physical destruction of the cell.

1.34.3.1

Intrinsic Pathway

Mitochondria are comprised of five distinct compartments (Figure 3): the outer membrane, the intermembrane space, the inner membrane, the cristae space, and the matrix.59,65 Cristae are formed by folds in the inner mitochondrial membrane, where it is estimated that 85–90% of cytochrome c stores are housed with the remainder found in the intermembrane space.19,116 The gatekeepers for the cristae are the cristae junctions composed of optic atrophy type I (OPAI) oligomers formed from OPAI monomers located on the inner mitochondrial membrane and truncated OPAI isoforms found in the inner membrane space. The average opening of a cristae junction, in healthy mitochondria, is 15–30 nm, which is wide enough to allow the passage of cytochrome c. However, OPAI oligomers at the junctions, which can be several 100 kDa in size, restrict the movement of cytochrome c into the inner membrane space. During apoptosis, the OPAI oligomer complex is disassociated by truncated Bid (tBid), which is generated by the cleaving of Bid by caspase-8 and,32 therefore, allows for the passage of cytochrome c into the matrix. Cytochrome c plays two diverse yet vitally important roles within the mitochondria. The first role is to aid in the maintenance of cell viability. In a healthy cell, cytochrome c is sequestered in the intermembrane space and operates by transferring electrons between Complex III (cytochrome bc1) and Complex IV (cytochrome c oxidase). However, in the presence of an

Apoptosis: The Signaling Pathways and Their Control

471

Matrix Inner membrane

Intermembrane space

Intercristal space Outer membrane

Figure 3 Mitochondria compartments. Mitochondria are comprised of five distinct compartments including the outer membrane, the intermembrane space (space formed between the outer and inner membrane), the inner membrane, intercristal space (space formed by folds in the inner membrane), and the matrix (the space within the inner membrane and outside of intercristal spaces).

apoptotic-inducing stress signal such as nutrient deprivation or chemical toxin exposure, cytochrome c is released into the cytosol, where it functions as an apoptotic catalyst. The release of cytochrome c occurs through the formation of a mitochondrial outer membrane pore (MOMP). Permeabilization of the outer mitochondrial membrane (OMM) occurs in a sequential manner and involves several members of the pro-apoptotic Bcl-2 family, namely Bid, Bax, and Bak. Upon activation, tBid binds to the OMM where it then binds Bax/Bak. Bax is primarily found in the cytoplasm and incorporates into the OMM following an apoptotic stimulus. Bak, on the other hand, resides primarily at the OMM. Once tBid is bound to Bax/Bak, Bax/Bak auto-oligomerizes to create the pore that allows for the release of cytochrome c from the mitochondria.60,93,111 The anti-apoptotic protein, Bcl-2, can bind tBid to suppress the interaction of tBid with the OMM to prevent Bax translocation and attenuate Bax and Bak auto-oligomerization.93,120 Bcl-xL, another anti-apoptotic protein, has the ability to bind membrane-bound tBid and Bax to competitively inhibit Bax oligomerization and suppress MOMP formation.6 By contrast, the pro-apoptotic protein, Bad, has the ability to break the tBid–Bcl-xL interaction in order to promote MOMP formation.98 Once cytochrome c has migrated to the cytosol, it interacts with the apoptotic-protease-activating factor-1 (Apaf-1) in the presence of adenosine triphosphate (ATP) and oligomerizes into an Apaf-1/cytochrome c octamer. Concurrently, pro-caspase-9 binds Apaf-1 resulting in its autocatalytic activation. The resulting cytochrome c/Apaf-1/caspase-9 multimer is referred to as the apoptosome. Activated caspase-9, in its bound or unbound form, then activates effector caspases, which generate a positive feedback loop to further activate caspase-9 and accelerate the cell‘s demise. In addition to cytochrome c, the mitochondria contain a warehouse of other pro-apoptotic molecules such as second mitochondrial activator of caspases (Smac)/direct IAP-binding protein with low pl (DIABLO), HtrA2/Omi, EndoG, AIF, and CAD, which are also released from the confinement of the mitochondria following apoptotic stimuli. The translocation of Smac/DIABLO and HtrA2/Omi is a caspase-catalyzed event and, therefore, occurs after the release of cytochrome c. Once in the cytosol, Smac/DIABLO and Htr2/Omi bind IAP family members, thus eliminating the inhibitive effects of IAP on caspases. EndoG, AIF, and CAD are released from the mitochondria, only after the commitment to apoptosis has been made, and translocate to the nucleus where they are involved with various aspects of DNA fragmentation.

1.34.3.2

ER Stress Pathway

The ER is comprised of a network of membranes responsible for the synthesis, folding, and transport of proteins regulated by molecular chaperones, such as the heat shock protein (Hsp) family, and is highly sensitive to alterations in redox state, cellular energy levels, and Ca2þ concentration. Perturbations in cellular homeostasis can cause an accumulation of unfolded or misfolded proteins within the ER lumen or prolonged Ca2þ signaling that can lead to apoptosis. Unfolded or misfolded proteins cause stress to the ER, which leads to a signal transduction pathway called the unfolded protein response (UPR). The UPR invokes proteins involved in the folding, chaperoning, and degradation pathways in an endeavor to properly fold accumulating proteins. However, when the stress level becomes overwhelming, apoptosis is activated. Three ER transmembrane proteins are known to be involved in this pathway: PKR-like ER kinase (PERK), activating transcription factor 6 (ATF6), and inositol-requiring enzyme 1 (IRE1). The UPR is initiated when glucose-regulated protein 78 (GRP78, BiP) becomes unbound from the transmembrane domains of the aforementioned proteins.5,97 The dissociation of GRP78, in return, activates PERK and IRE1. PERK then phosphorylates eukaryotic initiation factor 2, consequently impeding protein translation.39 The restricted rate of translation promotes cell survival by abating the flow of nascent proteins to the ER. However, activated PERK is also known to indirectly activate the pro-apoptotic transcription factor C/EBP homologous protein (CHOP).39 On the other hand, after GRP78 dissociation, ATF6 translocates to the Golgi apparatus for activation by proteolysis.40 It then proceeds to the nucleus to activate genes such as CHOP and the transcription factor X box-binding protein 1 (XBP1) as well as ER chaperone proteins GRP78 and GRP94. Unlike PERK activation, ATF6 activation only results in the further activation of prosurvival mechanisms. Activated IRE1 allows for the creation of a splice variant of XBP1 (sXBP1) previously activated by ATF6.55,121

472

Apoptosis: The Signaling Pathways and Their Control

sXBP1 acts upon p58IPK that binds to and inhibits PERK, thus providing a feedback loop that can halt the PERK-induced translational block.117 If ER function is successfully restored, the feedback loop allows for the termination of the UPR and the cell survives. However, if ER stress persists, pro-apoptotic proteins are processed in the ER, thereby resulting in apoptosis. ER-induced apoptosis may also result from the prolonged release of Ca2þ from the ER.104,105 Two transporters regulate Ca2þ homeostasis within the ER; the sarcoplasmic/endoplasmic reticulum calcium-ATPase pump imports Ca2þ from the cytosol to the ER and the inositol triphosphate receptor (IP3R) regulates the flux of Ca2þ from the ER into the cytosol and mitochondria. The persistent and prolonged release of Ca2þ from the ER causes an overload of Ca2þ in the mitochondria, which triggers the release of cytochrome c. Cytochrome c release is initiated by the dissipation of the inner mitochondrial transmembrane potential (Djm) and the subsequent opening of the permeability transition pore (PTP). The PTP is thought to consist of three components: cyclophilin D found inside the mitochondria, adenosine nucleotide translocase (ANT) found on the inner membrane, and the voltage-dependent anion transporter (VDAC) found on the outer membrane. A conformational change in ANT, facilitated by cyclophilin D, causes the PTP to open and allows the release of cytochrome c. Cytochrome c then becomes involved in the formation of the apoptosome as in the intrinsic pathway, thus resulting in apoptosis. It should be noted that although VDAC is responsible for the exchange of metabolites and ATP/adenosine diphosphate (ADP) across the OMM, it is believed to have no direct involvement in the PTP or the Bax-induced pores created during the intrinsic pathway of apoptosis. Interestingly, Bcl-2 and Bcl-xL have been found to confer resistance to Ca2þ-induced apoptosis by reducing the flux of Ca2þ from the ER, whereas Bax increases the Ca2þ load.9,12,85 In fact, Bcl-2 and Bcl-xL have been found to directly interact with IP3Rs to accomplish this.12,38,112 A mitochondrial-independent mechanism for ER-induced apoptosis centers on the release of active caspase-12. Caspase-12 can be activated through either the UPR or calpain cleavage resulting from an increase in cytoplasmic Ca2þ.74,75 Once active, caspase-12 translocates to the cytosol where it activates caspase-9, irrespective of apoptosome formation, thus invoking the distal steps of the intrinsic pathway while bypassing the mitochondria.76

1.34.3.3

Extrinsic Pathway

In mammalian cell cultures, apoptosis may also be initiated by members of the tumor necrosis factor (TNF)/TNF receptor (TNFR) superfamily94,114; however, it is not believed to be the predominant mode of apoptosis initiation. Although the TNF/TNFR superfamily members are involved in numerous biological functions including cellular proliferation, differentiation, and survival, this article focuses on the receptor‘s role in the apoptosis pathway. This signaling pathway involves cell surface DRs that transduce death signals from the extracellular environment to the intracellular signaling pathways. Over 40 members of the TNF/TNFR superfamily have been discovered and those involved in apoptosis include the TNF receptor 1 bound by the TNF-a ligand, the Fas/Apo1/CD95 receptor bound by FasL/CD95L, Apo3/DR3 bound by Apo3L/TWEAK, TRAIL-R1/DR4 and TRAIL-R2/DR5 both bound by Apo2L/TRAIL, and DR6 bound by amyloid precursor protein, known as APP.2,77 Of these six signaling mechanisms, we use FasR/FasL as a representative example of the DR cascade. Following receipt of an extracellular suicide signal, a trimer of Fas ligands binds to three Fas receptors resulting in a conformational change that facilitates the binding of the cytosolic adaptor protein called Fas-associated death domain (FADD). This binding occurs between the death domain of FasL and the death effector domain of FADD. Caspase-8 or caspase-10 zymogens are then recruited to form a second type of apoptosome called the death-inducing signaling complex (DISC).51 The formation of the DISC causes caspases to cluster, resulting in subsequent autocatalytic activation by induced proximity similar to the activation of caspase-9 in the intrinsic pathway.66,69 Activated caspase-8, in return, activates the effector caspases through proteolytic cleavage in a mitochondrial-independent manner. The anti-apoptotic protein FLIP (FADD interleukin-1b-converting enzyme (FLICE)-like inhibitory protein), however, can inhibit the activation of pro-caspase-8 in the DISC. In some instances, however, only a limited number of caspase-8 molecules are recruited to the DISC for autocatalytic activation. Active caspase-8 then indirectly activates effector caspases in a mitochondrial-dependent manner through cleavage of Bid, resulting in the release of pro-apoptotic molecules from the mitochondria in a manner consistent with the intrinsic pathway. Caspase-8 cleaves Bid at Asp75 to form two fragments – a C-terminal fragment and an N-terminal fragment. The C-terminal fragment, known as tBid, is then able to translocate to the mitochondria.57,61 In this pathway, cardiolipin, also known as diphosphatidylglycerol, plays an important role. This anionic phospholipid residing in the mitochondrial membrane functions as a docking station for tBid on the OMM to support Bax oligomerization and MOMP formation for the release of cytochrome c.62 Cardiolipin also recruits caspase-8 for insertion into the OMM.35 Once accumulated in the OMM, caspase-8 undergoes oligomerization,35 resulting in further activation of effector caspases and demise of the cell.

1.34.4

Apoptosis and Autophagy

The biological process of autophagy functions as a degradation and recycling center for cellular components. A cell will sometimes undergo autophagy as both a survival and death mechanism in order to maintain cellular homeostasis during periods of nutrient deprivation and is highly regulated from yeast to mammals due to the involvement of a family of autophagy-related genes (Atg). When starved of nutrients, the cell will attempt to survive through a catabolic process in which the cell recycles or degrades its intracellular components for removal of damaged organelles and the production of ATP and nutrients. In the latter stages of batch culture, it is believed that autophagy is triggered in an attempt to recycle nutrients for the cell‘s survival, but ultimately the persistent lack of nutrients causes cell death.50 The lack of specific nutrients, in particular the amino acids

Apoptosis: The Signaling Pathways and Their Control

473

Bad Bcl-2, Bcl-xL

Lysosome

PI3KC3

Beclin-1

Isolation membrane

Vesicle elongation

Autophagosome

Autolysosome

Figure 4 Autophagy. Nutrient deprivation triggers the upregulation of Beclin-1 located in the cytosol and trans-Golgi network. Beclin-1 then forms a complex with class II PI3 kinase (PI3KC3) to promote the formation of autophagosomes. Autophagosomes function to surround cytosolic areas or organelles and eventually imprison the contents. They then fuse with lysosomes allowing for the degradation of the internal components.

methionine, lysine, leucine, or arginine, has been found to invoke autophagy in CHO cell cultures, whereas the lack of other amino acids (serine, proline, isoleucine, glycine, glutamate, cysteine, asparagine, aspartate, or alanine) did not.99 Nutrient deprivation triggers the upregulation of an evolutionarily conserved mediator of autophagy termed Beclin-1, located in the cytosol and trans-Golgi network, which forms a complex with class II PI3 kinase (PI3KC3). The Beclin-1–PI3KC3 complex is responsible for the initiation of autophagy by promoting the formation of autophagosomes (Figure 4). Autophagosomes function to surround cytosolic areas or organelles and eventually imprison the contents. They then fuse with lysosomes allowing for the degradation of the internal components.28,34,109,124 It was recently discovered that mitochondria can undergo a selective form of autophagy known as mitophagy (Figure 5) in order to maintain their integrity during physiological changes in the cell. Researchers determined that Uth1p in yeast cells was required for mitochondrial clearance. Similarly, the yeast mitochondria-anchoring protein Atg32 functions as a receptor during mitophagy. There are no known mammalian homologs of Uth1p and Atg32; however, recent studies suggest that Nix, a BH3-only member of the Bcl-2 family, could perform a similar function in mammalian cells. Indeed, Nix recruits Atg8 homologs LC3 and GABARAP to damaged mitochondria to facilitate mitochondrial clearance. Although the mechanism is still unclear, Nix is able to disrupt

Isolation membrane LC3/ GABARAP

Nix

Membrane elongation

Bcl-xL Beclin-1

Bcl-xL

Nix

Beclin-1

Lysosome

Autolysosome

Autophagosome

Figure 5 Mitophagy. In mammalian cells, Nix recruits Atg8 homologs LC3 and GABARAP to damaged mitochondria to facilitate mitochondrial clearance. Nix is able to disrupt Beclin-1–Bcl-xL complexes, thereby recuiting Bcl-xL to the mitochondria. Bcl-xL in return disrupts the binding of Nix with LC3/GABARAP to release the isolation membrane. Beclin-1 recruits the isolation membrane and begins forming the autophagosome around the mitochondria. The autophagosomes eventually fuse with lysosomes to degrade the damaged mitochondria.

474

Apoptosis: The Signaling Pathways and Their Control

Beclin-1–Bcl-xL complexes, thereby recruiting Bcl-xL to the mitochondria and freeing Beclin-1 to initiate autophagosome formation around the mitochondria leading to mitochondrial clearance.34,46,79,96 Researchers have determined an interesting link between autophagy and apoptosis with the discovery that Beclin-1 also has the ability to form complexes with members of the anti-apoptotic Bcl-2 family including Bcl-2, Bcl-xL, Bcl-w, and Mcl-1.25 Beclin-1 is a substrate for caspase-mediated cleavage14 and possesses a novel BH3 domain, placing it in the category with other BH3-only members of the Bcl-2 family.27,81 In particular, Bcl-2 and Bcl-xL in complex with Beclin-1 have been shown to inhibit autophagy as well as apoptosis. On the other hand, pro-apoptotic BH3-only proteins, such as Bad, can inhibit Beclin-1–Bcl-2/Bcl-xL complexes to promote autophagy.118 Furthermore, mitophagy has been linked to apoptosis through the formation of the mitochondrial permeability transition pore (MPTP). As previously discussed, ER stress can trigger MPTP formation and the release of cytochrome c, which can then initiate apoptosis. It has been determined that MPTP formation, however, can also stimulate mitophagy in an effort to remove damaged mitochondria. Therefore, MPTP formation is important in initiating mitophagy in addition to apoptosis.91 The involvement of Bcl-2 family members and the MPTP in autophagy provides a concrete link between the two forms of PCD, although further work is needed to fully appreciate the interrelationship between autophagy and apoptosis and the potential to alter each of these pathways in mammalian cell culture systems.

1.34.5

Inhibition of Apoptosis

Apoptosis in mammalian cell culture for the production of biopharmaceuticals can be harmful for multiple reasons. Cell death causes a decrease in viability and viable cell numbers and may lead to lower-than-desired production levels as well as reduced product quality from the release of proteases resulting from secondary necrosis. Additionally, cellular debris may interfere with downstream purification processes. Therefore, it is beneficial to limit apoptosis in a mammalian cell culture production setting. There are several methods available to achieve this goal. The most common approach to improve culture viability is through the addition of plant hydrolysates or through the addition of limiting nutrients and growth factors, because deprivation of the latter is an apoptosis trigger.8,87,90,100,101,119,122 The addition of chemical IAP to cell culture has also been considered1,36,68,95; however, the addition of these types of components to the culture medium may not be desirable at large scale where reagent cost can become problematic. An alternative approach to inhibiting apoptosis involves the genetic modification of the cell either through the overexpression of anti-apoptotic genes or through the suppression or elimination of pro-apoptotic genes. There are several benefits to this approach of host cell engineering. First of all, no additional work will be necessary for the generation of production cell lines if a host cell line expressing the desired anti-apoptotic protein(s), or suppressing the desired pro-apoptotic gene(s), is established prior to transfection with the target gene product of interest. Second, cell lines engineered to be resistant to apoptosis may adapt to chemically defined, animal-protein-free medium better than the parent cell line, which is favorable for today‘s bioreactor processes. For example, the reagent sodium butyrate (NaBu) can be used for increasing protein production by enhancing transcription.82,84,103 However, it can cause rapid apoptosis in cell lines not engineered for increased survival; therefore, anti-apoptosis engineering of the cell lines could prove useful. Work in this field has centered on the overexpression of Bcl-2, the pioneer of the apoptosis family of inhibitors, in CHO, BHK, and hybridoma cell lines.15,26,36,44,67,68,107 Less widespread efforts have been put forth on the overexpression of other Bcl-2 and IAP family members in industrially relevant cell lines. Overexpression of Bcl-xL in a CHO cell line producing an antibody directed against a1b1 integrin resulted in a 50% increase in viability and a 90% increase in titer when cultured in chemically defined medium.13 Furthermore, the addition of NaBu to the culture medium allowed for more than a twofold increase in product titer in the Bcl-xL containing cell line.13 Stable overexpression of yet another Bcl-2 homolog, Mcl-1, in a serum-free CHO line in a commercial fed-batch process, resulted in an approximately 25% increase in antibody titer and viability at day 14.64 The viral Bcl-2 homolog, E1B-19K, was shown to be effective in inhibiting death and increasing antibody yields in NS0 cultures during batch and perfusion culture, when stably overexpressed.70,71 Caspase inhibitors such as wild-type XIAP and cytokine response modifier (CrmA), as well as mutants of each, have also been examined and shown to be effective at delaying apoptosis.95 However, these anti-apoptosis genes are farther downstream than Bcl-2 family proteins in the apoptosis pathway and, thus, may not be as effective at inhibiting apoptosis because the initial steps of the apoptotic cascade have already been activated by the time these inhibitors come into play. This was demonstrated by the overexpression of XIAP in recombinant CHO (rCHO) cells during the use of NaBu.49 However, combining an upstream apoptosis inhibitor, such as Bcl-xL, with the downstream XIAP gene lacking the C-terminal RING, proved to have a greater benefit in delaying apoptosis in CHO cells than either gene expressed individually.94 Several other studies have taken the combinatorial approach to delaying death. Aven is an inhibitor of apoptosis that binds both Apaf-1 and Bcl-xL independently of each other.10 When co-expressed with Bcl-xL, a synergistic effect was observed resulting in enhanced cell survival from various apoptotic insults.30 E1B-19K, a viral homolog of Bcl-2 and Bcl-xL, was also co-expressed with Aven and resulted in improved cell viability and recombinant protein titer and a decrease in caspase-3 activity when compared to a control cell line in a perfusion system at decreasing specific perfusion rates.78 This combination of genes has also been studied in CHO cells expressing antibodies in batch and fedbatch with similar positive effects on cell viability and product titers.29 Bcl-2 family members have also been combined with cell-cycle genes for various purposes. For the adaptation of adherent CHO cells to protein-free suspension culture, it was found that the cyclin-dependent kinase inhibitor p21CIP1, which arrests the cells in the G1 phase, allowed for a more rapid adaptation.4 When Bcl-2 was added in addition to p21CIP1, the cell viability was increased during the G1 arrest.3

Apoptosis: The Signaling Pathways and Their Control

475

Although much of the work in genetic modification has focused on the overexpression of Bcl-2 family members for the stable production of recombinant proteins, anti-apoptosis genes have also been used to inhibit apoptosis during transient gene expression. Recently, researchers have begun investigating the use of CHO cells for transient expression.20,33,73,92,106 CHO cells, which are the industry standard for recombinant protein production, are an appropriate choice when a product is needed with similar processing to future large-scale bioreactor runs. This work has been taken one step further with the addition of Bcl-xL. CHO-DG44 cells that were engineered to overexpress Bcl-xL were used for the transient production of a fusion protein. Not only was the viability extended compared to the nonengineered control, but also specific productivity and the total production level were increased.63 This improvement has been achieved for both an IgG fusion and membrane proteins.80 Further work has been performed in an attempt to generate improved anti-apoptosis genes. Many of these survival genes contain inherent sequences that allow for their degradation through methods such as caspase cleavage and ubiquitination in order to maintain cellular homeostasis. In cell culture, however, it is desired to maintain expression of the anti-apoptotic gene to achieve the maximum benefit in culture. To that effect, it has been demonstrated that removing the pro-apoptotic regions of anti-apoptotic genes leads to increased survival rates compared to their wild-type counterparts.30,31,95 The Hsp family is another family of proteins affiliated with the apoptosis cascade. These molecular chaperones, which orchestrate the proper folding, assembly, localization, and secretion of new proteins, may increase the cell‘s productivity and viability by inhibiting protein misfolding and aggregation and, consequently, ER-induced apoptosis.72 Hsp70 is a cytosolic chaperone protein that inhibits protein aggregation by targeting misfolded proteins to the cellular degradation machinery and displays anti-apoptotic tendencies when overexpressed in mammalian cell culture. Ishaque et al.43 found that overexpression of Hsp70 in a BHK line transfected with recombinant coagulation factor VIII (FVIII) led to decreased apoptosis during nutrient depletion and exposure to the chemical toxins staurosporine, etoposide, and camptothecin. When Hsp70 was transfected into NS0 cells, which are deficient in this stress protein, the onset of apoptosis was delayed and product quality was improved in the resulting hybridomas.54 Hsp70, in addition to another chaperone Hsp27, has also been transfected in CHO cells expressing recombinant human interferon-g (IFN-g). Whether expressed individually or in combination, the fed-batch cultures experienced delays in the initiation of caspase activation and apoptosis, as well as an increase in IFN- g yield.56 In contrast to overexpressing an anti-apoptotic gene, the knockdown or silencing of a pro-apoptotic gene may prove fruitful. This has been observed with the pro-apoptotic mitochondrial proteins, Bak and Bax. In one study, zinc-finger nuclease technology was used to eliminate these two genes in CHO cells resulting in a greater resistance to apoptosis and increased IgG production without compromising the culture‘s cell density.17 In another study, small interfering RNA (siRNA) constructs were used to generate a CHO cell line with significantly reduced Bak and Bax activity that resulted in increased culture viability and productivity.58 siRNA was used to knock-down Bad, Bid, and Bim in HEK 293 cells to extend culture longevity.83 In an interesting study performed by Yap and co-workers, transcriptional profiling using microarray analysis in both batch and fed-batch culture provided insight into which genes in the apoptotic cascade were up- and downregulated during apoptosis initiation. It was determined that the intrinsic and extrinsic pathways were the preferred signaling pathways as opposed to the ER pathway.113 In a follow up, siRNA technology was used to knock down the pro-apoptotic genes, Apoptosis-linked gene-2 (Alg-2) and Requiem. Alg-2 has been found to be active in both the ER and extrinsic pathways, whereas the role of Requiem in the apoptotic pathway is still being elucidated. Knockdown of these genes allowed for higher peak viable cell densities and greater recombinant protein yields.114 However, when siRNA technology was applied to caspase-3 in CHO cells in an attempt to combat the apoptotic effects of NaBu, neither antibody productivity nor inhibition of apoptosis was improved.47,102 Apoptosis and autophagy have been shown to occur both individually or simultaneously within a cell.41 In fact, rCHO cells have been found to undergo both modes of death during conditions of hyperosmotic stress.37 Therefore, the inhibition of autophagy in addition to apoptosis to extend culture longevity should be examined further. Overexpression of constitutively active Akt (protein kinase B) in CHO cells delayed both apoptosis and autophagy during nutrient deprivation.42 Furthermore, members of the anti-apoptotic Bcl-2 family, including Bcl-2, Bcl-xL, Bcl-w, and Mcl-1, have been found to inhibit autophagy to various degrees through interaction with Beclin-1.25 Interestingly, only ER-targeted Bcl-2 and Bcl-xL suppress autophagy.18,89 In fact, overexpression of Bcl-xL in CHO cells delayed both apoptosis and autophagy during nutrient deprivation from batch culture.50 Although Bcl-2 can disrupt Beclin-1-induced autophagy, the converse does not hold true in that Beclin-1 fails to abolish the anti-apoptotic ability of Bcl-2.16

1.34.6

Apoptosis Affects Metabolic Pathways

The principal component of cellular metabolism in mammalian cell culture is glucose. Glucose is transported from the culture medium across the plasma membrane and into the cell through facilitative diffusion regulated by a family of proteins called facilitative glucose transporters. There are five transporters that vary depending upon tissue type; however, for typical mammalian cell cultures, glucose transporter 1 (GLUT1) is predominately expressed. Once internalized, glucose enters the glycolytic pathway where one glucose molecule is converted to two pyruvate molecules and two ATPs. At this point, pyruvate typically moves into the mitochondria and enters the citric acid/Kreb‘s cycle for the production of CO2 and H2O through the use of key enzymes such as pyruvate carboxylase (PC). In mammalian cell lines, however, pyruvate can remain in the cytosol and be converted by lactate dehydrogenase (LDH) into lactate, which is released into the culture medium. Accumulation of lactate occurs during the exponential growth phase of the cell culture and is referred to as the Warburg effect.110 However, a slight consumption of lactate

476

Apoptosis: The Signaling Pathways and Their Control

has been observed during the stationary phase of CHO, NS0, and hybridoma.7,22,88,123 Lactate buildup in the medium causes acidification and results in the need for additional alkali to maintain optimum pH as well as increased osmolality and consequently an inhibition of cell growth. Ultimately, when lactate reaches toxic levels, it leads to apoptosis. Previous research has focused on preventing the accumulation of toxic levels of lactate. One approach to reducing lactate levels is to use feeding strategies that reduce or control the amount of glucose available in the culture medium.53,115 Another method involves the use of siRNA to downregulate the expression of LDH-A. In the rCHO cell lines tested, LDH activity, specific glucose consumption, and specific lactate production were all reduced without compromising cell growth or productivity.48 By contrast, overexpression of PC drives the cell‘s metabolism to more efficiently utilize pyruvate in the Kreb‘s cycle, thus reducing the need for LDH for the production of lactate. This was demonstrated in CHO-DG44 cells, where reduced lactate production was observed.48 Recently, a more direct link between the apoptotic and metabolic pathways has been observed in mammalian CHO cells overexpressing E1B-19K and Aven. Engineering of apoptosis pathways in this suspension CHO cell line altered the nutrient profile of the cultures significantly.23 The control cell line exhibited an accumulation of lactate during the exponential phase of growth, followed by a slight consumption of lactate in the stationary phase, as observed previously by other researchers.7,22,88,123 However, in the anti-apoptosis cell lines containing E1B-19K and Aven, the cultures experienced not only an increase in longevity but also complete elimination of lactate levels. This phenomenon of reduced lactate levels has also been observed for E1B-19K and Aven in BHK cells when run in perfusion.78 Furthermore, the engineered cell lines even consumed lactate exogenously added during the exponential phase at levels comparable to those in the control cultures.23 As a result of their altered physiology, the cells could be cultured effectively in a high-glucose medium, where viability, peak viable cell density, and antibody titer all significantly increased compared to the controls, which accumulated lactate at levels thought to be toxic. Similarly, Bcl-2 expression in Myc-transformed Rat1 cells was found to alter glucose metabolism by lowering the molar ratio of lactate production to glucose consumption.86 This change in metabolic physiology may be particularly beneficial in an industrial setting because the decrease in secreted lactate level will lower the pH of the culture medium below toxic levels and, therefore, help to lower osmolarity increases affiliated with maintaining a constant pH. Indeed, all of these conditions are advantageous for maintaining the product quality.11

1.34.7

Conclusion

The biopharmaceutical industry is responsible for the production of therapeutic products through the use of mammalian cells modified for suspension culture. However, apoptosis is an inherent problem of mammalian cells. Increasingly, research is being conducted to inhibit apoptosis as well as autophagy, which is now known to occur both independently and concurrently with apoptosis. There are a number of apoptosis regulators and executioners responsible for modulating the intrinsic, extrinsic, and ER stress pathways of apoptosis. Understanding these pathways and considering the importance of the mitochondria have ultimately led to many advances in cell engineering and genetic manipulation as it applies to apoptosis. Overexpression of anti-apoptotic Bcl-2 family proteins, caspase inhibitors, and Hsps has shown great promise in inhibiting apoptosis. Conversely, the knockdown or knockout of pro-apoptotic Bcl-2 family proteins, such as Bak and Bax, has also proved fruitful in combating apoptosis. Emerging research is also centered on improving the function of anti-apoptosis genes. These applications of anti-apoptosis technology in the industrial setting are helping to increase mammalian cell culture longevity and biotherapeutic product yields and quality. Interestingly, engineering cells to be more resistant to apoptosis have also been shown recently to affect cellular metabolism. More specifically, anti-apoptosis cell lines can enable cells to drive pyruvate more effectively into the citric acid cycle and therefore reduce lactate buildup. In this way, anti-apoptosis engineering may alter resistance to apoptosis, induction of autophagy, and metabolism of nutrients in order to improve performance and productivity of mammalian cells in industrial bioreactors.

References 1. Arden, N.; Ahn, S.-H.; Vaz, W.; et al. Chemical Caspase Inhibitors Enhance Cell Culture Viabilities and Protein Titer. Biotechnol. Prog. 2007, 23, 506–511. 2. Ashkenazi, A. Death Receptors: Signaling and Modulation. Science 1998, 281, 1305–1308. 3. Astley, K.; Al-Rubeai, M. The Role of Bcl-2 and its Combined Effect with P21CIP1 in Adaptation of CHO Cells to Suspension and Protein-free Culture. Appl. Microbiol. Biotechnol. 2008, 78, 391–399. 4. Astley, K.; Naciri, M.; Racher, A.; Al-Rubeai, M. The Role of P21cip1 in Adaptation of CHO Cells to Suspension and Protein-free Culture. J. Biotechnol. 2007, 130, 282–290. 5. Bertolotti, A.; Zhang, Y.; Hendershot, L. M.; et al. Dynamic Interaction of BiP and ER Stress Transducers in the Unfolded-protein Response. Nat. Cell Biol. 2000, 2, 326–332. 6. Billen, L. P.; Kokoski, C. L.; Lovell, J. F.; et al. Bcl-xl Inhibits Membrane Permeabilization by Competing with Bax. PLoS Biol. 2008, 6, e147. 7. Burky, J. E.; Wesson, M. C.; Young, A.; et al. Protein-free Fed-batch Culture of Non-GS NS0 Cell Lines for Production of Recombinant Antibodies. Biotechnol. Bioeng. 2007, 96, 281–293. 8. Burteau, C. C.; Verhoeye, F. C. R.; Mols, J. F.; et al. Fortification of a Protein-free Cell Culture Medium with Plant Peptones Improves Cultivation and Productivity of an Interferon-gamma-producing CHO Cell Line. In Vitro Cell. Dev. Biol. Anim. 2003, 39, 291–296. 9. Chami, M.; Prandini, A.; Campanella, M.; et al. Bcl-2 and Bax Exert Opposing Effects on Ca2þ Signaling, Which Do Not Depend on Their Putative Pore-forming Region. J. Biol. Chem. 2004, 279, 54581–54589. 10. Chau, B. N.; Cheng, E. H.; Kerr, D. A.; Hardwick, J. M. Aven, a Novel Inhibitor of Caspase Activation, Binds Bcl-xL and Apaf-1. Mol. Cell 2000, 6, 31–40. 11. Chen, K.; Liu, Q.; Xie, L.; et al. Engineering of a Mammalian Cell Line for Reduction of Lactate Formation and High Monoclonal Antibody Production. Biotechnol. Bioeng. 2001, 72, 55–61.

Apoptosis: The Signaling Pathways and Their Control

477

12. Chen, R.; Valencia, I.; Zhong, F.; et al. Bcl-2 Functionally Interacts with Inositol 1,4,5-trisphosphate Receptors to Regulate Calcium Release from the ER in Response to Inositol 1,4,5-trisphosphate. JCB (J. Cell Biol.) 2004, 166, 193–203. 13. Chiang, G. G.; Sisk, W. P. Bcl-x(L) Mediates Increased Production of Humanized Monoclonal Antibodies in Chinese Hamster Ovary Cells. Biotechnol. Bioeng. 2005, 91, 779–792. 14. Cho, D.-H.; Jo, Y. K.; Hwang, J. J.; et al. Caspase-mediated Cleavage of ATG6/Beclin-1 Links Apoptosis to Autophagy in HeLa Cells. Canc. Lett. 2009, 274, 95–100. 15. Chung, J. D.; Sinskey, A. J.; Stephanopoulos, G. Growth Factor and Bcl-2 Mediated Survival during Abortive Proliferation of Hybridoma Cell Line. Biotechnol. Bioeng. 1998, 57, 164–171. 16. Ciechomska, I. A.; Goemans, G. C.; Skepper, J. N.; Tolkovsky, A. M. Bcl-2 Complexed with Beclin-1 Maintains Full Anti-apoptotic Function. Oncogene 2009, 28, 2128–2141. 17. Cost, G. J.; Freyvert, Y.; Vafiadis, A.; et al. BAK and BAX Deletion Using Zinc-finger Nucleases Yields Apoptosis-resistant CHO Cells. Biotechnol. Bioeng. 2010, 105, 330–340. 18. Criollo, A.; Vicencio, J. E. M.; Tasdemir, E.; et al. The Inositol Trisphosphate Receptor in the Control of Autophagy. Autophagy 2007, 3, 350–353. 19. Delivani, P.; Martin, S. J. Mitochondrial Membrane Remodeling in Apoptosis: An inside Story. Cell Death Differ. 2006, 13, 2007–2010. 20. Derouazi, M.; Girard, P.; Van Tilborgh, F. E. E.; et al. Serum-free Large-scale Transient Transfection of CHO Cells. Biotechnol. Bioeng. 2004, 87, 537–545. 21. Deveraux, Q. L.; Reed, J. C. IAP Family Proteins – Suppressors of Apoptosis. Genes Dev. 1999, 13, 239–252. 22. Dezengotita, V. M.; Miller, W. M.; Aunins, J. G.; Zhou, W. Phosphate Feeding Improves High-cell-concentration NS0 Myeloma Culture Performance for Monoclonal Antibody Production. Biotechnol. Bioeng. 2000, 69, 566–576. 23. Dorai, H.; Kyung, Y. S.; Ellis, D.; et al. Expression of Anti-apoptosis Genes Alters Lactate Metabolism of Chinese Hamster Ovary Cells in Culture. Biotechnol. Bioeng. 2009, 103, 592–608. 24. Eckhart, L.; Ballaun, C.; Hermann, M.; et al. Identification of Novel Mammalian Caspases Reveals an Important Role of Gene Loss in Shaping the Human Caspase Repertoire. Mol. Biol. Evol. 2008, 25, 831–841. 25. Erlich, S.; Mizrachy, L.; Segev, O.; et al. Differential Interactions between Beclin 1 and Bcl-2 Family Members. Autophagy 2007, 3, 561–568. 26. Fassnacht, M.; Beuschlein, F.; Vay, S.; et al. Aminoglutethimide Suppresses Adrenocorticotropin Receptor Expression in the NCI-h295 Adrenocortical Tumor Cell Line. J. Endocrinol. 1998, 159, 35–42. 27. Feng, W.; Huang, S.; Wu, H.; Zhang, M. Molecular Basis of Bcl-xL‘s Target Recognition Versatility Revealed by the Structure of Bcl-xL in Complex with the BH3 Domain of Beclin-1. J. Mol. Biol. 2007, 372, 223–235. 28. Ferraro, E.; Cecconi, F. Autophagic and Apoptotic Response to Stress Signals in Mammalian Cells. Arch. Biochem. Biophys. 2007, 462, 210–219. 29. Figueroa, B.; Ailor, E.; Osborne, D.; et al. Enhanced Cell Culture Performance Using Inducible Anti-apoptotic Genes E1B-19K and Aven in the Production of a Monoclonal Antibody with Chinese Hamster Ovary Cells. Biotechnol. Bioeng. 2007, 97, 877–892. 30. Figueroa, B.; Chen, S.; Oyler, G. A.; et al. Aven and Bcl-xL Enhance Protection against Apoptosis for Mammalian Cells Exposed to Various Culture Conditions. Biotechnol. Bioeng. 2004, 85, 589–600. 31. Figueroa, B.; Sauerwald, T. M.; Mastrangelo, A. J.; et al. Comparison of Bcl-2 to a Bcl-2 Deletion Mutant for Mammalian Cells Exposed to Culture Insults. Biotechnol. Bioeng. 2001, 73, 211–222. 32. Frezza, C.; Cipolat, S.; Martins De Brito, O.; et al. OPA1 Controls Apoptotic Cristae Remodeling Independently from Mitochondrial Fusion. Cell 2006, 126, 177–189. 33. Galbraith, D. J.; Tait, A. S.; Racher, A. J.; et al. Control of Culture Environment for Improved Polyethylenimine-mediated Transient Production of Recombinant Monoclonal Antibodies by CHO Cells. Biotechnol. Prog. 2006, 22, 753–762. 34. Glick, D.; Barth, S.; Macleod, K. F. Autophagy: Cellular and Molecular Mechanisms. J. Pathol. 2010, 221, 3–12. 35. Gonzalvez, F.; Schug, Z. T.; Houtkooper, R. H.; et al. Cardiolipin Provides an Essential Activating Platform for Caspase-8 on Mitochondria. JCB (J. Cell Biol.) 2008, 183, 681–696. 36. Goswami, J.; Sinskey, A. J.; Steller, H.; et al. Apoptosis in Batch Cultures of Chinese Hamster Ovary Cells. Biotechnol. Bioeng. 1999, 62, 632–640. 37. Han, Y. K.; Kim, Y.-G.; Kim, J. Y.; Lee, G. M. Hyperosmotic Stress Induces Autophagy and Apoptosis in Recombinant Chinese Hamster Ovary Cell Culture. Biotechnol. Bioeng. 2010, 105, 1187–1192. 38. Hanson, C. J.; Bootman, M. D.; Distelhorst, C. W.; et al. Bcl-2 Suppresses Ca2þ Release through Inositol 1,4,5-trisphosphate Receptors and Inhibits Ca2þ Uptake by Mitochondria without Affecting ER Calcium Store Content. Cell Calcium 2008, 44, 324–338. 39. Harding, H. P.; Novoa, I.; Zhang, Y.; et al. Regulated Translation Initiation Controls Stress-induced Gene Expression in Mammalian Cells. Mol. Cell 2000, 6, 1099–1108. 40. Haze, K.; Yoshida, H.; Yanagi, H.; et al. Mammalian Transcription Factor ATF6 Is Synthesized as a Transmembrane Protein and Activated by Proteolysis in Response to Endoplasmic Reticulum Stress. Mol. Biol. Cell 1999, 10, 3787–3799. 41. Hwang, S. O.; Lee, G. M. Nutrient Deprivation Induces Autophagy as Well as Apoptosis in Chinese Hamster Ovary Cell Culture. Biotechnol. Bioeng. 2008, 99, 678–685. 42. Hwang, S. O.; Lee, G. M. Effect of Akt Overexpression on Programmed Cell Death in Antibody-producing Chinese Hamster Ovary Cells. J. Biotechnol. 2009, 139, 89–94. 43. Ishaque, A.; Thrift, J.; Murphy, J. E.; Konstantinov, K. Over-expression of Hsp70 in BHK-21 Cells Engineered to Produce Recombinant Factor VIII Promotes Resistance to Apoptosis and Enhances Secretion. Biotechnol. Bioeng. 2007, 97, 144–155. 44. Itoh, Y.; Ueda, H.; Suzuki, E. Overexpression of Bcl-2, Apoptosis Suppressing Gene: Prolonged Viable Culture Period of Hybridoma and Enhanced Antibody Production. Biotechnol. Bioeng. 1995, 48, 118–122. 45. Kadohara, K.; Nagumo, M.; Asami, S.; et al. Caspase-8 mediates mitochondrial release of pro-apoptotic proteins in a manner independent of its proteolytic activity in apoptosis induced by the protein synthesis inhibitor acetoxycycloheximide in human leukemia Jurkat cells. J. Biol. Chem. 2009, 284, 5478–5487. 46. Kanki, T. Nix, a Receptor Protein for Mitophagy in Mammals. Autophagy 2010, 6, 433–435. 47. Kim, N. S.; Lee, G. M. Inhibition of Sodium Butyrate-induced Apoptosis in Recombinant Chinese Hamster Ovary Cells by Constitutively Expressing Antisense RNA of Caspase-3. Biotechnol. Bioeng. 2002, 78, 217–228. 48. Kim, S. H.; Lee, G. M. Down-regulation of Lactate Dehydrogenase-A by SiRNAs for Reduced Lactic Acid Formation of Chinese Hamster Ovary Cells Producing Thrombopoietin. Appl. Microbiol. Biotechnol. 2007, 74, 152–159. 49. Kim, Y.-G.; Kim, J. Y.; Lee, G. M. Effect of XIAP Overexpression on Sodium Butyrate-induced Apoptosis in Recombinant Chinese Hamster Ovary Cells Producing Erythropoietin. J. Biotechnol. 2009, 144, 299–303. 50. Kim, Y.-G.; Kim, J. Y.; Mohan, C.; Lee, G. M. Effect of Bcl-xL Overexpression on Apoptosis and Autophagy in Recombinant Chinese Hamster Ovary Cells under Nutrient-deprived Condition. Biotechnol. Bioeng. 2009, 103, 757–766. 51. Kischkel, F. C.; Hellbardt, S.; Behrmann, I.; et al. Cytotoxicity-dependent APO-1 (Fas/CD95)-associated Proteins Form a Death-inducing Signaling Complex (DISC) with the Receptor. EMBO J. 1995, 14, 5579–5588. 52. Korsmeyer, S. J.; Wei, M. C.; Saito, M.; et al. Pro-apoptotic Cascade Activates BID, Which Oligomerizes BAK or BAX into Pores that Result in the Release of Cytochrome C. Cell Death Differ. 2000, 7, 1166–1173. 53. Kurokawa, H.; Park, Y. S.; Iijima, S.; Kobayashi, T. Growth Characteristics in Fed-batch Culture of Hybridoma Cells with Control of Glucose and Glutamine Concentrations. Biotechnol. Bioeng. 1994, 44, 95–103. 54. Lasunskaia, E. B.; Fridlianskaia, , II; Darieva, Z. A.; et al. Transfection of NS0 Myeloma Fusion Partner Cells with HSP70 Gene Results in Higher Hybridoma Yield by Improving Cellular Resistance to Apoptosis. Biotechnol. Bioeng. 2003, 81, 496–504. 55. Lee, K.; Tirasophon, W.; Shen, X.; et al. IRE1-mediated Unconventional mRNA Splicing and S2P-mediated ATF6 Cleavage Merge to Regulate XBP1 in Signaling the Unfolded Protein Response. Genes Dev. 2002, 16, 452–466.

478

Apoptosis: The Signaling Pathways and Their Control

56. Lee, Y. Y.; Wong, K. T. K.; Tan, J.; et al. Overexpression of Heat Shock Proteins (HSPs) in CHO Cells for Extended Culture Viability and Improved Recombinant Protein Production. J. Biotechnol. 2009, 143, 34–43. 57. Li, H.; Zhu, H.; Xu, C. J.; Yuan, J. Cleavage of BID by Caspase 8 Mediates the Mitochondrial Damage in the Fas Pathway of Apoptosis. Cell 1998, 94, 491–501. 58. Lim, S. F.; Chuan, K. H.; Liu, S.; et al. RNAi Suppression of Bax and Bak Enhances Viability in Fed-batch Cultures of CHO Cells. Metab. Eng. 2006, 8, 509–522. 59. Logan, D. C. The Mitochondrial Compartment. J. Exp. Bot. 2006, 57, 1225–1243. 60. Lovell, J. F.; Billen, L. P.; Bindner, S.; et al. Membrane Binding by TBid Initiates an Ordered Series of Events Culminating in Membrane Permeabilization by Bax. Cell 2008, 135, 1074–1084. 61. Luo, X.; Budihardjo, I.; Zou, H.; et al. Bid, a Bcl2 Interacting Protein, Mediates Cytochrome C Release from Mitochondria in Response to Activation of Cell Surface Death Receptors. Cell 1998, 94, 481–490. 62. Lutter, M.; Fang, M.; Luo, X.; et al. Cardiolipin Provides Specificity for Targeting of TBid to Mitochondria. Nat. Cell Biol. 2000, 2, 754–761. 63. Majors, B. S.; Betenbaugh, M. J.; Pederson, N. E.; Chiang, G. G. Enhancement of Transient Gene Expression and Culture Viability Using Chinese Hamster Ovary Cells Overexpressing Bcl-x(L). Biotechnol. Bioeng. 2008, 101, 567–578. 64. Majors, B. S.; Betenbaugh, M. J.; Pederson, N. E.; Chiang, G. G. Mcl-1 Overexpression Leads to Higher Viabilities and Increased Production of Humanized Monoclonal Antibody in Chinese Hamster Ovary Cells. Biotechnol. Prog. 2009, 25, 1161–1168. 65. Mannella, C. A. Structure and Dynamics of the Mitochondrial Inner Membrane Cristae. Biochim. Biophys. Acta 2006, 1763, 542–548. 66. Martin, D. A. Membrane Oligomerization and Cleavage Activates the Caspase-8 (FLICE/MACHalpha 1) Death Signal. J. Biol. Chem. 1998, 273, 4345–4349. 67. Mastrangelo, A. J.; Hardwick, J. M.; Bex, F.; Betenbaugh, M. J. Part I. Bcl-2 and Bcl-x(L) Limit Apoptosis upon Infection with Alphavirus Vectors. Biotechnol. Bioeng. 2000, 67, 544–554. 68. Mastrangelo, A. J.; Zou, S.; Hardwick, J. M.; Betenbaugh, M. J. Antiapoptosis Chemicals Prolong Productive Lifetimes of Mammalian Cells upon Sindbis Virus Vector Infection. Biotechnol. Bioeng. 1999, 65, 298–305. 69. Medema, J. P.; Scaffidi, C.; Kischkel, F. C.; et al. FLICE Is Activated by Association with the CD95 Death-inducing Signaling Complex (DISC). EMBO J. 1997, 16, 2794–2804. 70. Mercille, S.; Massie, B. Apoptosis-resistant NS/0 E1B-19K Myelomas Exhibit Increased Viability and Chimeric Antibody Productivity under Cell Cycle Modulating Conditions. Cytotechnology 1998, 28, 189–203. 71. Mercille, S.; Massie, B. Apoptosis-resistant E1B-19K-expressing NS/0 Myeloma Cells Exhibit Increased Viability and Chimeric Antibody Productivity under Perfusion Culture Conditions. Biotechnol. Bioeng. 1999, 63, 529–543. 72. Mosser, D. D.; Massie, B. Genetically Engineering Mammalian Cell Lines for Increased Viability and Productivity. Biotechnol. Adv. 1994, 12, 253–277. 73. Muller, N.; Derouazi, M.; Van Tilborgh, F. E. E.; et al. Scalable Transient Gene Expression in Chinese Hamster Ovary Cells in Instrumented and Non-instrumented Cultivation Systems. Biotechnol. Lett. 2007, 29, 703–711. 74. Nakagawa, T.; Yuan, J. Cross-talk between Two Cysteine Protease Families. Activation of Caspase-12 by Calpain in Apoptosis. JCB (J. Cell Biol.) 2000, 150, 887–894. 75. Nakagawa, T.; Zhu, H.; Morishima, N.; et al. Caspase-12 Mediates Endoplasmic-reticulum-specific Apoptosis and Cytotoxicity by Amyloid-beta. Nature 2000, 403, 98–103. 76. Nakanishi, K.; Maruyama, M.; Shibata, T.; Morishima, N. Identification of a Caspase-9 Substrate and Detection of its Cleavage in Programmed Cell Death during Mouse Development. J. Biol. Chem. 2001, 276, 41237–41244. 77. Nikolaev, A.; Mclaughlin, T.; O‘leary, D. D. M.; Tessier-Lavigne, M. APP Binds DR6 to Trigger Axon Pruning and Neuron Death via Distinct Caspases. Nature 2009, 457, 981–989. 78. Nivitchanyong, T.; Martinez, A.; Ishaque, A.; et al. Anti-apoptotic Genes Aven and E1B-19K Enhance Performance of BHK Cells Engineered to Express Recombinant Factor VIII in Batch and Low Perfusion Cell Culture. Biotechnol. Bioeng. 2007, 98, 825–841. 79. Novak, I.; Kirkin, V.; Mcewan, D. G.; et al. Nix Is a Selective Autophagy Receptor for Mitochondrial Clearance. EMBO Rep. 2010, 11, 45–51. 80. O‘Connor, S.; Li, E.; Majors, B. S.; et al. Increased Expression of the Integral Membrane Protein ErbB2 in Chinese Hamster Ovary Cells Expressing the Anti-apoptotic Gene Bcl-xL. Protein Expr. Purif. 2009, 67, 41–47. 81. Oberstein, A.; Jeffrey, P. D.; Shi, Y. Crystal structure of the Bcl-xl-beclin 1 Peptide Complex: Beclin 1 Is a Novel BH3-only Protein. J. Biol. Chem. 2007, 282, 13123–13132. 82. Oh, S. K.; Vig, P.; Chua, F.; et al. Substantial Overproduction of Antibodies by Applying Osmotic Pressure and Sodium Butyrate. Biotechnol. Bioeng. 1993, 42, 601–610. 83. Padmanabhan, A.; Liu, S.; Song, Z. Bad Plays a More Significant Role than Bid and Bim in Mediating Cell Death Signals in Batch Cultures of HEK 293 Cells. Biotechnol. Lett. 2008, 30, 819–827. 84. Palermo, D. P.; Degraaf, M. E.; Marotti, K. R.; et al. Production of Analytical Quantities of Recombinant Proteins in Chinese Hamster Ovary Cells Using Sodium Butyrate to Elevate Gene Expression. J. Biotechnol. 1991, 19, 35–47. 85. Palmer, A. E.; Jin, C.; Reed, J. C.; Tsien, R. Y. Bcl-2-mediated Alterations in Endoplasmic Reticulum Ca2þ Analyzed with an Improved Genetically Encoded Fluorescent Sensor. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 17404–17409. 86. Papas, K. K.; Sun, L.; Roos, E. S.; et al. Change in Lactate Production in Myc-transformed Cells Precedes Apoptosis and Can Be Inhibited by Bcl-2 Overexpression. FEBS (Fed. Eur. Biochem. Soc.) Lett. 1999, 446, 338–342. 87. Park, D. S.; Stefanis, L.; Yan, C. Y.; et al. Ordering the Cell Death Pathway. Differential Effects of BCL2, an Interleukin-1-converting Enzyme Family Protease Inhibitor, and Other Survival Agents on JNK Activation in Serum/nerve Growth Factor-deprived PC12 Cells. J. Biol. Chem. 1996, 271, 21898–21905. 88. Pascoe, D. E.; Arnott, D.; Papoutsakis, E. T.; et al. Proteome Analysis of Antibody-producing CHO Cell Lines with Different Metabolic Profiles. Biotechnol. Bioeng. 2007, 98, 391–410. 89. Pattingre, S.; Tassa, A.; Qu, X.; et al. Bcl-2 Antiapoptotic Proteins Inhibit Beclin 1-dependent Autophagy. Cell 2005, 122, 927–939. 90. Rebollo, A.; Gómez, J.; Aragón, A. M. D.; Lastres, P. Apoptosis Induced IL-2 Withdrawal Is Associated with an Intracellular Acidification. Exp. Cell 1995, 218, 581–585. 91. Rodriguez-Enriquez, S.; He, L.; Lemasters, J. J. Role of Mitochondrial Permeability Transition Pores in Mitochondrial Autophagy. Int. J. Biochem. Cell Biol. 2004, 36, 2463–2472. 92. Rosser, M. P.; Xia, W.; Hartsell, S.; et al. Transient Transfection of CHO-k1-s Using Serum-free Medium in Suspension: A Rapid Mammalian Protein Expression System. Protein Expr. Purif. 2005, 40, 237–243. 93. Ruffolo, S. C.; Shore, G. C. BCL-2 Selectively Interacts with the BID-induced Open Conformer of BAK, Inhibiting BAK Auto-oligomerization. J. Biol. Chem. 2003, 278, 25039–25045. 94. Sauerwald, T. M.; Figueroa, B.; Hardwick, J. M.; et al. Combining Caspase and Mitochondrial Dysfunction Inhibitors of Apoptosis to Limit Cell Death in Mammalian Cell Cultures. Biotechnol. Bioeng. 2006, 94, 362–372. 95. Sauerwald, T. M.; Oyler, G. A.; Betenbaugh, M. J. Study of Caspase Inhibitors for Limiting Death in Mammalian Cell Culture. Biotechnol. Bioeng. 2003, 81, 329–340. 96. Schwarten, M.; Mohrlüder, J.; Ma, P.; et al. Nix Directly Binds to GABARAP: A Possible Crosstalk between Apoptosis and Autophagy. Autophagy 2009, 5, 690–698. 97. Shen, J.; Chen, X.; Hendershot, L.; Prywes, R. ER Stress Regulation of ATF6 Localization by Dissociation of BiP/GRP78 Binding and Unmasking of Golgi Localization Signals. Dev. Cell 2002, 3, 99–111. 98. Shore, G. C.; Nguyen, M. Bcl-2 Proteins and Apoptosis: Choose Your Partner. Cell 2008, 135, 1004–1006. 99. Shvets, E.; Fass, E.; Elazar, Z. Utilizing Flow Cytometry to Monitor Autophagy in Living Mammalian Cells. Autophagy 2008, 4, 621–628. 100. Simpson, N. H.; Singh, R. P.; Perani, A.; et al. In Hybridoma Cultures, Deprivation of Any Single Amino Acid Leads to Apoptotic Death, Which Is Suppressed by the Expression of the Bcl-2 Gene. Biotechnol. Bioeng. 1998, 59, 90–98. 101. Singh, R. P.; Al-Rubeai, M.; Gregory, C. D.; Emery, A. N. Cell Death in Bioreactors: A Role for Apoptosis. Biotechnol. Bioeng. 1994, 44, 720–726. 102. Sung, Y. H.; Hwang, S.-J.; Lee, G. M. Influence of Down-regulation of Caspase-3 by SiRNAs on Sodium-butyrate-induced Apoptotic Cell Death of Chinese Hamster Ovary Cells Producing Thrombopoietin. Metab. Eng. 2005, 7, 457–466.

Apoptosis: The Signaling Pathways and Their Control

479

103. Sung, Y. H.; Song, Y. J.; Lim, S. W.; et al. Effect of Sodium Butyrate on the Production, Heterogeneity and Biological Activity of Human Thrombopoietin by Recombinant Chinese Hamster Ovary Cells. J. Biotechnol. 2004, 112, 323–335. 104. Szegezdi, E.; Logue, S. E.; Gorman, A. M.; Samali, A. Mediators of Endoplasmic Reticulum Stress-induced Apoptosis. EMBO Rep. 2006, 7, 880–885. 105. Szegezdi, E.; Macdonald, D. C.; Tríona, N. I. C.; et al. Bcl-2 Family on Guard at the ER. Am. J. Physiol. Cell Physiol. 2009, 296, 941–953. 106. Tait, A. S.; Brown, C. J.; Galbraith, D. J.; et al. Transient Production of Recombinant Proteins by Chinese Hamster Ovary Cells Using Polyethyleneimine/DNA Complexes in Combination with Microtubule Disrupting Anti-mitotic Agents. Biotechnol. Bioeng. 2004, 88, 707–721. 107. Tey, B. T.; Singh, R. P.; Piredda, L.; et al. Influence of Bcl-2 on Cell Death during the Cultivation of a Chinese Hamster Ovary Cell Line Expressing a Chimeric Antibody. Biotechnol. Bioeng. 2000, 68, 31–43. 108. Tsujimoto, Y. Role of Bcl-2 Family Proteins in Apoptosis: Apoptosomes or Mitochondria? Gene Cell. 1998, 3, 697–707. 109. Wang, J. Beclin 1 Bridges Autophagy, Apoptosis and Differentiation. Autophagy 2008, 4, 947–948. 110. Warburg, O.; Poesener, K.; Negelein, E. Über den stoffwechsel der carcinomzelle. Biochem. Z. 1924, 152, 309–344. 111. Wei, M. C.; Lindsten, T.; Mootha, V. K.; et al. tBID, a Membrane-targeted Death Ligand, Oligomerizes BAK to Release Cytochrome C. Genes Dev 2000, 14, 2060–2071. 112. White, C.; Li, C.; Yang, J.; et al. The Endoplasmic Reticulum Gateway to Apoptosis by Bcl-xl Modulation of the InsP3R. Nat. Cell Biol. 2005, 7, 1021–1028. 113. Wong, D. C. F.; Wong, K. T. K.; Lee, Y. Y.; et al. Transcriptional Profiling of Apoptotic Pathways in Batch and Fed-batch CHO Cell Cultures. Biotechnol. Bioeng. 2006, 94, 373–382. 114. Wong, D. C. F.; Wong, K. T. K.; Nissom, P. M.; et al. Targeting Early Apoptotic Genes in Batch and Fed-batch CHO Cell Cultures. Biotechnol. Bioeng. 2006, 95, 350–361. 115. Xie, L.; Wang, D. I. C. Fed-batch Cultivation of Animal Cells Using Different Medium Design Concepts and Feeding Strategies. 1994. Biotechnol. Bioeng. 2006, 95, 270–284. 116. Yamaguchi, R.; Perkins, G. Dynamics of Mitochondrial Structure during Apoptosis and the Enigma of Opa1. Biochim. Biophys. Acta 2009, 1787, 963–972. 117. Yan, W.; Frank, C. L.; Korth, M. J.; et al. Control of PERK EIF2alpha Kinase Activity by the Endoplasmic Reticulum Stress-induced Molecular Chaperone P58IPK. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 15920–15925. 118. Yee, K. S.; Wilkinson, S.; James, J.; et al. PUMA- and Bax-induced Autophagy Contributes to Apoptosis. Cell Death Differ. 2009, 16, 1135–1145. 119. Yeo, J. H. M.; Lo, J. C. Y.; Nissom, P. M.; Wong, V. V. T. Glutamine or Glucose Starvation in Hybridoma Cultures Induces Death Receptor and Mitochondrial Apoptotic Pathways. Biotechnol. Lett. 2006, 28, 1445–1452. 120. Yi, X.; Yin, X.-M.; Dong, Z. Inhibition of Bid-induced Apoptosis by Bcl-2. TBid Insertion, Bax Translocation, and Bax/Bak Oligomerization Suppressed. J. Biol. Chem. 2003, 278, 16992–16999. 121. Yoshida, H.; Matsui, T.; Yamamoto, A.; et al. XBP1 mRNA Is Induced by ATF6 and Spliced by IRE1 in Response to ER Stress to Produce a Highly Active Transcription Factor. Cell 2001, 107, 881–891. 122. Zhang, J.; Robinson, D. Development of Animal-free, Protein-free and Chemically-defined Media for NS0 Cell Culture. Cytotechnology 2005, 48, 59–74. 123. Zhou, W.; Rehm, J.; Hu, W. S. High Viable Cell Concentration Fed-batch Cultures of Hybridoma Cells through On-line Nutrient Feeding. Biotechnol. Bioeng. 1995, 46, 579–587. 124. Zustiak, M. P.; Pollack, J. K.; Marten, M. R.; Betenbaugh, M. J. Feast or Famine: Autophagy Control and Engineering in Eukaryotic Cell Culture. Curr. Opin. Biotechnol. 2008, 19, 518–526.

1.35 Design Principles of Self-Assembling Peptides and Their Potential Applications P Sadatmousavi, M Soltani, R Nazarian, T Mamo, S Lu, W Xu, J Wang, P Chen, and M Jafari, University of Waterloo, Waterloo, ON, Canada © 2011 Elsevier B.V. All rights reserved. This is a reprint of P. Sadatmousavi, M. Soltani, R. Nazarian, T. Mamo, S. Lu, W. Xu, J. Wang, P. Chen, M. Jafari, 1.37 - Design Principles of Self-assembling Peptides and Their Potential Applications, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 495-509.

1.35.1 1.35.2 1.35.2.1 1.35.2.2 1.35.2.2.1 1.35.2.2.2 1.35.2.2.3 1.35.2.3 1.35.2.4 1.35.3 1.35.3.1 1.35.3.1.1 1.35.3.1.2 1.35.3.2 1.35.3.3 1.35.3.4 1.35.3.5 References

Introduction Design Principles of Self-Assembling Peptides Introduction Ionic Complementarity Molecular Structure Physical/Biochemical Properties Peptide Self-Assembly and Control Hydrogen Bonding Complementarity Amino Acid Pairing Applications of Self-Assembling Peptides Peptide-Mediated Drug Delivery Complexation of Ellipticine With Self-Assembling Peptides and Its Release Into a Cell Membrane Mimic Cellular Toxicity and Uptake of EAK16-II–Ellipticine Complexes Peptide-Mediated siRNA Delivery Tissue Engineering Biosensors Nanofabrication

480 481 481 481 482 482 483 484 485 486 486 487 488 490 492 493 494 494

Glossary Amino acid pairing The ability of the amino acids to interact with one another through at least one among ionic complementarity, hydrogen-bonding, hydrophobic, and van der Waals interactions. Amphiphilic Tendency to have both hydrophobic and hydrophilic properties. Hydrophilic Of tendency to absorb water and having strongly polar groups. Hydrophobic Tending to repel or fail to mix with water. Ionic complementary Arrangement of amino acids with alternative charges (þ þ ) in a peptide chain. Peptide (n. biochemistry) A compound consisting of two or more amino acids linked in a chain, a carboxyl group of one amino acid being joined to the amino group of the next amino acid by a peptide bound; –OC–NH–. Self-assembling (n. biology) The spontaneous organization of molecules under thermodynamic equilibrium conditions into structurally well-defined and stable arrangement through a number of noncovalent interactions. Small interfere RNA (siRNA) A class of double-stranded RNA molecules with 20–25 nt in length. Therapeutic agent Any drug or pharmaceutical compound for treating a condition.

1.35.1

Introduction

In the past few decades, engineering advanced materials in diverse applications have become tremendously popular. Designing novel nanobiomaterials in a desired fashion is a key element in the biotechnology industry. Self-assembly is ubiquitous in nature at both the macroscopic and microscopic level. Molecular self-assembly, the spontaneous organization of molecules under thermodynamic equilibrium conditions into structurally well-defined and stable arrangements through a number of noncovalent interactions, is currently one of the most important design approaches. The key in engineering robust self-assembly systems is to artfully design molecular building blocks that are able to undergo spontaneous assembly. Among self-assembling molecules (DNA, lipids, surfactants, and protein/peptide), peptides have attracted great attention in biotechnologic efforts to construct functional nano-/microstructures for diverse applications. These self-assembling peptides are

480

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00031-8

Design Principles of Self-Assembling Peptides and Their Potential Applications

481

either compartments of naturally existing proteins or proceed from novel biomolecular design. In both the cases, they are able to self-associate in aqueous solutions. Furthermore, synthetic peptides can be manipulated to self-assemble by taking advantage of the intrinsic property of amino acids. Depending on the application, the peptides can be easily engineered into various stable structures such as fibers, rids, tubes, vesicles, and globules. Another unique property of self-assembling peptides is their ability to form complexes with binding motifs such as aptamers, oligonucleotides, small interfere RNA (siRNA), and hydrophobic and hydrophilic anticancer drugs. Moreover, these peptides can be engineered in such a manner as to be able to penetrate cell membranes and recognize specific cell receptors. With such features, self-assembling peptides have a wide range of applications in nanobiotechnology.1–4,7–8 One common class of self-assembling peptides is ionic-complementary peptides, which possess a special property of ionic complementarity that results from their repetitive charge distribution in their amino acid sequence. These peptides also contain hydrophobic and hydrophilic components in their sequences, resulting in their special amphiphilic structures. They have recently emerged as promising nanobiomaterials for various applications in biomedical engineering and nanomedicine.1 These peptides gained popularity among material scientist following the work of Zhang et al. in 1993. In this work, a new class of ioniccomplementary peptide was derived from a short segment of a Z-DNA-binding protein in yeast. This short segment contained an unusual 16-amino-acid sequence with repetitive polar and nonpolar residues, AEAEAKAKAEAEAKAK (now called EAK16-II). The peptides form a unique amphiphilic molecular structure and self-assemble into b-sheet-rich nanofibers and macroscopic membranes. Recent works have shown that this class of peptides has immense potential for application as nanocarriers for drug delivery, as scaffolds for tissue engineering and as novel materials for regenerative medicine.4 The self-assembling peptides interact with other peptides or compounds via noncovalent interactions: electrostatic interactions, van der Waals forces, and hydrogen bonding. These forces can be utilized in the rational design of self-assembling peptides, a challenging undertaking. For example, for designing a peptide of five amino acids, 205 or 3.2 million possible sequences exist. Therefore, the number of possible sequence arrangements is nearly endless. Based on the applications and desired features, any peptide architecture can be designed, but the challenge is finding the rule to construct the peptide sequence into a structure. Researchers have made significant progress in designing peptides with specific properties for specific purposes. Design rules are derived either by copying from nature (alpha-helix, beta-sheets) or based on derivatives (peptide amphiphiles or cyclic peptides).5 In this work, we show the principle of designing amino acid pairing (AAP)-based self-assembling peptides. In this article, we discuss the design principle of self-assembling peptides through description of their molecular structure, physical/biochemical properties, and the control of peptide self-assembly. The three main categories of self-assembling peptide design, AAP, ionic-complementarity, and hydrogen bonding complementarity, are reviewed. Finally, the major applications of these peptides, including drug delivery, siRNA delivery, tissue engineering, biosensors, and nanofabrication, are highlighted herein.

1.35.2

Design Principles of Self-Assembling Peptides

1.35.2.1

Introduction

There are 20 natural amino acids that can work as building blocks for various nanostructures. All of these amino acids have the same basic structure but different R-groups at the central carbon (Ca) position of the molecule. Due to different properties of the R-group, amino acids can be categorized as hydrophobic, hydrophilic, or charged. Furthermore, hydrophobic residues can be divided into two groups: the aliphatic residues, including alanine (A), cysteine (C), glycine (G), isoleucine (I), leucine (L), methionine (M), and valine(V), and the aromatic residues, including phenylalanine (F), tryptophan (W), and tyrosine (Y). The aliphatic residues generate a general hydrophobic environment; meanwhile, the aromatic residues can also contribute to p–p stacking. The uncharged hydrophilic residues, including asparagine (N), glutamine (Q), proline (P), serine (S), and threonine (T), will generate hydrogen bonding with one another via either –OH or –CONH groups, except for proline. There are also charged residues, including positively charged histidine (H), lysine (K), and arginine (R) with pKa values of 6.5, 10, and 12, respectively, and negatively charged aspartic acid (D) and glutamic acid (E) with similar pKa of 4.4. The charged residues can provide electrostatic interaction that can be designed to help or prevent self-assembly. Besides the above-mentioned general properties, several amino acids contain specialized properties that either contribute to structure modification or offer sites for chemical modification, for example, the hydrogen atom as the R-group in glycine (G) removes steric hindrances imposed by the R-group in other amino acids, resulting in a high degree of flexibility. By contrast, proline (P) introduces structural rigidity due to the circular conformation caused by the hydrocarbon chain being covalently linked to the amino terminus. The thiol group in cysteine (C) provides a unique target for chemical modification, gold surface binding, and interpeptide cross-linking. Tyrosine (Y), serine (S), and threonine (T) can also be used as targets for chemical or enzymatic modification; meanwhile, repeat sequences of histidine can bind metal ions.5 Ionic complementarity, hydrogen bonding, and AAP are the main strategies to design self-assembling peptides by exploiting amino acid interaction properties.

1.35.2.2

Ionic Complementarity

Ionic-complementary peptides have several features that can be utilized in the self-assembling design. First, their charge distribution can be altered through simple molecular design. For example, the sequence design based on the same components can generate different charge distributions of types I, II, and IV. The differences in charge distribution may determine the peptide secondary structure and influence peptide self-assembly. Second, specific peptide chain lengths are required to exhibit ionic complementarity. Third, the ionic-complementary peptide family can be easily expanded by replacing different amino acids in the sequence.1 The peptide sequence, such as type, number, and arrangement of amino acids in sequence, serves as a key determinant in forming

482

Design Principles of Self-Assembling Peptides and Their Potential Applications

the secondary structures and self-assembled nanostructures of the peptides. In addition, the solution pH will influence the charged residues in ionic-complementary peptides and further affect peptide self-assembly behavior. The intramolecular electrostatic interactions can be controlled by adjusting the pH, which results in the change of secondary structure from b-turn to b-sheet.1,6

1.35.2.2.1

Molecular Structure

The molecular structure of ionic-complementary peptides derives from the unique arrangement of alternating negatively and positively charged amino acid residues. This ordered charge sequence results in unique electrostatic interactions that guide their molecular self-assembly, in addition to the usual hydrogen bonds and van der Waals forces. In order to guide molecular self-assembly through ionic complementarity, the charges should be distributed following a certain pattern. The most widely studied patterns among ionic-complementary peptides are type I, þ, type II, þþ, and type IV, þþþþ. Thus, the amino acid units are arranged in such a way that the negatively and positively charged units alternate in groups of one, two, or four. These arrangements can also be combined or repeated to form additional ionic-complementary peptides, although this would result in a complicated form of self-assembly. In addition to the unique arrangement of the charges, self-assembling ionic-complementary peptides also have a special arrangement of hydrophobic and hydrophilic amino acid residues alternating in sequence and leading to a unique amphiphilic structure that can have ‘side-to-side’ interactions as opposed to the ‘head-to-tail‘ structure of surfactants. For instance, for a b-strand peptide, hydrophobic residues are on one side of the peptide backbone and hydrophilic ones are on the other side. Clearly, the molecular self-assembly of ionic-complementary peptides is quite different from the micelle formation of surfactants. As indicated above, a special class of ionic-complementary peptides has been studied upon the discovery of the peptide known as EAK16-II, the 16-amino-acid peptide segment derived from a Z-DNA-binding protein. This peptide is of the type II group but also has derivatives in the form of type I and type IV groups: EAK16-I and EAK16-IV (Figure 1). Most peptides in this family can self-assemble into b-sheet-rich fibril structures, which result from the combination of three different interactions: hydrogen bonding from the peptide backbone, the electrostatic interaction from the ionic-complementary residues, and the hydrophobic interaction from the hydrophobic side. Following detailed studies of the molecular structure of EAK16-II, other and similar self-assembling ionic-complementary peptides with different functionalities are also appearing. A good example is the peptide RADA16-I that has been designed to enhance the biocompatibility and cell adhesion for tissue scaffoldings.1 The molecular structure of ionic-complementary peptides has several unique properties depending on the charge distribution, the chain length, and the ability to replace similar amino acids. First, the order and type of charge distribution can determine the peptide secondary structure and self-assembly. Second, peptide chain lengths determine the ability to exhibit ionic complementarity. The minimum number of amino acids required is 4, 8, and 16 for forming type I, type II, and type IV peptides, respectively. Third, replacing certain amino acids can expand the ionic-complementary peptide family. For example, the replacement of the hydrophobic residue alanine (A) of EAK with phenylalanine (F) or leucine (L) will create two other similar peptides, EFK and ELK, respectively. Such simple replacements maintain the intrinsic properties of ionic complementarity and self-assembly.

1.35.2.2.2

Physical/Biochemical Properties

One of the main properties of ionic-complementary peptides is the spontaneous formation of unusually stable b-sheets in aqueous solution1. These b-sheets are very stable under a wide range of physicochemical conditions such as extreme pH, high temperature, and dilution, and they are resistant to digestion with several proteases – including trypsin, a-chymotrypsin, papain, protease K, and pronase. They have also been shown to resist degradation by acidic or basic environments ranging from pH 1.5 to 11 and are stable upon treatment with sodium dodecyl sulfate/urea1. The b-sheet structures of EAK16-II can withstand up to 7 M concentration of the denaturing agents, guanidine–HCl, and 8 M urea (unlike normal proteins that denature above 4 M)1. These b-sheets can also

A −

+

−

+

−

B

−

−

+

+

C

−

−

−

−

+

−

−

+

+

−

+

+

+

+

+

Figure 1 Molecular structure of EAK16-I (A), EAK16-II (B), and EAK16-IV (C). Adapted from Fung SY (2008). Self-Assembling Peptides as Potential Carriers for the Delivery of Hydrophobic Anticancer Agent Ellipticne. PhD Thesis, University of Waterloo.

Design Principles of Self-Assembling Peptides and Their Potential Applications

483

aggregate into very stable membrane-like supramolecular structures that are mechanically stable. The strong stability of the membrane is mainly attributed to the durable b-sheets. In addition to these stabilities, the peptides were also found not to be immunogenic when injected into rabbits, rats, and goats1. These properties of the peptides, that is, stability in serum, high resistance to proteolytic digestion, simple composition, and lack of cytotoxicity, make them suitable candidates for various biomedical applications. The usually stable b-sheet structure of the peptides is explained through analysis of the molecular structures and interactions in the self-assembling ionic-complementary peptides. The main component of the b-sheet formation is the hydrogen bonding between individual peptide backbones. Hydrogen bonding is a major component in protein folding and aggregation; hence, its role in b-sheet formation is crucial. In addition to hydrogen bonding, the electrostatic intermolecular interaction due to ionic complementarity is mainly responsible in stabilizing the b-sheets. These interactions are absent in other types of peptides; hence, they are major components that explain the unusual b-sheet formation. Similarly, the hydrophobic interactions that arise due to the nonpolar components of the amino acids also play a major role in stabilizing the b-structure. Overall, ionic-complementary peptides self-assemble into stable nano-/microstructures through the stacking of b-sheet layers stabilized through hydrogen bonding and electrostatic and hydrophobic interactions. These biochemical properties of the peptides are what make them unique for various biomedical and nanomedical applications.

1.35.2.2.3

Peptide Self-Assembly and Control

To regulate the formation of nano-/microstructure, formulate peptide–drug complexes, and further control the complex size, understanding and precise control of peptide self-assembly are vital. Many internal and external factors such as amino acid sequence, molecular size, peptide concentration, solution pH, ionic strength, solvent, presence of denaturation agents, temperature, time, surface and its property, and mechanical force are effective in peptide self-assembly. Some of these effects are discussed here. Peptide sequence, which is an internal factor, tunes peptide self-assembly. In determination of the secondary structures and self-assembled nanostructures of the peptides, type, number, and arrangement of amino acids in the sequence are very important. For instance, EFK8-I forms a b-sheet secondary structure in an aqueous solution. However, replacing phenylalanine (F) with alanine (A) at the same charge distribution results in random coils. Although the reason for this phenomenon is not known yet, one possibility is that the steric hindrance of the hydrophobic phenylalanine may help to form b-sheets. Not only the type of amino acid but also the length of the peptide sequence is significantly important in nano-/macrostructure formation. Comparison of EAK16-II, EAK12, and EAK8-II shows that the first one associates to form a macroscopic membrane if it is dissolved in a salt solution. The second one, under the same condition, can also form a membrane, but to a much lesser degree, less than 50%. The third one cannot form a membrane under these conditions. One possible reason for these phenomena is that peptides with a shorter chain length have fewer ionic-complementary pairs that can affect self-assembly. A different charge distribution because of the arrangement of amino acids can change peptide self-assembly to different nanostructures. For example, EAK16-I and EAK16-II have a fibril-like nanostructure, but EAK16-IV has globular aggregates at neutral pH and forms at neutral pH (Figure 2). It is shown that EAK16-IV has a tendency to form a b-turn secondary structure via strong intramolecular electrostatic attraction. The assembly of b-turns may cause the formation of globular aggregates. By contrast, EAK16-II prefers a stretched b-strand, causing the formation of linear fibrils. Monte Carlo simulations can be used to verify the experimental results. The peptide concentration is another factor and has significant effects in peptide self-assembly and the nanostructure formation. The self-assembly of amphiphilic molecules such as surfactants is usually concentration dependent. For these molecules, there is also a critical concentration that governs the assembly process. Similar to the micelle formation of surfactants at a concentration above the critical micelle concentration (CMC), self-assembling ionic-complementary peptides are amphiphilic, and there is a critical aggregation concentration (CAC). This finding is expected because the CAC has been reported for many protein and amphiphilic peptides. Although the concentration-dependent self-assembly of most biomolecules follows a nucleation and growth mechanism, the CMC characterizes a threshold concentration above which micellization of surfactants takes place. Surface tension measurements show that the self-assembling ionic-complementary peptide EAK16-II has a CAC of  0.1 mg ml1 (60 mM). Atomic Force Microscopy (AFM) studies of the nanostructure of EAK16-II show that when the peptide

A

B

C

Figure 2 AFM images of peptide self-assembled nanostructures from EAK16-I (A), EAK16-II (B), and EAK16-IV (C). Adapted from Biomacromolecules Journal with License Number: 2570921325399 from Sadatmousavi P (2009). Peptide-Mediated Anticancer Drug Delivery. MASc Thesis, University of Waterloo.

484

Design Principles of Self-Assembling Peptides and Their Potential Applications

A

B

C

Figure 3 AFM images of EAK16-II on mica in various solutions after 30 min: (A) 1-mM HCl; (B) pure water; and (C) 1-mM NaOH. Scan area is 2000 nm  2000nm. Adapted from ACS Journal with License Number: 2570921021417 from reference Yang H, Fung S, Pritzker M, and Chen P (2007). Surface-assisted assembly of an ionic-complementary peptide: Controllable growth of nanofibers. Journal of American Chemical Society 129: 12200–12210.

concentrations go beyond the CAC, nanofiber networks are formed, at the same time as isolated filaments and globules are viewed at concentrations below the CAC. The fibril size and network density are also related to the peptide concentration. Likewise, the critical concentration of EAK16-I is 0.3 mg ml1 (180 mM). AFM studies show that the assembly behavior of EAK16-II and EAK16-I differs. EAK16-I assembly experiences two nanostructure transitions. When the peptide concentration is about 0.05 mg ml1 (30 mM), the first transition from globules to fibril morphology occurs; as the concentration goes up to 0.3 mg ml1 (180 mM) or higher, the second transition, which involves dramatic increase in the fibril size, begins. The solution pH in biological systems plays an important environmental role in affecting protein and peptide structures. The ionic state of the charged residues and the net charge of peptides/proteins are influenced by any changes in pH value. Therefore, the self-assembly behavior of the peptide and protein folding/aggregation will be further influenced. The reaction of the structures of self-assembling ionic-complementary peptides to the solution pH differs based on the peptide charge distribution. It is shown that EAK16-IV can be found in different nanostructural transitions – from globular aggregates to fibril-like nanostructures – based on the pH value, globular nanostructures at a pH between 6.5 and 7.5, and nanofiber networks at pHs beyond this range. By contrast, nanofibers occur in EAK16-II at a pH between 4 and 11. Based on this information, one can conclude that the charge distribution is a significant parameter in determining self-assembling peptide nanostructures. For a type IV charge distribution, at neutral pH, strong intramolecular electrostatic attractions take place and may result in molecular binding and the arrangement of globular nanostructures. At extremes of pH, 4 or 11, either lysine residues or glutamic acid are neutralized, causing weakened intramolecular electrostatic interaction and lower propensity for the peptide to make a b-turn structure. Therefore, nanofiber formation is predominant. In many biomolecular processes such as protein aggregation and peptide self-assembly, the surface/interface is a significant factor. It is shown that a hydrophilic surface (mica) assists the assembly of the self-assembling ionic-complementary peptide EAK16-II. In fact, the assembly kinetics of the peptide is considerably increased on the surface. The two steps in the surfaceassisted peptide assembly that follows a nucleation and growth mechanism are as follows. First, there is the nanofiber and fiber cluster adsorption on the surface to serve as nuclei or seeds; second, there is the fiber elongation from the active ends of the seeds. Nanofibers or fiber clusters in a state of equilibrium or close to this state can frequently grow on the surface. Surface-induced peptide assembly can be used to better understand and control protein aggregation related to conformational diseases. It should be noted that the surrounding environment parameters such as pH are important factors in the adsorption and assembly kinetics of peptide nanofibers on the surface. Increasing a solution pH decreases the amount of nanofiber ‘seed‘ adsorption; however, the nanofiber growth rate under different solution conditions follows this order: pure water > 1-mM HCl > 1-mM NaOH >10-mM HCl z10mM NaOH z0 (Figure 3). Therefore, a solution pH can adjust the adsorption of the ‘seeds‘ to the surface similar to the fiber growth rate by affecting the peptide–surface and peptide–peptide interactions, correspondingly. Thus, a way to control peptide assembly and resulting nanostructures on surfaces is given. Usually, molecular assembly can be influenced by surface properties to produce ordered nanostructures. For instance, experiments show that EAK16-II forms well-patterned nanostructures on a hydrophobic surface (HOPG) but randomly deposited nanofiber networks on the hydrophilic mica surface (Figure 4). For the well-ordered nanofiber patterns on HOPG, there are preferential orientations at angles of 60 or 120 degrees to each other that are similar to those in the crystallographic structure of graphite. The adsorption of peptides on the HOPG surface, like EAK16-II assembly on mica, is also considerably influenced by the solution pH. The peptide does not adsorb and assemble on the HOPG surface, in acidic environments, but in pure water and basic environments, patterned nanofibers can be formed on HOPG.

1.35.2.3

Hydrogen Bonding Complementarity

Among the 20 natural amino acids, 13 have the capability to interact via a hydrogen bond. These peptides can be further classified as soluble and less soluble, and the bond pairing selection is based on the position of proton acceptor and donor, and their solubility. Table 1 lists the amino acids that can participate in hydrogen bonding and the position of the atoms in the amino acids that serve as acceptors and donors.

Design Principles of Self-Assembling Peptides and Their Potential Applications

A

Figure 4

485

B

AFM images of EAK16-II on mica (A) and on HOPG (B) in various solutions after 30 min.1 Table 1

List of amino acids that can participate in hydrogen bonding and the position of the atoms in the amino acids that serve as acceptor and donor

Amino acid

H acceptor position (i)

H acceptor position (ii)

H donor position (i)

H donor position (ii)

R W Y K H D T C S N E M Q

4 4 5 5 3 3 2 2 2 3 4 3 4

6

5 5 6 6 5 4 3 3 3 4 5

7

4

3

5

Adapted from Chen P, Yang H, and Fung SY (2009). Amino Acid Pairing-Based Self Assembling Peptides and Methods, WO/2009/ 026729, 5 March 2009 patent (WIPO) WO/2009/026729.

The factors that affect the formation of secondary structures are similar to the ionic-complementary peptide system. The following example considers glutamine (Q)–asparagine (N) and asparagine (N)–serine (S) amino acid pairs. Q–N and N–S amino acid pairs are typical of most hydrophilic amino acid pairs except the pairs involving charged amino acids (E, D, R, and K), with 4-position and 3-position hydrogen bonding, respectively. The peptide sequences in these two pairings are n–QN–c, n–QNQN–c, n–QQNN–c, n–NS–c, n–NSNS–c, n–NSNSN–c, and n– NSNSNSNS–c. The secondary structure components are characterized with Fourier-transform infrared (FTIR) and circular dichroism (CD) spectroscopy and shown in Figure 5.6 Both of the graphs show a mixture of secondary structures, including b-sheets, a-helices, and random coils. Two trends can also be found in the graphs: first, the b-sheet content increases with an increase in the length of peptide sequence; second, the b-sheet content in the type I amino acid arrangement is higher than that in type II. This section has provided certain principles of AAP design. The focus was on three different complementary peptides, the ioniccomplementary peptide, the AAP and hydrogen-bonding-complementary peptides. These peptides either have alternating positiveand negative-charged distribution or have a hydrogen-bonding donor and acceptor in their structure, resulting in complementarity. The intermolecular forces – hydrogen bonding, hydrophobic interaction, or electrostatic interaction – which contribute to peptide self-assembly to different secondary structures are determined by the design of amino acid sequences. In addition to the structure of peptide self-assembly, the rational design of amino acid sequence also has a significant effect on the performance of peptide functional nanomaterials.1,6

1.35.2.4

Amino Acid Pairing

The traditional material fabrication process is through a top-bottom approach, which is at the limits of the difficulty of size reduction, the complexity of constructing functional materials, and the excess consumption of materials and energy.4 A bottom-top approach is based on molecular self-assembly and provides an alternative way to develop structurally well-defined novel materials. Peptides can interact with other peptides by different intermolecular forces: ionic interaction, hydrogen bonding, hydrophobic and van der Waals interactions, and p–p stacking. The strategy for designing self-assembling peptides called AAP is based on pairing

486

Design Principles of Self-Assembling Peptides and Their Potential Applications

10 β-sheet

–10

1.2

QNQN

Absorbance

QQNN

0.8

NS4

0.6

CD(mdeg)

QN

1

–30 –50 –70

QN

0.4

QNQN

–90

0.2

QNNN-300uM NSNSNSNS

0 1600

1620 1640 1660 1680 Wavenumber (cm–1)

–110 190 195 200 205 210 215 220 225 230 235 240 245 250 Wavelength (nm)

1700

CD spectra

FTIR spectra

Figure 5 Secondary structure of selected peptides detected by FTIR (left) and CD (right). Adapted from Chen P, Yang H, and Fung SY (2009). Amino Acid Pairing-Based Self Assembling Peptides and Methods, WO/2009/026729, 5 March 2009 patent (WIPO) WO/2009/026729.

Amino acid pairing (AAP)

Hydrogen bond pairing (HBP)

Others

All pairing (AP) Ionic pairing (IP) Van der waals pairing (VdWP)

QN and NS pairs

EAK16s

AC8

Figure 6 Amino acid pairing strategy. Adapted from Chen P, Yang H, and Fung SY (2009). Amino Acid Pairing-Based Self Assembling Peptides and Methods, WO/2009/026729, 5 March 2009 patent (WIPO) WO/2009/026729.

amino acids to interact with each other through at least one of the above interactions. AAP-based peptide design provides complementary interactions that achieve certain stereochemical and physicochemical stability, resulting in pair affinity and minimum pairing free energy. Figure 6 shows an AAP strategy for designing different self-assembling peptide systems. Among all these systems, ionic-complementary peptides are characterized by alternating positively and negatively charged residues; hydrogen bonding peptides are characterized by alternating hydrogen bonding pairs consisting of proton donors and acceptors; and all pairing in a peptide contains all of hydrogen bonding, electrostatic, and hydrophobic bonding amino acid pairs. These systems are divided based on the different amino acids sequences.6 Except when determining the interaction between peptides, the sequence of the peptides also plays a key role on the secondary structure and nanostructure of the self-assembled peptide aggregation, which influences their applicability in drug delivery, chemical sensing, biofuel cells, and models for the study of protein aggregation disease. Most ionic-complementary peptides can form stable b-sheets spontaneously in aqueous solutions; still, there are peptides that form a-helices and random coils because of the different sequences, although they may have the same charge distribution.

1.35.3

Applications of Self-Assembling Peptides

1.35.3.1

Peptide-Mediated Drug Delivery

To date, several drug delivery systems, such as liposome, polymers, micelles, and peptides, have been established for treatment and prevention of human diseases. They have the ability to encapsulate therapeutic agents and release entrapped drugs into body cells or affected tissues. An ideal delivery system should be able to control pharmacokinetics and pharmacodynamics, lower nonspecific toxicity, avoid immunogenicity, stabilize drugs for long circulation, target specific cells, and enhance therapeutic efficacy. Among

Design Principles of Self-Assembling Peptides and Their Potential Applications

487

many emerging drug carriers, peptides may overcome many challenges in delivery and increase drug biocompatibility, efficiency, and potency to bind with targeting and long circulating moieties. The natural properties of peptides, such as cell penetrating, targeting, and therapeutic characteristics, make them attractive for drug-delivery systems. Research on peptide-mediated delivery systems has shown very promising abilities to deliver therapeutic agents such as hydrophobic anticancer drugs, small molecules, and oligonucleotides. Self-assembling ionic-complementary peptides are novel biomolecules that can be used as carriers for drug delivery. They can be self-assembled to a desired structure through weak interactions such as hydrogen bonding, and electrostatic and hydrophobic interactions. Amphiphilicity is their unique property, which enables them to encapsulate hydrophobic compounds as well as hydrophilic gene therapeutics. In addition, they are relatively biocompatible and do not induce immunogenicity when introduced into animals. The most important advantage of peptide-based carriers is the ease with which their sequences can be manipulated and designed for various ends such as targeting and cell penetration. The ionic-complementary peptides EAK16-II and EAK16-IV, for instance, are able to deliver hydrophobic compounds. These peptides can spontaneously organize themselves into b-sheet structures that may stabilize the hydrophobic molecule, Ellipticine, in an aqueous solution.1,7 The typical self-assembling ionic-complementary peptide, EAK16-II, with the sequence of AEAEAKAKAEAEAKAK (Mw ¼ 1657 g ml1), has shown encapsulation of pyrene in a model hydrophobic compound. The encapsulated pyrene in the complexes is released into liposome as a cell membrane mimic. The release has occurred in a few hours, and the rate is proportional to the peptide–pyrene ratio. The results of the hydrophobic model show the possibility of the peptide being used to encapsulate hydrophobic anticancer drugs. In this research, Ellipticine has been selected as the hydrophobic anticancer agent for the following reasons. First, it has an anticancer activity that intercalates with DNA and inhibits topoisomerase II. Second, the fluorescence property of ellipticine makes it easy for researchers to monitor the interaction of ellipticine with peptides and the drug uptake by cells. Third, ellipticine is extremely hydrophobic, with a low water solubility of  0.62 mM. The hydrophobic side of peptides interacts with ellipticine and stabilizes it in an aqueous solution. Fourth, the severe side effects of ellipticine and its derivatives in clinical trials suggest a delivery system for this drug.

1.35.3.1.1

Complexation of Ellipticine With Self-Assembling Peptides and Its Release Into a Cell Membrane Mimic

Self-assembling peptide EAK16-II has been shown to readily encapsulate the hydrophobic anticancer drug, ellipticine. This section describes the effects of time and concentration on peptide–ellipticine complex formation. The release kinetics of ellipticine from a stable peptide–ellipticine complex are studies by monitoring the change of ellipticine fluorescence into egg phosphatidylcholine (EPC) vesicles as a cell membrane mimic. To prepare peptide–ellipticine complexes, a certain amount of ellipticine crystals are added to fresh peptide solutions (0.05–0.5 mg ml1) to obtain an ellipticine concentration of 0.1–1 mg ml1. Peptide powder is dissolved in pure water (18.2 MU, Milli-Q A10 synthesis) and then sonicated for 10 min. Ellipticine crystals are added into the peptide solution and stirred on a magnetic stir plate at 900 rpm for the duration of the complexation. To study the time dependence of the formation of peptide– ellipticine complexes, at specific times, the mixture is transferred into quartz cell to attain the fluorescence spectra of ellipticine on a steady-state spectrofluorometer (Photon Technology International), because ellipticine is fluorescent. The control samples, ellipticine in pure water and a peptide solution without ellipticine, are also prepared. The self-assembling procedure of peptides may interfere with the complexation of ellipticine and peptides due to peptide–peptide association. Therefore, the control peptide solution was stirred for 30 h at 900 rpm and was characterized by static light scattering. Fluorescence spectra of the ellipticine–peptide complex depend on the concentrations and show a peak located around 520 nm, which is a characteristic of protonated ellipticine, and a peak around  468 nm, which corresponds to microcrystals (Figure 7(A)). This latter peak increases with time and reaches the maximum after 6 h, then decreases. At the same time, the ellipticine form rises to a peak after 9 h, but there is no evidence of this peak initially. The intensity changes of the two peaks are plotted in Figure 7(B). The plot indicates that ellipticine is in a protonated state initially and turns to a crystalline state gradually. Ellipticine has the pKa  6, so it can be protonated in a weak acidic environment. The fresh EAK16-II peptide has a pH value of around 4.6, which can cause the protonation state of ellipticine. Glutamic acid residues in EAK16-II are negatively charged; hence, they can stabilize the protonated form of ellipticine. This protonated state of ellipticine diminishes after 6 h, before equilibrium is reached. This situation may be due to the peptide self-assembly over time under constant mechanical stirring. Through formation of the EAK16-II assemblies, the negatively charged glutamic acids are consumed, as they are complementary to the positively charged lysine residues. Meanwhile, the pH of the solution turns to  6.4, which is close to ellipticine pKa.7 The next study is a peptide concentration‘s effect on the complex formation at a fixed ellipticine concentration of 1.0 mg ml1 and peptide concentration ranges from 0.05 to 0.5 mg ml1. Figure 8 (A) and (B) shows the normalized fluorescence intensities of ellipticine over time at 468 and 520 nm, respectively. The fluorescence intensity at 468 nm increases initially and then reaches the equilibrium for all peptide concentrations. The equilibration time depends on the peptide concentration. The equilibration time increases with the increasing peptide concentration. Protonated ellipticine is also highly dependent on the peptide concentration. Protonated ellipticine stays longer (40 h) in a higher concentration of peptide 0.5 mg ml1. The CAC of EAK16-II, as mentioned before, is  0.1 mg ml1. At the concentrations below CAC, there is no significant protonated ellipticine, whereas at CAC and above, protonated ellipticine is more stable.7 The ellipticine concentration effect in complex formation is also investigated by fixing an EAK16-II concentration at 0.2 mg ml1 and three ellipticine concentrations at 1.0, 0.5, and 0.1 mg ml1. For both 468 and 520 nm fluorescence intensities, at the constant peptide concentration, the ellipticine concentration does not have a significant effect on the overall equilibration time. This time for

488

Design Principles of Self-Assembling Peptides and Their Potential Applications

A

0.05 0.5 h 1h 2h 4h 6h 8.1 h 9h 10 h 14 h 15 h

0.04

I/Is

0.03 0.02 0.01 0 370

420

470 520 Wavelength (nm)

570

B

0.04 468 nm 520 nm

I/IS

0.03

0.02

0.01

0

0

10

20

30

40

50

60

70

Time (h) Figure 7 The ellipticine fluorescence from the peptide–ellipticine suspension over time. (A) Fluorescence spectra of ellipticine as a function of time and (B) the normalized fluorescence intensities at 468 nm (diamonds) and 520 nm (squares) as a function of time. The ellipticine concentration is 1.0 mg ml1 and the peptide concentration is 0.2 mg ml1. Adapted from Fung SY, Yang H, Bhola PT, et al. (2009). Self-assembling peptide as a potential carrier for hydrophobic anticancer drug: Complexation, release and in vitro delivery. Advanced Functional Materials 19: 74–83. Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission.

the peptide–ellipticine formation for both states is  10 h. A particular complex with EAK16-II at 0.5 mg ml1 and ellipticine at 0.1 mg ml1 is examined in this study, which shows the prolonged stabilized protonated ellipticine. The very low intensity of the crystalline form is apparent, which indicates the negligible amount of ellipticine crystals in this particular formulation.7 The release of ellipticine from the peptide complex and the release kinetics was investigated by monitoring the release of ellipticine into EPC vesicles as cell membrane mimics. Four peptide concentrations (0.05, 0.1, 0.2, and 0.5 mg ml1) are used to stabilize ellipticine at 0.1 mg ml1 to study release kinetics. As discussed above, the complex with peptides at 0.5 mg ml1 is the exception in terms of the appearance and stabilization of protonated ellipticine for a prolonged time. The other complexes stabilize ellipticine in crystalline form. Ellipticine forms a neutral state ( 436 nm) in the presence of EPC, unlike in a peptide solution. Therefore, to investigate the release of ellipticine from a peptide complex into EPC vesicles, we have monitored the fluorescence signal at  436 nm. This signal is distinguishable from protonated ( 520 nm) and crystalline ( 468 nm) states of ellipticine. Figure 9 shows four kinetic profiles of ellipticine from different peptide–ellipticine complexes in EPC vesicles [EPTv] with time (h). All profiles exhibit a similar trend, with a fast rise initially and then a steadily approaching plateau. The complex with 0.5 mg ml1 EAK16-II shows the highest initial values of ellipticine in EPC and implies the burst release within 30 s. However, the other complexes have a slower release, with lower initial values of transfer profile. This finding is reasonable due to different states of ellipticine encapsulation by these peptides. The protonated ellipticine molecules can migrate into the lipid bilayer more easily than ellipticine microcrystals. The migration of microcrystals involves the dissolving of ellipticine, which is considerably slower than for protonated. There is a significant difference between the particle size and shape of different states of ellipticine in the complexes.7

1.35.3.1.2

Cellular Toxicity and Uptake of EAK16-II–Ellipticine Complexes

As discussed above, the ionic-complementary self-assembling peptide EAK16-II has shown great promise to encapsulate the anticancer drug ellipticine and release it into EPC vesicles as a cell membrane mimic. This section describes the cellular toxicity of the complexes and their uptake by cancer cells. Two cancer cell lines, the non–small cell lung cancer cell A549 and breast cancer cell MCF-7, were applied in this study. The complexes were prepared at different peptide-to-ellipticine ratios and then tested on cultured cells to evaluate cell viability using a 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay.

Design Principles of Self-Assembling Peptides and Their Potential Applications

489

A

+

+ + + + + + + +

0.02

+ + + + ++

−1 −1

EAK

0.2 mg ml

−1

0.1 mg ml

0.01

EAK

−1

0.08 mg ml

−1

−1

0

10

20

30

EAK

EAK

0.05mg ml 0 mg ml

0

EAK

0.5 mg ml

+ +

I/Is (468 nm)

+

0.03

EAK

40

50

Time (h) B

−1

0.5 mg ml

0.05

−1

0.2 mg ml

−1

0.1 mg ml

EAK EAK EAK

−1

+ +

0.08 mg ml

−1

0.05mg ml −1

0 mg ml

0.03

EAK

EAK

EAK

+

+

0.02 + ++ + + + + + + + + + + +

I/Is (520 nm)

0.04

0.01 0

0

10

20

30

40

50

Time (h) Figure 8 Effect of peptide concentration on the complex formation. The normalized fluorescence intensities of peptide–ellipticine suspensions as a function of time at 468 nm (A) and 520 nm (B). The ellipticine concentration was fixed at 1.0 mg ml1 with various EAK16-II concentrations ranging from 0 to 0.5 mg ml1. Adapted from Fung SY, Yang H, Bhola PT, et al. (2009). Self-assembling peptide as a potential carrier for hydrophobic anticancer drug: Complexation, release and in vitro delivery. Advanced Functional Materials 19: 74–83.

5

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + ++ ++

+

3

−1

0.5 mg ml

+

2

+ +

+

−1

++++ ++

1 0

0

EAK

−1

0.2 mg ml EAK

++

[EPTv] (µm)

4

0.1 mg ml

EAK

−1

0.05mg ml EAK

1

2

3

4

Time (h) Figure 9 The transfer profiles of ellipticine from different peptide–ellipticine complexes to the EPC vesicles. The complexes were made of 0.1 mg ml1 ellipticine with various EAK16-II concentrations: 0.05 (triangles), 0.1 (crosses), 0.2 (squares), and 0.5 mg ml1 (circles). The excitation and emission wavelengths are 295 and 436 nm, respectively. Adapted from Fung SY, Yang H, Bhola PT, et al. (2009). Self-assembling peptide as a potential carrier for hydrophobic anticancer drug: Complexation, release and in vitro delivery. Advanced Functional Materials 19: 74–83.

In this study, samples are prepared by the same method as described in the previous section. EAK16-II is prepared in five different concentrations 0.02, 0.1, 0.2, 0.5, and 1.0 mg ml1, while the ellipticine concentration is fixed at 0.1 mg ml1. The mixtures are stirred at 900 rpm for 24 h prior treatment. The cells are cultured in Dulbecco‘s Modified Eagle Medium (DMEM) containing 10% Fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37  C with 5% CO2. They are seeded into flat-bottom 96-well plates and incubated for 24 h. After 1 day, the old culture media is replaced with 150 ml fresh culture media, followed by the addition of 50-ml treatment. The plates are incubated for various incubation times before the cell viability assay.

490

Design Principles of Self-Assembling Peptides and Their Potential Applications

A

0.8 MCF-7 24 h

1:1 2:1

A549 24 h

Viability

0.6

0.4 1:5

5:1 10:1

0.2

0

0

200

100

300 −1

EAK16-II concentration (µg ml ) B

0.8 16x 8x

Viability

0.6

MCF-7 24 h

4x

A549 24 h

0.4

2x

0.2

0

0

10

20

30

−1

[EPT] (µg ml

)

Figure 10 Viability of MCF-7 and A549 cells treated with the complexes for 24 h at different peptide-to-ellipticine ratios (A) and upon serial dilution (B). The complex at 5:1 ratio was used for the serial dilution. The complexes were prepared with a fixed ellipticine concentration of 0.1 mg ml1 with various EAK16-II concentrations of 0.02–1.0 mg ml1. Adapted from Fung SY, Yang H, Bhola PT, et al. (2009). Self-assembling peptide as a potential carrier for hydrophobic anticancer drug: Complexation, release and in vitro delivery. Advanced Functional Materials 19: 74–83.

The cell viability of the incubated cells in the presence of complexes at different peptide-to-ellipticine ratios is shown in Figure 10(A) for both cancer cell lines. This figure indicates that the cellular toxicity of both ratios of 10:1 and 5:1 is higher than that for the other complexes (more than 0.75). At the ratios below 5:1 and at the ellipticine control, the efficacy decreases significantly due to the toxicity of the protonated ellipticine compared to that of the microcrystals. Protonated ellipticine interacts with negatively charged phospholipids in cell membranes, and the hydrophobic part of ellipticine helps it pass through the lipid bilayer. Note that the toxicity of the complexes in MCF-7 is more than A549 cells, perhaps due to the higher activity of the protonated ellipticine on MCF-7. The stability of a given formulation upon dilution is an important factor in determining its applicability in clinical usage. As the complexes at a 5:1 ratio show good anticancer activity against both cancer cells, a serial dilution of this complex in pure water is prepared and used to test their stability relation to the cellular toxicity. Figure 10(B) shows the toxicity of the serial dilution of the 5:1 ratio complex on both cell lines (2, 4, 8, and 16 times). Until a 2 times dilution, the complex is stable and has good activity; however, further dilutions considerably reduce the complex activity, indicating the instability of the complexes upon dilution in water. In order to evaluate time-dependent toxicity of EAK16-II–ellipticine complexes, both cell lines are incubated for 4, 8, 12, 24, and 48 h with complexes. The results are shown in Figure 11 for ellipticine control samples as well as complexes at a 5:1 and 1:1 ratio of peptide-to-ellipticine. The cell viability is significantly higher for complex 1:1 and the ellipticine control compared to 5:1 over the same time. Almost 0% viability is observed for both cell lines after 2 days treatment. All these observations indicate the capability of the EAK16-II–ellipticine complex to cause cell death in cancer cells. This complex is a potential biomolecule for use in drug delivery of various types, such as cancer therapy.

1.35.3.2

Peptide-Mediated siRNA Delivery

The discovery of RNA interference (RNAi) has revolutionarily changed genome and medical research. RNAi is a naturally occurring mechanism, whereas siRNA molecules turn off specific genes in a living cell to block the production of certain proteins before they

Design Principles of Self-Assembling Peptides and Their Potential Applications

491

A

1 1:1 5:1

0.8 Viability

EPT-H2O

0.6 0.4 0.2 0 0

10

20

30

40

50

Time (h) B

1

Viability

1:1 0.8

5:1

0.6

EPT-H2O

0.4 0.2 0

0

10

20

30

40

50

Time (h) Figure 11 Time-dependent toxicity of the EAK16-II–ellipticine complexes against MCF-7 (A) and A549 (B) cells. EPT-H2O: ellipticine control (in pure water). Adapted from Fung SY, Yang H, Bhola PT, et al. (2009). Self-assembling peptide as a potential carrier for hydrophobic anticancer drug: Complexation, release and in vitro delivery. Advanced Functional Materials 19: 74–83. Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission.

are made. This attractive technology can be used to elucidate the function of novel genes or treat human diseases by silencing genes that lead to disease in the first place. However, RNAi therapy, as is the case for most antisense or nucleic acid-based strategies, has achieved only limited success in clinical studies compared to its promising therapeutic potential in the lab. Because of its small size and hydrophilic property, a great amount of siRNA is excreted through the reticuloendothelial system when administered. Moreover, naked siRNA is subject to degradation by endogenous enzymes during circulation or within the cell. As the potency of the siRNA drugs is weakened, an increased dosage is needed to compensate for this effect. This need for increased dosages may be the main obstacle to the clinical use of nucleic acid-based drugs because genetic materials are rather difficult to produce on a large scale and may pose severe safety risks due to their carcinogenic potential and immunogenic effects. Repeated administration is, thus, contraindicated. With the increased potential of RNA interference as a therapeutic strategy, new methods for siRNA delivery are urgently needed. In the light of these problems, the carrier-mediated delivery system has emerged as a promising approach for improving the cellular delivery of siRNA. The carriers, covalently conjugated or associated with siRNA, are designed to enhance intracellular uptake and protect the siRNA incorporated against enzymatic or nonenzymatic degradation, while being biocompatible and biodegradable. Typically, a viral vector is the most effective carrier for gene delivery due to its selective-targeting property. However, safety risks due to inflammatory and immunogenic effects limit the clinical application based on viral vector strategies. Compared to viral vectors, nonviral carriers have lower immunogenicity. Actually, cationic liposomes, synthetic polymers, and polymeric nanoparticles represent a mature technology for both drug and gene delivery. Moreover, they can be easily modified and produced on a large scale at low cost. The real weakness of nonviral vectors is their very low levels of transfection rate. Therefore, a large number of vectors are needed, which may, in turn, cause severe toxicity problems. A peptide-based siRNA delivery system has been demonstrated to improve the cellular uptake of nucleic acids both in cultured cells and in vivo, representing an attractive method to solve the problem of poor membrane permeability. The underlying principle for peptide-mediated siRNA delivery initially derived from the general knowledge that the active sites of enzymes, receptors, and antibodies usually consist of 5–20 amino acids. Thus, it is reasonable to synthesize small peptides that could mimic the functional sites of proteins, especially the sites responsible for cell penetration and cell targeting, rendering peptide-mediated siRNA delivery system as efficient as viruses but without their limitations. Furthermore, the main attraction of peptides as siRNA carriers is their

492

Design Principles of Self-Assembling Peptides and Their Potential Applications

versatility. The different arrangement and combination of 20 natural amino acids, each of various sizes, hydrophobicity, and other physical properties, could yield multifunctional peptides with a large load capacity, high transfection efficiency, and specific targeting ability. In the 1990s, several proteins, termed ‘cell-penetrating peptides‘, were discovered to translocate spontaneously through cellular membranes, thus opening the way to the chemical synthesis of many mutations of this kind of peptide. However, we have been studying a new class of peptide – the AAP peptide – as siRNA carriers in the past 10 years. The AAP peptide primarily consists of two main parts: an AAP domain, which is responsible for self-assembly, and a cell penetration domain, usually an arginine and lysine-rich region, which is a functional group for cellular internalization. A linker segment is also applied to attach the main part. It is suggested that positively charged amino acids interact electrostatically with cellular membranes, which consist of negatively charged phospholipids. This kind of peptide is designed to be ionic complementary and geometrically matches with the phosphate backbone of siRNA. The mechanisms for the self-assembling of peptides generally involve electrostatic interaction, hydrophobic interactions, p–p stacking, and hydrogen bonding. These forces, combined with stereochemical and physicochemical stability, contribute to the final stable state of the peptide–siRNA complex. Advantageously, the AAP peptide can directly interact with its siRNA cargo through a stable noncovalent association, which makes siRNA easier to release in the cytosol than covalent-associated complexes. This feature distinguishes the AAP peptide vector from traditional drug-delivery vectors, which only enhance cell adhesion or cell membrane penetration and do not participate in the construction of the complex. In order to test the efficiency of a designed AAP complex in transfecting siRNA into mammalian cells, we run a high-throughput screening experiment on transformed cell lines, for example, mouse endothelial cells C166-GFP. These cells were transfected with a plasmid reporter, pEGFP-N1, which encodes the enhanced green fluorescence protein (eGFP). This signal can be detected by flow cytometer. If the designed peptide can deliver corresponding siRNA successfully into the cytosol, RNA interference will be performed to silence the specific mRNA encoding eGFP. As the production of eGFP protein is prevented in the first place, the fluorescence intensity should decrease after transfection until the siRNA is eventually degraded by enzymes in the cell. The experiment was performed with naked siRNA and siRNA-Lipofectamine2000 (a common transfection reagent) as normal and positive controls. Untreated cells were also tested under flow cytometer to obtain the base line for eGFP fluorescence. The silencing effect was monitored over an adequate time, usually 48 h, because GFP, which already exists in the cytosol before siRNA transfection, would still give off fluorescence until eventually degraded by proteases. Among all the peptides screened, half a dozen showed varying degrees of transfection efficiency compared to Lipofectamine2000. In particular, C1, one of these peptides, demonstrated an effect comparable to that of the positive control. Figure 12(A) shows the flow cytometry results for the untreated cells, cells treated with the positive control Lipofectamine2000siRNA, and cells treated with C1-siRNA complexes in 48 h. It can be seen that there is no significant difference between the fluorescence intensity of the C1 complex and the Lipofectamine2000 complex. The percentage of silencing is also shown in Figure 12(B). Toxicity is often a major obstacle when these carriers are in clinical use. There has been only limited successful use of liposome as a vector in vivo due to its toxicity. Therefore, an MTT assay was conducted to investigate the viability of cells after treatment by complexes. Figure 12(C) exhibits that the viability of cells treated by a C1 complex is higher than that for a Lipofectamine2000 complex, indicating that our peptide has a better biocompatibility than Lipofectamine2000.8

1.35.3.3

Tissue Engineering

Novel biological materials (especially biologically compatible scaffolds used as the substrate for cell growth, differentiation, and biological function) are important in tissue engineering. Several characteristics are necessary for a material to be an ideal biologically compatible scaffold to support cell attachment and growth, including units obtained from biological bases, fundamental components simply designed and adapted for particular purposes, biodegradable scaffolds, minimum cytotoxicity, logical cell–substrate interactions, fewest immune responses and inflammation, simple material generation, purification, and processing, ease of mobility, and chemical compatibility with aqueous solutions and physiological environments. Experiments show that ionic-complementary peptides such as EAK16-II and RAD16-II are good candidates for tissue engineering. These two peptides have been used to form matrices that support mammalian cell attachment. The ioniccomplementary peptide series can simply be adapted for different biological purposes, such as cell adhesion and cell membrane penetration. The main property of the peptide matrices is their cell adhesion, which is necessary in tissue engineering. Therefore, the peptide RAD16-II was considered to imitate the amino acid sequence RGD, which has been recognized as an adhesive identification sequence in fibronectin. The application of ionic-complementary peptides for tissue engineering is promising because of their ability to be used as substrates for neurite outgrowth and synapse formation. RAD16-I and RAD16-II also have illustrated some abilities for the above-mentioned application. To support neuronal cell attachment, differentiation, and extensive neurite outgrowth, these self-assembled scaffolds can be applied. It has been shown that they are also permissive substrates for functional synapse formation between the attached neurons. A lot of research has been done on RAD16-I matrices to demonstrate their potential in tissue engineering. For instance, in three-dimensional systems, they have been used to culture liver-derived stem cells. These stem cells have then been differentiated into functional hepatocyte-like cells. The migrating hippocampal neural cells can be entrapped by using the three-dimensional RAD16-I peptide nanofiber scaffolds. Culturing hippocampal slices on such a nanofiber layer, approximately 500-mm thick, between each tissue slice and the scaffold, a more extensive interface region was shaped. It should be mentioned that this simple technique can be applied both in developing technology for neural progenitor cell isolation or enrichment in vitro, and for expanding cells for the cell-based therapies of regenerative medicine.

Design Principles of Self-Assembling Peptides and Their Potential Applications

A

493

B

90

200

Untreated 80 70 Percentage of Silencing

#cells

150

Lipo 100 C1 50

60 50 40 30 20 10

0 100

101

102

103

0 Untreated

FL1-H:: FL1-H C

Naked siRNA

Lipo

C1

1.2 Naked

Complex with siRNA

1

Viability

0.8

0.6

0.4

0.2

0 Untreated

siRNA

Lipofectamin

C1

Figure 12 (A) Flow Cytometry results for (red) untreated cells, (green) treated cell with positive control Lipofectamine2000, (blue) treated cells with peptide-siRNA complex. (B) Percentage of silencing. (C) Viability of the cells after treatment with complexes. Adapted from Jafari M and Chen P (2009). Peptide Mediated siRNA Delivery. Current Topics in Medicinal Chemistry 9: 1088–1097. Copyright Bentham Science Publishers Ltd. Reproduced by permission.

As the materials have to concurrently stimulate high rates of cell division and the high rates of cell synthesis of phenotypically specific ECM macromolecules before the repair achieves a steady state of tissue maintenance, researchers have had difficulty in deciding on proper scaffolds for cartilage repair. Many members of ionic-complementary peptides are likely candidates for tissue scaffolds. For instance, it has been shown that EFK16-II hydrogel formation can be thermally and photochemically triggered through use of stimuli-responsive liposomes to release salts at a specific temperature or in response to near-infrared light exposure. Ionic-complementary peptides can be used as a scaffolding material in many applications in tissue engineering such as responsive gel formation, cartilage repair, and neurite outgrowth. These peptide matrices are both highly biocompatible and biodegradable. Their mechanical properties are also comparable, and their biological functionalities are diverse. Both the molecular design and the ability of these small peptide building blocks to self-assemble make them significant tools for future tissue engineering.4

1.35.3.4

Biosensors

Understanding and controlling molecular assembly at surfaces have become increasingly important for many nanotechnology and biological applications. Surface/substrate-assisted assembly has been explored in efforts to construct such nanoscopic devices as (bio)sensors, information storage units, optical computers, and solar cells. Recent developments in biomedical studies have exhibited the role of some surfaces to induce protein/peptide aggregations. However, due to the complexity of the biomolecules as well as the difficulties in the experimental nanoscale in situ examination of the resulting nanostructures, our understanding of the surface/substrate-assisted assemblies and aggregation kinetics of biomolecules remains incomplete. Utilizing in situ AFM,

494

Design Principles of Self-Assembling Peptides and Their Potential Applications

Figure 13 Fabrication of the Ag nanowires inside peptide nanotubes. Adapted from AAAS with License Number: 2298820878365 from Meital R and Ehud Gazit (2003). Casting metal nanowires within discrete self-assembled peptide nanotubes. Science 300: 625–627.

Yang and co-workers9 elucidated the surface/substrate effects on the nanoassemblies of a model ionic-complementary peptide, EAK16-II, on a negatively charged mica surface in order to better understand the assembly kinetics and further control the peptide aggregation on a surface. They have shown that modulating the peptide/peptide and peptide/surface interactions by adjusting the solution pH leads to rational control of the surface-assisted assembly.9 A biosensor is an analytical device applying for selectively detecting and measuring the amount of specific compounds in a given external environment. It consists of two major parts: a biochemical recognition system (bioreceptor) and a physicochemical transducer.10 The glucose biosensor is one of the most intensely studied biosensors because glucose detection is essential for both blood sugar monitoring of diabetes patients and quality control in food-processing industries. One challenge in the fabrication of an amperometric glucose biosensor is how to immobilize the active enzyme (usually glucose oxidase) onto the surface of electrodes. Various materials have been used for this purpose, such as polymers, sol–gels and self-assembled monolayers. Yang et al. have illustrated that the ionic-complementary peptide EAK16-II forms stable nanofibers on HOPG surfaces due to hydrophobic interactions. In addition, their findings have indicated that an EAK16-II-modified HOPG platform has an exciting potential for biosensing.10

1.35.3.5

Nanofabrication

In recent years, considerable advances have exposed the potential of certain self-assembling peptides in nanofabrication, in particular the possibility of fabricating conductive nanowires utilizing self-assembling peptide nanostructures as templates. Due to their natural biomineralization properties, self-assembling peptides are able to assemble into stable well-ordered nanostructures with the capability of being easily modified in order to bind with a range of inorganic compounds.2 Using these peptide sequences in conjunction with self-assembling peptides, various metallic nanofibers can be fabricated. Reches et al. have reported the use of aromatic FF (Phe–Phe) to form a template for Ag nanowire fabrication. They have shown that this peptide can self-assemble into well-ordered hollow, discrete, stiff nanotubes, and can serve as a template for the fabrication of Ag nanowires.11 By entering the silver ions in nanotubes at suitable conditions of solvents, Transmission electron microscopy (TEM) analysis has indicated the formation of silver assemblies within the majority of the tubes. They have also revealed that enzyme degradation of the peptide mold resulted in the attainment of individual silver nanowires that are 20 nm in diameter. The above-mentioned mechanism can be observed in Figure 13.

References 1. Fung, S. Y. Self-assembling Peptides as Potential Carriers for the Delivery of Hydrophobic Anticancer Agent Ellipticne, University of Waterloo, 2008. PhD Thesis. 2. Yang, H. Assembly of Ionic-complementary Peptide on Surfaces and its Potential Applications, University of Waterloo, 2007. PhD Thesis. 3. Hong, Y.; Legge, R. L.; Zhang, S.; Chen, P. Effect of Amino Acid Sequence and PH on Nanofiber Formation with Self-assembling Peptides EAK16-ii and EAK16-iv. Biomacromolecules 2003, 4, 1433–1442. 4. Chen, P.; Yang, H.; Fung, S. Y. Amino Acid Pairing-based Self-assembling Peptides and Methods, WO/2009/026729, 5 March 2009, 2009. 5. Ulijin, R. V.; Smith, A. M. Designing Peptide Based Nanomaterials. Chem. Soc. Rev. 2007, 37, 664–675. 6. Chen, P.; Yang, H.; Fung, S. Y. Amino Acid Pairing-based Self Assembling Peptides and Methods, WO/2009/026729, 5 March 2009 (??), 2009. 7. Fung, S. Y.; Yang, H.; Bhola, P. T.; et al. Self-assembling Peptide as a Potential Carrier for Hydrophobic Anticancer Drug: Complexation, Release and in Vitro Delivery. Adv.Funct. Mater. 2009, 19, 74–83. 8. Jafari, M.; Chen, P. Peptide Mediated SiRNA Delivery. Curr. Topics Med. Chem. 2009, 9, 1088–1097. 9. Yang, H.; Fung, S.; Pritzker, M.; Chen, P. Surface-assisted Assembly of an Ionic-complementary Peptide: Controllable Growth of Nanofibers. J. Am. Chem. Soc. 2007, 129, 12200–12210. 10. Yang, H.; Fung, S.; Pritzker, M.; Chen, P. Modification of Hydrophilic and Hydrophobic Surfaces Using an Ionic-complementary Peptide. PLoS One 2007, 2, e1325, 1–11. 11. Meital, R.; Gazit, E. Casting Metal Nanowires within Discrete Self-assembled Peptide Nanotubes. Science 2003, 300, 625–627.

1.36

Rational Design of Strategies Based on Metabolic Control Analysis

E Saavedra, S Rodrı´guez-Enrı´quez, H Quezada, R Jasso-Cha´vez, and R Moreno-Sa´nchez, Instituto Nacional de Cardiología, México DF, Mexico © 2011 Elsevier B.V. All rights reserved. This is a reprint of E. Saavedra, S. Rodríguez-Enríquez, H. Quezada, R. Jasso-Chávez, R. Moreno-Sánchez, 1.38 - Rational Design of Strategies Based on Meta olic Control Analysis, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 511–524.

1.36.1 Introduction 1.36.2 Fundamentals of Metabolic Control Analysis 1.36.2.1 Metabolic Modeling 1.36.3 Modulation of Clinically and Biotechnologically Relevant Metabolism 1.36.3.1 MCA for the Identification of Drug Targets in Human Parasitic Diseases 1.36.3.2 MCA for Drug-Target Identification in Cancer Cells 1.36.3.2.1 Antimitochondrial Therapy Against Malignant Neoplasias 1.36.3.2.2 OxPhos Inhibitors 1.36.3.2.3 OxPhos Uncouplers 1.36.3.2.4 mtDNA Destabilizing Agents 1.36.3.3 Enhanced Production of Amino Acids 1.36.3.3.1 Metabolic Engineering for L-tryptophan Production 1.36.3.3.2 Metabolic Engineering for L-phenylalanine Production 1.36.3.4 Enhanced Production of the Biogas Methane 1.36.4 Concluding Remarks Acknowledgments References

495 496 497 498 498 500 501 501 502 502 502 502 503 504 506 507 507

Glossary Concentration control coefficient It is the quantitative degree of control that a given step in a metabolic pathway, or cellular process, exerts on a given steady-state metabolite concentration. This coefficient has a positive sign for those reactions producing a metabolite X, whereas it is negative for those others consuming it. Delocalized lipophilic cations They are compounds with strongly hydrophobic nature and delocalized positive charge, properties that allow them to freely penetrate biological membranes and concentrate in negatively charged subcellular compartments such as mitochondria. Elasticity coefficient It is the quantitative measurement of the sensitivity of an enzyme (or transporter), or group of enzymes, toward the variation in a given ligand (i.e., substrate, product, inhibitor, or activator) during the functioning of a metabolic pathway, or cellular process, under steady-state conditions. Flux control coefficient It is the quantitative degree of control that a given step (i.e., enzyme, transporter, receptor, or kinase) in a metabolic pathway, or cellular process, exerts on flux (i.e., rate of end-product formation). A practical concept is the percentage of change in flux attained when a 1% change in the activity of a particular enzyme is achieved. Multiple targeting Strategy to inhibit or enhance a particular cellular function consisting in the simultaneous modification of several steps in a single metabolic pathway or various pathways involved in a cellular process. Metabolic modeling The set of coordinated simple and complex rate equations defining enzyme- (or transporter-) catalyzed reactions that are able to predict the behavior of the entire system under near-physiological conditions. Rate-limiting step It refers to the single component (i.e., enzyme, transporter, receptor, or kinase) in a given metabolic pathway or cellular process, which is usually assumed to exert full control over the flux, function, and/or metabolite concentration, ruling out the contribution of other components to the control of the system. Other synonyms are “key step” or “bottleneck.”

1.36.1

Introduction

The traditional experimental approaches used for changing the flux or the concentration of a particular metabolite of a metabolic pathway have been mostly based on the inhibition or overexpression of the presumed rate-limiting step. However, the attempts to manipulate a metabolic pathway by following such an approach have proved to be unsuccessful. The approaches

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00032-X

495

496

Rational Design of Strategies Based on Metabolic Control Analysis

used, in the identification of the pacemaker, key enzyme, “bottleneck,” rate-limiting step, or regulatory enzyme,1,2 have been the following: 1. Inspection of the metabolic pathway architecture. Due to cell economy and for reaching the highest efficiency, pathway control must reside in the enzyme localized at the beginning of a pathway or after a branch. 2. Determination of nonequilibrium reactions. Control resides in the step catalyzing the reaction with the lowest ratio between the mass action ratio (G; [products][substrates]) and its equilibrium constant (Keq): G/Keq 0.2%) variation in flux, then this enzyme exerts flux control. By contrast, if a 1% change in vi promotes a negligible change in flux (80–90%) invariably demonstrate that any protein from any metabolic pathway is essential and is a rate-limiting step because complete shutting down of the entire pathway or cellular process is attained. However, such high levels of enzyme inhibition are not usually reached in real therapeutics because drug inhibition regularly follows hyperbolic kinetics and, hence, total inhibition can only be achieved with massive doses, bringing about severe and multiple side effects. Hence, KO and RNAi strategies do not help to pinpoint drug targets with high therapeutic potential, thus precluding the use of these genetic methodologies for target validation. Therefore, the question that needs to be addressed for the relevant scientific task of drug targeting should be “How much is a metabolic pathway from the parasite affected when a given enzyme is partially inhibited?” or “How rate-controlling is an enzyme or transporter of a parasite metabolic pathway or cellular process?” Instead of tackling the problem by studying individual molecules, an integral approach to determine the actual impact of enzyme inhibition on pathway fluxes, metabolites, and cellular function seems necessary. Thus, “quantitative” measurements are required to identify the drug target(s), whose inhibition has greater impact on a parasitic pathway or cellular function. MCA provides a theoretical and experimental framework for the identification of the drug targets with the highest therapeutic potential in parasitic metabolic pathways. In principle, the enzymes with the highest flux or metabolite concentration control coefficients should be the most appropriate targets for SBDD. An enzyme with a flux control coefficient of 1 (a true rate-limiting step) should be the ideal drug target because, by decreasing its enzyme activity by 1%, a corresponding 1% decrease in pathway flux is obtained (Figure 1(a)). However, as previously discussed, MCA studies have experimentally demonstrated that the control of a pathway is shared by all pathway enzyme components, each one with different control degrees (control coefficients). Hence, drug targets should be searched among the steps with the highest control coefficients (>0.2) (Figure 1(a)), not being single-targeted but in combinations of two or more simultaneous targets (Figure 1(b)) for a successful alteration of the parasite’s metabolic pathways. Noncontrolling enzymes (with control coefficients 70%) has to be attained to significantly affect pathway functioning (Figure 1(a)).3 This is probably the case for TryR because 50–90% inhibition induced by genetic means did not have dramatic effects on T(SH)2 concentration. To date, reports of MCA applied to the study of parasite metabolic pathways for drug-target identification are scarce. A comprehensive analysis has been done for glycolysis in T. brucei. The infective stage (trypomastigote) of this parasite relies on glycolysis for ATP supply; thus, this pathway has been studied for therapeutic purposes. With the availability of a reasonable amount of information on the kinetic properties of purified enzymes as well as enzyme activities, metabolite concentrations, and fluxes within parasites, a kinetic model was built, which helped to determine the flux control coefficients of the pathway enzymes. It turns out that contrary to what is normally found in biochemistry textbooks that HK, ATP-PFK, and PYK are the rate-limiting steps of glycolysis and to what has been obtained by modeling glycolysis in erythrocytes by several groups, in which HK and the ATP demand are the main controlling steps, none of these enzymes was relevant in the parasitic pathway. Instead, the flux control mainly resided in glucose transporter (HXT) (50%), which, under other conditions, shifted to aldolase (ALDO), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), phosphoglycerate kinase (PGK), and glycerol-3-phosphate dehydrogenase (Gly3PDH).6 Therefore, it follows from these MCA studies that specific drug targeting of the parasite glucose transporter, together with other glycolytic middle-part enzymes, may have a better chance to succeed in the clinical treatment of this infectious disease. A kinetic model was built by our group for glycolysis in E. histolytica7 based on the full characterization of the amebal enzymes. Glycolysis is the main pathway for ATP generation because of the absence of functional mitochondria and OxPhos in the parasite; thus, drug targets have been explored in its carbohydrate metabolism. Amebal glycolysis deviates in several aspects from that in the human host. It contains two pyrophosphate-dependent enzymes, PPi-PFK and pyruvate phosphate dikinase (PPDK), which functionally replace the allosterically modulated ATP-PFK and PYK. Thus, with the exception of HK, amebal glycolytic enzymes are not allosteric and all of them catalyze reversible reactions, which suggested a different control distribution to that found in human cells. Because of their peculiarities, PPi-dependent enzymes have long been considered potential drug targets; however,

500

Rational Design of Strategies Based on Metabolic Control Analysis

they have low flux control coefficients. In fact, HK and HXT, together with the cofactor-independent phosphoglycerate mutase (PGAM), were the main controlling steps in amebal glycolysis. Moreover, it was predicted that specific HK and PGAM targeting should drastically diminish this essential pathway in the ameba and, hence, parasite survival.3,7 Metabolic modeling can also be used to improve the selectivity and effectiveness of drugs against parasitic human diseases. Named differential control analysis,8 the rationale behind this is to perform an analysis of the control distribution of a common metabolic pathway in both host and parasite, in order to identify the steps with the highest control in the parasite but low control in the host, thus allowing for the identification of the ideal drug targets. Thus, the selectivity of a drug was defined as the ratio of the response coefficient of the flux toward an inhibitor in the parasite divided by the corresponding response coefficient in the host as follows8: JðparasiteÞ

selectivity ¼

Rdrug

JðhostÞ

Rdrug

JðparasiteÞ

¼

Ci

iðparasiteÞ

$εdrug JðhostÞ

Ci

iðhostÞ

$εdrug

where RJdrug is the response coefficients of the flux J toward the drug (enzyme inhibitor), CJi are the flux control coefficients of enzyme i from the parasite and the host, and εidrug are their corresponding enzyme elasticity coefficients toward the inhibitor. If the parasitic and host enzymes have the same elasticities for the inhibitor and encounter the same inhibitor concentration, then the selectivity will only depend on the ratio of the flux control coefficients exerted by the parasite and host enzymes. The effectiveness of a drug to affect a metabolic pathway in the parasite depends on the response coefficient of the pathway flux to the inhibitor, which, in turn, does not only depend on how efficient a drug is to inhibit a particular parasite enzyme, but also depend on the control coefficient that the latter has on the entire pathway.8 Thus, an inhibitor can be very potent (in the nanomolar concentration) in inhibiting an isolated enzyme under in vitro conditions, but, if this enzyme has low control on the entire pathway flux or cellular function, then negligible perturbation of the parasite metabolic flux or function can be expected, and thus, no effect on the disease treatment. MCA through metabolic modeling can also help to identify the most adequate type of inhibitor (competitive, noncompetitive, uncompetitive, or mixed) that has to be designed for a particular enzyme in order to attain a significant perturbation of the parasitic pathway flux to obtain a significant biological effect. In general, SBDD leads to the synthesis of competitive inhibitors. However, these drugs are predicted to be the poorest in affecting a pathway flux because the immediate system response is the substrate accumulation of the inhibited enzyme, which may lead to eventually overcoming inhibition by binding competition. Thus, a computational analysis has been introduced, based on the selectivity concept described above (renamed as flux selectivity), to study the type of inhibitors and the concentration required to attain significant flux decrease by inhibiting each pathway enzyme. This analysis has indicated that for theoretical pathways with components following simple Michaelis–Menten (hyperbolic) kinetics, and arbitrarily chosen kinetic parameters, noncompetitive and uncompetitive inhibitors perform better than competitive inhibitors. Recently, a target identification computational program has been introduced for the systematic search of drug targets, effective inhibitor concentrations, and type of inhibitor in kinetic models. The use of functional genomics on carefully annotated parasite genomes for metabolic reconstruction in different parasites now appears feasible and may assist in drug-target identification for single or combined therapy. Integral, systems biology analysis for drug development against infectious diseases has been initiated, and also, application of new approaches such as metabolomics and fluxomics, in addition to the widely available transcriptomics and proteomics, together with powerful bioinformatic tools for the integration of such enormous amounts of information, will certainly help in the better understanding of the parasite metabolic networks and their control mechanisms and in the identification of the better drug targets for therapeutic intervention.

1.36.3.2

MCA for Drug-Target Identification in Cancer Cells

Search for the “magic cancer bullet” has traditionally focused on a single gene or protein assumed to be the “rate-limiting step” for tumor development and growth. Growth factor receptors, protein kinases, transcriptional factors, or metabolic enzymes are currently being targeted individually. However, many such attempts to block a metabolic or signal transduction pathway specifically by targeting a single rate-limiting molecule have also proven to be unsuccessful. Experimental MCA of cancer cells has provided an explanation for this phenomenon: several steps share the control of energy metabolism (for glycolysis: GLUT, HK, and branches for glycogen synthesis and ATP demand; for oxidative phosphorylation: respiratory complex I and ATP demand), that is, there is no single rate-limiting step.9 Targeting a step that does not exist is unlikely to be a successful, good paradigm for continued research into drug design for cancer treatment. The challenge is to move away from the design of drugs that specifically inhibit a single controlling step toward (1) unspecific, multi-site drugs or (2) drug mixtures in which several different proteins in the most exacerbated and unique pathways and cellular processes in cancer cells are targeted simultaneously. Indeed, the strategy selected by cancer cells to achieve an increased glycolytic flux to maintaining an adequate ATP supply is also multitargeted9: 1. most glycolytic enzymes and transporters are overexpressed; 2. there is an isoenzyme expression shift, from high inhibitor-sensitivity form to low inhibitor-sensitivity form; 3. there is a decrease in branching fluxes: partial inactivation of pyruvate dehydrogenase (PDH) complex leads to diminished pyruvate leak towards mitochondria; and 4. the plasma membrane transporter responsible for secreting lactate is also overexpressed.

Rational Design of Strategies Based on Metabolic Control Analysis

501

Therefore, cancer cells do not overexpress solely one rate-limiting step to increase glycolytic pathway flux and metabolite concentration but follow a variety of simultaneous strategies to achieve the desired objective. Current clinical treatment of cancer patients has anticipated the multi-targeted strategy: combinations of several antineoplastic drugs are usually administered for attaining high healing rates or increased index of survival of patients with any of the five major groups of tumors: lung, breast, prostate, ovarian, and colorectal. However, most anticancer drugs are rather unspecific as they also perturb other sites in both normal and cancer cells.10 Perhaps, this multi-site therapy might be improved by targeting the truly controlling steps of the most relevant cellular processes in tumors, thus decreasing the risk for undesired effects on healthy tissues. The differential control analysis described above may be helpful for selective drug targeting in cancer cells. Therefore, the most promising strategy for cancer treatment seems to be that of a multi-targeted, MCA-advised therapy. Thus, specific, potent, and cell-permeable inhibitors of the controlling steps of tumor glycolysis and OxPhos may prove to be suitable targets, along with specific inhibitors for other cancer cell processes. Drugs that target simultaneously, glycolysis and OxPhos, can be advantageous for cancer treatment as it has been described that there may be cellular subpopulations within solid tumors, some with a predominant glycolytic phenotype localized away from blood vessels in hypoxic regions and others with a predominant mitochondrial metabolism localized near blood vessels.

1.36.3.2.1

Antimitochondrial Therapy Against Malignant Neoplasias

Most cancer cells show an increased glycolytic activity, in comparison with the tissues of origin. Increased glycolysis in several neoplasias might certainly be the result of impaired mitochondrial respiration imposed by mitochondrial DNA (mtDNA) mutations or by the presence of damaged or diminished respiratory complexes, which, in turn, promote a deficient OxPhos flux (Warburg hypothesis).9 However, glucose deprivation or addition of glycolytic inhibitors (deoxyglucose (2-DOG) or 3-mercaptopicolinate) to fast-growth human and rodent carcinomas does not arrest tumor progression, until some mitochondrial inhibitors (rhodamines 123 or 6G, doxorubicin) are applied alone or in combination with antiglycolytic drugs. This observation indicates that at least in these tumor types, OxPhos is playing an important role in sustaining tumor proliferation. Nowadays, numerous cancer types have been experimentally evaluated in terms of their energy capacities, and the results suggest that the Warburg hypothesis cannot be applied to all tumor types. Therefore, the contribution of mitochondria to tumor development and, in consequence, its clinical importance as a hallmark for antineoplastic therapy is emerging. In this regard, several antineoplastic compounds have been identified that induce apoptosis through mitochondrial destabilization, cytochrome c release, and, finally, caspase activation through diverse mechanisms involving reactive oxygen species (ROS) overproduction. These drugs perturb the mitochondrial function, by acting as potent OxPhos inhibitors, uncouplers, and/or destabilizing mtDNA agents, in addition to affecting other pathways and cellular processes. It seems clear that such pathway targets are the main controlling pathways for ATP supply, growth, and survival in cancer cells. Identifying and targeting these key pathways may improve the current clinical treatment of cancer. These multiple-site anticancer drugs that also target mitochondria are listed below.10

1.36.3.2.2

OxPhos Inhibitors

Doxorubicin (Doxo), ditercalinium, cyclophosphamide, camptothecin, epirubicin, cyclophosphamide, and metal-based drugs (cisplatin, oxaliplatin, and casiopeinas) form DNA adducts, destabilizing DNA structure; however, all of them also seem to target mitochondria. In heart, liver, and Ehrlich tumor mitochondria, Doxo decreases oxygen consumption and inhibits cyclooxygenase (COX) and ATP synthase activities; in Raji tumor cells, the acyl-CoenzymeA (CoA) metabolism enzymes are downregulated after Doxo treatment. In healthy liver and heart mitochondria, Doxo strongly binds to anionic membrane phospholipids such as cardiolipin affecting the mitochondrial membrane fluidity and, in consequence, the functionality of several mitochondrial membrane enzymes (respiratory complexes I, III, and IV). In isolated normal mitochondria, Doxo also disturbs mitochondrial Ca2þ homeostasis and produces massive free radicals production, inducing mitochondria-dependent apoptosis. Nanoparticles of Doxo þ taxol, administered to mice bearing human neoplasias with glycoprotein P (GlycoP) overexpression, induced marked mitochondrial damage and decreased GlycoP activity. In some leukemia, ditercalinium inhibits cellular oxygen consumption with concomitant ATP diminution. In consequence, both pyrimidine and purine nucleotide syntheses are drastically abolished. Cyclophosphamide diminishes the activity of several enzymes of Krebs cycle (isocitrate, succinate, and malate dehydrogenases) and respiratory complexes I, II, III, and IV in heart cells; therefore, functionality of OxPhos is severely decreased. In tumor mitochondria, cisplatin, oxaliplatin, and casiopeinas disturb mitochondrial metabolism by (1) inhibiting Krebs cycle or respiratory chain enzymes; (2) disturbing the mitochondrial membrane potential; and (3) altering mitochondrial morphology. In renal and liver mitochondria, cisplatin also diminishes mitochondrial Ca2þ uptake, lowering the activity of the Krebs cycle calcium-dependent dehydrogenases. The redox-silent vitamin E derivative a-tocopheryl succinate (a-TOS) efficiently diminishes the proliferation of several solid malignant neoplasias such as mesothelioma, neuroblastoma, and prostate cancer, with no apparent toxicity against normal cells. In mitochondria, the drug inhibits respiratory complex II by interfering with ubiquinone binding, which results in increased ROS concentration, triggering mitochondrial destabilization and apoptosis. Some polyketide derivatives such as apoptolidin show potent inhibitory activity against some drug-resistant tumors and inhibit ATP synthase. Several delocalized lipophilic cations (DLCs) have been employed in clinical trials phase I due to their high selectivity toward tumor cells. DLCs include rhodamine derivatives (123, 3B, and 6G), F16, dequalinium chloride, AA1, and MKT-077. DLCs are greatly concentrated (10-fold more vs. normal ones) in tumor cells and mitochondria in response to their higher transmembrane

502

Rational Design of Strategies Based on Metabolic Control Analysis

electrical potentials (negative inside), strongly inhibiting succinate dehydrogenase (SDH), COX, and ATP synthase activities and thus inducing decreased OxPhos and mtDNA fragmentation.

1.36.3.2.3

OxPhos Uncouplers

Several agents used in adjuvant chemotherapy (tirapazamine, tetrahydrocannabinol, sulindac, and indomethacin) and other anticancer drugs (sulfonylurea-based, camptothecin, 1,4-anthracenediones, tetrahydrocannabinol, 5-fluorouracil, and gemcitabine) disrupt mitochondrial metabolism by virtue of their uncoupling effect (i.e., by collapsing the Hþ electrochemical gradient across the inner mitochondrial membrane). In turn, paclitaxel promotes mitochondrial membrane permeability transition pore opening and also apoptosis, cellular senescence, and increased ROS levels.

1.36.3.2.4

mtDNA Destabilizing Agents

Some anticancer drugs (Doxo, ditercalinium, and camptothecin) directly interact with mtDNA and/or affect mtDNA biogenesis through DNA polymerase gamma inhibition. Another mechanism associated with the diminution in mitochondria content and mtDNA is mitophagy activation. In this regard, imatinib increases the formation of autophagic vesicles and induces mitochondrial membrane disruption, and, finally, apoptosis.

1.36.3.3

Enhanced Production of Amino Acids

Amino acids, together with ethanol and antibiotics, represent the three main categories of fermentation products with commercial impact in the world market. Industrial fermentations are the main source of the amino acids L-glutamate, L-lysine, L-tryptophan, L-phenylalanine, and L-threonine; and the producer microorganisms are selected strains of Corynebacterium glutamicum and Escherichia coli. In the food industry, L-glutamate is used as a flavor enhancer; L-phenylalanine and L-aspartate are used for production of the sweetener aspartame; and L-lysine, DL-methionine, L-threonine, and L-tryptophan are used as feed additives. Enantiomerically pure amino acids are used as building blocks for pharmaceutical, cosmetics, and agricultural products.11 Strain improvement and advances in fermentation technology have contributed significantly to the recent growth in the amino acid market. Better understanding of C. glutamicum and E. coli physiology and development of genetic manipulation tools have allowed for the use of metabolic engineering strategies to significantly improve amino acid production. Strain improvement processes have changed from rounds of random mutagenesis and selection to rounds of genetic modification and physiological studies to thus identifying targets for further improvements.12,13 The former, early strategy rendered improved amino acidproducing strains but affected other characteristics that are relevant for industrial fermentations such as growth and sugar consumption rates or stress tolerance, whereas, for the latter, more recent strategy, the accumulated knowledge has permitted the construction of hyper-producing strains in a wild-type genetic background with robust growth and stress tolerance. Further, target identification is not a trivial issue. Genetic changes may lead to metabolic imbalances and reduction of growth, sugar consumption rates, and biomass yield. In the last decades, identification of targets for strain improvement has been done by overexpressing or inactivating selected genes and studying the physiological effects of such genetic modifications. Target identification has also been possible by metabolic flux analysis and metabolic modeling; these approaches describe the flux distribution through central and biosynthetic pathways, thus allowing for recognition of which fluxes should be restricted or augmented to direct the carbon flux toward the desired metabolite. Thus, the selection of the modified genes, proteins, or fluxes now seems not to be governed by the belief in the existence of a rate-limiting step (Table 1). Contradictorily, the successful strategy for the generation of hyper-producing strains has not been based on single modifications; instead, a proper set of genetic and biochemical changes has been simultaneously made to construct hyper-producing strains (Table 1). In general, to increase amino acid production, the following strategies have been implemented11–13: 1. 2. 3. 4. 5.

several pathway enzymes have to be simultaneously overexpressed; some of the overexpressed enzymes have to contain mutations that confer insensitivity to feedback inhibition; increase in the balanced supply of precursors; the pathway leaks have to be attenuated or fully blocked; the specific transporters that expel the respective amino acids have to be overexpressed to overcome re-uptake and utilization of the produced amino acid – these modifications attenuate end-product inhibition of the last enzymatic reactions, avoid reaching thermodynamic equilibrium, and facilitate downstream processes of purification; 6. elimination of the mechanisms that limit transcription of the relevant operons and translation of the biosynthetic enzymes (transcriptional repression and attenuation); and 7. further optimization of the culture conditions. In the following section, we describe the specific variety of strategies used to overproduce some amino acids.

1.36.3.3.1

Metabolic Engineering for L-tryptophan Production

Fermentations for L-tryptophan production use C. glutamicum-engineered strains. Initial attempts to generate hyper-producing strains were made by applying rounds of random mutagenesis and selection on appropriate media containing amino acid analogs. The final titers were around 4.4 g l1. Thus, the use of a single strategy to increase the production of tryptophan only rendered a marginal increase in final titers. These relatively low tryptophan productions contrast with the final titer of 58 g l1 reached

Rational Design of Strategies Based on Metabolic Control Analysis Table 1

503

Selected examples of the combination of metabolic engineering strategies to construct amino acid-producing strains

a

Elimination of attenuation of transcription and genetic controls includes gene overexpression with plasmids and inducible promoters. For L-glutamate production, optimization of fermentation conditions means the presence of inducers.

b

with the simultaneous use of six strategies. For this last work, a tryptophan-producing mutant of C. glutamicum derived through multiple rounds of mutagenesis was used; in particular, this strain was auxotrophic for tyrosine and phenylalanine, which decreases leaks of intermediates toward competing pathways and permits to control the feeding of this growth-limiting nutrient in fed-batch fermentations. In addition, several genes were amplified in order to increase the balanced supply of precursors and the activity of several enzymes of the pathway as well as to attenuate feedback inhibition, as detailed below: 1. the transketolase gene was overexpressed to increase the supply of erythrose 4-phosphate, one of the two precursors for 3-deoxy-D-arabino-heptulosonate 7-phosphate (3DAHP) synthesis, the first intermediate in the common pathway for aromatic amino acid biosynthesis (the other precursor is phosphoenolpyruvate); 2. a gene coding for a tryptophan-feedback resistant form of 3DAHP synthase, the first enzyme of the aromatic amino acid pathway, was overexpressed; 3. the 3-phosphoglycerate dehydrogenase gene was also overexpressed – this enzyme produces the key precursor (3-phosphohydroxypyruvate) for L-serine biosynthesis; L-serine is required in the last step of tryptophan biosynthesis; and 4. the tryptophan biosynthetic six-gene cluster, containing feedback resistant forms of anthranilate synthase and anthranilate phosphoribosyl transferase, was also overexpressed. It can be predicted that the successful simultaneous manipulations for the overproduction of this essential aromatic amino acid targeted the controlling steps of the pathway: the first and several end-segment enzymes. As a consequence, other steps very likely exert negligible flux and concentration control, in particular the plasma membrane permease that expels the amino acid, and the ATP and reducing equivalents supplying enzymes. On the other hand, some research groups have used a tryptophan-producer strain with beneficial but unknown mutations. Then, it would be interesting to test whether the use of other strategies, which have been proved to be beneficial in C. glutamicum or in E. coli, could further increase the amino acid yield, as long as they target steps with high control. These other strategies may be (1) prevention of tryptophan degradation; (2) use of nonionic detergents to promote the crystallization of secreted L-tryptophan; (3) prevention of tryptophan re-uptake; and (4) increased supply of phosphoenolpyruvate and reducing equivalents.

1.36.3.3.2

Metabolic Engineering for L-phenylalanine Production

Fermentations for L-phenylalanine production use E. coli engineered strains. The strategies used are similar to those described for tryptophan production: use of feedback-resistant and overexpressed enzymes, increased supply of precursors (erythrose 4-phosphate and phosphoenolpyruvate), use of classically obtained strains by random mutagenesis, L-tyrosine auxotrophy, elimination of genetic controls, and optimization of fermentation conditions. As expected, the best yields were reached when several strategies were used simultaneously. One example is the use of a C. glutamicum strain obtained by multiple rounds of random mutation and selection that overexpressed the three branch-point enzymes, which were insensitive to feedback inhibition: 3DAHP synthase, chorismate mutase, and prephenate dehydratase. This strain reached final titers of 28 g L-phenylalanine l1. Another remarkable example is a wild-type E. coli strain, which does not secrete L-phenylalanine and which was used as starting material to develop a hyper-producing strain that reached the final concentration of 45.5 g l1. The use of a defined wild-type strain,

504

Rational Design of Strategies Based on Metabolic Control Analysis

instead of a producer strain derived through multiple rounds of mutagenesis, may help to identify the precise molecular nature of each genetic modification. Sprenger13 has described the combined utilization of the following five strategies: 1. Auxotrophy for L-tyrosine. This modification decreases leaks of prephenate toward L-tyrosine biosynthesis, thus allowing for a suitable L-tyrosine feeding regulation in the production phase of fermentation, which serves to limit biomass growth (this was an important parameter for optimization of process conditions). 2. Overexpression of feedback-resistant isoforms of key enzymes. Examples of this strategy are the overexpression of a 3DAHP synthase isoform insensitive to L-tyrosine inhibition and of the bifunctional enzyme chorismate mutase–prephenate dehydratase isoform insensitive to L-phenylalanine inhibition. 3. Elimination of attenuation of transcription and genetic controls. The expression of the genes coding for the feedback-resistant enzymes was under the transcriptional control of an inducible promoter on a medium-copy-number vector; in this way, gene dose was increased with transcriptional regulation becoming independent of the repressor TyrR and of the attenuation of transcription. The use of these three strategies promoted L-phenylalanine final titers of 25–30 g l1. 4. Prevention of accumulation of intermediates in the biosynthetic pathways. In spite of the significant L-phenylalanine production, byproducts are inevitably formed. Accumulation of the common pathway intermediates 3DAHP, shikimate, and 3dehydroshikimate suggested that dehydroquinate synthase and shikimate kinase activities were exerting flux control toward L-phenylalanine. However, overexpression of these two enzymes significantly lowered formation of byproducts but did not elevate L-phenylalanine production, indicating that these enzymes did not control flux but metabolite concentration. This last observation illustrates that modification of metabolic fluxes is not achieved by simple sum of strategies; instead, MCA and physiological studies are required to understand how a pathway is controlled and thus detect what enzymes and transporters may have control on pathway flux and metabolite concentration. 5. Optimization of fermentation conditions. Surprisingly, further increases in production were achieved by the use of a wild-type strain, and, hence, in the presence of the L-tyrosine feedback-sensitive isoform of 3DAHP synthase. Under strict limited L-tyrosine feeding, a final titer of 38 g l1 with virtually no byproduct formation was achieved. Further improvement was accomplished by extending the fermentation time and using a new reactive extraction procedure, reaching an L-phenylalanine production of 45.5 g l1. In turn, by using an L-tyrosine-deficient mutant expressing feedback-resistant isoforms of the bifunctional enzyme chorismate mutase-prephenate dehydratase and 3DAHP synthase into a temperature-controllable expression vector, an 1 L-phenylalanine production of 46 g l was attained. This genetic tool used in the E. coli strain MWPWJ 304 bearing the plasmid pMW16, which is used for industrial scale L-phenylalanine production, yielded final titers of 51 g l1.11 It would be interesting to test whether L-phenylalanine production can be further improved by combining other strategies with potentially high control such as increasing the supply of the precursors erythrose 4-phosphate and phosphoenolpyruvate and preventing L-phenylalanine re-uptake.

1.36.3.4

Enhanced Production of the Biogas Methane

Methanogenesis is the pathway by which some microorganisms synthesize ATP necessary for cell duplication with the concomitant generation of methane as end product. Methanogens are strict anaerobes belonging to the Archaea phylogenetic domain. Therefore, these microorganisms are found in anaerobic environments such as the rumen and lower intestinal tract of animals, anaerobic digesters of sewage treatment plants, landfills, the sediments of freshwater wetlands, rice paddies, ponds, streams, lakes, and in the sea. It is well documented that only a few carbon sources can be utilized by methanogens. CO2 þ H2, methanol, methylamines, CO, formate, and acetate are the only substrates known to be used for growth and methane production, but it is from acetate that 75% of total methane in the earth is produced. Methanosarcinales are a select group of methanogens with the ability to produce methane with any of the substrates mentioned above. The marine archaeon Methanosarcina acetivorans genome has been completely sequenced, and genetic tools are available for manipulating its genetic background and metabolism in order to obtain methane overproducing strains. Although proteome analysis has identified the expressed enzymes when acetate is present in the culture media of M. acetivorans, the thermodynamic, kinetic, and structural properties of the enzymes and transporters involved in the aceticlastic pathway have not been determined, and hence, no studies on flux and concentration control of methane production have been carried out. Therefore, it is first required to fully elucidate the pathway architecture, before embarking on the more complex experimental analysis of the methanogenesis control structure. The reactions involved in the pathway are described below. Analysis of the kinetic data of the aceticlastic pathway enzymes reported for Methanosarcina sp. may help to initiate the identification of the steps that may have the main control on methane production. Acetate-utilizing methanogenic organisms cleave acetyl-CoA to produce a methyl group that is reduced to methane, with electrons derived from the oxidation of the acetyl-CoA carbonyl group to CO2: CH3 COO þ Hþ /CH4 þ CO2

0

DGo ¼ 36 kJ mol1

Acetate kinase reversibly catalyzes the first reaction in acetate activation: CH3 COO þ ATP4CH3 CO2 PO2 3 þ ADP

Rational Design of Strategies Based on Metabolic Control Analysis Table 2

505

Kinetic constants of the methanogenesis enzymes Forward reaction

Enzyme

Km (mM)

Vmax (mmol min1 mg1) or kcat (min1)

Acetate ¼ 22 kcat ¼ 0.73  105 ATP ¼ 2.2 PhosphoCoA ¼ 0.065 kcat ¼ 3.1  105 transacetylase Acetyl-P ¼ 0.18 CODH/Ac-CoA AcCoA ¼ 0.2 CO ¼ 5 kcat ¼ 15 synthase CoMCH3H4MTP ¼ 0.05 Vmax ¼ 0.6–12 methylTHMPT CoM ¼ 0.25 transferase Methyl-CoM MethylCoM ¼ Vmax ¼ 2.5 reductase 1.5 CoB ¼ 4.0 Carbonic CO2 ¼ 0.8 anhydrase E. coli-like Fdred ¼ 0.0075 Vmax ¼ 90 hydrogenase F420H2 F420H2 ¼ 0.005 kcat ¼ 1530 dehydrogenase Heterodisulfide Hydroxyphenazine kcat ¼ 4200 reductase H2 ¼ 0.09 CoBCoM ¼ 0.14 Acetate kinase

Reverse reaction

kcat/Km (min1 M1)

Km (mM)

5.5  104

Acetyl-P ¼ 0.5 ADP ¼ 0.1 AcCoA ¼ 0.09 Pi ¼ 0.74 CoA 90% of CF fatalities. Biofilm formation by P. aeruginosa has been the subject of intense investigation for a number of years. Consequently, we know more about the environmental cues and genetic elements influencing biofilm formation in this organism than any other. Much of this article focuses on biofilm development in P. aeruginosa and closely related bacteria.

1.39.2

Model Systems for Growing and Analyzing Biofilms

Over the past two decades, there have been significant advances in imaging techniques and molecular tools for studying biofilm development. A variety of technologies are available for growing and analyzing biofilms in the laboratory. Although the biofilms formed in experimental systems differ from those in environmental and medical settings, their analysis has provided a wealth of information regarding factors influencing biofilms. It is expected that the relevance of these findings will extend beyond the laboratory to biofilms in nature. Model biofilm setups can be classified as either static or flowing systems. Examples of each are discussed below along with microscopic imaging techniques for visualizing biofilms.

Figure 1 Biofilm-contaminated water pipeline. Biofilm microbes are very resistant to chlorine and can be released into the drinking water when there is a pressure change in the water distribution system. Photo courtesy of Early Warning Inc.

Biofilms 1.39.2.1

531

Biofilms Formed Using Static Systems

Static biofilms have a number of advantages over flowing systems. Static systems require no specialized equipment, and they are inexpensive and simple to set up. Frequently, static biofilms are grown in 96-well microtiter plates, a format which is amenable to high-throughput screening. This enables a mutant library or an array of culture conditions to be easily screened to determine their impact on biofilm formation. A modification of the 96-well plate assay has been designed for analyzing the effect of antimicrobial compounds on biofilms. This system employs a specialized lid with 96 pegs on which the biofilms form. After biofilm development, the lid is transferred to a 96-well plate containing dilutions of antimicrobial compounds. Removal of cells from the pegs and viable plate counting enables researchers to determine how effective these substances are at killing biofilm bacteria. The main drawback of the tray system is that bacteria are not supplied with fresh nutrients and they are not well aerated. These factors limit analysis to the early stages of biofilm development. Another type of static biofilm system is the colony biofilm assay. This technique is ideal for mimicking natural biofilms not typically bathed in liquid. Colony biofilms are established by inoculating bacteria onto a semi-permeable membrane atop of an agar plate. Once developed, the biofilms can be tested for sensitivity to antibiotics, ultraviolet (UV) radiation, and oxidative or other environmental stresses.

1.39.2.2

Growth and Microscopic Analysis of Biofilms Formed in Continuous-Culture (Flowing) Systems

A number of continuous-culture devices have been used to study biofilm development including the Robbins device, the rototorque bioreactor, the constant-depth film fermentor, and the flow cell apparatus.4 Flow cells offer several advantages over other systems including their simple design and the ability to directly image biofilms in a nondestructive manner. Not surprisingly, they have become the system of choice for many researchers. To image biofilms formed in a flow cell, standard light and epifluorescence microscopy can be used; however, as the biofilm increases in thickness, it becomes increasingly more difficult to acquire highresolution images. This limitation can be overcome with the use of confocal laser scanning microscopy (CLSM), which generates high-resolution images of thick specimens. During CLSM, images are acquired from selected depths using a process known as optical sectioning. Initially, the object is scanned with a laser over multiple planes interspersed by short distances. Through assembling a stack of two-dimensional images from successive focal planes, the computer is able to generate a virtual three-dimensional image of the biofilm. To view cells growing as a biofilm, they are made fluorescent by treating with dyes or tagging with fluorescent reporters. A variety of fluorescent dyes are available for visualizing biofilms. For example, dyes are commonly used as indicators of cell viability after a biofilm has been exposed to an antimicrobial agent. This approach uses cell membrane integrity together with nucleic acid probes with different membrane permeabilities to assess viability. The commercially available Live/Dead BacLight Bacterial Viability Kit (Invitrogen) is composed of two reagents that bind nucleic acids. The green fluorescing compound, Syto 9, readily permeates membranes; consequently, living cells are stained green. The red fluorescing compound, propidium iodide, cannot permeate intact membranes and so dead cells are stained red. To analyze the viability of flow cell biofilms, the two-component dye solution is injected into the chamber and allowed to incubate in the absence of flow. Approximately 15 min later, the residual stain is removed by allowing fresh media to perfuse the biofilm. The flow cells can then be analyzed using SCLM to determine the ratio of green cells to red cells as an indicator of cell viability (Figure 2). Although such stains are useful, there are inherent problems associated with them. For instance, the biofilm matrix can limit diffusion, so cells within the biofilm interior may be excluded. Furthermore, these compounds are usually toxic to bacteria, making staining a terminal step. For this reason, real-time analysis of biofilm development is not possible using conventional staining techniques. The discovery that the gene encoding green fluorescent protein (GFP) from the jellyfish Aequorea victoria can be expressed in heterologous organisms has led to its widespread use as a fluorescent tag for eukaryotic and prokaryotic cells as well as a reporter

Figure 2 Cells within a Pseudoalteromonas tunicata biofilm visualized using a Live/Dead bacterial stain. Live cells are green and dead cells are red. These structures are extremely tolerant to antimicrobial compounds and other environmental assaults. Cells in the center of the biofilm have reduced access to nutrients and higher exposure to metabolic waste products, which can result in cell death as indicated by red-stained cells. Photos courtesy of Dr. Jeremy Webb, University of Southampton, Southampton, UK.

532

Biofilms

for gene expression (Figure 3). GFP is composed of 238 amino acids, which upon absorbing blue light emits a bright green fluorescence. GFP has a number of favorable properties. For instance, apart from molecular oxygen that is needed for the internal rearrangement of the protein into its fluorescent form, no special substrates are required. What’s more, GFP expression is not toxic to cells so biofilm development can be microscopically imaged over time. To increase the range of potential applications, several derivatives of GFP have been engineered. A single point mutation (S65T) dramatically improved the photostability and spectral properties of GFP. The excitation maximum of this enhanced GFP was shifted to 488 nm with a peak emission at 507 nm, resulting in spectral characteristics that are compatible with fluorescein filter sets. Additional derivatives, such as gfpmut3, have been created with 20 times increased fluorescence compared with the native GFP. Red-shifted or blue-shifted variants have provided proteins with diverse spectral properties including cyan fluorescent, yellow fluorescent, and blue fluorescent protein derivatives. The availability of these GFP variants allows the comparison of two or more protein localizations or simultaneous monitoring of gene expression from different promoters. For temporal gene expression analysis, a series of short half-life GFP derivatives have been engineered. The wild-type GFP protein is extremely stable with a half-life of several days. Fusing a gene of interest to a promoterless copy of the wild-type gfp allows one to determine if and when a gene is turned on. However, using the wild-type GFP, it is not possible to detect downshifts in gene expression because cells will fluoresce green long after a gene has been turned off. To circumvent this problem, various groups have generated unstable variants by the addition of a short peptide tail to the C-terminal end of the wild-type GFP.1 These tail sequences render the GFP proteins more sensitive to degradation by endogenous cellular proteases, greatly reducing their half-life (Figure 4). The unstable GFP variants have proved to be very useful for temporal analysis of gene expression in biofilms as well as other applications.

A

B

Figure 3 Comparison of Pseudomonas aeruginosa PAO1 expressing a rhlI-wild-type gfp promoter fusion ((A) GFP half-life ¼ several days) versus the same promoter fused to an unstable gfp ((B) GFP half-life < 1 h). All of the cells expressing the stable GFP are intensely fluorescent, whereas cells harboring the unstable GFP fusion show varying degrees of fluorescence, which correlates with differences in gene expression.

A

B

C

D

E

F

Figure 4 Scanning confocal laser micrographs of a Pseudomonas aeruginosa PAO1 biofilm. The bacteria contain a rhlI–gfp fusion so cells that appear green are expressing the rhlI gene. The biofilm was counterstained with a combination of propidium iodide and Syto 85, which stains the total cell population red. Panels (A)–(C) are optical sections of the biofilm at depths of 0 mM (substratum (A)), 9 mM (B), and 18 mM (C) showing both the green and red cells. Panels (D)–(F) represent composite images of sections taken throughout the biofilm. Panel (D) shows cells expressing rhlI–gfp; panel (E) shows the total cell population; and panel (F) is an overlay of two micrographs illustrating where gene expression is occurring in the biofilm population.

Biofilms

1.39.3

533

Heterogeneity in Biofilms

Unlike planktonic cultures, bacteria growing as a biofilm exhibit a high degree of physiological heterogeneity. This is brought on by heterogeneity in environmental conditions, to which the cells must respond accordingly. It is easy to imagine that bacteria at the surface of the biofilm experience a very different set of conditions compared with those in the biofilm interior. A combination of bacterial metabolic activity and diffusion properties leads to gradients of nutrients, waste products, and signaling compounds.11 Therefore, a biofilm should be considered to be a group of cells in a wide range of physiological states rather than a uniform culture. It has been proposed that most of the physiological heterogeneity which arises in biofilms is mediated by chemical heterogeneity. Oxygen penetration through the biofilm is one parameter that has been extensively studied using microelectrodes. Looking at biofilms formed by aerobic or facultatively anaerobic bacteria, oxygen concentrations decline as the probe moves from the bulk fluid to the inner depths of the biofilm. Declining oxygen concentration is the result of actively respiring bacteria at the outer layers of the biofilm rather than limited diffusion. Similarly, a decline in nutrient concentration is observed with increasing biofilm depth because of consumption by microbes closer to the nutrient source. The opposite situation arises for waste product buildup where cells in the biofilm interior have increased exposure compared with those in the outermost layers. In nature, biofilms are rarely composed of a single species; instead, a consortium of bacteria is typically present. In this scenario, consumption of metabolic waste products may result if they can serve as a nutrient source for another organism. Therefore, as cells respond to the prevailing chemical conditions through physiological adaptation, distinct microniches are established throughout the biofilm. In addition to the chemical and physiological heterogeneity discussed above, researchers have discovered that when bacteria grow as a biofilm, genetic variants arise at a high frequency. Biofilms produced by a wide range of Gram-negative and Grampositive bacteria have been found to contain variant subpopulations. These variants are usually identified based on a change in colony morphology. In P. aeruginosa, for example, small rough variants have been isolated. This phenotype emerges because of overexpression of two gene clusters, called psl and pel, which are involved in polysaccharide synthesis. The presence of natural variants within the biofilm is thought to provide a diverse population better able to cope with antimicrobial assault and fluctuating environmental conditions. Genetic variation, therefore, is another factor contributing to biofilm heterogeneity.

1.39.4

Stages of Biofilm Development

Although once believed to be a passive process, scientists now understand that elaboration of complex biofilm communities proceeds through multiple steps (Figure 5). Even more remarkable is the fact that the various stages of biofilm development are conserved across a wide range of prokaryotes. As an overview of the process, biofilm development begins with bacterial attachment to a substratum. From here, bacteria divide and organize themselves into microcolonies. These microcolonies enlarge and become

1

2

3

4

5

Figure 5 The stages of biofilm development. Biofilm formation starts with attachment of bacteria to a surface (panel 1). Initially, cells undergo a reversible attachment phase during which they may leave the substratum. This is followed by irreversible attachment at which point cells become permanently associated with the surface. During the next stage, cells form clusters or microcolonies (panel 2). As the biofilm matures, its residents begin to produce an adhesive matrix composed of exopolysaccharide, protein, and DNA that enable cells to stick to one another to form a multilayer biofilm (panels 3 and 4). At various points, cells may leave the biofilm to resume the planktonic mode of growth, a process known as dispersion or detachment (panel 5). From Monroe D (2007). Looking for Chinks in the armor of bacterial biofilms. PLoS Biology 5(11): e307. https://doi.org/10. 1371/journal.pbio.0050307.

534

Biofilms

encased in an extracellular matrix, forming a structure generally referred to as a macrocolony or mature biofilm. At any time, cells can break free from the biofilm and resume the planktonic mode of growth, thus completing the developmental cycle. We will now examine each of these steps in more detail.

1.39.4.1

Step 1: Initial Attachment

A number of factors can influence the attachment process; these include environment-associated and bacteria-mediated components.

1.39.4.1.1

Physical and Chemical Factors

Almost immediately after placing an object like a glass slide into a liquid environment, proteins and polysaccharide molecules become adsorbed to the surface forming what is known as a ‘conditioning film‘. It is to this film that microbes first attach. Although conditioning films develop on both hydrophobic (e.g., plastic) and hydrophilic surfaces (e.g., glass), microbes preferentially adhere to hydrophobic materials. Microbes also adhere to rough surfaces more readily than smooth. In terms of the chemical nature of the bulk fluid, factors such as pH, ionic strength, and cation concentration have all been found to influence surface attachment.

1.39.4.1.2

Bacterial Factors

Temporal analysis of biofilm formation reveals that attachment to a surface occurs as a two-step process.8 Initially, cells undergo a reversible attachment stage and as the name implies, this can be transient with a number of cells leaving the surface to resume a planktonic lifestyle. After a period of time, a more stable interaction is established and the remaining population becomes irreversibly attached. Taking a closer look at these processes, when bacteria in the bulk fluid are in close proximity to a surface, both positive and negative forces of attraction come into play. Cells are positively attracted to the surface via van der Waals forces. Cells also experience a repulsive force resulting from the net negative charge borne by the bacteria as well as most surfaces. One important feature that increases the chances of bacterial cells making physical contact with a surface is motility. Appendages such as flagella and retractable pili can facilitate transient attachment in different ways. Flagellar motility, for example, increases the chances of a bacterium moving in close proximity to a substratum. Then, because flagella and pili are long appendages, they enable the bacteria to overcome surface-associated repulsive forces. These structures may also play a role in tethering the bacterium to the substratum. Some stabilizing factors that are involved in the transition from reversible to irreversible attachment have been identified. In P. fluorescens, for example, a genetic locus called lap (for large adhesion protein) is associated with the reversible-to-irreversible transition. This locus codes for an adhesion system consisting of the large adhesion protein (LapA) that is secreted onto the surface by an ATP-binding cassette transporter (LapBCE). Flow cell experiments demonstrated that lap mutants are unable to transition from the reversible to irreversible stage of attachment. Genes encoding the lap system have been identified in P. putida but not P. aeruginosa. Another protein, called SadB, has been implicated in the ability of P. aeruginosa to progress from transient to permanent attachment.

1.39.4.2

Step 2: Bacterial Migration and Microcolony Formation

The next stage after permanent attachment is microcolony formation. There is evidence to support a role for two types of surface translocation, known as twitching and swarming, in the formation of microcolonies. Twitching motility is mediated by the extension, attachment, and retraction of type IV pili filaments. Analysis of biofilms formed by P. aeruginosa and P. putida mutants lacking type IV pili reveals that development is arrested at the monolayer stage with no evidence of cell clustering (microcolony formation). Thus, it has been proposed that bacteria move along the surface via type-IV-mediated twitching motility resulting in the formation of microcolonies. Swarming motility differs from twitching in that it is dependent on a functional flagellum, biosurfactant production, and in some instances type IV pili. Studies have shown that under conditions favoring swarming motility, P. aeruginosa cells continue to migrate across the surface forming a uniform mat. Conversely, conditions that limit swarming result in a biofilm punctuated by microcolonies.

1.39.4.3

Step 3: Mature Biofilm Formation

As the biofilm matures, its residents begin to produce an adhesive matrix that enables cells to stick to one another to form a multilayer biofilm. The matrix, or extrapolymeric substance (EPS), is primarily composed of exopolysaccharide, protein, and DNA. In addition to EPS production, a second transition occurs during biofilm maturation, namely downregulation of flagella synthesis. Consequently, there is an inverse relationship between production of the EPS and flagella. This finding is supported by transcriptomic studies showing that genes involved in synthesis of the biofilm matrix and flagella are inversely expressed. The molecular mechanism behind this phenomenon has been elucidated in P. aeruginosa. Inverse regulation of matrix and flagellar components is mediated by the alternative sigma factor AlgT. AlgT functions to activate genes involved in synthesis of the biofilm matrix, while at the same time it represses flagellar gene expression through an indirect mechanism. Carbon source can have a profound effect on the overall architecture of the mature biofilm. In flow cells irrigated with glucose minimal media, P. aeruginosa forms biofilms consisting of mushroom-like structures with intervening water channels. Formation of these mushrooms has been proposed to follow a sequential process whereby a nonmotile subpopulation of cells develops into stalks. A motile group of cells then migrates onto the stalks forming the cap. However, in citrate media, a very different biofilm

Biofilms

535

develops. Under these conditions, flat undifferentiated biofilms are observed and this is thought to result from the extensive twitching motility that occurs in the presence of citrate. Migration of cells over the entire substratum ultimately leads to the development of a uniform mat-like structure. Interestingly, the effect of a given carbon source on biofilm structure is variable between even closely related organisms. For example, in citrate minimal media, P. putida develops heterogeneous biofilms.

1.39.4.4

Step 4: Biofilm Dispersion

The last step in biofilm development is dispersion. A multilayer biofilm can be viewed as a collection of heterogeneous microenvironments. As such, bacterial residents within the biofilm experience a myriad of different conditions depending on their location. Cells in the interior have reduced access to nutrients and possibly increased exposure to toxic byproducts compared with those positioned at the surface. As a result, communal living may become so inhospitable that resuming the planktonic lifestyle becomes the only viable option. There is a great deal of interest in understanding the signals and molecular mechanisms underlying biofilm dispersal as it may reveal a means of eradicating biofilms. In general, there are two ways in which biofilms disperse. The first, passive dispersion through the action of shear forces will not be discussed. The second mechanism involves active detachment, which is the focus of the following section. If we look at environmental conditions that can trigger a dispersal response, changes in oxygen tension as well as nutrient availability can stimulate this event. With respect to oxygen availability, as the biofilm thickness increases, the cells in the innermost surface experience decreased levels of oxygen. The depletion of oxygen and the buildup of reactive intermediates resulting from anaerobic metabolism have been shown to activate the dispersal response. In terms of nutrient availability, nutrient upshifts and downshifts can both bring about a dispersion event. For example, increasing carbon substrate availability results in dissolution of preformed P. aeruginosa biofilms. Increased expression of flagella genes was observed in the dispersed P. aeruginosa cells. Thus, it appears that for P. aeruginosa the decision to adopt a sessile lifestyle is accompanied by decreased flagellar synthesis, which is then reversed when it is time to leave the biofilm. Examination of P. aeruginosa biofilms formed in flow cells has revealed the presence of hollow cores in the mushroom-like structures. The hollowing is due to bacteria evacuating the macrocolony via flagellar motility, while nonmotile cells remain as part of the exterior wall. This phenomenon, termed ‘seeding dispersal‘, occurs only after the stalks have reached a diameter of 80 mm or greater. It would appear that when the mushrooms have reached a certain size, conditions within the interior become unfavorable, setting off an evacuation event. In P. putida, both carbon starvation and nutrient upshift can lead to biofilm dispersion. Microscopic analysis showed that the P. putida cells are swimming rapidly during dispersion. As nonmotile flagellar mutants also undergo this process, it is not mediated by flagellar motility. Instead, it appears that breaking down the cohesive matrix of the biofilm is essential for dispersion of this organism. Detachment can also be brought about through the action of a group of compounds known as surfactants. P. aeruginosa produces rhamnolipid surfactants that play a number of roles in the biofilm developmental cycle. For example, these molecules are essential for maintaining the open channel structures found within macrocolonies. Rhamnolipids can also mediate biofilm dispersal. In P. aeruginosa cells overexpressing rhamnolipids, a hyper-detachment phenotype is observed. Early induction of rhamnolipids leads to central hollowing in the stalk-like structures after only 3 days of biofilm maturation, compared with 10–12 days for the wild-type strain. P. putida produces a group of surfactant molecules called putisolvins. Mutants no longer producing putisolvins generate much thicker biofilms than the wild type. Interestingly, purified putisolvin can break down established biofilms from a number of organisms, suggesting that there may be an application for these molecules in biofilm eradication.

1.39.5

Regulation of Biofilm Development

High-resolution microscopic imaging of biofilms has revealed a complex architecture containing towers or stalk-like structures surrounded by interstitial void spaces. With this in mind, it is not surprising that erecting these elaborate structures is under regulatory control. Regulatory mechanisms affecting these processes include quorum sensing (QS), second messenger signaling, and twocomponent signal transduction.

1.39.5.1

Quorum Sensing

QS is a means by which bacteria are able to regulate gene expression according to population density. QS circuits rely on selfgenerated signaling molecules, called autoinducers (AIs), which accumulate in the extracellular environment. The premise of QS or cell-to-cell communication is based on the fact that when a single bacterium releases AIs into the environment, their concentration is too low to be detected. However, when sufficient bacteria are present in a defined environment, the AI reaches a threshold level that enables binding to a cognate receptor. QS systems exist in both Gram-positive and Gram-negative bacteria with significant differences between the two. For example, in Gram-positive QS systems, modified oligopeptides are used as the AI signals that are detected by a two-component signal transduction system. After a threshold level of the signal has accumulated, the AI molecule binds extracellularly to a sensor kinase, which undergoes autophosphorylation and phosphotransfer to the cognate response regulator. The phosphorylated response regulator can now activate or repress target genes. In Gram-negative QS, the AIs are usually Nacyl-homoserinelactones (AHLs) synthesized by LuxI-type proteins. Most AHL signals can freely diffuse across bacterial membranes.

536

Biofilms

When a threshold level of AI is reached, it binds to and activates a LuxR-type regulatory protein that can then activate or repress gene transcription. Studies have shown that biofilm formation can be affected either negatively or positively by QS; examples of each are described below.

1.39.5.1.1

Staphylococcus aureus

In the Gram-positive bacterium S. aureus, biofilm formation is negatively regulated by QS.7 The QS signal, called AIP, is encoded by the agrD gene, which after synthesis gets actively secreted by the AgrB transporter. Once the extracellular concentration of the signal reaches a threshold level in a defined environment, it binds to the sensor kinase AgrC resulting in autophosphorylation and phosphotransfer to the response regulator AgrA. AgrA activation leads to increased expression of a g-hemolysin peptide, which is believed to inhibit biofilm formation through its biosurfactant properties, as well as increased production of two proteases that mediate biofilm detachment. Mutants that are no longer able to produce AIP generate much thicker biofilms, which is consistent with the idea that in S. aureus, QS negatively impacts biofilm development.

1.39.5.1.2

Pseudomonas aeruginosa

In the Gram-negative bacterium P. aeruginosa, biofilm formation is positively controlled by the QS circuitry.2 P. aeruginosa has two AHL-based QS systems, Las and Rhl. The Las circuit consists of a transcriptional activator, LasR, and its cognate AHL signal, N-(3oxododecanoyl) homoserine lactone (3O-C12-HSL), synthesized by the AHL synthase LasI. Similarly, the Rhl system is comprised of the transcriptional activator RhlR together with its cognate AHL, N-butyryl homoserine lactone (C4-HSL), synthesized by RhlI. The first link between QS and biofilm formation was made when analysis of a lasI (AI deficient) mutant revealed flat, uniform biofilms that developed in a flow-chamber system. This was in stark contrast to the parental biofilm, where mushroom- and stalk-like structures were observed. Furthermore, upon exposure to the detergent sodium dodecyl sulfate (SDS), the lasI biofilm quickly dispersed, whereas the wild-type biofilm remained intact. Therefore, in addition to being structurally altered, the mutant biofilm was functionally impaired in its ability to protect against biocidal agents. It has been shown that up to 11% of the P. aeruginosa genome is under QS control. Thus, understanding exactly which QS-controlled genes are involved in P. aeruginosa biofilm development becomes a challenge. In spite of this, several connections between QS and biofilm formation have been made, which are discussed below. 1.39.5.1.2.1 QS and the Biofilm Matrix The cohesive matrix surrounding cells in a P. aeruginosa biofilm is typically composed of polysaccharides, proteins, and DNA. The pel operon, one of five exopolysaccharide gene clusters in the P. aeruginosa genome, encodes the biosynthetic machinery for a glucoserich exopolysaccharide. Evidence to suggest that the pel operon is under QS control came from analysis of QS mutants that were greatly diminished in their ability to produce both air–liquid biofilm (pellicle) and surface-associated biofilms. The molecular basis for this was revealed with the finding that the Las QS system and, to a lesser extent, the Rhl QS system are involved in transcription of the pel biosynthetic operon. A second component of the EPS, extracellular DNA, is also under QS control. Early on in P. aeruginosa biofilm development, extracellular DNA is an important cohesive component of the matrix. Studies have shown that extracellular DNA is liberated through the lysis of a small population of cells by two different pathways: A QS-dependent pathway generates the majority of DNA, whereas a second QS-independent pathway liberates small amounts of DNA. Analysis of biofilms formed by a lasIrhlI mutant revealed that there was less extracellular DNA present in the biofilm matrix. Furthermore, the mutant biofilm was much more sensitive to SDS treatment. We can see that QS regulates production of two key components of the EPS matrix, namely Pel polysaccharide and DNA. 1.39.5.1.2.2 QS Control of Rhamnolipids As mentioned earlier, P. aeruginosa produces rhamnolipids that are amphipathic glycolipids with biosurfactant properties. Rhamnolipids are the product of rhlAB and rhlC, and all three of these genes are under QS control. In fact, the Rhl QS system was so named because of its control over rhamnolipid production. These molecules keep water channels open and play a role in biofilm dispersal. In addition, it was discovered that when grown in flow-cell chambers under conditions that promote mushroom formation, the mushroom caps formed by rhlA mutants are much smaller than the wild type. Thus, it appears that rhamnolipid synthesis facilitates mushroom cap formation in P. aeruginosa biofilms. Because rhamnolipids are under QS control, this is another way in which intercellular communication affects formation of these adherent communities. 1.39.5.1.2.3 QS Inhibition as a Means of Attenuating Virulence and Biofilm Formation As the number of infections caused by antibiotic-resistant bacteria continues to rise, so does the demand for new ways of combating microbial diseases. One strategy that has received a great deal of attention is attenuation of bacterial virulence through the use of antipathogenic compounds. Antipathogenic drugs are designed to disarm the regulatory circuitry controlling expression of virulence factors. In the presence of these compounds, pathogenic bacteria are rendered avirulent allowing them to be cleared by the host immune system. The advantage of this approach over growth inhibitory compounds such as antibiotics is avoidance of the selective pressure that leads to bacterial resistance. Expression of virulence factors by pathogenic bacteria is typically under QS regulation and so QS has emerged as an ideal target for antipathogenic compounds. As discussed earlier, for many bacteria QS controls formation of biofilms and these structures contribute to pathogenicity. Therefore, it seems reasonable that QS inhibition may result in attenuation of bacterial pathogens. For Gram-negative QS, several strategies for shutting down the regulatory circuits have been recognized. The most thoroughly

Biofilms

537

investigated approach, which appears to hold promise, is the use of small molecules to block reception of the AHL signal. There are two main ways in which AHL antagonists can inhibit activation of the LuxR-type protein. First, molecules that are structurally similar to the cognate AHL but incapable of activating the LuxR protein may occupy the AHL-binding site. Second, noncompetitive antagonists may bind alternative sites causing conformational changes in the receptor protein that prevent AHL binding. A number of in vitro studies have demonstrated that AHL antagonists can effectively inhibit QS in different bacteria. Interestingly, many higher organisms naturally interfere with QS through liberating small molecules that act as AHL antagonists. The Australian red macroalga (seaweed) Delisea pulchra, for example, produces a series of over 30 halogenated furanone compounds capable of interfering with bacterial signaling. These furanone molecules inhibit QS through occupying the AHL-binding site on the LuxR receptor. Natural furanones impact P. aeruginosa QS only nominally; consequently, synthetic derivatives were generated and two compounds (C30 and C56) were identified as being potent inhibitors of P. aeruginosa QS.6 These molecules were found to suppress virulence factor production in planktonic cultures. In addition, furanone exposure resulted in biofilms with a flat, undifferentiated appearance that exhibited increased sensitivity to the antibiotic tobramycin and the detergent SDS. Using microarrays, it was discovered that these furanones specifically repress QS-controlled genes in P. aeruginosa. Transcription of the lasR/lasI and rhlR/rhlI regulatory genes was not affected by furanone treatment, which supports the idea that furanones affect these QS regulators at the posttranscriptional level. When the effect of these molecules was tested in a pulmonary mouse model of infection, treatment with C30 and C56 increased the survival time for mice with lethal P. aeruginosa infections. Moreover, these molecules reduced the number of bacteria found in the lungs and the overall lung damage. Unfortunately, the halogenated furanones used in these studies are too toxic for human use. However, their ability to control P. aeruginosa in animal infection models is significant because it establishes QS as a viable target for managing undesirable bacterial activities, such as biofilm formation associated with persistent infections.6

1.39.5.2

Bis-(30 ,50 )-Cyclic-Dimeric-Guanosine Monophosphate

In the last two decades, bis-(30 ,50 )-cyclic-dimeric-guanosine monophosphate (c-di-GMP) has been recognized as a ubiquitous second messenger that regulates a number of bacterial functions. In particular, this cyclic dinucleotide is the primary regulator controlling the transition between the single-cell planktonic and the surface-attached biofilm state in Gram-negative bacteria. A rise in c-di-GMP levels leads to increased production of components required for biofilm establishment, whereas decreased c-di-GMP typically leads to enhanced motility factor expression. The level of c-di-GMP is controlled by two groups of enzymes with opposing functions. Diguanylate cyclases (DGCs) synthesize c-di-GMP; phosphodiesterases (PDEs) catalyze its breakdown. The DGC and PDE enzymes have been found in most of the bacterial genomes sequenced to date and are easily recognized by their conserved GGDEF (Gly–Gly– Asp–Glu–Phe) and EAL (Glu–Ala–Leu) domains, respectively. Many DGC and PDE proteins have additional domains that are believed to sense an array of environmental signals. Direct binding of effector molecules to the sensory input domains affects the enzymatic activity of the DGC/PDE proteins. In this manner, a diverse range of environmental signals can be transduced through c-di-GMP signaling, with the end result being either a planktonic or sessile existence. Presently, there is very little known about the mechanisms whereby c-di-GMP regulates cellular functions such as biofilm formation. That being said, a few targets of c-diGMP have been identified, some of which are associated with polysaccharide biosynthetic machinery. For example, in the first bacterium in which c-di-GMP was discovered, Gluconacetobacter xylinus, c-di-GMP was found to act as an allosteric activator of BcsA. BcsA forms part of the cellulose synthase enzyme required for cellulose synthesis. BcsA contains a C-terminal PilZ domain as the c-di-GMPbinding site. PilZ domains are one of the well-established downstream targets of c-di-GMP, and direct binding of c-di-GMP by PilZ domains has been demonstrated experimentally. In P. aeruginosa, the PilZ domain protein Alg44 is involved in production of the exopolysaccharide alginate. Experimental evidence indicates that alginate synthesis is dependent on c-di-GMP binding to Alg44. PelD is a component of the P. aeruginosa pel operon involved in pellicle formation. Like the other two examples, PelD directly binds c-di-GMP; however, it does not have a PilZ domain. Mutants of PelD that are unable to bind c-di-GMP are deficient in pellicle formation, indicating that interaction between PelD and c-di-GMP is a requirement for PEL polysaccharide biosynthesis. Taken together, cdi-GMP binding to cellulose (BcsA), alginate (Alg44), and PEL (PelD) biosynthetic components suggests that interaction of this nucleotide with the exopolysaccharide machinery is a common mechanism underlying c-di-GMP control of biofilm formation.7

1.39.5.3

Two-Component Regulatory Systems

Two-component regulatory systems are a common mechanism whereby bacteria can sense a range of stimuli and make an appropriate adaptive response. In their most basic form, two-component systems are comprised of two proteins: an inner membranespanning histidine kinase and a cytoplasmic response regulator. The histidine kinase either directly or indirectly detects a signal, via its N-terminal input domain, which stimulates autophosphorylation at a conserved histidine residue. In many cases, the nature of the signaling molecule is unknown. The signal is then relayed to the response regulator through phosphotransfer to a conserved aspartate. Phosphorylation of the response regulator alters its affinity for certain DNA sequences, which usually elicits a change in gene expression and an appropriate physiological response. A number of variations of this classical two-component theme exist. For example, there are hybrid sensor kinase-response regulators in which a single protein undergoes autophosphorylation and phosphotransfer to a conserved aspartate residue located in the C-terminus. Moreover, unorthodox histidine kinases exist that contain an additional histidine phosphotransfer domain fused downstream of the receiver domain. These sensor kinases are believed to undergo a complex phosphorelay cascade. Considering that these two-component systems are an integral part of bacterial adaptation, it is not surprising that they play a role in biofilm formation.7

538

Biofilms

One example of a two-component system that regulates production of EPS and biofilm formation is the GacS/GacA system. GacS is classified as an unorthodox histidine kinase with two histidine kinase regions flanking a single phosphotransfer domain. Once the phosphoryl group is transferred from GacS to its cognate response regulator GacA, GacA activates transcription of one or more regulatory RNAs. In P. aeruginosa, where a link between Gac and biofilm formation has been firmly established, this regulatory system controls expression of two small RNAs, called RsmY and RsmZ. These regulatory RNAs function to antagonize an mRNA-binding protein called RsmA. RsmY and Z are able to titrate RsmA away from target mRNAs resulting in derepression of a number of extracellular products under Gac–Rsm control. In P. aeruginosa, there appears to be an inverse relationship between expression of genes encoding the TTSS and those involved in biofilm formation. Regulation of this phenotypic shift involves the Gac–Rsm system, which acts in concert with two other histidine kinases, RetS and LadS. RetS is a hybrid sensor kinase with one kinase domain followed by two response regulator receiver domains. Analysis of a retS mutant revealed that genes required for acute infection (TTSS, lipase, exotoxin A, and type IV pili genes) were downregulated. Conversely, the psl and pel EPS operons involved in biofilm formation were expressed at a higher level. It was concluded that RetS functions to activate genes required for acute infection while repressing genes associated with chronic infection. Like RetS, LadS is a hybrid sensor kinase, but it has only a single response regulator domain. A ladS mutant produced very little biofilm, which was attributed to the fact that LadS positively regulates the pel EPS operon. The ladS mutant also exhibited increased TTSS gene expression and cytotoxicity compared with the wild type. Thus, it appears that LadS works in opposition to RetS, activating expression of genes involved in biofilm formation and repressing those required for acute infection. The link to the Gac–Rsm system comes from the discovery that a retS mutant exhibits increased rsmZ transcription while expression of rsmZ is dramatically decreased in the ladS mutant background. Collectively, these findings have led to the proposal that LadS, RetS, and GacS coordinately regulate expression of genes involved in acute and chronic infections. GacS and LadS increase rsmZ transcription, which results in biofilm formation and the chronic infection state. RetS, on the other hand, represses rsmZ transcription leading to an acute infection. Many of the details regarding how these regulators control the switch from acute to chronic infections still need to be established. Nonetheless, the idea that two-component regulatory systems such as these could be targets for therapeutic intervention of bacterial infections is intriguing.

1.39.6

Biofilm Infections

The Centers for Disease Control estimates that over 65% of microbial infections in developed countries are caused by biofilms. Biofilm infections are on the rise largely due to the increased use of implanted medical devices. Once inside the body, devices such as catheters, pacemakers, and prosthetic heart valves provide an ideal surface on which biofilms can form. Non-device-related biofilm infections include tooth decay and gingivitis, middle ear infections, prostatitis, and kidney disease. Perhaps the beststudied disease in which biofilms are a major contributing factor is chronic lung infections caused by P. aeruginosa in patients suffering from CF. CF is an inherited disease caused by mutations in the CF transmembrane conductance regulator gene, which leads to altered electrolyte secretion. CF patients produce a very viscous respiratory mucus that impairs mucociliary clearance leaving them vulnerable to lung infections. Early on, the lungs of CF patients are colonized by S. aureus and Haemophilus influenza, but these infections are typically controlled with antibiotics. The next bacterium to colonize the CF lung is P. aeruginosa. During chronic infection, P. aeruginosa establishes itself as an alginate-encased biofilm, at which point the bacteria are said to be in a mucoid state. Mucoid P. aeruginosa infections are notoriously recalcitrant to immune system clearance and antibiotic killing and so once established, P. aeruginosa cannot be eradicated from the lung.5 Progressive pulmonary deterioration is the leading cause of mortality in CF patients.

1.39.7

Pathogenicity and Antibiotic Resistance of Biofilms

The contribution of biofilms to pathogenicity is not unique to P. aeruginosa; in almost all cases, biofilm formation enables microbes to survive within a host.3,5 The biofilm EPS forms a protective barrier that shields its inhabitants from key immune components including antibodies and white blood cells. In addition, when bacteria are growing as a biofilm, they can resist extraordinarily high levels of antibiotics compared with free-floating cells. In some instances, biofilms have been reported to survive exposure to 1000 times the antibiotic dose that is able to sterilize planktonic cultures. Three general mechanisms have been proposed to explain biofilm resistance including restricted penetration of the antimicrobial, decreased growth rate, and the presence of an antibiotic-tolerant subpopulation, sometimes called ‘persisters‘. In some instances, genetic determinants of biofilm resistance have been identified. Each of these mechanisms is discussed in more detail in the following sections.

1.39.7.1

Restricted Penetration

In order for an antimicrobial agent to inactivate the constituents within the biofilm, it must be able to diffuse through the matrix. The EPS matrix represents a diffusional barrier; consequently, if the antimicrobial agent can be broken down, a synergistic effect between restricted penetration and degradation occurs. This situation has observed with b-lactam antibiotic inactivation by b-lactamases and hydrogen peroxide breakdown by catalases. For positively charged substances such as aminoglycoside antibiotics and some metal ions, the negatively charged EPS is able to bind these compounds and prevent their entry. It is important to note that

Biofilms

539

there are a number of examples of smaller molecules, such as fluoroquinolone antibiotics, which rapidly diffuse through the biofilm matrix. Thus, restricted penetration due to the EPS cannot explain resistance to all biocidal agents. In addition to the biofilm matrix, intracellular components can prevent antibiotics from reaching their targets. A screen for genetic determinants of P. aeruginosa biofilm resistance revealed a gene, called ndvB, that is required for the synthesis of periplasmic glucans.10 Biofilms formed by the ndvB mutant were significantly more sensitive to a number of antibiotics including tobramycin, gentamicin, ciprofloxacin, chloramphenicol, and ofloxacin. This was found to be a biofilm-specific trait because the mutant and wild-type planktonic cultures exhibited equivalent antibiotic sensitivity. With this in mind, it is not surprising that ndvB was found to be preferentially expressed in biofilm cells. The discovery that periplasmic glucans physically interact with antibiotics such as tobramycin has led to the proposal that the role of these glucans is to bind and sequester antibiotics in biofilm-grown cells. Sequestration prevents the passage of antibiotics through the periplasm to their cytoplasmic site of action. This necessitates the use of high levels of antibiotics so that a sufficient amount can bypass the perplasmic glucans and reach the cytoplasm.

1.39.7.2

Decreased Growth Rate

There is a great deal of heterogeneity in the metabolic activities of cells within a biofilm. Cellular activity ranges from bacteria that are metabolically active to ones that are dormant. Certain antibiotics, such as penicillin, are only effective against cells that are dividing. Others are able to kill nongrowing cells, but these antibiotics presumably require a low level of metabolic activity because their effect is through interruption of normal cellular processes. The presence of a subpopulation of cells in a metabolically quiescent state, therefore, is arguably a major contributor to biofilm resistance.

1.39.7.3

Tolerant Subpopulation of Persister Cells

Antibiotic-tolerant bacteria, also known as persisters, are present in both planktonic and biofilm populations.9 These cells differ from antibiotic-resistant organisms in that they are not able to grow in the presence of high concentrations of antibiotics, as is the case with resistant bacteria. In fact, quite the opposite is true; persister cells are believed to be in a state of dormancy. By shutting down their drug targets, they become impervious to the deadly effects of antibiotics. It has been observed that the number of persisters increases with increasing population density and reaches a maximum of about 1% in stationary phase planktonic cultures. A substantial number of persisters exist in biofilm populations as well. Gene expression profiling has revealed that genes involved in cellular energetics and flagellar synthesis are downregulated in persisters, which is consistent with the notion that these cells are in a state of dormancy.9 Genes encoding toxin/antitoxin (TA) modules show elevated expression in persister cells. Two of the toxin components, RelE and MazF, were found to inhibit translation resulting in stasis, the effect of which can be reversed by expression of the cognate antitoxin (RelB and MazE, respectively). Importantly, overexpression of the toxins RelE, MazF, and HipA resulted in cells that were tolerant to a number of antibiotics. Therefore, it has been proposed that enhanced expression of toxins that hinder essential cellular functions contributes to multidrug tolerance in bacteria. The fact that TA modules have been discovered in the chromosomes of all free-living bacteria adds additional support for their role in persister formation.9

1.39.8

Antibiotics Act as Signals That Stimulate Biofilm Formation

In addition to being able to resist the effects of antibiotics, another link between biofilms and antimicrobials exists. Growth in the presence of subinhibitory concentrations of certain antibiotics can actually stimulate biofilm formation. At subinhibitory levels, aminoglycosides such as tobramycin are able to induce biofilm formation in P. aeruginosa and E. coli. Similarly, tetracycline and norfloxacin augmented P. aeruginosa biofilm development. This phenomenon is not related to motility because exposure to these antibiotics affected P. aeruginosa motility in different ways. For example, swimming and swarming motility were found to increase in the presence of tobramycin and decrease when grown with ciprofloxacin. On the other hand, both forms of motility were unaffected by exposure to subinhibitory concentrations of tetracycline. A genetic determinant underlying tobramycin-mediated biofilm induction in P. aeruginosa has been discovered. The gene, designated arr (aminoglycoside response regulator), is essential for enhanced biofilm formation in the presence of this antibiotic. Arr is predicted to be a PDE whose activity is increased in the presence of tobramycin, resulting in c-di-GMP inactivation and elevated biofilm development. Furthermore, Arr plays a role in biofilm resistance. Biofilms formed by the arr-deficient strain were 100-fold more susceptible to killing by tobramycin, whereas the mutation had no effect on planktonic sensitivity. Thus, it appears that arr is a genetic determinant of P. aeruginosa biofilm resistance, much like the ndvB allele, discussed earlier. Taken together, these findings indicate that bacteria such as E. coli and P. aeruginosa can respond to subinhibitory concentrations of drugs by producing antibioticresistant biofilms. The soil is rich with antibiotic-producing microbes and so it has long been assumed that the role of antibiotics is to inhibit competing microbes in natural settings. However, antibiotic concentrations in the soil are typically lower than that required for a killing effect. This has led to idea that these compounds have concentration-dependent roles. At higher concentrations, antibiotics function as weapons for destroying competing microbes, whereas at lower concentrations, they act as signaling molecules enabling communities to coordinate an adaptive response.

540

1.39.9

Biofilms

Concluding Remarks

The last two decades have seen a tremendous surge in biofilm research. The realization that over 99% of microbial biomass exists as a biofilm and that sessile cells differ dramatically from their planktonic counterparts means that the conventional approach of studying bacteria growing as liquid cultures is an inaccurate representation of microbial life in the real world. Also, fueling the intense interest in biofilms is the fact that these structures pose huge problems in industrial and medical settings. Billions of dollars are lost every year due to damaged equipment, energy losses, and medical infections, all caused by biofilms. For these reasons, it is hoped that elucidating the environmental cues and genetic elements involved in biofilm formation will provide a platform from which effective control strategies can be designed. Thanks to the availability of sophisticated three-dimensional imaging techniques and apparatus for growing model biofilms, our understanding of biofilms has come a long way in recent years. Rather than being a passive process with cells simply piling up on one another, the earliest SCLM images revealed an astonishing level of complexity. Mushroom-like structures with intervening water channels believed to facilitate nutrient and waste exchange were observed. In light of their structural complexity, it is not surprising that building these communities requires cooperation and communication between biofilm residents. It has also been recognized that the physiology of adherent cells is vastly different from that of planktonic cells, particularly, their ability to tolerate exposure to antibiotics, biocides, and immune factors. To better understand these differences, several groups have undertaken global expression analysis to try to identify genes and gene products that are up- or downregulated in sessile cells. However, comparisons of transcriptomic and proteomic findings have failed to reveal a defined genetic program of biofilm development.8 Rather, it appears that multiple pathways for biofilm establishment exist, which enable bacteria to optimize surface colonization in response to the prevailing conditions. Undoubtedly, identification of biofilm-specific determinants/pathways is complicated by the degree of heterogeneity found within these communities. Be that as it may, analysis of a wide range of prokaryotes has shown that biofilm formation proceeds through a series of highly conserved developmental steps. These steps include attachment, formation of microcolonies, development of the mature biofilm, and dispersal. The bacterial determinants involved in each of these stages, however, are not necessarily conserved. Even within a given strain, depending on the environmental conditions, there appears to be quite a bit of flexibility with respect to the bacterial component involved. Given the proclivity of bacteria to adhere to surfaces and the selective advantages conferred by the biofilm lifestyle, it is not surprising that bacteria have evolved numerous pathways for establishing biofilms.

References 1. Anderson, J. B.; Sternberg, C.; Poulsen, L. K.; et al. New Unstable Variants of Green Fluorescent Protein for Studies of Transient Gene Expression in Bacteria. Appl. Environ. Microbiol. 1998, 64, 2240–2246. 2. de Kievit, T. R. Quorum Sensing in Pseudomonas aeruginosa Biofilms. Environ. Microbiol. 2009, 11, 279–288. 3. Donlan, R. M.; Costerton, J. W. Biofilms: Survival Mechanisms of Clinically Relevant Microorganisms. Clin. Microbiol. Rev. 2002, 15, 167–193. 4. Doyle, R. J., Ed.; Methods in Enzymology; BiofilmsVol. 310; Academic Press: San Diego, CA, 1999. 5. Hall-Stoodley, L.; Costerton, J. W.; Stoodley, P. Bacterial Biofilms: From the Natural Environment to Infectious Diseases. Nat. Rev. Microbiol. 2004, 2, 95–108. 6. Hentzer, M.; Givskov, M. Pharmacological Inhibition of Quorum Sensing for the Treatment of Chronic Bacterial Infections. J. Clin. Invest. 2003, 112, 1300–1307. 7. Karatan, E.; Watnick, P. Signals, Regulatory Networks, and Materials that Build and Break Bacterial Biofilms. Microbiol. Mol. Biol. Rev. 2009, 73, 310–347. 8. Kjelleberg, S., Givskov, M., Eds.; The Biofilm Mode of Life: Mechanisms and Adaptations, Horizon Bioscience: Norfolk, 2007. 9. Lewis, K. Multidrug Tolerance of Biofilms and Persister Cells. In Current Topics in Microbiology and Immunology; Romeo, T., Ed.; Bacterial Biofilms, Vol. 322; Springer: Berlin, Heidelberg, 2008; pp 107–131. 10. Mah, T. F.; Pitts, B.; Pellock, B.; et al. A Genetic Basis for Pseudomonas aeruginosa Biofilm Antibiotic Resistance. Nature 2003, 426, 306–310. 11. Stewart, P. S.; Franklin, M. J. Physiological Heterogeneity in Biofilms. Nat. Rev. Microbiol. 2008, 6, 199–210.

1.40

Flow Cytometry

B-F Alfonso and M Al-Rubeai, University College Dublin, Dublin, Ireland © 2011 Elsevier B.V. All rights reserved. This is an update of B.-F. Alfonso, M. Al-Rubeai, 1.42 - Flow Cytometry, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 559-578.

1.40.1 Introduction 1.40.2 Principles and Instrumentation 1.40.3 Data Representation 1.40.4 Common Applications 1.40.4.1 Fluorescent Proteins 1.40.4.2 Cell Proliferation 1.40.4.3 Cell Physiology 1.40.4.3.1 Intracellular pH 1.40.4.3.2 Membrane Potential 1.40.4.3.3 Cytosolic Ca2þ Concentrations 1.40.4.3.4 Ros Generation 1.40.4.4 Membrane Integrity, Apoptosis, and Necrosis 1.40.4.5 Cell Cycle, DNA, and RNA Analysis 1.40.4.6 Immunophenotyping 1.40.4.7 Cell Sorting 1.40.4.8 Imaging Flow Cytometer References Relevant Websites

541 542 544 550 550 550 550 551 551 551 551 552 554 556 557 558 559 560

Glossary Flow cytometry Flow cytometry is a method for the qualitative and quantitative measurement of biological and physical properties of cells and other particles in suspension. A flow cytometer can provide information on the intrinsic and extrinsic characteristics of the analyzed cells including size, shape, density, DNA, RNA and protein content, internal or external receptors, membrane structure, apoptosis and necrosis, calcium flux, and intracellular pH. This article describes the principles of flow cytometry and explain how it works. It summaries some of the various applications and methods currently available for the analyses and sorting of cells in biology and biotechnology.

1.40.1

Introduction

Humans tend to qualify and quantify everything, establishing categories, groups, and subgroups to improve their understanding and control of the environment. This process has caused scientists to develop numerous systems of classification, which have generally started with macroscopic observations of entities and phenomena. These observations were continuously refined as technological progress allowed progressive microscopic observations; thus, microbiology developed its own system of classification based on increasingly precise observations of microorganisms’ morphological, functional, and physiological characteristics. Nowadays, scientists continue to focus their attention on developing more efficient and less time-consuming methods to investigate phenomena, which have led them to redesign some laboratory equipment and to introduce new laboratory technologies and instruments. Flow cytometry (FC), a qualitative and quantitative analysis technology, can characterize cell populations at a single-cell level. FC is a multiparameter technology that measures fluorescent signals from the scattered light produced by cells when they pass one by one through a light source. These signals are correlated to structural and/or functional cell parameters. Unlike other biochemical techniques that generally give the mean value of a large number of cells, FC allows the study of a variety of organisms or particles, whole cells, chromosomes, organelles, or protoplasts. These measures are obtained at a sample rate of thousands of particles per second, with a very low variation coefficient (lower than 7% in plants and even lower in animal cells). In this way, FC makes it possible to distinguish different subpopulations of cells – for example, it allows the analysis of cell cultures with asynchronous growth, where the different phases of the cell cycle are clearly distinguishable. FC allows scientists to carry out different studies on the cell cycle in connection with the effects of drugs and radiations, nuclear DNA amount, and ploidy determination and also allows the measurement of different cellular parameters (intracellular pH, Caþ2 concentrations, membrane potential, fluidity,

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00048-3

541

542

Flow Cytometry

etc.), the detection of a wide variety of antigens; FC even allows the physical sorting of particles, like organelles and chromosomes. This technique has been used to study a wide range of organisms, including plant cells, yeasts, and bacteria. Over the years, it has also become important in mammalian cell culture, mainly as a research tool. Nevertheless, its main usage and applications are related to clinical laboratories, which use FC for the detection of leukemia, lymphomas, etc., as it can discriminate different subtypes of lymphocytes expressing different antigens, by using fluorescent monoclonal antibodies linked to these antigens.1–3,5,8,10–12,14,15 Flow cytometers and probes, broadly available, allow the easy use of the fluorescence. Thus, it provides a sensitive, specific, quantitative, rapid (using microquantities of reagents and avoiding radioactive products), multiparametric (able to use several different fluorochromes at the same time), and vital analysis (in the case of a nondestructive staining).6 Some instruments, such as cell sorter, have facilitated the physical separation of cells according to the values of the measured parameters. Continuous improvements in lasers, computers, processors, software, etc., together with the increased availability of fluorescent dyes and antibodies have allowed the development of more powerful and efficient flow cytometers.11

1.40.2

Principles and Instrumentation

As its name implies, FC is a method for the qualitative and quantitative measurement of the biological and physical properties of cells and other particles suspended within a high-velocity fluid stream when these events pass in a single file through a laser beam. Many people use the term ‘FACS’ instead of ‘FC’ but these terms are not synonymous. FACS, a registered trademark of Becton Dickinson, refers to fluorescence-activated cell sorter, so that the term FACS should not be used to mean FC. A flow cytometer allows the rapid analysis of large number of cells. Instruments currently on the market can measure up to 100,000 events per second. These instruments are able to qualify and quantify biological and physiological characteristics of every one of these single events. Thus, the heterogeneity of populations can be revealed and different subsets of cells can be identified and quantified. One of the most advanced instruments available now in the market, the Beckman Coulter Atrios, can use seven lasers to analyze up to 48 different labels simultaneously. By following the necessary quality control procedures, one obtains data that are extremely reproducible and accurate. Cell sorters, a particular type of flow cytometers, can physically separate a selected cell population from the rest of the sample. These factors, along with the simplicity and speed of sample preparation, make this technology one of the most easy and accurate systems used for cell biology studies. Although some flow cytometers permit the analysis of tissues, even such small organisms as zebrafishes or Caenorhabditis elegans, these instruments are mostly suitable for the analysis of cells in suspension. This means that, before analysis, tissues and/or adherent cell lines must be disaggregated in order to obtain a single-cell suspension. The most common methods for doing so include enzymatic digestion, trypsinization, and/or the mechanical chopping and filtration of the tissues to remove clumps of cells that could clog the system. A flow cytometer will provide information about intrinsic and extrinsic characteristics of the analyzed cells. A cell’s intrinsic characteristics include its size, shape, density, granularity, or the presence of some pigments. Plants and some microorganisms contain chlorophylls and carotenoids that contain inherent fluorescence. All cells autofluorescence at higher or lower intensities, but this can be altered by some biological processes, such as aging. A cell’s extrinsic characteristics include those measurements accomplished by adding external fluorescent labels: DNA content, composition, or synthesis; internal or external receptors, such as antibodies; membrane structural modifications, apoptosis and/or necrosis processes; and physiological parameters such as calcium flux, changes of intracellular pH, enzymatic procedures, drug kinetics, etc. How does a flow cytometer work? A flow cytometer comprises a sample flow, a sheath fluid that envelops the sample flow, a light excitation source, a flow chamber where light excites labeled particles, a set of optical filters to collect particle-emitted fluorescence, an acquisition system for the light-generated signals (detectors), and a data-conversion system able to interpret the results.15 The process is as follows: cells and/or particles in suspension are sucked using a sample injection probe (SIP), and they are introduced into an area called the flow cell or flow chamber, which is the heart of the flow cytometer. Most systems use compressors to deliver the sample, although syringes and peristaltic pumps are also used. Typically, air pressure, supplied by a compressor, drives the sheath fluid (water or buffer) through the flow chamber. The same pressure is used to force the sample into the sheath. The sheath fluid envelops this sample and allows its advance. The flow chamber has a conical morphology, designed following hydrodynamic focusing principles, that focuses the sample in the middle of the sample stream. The speed of cell delivery and analysis depends on the diameter of the flow, which is controlled by a pressure regulator. Currently, some instruments can measure up to 100,000 events per second, but the sample being analyzed controls this speed, that is, for system calibration or cell-cycle analysis, the speed must be around 100–200 events per second, whereas for whole-blood analysis, the speed usually ranges from 400 to 20,000 events per second. To maintain the laminar flow and the focusing of cell-sized particles (0.1–50 mm diameters) requires a stream velocity of approximately 10 m s1 which means that a 10-mm particle will traverse its own diameter in 1ms. At this speed, the need for a rapid interrogation system is obvious. At the interrogation point, the light sources (lasers, arc lamps, light-emitting diodes (LEDs), or lasers) illuminate the focused cells and particles. Nowadays, the most popular light sources are air-cooled and solid-state lasers, because they provide a very sensitive signal, require little space, and have a relatively low cost. If a cell is placed in the light beam, it will emit fluorescence in all directions, but some light of the laser will be dispersed and can be detected by an optical sensor. The intensity of this light varies as a function of different physical (diffusion, cell volume, roundness, and granularity/ruggedness) or biological properties (specific fluorochromes of some cellular components such as DNA, proteins, etc.).1–3,5,8–12,14,15 How can the system distinguish the different emissions produced by the excitation of the cell? Although cells may look the same, they differ in size, density, granularity, functionality, etc. Some markers can be used to identify and characterize subpopulations of

Flow Cytometry

543

cells based on these characteristics. Fluorophores, natural or artificial, can be excited at a particular wavelength; when this happens, an electron is excited from ground state, by the absorption of light, and it temporarily jumps to a higher energy state. When the electron decays, energy is released as a photon, so light is emitted at a longer wavelength than the one that excited the electron. Different fluorophores have different excitation/emission properties and can be used to label specific target molecules or subcellular constituents to enable specific discrimination. These fluorophores can label cell components directly (like propidium iodide (PI), which binds to nucleic acids), can be attached to antibodies (i.e., FITC-CD4), and can be used to control environmental changes (e.g., calcium-sensitive dyes, like Fluo-4 or Indo-1). Some of the typical fluorochromes for immunophenotyping are shown in Table 1 along with their excitation and emission wavelengths. Table 2 shows other fluorochromes commonly used for specific purposes. The fluorescence emitted by each fluorochrome is collected by a unique fluorescence detector, and the specificity of detection is controlled by the wavelength selectivity of optical filters and mirrors. Most of the optics of actual flow cytometers have an orthogonal design: a photodiode is placed opposite to the source of light to measure the forward angle light scatter (FALS) also known as forward scatter (FSC or FS), and a set of optical filters and photomultipliers are deployed at a 90 angle to collect the scattered light. Cells of the same size can be detected and distinguished if they have different internal refraction properties.8,10,11,15,16 For the detection and measurement of this signal, a bar which blocks the laser beam is placed on the far side of the flow chamber opposite to the excitation source; a photodiode is placed after this bar to measure both low-angle diffused light and forward-angle diffuse light (FALS). The measurements made by this detector are displayed as FALS, FSC, or FS (depending on the supplier) and, as mentioned before, is usually considered as relative to the size of the cell. The rest of the scattered light from each laser will pass through a specific set of pinholes situated at the same level as the excitation light at a 90 degree angle. A set of optical filters and dichroic mirrors allows the collection and selection of these optical signals, directing them toward different detectors, called photomultiplier tubes (PMTs). Three types of filters are used: long-pass (LP) filters typically permit the transmission of all light above a set wavelength, which is specified by the number displayed on the side of the filter, for example, a 500LP is an LP filter that allows wavelengths longer than 500 nm to pass; short-pass (SP) filters will filter shorter wavelengths than the ones displayed, for example, 500SP; and band-pass (BP) filters permit the transmission of light between two defined wavelengths, therefore obtaining a more defined signal, for example, a 500BP/40 or 500/40 filters wavelengths from 480 to 520 nm. Dichroic filters or dichroic mirrors reflect light up to one defined wavelength and transmit light beyond a specified wavelength (a dichroic LP filter, or DLP) or vice versa (a dichroic SP filter, or DSP). Depending on the instrument, the filtration of the scattered light can go from the highest to the lowest or from the lowest to the highest fluorescence.1–3,5,8–12,14,15 After the filtration of the signal, the light arrives at different sensitive detectors, or PMTs, which are used to detect fluorescent signals and weak side-scattered light. These PMTs transform the light pulses (photons) into pulses of electrical current according to the diffusion or fluorescence of the particle. These electrical signals are amplified and converted into digital data that a computer will process and, with specific software, generate graphics showing the intensity value generated from each measured particle, Table 1

List of common fluorochromes for immunophenotyping

Fluorochrome

Max. excitation (nm)

Max. emission (nm)

Laser wavelengths (nm)

AMCA Alexa 350 Marina Blue Cascade Blue Pacific Blue Alexa 405 Fluorescein (FITC) Alexa 488 Alexa 532 TRITC Phycoerythrin (PE) PE-Texas Red (ECD) PE-Cyanine-5 (PE-Cy5) Peridin-chlorophyll (PerCP) PE-Cyanine-5.5 (PE-Cy5.5) PE-Cyanine-7 (PE-Cy7) Alexa 633 Alexa 647 Allophycocyanin (APC) Cyanine-5 (Cy5) APC-Cyanine-7 (APC-Cy7) Alexa 660 Cyanine-5.5 (Cy5.5) Alexa 680 Alexa 700

345 350 365 395 405 405 495 500 532 545 565 565 565 490 565 565 630 647 650 650 650 660 675 680 700

440 445 460 420 455 440 520 520 555 580 575 615 670 670 695 770 650 670 660 665 770 690 695 700 720

334–364, 351–356 334–364, 351–356 334–364, 351–356, 405, 407 405, 407 405, 407 405, 407 488 488 514 568 488, 514, 568 488, 514 488, 514 488 488, 514 488, 514 633, 635, 647 633, 635, 647 633, 635, 647 633, 635, 647 633, 635, 647 633, 635, 647 633, 635, 647 633, 635, 647 633, 635, 647

544

Flow Cytometry Table 2

List of common fluorochromes for immunophenotyping

Application

Fluorochrome

Max. excitation (nm)

Max. emission (nm)

Laser wavelengths (nm)

Calcium

Indo-1 (positive) Indo-1 (negative) Fura Red Fluo-3 (Fluo-4) DiOC6 JC1 Rhodamine 123 Mitotracker Green Hoechst 33342 DAPI Acridine Orange (AO) Propidium Iodide (PI) 7-AAD To-Pro-3 SNARF-1 eCFP eGFP eYFP Ds-Red HcRed

325 345 485 500 488 493 488 488 355 360 495 535 545 640 548/579 430 495 520 555 590

400 485 675 540 500 530 525 515 455 460 535 620 650 655 530/640 475 510 535 585 620

334–364, 351–356 334–364, 351–356 458, 488 488 488 488 488 488 334–364, 351–356 334–364, 351–356, 405, 407 488 488, 514, 568. 633, 647 488, 514, 568 633, 635, 647 514, 568 458 488 514 514, 568 568

Membrane potential Mitochondrial activity Mitochondrial mass Nucleic acid

pH Reporter gene

making these measurements easy to understand. In this way, the fluorescence intensity (which depends on how many of the cell’s different components are labeled with fluorescent dyes) is recorded and quantified for each individual cell. Part of the reflected fluorescence corresponds to the emission light scattered at a right angle. This light, called sideward scatter (SSC or SS) or right-angle light scatter (RALS), is roughly proportional to the granularity or the density of the cell.1–3,5,8–12,14,15 New instruments, like the Accuri C6, have a slightly different optical distribution: the detector of the FALS is placed opposite to the source of light, at a right angle to the detector of the RALS. Nevertheless, PMTs and filters are displayed at different angles around the flow cell. This distribution decreases the number of filters needed, therefore increases the detection sensitivity. Multicolor analysis has some advantages: it saves time, saves reagents and samples, produces an exponential increase in information – an increase that allows the identification of new/rare populations ( 0). Examples of measured sensing signals recorded for single CHO cells experiencing nDEP, no DEP, and pDEP is shown on the right. Reprinted from [55] with the permission of AIP Publishing, copyright© 2016 AIP Publishing LLC.

conductivity se ¼ 0.17 S m1 and DEP frequency at 6 MHz, viable CHO cells exhibit pDEP (4 > 0) whereas apoptotic ones exhibit nDEP (4 < 0).27 Therefore, viability can be estimated by the number of events with a positive force index over the total number of events. The DEP cytometer approach has also been used to study the effect of inhibitors,54 nutrient depravation,89,90 and stress due to heat treatment and to determine a comprehensive dielectric model for CHO cells.55

1.43.5.2.2

DEP Cytometry and Cell Viability Assays

The application of a DEP cytometer for monitoring the viability of CHO cells over several days in a bioreactor is described by Braasch et al.28 CHO cells were cultured in a 3L glass bench-top bioreactor and their viability was measured over a 168 h period (every 24 h from inoculation to the end of the culture) using trypan blue exclusion, fluorescent cytometry, and DEP cytometry. Trypan blue assay was performed using a Cedex XS cell analyzer (Innovatis, Germany). Two different fluorescent-based assays, ViaCount and Nexin Annexin V, were performed using a Guava 8HT system (EMD Millipore, Danvers, MA). The Nexin Annexin V assay detects the translocation of phosphatidylserine (PS) to the outer surface of the cell membrane. PS externalization is associated with the onset of apoptosis. For DEP cytometry, a sample of cells was diluted and suspended in a medium with conductivity se ¼ 0.17 S m1. The force indices of more than 600 single cells were measured every 24 h from inoculation to the end of the culture using DEP at 6 MHz. Each measurement takes approximately 15 min. The viability was determined using an automated tool which counts the total number of single cell events as well as the pDEP (viable) and nDEP (non-viable) cells. The results of the assays are shown in Fig. 12. The results show that the DEP and Nexin Annexin V assays detect the decline in the cell viability earlier than the membrane integrity assays, ViaCount and trypan blue. As illustrated in Fig. 5, DEP measurement at 6 MHz is dominantly sensitive to changes in the cells internal conductivity (related to the cells ionic content). The result suggests that the onset of apoptosis, conventionally detected

Figure 12 Comparison of cell viability determined for a CHO batch culture using trypan blue exclusion assay (Cedex analyzer), ViaCount and Nexin Annexin V assays (Guava flow cytometer) and DEP cytometry.28 The culture was sampled every 24 h from 0 to 168 h. The values plotted are the average of four replicates for each sampling point with error bars representing the standard error of the mean. Each DEP value is based on measurement of more than 600 cells.

596

Dielectric Properties of Cells

Figure 13 Time-lapse histograms of the force indices of CHO cells, grown in a batch culture, measured by the DEP cytometer27. It shows a population of viable cells at the beginning of the culture and the emergence and progression of a second population of cells after 96 h. This second subpopulation corresponds to the cells that have progressed to apoptotic or other non-viable states.

by an Annexin V assay, can be detected by the DEP technique which relies on the intrinsic dielectric properties of cells and employs no external marker molecule. By measuring the response of single cells, the DEP cytometer also provides information about the emergence and progression of apoptotic cells. The histogram of the cells measured by the DEP cytometer over the 168 h period is shown in Fig 13. At the beginning of the culture a single population of cells with a positive average force index (viable cells) is observed. This population remains up to 96 h of the culture. After 96 h a second population with a negative force index (non-viable cells) emerges. As the culture proceeds, this second population becomes more prominent. At the end of the culture one single population with a negative average force index is observed. Comparison with the Nexin assay verifies that the second population, appearing after 96 h, corresponds to the population of cells that have progressed to apoptotic or other non-viable states. The obtained results demonstrate the promising potential of the DEP cytometry approach for monitoring bioprocesses. The technique is label-free and non-invasive and requires minimal sample preparation.

References 1. Asami, K.; Takahashi, Y.; Takashima, S. Dielectric Properties of Mouse Lymphocytes and Erythrocytes. Biochim. Biophys. Acta Mol. Cell Res. January 1989, 1010 (1), 49–55. 2. Ermolina, I.; Polevaya, Y.; Feldman, Y.; Ginzburg, B. Z.; Schlesinger, M. Study of Normal and Malignant White Blood Cells by Time Domain Dielectric Spectroscopy. IEEE Trans. Dielectr. Electr. Insul. April 2001, 8 (2), 253–261. 3. Toner, M.; Irimia, D. Blood-on-a-chip. Annu. Rev. Biomed. Eng. 2005, 7, 77–103. 4. Yang, J.; Huang, Y.; Wang, X.; Wang, X. B.; Becker, F. F.; Gascoyne, P. R. Dielectric Properties of Human Leukocyte Subpopulations Determined by Electrorotation as a Cell Separation Criterion. Biophys. J. 1999, 76 (6), 3307–3314. 5. Flanagan, L. A.; Lu, J.; Wang, L.; Marchenko, S. A.; Jeon, N. L.; Lee, A. P.; Monuki, E. S. Unique Dielectric Properties Distinguish Stem Cells and Their Differentiated Progeny. Stem Cell. 2008, 26 (3), 656–665. 6. Zhou, Y.; Basu, S.; Laue, E.; Seshia, A. Single Cell Studies of Mouse Embryonic Stem Cell (mESC) Differentiation by Electrical Impedance Measurements in a Microfluidic Device. Biosens. Bioelectron. 2016, 15 (81), 249–258. 7. Gascoyne, P.; Mahidol, C.; Ruchirawat, M.; Satayavivad, J.; Watcharasit, P.; Becker, F. F. Microsample Preparation by Dielectrophoresis: Isolation of Malaria. Lab Chip May 2002, 2 (2), 70–75. 8. Prieto, J. L.; Su, H.-W.; Hou, H. W.; Vera, M. P.; Levy, B. D.; Baron, R. M.; Han, J.; Voldman, J. Monitoring Sepsis Using Electrical Cell Profiling. Lab Chip November 2016, 16, 4333–4340. 9. Gascoyne, P. R. C.; Noshari, J.; Anderson, T. J.; Becker, F. F. Isolation of Rare Cells from Cell Mixtures by Dielectrophoresis. Electrophoresis 2009, 30 (8), 1388–1398. 10. Huang, Y.; Hölzel, R.; Pethig, R.; Wang, X. B. Differences in the AC Electrodynamics of Viable and Non-viable Yeast Cells Determined through Combined Dielectrophoresis and Electrorotation Studies. Phys. Med. Biol. 1992, 37 (7), 1499–1517. 11. Doh, I.; Cho, Y.-H. A Continuous Cell Separation Chip Using Hydrodynamic Dielectrophoresis (DEP) Process. Sens. Actuators A Phys. May 2005, 121 (1), 59–65. 12. Kim, S. H.; Yamamoto, T.; Fourmy, D.; Fujii, T. An Electroactive Microwell Array for Trapping and Lysing Single-bacterial Cells. Biomicrofluidics June 2011, 5 (2), 24114. 13. Hu, N.; Yang, J.; Qian, S.; Joo, S. W.; Zheng, X. A Cell Electrofusion Microfluidic Device Integrated with 3D Thin-film Microelectrode Arrays. Biomicrofluidics September 2011, 5 (3), 34121.

Dielectric Properties of Cells

597

14. Park, S.; Ben Bassat, D.; Yossifon, G. Individually Addressable Multi-chamber Electroporation Platform with Dielectrophoresis and Alternating-current-electro-osmosis Assisted Cell Positioning. Biomicrofluidics March 2014, 8 (2), 24117. 15. Justice, C.; Brix, A.; Freimark, D.; Kraume, M.; Pfromm, P.; Eichenmueller, B.; Czermak, P. Process Control in Cell Culture Technology Using Dielectric Spectroscopy. Biotechnol. Adv. January 2011, 29 (4), 391–401. 16. Carvell, J. P.; Dowd, J. E. On-line Measurements and Control of Viable Cell Density in Cell Culture Manufacturing Processes Using Radio-frequency Impedance. Cytotechnology March 2006, 50 (1–3), 35–48. 17. Petiot, E.; Ansorge, S.; Rosa-Calatrava, M.; Kamen, A. Critical Phases of Viral Production Processes Monitored by Capacitance. J. Biotechnol. 2017, 242, 19–29. 18. Ansorge, S.; Lanthier, S.; Transfiguracion, J.; Henry, O.; Kamen, A. Monitoring Lentiviral Vector Production Kinetics Using Online Permittivity Measurements. Biochem. Eng. J. 2011, 54, 16–25. 19. Noll, T.; Biselli, M. Dielectric Spectroscopy in the Cultivation of Suspended and Immobilized Hybridoma Cells. J. Biotechnol. 1998, 63 (3), 187–198. 20. Dowd, J. E.; Jubb, A.; Kwok, K. E.; Piret, J. M. Optimization and Control of Perfusion Cultures Using a Viable Cell Probe and Cell Specific Perfusion Rates. Cytotechnology 2003, 42 (1), 35–45. 21. Cannizzaro, C.; Gügerli, R.; Marison, I.; Von Stockar, U. On-line Biomass Monitoring of CHO Perfusion Culture with Scanning Dielectric Spectroscopy. Biotechnol. Bioeng. 2003, 84 (5), 597–610. 22. Ansorge, S.; Esteban, G.; Schmid, G. Multifrequency Permittivity Measurements Enable On-line Monitoring of Changes in Intracellular Conductivity Due to Nutrient Limitations during Batch Cultivations of CHO Cells. Biotechnol. Prog. 2010, 26 (1), 272–283. 23. Opel, C. F.; Li, J.; Amanullah, A. Quantitative Modeling of Viable Cell Density, Cell Size, Intracellular Conductivity, and Membrane Capacitance in Batch and Fed-batch CHO Processes Using Dielectric Spectroscopy. Biotechnol. Prog. January 2010, 26 (4), 1187–1199. 24. Downey, B. J.; Graham, L. J.; Breit, J. F.; Glutting, N. K. A Novel Approach for Using Dielectric Spectroscopy to Predict Viable Cell Volume (VCV) in Early Process Development. Biotechnol. Prog. March 2014, 30 (2), 479–487. 25. Lee, H. W.; Carvell, J.; Brorson, K.; Yoon, S. Dielectric Spectroscopy-based Estimation of VCD in CHO Cell Culture. J. Chem. Technol. Biotechnol. February 2015, 90 (2), 273–282. 26. Kiviharju, K.; Salonen, K.; Moilanen, U.; Meskanen, E.; Leisola, M.; Eerikäinen, T. On-line Biomass Measurements in Bioreactor Cultivations: Comparison Study of Two On-line Probes. J. Ind. Microbiol. Biotechnol. July 2007, 34 (8), 561–566. 27. Nikolic-Jaric, M.; Cabel, T.; Salimi, E.; Bhide, A.; Braasch, K.; Butler, M.; Bridges, G. E.; Thomson, D. J. Differential Electronic Detector to Monitor Apoptosis Using Dielectrophoresis-induced Translation of Flowing Cells (Dielectrophoresis Cytometry). Biomicrofluidics 2013, 7 (2), 24101. 28. Braasch, K.; Nikolic-Jaric, M.; Cabel, T.; Salimi, E.; Bridges, G. E.; Thomson, D. J.; Butler, M. The Changing Dielectric Properties of CHO Cells Can Be Used to Determine Early Apoptotic Events in a Bioprocess. Biotechnol. Bioeng. 2013, 110 (11), 2902–2914. 29. Pethig, R. Dielectric and Electronic Properties of Biological Materials, John Wiley & Sons: New York, 1979. 30. Stratton, J. A. Electromagnetic Theory, Wiley-IEEE Press, 2007. 31. Cole, K. S.; Cole, R. H. Dispersion and Absorption in Dielectrics I. Alternating Current Characteristics. J. Chem. Phys. 1941, 9 (4), 341–351. 32. Schwan, H. P. Electrical Properties of Tissue and Cell Suspensions. Adv. Biol. Med. Phys. 1957, 5, 147–209. 33. Schwan, H. P. Determination of Biological Impedances. In Physical Techniques in Biological Research: Electrophysiological Methods Part B; Nastuk, W. L., Ed., Elsevier, 1963; pp 323–407. 34. Markx, G. H.; Davey, C. L. The Dielectric Properties of Biological Cells at Radiofrequencies: Applications in Biotechnology. Enzym. Microb. Technol. 1999, 25 (3–5), 161–171. 35. Pethig, R.; Kell, D. B. The Passive Electrical Properties of Biological Systems: Their Significance in Physiology, Biophysics and Biotechnology. Phys. Med. Biol. 1987, 32 (8), 933–970. 36. Arnold, W. M.; Zimmermann, U. Electro-rotation: Development of a Technique for Dielectric Measurements on Individual Cells and Particles. J. Electrost. 1988, 21 (2–3), 151–191. 37. Sun, T.; Morgan, H. Single-cell Microfluidic Impedance Cytometry: A Review. Microfluid. Nanofluidics 2010, 8 (4), 423–443. 38. Pethig, R. Dielectrophoresis: Status of the Theory, Technology, and Applications. Biomicrofluidics 2010, 4 (2), 1–36. 39. Maxwell, J. C. A Treatise on Electricity and Magnetism, 3rd ed.; Clarendon Press: Oxford, 1891. 40. Wagner, K. W. The After-effect in Dielectrics. Arch. Elektrotechnik 1914, 2 (9), 371–387. 41. Fricke, H. A Mathematical Treatment of the Electric Conductivity and Capacity of Disperse Systems II. The Capacity of a Suspension of Conducting Spheroids Surrounded by a Non-conducting Membrane for a Current of Low Frequency. Phys. Rev. 1925, 25 (5), 678–681. 42. Irimajiri, A.; Hanai, T.; Inouye, A. A Dielectric Theory of ‘multi-stratified Shell’ Model with its Application to a Lymphoma Cell. J. Theor. Biol. 1979, 78 (2), 251–269. 43. Sillars, R. W. The Properties of a Dielectric Containing Semiconducting Particles of Various Shapes. Inst. Electr. Eng. - Proc. Wirel. Sect. Inst. 1937, 12 (35), 139–155. 44. Fricke, H. The Maxwell-Wagner Dispersion in a Suspension of Ellipsoids. J. Phys. Chem. 1953, 57 (9), 934–937. 45. Bruggeman, D. A. G. Berechnung verschiedener physikalischer Konstanten von heterogenen Substanzen. I. Dielektrizitätskonstanten und Leitfähigkeiten der Mischkörper aus isotropen Substanzen. Ann. Phys. 1935, 416 (7), 636–664. 46. Hanai, T. Theory of the Dielectric Dispersion Due to the Interfacial Polarization and its Application to Emulsions. Kolloid Z. 1960, 171 (1), 23–31. 47. Hanai, T.; Sekine, K. Theory of Dielectric Relaxations Due to the Interfacial Polarization for Two-component Suspensions of Spheres. Colloid Polym. Sci. 1986, 264 (10), 888–895. 48. Asami, K. Characterization of Heterogeneous Systems by Dielectric Spectroscopy. Prog. Polym. Sci. October 2002, 27 (8), 1617–1659. 49. Morgan, H.; Sun, T.; Holmes, D.; Gawad, S.; Green, N. G. Single Cell Dielectric Spectroscopy. J. Phys. D Appl. Phys. 2007, 40 (1), 61–70. 50. Broche, L. M.; Labeed, F. H.; Hughes, M. P. Extraction of Dielectric Properties of Multiple Populations from Dielectrophoretic Collection Spectrum Data. Phys. Med. Biol. May 2005, 50 (10), 2267–2274. 51. Gagnon, Z.; Gordon, J.; Sengupta, S.; Chang, H. C. Bovine Red Blood Cell Starvation Age Discrimination through a Glutaraldehyde-amplified Dielectrophoretic Approach with Buffer Selection and Membrane Cross-linking. Electrophoresis 2008, 29 (11), 2272–2279. 52. Salimi, E.; Thomson, D. J.; Bridges, G. E. Membrane Dielectric Dispersion in Nanosecond Pulsed Electroporation of Biological Cells. IEEE Trans. Dielectr. Electr. Insul. August 2013, 20 (4), 1256–1265. 53. Mulhall, H. J.; Cardnell, A.; Hoettges, K. F.; Labeed, F. H.; Hughes, M. P. Apoptosis Progression Studied Using Parallel Dielectrophoresis Electrophysiological Analysis and Flow Cytometry. Integr. Biol. (Camb). November 2015, 7 (11), 1396–1401. 54. Saboktakin Rizi, B.; Braasch, K.; Salimi, E.; Butler, M.; Bridges, G. E.; Thomson, D. J. Monitoring the Dielectric Response of Single Cells Following Mitochondrial Adenosine Triphosphate Synthase Inhibition by Oligomycin Using a Dielectrophoretic Cytometer. Biomicrofluidics November 2014, 8 (6), 64114. 55. Salimi, E.; Braasch, K.; Butler, M.; Thomson, D. J.; Bridges, G. E. Dielectric Model for Chinese Hamster Ovary Cells Obtained by Dielectrophoresis Cytometry. Biomicrofluidics 2016, 10, 14111. 56. Salimi, E.; Braasch, K.; Butler, M.; Thomson, D. J.; Bridges, G. E. Dielectrophoresis Study of Temporal Change in Internal Conductivity of Single CHO Cells after Electroporation by Pulsed Electric Fields. Biomicrofluidics 2017, 11 (1), 14111. 57. Asami, K.; Yonezawa, T. Dielectric Behavior of Wild-type Yeast and Vacuole-deficient Mutant over a Frequency Range of 10 KHz to 10 GHz. Biophys. J. 1996, 71 (4), 2192–2200. 58. Ferrier, G. A.; Romanuik, S. F.; Thomson, D. J.; Bridges, G. E.; Freeman, M. R. A Microwave Interferometric System for Simultaneous Actuation and Detection of Single Biological Cells. Lab Chip 2009, 9 (23), 3406–3412.

598

Dielectric Properties of Cells

59. Irimajiri, A.; Doida, Y.; Hanai, T.; Inouye, A. Passive Electrical Properties of Cultured Murine Lymphoblast (L5178Y) with Reference to its Cytoplasmic Membrane, Nuclear Envelope, and Intracellular Phases. J. Membr. Biol. September 1978, 38 (3), 209–232. 60. Polevaya, Y.; Ermolina, I.; Schlesinger, M.; Ginzburg, B. Z.; Feldman, Y. Time Domain Dielectric Spectroscopy Study of Human Cells. II. Normal and Malignant White Blood Cells. Biochim. Biophys. Acta 1999, 1419 (2), 257–271. 61. Jones, T. B.; Jones, T. B. Electromechanics of Particles, Cambridge University Press, 2005. 62. Gawad, S.; Schild, L.; Renaud, P. H. Micromachined Impedance Spectroscopy Flow Cytometer for Cell Analysis and Particle Sizing. Lab Chip 2001, 1 (1), 76–82. 63. Foster, K. R.; Schwan, H. P. Dielectric Properties of Tissues and Biological Materials: a Critical Review. Crit. Rev. Biomed. Eng. 1989, 17 (1), 25–104. 64. Sun, T.; Bernabini, C.; Morgan, H. Single-Colloidal Particle Impedance Spectroscopy: Complete Equivalent Circuit Analysis of Polyelectrolyte Microcapsules. Langmuir 2010, 26 (6), 3821–3828. 65. Valero, A.; Braschler, T.; Renaud, P. A Unified Approach to Dielectric Single Cell Analysis: Impedance and Dielectrophoretic Force Spectroscopy. Lab Chip 2010, 10 (17), 2216–2225. 66. Denzi, A.; Merla, C.; Camilleri, P.; Paffi, A.; D’Inzeo, G.; Apollonio, F.; Liberti, M. Microdosimetric Study for Nanosecond Pulsed Electric Fields on a Cell Circuit Model with Nucleus. J. Membr. Biol. 2013, 246 (10), 761–767. 67. Coulter, W. H. High Speed Automatic Blood Cell Counter and Cell Size Analyzer. In Proceedings of the National Electronics Conference; 1956. 1956. 68. Hoffman, R. A.; Johnson, T. S.; Britt, W. B. Flow Cytometric Electronic Direct Current Volume and Radiofrequency Impedance Measurements of Single Cells and Particles. Cytometry 1981, 1 (6), 377–384. 69. Sun, T.; Holmes, D.; Gawad, S.; Green, N. G.; Morgan, H. High Speed Multi-frequency Impedance Analysis of Single Particles in a Microfluidic Cytometer Using Maximum Length Sequences. Lab Chip 2007, 7 (8), 1034. 70. Gawad, S.; Sun, T.; Green, N. G.; Morgan, H. Impedance Spectroscopy Using Maximum Length Sequences: Application to Single Cell Analysis. Rev. Sci. Instrum. 2007, 78 (5). 71. Gawad, S.; Cheung, K.; Seger, U.; Bertsch, A.; Renaud, P. Dielectric Spectroscopy in a Micromachined Flow Cytometer: Theoretical and Practical Considerations. Lab Chip 2004, 4 (3), 241–251. 72. Heidmann, I.; Schade-Kampmann, G.; Lambalk, J.; Ottiger, M.; Di Berardino, M. Impedance Flow Cytometry: A Novel Technique in Pollen Analysis. PLoS One 2016, 11 (11), 1–15. 73. Gascoyne, P.; Satayavivad, J.; Ruchirawat, M. Microfluidic Approaches to Malaria Detection. Acta Trop. February 2004, 89 (3), 357–369. 74. Foster, K. R.; Sauer, F. A.; Schwan, H. P. Electrorotation and Levitation of Cells and Colloidal Particles. Biophys. J. 1992, 63 (1), 180–190. 75. Hölzel, Ralph. Non-invasive Determination of Bacterial Single Cell Properties by Electrorotation. Biochim. Biophys. Acta 1999, 1450 (1), 53–60. 76. Holzel, R. Electrorotation of Single Yeast Cells at Frequencies between 100 Hz and 1.6 GHz. Biophys. J. August 1997, 73 (2), 1103–1109. 77. Ziervogel, H.; Glaser, R.; Schadow, D.; Heymann, S. Electrorotation of Lymphocytes-The Influence of Membrane Events and Nucleus. Biosci. Rep. 1986, 6 (11), 973–982. 78. Hu, X.; Arnold, W. M.; Zimmermann, U. Alterations in the Electrical Properties of T and B Lymphocyte Membranes Induced by Mitogenic Stimulation. Activation Monitored by Electro-rotation of Single Cells. Biochim. Biophys. Acta 1990, 1021 (2), 191–200. 79. Han, S.-I.; Joo, Y.-D.; Han, K.-H. An Electrorotation Technique for Measuring the Dielectric Properties of Cells with Simultaneous Use of Negative Quadrupolar Dielectrophoresis and Electrorotation. Analyst 2013, 138 (5), 1529–1537. 80. Gimsa, J.; Schnelle, T.; Zechel, G.; Glaser, R. Dielectric Spectroscopy of Human Erythrocytes: Investigations under the Influence of Nystatin. Biophys. J. 1994, 66 (4), 1244–1253. 81. Sukhorukov, V. L.; Zimmermann, U. Electrorotation of Erythrocytes Treated with Dipicrylamine: Mobile Charges within the Membrane Show Their Signature in Rotational Spectra. J. Membr. Biol. 1996, 153 (2), 161–169. 82. Georgieva, R.; Neu, B.; Shilov, V. M.; Knippel, E.; Budde, A.; Latza, R.; Donath, E.; Kiesewetter, H.; Bäumler, H. Low Frequency Electrorotation of Fixed Red Blood Cells. Biophys. J. 1998, 74 (4), 2114–2120. 83. Pethig, R. Review Article-dielectrophoresis: Status of the Theory, Technology, and Applications. Biomicrofluidics January 2010, 4 (2). 84. Gagnon, Z. R. Cellular Dielectrophoresis: Applications to the Characterization, Manipulation, Separation and Patterning of Cells. Electrophoresis 2011, 32 (18), 2466–2487. 85. Zeiser, A.; Bedard, C.; Voyer, R.; Jardin, B.; Tom, R.; Kamen, A. On-line Monitoring of the Progress of Infection in Sf-9 Insect Cell Cultures Using Relative Permittivity Measurements. BIT Biotechnol. Bioeng. 1999, 63 (1), 122–126. 86. Holland, T.; Blessing, D.; Hellwig, S.; Sack, M. The In-line Measurement of Plant Cell Biomass Using Radio Frequency Impedance Spectroscopy as a Component of Process Analytical Technology. Biotechnol. J. August 2013, 8 (10), 1231–1240. 87. Nikolic-Jaric, M.; Romanuik, S. F.; Ferrier, G. A.; Cabel, T.; Salimi, E.; Levin, D. B.; Bridges, G. E.; Thomson, D. J. Electronic Detection of Dielectrophoretic Forces Exerted on Particles Flowing over Interdigitated Electrodes. Biomicrofluidics 2012, 6 (2), 1–16. 88. Nikolic-Jaric, M.; Romanuik, S. F.; Ferrier, G. A.; Bridges, G. E.; Butler, M.; Sunley, K.; Thomson, D. J.; Freeman, M. R. Microwave Frequency Sensor for Detection of Biological Cells in Microfluidic Channels. Biomicrofluidics 2009, 3 (3), 34103. 89. Afshar, S.; Fazelkhah, A.; Salimi, E.; Butler, M.; Thomson, D.; Bridges, G. Change in the Dielectric Response of Single Cells Induced by Nutrient Deprivation over a Wide Frequency Range, IEEE MTT-S Int. Microw. Symp. Dig., 2017; pp 872–875, 4872. 90. Salimi, E.; Braasch, K.; Fazelkhah, A.; Afshar, S.; Saboktakin-Rizi, B.; Mohammad, K.; Butler, M.; Bridges, G. E.; Thomson, D. J. Single Cell Dielectrophoresis Study of Apoptosis Progression Induced by Controlled Starvation. Bioelectrochemistry 2018, 124, 73–79.

1.44

Cell Isolation From Tissue

MR Mirbolooki, H Bozorgmanesh, C Foster, III, W Kuhtrieber, and JRT Lakey, University of California at Irvine, Orange, CA, United States © 2011 Elsevier B.V. All rights reserved. This is a reprint of M.R. Mirbolooki, H. Bozorgmanesh, C. Foster, W. Kuhtrieber, J.R.T. Lakey, 1.44 - Cell Isolation from Tissue, Editor: Murray MooYoung, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 591-598.

1.44.1 1.44.2 1.44.3 1.44.4 1.44.5 1.44.6 1.44.6.1 1.44.6.2 1.44.6.2.1 1.44.6.2.2 1.44.6.2.3 1.44.6.2.4 1.44.6.2.5 1.44.6.2.6 1.44.6.2.7 1.44.6.2.8 1.44.6.2.9 1.44.6.2.10 1.44.7 1.44.7.1 1.44.7.2 1.44.7.3 1.44.7.4 1.44.8 1.44.8.1 1.44.8.2 1.44.8.3 1.44.9 References

Introduction Tissue/Organ Procurement Tissue/Organ Preservation Tissue/Organ Rinsing Tissue/Organ Fragmentation Cell Dissociation Nonenzymatic Dissociation Enzymatic Dissociation Stromal Cells Adipocytes Thymus and Umbilical Cord Breast Tissue Colorectal Cancer Cells Epithelial Cells From Intestine Hepatocytes Cells From Skin Hair Follicles Synovia and Cartilage Purification Regular Centrifugation Gradient Centrifugation Adherence Fluorescence-Activated Cell Sorting Cell Yield, Viability, and Purity Assessment Cell Yield Cell Viability Purity Conclusions

599 600 600 601 601 601 601 601 601 602 602 602 602 602 603 603 603 603 603 603 604 604 604 604 604 604 605 605 605

Glossary Cell isolation Separation of certain cells from a tissue/organ. Cytotoxicity The degree to which something is toxic to living cells. Endonuclease An enzyme that cleaves the phosphodiester bond within a polynucleotide chain. Lyophilized Dried by freezing in a high vacuum. Organ perfusion Injection of a fluid into a blood vessel of the procured organ or tissues. Organ preservation Transporting of donor organs, after surgical removal, for processing and transplant. Organ procurement Obtaining organs for transplantation or cell isolation. Reconstitution To bring back a liquid in a concentrated or powder form to its normal strength by adding water.

1.44.1

Introduction

Over the past decades, interest has grown in identifying and characterizing cells isolated from a range of types of human tissues and organs. Experimental models and clinical trials have provided great opportunities to investigate the remarkable biological and clinical properties of these cells. For instance, stromal cells1 found in many adult tissues have been an attractive stem cell source for the

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00036-7

599

600

Cell Isolation From Tissue

regeneration of damaged tissues in clinical applications; cultured skin substitutes (CSS)2 composed of dermal fibroblasts and epidermal keratinocytes have been used as an adjunctive therapy in the treatment of large burn wounds; and transplantation of limbal tissue has replenished the stem cells population to support the regeneration of the entire corneal surface epithelium.3 Isolation of cells from organs such as liver and pancreas has been in the center of interest in many research laboratories. The engraftment of transplanted hepatocytes in the liver has drawn significant interest from the field of cell therapy,4 and islet transplantation has shown the potential to prevent chronic pancreas transplantation complications in diabetics, while providing physiologic glucose control.5 In spite of the advances in the field, there is a confusing inconsistency in the methods of cell isolation and they seem to be far from optimum. Therefore, the choice of one technique over another has often been arbitrary and based more on individual experiences rather than on an understanding of why a certain method works and what modifications could lead it to an even better outcome. The main goal of a cell isolation procedure is to optimize the yield of functionally viable, isolated cells. Many factors affect cell isolation, as it is a complex procedure. Type of tissue, donor’s factors such as body mass index (BMI) and age, warm/cold ischemia during organ procurement/preservation, isolation solution, enzyme(s), variances in any enzyme preparation, concentration(s) of enzyme(s), and digestion time are among the factors that affect the cell isolation outcome. Scientists searching the literature for the protocols for using digestive enzymes and optimal conditions for tissue/cell isolation are often confronted with conflicting data. As a first step, it is important to develop a consistent and reproducible method of isolating intact and viable cells. This article summarizes different tissue/cell isolation protocols to achieve a logical approach for establishing a specific cell isolation standard operating procedure.

1.44.2

Tissue/Organ Procurement

Many factors limit the success of a cell isolation procedure. Donor factors such as cause of death6 and donor age,7 along with the surgical procurement procedure, are among the key factors to be considered at the time of tissue/organ acceptance for the purpose of cell isolation and transplantation. If perfusion is not required to obtain the organ/tissue, the procedure is relatively simple and easy. Tissues are procured either from cadavers or from live donors who are undergoing surgery for another reason. Human subcutaneous adipose tissue can be obtained from healthy women undergoing abdominal dermolipectomy for plastic surgery.6 Bone marrow aspirates can be obtained from the femoral shaft of patients undergoing total hip replacement or puncturing the iliac crest at an orthopedic department.7 Even discarded tissue like human thymus can be obtained from young children undergoing corrective cardiac surgery.8 However, if perfusion is required to obtain an organ like liver or pancreas, the procedure is more complicated and many factors including warm/cold ischemia could affect the outcome of cell isolation. The human pancreas is one of the most challenging organs to procure for vascularized clinical transplantation, based on a need to preserve the integrity of its capsule and of the vascular pedicles for implantation while avoiding injury to the inflow to the liver.9 Different challenges persist when the pancreas is procured for the purpose of islet isolation, because precise procurement techniques are essential for successful isolation of large numbers of viable islets. Surgical expertise, procurement technique, and minimal ischemia time have a major impact on the recovery of functionally viable islets and posttransplant clinical outcomes. In these cases, the donor organs are typically perfused in situ with cold University of Wisconsin (UW) solution before explantation. Cold ischemia time during procurement is defined as the time after clamping of aorta until excision of the organ, which varies from 20 min to 4 h. In some cases, the organ is placed on ice and perfused with ice-cold UW solution by means of multiple catheters inserted into the vessels on the cut surface of the resected fragment immediately after excision.10 Although the most common perfusion solution is UW in multiple organ procurement, perfusion can be done with other solutions including the simple option of Hanks‘ balanced salt solution (HBSS) lacking calcium and magnesium and containing 10 mM HEPES and 5 mM ethyleneglycoltetraacetic acid (EGTA).11

1.44.3

Tissue/Organ Preservation

During organ procurement, blood supply and hence oxygen supply is necessarily interrupted. Therefore, cells cannot continue to meet the energy demands of the active ion-transporting systems and this leads to cell death.12 Research efforts by one of the pioneers of organ preservation Folkert O. Belzer and his colleague James H. Southard resulted in the development of a preservation solution in the late 1980s based on five philosophies.13 They developed a UW solution containing impermeants (raffinose, lactobionate) to minimize hypothermia-induced cell swelling, buffers (phosphate) to prevent intracellular acidosis, a colloid (hydroxyethyl starch) to prevent the expansion of interstitial space during the flush-out period, free radical inhibitors and scavengers (glutathione, allopurinol) to prevent injury from oxygen-free radicals during ischemia and after reperfusion, and energy precursors (Mgþ, adenosine) for energy metabolism during reperfusion period. Histidine–tryptophane–ketoglutarate (HTK) solution is another common preservation solution. It was developed in the 1970s by Bretschneider as a cardioplegia solution14 and is being used increasingly for both kidney15 and liver16 transplantation. HTK contains less potassium and sodium and a strong histidine buffer that increases the osmotic effect of mannitol, which is also included in this solution. Tryptophan is added as a membrane stabilizer, and ketoglutarate is added as a metabolism substrate. These solutions are designed for long-term (12–24 h) organ preservation. For shorter preservation, the media can be simple. Endometrial tissue, for instance, is transported to the laboratory in an isolation medium consisting of Dulbecco‘s modified Eagle medium, high glucose (DMEM-H) culture medium and 5% fetal bovine serum (FBS) plus antibiotics.17

Cell Isolation From Tissue

1.44.4

601

Tissue/Organ Rinsing

The tissue pieces may need to be washed by an appropriate buffer like phosphate buffered saline (PBS)7 or Hank‘s buffered saline solution (HBSS)18,19 to remove contaminating debris and red blood cells. The washing solution may also contain antibiotics.18,19 For some tissues, 2.5 mg/ml amphotericin B20 is added to cover a broad range of bacteria. The washing step usually takes 5–10 min; however, tissue may be incubated in buffer for a longer time (30 min) to more efficiently reduce contamination of the tissue with blood cells and soluble factors.21 Tissue may even be disinfected and dissociated at the same time in DMEM containing penicillin, streptomycin, and collagenase.19 In some specific tissues like colonic mucosal specimens, samples are thoroughly washed in PBS containing 100 IU ml1 penicillin–streptomycin and 2 mM dithiothreitol (DTT, wash buffer) to remove debris.22 They might also be washed in 6.5 mM DTT to remove mucus contamination. After gentle removal of the DTT solution, tissue fragments are rinsed once with HBSS.23 The washing solution may be supplemented by 10% FBS and higher concentrations of antibiotics (200 IU ml1) for prostate tissues.24 For certain tissues including the brain ventricular zone,25 intestinal mucosa,26 and umbilical cord27 tissues, washing solution is Ca/Mg-free HBSS. The corneoscleral tissue (HCECs) is rinsed three times with DMEM containing specific antibiotics (50 mg/ml gentamicin and 1.25 mg ml1 amphotericin B).28

1.44.5

Tissue/Organ Fragmentation

Generally, the tissues need to be cut into multiple pieces with sterile scissors or scalpel.18 The size of the pieces varies based on the tissue type and the purpose of cell isolation. Whereas adipose tissue was cut into small pieces with average weight of 20–50 mg in one study,6 the same tissue was cut into 10–20 mg pieces in another study.21 In some cell isolation procedures like breast cancer cell isolation, the tissue is cut into small pieces, and the pieces are then minced with a blade to yield 2–3 mm3 pieces.29 In some organs like pancreas, the tissue is fragmented into larger pieces after enzyme perfusion.

1.44.6

Cell Dissociation

1.44.6.1

Nonenzymatic Dissociation

Cells attach to surfaces and to each other by cell surface adhesion molecules. Cell adhesion molecules (CAMs) can be classified into four major families30: cadherins, integrins, selectins, and immunoglobulin (Ig) CAMs. The first three are calcium and/or magnesium dependent, whereas members of the last group do not require divalent ions. Removal of calcium and magnesium causes the cells to dissociate from the surface and/or from each other. Extracellular calcium is also needed for the formation of tight junctions between cells.31 Nonenzymatic cell dissociation media do not contain divalent ions and are often supplemented with mixtures of chelators that remove additional divalent ions from the tissues. The commonly used chelators are ethylenediaminetetraacetic acid (EDTA), EGTA, and citrate. These chelators can bind calcium and magnesium from various CAMs. Chelating agents are dissolved at concentrations of 0.1–2% in Ca/Mg-free HBSS or PBS. The solutions may also contain stabilizing agents such as glycerol. Several companies market proprietary nonenzymatic dissociation solutions and claim that they are more effective as compared to the simple chelator-containing solutions. Well-known products are from Sigma, ATCC, Cellgro (Mediatech), Millipore, and Invitrogen. Nonenzymatic dissociation solutions can typically be stored for long periods of time at room temperature. Cytotoxicity of nonenzymatic solutions is highly dependent on the cell types being treated.32 While on the one hand it is known that chelators such as EDTA can have cytotoxic effects, on the other hand, cells can be exposed for longer periods of time to such solutions as compared to trypsin without the risk of damage associated with protein overdigestion.

1.44.6.2

Enzymatic Dissociation

A major obstacle to successful cell dissociation has been the inconsistent enzymatic activity and stability of the enzyme preparations. Variability in enzyme blends has been considered the most important determinant of the success or failure in isolated cell yields, and this variation in potency has been observed between, and even within, enzyme lots.33 Enzymes are typically available as lyophilized powders. They may be stored at 2–8  C, and special care is required when opening the enzyme vials. They should not be opened in humid areas. Any vial has to be brought to room temperature before opening. Ideally, the vials should be taken from the refrigerator at least a half hour before opening and should be left in a dessicator. Before opening the vials, it must be made sure that it is not at all cool to the touch. Once diluted with media or buffer, proteolytic enzymes can undergo autolysis; hence enzymes should be dissolved only just prior to use. Reconstituted enzymes should not be stored at 2–8  C; if necessary, they can be aliquoted and frozen at –20  C. Repeated freeze–thaw cycles should be avoided. All enzymes, upon reconstitution, can be sterile filtered through a 0.22-mm-poresize membrane. As there are variations in the isolation procedures of different types of cells, we will address the differences in enzymes usage, digestion time, enzyme blockade, and collection media between various types of cells isolated from the tissues.

1.44.6.2.1

Stromal Cells

Stromal cells are connective tissue cells that form the supportive structure in which the functional cells of the tissue reside. Stromal cells can be isolated from a variety of tissues, such as bone marrow, periosteum, trabecular bone, synovium, skeletal muscle,

602

Cell Isolation From Tissue

deciduous teeth, and adipose tissues. Collagenase type I (from Clostridium histolyticum) is a crude collagenase preparation that can be used for the isolation of stromal cells. The preparation contains average amounts of caseinase, clostripain, and tryptic activities. To separate the human adipose-derived stromal cell (ASC) fraction from adipocytes, Traktuev et al. digested the tissues in collagenase type I solution under agitation for 2 h at 37  C and centrifuged at 300 g for 8 min. They resuspended the pellet in DMEM–Ham‘s Nutrient Mixture F12 (DMEM–F12) containing 10% FBS, filtered it through 250-mm filters and centrifuged at 300 g for 8 min.1 There are different versions of stromal cells isolation protocols available in the literature with minor differences. For instance, adipose tissues are treated with collagenase type I (1 mg ml1 in HBSS with 1% bovine serum albumin (BSA)) for just 60 min at 37  C with intermittent shaking34 or for 30–60 min at the same temperature with gentle agitation.7 The pellet is centrifuged at the same g-force (300 g) but for a shorter time (5 min)34 or at a higher g-force (400 g) and for a longer time (10 min).20 The activity of the collagenase is neutralized with DMEM-LG containing 10% fetal calf serum (FCS).7 The pellets are resuspended in a red blood cell lysis buffer (2.06 g l1 Tris base, 7.49 g l1 NH4Cl, pH 7.2) for 10 min at room temperature. The suspended cells are passed first through 100-mm and then through 40-mm cell sieves.20 The prostatic stromal cells are also digested with collagenase type I (2 mg ml1) for 2.5 h at 37  C on a shaking rotor. The tissue digest is vigorously pipetted and epithelial clumps settled from stromal cells for 15 min without centrifugation.24

1.44.6.2.2

Adipocytes

Adipocytes are traditionally isolated with collagenase type II at different concentrations based on the tissue‘s site. It is another form of the purified collagenase enzymes prepared to contain higher clostripain activity. Collagenase type I is also used in adipocyte dissociation. The orbital preadipocyte is digested with collagenase type I (2 mg ml1) in HBSS for 45 min at 37  C. The samples are centrifuged at 500 g for 1 min, and the supernatant containing connective tissue debris, collagenase, and lipid is removed leaving a pellet containing preadipocytes.18 However, it is digested in 1 mg ml1 collagenase type II, filtered (with 150-mm nylon mesh), and centrifuged at 200 g for 35 s. In some experiments, adipocytes are digested at 37  C in PBS containing 2% BSA and 2 mg ml1 collagenase for 45 min and filtered through 25 mm filters.35 In others, adipose tissue explants are digested in DMEM containing 0.5 mg ml1 collagenase type II and 1% BSA for 30–40 min at 37  C under constant shaking. At the end of the incubation period, the reaction is stopped by dilution with DMEM and filtered on a silk screen in order to retain undigested explants.6 Adipose tissue may even be dissociated for just 5–10 min in DMEM containing antibiotics, 2 mg ml1 collagenase, and 20 mg ml1 BSA.19

1.44.6.2.3

Thymus and Umbilical Cord

Collagenase type II is also employed for thymus and umbilical cord dissociations. Thymus is cut into small fragments, suspended in 10 ml RPMI 1640 (Rosewell Park Memorial Institute) medium containing 2% FCS, collagenase (1 mg ml1, type II), and deoxyribonuclease (DNAse) (0.02 mg ml1, grade II bovine pancreatic DNAse I), and then digested with intermittent agitation for 15 min at 37  C followed by 5 min at room temperature with constant agitation. To disrupt the dendritic cell (DC)–T cell complexes, EDTA is added (to 0.01 M final) to the digest, and incubation with agitation is continued for another 5 min. The suspension is then passed through a stainless steel sieve to remove aggregates and stromal material.8 Human umbilical cord segments are washed and flushed with Ca/Mg-free PBS to remove clotted blood. Sixty milliliters of a 0.1% solution of collagenase type II dissolved in Ca/Mg-free PBS are gently infused into the umbilical vein and incubated at 37  C for 20 min. The collagenase–endothelial cell suspension is then allowed to drain slowly into a 50-ml tube. This tube is centrifuged for 10 min at 1000 250 g.27

1.44.6.2.4

Breast Tissue

Breast tissue is digested by collagenase type III with very different concentrations. It is lower in secondary proteolytic contaminant activities but contains typical collagenase activity. After washing with HBSS twice, minced tissue is dissociated with 200 IU ml1 of collagenase type III in Medium 199 at 37  C for about 2 h. During incubation, tissue is pipetted every 30 min. Dissociation is stopped by adding 5% FBS, and the cells are diluted with Medium 199 and then filtered sequentially through a sterile 100-mm nylon mesh and a 40-mm cell strainer to obtain a single cell suspension. Cells are then washed twice with HBSS and 2% heat-inactivated calf serum.29 Breast biopsies are cut and rotated for 24 h in a serum-free medium, DMEM–F12 supplemented with 2 mM glutamine and 50 mg ml1 gentamicin, and a high concentration of collagenase type III (900 IU ml1). The fibroblasts are isolated by differential centrifugation of the collagenase digest and plated in DMEM–F12 in T-25 flasks at 37  C.

1.44.6.2.5

Colorectal Cancer Cells

To separate colorectal cancer cells, collagenease type III is also used in combination with DNase I. DNase I is an endonuclease that cleaves DNA preferentially at phosphodiester linkages adjacent to a pyrimidine nucleotide. After gentle removal of the DTT solution, tissue fragments are rinsed once with HBSS, resuspended in serum-free RPMI medium 1640 (2 mM L-glutamine, 120 mg ml1 penicillin, 100 mg ml1 streptomycin, 50 mg ml1 ceftazidime, 0.25 mg ml1 amphotericin-B, 20 mM HEPES) with 200 IU ml1 collagenase type III and 100 IU ml1 DNase I, and incubated for 2 h at 37  C to obtain enzymatic disaggregation. Cells are then resuspended by pipetting and serially filtered by using sterile gauze and 70- and 40-mm nylon meshes.23

1.44.6.2.6

Epithelial Cells From Intestine

Epithelial cells are dissociated by collagenase type IV, which has low tryptic activity. This type of collagenase is usually used when the receptors need to remain integrated. The mucosal layer of intestinal specimens is cut into small pieces and treated with 0.1 mM (small intestine) or 0.2 mM (large intestine) DTT with shaking to free the cells from the tissue. The supernatant containing epithelial

Cell Isolation From Tissue

603

cells from the luminal/villous compartment and intraepithelial lymphocytes is collected. The remaining tissue pieces are treated subsequently with 72.5 IU ml1 of collagenase type IV in heat-inactivated human AB serum, with vigorous shaking at 37  C for 30 min. The cell suspension is then passed through a stainless steel sieve, resulting in a cell fraction containing epithelial cells mainly from the crypt compartment, lamina propria leucocytes, and stromal cells. Cell fractions are then washed in Tris-buffered HBSS containing 0.2% human serum albumin.36

1.44.6.2.7

Hepatocytes

Both collagenase types I and IV have been used to isolate hepatocytes. Following 30 min of initial perfusions, the warmed liver is perfused at a flow rate of 60 ml min111 or 100 ml min110 with 1.7 l of HBSS (containing calcium and magnesium) supplemented with 0.5% BSA, 0.05% collagenase (type IV), penicillin, and streptomycin. Following the collagenase perfusion, softened sections of the liver are dissected and placed into a sterile beaker and chopped with scissors, and then 500 ml of HBSS containing 0.5% BSA, 0.02% collagenase type IV11 or 0.05% collagenase type I,10 penicillin, and streptomycin is added to the mixture. The tissue is incubated at 37  C for 10 min with gentle shaking, and the released cells are filtered first through sterile gauze and then through a 250-mm nylon mesh and are collected into 250-ml centrifuge bottles.11 Hepatocytes are pelleted by centrifugation at either 500 g for 5 min at 4  C37 or 50 g for 3 min and are washed twice with HBSS containing 0.5% BSA, penicillin, and streptomycin.11

1.44.6.2.8

Cells From Skin

Dispase, a neutral protease, has been proven to be a rapid, effective, but gentle, agent for separating intact epidermis from the dermis. First, the dermis and epidermis of the skin are enzymatically separated by incubation with dispase for 2.75 h. The dermal strips are placed in endothelial cell growth medium supplemented with 10% FBS and scraped with sterile angled scissors to release microvascular endothelial cells (HDMEC). The cell suspension is centrifuged and the HDMEC are inoculated in endothelial cell growth medium into flasks coated with attachment factor. At this time collagenase may be added for inducing further cell dissociation. The scraped dermal tissue strips are minced and incubated for 1 h with collagenase for fibroblast isolation.2 The epidermal pieces are digested in trypsin–EDTA solution and the keratinocytes are inoculated into flasks containing lethally irradiated NIH 3T3 cells in a keratinocyte growth medium. Trypsin enzyme degrades protein and it is often referred to as a proteolytic enzyme. The primary basal keratinocytes can be isolated from neonatal foreskins in 4 mg ml1 of dispase and 3 mg ml1 of collagenase in PBS at 37  C for 1.5 h, and the resultant cell suspension is filtered through a 70-mm cell strainer.38

1.44.6.2.9

Hair Follicles

Hair follicles are also isolated by dispase from scalp tissues (0.5–2 cm2 or less). Tissues are rinsed, trimmed to remove excess adipose tissues, cut into small pieces, and subjected to enzymatic dissociation in 12.5 mg ml1 dispase in DMEM for 24 h at 4  C. After treatment, the epidermis is peeled off from the dermis, and hair follicles are plucked from the dermis. Hair follicles are rinsed thoroughly with PBS to prevent contaminating epidermal or dermal cells and are examined under an inverted microscope.39

1.44.6.2.10

Synovia and Cartilage

Synovia are minced and digested with 1.5 mg ml1 collagenase–dispase, 1 mg ml1 hyaluronidase, and 0.15 mg ml1 DNase I for 3–4 h at 37  C.40 The septal cartilage specimens are minced into 1- to 3-mm3 cubes, placed into a spinner flask, and incubated at 37  C in a digestion medium (collagenase type II (2.00 mg ml1), hyaluronidase (0.10 mg ml1), and DNase I (0.15 mg ml1) in DMEM–F12 (1:1) medium) for 18–36 h. After digestion, the dispersed cells are filtered through a 40-mm nylon cell strainer to remove any remaining undigested clumps. The cells are then suspended with PBS and centrifuged at a low speed (1000 g for 7 min) twice to remove any remaining enzymes.41

1.44.7

Purification

It is believed that there are several advantages in transplanting highly purified isolated cells, including improved engraftment, increased safety, and reduced graft immunogenicity.42 Even in research settings, purified cells provide more consistent and reliable data. The purification of cells from tissue extract is performed in four different ways.

1.44.7.1

Regular Centrifugation

Some cells are purified by just simple centrifugation. Preadipocytes43 are centrifuged at 90 g for 1 min, and the top adipocyte layer is removed and then centrifuged again at 90 g for 5 min. The pellet containing preadipocytes is removed, and the cells are washed with DMEM–F12. Isolated stromovascular (SV) cells are separated from adipocytes and the medium by centrifugation in 15-ml tubes very gentle for 1 min at 400 g21 or stronger for 10 min at 600 g.35 The SV cells are defined as those cells isolated by collagenase digestion that do not float. After liver digestion, cell suspension is centrifuged three times in low gravitational force (50 g, 5 min) to separate hepatocytes from nonparenchymal cells.10

604 1.44.7.2

Cell Isolation From Tissue Gradient Centrifugation

SV Cell suspensions (15 ml) are applied to Histopaque-1077 gradients (15 ml) in 50-ml tubes. After centrifugation (400 g, 30 min20 or 8–10 min17), the cells at the gradient interface are collected, washed in HBSS, and passed through a 30-mm mesh.20 The bone marrow aspirates are diluted 1:5 with 2 mM EDTA–PBS. The mononuclear cell (MNC) fraction is isolated by density gradient centrifugation at 435 g for 30 min at room temperature using Ficoll-Hypaque Plus solution.44,45 Currently, the purification of islets from exocrine tissue is performed by continuous Ficoll gradients using a refrigerated COBE 2991 cell processor.46 The cells are recovered from the digest by centrifugation. Then the pellet is immediately resuspended in Nycodenz medium (1.068 g cm3 and iso-osmotic with human serum), and a low-density fraction is collected after centrifugation at 1700 g for 10 min.8

1.44.7.3

Adherence

The cellular pellet is washed with DMEM–F12 containing 15% (v/v) FCS and seeded on 48-well plates. Cells are left overnight to attach and all unattached cells (including red blood cells) are washed the following day with HBSS.44,45 Nonadherent cells are removed 12–18 h after initial plating by intensely washing the plates. Adipose tissue (AT)-derived fibroblastoid adherent cells are harvested at subconfluence using trypsin.7

1.44.7.4

Fluorescence-Activated Cell Sorting

The fluorescene-activated cell sorter is a machine that can rapidly separate the cells in a suspension on the basis of size and the color of their fluorescence. This apparatus can sort as many as 300 000 cells per minute. For dendritic cell purification, the cells are then incubated for 25 min with a mixture of monoclonal antibodies (mAbs), including anti-CD3, anti-CD8, anti-CD7, anti-CD15, antiCD19, anti-CD20, and anti-glycophorin A, in EDTA–SS containing 2% human serum. After incubation, the cells coated with mAbs are removed by two cycles of sheep anti-mouse immunoglobulin-coupled magnetic beads. The first cycle is at a 3:1 and the second at a 6:1 bead-to-cell ratio. The cells are then kept overnight at 4  C in EDTA–SS containing 10% FCS. The next morning, the cells are incubated for 25 min at 4  C with Cy5-conjugated anti-HLA-DR and biotinylated anti-CD11b in EDTA–SS containing 2% human serum. After two washes, the cells are incubated with streptavidin–Texas Red. DC populations are then sorted by means of a FacStar Plus.8

1.44.8

Cell Yield, Viability, and Purity Assessment

1.44.8.1

Cell Yield

Cell yield can be estimated by counting trypan blue-stained samples, using a hemocytometer.11,37 However, total colonocyte yield is determined as the final wet weight of the purified cell pellet as compared to the initial wet weight of the washed epithelium or tumor prior to cell isolation.22 The yield of isolated islets is evaluated with a light microscope after dithizone (DTZ) staining. DTZ is a zincchelating agent known to selectively stain pancreatic beta cells because of their high zinc content. Determination of the islet mass is important for the normalization of islet experiments in the laboratory and for the precise dosing of islets for transplantation. Therefore, the common microscopic analysis is based on individual islet sizing, calculation of the frequency distribution, and conversion into islet equivalent (IEQ), which is the volume of a spherical islet with a diameter of 150 mm. However, islets are of irregular form, which makes this determination user-dependent, and the analysis is irreproducible once the original sample is discarded. Recently, Lembert et al. showed that areal–density measurements allow a rapid and reproducible estimation of IEQ without counting individual islets. It can be performed in a single-step analysis without computer programming and is valuable for online determination of islet yield preceding transplantation.47 An improved method of islet volume determination using digital image analysis (DIA) has been also developed to remove operator bias and automate the islet counting process. It was found that volumes determined by DIA correlated more closely with insulin content and DNA content than did conventionally determined volumes. The quantification of isolated islet tissue volume using DIA has been shown to be rapid, consistent, and objective. In the laboratory, use of this method as the standard for islet volume measurement will allow more meaningful comparison of experimental results between centers. In the clinic, its use will allow more accurate dosing of transplanted tissue.48 The islets need to be stained with DTZ, a zinc chelating agent, which is known to selectively stain the islets of Langerhans in the pancreas.

1.44.8.2

Cell Viability

The viability of cell lots after isolation is typically assessed via trypan blue staining, in which dead cells stain blue.10,44,45 The viability of cells can be confirmed using two different dyes: trypan blue exclusion and [3H]leucine uptake.22 To assess viability, Annexin V, to identify early apoptotic cells, and 7-AAD, to identify late apoptotic and necrotic cells, may be used. One of the most important factors in cell transplantation is isolation of a large number of cells with good viability. Cell viability could be assessed through functional assays including static incubation and perifusion tests of glucose-stimulated insulin secretion for isolated islets. For clinical islet transplantation, viability assessment of isolated islets must be simple, rapid, sensitive, and prospective. Fluorescein diacetate (FDA) causes live cells to fluoresce green under blue light excitation, and ethidium bromide (EB) causes dead

Cell Isolation From Tissue

605

cells to fluoresce red. Discrimination of living from dead islets by insulin secretion correlated well with viability as determined by FDA/EB staining. The FDA/EB assay prospectively and easily provides a rapid, accurate, and objective measurement of the proportion of living cells and dead cells in isolated islets for clinical islet transplantation.49 Double staining with FDA and propidium iodide (FDA/PI) is the current international standard to determine islet viability. However, a study by our group that evaluated the SYTO-13/EB (SYTO/EB) and FDA/PI techniques suggests that FDA/PI staining may overestimate islet viability and demonstrates consistently elevated values as compared to SYTO/EB. The discrepancies found between FDA/PI scoring and visual quality, when compared with alternative stains, suggest that the FDA/PI stain may not be the optimal approach to assess islet viability.49

1.44.8.3

Purity

Stromal cell populations are routinely assessed by phase microscopy and by immunohistochemistry using antibodies for cytokeratins and vimentin for purity. Nonmalignant and malignant cell population purity is confirmed by morphological evaluation after hematoxylin/eosin staining and by cytokeratin immunohistochemistry (anti-Pan-Keratin clone AE1/AE3) and parallel histological analysis.22 The purity of isolated islets is evaluated with a light microscope after DTZ staining. DTZ selectively stains pancreatic beta cells to orange. The exocrine part of the tissue remains yellow under the microscope. The purity of cells can also be analyzed by flow cytometry.8,26,50

1.44.9

Conclusions

Clinical outcomes of cell transplantation are influenced by numerous variables in the isolation process and pretransplant culture. Many technical challenges in these procedures must be addressed if clinical cell transplantation is to improve. Cooperation between tissue/organ transplant centers, the procurement team, and the isolation laboratory is crucial to ensure that the available cadaveric tissue/organs are referred appropriately and expediently. Cell yields remain quite variable (typically 25–75% of the potential cell mass). Moreover, clinical results vary considerably across centers in spite of comprehensive efforts to standardize isolation/purification procedures and establish strict quality control criteria in accordance with World Health Organization (WHO) good manufacturing practice (GMP) guidelines. Production of high-quality cells is expensive, labor-intensive, and time-consuming. The process has a steep learning curve and is yet to be standardized. Despite the efforts to manufacture highly purified and standardized collagenase blends, the heterogeneity of the preparations, quality and nature of donor organs, and prolonged cold ischemia times hamper a process that is inherently difficult to control. To consistently maximize cell yield and viability for transplantation, one solution may be to optimize the cell isolation procedure. Obtaining low yield–low viability cells is mostly due to overdigestion of the tissue. It is required to either change to lessdigestive-type enzymes or decrease the enzyme concentration. However, obtaining low yield–high viability cells is mostly due to underdigestion of the tissue. It is required to increase either the enzyme concentration or the incubation time. If yield remains poor, evaluating a more digestive type enzyme and/or the addition of secondary enzyme(s) might be necessary. Based upon the enzyme(s) used, setting up the preliminary dissociation conditions similar to those of the closest available reference(s) in the literature will speed up the process of optimization. After optimizing the primary enzyme‘s concentration and incubation conditions, evaluating a secondary enzyme(s) may help as well. For accurate evaluation of a particular procedure‘s performance, cell yield and viability should be quantified and compared. Based upon these results, the method may be judged suitable for clinical use or reoptimized.51

References 1. Traktuev, D. O.; Merfeld-Clauss, S.; Li, J.; et al. A Population of Multipotent CD34-positive Adipose Stromal Cells Share Pericyte and Mesenchymal Surface Markers, Reside in a Periendothelial Location, and Stabilize Endothelial Networks. Circ. Res. 2008, 102, 77–85. 2. Supp, D. M.; Wilson-Landy, K.; Boyce, S. T. Human Dermal Microvascular Endothelial Cells Form Vascular Analogs in Cultured Skin Substitutes after Grafting to Athymic Mice. Faseb. J. 2002, 16, 797–804. 3. Espana, E. M.; Romano, A. C.; Kawakita, T.; et al. Novel Enzymatic Isolation of an Entire Viable Human Limbal Epithelial Sheet. Investig. Ophthalmol. Visual Sci. 2003, 44, 4275–4281. 4. Gupta, S.; Malhi, H.; Gagandeep, S.; Novikoff, P. Liver Repopulation with Hepatocyte Transplantation: New Avenues for Gene and Cell Therapy. J. Gene Med. 1999, 1, 386–392. 5. Ryan, E. A.; Paty, B. W.; Senior, P. A.; et al. Five-year Follow-up after Clinical Islet Transplantation. Diabetes 2005, 54, 2060–2069. 6. Gesta, S.; Lolmède, K.; Daviaud, D.; et al. Culture of Human Adipose Tissue Explants Leads to Profound Alteration of Adipocyte Gene Expression. Horm. Metab. Res. 2003, 35, 158–163. 7. Kern, S.; Eichler, H.; Stoeve, J.; et al. Comparative Analysis of Mesenchymal Stem Cells from Bone Marrow, Umbilical Cord Blood, or Adipose Tissue. Stem Cell. 2006, 24, 1294–1301. 8. Vandenabeele, S.; Hochrein, H.; Mavaddat, N.; et al. Human Thymus Contains 2 Distinct Dendritic Cell Populations. Blood 2001, 97, 1733–1741. 9. Sollinger, H. W.; Young, C. J. Procurement of Pancreatic Islets. In Pancreatic Islet Transplantation; Lanza, R. P., Chick, W. L., Eds., R.G. Landes Co: Austin, TX, 1994; p 1. 10. Dandri, M.; Burda, M. R.; Török, E.; et al. Repopulation of Mouse Liver with Human Hepatocytes and in Vivo Infection with Hepatitis B Virus. Hepatology 2001, 33, 981–988. 11. Duanmu, Z.; Locke, D.; Smigelski, J.; et al. Effects of Dexamethasone on Aryl (SULT1A1)- and Hydroxysteroid (SULT2A1)-sulfotransferase Gene Expression in Primary Cultured Human Hepatocytes. Drug Metabol. Dispos. 2002, 30, 997–1004. 12. Boutilier, R. G. Mechanisms of Cell Survival in Hypoxia and Hypothermia. J. Exp. Biol. 2001, 204, 3171–3181.

606

Cell Isolation From Tissue

13. Belzer, F. O.; Southard, J. H. Principles of Solid-organ Preservation by Cold Storage. Transplantation 1988, 45, 673–676. 14. Bretschneider, H. J. Myocardial Protection. J. Thorac. Cardiovasc. Surg. 1980, 28, 295–302. 15. de Boer, J.; De Meester, J.; Smits, J. M.; et al. Eurotransplant Randomized Multicenter Kidney Graft Preservation Study Comparing HTK with UW and Euro-Collins. Transpl. Int. 1999, 12, 447–453. 16. Hatano, E.; Kiuchi, T.; Tanaka, A.; et al. Hepatic Preservation with Histidine-tryptophan-ketoglutarate Solution in Living-related and Cadaveric Liver Transplantation. Clin. Sci. (Lond.) 1997, 93, 81–88. 17. Arnold, J. T.; Kaufman, D. G.; Seppälä, M.; Lessey, B. A. Endometrial Stromal Cells Regulate Epithelial Cell Growth in Vitro: A New Co-culture Model. Hum. Reprod. 2001, 16, 836–845. 18. Bujalska, I. J.; Durrani, O. M.; Abbott, J.; et al. Characterisation of 11beta-hydroxysteroid Dehydrogenase 1 in Human Orbital Adipose Tissue: A Comparison with Subcutaneous and Omental Fat. J. Endocrinol. 2007, 192, 279–288. 19. Rodriguez, A. M.; Pisani, D.; Dechesne, C. A.; et al. Transplantation of a Multipotent Cell Population from Human Adipose Tissue Induces Dystrophin Expression in the Immunocompetent Mdx Mouse. J. Exp. Med. 2005, 201, 1397–1405. 20. Boquest, A. C.; Shahdadfar, A.; Frønsdal, K.; et al. Isolation and Transcription Profiling of Purified Uncultured Human Stromal Stem Cells: Alteration of Gene Expression after in Vitro Cell Culture. Mol. Biol. Cell 2005, 16, 1131–1141. 21. Fain, J. N.; Madan, A. K.; Hiler, M. L.; et al. Comparison of the Release of Adipokines by Adipose Tissue, Adipose Tissue Matrix, and Adipocytes from Visceral and Subcutaneous Abdominal Adipose Tissues of Obese Humans. Endocrinology 2004, 145, 2273–2282. 22. Emenaker, N. J.; Basson, M. D. Short Chain Fatty Acids Differentially Modulate Cellular Phenotype and C-myc Protein Levels in Primary Human Nonmalignant and Malignant Colonocytes. Dig. Dis. Sci. 2001, 46, 96–105. 23. Dalerba, P.; Dylla, S. J.; Park, I. K.; et al. Phenotypic Characterization of Human Colorectal Cancer Stem Cells. Proc. Natl Acad. Sci. U.S.A. 2007, 104, 10158–10163. 24. Le, H.; Arnold, J. T.; McFann, K. K.; Blackman, M. R. DHT and Testosterone, but Not DHEA or E2, Differentially Modulate IGF-i, IGFBP-2, and IGFBP-3 in Human Prostatic Stromal Cells. Am. J. Physiol. Endocrinol. Metab. 2006, 290, E952–E960. 25. Roy, N. S.; Benraiss, A.; Wang, S.; et al. Promoter-targeted Selection and Isolation of Neural Progenitor Cells from the Adult Human Ventricular Zone. J. Neurosci. Res. 2000, 59, 321–331. 26. Kanai, T.; Totsuka, T.; Uraushihara, K.; et al. Blockade of B7-h1 Suppresses the Development of Chronic Intestinal Inflammation. J. Immunol. 2003, 171, 4156–4163. 27. Takano, M.; Meneshian, A.; Sheikh, E.; et al. Rapid up Regulation of Endothelial P-selectin Expression via Reactive Oxygen Species Generation. Am. J. Physiol. Heart Circ. Physiol. 2002, 283, H2054–H2061. 28. Li, W.; Sabater, A. L.; Chen, Y. T.; et al. A Novel Method of Isolation, Preservation, and Expansion of Human Corneal Endothelial Cells. Investig. Ophthalmol. Visual Sci. 2007, 48, 614–620. 29. Liu, R.; Wang, X.; Chen, G. Y.; et al. The Prognostic Role of a Gene Signature from Tumorigenic Breast-cancer Cells. N. Engl. J. Med. 2007, 356, 217–226. 30. Makrilia, N.; Kollias, A.; Manolopoulos, L.; Syrigos, K. Cell Adhesion Molecules: Role and Clinical Significance in Cancer. Canc. Invest. 2009, 27, 1023–1037. 31. Gonzalez-Mariscal, L.; Contreras, R. G.; Bolívar, J. J.; et al. Role of Calcium in Tight Junction Formation between Epithelial Cells. Am. J. Physiol. 1990, 259, C978–C986. 32. Hugenschmidt, S.; Planas-Bohne, F.; Taylor, D. M. On the Toxicity of Low Doses of Tetrasodium-ethylenediamine-tetraacetate (Na-edta) in Normal Rat Kidney (NRK) Cells in Culture. Arch. Toxicol. 1993, 67, 76–78. 33. Barnett, M. J.; Zhai, X.; LeGatt, D. F.; et al. Quantitative Assessment of Collagenase Blends for Human Islet Isolation. Transplantation 2005, 80, 723–728. 34. Jeon, E. S.; Song, H. Y.; Kim, M. R.; et al. Sphingosylphosphorylcholine Induces Proliferation of Human Adipose Tissue-derived Mesenchymal Stem Cells via Activation of JNK. J. Lipid Res. 2006, 47, 653–664. 35. Planat-Benard, V.; Silvestre, J. S.; Cousin, B.; et al. Plasticity of Human Adipose Lineage Cells toward Endothelial Cells: Physiological and Therapeutic Perspectives. Circulation 2004, 109, 656–663. 36. Fahlgren, A.; Hammarström, S.; Danielsson, A.; Hammarström, M. L. Increased Expression of Antimicrobial Peptides and Lysozyme in Colonic Epithelial Cells of Patients with Ulcerative Colitis. Clin. Exp. Immunol. 2003, 131, 90–101. 37. Malhi, H.; Irani, A. N.; Gagandeep, S.; Gupta, S. Isolation of Human Progenitor Liver Epithelial Cells with Extensive Replication Capacity and Differentiation into Mature Hepatocytes. J. Cell Sci. 2002, 115, 2679–2688. 38. Li, A.; Pouliot, N.; Redvers, R.; Kaur, P. Extensive Tissue-regenerative Capacity of Neonatal Human Keratinocyte Stem Cells and Their Progeny. J. Clin. Investig. 2004, 113, 390–400. 39. Yu, H.; Fang, D.; Kumar, S. M.; et al. Isolation of a Novel Population of Multipotent Adult Stem Cells from Human Hair Follicles. Am. J. Pathol. 2006, 168, 1879–1888. 40. Liagre, B.; Vergne-Salle, P.; Corbiere, C.; et al. Diosgenin, a Plant Steroid, Induces Apoptosis in Human Rheumatoid Arthritis Synoviocytes with Cyclooxygenase-2 Overexpression. Arthritis Res. Ther. 2004, 6, R373–R383. 41. Dunham, B. P.; Koch, R. J. Basic Fibroblast Growth Factor and Insulinlike Growth Factor I Support the Growth of Human Septal Chondrocytes in a Serum-free Environment. Arch. Otolaryngol. Head Neck Surg. 1998, 124, 1325–1330. 42. Gores, P. F.; Sutherland, D. E. Pancreatic Islet Transplantation: Is Purification Necessary? Am. J. Surg. 1993, 166, 538–542. 43. Quinkler, M.; Sinha, B.; Tomlinson, J. W.; et al. Androgen Generation in Adipose Tissue in Women with Simple Obesity – a Site-specific Role for 17beta-hydroxysteroid Dehydrogenase Type 5. J. Endocrinol. 2004, 183, 331–342. 44. Seboek, D.; Linscheid, P.; Zulewski, H.; et al. Somatostatin Is Expressed and Secreted by Human Adipose Tissue upon Infection and Inflammation. J. Clin. Endocrinol. Metab. 2004, 89, 4833–4839. 45. Oberlin, E.; Tavian, M.; Blazsek, I.; Péault, B. Blood-forming Potential of Vascular Endothelium in the Human Embryo. Development 2002, 129, 4147–4157. 46. Ricordi, C. The Automated Method for Islet Isolation. In Methods in Cell Transplantation; Ricordi, C., Ed., RG Landes: Austin, TX, 1995; pp 99–112. 47. Lembert, N.; Wesche, J.; Petersen, P.; et al. A Real Density Measurement Is a Convenient Method for the Determination of Porcine Islet Equivalents without Counting and Sizing Individual Islets. Cell Transplant. 2003, 12, 33–41. 48. Stegmann, J. P.; O‘Neil, J. J.; Nicholson, D. T.; Mullon, C. J. Improved Assessment of Isolated Islet Tissue Volume Using Digital Image Analysis. Cell Transplant. 1998, 7, 469–478. 49. Miyamoto, M.; Morimoto, Y.; Nozawa, Y.; et al. Establishment of Fluorescein Diacetate and Ethidium Bromide (FDAEB) Assay for Quality Assessment of Isolated Islets. Cell Transplant. 2000, 9, 681–686. 50. Panchision, D. M.; Chen, H. L.; Pistollato, F.; et al. Optimized Flow Cytometric Analysis of Central Nervous System Tissue Reveals Novel Functional Relationships Among Cells Expressing CD133, CD15, and CD24. Stem Cell. 2007, 25, 1560–1570. 51. Rønnov-Jessen, L.; Villadsen, R.; Edwards, J. C.; Petersen, O. W. Differential Expression of a Chloride Intracellular Channel Gene, CLIC4, in Transforming Growth Factor-beta1mediated Conversion of Fibroblasts to Myofibroblasts. Am. J. Pathol. 2002, 161, 471–480.

Nanobiotechnologyq

1.45

KK Jain, Jain PharmaBiotech, Basel, Switzerland © 2019 Elsevier B.V. All rights reserved. This is an update of K.K. Jain, 1.45 - Nanobiotechnology, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 599-614.

1.45.1 1.45.2 1.45.3 1.45.3.1 1.45.3.2 1.45.3.3 1.45.3.4 1.45.3.5 1.45.4 1.45.4.1 1.45.4.2 1.45.4.3 1.45.5 1.45.5.1 1.45.5.2 1.45.5.3 1.45.5.4 1.45.5.5 1.45.6 1.45.7 1.45.7.1 1.45.7.2 1.45.8 1.45.8.1 1.45.8.2 1.45.8.3 1.45.9 1.45.9.1 1.45.9.2 1.45.9.3 1.45.9.4 1.45.10 1.45.11 1.45.12 1.45.13 1.45.13.1 1.45.13.2 1.45.13.3 1.45.14 References

1.45.1

Introduction Nanoparticles Role of Nanobiotechnology in Molecular Diagnostics Nanoparticles for Molecular Diagnostics Nanoparticles as Biolabels Paramagnetic and Superparamagnetic Nanoparticles Role of Nanobiotechnology in Discovery of Biomarkers Nanobiotechnology and Cytogenetics Pharmaceutical Applications of Nanobiotechnology Role of Nanobiotechnology in Drug Discovery and Development Nanobiotechnology-Based Drug Delivery Nanobiotechnology and Drug Delivery Devices Role of Nanobiotechnology in Biological Therapies Role of Nanobiotechnology in Cell Therapy Role of Nanobiotechnology in Gene Therapy Role of Nanobiotechnology in Vaccines Role of Nanobiotechnology in Antisense Therapy Role of Nanobiotechnology in RNA Interference Clinical Nanomedicine Nanooncology Nanotechnology-Based Drug Delivery in Cancer Combination of Diagnosis and Therapy in Cancer Nanoneurology Role of Nanobiotechnology in Diagnosis of Neurological Disorders Nanoparticle-Based Drug Delivery to the Brain Nanoparticles for Neuroprotection Nanocardiology Combining Diagnosis with Therapy of Cardiovascular Disorders Targeted Drug Delivery in Cardiovascular Disorders Role of Nanobiotechnology in Management of Restenosis Tissue Engineering and Regeneration of the Cardiovascular System Nanosurgery Nanorobotics Role of Nanobiotechnology for the Development of Personalized Medicine Safety Issues of Nanoparticles Fate of Nanoparticles in the Human Body Impact of Nanoparticles on Human Health Measures to Resolve Toxicity Issues of Nanoparticles Prospects of Nanobiotechnology

607 610 610 610 611 611 612 613 613 613 613 614 614 614 615 615 616 616 617 617 617 617 619 619 619 619 620 620 620 620 620 621 621 621 622 622 622 623 623 623

Introduction

Nanotechnology (‘nano’ in Greek means dwarf) is the creation and utilization of materials, devices, and systems through the control of matter on the nanometerdlength scale, that is, at the level of atoms, molecules, and supramolecular structures. Nanotechnology, as defined by the National Nanotechnology Initiative (http://www.nano.gov/), is the understanding and control of matter at dimensions of approximately 1–100 nm, where unique phenomena enable novel applications. Encompassing nanoscale science,

q

Change History: November 2018. KK Jain made updates to tables, revised the figure, author biography updated, new reference added.

Comprehensive Biotechnology, 3rd edition, Volume 1

https://doi.org/10.1016/B978-0-444-64046-8.00050-1

607

608

Nanobiotechnology

engineering, and technology, nanotechnology involves imaging, measuring, modeling, and manipulating matter at this length scale. It is the popular term for the construction and utilization of functional structures with at least one characteristic dimension measured in nanometers (nm) a nanometer is one-billionth of a meter (10 9 m). This is roughly four times the diameter of an individual atom and the bond between two individual atoms is 0.15 nm long. Proteins are 1–20 nm in size. The definition of ‘small’, another term used in relation to nanotechnology, depends on the application, but can range from 1 nm to 1 mm. Dimensions of various objects in nanoscale are shown in Table 1. Nature constructs complex, efficient, self-organizing, and self-regulating molecular machinery and systems for all processes in living organisms. Nature has made highly precise and functional nanostructures for billions of years: DNA, proteins, membranes, etc., but it is only since the 1980s that humans have been able to manufacture such precise synthetic nanostructures at will. Nanostructures are made from thousands of atoms that are precisely defined in space. They have an unlimited number of compositions, sizes, shapes, and, most importantly, functionality. Classification of various nanobiotechnologies is shown in Table 2. Beyond structure and function, nanobiotechnology extends to the visualization and manipulation on a nanoscale. In its most basic form, atomic force microscopy (AFM) images topography by precisely scanning a probe across the sample to ‘feel’ the contours of the surface. The interaction between the needle and the surface is measured and an image is reconstructed from the data collected in this manner. Cantilevers, the basic technology in AFM, transform a chemical reaction into a mechanical motion on the nanometer scale. Measurements of a cantilever are in micrometer scale but deflections are in nanometers. This motion can be measured directly by deflecting a light beam from the cantilever surface. With AFM, it is possible to reach an extremely high resolution. In addition to its superior resolution and routine three-dimensional measurement capability, AFM offers several other clear advantages over traditional microscopy techniques. For example, scanning and transmission electron microscopy (SEM and TEM) image biologically inactive, dehydrated samples and generally require extensive sample preparation such as staining or metal coating. AFM eliminates these requirements and, in many cases, allows the direct observation of native specimens and ongoing processes under native or near-native conditions. Further adding to its uniqueness, AFM can directly measure nanoscale interactive forces, for example, ligand-receptor binding. Several other microscope systems have now been developed for application in nanobiotechnology. Nanofluidics, the next stage in the reduction of volume from microfluidics, implies extreme reduction in quantity of fluid analyte in a microchip. The benefits of operating in the nanoliter space include reducing solvent, waste disposal costs, and human exposure by factors of 1000. Optical spatially resolved flow measurements in nanochannels are difficult to visualize. There is a need for refinement of microscale flow visualization methods and the development of direct measurement methods for nanoflow. Various nanobiotechnologies and nanostructures are in development in academic centers and well as in the commercial sector, which are described in detail in a special report on this topic.1 Various nanomachines and other nano-objects that are currently under investigation in medical research and diagnostics will soon find applications in the practice of medicine. Nanobiotechnologies are being used to create and study models of human disease, particularly immune disorders. The application of nanobiotechnology to medicine is termed nanomedicine.2 As a broad term, nanomedicine also covers the applications of nanobiotechnologies in molecular diagnostics, which is also referred to as nanodiagnostics. Application in the pharmaceutical industry for drug discovery and drug delivery can be termed ‘nanobiopharmaceuticals’. The prefix ‘nano’ has been applied to various therapeutic areas of medicine, for example, nanooncology for the application of nanobiotechnologies in cancer, and nanoneurology to indicate applications for neurological disorders. The relationship of nanobiotechnology to healthcare and related technologies is depicted graphically in Fig. 1. Although current applications of nanotechnology dominate the communications and engineering industries, future applications in healthcare and life sciences are expanding rapidly. The most promising application relevant to healthcare is in the development of personalized medicine, which simply means the selection of best treatment for an individual patient.3 Refinement of molecular diagnostics, combination of diagnostics with therapeutics, and targeted drug delivery play important roles in this application.

Table 1

Dimensions of various objects in nanoscale

Object

Dimension

Width of a hair Red blood cell Vesicle in a cell Bacterium Virus Exosomes (nanovesicles shed by dendritic cells) Width of DNA Ribosome A base pair in human genome Proteins Amino acid (e.g., tryptophan, the largest) Aspirin molecule An individual atom

50,000 nm 2500 nm 200 nm 1000 nm 100 nm 65–100 nm 2.5 nm 2–4 nm 0.4 nm 1–20 nm 1.2 nm (longest measurement) 1 nm 0.25 nm

Nanobiotechnology Table 2

609

Classification of basic nanobiotechnologies

Nanoparticles Fluorescent nanoparticles Fullerenes Gold nanoparticles Lipoparticles Magnetic nanoparticles Nanocrystals Nanoparticles assembly into micelles Nanoshells Paramagnetic and superparamagnetic nanoparticles Polymer nanoparticles Quantum dots Silica nanoparticles Nanofibers Nanowires Carbon nanofibers Dendrimers Polypropylenimine dendrimers Composite nanostructures Cochleates DNA-nanoparticle conjugates Nanocapsules enclosing other substances Nanoemulsions Nanolipisomes Nanoshells Nanovesicles Nanoconduits Nanotubes Nanopipettes Nanoneedles Nanochannels Nanopores Nanofluidics Nanostructured silicon Nanoscale motion and manipulation at nanoscale Cantilevers Femtosecond laser systems Nanomanipulation Surface plasmon resonance Visualization at nanoscale Atomic force microscopy Magnetic resonance force microscopy and nanoscale MRI Multiple single-molecule fluorescence microscopy Nanoparticle characterization by Halo LM10 technology Nanoscale scanning electron microscopy Near-field scanning optical microscopy Optical Imaging with a Silver Superlens Partial wave spectroscopy Photoactivated localization microscopy Scanning probe microscopy Super-resolution nanoscopy for in vivo cell imaging Ultra-nanocrystalline diamond Visualizing atoms with high-resolution transmission electron microscopy

Safety issues of nanoparticles are a concern and studies are in progress to resolve some of these. Risk evaluation presents challenges due to a lack of data, the complexity of nanomaterials, measurement difficulties, and undeveloped hazard-assessment frameworks. The success of nanoparticles is due to their small size, which enables us to get them into parts of the body where large particles cannot enter because of their size, and this is a cause for fear that nanoparticles may cross some of the barriers such as the blood–brain barrier and lodge in the brain. We are exposed to spontaneously generated nanoparticles in the atmosphere and many of these are excreted from the body. Although some of the nanoparticles are nontoxic, others are toxic; the current trend is to use nontoxic degradable polymer nanoparticles.

610

Nanobiotechnology

Figure 1

Relationship of nanobiotechnology to other technologies for the development of personalized medicine.

1.45.2

Nanoparticles

Nanoparticles can be made of different materials and each contains tens to thousands of atoms and exists in a realm that straddles the quantum and the Newtonian. At those sizes, every particle has new properties that change depending on its size. As matter is shrunk to the nanoscale, electronic and other properties change radically. Nanoparticles may contain unusual forms of structural disorder that can significantly modify materials properties and thus cannot solely be considered as small pieces of bulk material. Two nanoparticles, both made of pure gold, can exhibit markedly different behavior different melting temperature, different electrical conductivity, and different color if one is larger than the other. Some applications of nanoparticles take advantage of the fact that more surface area is exposed when material is broken down to smaller sizes. For magnetic nanoparticles, the lack of blemishes produces magnetic fields remarkably strong considering the size of the particles. Nanoparticles are also so small that, in most of them, the atoms line up in perfect crystals without a single blemish. Selected nanoparticles with their applications are listed in Table 3. Most of these have applications in drug delivery.

1.45.3

Role of Nanobiotechnology in Molecular Diagnostics

Nanomolecular diagnostics is the use of nanobiotechnology in molecular diagnostics, and can be termed ‘nanodiagnostics’. Numerous nanodevices and nanosystems for sequencing single molecules of DNA are feasible. Given the inherent nanoscale of receptors, pores, and other functional components of living cells, the detailed monitoring and analysis of these components will be made possible by the development of a new class of nanoscale probes. Biological tests measuring the presence or activity of selected substances become quicker, more sensitive, and more flexible when certain nanoscale particles are put to work as tags or labels. Nanotechnology will improve the sensitivity and integration of analytical methods to yield a more coherent evaluation of life processes. The applications of nanobiotechnology in molecular diagnostics have been reviewed elsewhere.4 Applications of nanobiotechnologies in molecular diagnostics are listed in Table 4.

1.45.3.1

Nanoparticles for Molecular Diagnostics

Gold nanoparticles assemble onto a sensor surface only in the presence of a complementary target. If a patterned sensor surface of multiple DNA strands is used, the technique can detect millions of different DNA sequences simultaneously. The current nonoptimized detection limit of this method is 20 fmol. Gold nanoparticles were found to be particularly good labels for sensors because a variety of analytical techniques can be used to detect them, including optical absorption, fluorescence, Raman scattering, atomic and magnetic force, and electrical conductivity. There is considerable interest in the use of quantum dots (QDs) as inorganic fluorophores, because they offer significant advantages over conventionally used fluorescent markers. For example, QDs have broad excitation spectra – from ultraviolet to red – that can be tuned depending on their size and composition. At the same time, QDs have narrow emission spectra, making it possible to

Nanobiotechnology Table 3

611

Nanoparticles

Structure

Size

Description/applications

Carbon magnetic nanoparticles Ceramic nanoparticles

40–50 nm  35 nm

Dendrimers Gold nanoparticles N-(2-hydroxyl) propyl-3-trimethyl ammonium chitosan chloride (HTCC) Micelle/nanopill

1–20 nm 2–4 nm 110–180 nm

Low-density lipoproteins Nanoemulsions Nanoliposomes

20–25 nm 20–25 nm 25–50 nm

Nanoparticle composites

 40 nm

Nanospheres Nanosponges

50–500 nm 10 nm

Nanostructured organogels

 50 nm

Nanotubes

Single wall 1–2 nm Multiwall 20–60 nm 25–100 nm 50–200 nm 2–10 nm 10–100 nm

For drug delivery and targeted cell destruction Accumulate exclusively in the tumor tissue and allow the drug to act as a sensitizer for photodynamic therapy without being released Holding therapeutic substances such as DNA in their cavities Enable externally controlled drug release Ability to overcome biological barriers and facilitate the delivery of complex drugs such as insulin, vaccines, plasmid DNA and genes. Made from two polymer molecules – one hydrophilic and the other hydrophobic – that self-assemble into a sphere called a ‘micelle’ that can deliver drugs to specific structures within the cell Drugs solubilized in the lipid core or attached to the surface Drugs in oil and/or liquid phases to improve absorption Incorporate fullerenes to deliver drugs that are not water-soluble and tend to have large molecules Attached to guiding molecules such as monoclonal antibodies for targeted drug delivery Hollow ceramic nanospheres created by ultrasound A long, linear molecule scrunched into a sphere 10 nm in diameter with a large number of surface sites for drug molecules attachment Mixture of olive oil, and liquid solvents, and adding a simple enzyme to chemically activate a sugar; used to encapsulate drugs Resemble tiny drinking straws and might offer advantages over spherical nanoparticles for some applications Spheres containing the drugs in lipid bilayer Enclosing drugs Combine imaging with therapeutics As drug carriers for intravenous injections to evade the reticuloendothelial system as well as to penetrate the very small capillaries within the body tissues improving distribution

Nanovesicles Polymer nanocapsules Quantum dots Superparamagnetic iron-oxide nanoparticles

25–200 nm

resolve the emissions of different nanoparticles simultaneously and with minimal overlap. Finally, QDs are highly resistant to degradation, and their fluorescence is remarkably stable. QDs have been used as possible alternatives to the dyes for tagging viruses and cancer cells. A major challenge is that QDs have an oily surface, whereas the cellular environment is water based. Attempts are being made to modify the surface chemistry of QDs so that they interact with hydrophilic molecules such as proteins and DNA. The current goal is to develop QDs that can target a disease site and illuminate it. This can, in future, lead to an integrated system that will also use the QDs to diagnose as well to deliver drug therapies to the disease site.

1.45.3.2

Nanoparticles as Biolabels

Nanoparticles usually form the core in nanobiomaterials. However, in order to interact with a biological target, a biological or molecular coating or layer acting as an interface needs to be attached to the nanoparticle. Coatings that make the nanoparticles biocompatible include antibodies, biopolymers, or monolayers of small molecules. Potential benefits of using nanoparticles and nanodevices include an expanded range of label multiplexing. Different types of fluorescent nanoparticles and other nanostructures have been promoted as alternatives for the fluorescent organic dyes that are traditionally used in biotechnology. These include QDs and dye-doped polymer and silica nanoparticles.

1.45.3.3

Paramagnetic and Superparamagnetic Nanoparticles

Paramagnetic particles are important tools for cell sorting, protein separation, and single-molecule measurements. Superparamagnetic iron oxide nanoparticles (SPIONs) with appropriate surface chemistry have been widely used experimentally for numerous in vivo applications such as magnetic resonance imaging (MRI) contrast enhancement, tissue repair, immunoassay, detoxification of biological fluids, hyperthermia, drug delivery, cell separation, etc. These applications require that these nanoparticles have high magnetization values and a size smaller than 100 nm with overall narrow particle size distribution, so that the particles have uniform physical and chemical properties. In addition, these applications need special surface coating of the magnetic particles, which not only has to be nontoxic and biocompatible but also allows a targetable delivery with particle localization in a specific area. The nature of surface coatings of the nanoparticles not only determines the overall size of the colloid but also plays a significant role in biokinetics and biodistribution of nanoparticles in the body. Magnetic nanoparticles can bind to drugs, proteins, enzymes,

612

Nanobiotechnology Table 4

Applications of nanotechnologies in molecular diagnostics

Nanotechnology to improve polymerase chain reaction (PCR) Nanotechnology-based biochips and microarrays Nanotechnology on a chip: nanobiochips Nanotechnology in microfluidics: nanoarrays Protein nanobiochips/nanoarrays Nanotechnology-based cytogenetics Study of chromosomes by atomic force microscopy (AFM) Quantum dot fluorescent in situ hybridization (FISH) Nanoscale single molecule identification Nanotechnology-based DNA sequencing Nanopore-based DNA sequencing Nanoparticle-based DNA sequencing Nanoparticle technologies Carbon nanotubes Dendrimers Gold particles Nanobarcodes Magnetic nanoparticles Quantum dot technology Nanoparticle probes Nanowires Nanotechnology-based biomarkers Nanoparticles for discovering biomarkers Nanoproteomics and biomarkers Nanosensors for measuring biomarkers in blood Exosome-based molecular diagnostics DNA nanomachines for molecular diagnostics Nanoparticle-based immunoassays DNA-protein and nanoparticle conjugates Nanobiosensors Cantilever arrays Living spores as nanodetectors Quartz nanobalance DNA sensor Nanosensor glucose monitor Photostimulated luminescence in nanoparticles Optical biosensors: surface plasmon resonance technology Imaging applications of nanoparticles Computer tomography image enhancement by nanoparticles Functionalized carbon nanotubes as ultrasound contrast agents Nanoparticles as contrast-enhancing agents for magnetic resonance imaging

antibodies, or nucleotides and can be directed to an organ, tissue, or tumor using an external magnetic field, or can be heated in alternating magnetic fields for use in hyperthermia. The magnetic labeling of cells provides the ability to monitor their temporal spatial migration in vivo by MRI. Various methods have been used to magnetically label cells using SPIONs. The magnetic tagging of stem cells and other mammalian cells has the potential for guiding future cell-based therapies in humans and for the evaluation of cellular-based treatment effects in disease models.

1.45.3.4

Role of Nanobiotechnology in Discovery of Biomarkers

A biomarker is a characteristic that can be objectively measured and evaluated as an indicator of a physiological as well as a pathological process, or a pharmacological response to a therapeutic intervention. Examples of classical biomarkers include measurable alterations in blood pressure, blood lactate levels following exercise, and blood glucose in diabetes mellitus. Any specific molecular alteration of cellular DNA, RNA, metabolite, or protein levels can be referred to as a molecular biomarker.5 From a practical point of view, the biomarker would specifically and sensitively reflect a disease state and could be used for diagnosis as well as for disease monitoring during and following therapy. Currently available molecular diagnostic technologies have been used to detect biomarkers of various diseases such as cancer, metabolic disorders, infections, and diseases of the central nervous system. Nanotechnology has further refined the detection of biomarkers. Some biomarkers also form the basis of innovative molecular diagnostic tests.

Nanobiotechnology

613

The physicochemical characteristics and high surface areas of nanoparticles make them ideal candidates for developing biomarker-harvesting platforms. Given the variety of nanoparticle technologies that are available, it is feasible to tailor nanoparticle surfaces to selectively bind a subset of biomarkers and sequester them for later study using high-sensitivity proteomic tests. Biomarker harvesting is an underutilized application of nanoparticle technology and is likely to undergo substantial growth. Functional polymer-coated nanoparticles can be used for the rapid detection of biomarkers and DNA separation. A magnetic nanosensor, which is up to 1000 times more sensitive than any device now in clinical use, can accurately detect protein biomarkers over a range of concentrations 3 times broader than any existing method.6 The nanobiosensor chip can also search for up to 64 different proteins simultaneously and has been shown to be effective in the early detection of tumors in mice, suggesting that it may open the door to significantly earlier detection of even the most elusive cancers in humans. The magnetic nanobiosensor can successfully detect cancerous tumors in mice when levels of cancer-associated proteins are still well below concentrations detectable through the current standard method, enzyme-linked immunosorbent assay (ELISA). The sensor can also be used to detect biomarkers of diseases other than cancer.

1.45.3.5

Nanobiotechnology and Cytogenetics

The term ‘cytogenetics’ has been classically used for studies of the cellular aspects of heredity and was used mainly to describe the chromosome structure and identify abnormalities related to disease. Cytogenetics can now be included under the term ‘cytomics’, which means that the structural and functional information is obtained by the molecular cell phenotype analysis of tissues, organs, and organisms at the single-cell level.7 Molecules in the cell are organized in nanometer-scale dimensions. Visualizing the dynamic change in these molecules and studying the function of cells is one of the challenges in nanobiology. A single molecule is the ultimate nanostructure. Single-molecule microscopy and spectroscopy are some of the techniques used to study single molecules.

1.45.4

Pharmaceutical Applications of Nanobiotechnology

Among the new technologies, nanobiotechnology has evoked considerable interest for application in the pharmaceutical industry. Important applications of nanobiotechnology are in the areas of drug discovery, drug development, and drug delivery, and these are collectively referred to as nanopharmaceuticals.

1.45.4.1

Role of Nanobiotechnology in Drug Discovery and Development

Nanobiotechnology, particularly the use of nanoparticles, has made significant contributions to drug discovery and development.8 The multivalent attachment of small molecules to nanoparticles can increase specific binding affinity and reveal new biological properties of such nanomaterials. Multivalent drug design has yielded antiviral and anti-inflammatory agents several orders of magnitude more potent than monovalent agents. In addition to the use of nanobiotechnology for drug discovery, some drugs are being developed from nanomaterials. Well-known examples of these are dendrimers, fullerenes, and nanobodies. Dendrimer conjugation with low-molecular-weight drugs has been of increasing interest recently for improving pharmacokinetics, targeting drugs to specific sites, and facilitating cellular uptake. A key attribute of the fullerene molecules is their numerous points of attachment, allowing for precise grafting of active chemical groups in three-dimensional (3D) orientations. This attribute, the hallmark of rational drug design, allows for positional control in matching fullerene compounds to biological targets. Fullerene antioxidants bind and inactivate multiple circulating intracellular free radicals, giving them unusual power to stop free radical injury and to halt the progression of diseases caused by excess free radical production. Nanobodies, derived from naturally occurring single-chain antibodies, are the smallest fragments of naturally occurring heavy-chain antibodies that have evolved to be fully functional in the absence of a light chain. Like conventional antibodies, nanobodies show high target specificity and low inherent toxicity; however, like small molecule drugs they can inhibit enzymes and can access receptor clefts. An increasing use of nanobiotechnology by the pharmaceutical and biotechnology industries is anticipated. Nanotechnology will be applied at all stages of drug development from formulations for optimal delivery to diagnostic applications in clinical trials. In the future, it may be possible to fully model an individual cell’s structure and function by computers connected to nanobiotechnology systems. Such a detailed virtual representation of cell functions might enable scientists to develop novel drugs with unprecedented speed and precision without any experiments in living animals.

1.45.4.2

Nanobiotechnology-Based Drug Delivery

One of the major problems with drugs is solubility, which is an essential factor for drug effectiveness, independent of the route of administration. It is also a major challenge for pharmaceutical companies developing new pharmaceutical products as nearly 50% of the new chemically based drugs are insoluble, or poorly soluble, in water. Many, otherwise promising, compounds never reach the market. Others reach the market but in a suboptimal formulation, possibly providing low or unpredictable bioavailability, or

614

Nanobiotechnology

posing an increased risk of side effects. Enhanced solubility technology can be used to reformulate such drugs and increase their commercial potential. Nanobiotechnology provides the following solutions to the problems of drug delivery:

• • • • • • •

the particle size is reduced to the nanometer-size range to increase the surface area, thereby increasing the rate of dissolution; improving solubilization of the drug; using noninvasive routes of administration eliminates the need for administration of drugs by injection; development of novel nanoparticle formulations with improved stabilities and shelf lives; development of nanoparticle formulations for improved absorption of insoluble compounds and macromolecules enables improved bioavailability and release rates, potentially reducing the amount of dose required and increasing safety through reduced side effects; manufacture of nanoparticle formulations with controlled particle sizes, morphology, and surface properties would be more effective and less expensive than other technologies; and nanoparticle formulations that can provide sustained-release profiles up to 24 h can improve patient compliance with drug regimens.

The direct coupling of drugs to targeting ligand restricts the coupling capacity to a few drug molecules, but the coupling of drug– carrier nanosystems to ligands allows the import of thousands of drug molecules by means of one receptor-targeted ligand. Nanosystems offer opportunities to achieve drug targeting with newly discovered disease-specific targets. Biodegradable polymer nanoparticles have been used frequently as drug delivery vehicles due to biodegradability, better encapsulation, controlled release, and low toxicity. Various nanoparticulate systems, general synthesis and encapsulation process, and controlled release and improvement of the therapeutic effect of nanoencapsulated drugs have been reviewed.9

1.45.4.3

Nanobiotechnology and Drug Delivery Devices

There is a need to improve devices introduced into the human body. Some drug delivery devices are implanted in the body for the release of therapeutic substances. The lining of these devices can be improved by nanotechnology. For example, implants can be coated by nanolayers of polymers. DNA-containing polymeric nanocontainers or nanotraps can preserve the full activity of an encapsulated enzyme against hostile outside environments and the release can be controlled according to demand. Nanocontainers with asymmetric membranes can enable direct protein insertion and chemical modification. Nanoporous materials with ordered and controlled pore structures, high surface area, and pore volume are particularly suited for implantable drug delivery systems. Considerable progress has been made in electrochemically engineered nanopores/nanotube materials such as nanoporous alumina and nanotubular titanium.10 Nanoporous devices can be used for cell encapsulation in hormonal therapy. Biosensors mounted on these devices can be used for noninvasive signal detection. An implanted titanium drug delivery device with a silicon nanopore membrane can control the release by diffusion of an encapsulated drug at a nearly constant rate. It can achieve nearly zero-order drug kinetics over long periods of time. It would be suitable for the delivery of protein and peptide drugs and it avoids the poor pharmacokinetics associated with injections, providing an optimized method of delivery for these compounds. The drug can be formulated as a dry powder or a concentrated suspension, and maintains its stability. The drug is protected from the immunological reaction of the body by the nanopore membrane, which releases the drug but excludes entry of unwanted cells.

1.45.5

Role of Nanobiotechnology in Biological Therapies

Biological therapies are playing an important role in current therapeutics and many are in development with a promising future. These include cell therapy, gene therapy, vaccines, RNA interference, and antisense therapeutics. An important issue in biological therapies is delivery. Nanoparticles play an important role in the delivery of biological therapies.

1.45.5.1

Role of Nanobiotechnology in Cell Therapy

Cell therapy is the prevention or treatment of human disease by the administration of cells that have been selected, multiplied, and pharmacologically treated or altered outside the body (ex vivo). The scope of cell therapy can be broadened to include methods, pharmacological as well as nonpharmacological, to modify the function of intrinsic cells of the body for therapeutic purposes. Most of the interest in cell therapy centers around stem cells. Examples of applications of nanobiotechnology in cell therapy and related areas of tissue engineering include the following:

• • • •

QDs are semiconductor nanocrystals that serve as promising alternatives to organic dyes for cell labeling; the use of SPIO nanoparticles with MRI to monitor the cells introduced therapeutically into the body; self-assembling peptide nanofibers are designed for prolonged delivery of insulin-like growth factor 1 to the myocardium for improving its function following myocardial infarction; bone formation from mesenchymal stem cells is facilitated on a novel nanofibrous scaffold;

Nanobiotechnology

• •

615

carbon nanotubes are used to aid stem cell therapy of neurological disorders; and nanoparticles, when injected into the myocardium after infarction, promote the formation of new blood vessels and may play a role in the recovery of myocardial function following this procedure.

Tracking of stem cells labeled with nanoparticles offers significant advantages over other cell-labeling technologies in development. Laboratory tests showed that the cells retained their usual surface markers, and that they were still functional after the labeling process. The labeled cells have been shown to migrate to and incorporate into blood vessels that form around tumors in experimental animals. These could be translated into clinical applications to enable physicians to directly track cells used in medical treatments using unique signatures from the ingested nanoparticle beacons. They could prove useful for monitoring tumors, and diagnosing as well as treating cardiovascular problems.

1.45.5.2

Role of Nanobiotechnology in Gene Therapy

Gene therapy is defined as the transfer of defined genetic material to specific target cells of a patient for the ultimate purpose of preventing or altering a disease. An important component of gene therapy is the carrier or the delivery vehicle (vector) to deliver the healthy gene to a patient’s cells. Vectors are usually viral, but several nonviral techniques are being used as well. Genes and DNA are now being introduced without the use of vectors, and various techniques are being used to modify the function of genes in vivo without gene transfer. Nanoparticles and other nanostructures can be used for nonviral gene delivery. Degradable nanoparticles are the only nonviral vectors that can provide a targeted intracellular delivery with controlled release properties. Furthermore, the potential advantage of degradable nanoparticles over their nondegradable counterparts is the reduced toxicity and the avoidance of accumulation within the target tissue after repeated administration. Biocompatible, inorganic nanoparticles of carbonate apatite have unique features that are useful for delivery, as well as for the expression of genetic material in mammalian cells. Carbonate apatite nanoparticles adsorb DNA to carry it effectively across the cell membrane. They also possess a high dissolution rate in endosomal acidic pH, leading to the rapid release of the bound DNA for a subsequent high level of protein expression. Carbonate apatite is a natural component of the body, and is usually found in the hard tissues, such as bone and teeth. Moreover, because of their nanosize dimensions and sensitivity to low pH, particles of carbonate apatite are quickly degraded when taken up by cells in their acidic vesicles, without any indication of toxicity. Apatite nanoparticles are promising candidates for nonviral gene delivery and are superior to polymer- or lipid-based systems that are generally nonbiodegradable and inefficient. Polymer beacons have been developed that enable the delivery of nucleic acids to be visualized at the nanoscale. These delivery beacons effectively bind and compact plasmid DNA (pDNA) into nanoparticles and protect nucleic acids from nuclease damage. Because of their versatility, these delivery beacons possess remarkable potential for tracking and understanding nucleic acid transfer in vitro, and have promise as in vivo vectors for gene therapy and agents for combining diagnostics and therapeutics. Novel multifunctional DNA carriers have been described, which self-assemble with DNA to form structured nanoparticles that possess virus-like functions for cellular trafficking. The nanoparticles are internalized in a cell-specific fashion and subsequently exit the endosome into the cytoplasm. The nanoparticles interact with cellular nuclear transport proteins and are actively trafficked into the cell nucleus of nondividing cells, resulting in three- to fourfold higher reporter gene expression in cells, in addition to lower cytotoxicity than lipid and polyethyleneimine vectors.

1.45.5.3

Role of Nanobiotechnology in Vaccines

Nanobiotechnology will play an important role in vaccinology, referred to as nanovaccinology. Use of nanoparticles in vaccine formulations enables not only improved antigen stability and immunogenicity but also targeted delivery and slow release. Several nanoparticle vaccines varying in composition, size, shape, and surface properties have been approved for human use and the number of candidates is increasing. Nanoparticles can operate both as a delivery system to enhance antigen processing and as an immunostimulatory adjuvant to induce and amplify protective immunity because of their ability to activate the inflammasome and induce the maturation of interleukin 1b.11 Nanoparticles can be excellent adjuvants due to their biocompatibility and their physicochemical properties (e.g., size, shape, and surface charge), which can be tailored to obtain different immunological effects. In contrast to alum, which is conventionally used as an adjuvant, antigens covalently coupled to nanobeads, measuring 40 nm in diameter, induce substantial cell-mediated responses along with moderate humoral responses. No adverse reactions have been noted at the site of immunization in experimental animals. Thus, nanobead adjuvants in veterinary species may be useful for the induction of immunity to viral pathogens, where a cell-mediated response is required. These vaccines have potential usefulness for intracellular pathogens in humans. Most adjuvants only stimulate antibodies against a disease. The nanobead technology gives the immune system a further boost, also producing T-cells, which are needed to eliminate viruses or cancer. The size of 40 nm is critical, as it is a size like that of many viruses, where the nanobeads are taken up abundantly by the immune system and tricked into producing high levels of many types of T-cells. A targeted synthetic vaccine platform creates fully integrated synthetic nanoparticle vaccines engineered to mimic the properties of natural pathogens to elicit a maximal immune response. Antigens and adjuvants are delivered within the same biodegradable nanoparticle, directly to antigen-presenting cells. This approach maximizes the immune response, while minimizing undesirable off-target effects.

616

Nanobiotechnology

Nanotechnology offers the possibility of important advances in the design of new vaccines by facilitating the targeting to the cells of the mononuclear phagocyte system, which are the major antigen-presenting cells (APCs) because of the nanoparticulate nature. It is possible to precisely target the selected APCs by inserting specific molecules on the nanoparticle surface (ligands, receptor). Thus, nanoparticles will facilitate personalized vaccinology, i.e., the design of vaccination strategies to obtain maximal efficacy and minimal collateral effects, based on the individual immune status.

1.45.5.4

Role of Nanobiotechnology in Antisense Therapy

Antisense molecules are synthetic segments of DNA or RNA, designed to mirror specific messenger RNA (mRNA) sequences and block protein production. One way to target the genetic material is to block mRNA by using ‘antisense DNA’, which prevents the message from ever becoming a protein. The use of antisense drugs to block abnormal disease-related proteins is referred to as antisense therapeutics. Synthetic short segments of DNA or RNA are referred to as oligonucleotides (ODNs). Whereas typical drugs target the proteins, it is possible, through antisense gene therapy, to target the genetic material itself before it is ever made into copies of harmful proteins. Antisense drugs have the promise to be more effective than conventional drugs, but one of the problems with antisense therapy is delivery. The efficacy of antisense ODNs is limited by the poor stability of the natural oligomers and the low efficacy of their cellular uptake. Nanotechnology has been used to improve this situation. Gold nanoparticle–ODN complexes have been used as antisense intracellular gene regulation agents to control protein expression in cells. Once inside cells, the DNA-modified nanoparticles act as mRNA ‘sponges’ that bind to their targets and prevent them from being converted into proteins. The advantages of attaching multiple strands of antisense DNA to the surface of a gold nanoparticle over conventional antisense ODNs are:

• • • •

the DNA becomes more stable and can bind to the target mRNA more effectively than DNA that is not attached to a nanoparticle; they are less susceptible to degradation by nuclease activity; they exhibit greater than 99% cellular uptake; and they can introduce ODNs at a higher effective concentration than conventional transfection agents, and are nontoxic to the cells under the conditions studied.

Polymethacrylate nanoparticles appeared to be a promising vehicle for the delivery of antisense ODNs. The uptake of ODN is significantly increased when loaded by nanoparticles, which also depends on the nanoparticle concentration. A slight cytotoxicity can occur when high doses of nanoparticles are used.

1.45.5.5

Role of Nanobiotechnology in RNA Interference

Therapeutics designed to engage RNA interference (RNAi) pathways are now well recognized. Small interfering RNAs (siRNAs), which are approximately 21-base-pair double-stranded RNAs, can elicit RNAi. siRNA therapeutics are hindered by poor intracellular uptake, limited blood stability, and nonspecific immune stimulation. Nanotechnology has been applied to address the delivery problems of siRNA. Chitosan-coated poly(isobutylcyanoacrylate) nanoparticles have been used to deliver siRNA. The reduction in size of the chitosan-decorated nanoparticles has enabled the protection of siRNA from in vivo degradation, leading to significant tumor growth inhibition after intratumoral administration. Stable nucleic acid–lipid particles (SNALPs) are specialized lipid nanoparticles that fully encapsulate and systemically deliver a variety of nucleic acid molecules such as siRNA, aptamers, and pDNA. Because the SNALPs are small ( 100 nm), have a uniform size and low surface charge, are stable, and do not aggregate, they remain intact in circulation for many hours. These features of SNALP allow the particles to accumulate at target sites. This technology utilizes a mechanism referred to as the ‘enhanced permeability and retention effect’, which occurs because these nucleic acid-containing particles have a long circulation time in the blood, resulting in accumulation at sites of vascular leak such as those found at sites of tumor cell growth, infection, inflammation, and in the normal liver. Highly branched, dendritic polymers including poly(amidoamine) (PAMAM) can bind to DNA and RNA molecules and mediate modest cellular delivery of these nucleic acids. PAMAM dendrimers have been evaluated for the successful delivery of siRNA or antisense molecules. The next generation of dendrimers for siRNA delivery will need to integrate charge reduction of dendrimers without compromising their endosomal buffering capacity. An important consideration in using RNAi for identifying genotype/phenotype correlations is the uniformity of gene silencing within a cell population. Variations in transfection efficiency, delivery-induced cytotoxicity, and off-target effects at high siRNA concentrations can confound the interpretation of functional studies. To address this problem, a novel method of monitoring siRNA delivery has been developed that combines unmodified siRNA with semiconductor QDs to form multicolor biological probes. Co-transfection of siRNA with QDs using standard transfection techniques, thus leveraging the photostable fluorescent nanoparticles to track delivery of nucleic acid, sorts cells by degree of transfection and purifies homogenously silenced subpopulations. Compared to alternative RNAi tracking methods (co-delivery of reporter plasmids and end labeling the siRNA), QDs exhibit superior photostability and tunable optical properties for an extensive selection of nonoverlapping colors. A phase I clinical trial involved the systemic administration of siRNA to patients with solid cancers using a targeted nanoparticle delivery system.12 Tumor biopsies from melanoma patients obtained after treatment showed the presence of intracellular nanoparticles localized in tumor cells in amounts corresponding to the dose levels of the nanoparticles administered. The presence of an mRNA fragment demonstrates that siRNA-mediated mRNA cleavage occurs specifically at the site predicted for an RNAi mechanism.

Nanobiotechnology Table 5

617

Areas of applications of clinical nanomedicine

Diagnostics Development of new nanotechnology-based assays Extending limits of detection by refining currently available molecular diagnostic technologies Molecular imaging using nanocontrast materials Nanobiosensors Nanoendoscopy Ultrasound imaging using nanoparticles Application in various specialties of medicine Nanocardiology Nanodermatology Nanodentistry Nanogerontology Nanohematology Nanoimmunology Nanomicrobiology Nanonephrology Nanoneurology Nanooncology Tissue engineering and regenerative medicine Nanofiber scaffolds with stem cell transplants Nanomaterial-based regeneration of tissues Nanomaterials for combining tissue engineering and drug delivery Transplantation medicine Exosomes from donor dendritic cells for drug-free organ transplants Nanosurgery Minimally invasive surgery: miniaturized nanosensors implanted in catheters Nanosurgery by integration of nanoparticles and external energy, e.g., lasers Nanorobotic treatments Vascular surgery by nanorobots introduced into the vascular system Nanorobots for detection and destruction of cancer Combination of diagnostics with therapy Development of personalized medicine

1.45.6

Clinical Nanomedicine

Application areas of clinical nanomedicine are listed in Table 5 and include practically all diseases. The prefix ‘nano’ can be added to each specialty, for example, nanoneurology. The most important applications are in cancer, neurological disorders, and cardiovascular disorders, which will be discussed further.

1.45.7

Nanooncology

Nanooncology includes both the diagnosis and treatment of cancer using nanotechnology.13,14 Nanobiotechnology plays an important role in drug delivery for cancer.

1.45.7.1

Nanotechnology-Based Drug Delivery in Cancer

Two nanotechnology-based products are already approved for the treatment of cancer: Doxil (a liposome preparation of doxorubicin) and Abraxane (paclitaxel in nanoparticle formulation). Approximately 150 drugs in development for cancer are based on nanotechnology. Drug delivery in cancer is important for optimizing the effect of drugs and reducing toxic side effects. Several nanobiotechnologies, mostly based on nanoparticles, have been used to facilitate drug delivery in cancer. A classification of the nanotechnologies for drug delivery in cancer is shown in Table 6.

1.45.7.2

Combination of Diagnosis and Therapy in Cancer

Nanoparticle-based diagnostics and therapeutics hold great promise because multiple functions can be built into the particles. One such function is an ability to home to specific sites in the body.

618

Nanobiotechnology Table 6

Classification of nanobiotechnology approaches to drug delivery in cancer

Nanoparticles Nanoparticle formulations of anticancer drugs, for example, paclitaxel Exosomes for cancer drug delivery Nanoencapsulation and enclosure of anticancer drugs Enclosing drugs in lipid nanocapsules Encapsulating drugs in hydrogel nanoparticles Micelles for drug delivery in cancer Targeted delivery of anticancer therapy Targeted drug delivery with nanoparticles PEGylated nanoliposomal formulation Folate-linked nanoparticles Carbon magnetic nanoparticles for targeted drug delivery in cancer Targeted drug delivery with nanoparticle–aptamer bioconjugates Nanodroplets for site-specific cancer treatment Lipid-based nanocarriers Targeted antiangiogenic therapy using nanoparticles Nanoparticles for delivery of drugs to brain tumors Combination of nanoparticles with radiotherapy Combination with boron neutron capture therapy Nanoengineered silicon for brachytherapy Combination with physical modalities of cancer therapy Combination with laser ablation of tumors Combination with photodynamic therapy Combination with thermal ablation Combination with ultrasound Nanoparticle-mediated gene therapy p53 gene therapy of cancer Immunolipoplex for delivery of p53 gene Intravenous delivery of FUS1 gene Strategies combining diagnostics and therapeutics Nanoshells as adjuncts to thermal tumor ablation Perfluorocarbon nanoparticles Nanocomposite devices

Photothermal therapy is based on the enhancement of electromagnetic radiation by noble-metal nanoparticles due to strong electric fields at the surface. The nanoparticles also absorb the laser light more easily, so that the coated malignant cells only require half the laser energy to be killed compared to the benign cells. This makes it relatively easy to ensure that only the malignant cells are being destroyed. These unique properties provide the potential of designing novel optically active reagents for simultaneous molecular imaging and photothermal cancer therapy. Gold nanorods with suitable aspect ratios (length divided by width) can absorb and scatter strongly in the near-infrared (NIR) region (650–900 nm). Changing the spheres into rods lowers the frequency to which the nanoparticles respond from the visible light spectrum used by the nanospheres to the NIR spectrum. As these lasers can penetrate deeper under the skin than lasers in the visible spectrum, they can reach tumors that are inaccessible to visible lasers. Tumor-targeting dendrimers contain both an imaging agent and a therapeutic agent. A DNA-linked dendrimer platform enables the delivery of drugs, genetic materials, and imaging agents to cancer cells, offering the potential for developing combinatorial therapeutics. A dendrimer linked to a fluorescent imaging agent and paclitaxel can identify tumor cells and kill them simultaneously. The completeness of tumor removal during surgery depends on the surgeon’s ability to differentiate tumor from normal tissue. A method has been developed to visualize tumors during surgery using activatable cell-penetrating peptides, in which the fluorescently labeled, polycationic cell-penetrating peptide is coupled via a cleavable linker to a neutralizing peptide.17 The application of this technique to experimental tumors in animals showed that the long-term tumor-free survival rate was higher for mice that underwent surgery using the tagged nanoparticles than those that underwent the same surgery with traditional imaging methods. The basic rationale for using nanobiotechnology in oncology is that nanoparticles have optical, magnetic, or structural properties that are not available in larger molecules or bulk solids. When linked with tumor-targeting ligands such as monoclonal antibodies (MAbs), peptides, or small molecules, these nanoparticles can be used to target tumor antigens (biomarkers) as well as tumor vasculatures with high affinity and specificity. Nanoparticles ranging in size from 5 to 100 nm have large surface areas and functional groups for conjugating to multiple diagnostic and therapeutic anticancer agents. Recent advances have led to bioaffinity nanoparticle probes for molecular and cellular imaging, targeted nanoparticle drugs for cancer therapy, and integrated nanodevices for early cancer detection and screening. These developments have provided opportunities for personalized oncology in which biomarkers are used to diagnose and treat cancer based on the molecular profiles of individual patients.

Nanobiotechnology

1.45.8

619

Nanoneurology

Nanoneurology is the application of nanobiotechnology in neurology. Nanobiotechnology provides methods for studying the pathomechanism of neurological disorders as well as their diagnosis and treatment.15,16

1.45.8.1

Role of Nanobiotechnology in Diagnosis of Neurological Disorders

Activated macrophages, acting in concert with other immune-competent cells, are an index of inflammatory/immune reaction in central nervous system (CNS) disorders such as multiple sclerosis, ischemic stroke lesions, and tumors. The MRI detection of brain macrophages defines precise spatial and temporal patterns of macrophage involvement that helps to characterize individual neurological disorders. Macrophage tracking by MRI with iron-oxide nanoparticles has been developed during the last decade for numerous diseases of the CNS. Experimental studies on animal models were confirmed by clinical applications of MRI technology on brain macrophages. This approach is being explored as an in vivo biomarker for the clinical diagnosis of cerebral lesion activity, in experimental models for the prognosis of disease development, and to determine the efficacy of immunomodulatory treatments under clinical evaluation. Comparative brain imaging follow-up studies of blood–brain barrier (BBB) leakage by MRI with gadolinium chelates, microglia activation by positron emission tomography (PET) with radiotracer ligand PK11195, and MRI detection of macrophage infiltration provide more precise information about the pathophysiological cascade of inflammatory events in cerebral diseases. Such multimodal characterization of the inflammatory events should help in the monitoring of patients, in defining precise time intervals for therapeutic interventions, and in developing and evaluating new therapeutic strategies. Delivery of the therapeutics to the brain is one of the important issues in neurotherapeutics. Currently, most of the strategies are directed at overcoming the BBB. Very small nanoparticles may just pass through the BBB, but this uncontrolled passage is not desirable. Most of the strategies for the passage of drugs across the BBB can be enhanced by nanotechnology and some examples include the following:

• • • • • • •

Nanoparticles open the tight junctions between endothelial cells and enable the drug to penetrate the BBB either in free form or together with the nanocarrier. Nanoparticles are transcytosed through the endothelial cell layer and allow the direct transport of their therapeutic cargo. Nanoparticles are endocytosed by endothelial cells and release the drug inside the cell, as a precursor step to the transport of active ingredients, which occurs by exocytosis at the abluminal side of the endothelium. Nanoparticles, which combine an increased retention at the brain capillaries with adsorption onto the capillary walls, improve delivery to the brain by creating a concentration gradient that promotes transport across the endothelial cell layer. Drug transport is enhanced by the solubilization of the endothelial cell membrane lipids by a surfactant, which leads to membrane fluidization (surfactant effect). Coating agents (such as polysorbates) inhibit the transmembrane efflux systems, that is, P-glycoprotein. Nanoparticles induce local toxic effects at the brain vasculature, which leads to a limited permeabilization of the brain endothelial cells.

BBB represents an insurmountable obstacle for many drugs, including antibiotics, anticancer agents, and a variety of CNS-active drugs, especially neuropeptides. One of the possibilities to overcome this barrier is drug delivery to the brain using nanoparticles. The mechanism of the nanoparticle-mediated transport of drugs across the BBB at present is not fully elucidated. The most likely mechanism is endocytosis by the endothelial cells lining the brain blood capillaries. Other processes such as tight junction modulation or P-glycoprotein inhibition also may occur. Moreover, these mechanisms may run in parallel or may be cooperative, thus enabling drug delivery to the brain.

1.45.8.2

Nanoparticle-Based Drug Delivery to the Brain

The use of nanoparticles to deliver drugs to the brain across the BBB may provide a significant advantage to current strategies. The primary advantage of nanoparticle carrier technology is that nanoparticles mask the BBB-limiting characteristics of the therapeutic drug molecule. Furthermore, this system may slow down drug release in the brain, decreasing peripheral toxicity. Various factors that influence the transport include the type of polymer or surfactant, nanoparticle size, and the drug molecule. Polymeric nanoparticles have been shown to be promising carriers for CNS drug delivery due to their potential both in encapsulating drugs, hence protecting them from excretion and metabolism, and in delivering active agents across the BBB without inflicting any damage to the barrier. Currently, reports evaluating nanoparticles for brain delivery have studied anesthetic and chemotherapeutic agents. Polymer nanoparticle technology appears to have significant promise in delivering therapeutic molecules across the BBB.

1.45.8.3

Nanoparticles for Neuroprotection

Some nanoparticles have a neuroprotective effect, which along with neuroregeneration, is important for management of neurological disorders.18 QD technology has been used to gather information about how the CNS environment becomes inhospitable to neuronal regeneration following injury or degenerative events by studying the process of reactive gliosis. Glial cells, housekeeping

620

Nanobiotechnology

cells for neurons, have their own communication mechanisms that can be triggered to become reactive following injury. QDs are being used to build data capture devices that are easy to use by neuroscientists for tracking glial activity. Other research is looking at how QDs might spur the growth of neurites by adding bioactive molecules to the QDs, in a way to provide a medium that will encourage this growth in a directed manner. Poly(D,L -lactide co-glycolide; PLGA) nanoparticles loaded with superoxide dismutase have neuroprotective effect after H2O2induced oxidative stress, which appears to be due to the stability of the encapsulated enzyme and its better neuronal uptake after encapsulation.

1.45.9

Nanocardiology

Nanocardiology is the application of nanobiotechnology to the diagnosis and treatment of cardiovascular diseases. Important applications are in areas of diagnosis, targeted drug delivery, and reconstructive surgery of cardiovascular disorders.

1.45.9.1

Combining Diagnosis with Therapy of Cardiovascular Disorders

Perfluorocarbon (PFC) nanoparticles provide an opportunity for combining molecular imaging and local drug delivery in cardiovascular disorders. Ligands such as MAbs and peptides can be cross-linked to the outer surface of PFCs to enable active targeting to biomarkers expressed within the vasculature. PFC nanoparticles are naturally constrained by size to the circulation, which minimizes unintended binding to extravascular, nontarget tissues expressing similar epitopes. The utility of targeted PFC nanoparticles has been demonstrated for a variety of applications in animal models, including the diagnosis of ruptured plaques and the quantification and antiangiogenic treatment of atherosclerotic plaques. PFC nanoparticles and liposomes can be used as integrin-targeted paramagnetic contrast agents for molecular MRI of focal angiogenesis. Site-targeted PFC nanoparticles also offer the opportunity for local drug delivery in combination with molecular imaging. The diagnosis and treatment of unstable plaque constitute an area in which nanotechnology could have an immediate impact. Fibrin-specific PFC nanoparticles may allow the detection and quantification of unstable plaque in susceptible patients, which may be an important feature of future strategies to prevent heart attacks or stroke. Research is underway using probes targeted to plaque components for noninvasive detection of patients at risk. In an extension of this approach, targeted nanoparticles, multifunctional macromolecules, or nanotechnology-based devices could deliver therapy to a specific site, localized drug release being achieved either passively (by proximity alone), or actively (through supply of energy as ultrasound, NIR, or magnetic field). Targeted nanoparticles or devices could also stabilize vulnerable plaque by removing material such as oxidized low-density lipoproteins. Devices able to attach to unstable plaques and warn patients and emergency medical services of plaque rupture would facilitate timely medical intervention.

1.45.9.2

Targeted Drug Delivery in Cardiovascular Disorders

High-affinity ligand–receptor interactions have been exploited in the design and engineering of targeting systems that use nanodevices for site-specific cardiovascular drug delivery.19 An example of application is atherothrombosis, a condition in which platelet activation, adhesion, and aggregation are closely associated with vascular thrombotic events. Therefore, most antithrombotic therapies have focused on drugs that impede platelet–activation pathways or block ligand-binding platelet integrins. Despite the reasonable clinical efficacy of these therapies, the magic bullet, a single drug and delivery system that selectively targets pathologically thrombotic environment without affecting hemostatic balance remains elusive. The use of a combination of anti-integrin/ anticoagulant/anti-inflammatory drugs, carried by a nanoscale device, might be necessary to treat the multifactorial nature of pathological thrombogenesis. Nanoliposomes can be used as platelet-targeted devices for the delivery of cardiovascular therapeutics.

1.45.9.3

Role of Nanobiotechnology in Management of Restenosis

Restenosis – the reclosure or narrowing of an artery – can be a concern with coronary stents and with balloon angioplasty procedures. By using antiproliferative compounds that elute from the surface of a stent, the latest generation of drug-eluting stents (DES) has enabled a significant reduction in restenosis rates. Local delivery of antiproliferative drugs encapsulated in biodegradable nanoparticles, as well as nanocoating techniques for DES, have shown promise as strategies for preventing restenosis. There is need for new materials that cause the endothelial cells to attach better to the stents without forming much dangerous scar tissue. Ideally, endothelial cells should quickly attach to stents and form a coating that is only one cell-layer thick. The nanometer-scale bumps on titanium stents mimic surface features of proteins and natural tissues, prompting cells to stick better as compared to stents made of ordinary titanium.

1.45.9.4

Tissue Engineering and Regeneration of the Cardiovascular System

Nanotechnology may facilitate repair and replacement of blood vessels, myocardium, and myocardial valves. It may be also used to stimulate regenerative processes such as therapeutic angiogenesis for ischemic heart disease. Cellular function is integrally related to

Nanobiotechnology

621

morphology, so the ability to control cell shape in tissue engineering is essential to ensure proper cellular function in final products. Precisely constructed nanoscaffolds and microscaffolds are needed to guide tissue repair and replacement in blood vessels and organs. Nanofiber meshes may enable vascular grafts with superior mechanical properties to avoid patency problems common in synthetic grafts, particularly small-diameter grafts. Cytokines, growth factors, and angiogenic factors can be encapsulated in biodegradable nanoparticles and embedded in tissue scaffolds and substrates to enhance tissue regeneration. Scaffolds capable of mimicking cellular matrices should be able to stimulate the growth of new heart tissue and direct revascularization.

1.45.10 Nanosurgery The evolution of surgery from macrosurgery to microsurgery is undergoing further miniaturization, leading to nanosurgery. Nanotechnology will play an important role in the construction of miniaturized biosensing devices. These sensors improve outcomes, lower risk, and help control costs by providing the surgeon with real-time data about:

• • • •

instrument force and performance; tissue density, temperature, or chemistry; better or faster methods of preparing tissue or cutting tissue; and extracting tissue and fluids.

The introduction of lasers in surgery has already refined surgery and experimental biological procedures to enable manipulations beyond the capacity of the human hand-held instruments. Laser microsurgery is now evolving into nanoscale laser surgery. Femtosecond (one millionth of a billionth of a second) laser pulses can selectively cut a single strand in a single cell in a worm and selectively knock out the sense of smell. One can target a specific organelle inside a single cell (e.g., a mitochondrion, or a strand on the cytoskeleton) and zap it out of existence without disrupting the rest of the cell. Lasers can neatly zap specific structures without harming the cell or hitting other mitochondria only a few hundred nanometers away. It is possible to carve channels slightly less than 1 mm wide, well within a cell’s diameter of 10–20 mm. By firing a pulse for only 10–15 fs in beams only 1 mm wide, the photon density in each burst becomes incredibly intense: 100 quadrillion W m 2, 14 orders of magnitude greater than outdoor sunlight. That searing intensity creates an electric field strong enough to disrupt electrons on the target and create a micro-explosion. However, because the pulse is so brief, the actual energy delivered into the cell is only a few nanojoules. To achieve that same intensity with nanosecond or millisecond pulses would require so much more energy that the cell would be destroyed. The technology might be scaled up to do surgery without scarring or perhaps to deliver drugs through the skin. NIR femtosecond laser pulses have been applied in a combination of microscopy and nanosurgery on fluorescently labeled structures within living cells. Femtolasers are already in use in corneal surgery.

1.45.11 Nanorobotics Robotics is already developing for applications in life sciences and medicine. Robots can be programmed to perform routine surgical procedures. Nanobiotechnology introduces another dimension in robotics, leading to the development of nanorobots also referred to as ‘nanobots’. Instead of performing procedures from outside the body, nanobots will be miniaturized for introduction into the body through the vascular system or at the end of catheters into various vessels and other cavities in the human body. A surgical nanobot, programmed by a human surgeon, could act as an autonomous on-site surgeon inside the human body. Various functions such as searching for pathology, diagnosis, and removal or correction of the lesion by nanomanipulation can be performed and coordinated by an on-board computer. Such concepts, once science fiction, are now considered to be within the realm of possibility. Nanobots will have the capability to perform precise and refined intracellular surgery, which is beyond the capability of manipulations by the human hand.

1.45.12 Role of Nanobiotechnology for the Development of Personalized Medicine Personalized medicine, also referred to as individualized therapy, simply means the development and prescription of specific treatments and therapeutics best suited for an individual, taking into consideration both genetic and environmental factors that influence response to therapy.3 Genomic/proteomic technologies have facilitated the development of personalized medicines, but other technologies such as metabolomics are also contributing to this effort. Nanobiotechnology has refined many of these technologies. Personalized medicine is the best way to integrate new biotechnologies into medicine for improving the understanding of the pathomechanism of diseases and management of patients. Advances in nanobiotechnology will facilitate the development of personalized medicine by20:

• • •

improving the sensitivity and extending the present limits of molecular diagnostics/molecular imaging; integration of information from nanotechnology-based detection of biomarkers, point-of-care devices, nanochips, and nanobiosensors; and integration of diagnosis and therapy.

622

Nanobiotechnology

Pharmacogenetic assays based on nanotechnology have been approved by the Food and Drug Administration (FDA). Two examples are (1) Verigene (Nanosphere, Inc.) a nucleic acid test to detect variants of cytochrome P450 C29 (CYP2C9) and vitamin K epoxide reductase complex subunit 1 (VKORC1) genes, responsible for sensitivity to the anticoagulant warfarin and (2) Verigene F5/F2/ MTHFR (methylene-tetrahydrofolate reductase) nucleic acid test, which detects disease-associated gene mutations that can contribute to blood coagulation disorders and difficulty in metabolizing folate. Nanotechnology-based devices that provide site-specific drug delivery and efficient monitoring of drug effects may contribute to the development of personalized medicine. Examples have been given in various therapeutic areas in the preceding sections. Personalization of cancer therapies is based on a better understanding of the disease at the molecular level, and nanotechnology will play an important role in this area. Biomarkers, discovered by application of nanobiotechnology, are used to diagnose and treat cancer based on the molecular profiles of individual patients. The most important feature of personalization of cancer therapy is the use of the same nanoparticles for diagnosis as well as therapy. The future of cardiovascular diagnosis is already being impacted by nanosystems that can both diagnose pathology and treat it with targeted delivery systems. The potential dual use of nanoparticles for both imaging and site-targeted delivery of therapeutic agents to cardiovascular disease offers great promise for individualizing therapeutics. Image-based therapeutics with site-selective agents should enable verification that the drug is reaching the intended target and a molecular effect is occurring.

1.45.13 Safety Issues of Nanoparticles Natural nanoparticles existed in our environments long before the era of nanotechnology, but attention to safety issues is due to commercial development and application of nanoparticles, which may release nanoparticles into the atmosphere. The therapeutic success of nanoparticles is due to their small size, which enables us to introduce them into parts of the body where usual inorganic materials cannot enter because of their large particle size. This is also a cause for concern of toxicity of small particles that can enter parts of the body at unintended target sites. Toxicity is only an issue for in vivo use of nanoparticles, but not for in vitro diagnostics.

1.45.13.1 Fate of Nanoparticles in the Human Body Following inhalation, ultrafine and fine particles can penetrate through the different tissue compartments of the lungs, and eventually reach the capillaries and circulating cells or constituents, for example, erythrocytes. These particles are then translocated by the circulation to other organs including the liver, the spleen, the kidneys, the heart, and the brain, where they may be deposited. Kinetic studies to determine the influence of particle size on the in vivo tissue distribution of spherical-shaped gold nanoparticles in the rat show that most of the gold nanoparticles are concentrated in the liver and the spleen. A clear difference is observed between the distribution of the 10-nm particles and the larger particles. The 10-nm particles are present in various organ systems including blood, liver, spleen, kidney, testis, thymus, heart, lung, and brain, whereas the larger particles are only detected in blood, liver, and spleen. Smaller particles apparently circulate for much longer and, in some cases, can cross the BBB to lodge in the brain. They can leak out of capillaries and enter the fluids between cells. Therefore, they can go to places in the body that an average inorganic mineral cannot. Such effects may not be a cause for concern in case of targeted delivery of nanoparticle-based therapy in cancer. The eventual decision to use nanoparticle-based therapy may depend on a risk-versus-benefit assessment.

1.45.13.2 Impact of Nanoparticles on Human Health Modern humans breathe in considerable numbers of nanoparticles daily in traffic fumes, and even from cooking. Nanoparticles are used increasingly in industrial processes and have been hypothesized to be an important contributing factor to the toxicity and adverse health effects of particulate air pollution. Small size, a large surface area, and an ability to generate reactive oxygen species play a role in the ability of nanoparticles to induce lung injury. The biological impacts of nanoparticles are dependent on size, chemical composition, surface structure, solubility, shape, and aggregation. These parameters can modify cellular uptake, protein binding, translocation from portal of entry to the target site, and the possibility of causing tissue injury. The effects of nanoparticles depend on the routes of exposure that include gastrointestinal tract, skin, lung, and systemic administration for diagnostic and therapeutic purposes. Interactions of nanoparticles with cells, body fluids, and proteins play a role in their biological effects and ability to distribute throughout the body. Nanoparticle binding to proteins may generate complexes that are more mobile and can enter tissue sites that are normally inaccessible. Accelerated protein denaturation or degradation on the nanoparticle surface may lead to functional and structural changes, including interference in enzyme function. Nanoparticles also encounter several defenses that can eliminate, sequester, or dissolve them. The effects of particles on human health have been studied by toxicologists previously. The effects of larger particles generated by wearing down of implants in the body and aerosolized particles of all sizes on have been studied. Many of the inhaled particles enter the systemic circulation and are excreted in urine if they are 1 and the Number of Stirrers 2.08.4.2.3 Influence of Viscosity 2.08.4.3 Two-Phase Stirred Tanks 2.08.4.3.1 Influence of Aeration 2.08.4.3.2 Influence of HL/ T > 1 and the Number and Type of Stirrers 2.08.4.3.3 Conclusion 2.08.4.4 Bubble Column 2.08.5 The Airlift 2.08.6 Comparison of the Reactor Types 2.08.7 Gas-Phase Mixing 2.08.7.1 Stirred Tank 2.08.7.2 Bubble Column 2.08.7.3 Airlift 2.08.8 The Meaning of Mixing 2.08.8.1 Characteristic Times 2.08.8.2 Heat Production 2.08.8.3 C-Substrate 2.08.8.4 Oxygen Mass Transfer 2.08.8.5 Oxygen Depletion in Stirred Tank 2.08.8.6 Oxygen Depletion in Bubble Column 2.08.9 Conclusions References Relevant Website

78 78 79 79 80 80 81 81 81 82 82 82 82 83 83 83 83 84 84 84 85 85 85 86 88 88 89 90 91 91 91 91 91 91 92 92 92 93 93 93 94 94

Glossary Characteristic time Reactor parameter characterizing the rate at which a (sub)process occurs in the reactor, such as heat production, oxygen mass transfer, and C-substrate consumption. Critical time The time value above which problems will occur due to variation in temperature, oxygen concentration, and Csubstrate concentration. Degree of homogeneity The extent to which homogenization has progressed compared to the initial nonhomogeneous state.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00070-7

77

78

Mixing in Bioreactor Vessels

Ideally mixed The single-phase contents of a vessel are ideally mixed if the mixing process is much faster than the other relevant subprocesses (such as substrate conversion and residence time). Mixing time The time required to achieve a certain degree in the homogeneity of a considered volume of fluid in batch at a certain scale of mixing, starting from a nonhomogeneous state. Mixing time number The dimensionless group consisting of mixing time, power dissipation, and characteristic length. Scale of mixing The smallest scale at which inhomogeneities are considered.

2.08.1

Introduction

The process of conversion of substances by microorganisms or by chemical reaction necessitates the transport of substrate and product toward and from the location of the conversion. One of the important possible limitations of the conversion process is the transport by mixing in each phase present. In a continuous process or fed-batch process, the feed (substrate, wastewater) must be transported from the injection point through the bulk liquid to the organisms. The products of the conversion, at least those that may hinder the conversion process, must be removed from the microenvironment of the organisms. In a batchwise-operated process, transport is needed to avoid local depletion of substrate, such as oxygen. The way in which transport is desired to occur is dictated by the process type and realized by the choice of reactor type. The way in which the transport actually will happen is determined by technical and physical limitations in reaching the desired mixing behavior. The two extremes in bulk liquid mixing behavior are the ideal stirred vessel (or ideal mixer) and the ideal plug flow. Ideal means that these cases can be described mathematically in a simple way, but that they cannot be realized in practice, only approached.

2.08.1.1

The Ideal Stirred Vessel

In the ideal stirred vessel, there are no concentration differences. The substance considered is homogeneously distributed in the vessel at all times. If there is a feed of concentrated substrate, then this substance is distributed instantly and completely at the moment of entry (Figure 1). Mathematically this means that the concentration c is independent of the location within the vessel. Consequently, the concentration of the outgoing flow is equal to the concentration within the vessel (cout ¼ c). Therefore, a macroscopic mass balance of a component over the vessel volume V is not a function of location. dcV ¼ Fin cin  Fout c þ Fm;transfer  rV dt accumulation ¼ transport  conversion

(1)

The change in the amount of substance cV in time (accumulation) (mol s1 or kg s1) is the summation of inflow Fincin (feed), outflow  Foutc, conversion (consumption  rV or production þ rV) of the substance, and possibly transfer (Fm,transfer) of the substance from (or to) another phase. In multiphase systems (e.g., aeration: air and water), a mass balance is needed for each phase (for air and water). A special case of the ideal stirred vessel is when the vessel is operated batchwise (Fin ¼ 0; Fout ¼ 0) (single-phase system: Ftransfer ¼ 0, considered component is a substrate in solution): dcV ¼ rV dt

(2)

Concentration c in

In Out

cout

c

In Figure 1

Location

Out

Example of the distribution of the concentration of a substance in the ideal stirred tank in steady state.

Mixing in Bioreactor Vessels

79

If the volume does not change as a result of the conversion or otherwise dc ¼ r dt

2.08.1.2

(3)

The Ideal Plug Flow

A vessel with ideal plug flow behavior is visualized as a flow in a tube (Figure 2). The velocity profile of the fluid in the tube is fully flat and there is totally no dispersion of the considered substance. Consider now a (very small) volume element of fluid that travels through the tube (the vessel) without exchange with its surroundings. The process in this volume is mathematically identical with that in the ideal stirred batch vessel (Eq. (3)). (The volume of the element considered is so small that diffusion is fast enough to maintain a homogeneous distribution within the element.) Since dx ¼ v dt Eq. (3) gives v

dc ¼ r dx

(4)

In a plug flow vessel, the amount of substance depends fully on location (Figure 2) and is only a function of time if the entry value is a function of time. Plug flow without conversion may be considered as a time ‘delay’, where the delay is equal to the residence time in the plug flow vessel. The derivation of Eq. (4) follows also from a mass balance of the component considered over a section located between x and x þ Dx (see Figure 3) for constant flow velocity (accumulation ¼ transport  conversion) ðA DxÞ

dc ¼ Av cðxÞ  Av cðx þ DxÞ  A Dx r dt

or for Dx / 0 (and v independent of x)


 dc dc ¼v  r dt dx

(5)

(6)

In steady state (dc/dt ¼ 0, hence also the entry concentration is constant), Eq. (4) results.

2.08.2

Characterization of Mixing

Although the ideal cases of mixing behavior discussed in the previous session cannot be realized in practice, special design can lead to reactors whose mixing behavior approximate the ideal case sufficiently to allow the ideal description. Small-scale laboratory fermenters with nonviscous medium and rigorously stirring may be considered to be ideally mixed. Slender packed columns

Concentration

c in

In Out

cout In

Location

Out

Figure 2

Example of the concentration distribution of a converted substance in an ideal plug flow reactor in steady state.

Figure 3

Control volume of ideal plug flow.

80

Mixing in Bioreactor Vessels

may be considered to be in ideal plug flow. Aeration tanks are generally considered to be well mixed; settlers are designed to approximate plug flow. If the mixing process is much faster than the other relevant subprocesses (such as oxygen transfer, substrate conversion, and residence time), then the vessel may be considered to be ideally mixed. On the other hand, if mixing is much slower, nonbatch systems may be considered to be in ideal plug flow. Comparing subprocesses in this way is the subject of regime analysis. Here, it is sufficient to state that subprocesses may be characterized each by a time, a characteristic time, which is a measure of the rate at which the subprocesses run. For example, for oxygen transfer, this is the time it takes to saturate a depleted amount of liquid to a defined extent, given by 1/kLa. For bulk liquid mixing, it is the time it takes to homogenize an added amount of substance in a given amount of liquid to a certain degree. If the characteristic mixing time, the time of the mixing process in the whole reactor, is much smaller than that of the other relevant subprocesses in that phase, then the simple model of the ideal stirred vessel may be used for the whole reactor. If the mixing time is much larger, then problems might arise, and for description of the process, a more complex model is needed.

2.08.2.1

Mixing Time

To arrive at a proper definition of the mixing time, first homogenization must be defined. Convective (bulk) mixing is effective to a certain scale. For smaller dimensions, diffusion (micromixing) takes over. Since the latter process cannot be influenced by the input of power, the corresponding scale may be used as the ‘scale of mixing’: the scale on which the inhomogeneities are considered. However, when mixing time is measured using a tracer, the scale of mixing is determined by the size of the probe. The probe size is generally larger than the scale of micromixing. To what extent homogenization has progressed is indicated by the ‘degree of homogeneity’  
CN  C   100% (7) m ¼ 1   CN  After the concentration c of the substance considered has reached its final value cN, the degree of homogeneity is 100%. The mixing time tm is defined as the time required to achieve a certain degree in homogeneity of a considered volume of fluid in a batch vessel at a certain scale of mixing starting from a nonhomogeneous state. The mixing time is measured by following the concentration of an inert tracer in time at one or more locations within the vessel after adding a certain amount pulsewise. Practical mixing times are defined for degrees of homogeneity m between 50 and 95%, the latter being most common and used in this text (0.95cN < c < 1.05cN; see Figure 4).

2.08.3

Mixing Models

Mixing is caused by bulk flow and turbulent dispersion. Two main types of mixing models originate from this background: the bulk flow model, starting from the bulk flow path with dispersion superimposed on the average flow, and the turbulence model, starting from the mixing action of the energy-dissipating turbulent eddies.

Normalized signal c 1.6 c∞ 1.4 tc 1.2

Definition 95% mixing time

1 105%

1 95% 0.8

2

0.6 0.4 0.2 0 0 Figure 4

t95.1

t95.2

Time

Mixing time: Visualization of the definition; 1 and 2 indicate two possible response curves after adding some substance to a batch vessel.

Mixing in Bioreactor Vessels 2.08.3.1 2.08.3.1.1

81

Bulk Flow Models Stirred Tank

The flow path in a one-phase stirred tank with a radial pumping stirrer is given in Figure 5. Each of these circulation loops can be modeled with a closed-loop reactor as shown in Figure 6. A probe (pH electrode, conductivity, etc.) can detect the recurrent tracer (acid, salt, etc.) concentration as shown in Figure 4. It is evident that a circulation loop exists in Figure 6. In the loop, a circulation time tc can be defined (see Figure 4, curve 1) as the average time needed for the liquid to pass one circulation. Although less pregnant, a similar response curve can be found in a stirred tank after tracer addition in the stirrer region. Based on stirrer pumping capacity and total volume, a mixing time relation can be derived for the fully turbulent case2:  T 3
 1 D HL (8) tm f N NP0:33 T where NP ¼ impeller power number defined by Ps ¼ NPrLN3D5, rL ¼ liquid density, D ¼ stirrer diameter, T ¼ tank diameter, HL ¼ liquid height, Ps ¼ stirrer power consumption, N ¼ stirrer speed. Eq. (8) predicts that for Re > 5000, for which NP is constant, Ntm has a constant value.

2.08.3.1.2

Bubble Column

Following the same lines of Ref. 2 with velocity data measured in bubble columns  c 2 0:33 HL tm f gvGs T T

(9)

where g ¼ gravitational acceleration, vGsc ¼ pressure-corrected superficial gas velocity.

H

T Figure 5

Schematic representation of the secondary flow path in a turbine stirred vessel.1

Tracer pulse at t = 0 Figure 6

Modeled flow path for a turbine stirred tank.

Tracer detection

82

Mixing in Bioreactor Vessels

2.08.3.1.3

Airlift

The airlift is a circulation loop in itself. The general formula for the flow velocity is given as3,4 ZHL

g DεðzÞdz ¼ 0:5Kf ðvLs Þ2

(10)

0

where Kf ¼ friction coefficient, a value to be determined for each airlift; Dε(z) ¼ holdup difference riser and downcomer; z ¼ length coordinate. In this formula, ε and nLs depend on each other. (Iteration procedures are shown by Verlaan et al.4 to calculate the values of, Dε and nLs.)

2.08.3.2 2.08.3.2.1

Turbulence Models Stirred Tank

The basis for calculation of mixing time is the relation5
1 3 et L2c =

tm f

(11)

where et ¼ local energy dissipation rate, Lc ¼ integral scale of turbulence. The problem that is included in this formula is the uneven distribution of the energy dissipation throughout the vessel being much larger near the stirrer than, for instance, near the vessel wall. Mixing is complete only when all parts of the vessel are mixed. The position of the lowest energy dissipation rate therewith determines the mixing time.5 Groen1 also used energy dissipation as a starting point. He regards the stirrer dimension as the characteristic length scale of the turbulent eddies and includes also the energy provided by the gas flow. He derives with flow paths as given in Figure 7:

 T 2=3 Hs 4=3 HL 2 tm f 1=3 (12) T T e where e ¼ total, stirrer plus air, specific energy dissipation. For a one-phase stirred tank and a given ratio of stirrer diameter and stirrer height, this may be transformed into  T 5=3
2 1 D HL tm f N N 1=3 T

(13)

P

2.08.3.2.2

Bubble Column

Groen1 could also apply his model straightforwardly to a bubble column:
 T 2=3 HL 2 tm f 1=3 T e

(14)

wherein c e ¼ gvGs

H

H

T Figure 7 Schematic representation of the flow paths in stirred vessels (single- and multi-impeller) and bubble columns (heterogeneous and homogeneous regime).1

Mixing in Bioreactor Vessels 2.08.3.2.3

83

Airlift

Energy dissipation models for an airlift are less useful because bulk circulation flow models provide the opportunity to calculate the circulating velocities. Dispersion then can be included by the relation between circulation time and mixing time, which is assumed to be tm ¼ ðfrom 4  7Þtc

2.08.3.3

(15)

Mixing Time Number 1=3

1=3

The mixing time relations shown in the previous sections contain either factor e1=3 or NP . At given stirrer speed and diameter, NP is equivalent to e1=3 . A dimensionless mixing number Nmix, including scale effects, can be defined as1 Nmix ¼

tm e1=3 T 2=3

(16)

The mixing number should be dependent on geometric data of the vessels only, because the dependence on stirrer speed and power number for stirred tanks and on dissipated energy for bubble columns is included in the mixing number itself. With Eq. (16) the mixing time relations for both a stirred tank and a bubble column1 transform to
2 HL Nmix f (17) T

2.08.4

Experimental Verification

2.08.4.1

Inherent Variations in Measured Mixing Times

It is stated already that the actual value of the mixing time depends on the required homogeneity. Usually 95% is taken. Given this criterion, the reproducibility of the experiments can be within 10%. However, another criterion, and another measurement scale, leads to another, and again reproducible, mixing time value. An even much more important cause for the large differences in measured mixing times is the dependence on the position of the tracer addition and the measurement point. For a single turbine stirred tank, addition of the tracer at the top of the vessel and measurement near the vessel bottom makes considerably longer measured mixing time values compared to the case of a tracer added in the stirrer region. For a bubble column or a stirred tank with multiple stirrers, halfway the bubble column or stirred tank makes considerably shorter mixing times. As an example, Figure 8 shows that the measured mixing time can differ by a factor of 2 depending on the positions of tracer addition and measurement point. The lesson learned is that we cannot expect exactly the same mixing time values when comparing measured values of different authors in the same vessel geometry; too many (reproducible) possibilities for differences do exist.

Relative reactor height 2

H T 1.5

1

Impellers

0.5

0 0

10 s

20 s

Mixing time Figure 8 Effect of position of addition on the mixing time in a nonaerated stirred vessel (T ¼ 0.72 m) with two Rushton-type impellers (N ¼ 110 min1, D ¼ T/2). Adapted from Cronin DG, Nienow AW, and Moody GW (1994). An experimental study of mixing in a protofermenter agitated by dual Rushton turbines. Transactions of the Institute of Chemical Engineers, Food and Bioproducts Proceedings 72(C1): 35–40.

84

Mixing in Bioreactor Vessels

2.08.4.2 2.08.4.2.1

Single-Phase Stirred Tanks Influence of T/D Value

Voncken et al.,7 already in 1964, had thoroughly researched the influence of T/D. They measured the (T/D)2 relation. Many authors5 have found relations around the value of 2. All these data together can be summarized within the variations possible as
2 T (18) tm f D The relation comes very near to the relation predicted by using flow patterns and turbulence theory.1

2.08.4.2.2

Influence of HL/ T > 1 and the Number of Stirrers

2  Groen measured the HL=T relation for a large variety of scales. Several authors reported relation values of 1.5 up to 2.5. Within the accuracy of mixing time measurements, we assume
2 HL (19) tm f T 1

2.08.4.2.2.1 Single-Phase Mixing Time Equation Nienow5 has worked out Eq. (19). Correcting the data of Nienow5 with a factor of 2, assuming that to be the difference between tracer addition near the stirrer and at the top of the vessel, and combining the T/D relation and the HL/T relation with his data and the data of Groen1 gives. Single-phase, turbulent tracer added on top tm

 T  2
2 10 D HL ¼ N N 1=3 T P

(20)

or, rewritten as a function of the dimensionless mixing time number
1=3
2 tm e1=3 T HL Nmix ¼ 2=3 ¼ 11 D T T

(21)

Figure 9 shows Nmix calculated with Eq. (21) for a large variety of scales. This describes the experimental results very well. The data in Figure 9 are all measured for a stirrer in each H ¼ T compartment. The situation for other geometries is more complicated. Groen measured for instance a distinct influence of the number of stirrers. The mixing time in a tall vessel equipped with one stirrer is lower than that in the same vessel with a stirrer in each H ¼ T compartment. In Figure 10, the diamonds show measured tm values for a large-scale fermenter dependent on height, with a clearance between impellers of 2D (5.4 m3 per impeller) at constant P/V. As expected, the mixing time increases quadratically with height (line confirming quadratic relation of Eq. (21)) and also a large number of impellers increase the mixing time at constant volume (diamonds have a higher mixing time than squares at the same volume). This means that, at the same stirrer speed, doubling the power input (by doubling the number of stirrers per volume) increases the mixing time instead of decreasing. This elucidates that the effect of compartmentalization with more impellers is stronger than the effect of power input. The type of stirrers (three turbines versus two Scaba and one turbine or multiple stirrers) also influences the mixing time.9,10 In general, compartmentalization is considerably less, but not fully absent when using axial Nmix experimental

Working volume; Number of stirrers

100 80 130 m3 ; 1, 3

60

3 2.40 m ; 2 3 1.80 m ; 2

40

3 0.10 m ; 1,3

0.10 m3; 2

20

3 0.045 m ; 1,3

0

0

20

40

60

80 100 Nmix, calculated

Figure 9 Mixing time number model verification for nongassed vessels and Rushton impeller(s). Measurements were done at several scales and geometries as indicated. In the case of reduced number of impellers, the working volume was adapted accordingly. T/D ¼ 2 (130 m3/45 m3); 2.5 (2.4 m3); 2.9 (1.8 m3); 2.27 (all others). Blade height is 0.25D, but 0.30D (2.4 m3) or 0.33D (1.8 m3). Volumes are water volumes.1

Mixing in Bioreactor Vessels

85

Mixing time at 115 rpm 200 150 100 50 0 0

1

2 3 Number of stirrers

4

5

Figure 10 Mixing times in multi-impeller systems. VL ¼ 22 m3 reactor with four Rushton turbine impellers with clearance 2D (H ¼ 6.55 m, T ¼ 2.07 m, D ¼ 0.7 m, 4 baffles, water). Diamonds: mixing time for 1 impeller per 5.5 m3; squares: mixing time for 1 (lower) impeller per 11 m3. Data for the same tank at lower liquid volumes (VL ¼ 16.5, 11, and 5.5 m3) are indicated by the length of the stirrer axis. Unpublished data and Enfors SO, Jahic M, Rozkov A, et al. (2001). Physiological responses to mixing in large scale bioreactors. Journal of Biotechnology 85: 175–185. Table 1

Calculated mixing times for a stirred tank for different scales and aspect ratios ¼ 2:5, e ¼ 2 W kg1

T D

VL (m3)

0.01

10

100

1000

HL 5T 2T T

79 15 4

364 71 21

607 119 35

1011 198 58

pumping impellers. Yet, it is difficult to give definite answers. The lower stirrer is nearly always of the radial pumping type, and with axially downward pumping stirrers at the higher positions maybe the mixing time is decreased. One should realize that the type(s) of stirrer(s) at HL > T, the mixing intensity and scale criteria, and the position of the tracer addition/measurement points all influence the mixing time, with an estimated range of about 0.5 up to 2 times the value of Eqs. (20) and (21). Table 1 shows mixing time values for a range of volumes from 1 l up to 1000 m3 and for different geometries. A general value of stirrer power P/V ¼ 2000 W m3 and T/D ¼ 2.5 is used. It is clear that scale-up inevitably leads to larger mixing time values. Large-scale mixing times of the order of 100 s are found. To bring back 100 s to 10 s, a factor of 103 increase in the required power value is needed or a power of 2 MW m3 due to the 1/3 power for e in the Nmix formula. In particular, tall vessels (with multiple stirrers) lead to extreme mixing time values. The data of Table 1 also show that the scale of operation and HL/T values are overpowering all other effects such as the earlier mentioned possible variations of a factor of 0.5 up to 2. The mixing time can easily become larger than the characteristic time for mass transfer (order of 10 s) or substrate uptake. The consequences of this will be discussed later.

2.08.4.2.3

Influence of Viscosity

Viscosity can have a dramatic effect on the mixing time. The effect is dependent on the stirrer Reynolds number: Rest ¼

rND2 h

Figure 11 shows that above a value of 5000, the Ntm value is constant. However, below a value of 1000, the Ntm value can increase easily by a factor of 10 or more. This will certainly lead to complications because in that case nearly all characteristic times will be lower than the mixing time. The Reynolds number will increase at scale-up for the usual criteria such as constant Ps/VL values or constant tip speed. This has a positive effect. In the case of viscosity limitations at a small scale of, for instance at, 1 liter scale and Rest around 1000, it might be possible that at a large scale the Rest number increases to a value above 5000 where no viscosity problems occur anymore. In the case of high-viscosity and pseudoplastic liquids, the liquid in the top of a tall vessel can come to a complete standstill. Mixing in the stirrer region is not that bad in that case, but the mixing time for the vessel as a whole is infinite.

2.08.4.3 2.08.4.3.1

Two-Phase Stirred Tanks Influence of Aeration

Aeration influences the bulk flow field and the behavior of the stirrer. Figure 12 shows the different bulk flow patterns. At low gas flow rates and high stirrer speeds, the flow pattern resembles largely the flow pattern of the unaerated case resulting in total gas

86

Mixing in Bioreactor Vessels

10 000

Ntm

1000

100

10 100

1000

10 000

100 000 Rest

Figure 11

Ntm values for a Rushton-type turbine impeller.

N

FG Figure 12 Flow patterns in an aerated stirred vessel with Rushton-type impeller. Left to right: increasing gas flow rate or decreasing stirring speed. Adapted from Gogate PR, Beenackers AACM, and Pandit AB (2000). Multi impeller systems with a special emphasis on bioreactors: A critical review. Biochemical Engineering Journal 6: 109–144.

recirculation. Increasing the gas flow rate changes the flow pattern. The increasing energy content of the gas pushes the flow from the stirrer upward. Here, we get a situation of limited gas recirculation. Finally, a situation can occur where the gas is not flown out anymore and the circulation loop is totally reversed, with the gas and liquid rising up along the shaft. This is called flooding. A second effect of aeration is the formation of cavities behind the stirrer blades.2,11 Figure 13 shows three distinct types. Vortex clinging cavities appear with a gas recirculating bulk flow. Large cavities are coupled to only partial gas recirculation. Stirrer power consumption and pumping capacity decrease at cavity formation and they might also influence the flow and mixing time. Altogether, mixing time determination in gassed stirred tanks might be complicated. A fine example is shown in Figure 14. The unaerated data fulfill the Ntm ¼ constant criterion. However, a number of data show a distinct deviation from this curve. A careful observation of the flow conditions and the use of the total energy dissipation by means of the mixing number give the solution as shown in Figure 15. Flooding leads to lower mixing times. The bulk flow is reversed at flooding and will be more similar to gulf stream flow. Intensive aeration (3–3 structure) leads to mixing numbers of the same value as in the unaerated case. Figure 15 shows that limited aeration leads to higher Nmix values. Most large-scale bioreactors operate in the range of high gas load because of efficient mass transfer2 and the mixing time in that case will be about the same as that for the unaerated case. Figure 16 shows this for measurements at different scales. In particular, this is true at large scales. At a small scale (below 50 l), the effect of aeration seems to become negative, and the mixing time increases up to a factor of about 2. Altogether, the mixing time in aerated stirred tanks has the same value as for the unaerated case, considering that depending on aeration rate, stirrer speed, and scale, a variation of a factor in the range of 0.5 up to 2 can occur.

2.08.4.3.2

Influence of HL/ T > 1 and the Number and Type of Stirrers

Groen1 and Vrabel et al.9 have shown that, in general, the quadratic relation works out well. From the same sources, it is clear that differences up to a factor of 2 occur when certain combinations of stirrers are used, in particular, downward pumping stirrers in the upper compartments and a turbine in the lower compartment.

Mixing in Bioreactor Vessels

87

Direction of rotation Disk Blade

Unaerated Disk

Direction of rotation

Blade

Vortex cavity Clinging cavity

Disk

Direction of Blade rotation

Large cavity

Disk

Direction of Blade rotation

Figure 13 Cavity shapes. Warmoeskerken MMCG (1986). Gas-Liquid Dispersion Characteristics of Turbine Agitators. PhD Thesis, Delft University of Technology. http://repository.tudelft.nl/view/ir/uuid:2bc4c2fc-f89e-497a-ab7a-ed8a7a2d15df/.

VGs =

95% mixing time

40 s

no air 0.015 ms–1 0.029 ms–1 0.044 ms–1 0.055 ms–1 Ntm = 56

30

20

10

0 1

2

3

4

5 s–1 Stirrer speed

Figure 14 Mixing times influenced by gas superficial velocity in a 4 m3 stirred tank with two impellers. Groen DJ (1994). Macromixing in Bioreactors. PhD Thesis, Delft University of Technology. http://repository.tudelft.nl/view/ir/uuid:3ac019f1-d19a-4853-9a29-554f1149bd5b/.

50

Mixing number Nmix 3–3 structure Flooding Gas recirculation Nonaerated

40 30 20 10 0 0

0.1

0.2

0.3 Gas flow number

FG

0.4

ND 3

Figure 15 The effect of aeration and cavity structures on the mixing number. Four regimes emerge: nonaerated; complete gas recirculation; impeller loading conform 3–3 structure; flooding. The mixing time in the 3–3 structure regime is equal to that of the unaerated case at equal total power input. Groen DJ (1994). Macromixing in Bioreactors. PhD Thesis, Delft University of Technology. http://repository.tudelft.nl/view/ir/uuid: 3ac019f1-d19a-4853-9a29-554f1149bd5b/.

88

Mixing in Bioreactor Vessels

Gassed over ungassed mixing number ratio 2

Gas recirculation

N mix, G N mix, u

3–3 structure Flooding

1.5

1

0.5

0 0.01

0.1

1

100 m3

10

1000

Scale Figure 16 The influence of scale and flow regime on the mixing number. Groen DJ (1994). Macromixing in Bioreactors. PhD Thesis, Delft University of Technology. http://repository.tudelft.nl/view/ir/uuid%3A3ac019f1-d19a-4853-9a29-554f1149bd5b/.

2.08.4.3.3

Conclusion

The mixing time in aerated stirred tanks has about the same value as for the unaerated case, considering that depending on aeration rate a variation of a factor in the range of 0.5 up to 2 can occur. As shown earlier in Table 1, scale of operation and H/T values are overpowering this effect. Stirred tank. Single- and two-phase, turbulent tracer added on top tm ¼

 T  2
2 10 D HL N N 1=3 T P

with NP ¼ unaerated power number or

1

Nmix ¼

tm e3u 2

T3

¼ 11

ð20Þ


1
2 T 3 HL D T

ð21Þ

with eu ¼ unaerated power dissipation.

2.08.4.4

Bubble Column

Measurements at different scales including commercial-scale bubble columns are shown in Figure 17. Till HL/T ¼ 3 the mixing number is about 16, independent of the height of the column, with a scatter of a factor of 0.5 up to 2.

Nmix

(HL/T)2

500 0.12 m3

100

4.0 m3 120 m3

N mix = 16

10 5

1

2

3

10

4 5

20

HL/T Figure 17 Measured mixing numbers in bubble columns: aspect ratio and scale. Data from Groen DJ (1994). Macromixing in Bioreactors. PhD Thesis, Delft University of Technology. http://repository.tudelft.nl/view/ir/uuid:3ac019f1-d19a-4853-9a29-554f1149bd5b/.

Mixing in Bioreactor Vessels

89

At H/T > 3, the situation changes considerably; here a relation with H/T2 appears. Thus the conclusion for a bubble column should be.

2

tm ¼ 16 

Nmix ¼

T3  c 1=3

gvGs

tm e1=3 ¼ 16 T 2=3

and

HL 3 T

(25)

tm  ¼ 1:6 

Gs

Nmix ¼

(22)

These equations are in accordance with those predicted by the bulk flow as well as the turbulence theories. However, the (H/T)1 relation in Eq. (9) certainly does not hold and should be replaced by the relations given above. As can be expected, mixing time increases inevitably with scale, at a constant pressure-corrected superficial gas velocity 2 (pneumatic power input) with a factor of T 3 . All data until now are for the heterogeneous flow regime. The homogeneous flow regime leads to larger mixing times.1,2 As this regime will hardly be present in any commercial fermenter, this will not be discussed here and the interested reader can refer the literature on this. Table 2 shows calculated values dependent on scale and H/T for a vGsc value of 0.05 m s1. As for the stirred tanks, scale and H/T overrule any other effect. The increase from H ¼ T compared to H ¼ 2T follows from the formula, due to the larger diameter at H ¼ T at a given volume.

2.08.5

The Airlift

The airlift reactor is developed from the viewpoint of controlled flow and intensified circulation caused by the density difference between the riser column filled with air bubbles and a downcomer column filled with liquid only or less air. A schematic representation of airlift types (external and internal) is given in Figure 18. Controlling the flow is interesting in particular on very large scales. Extensive reviews of airlift type reactors are given by Merchuk12 and Gumery et al.13 Table 2

Calculated mixing times for bubble columns (Eqs. 23 and 25) at different scales and HL/T values vGsc ¼ 0.05 m s1

VL (m3)

0.01

10

100

1000

HL 5T 2T T

13 7 8

62 30 35

103 51 59

172 84 98

G

G Figure 18

Schematic representation of external and internal airlift loop reactor.

G

90

Mixing in Bioreactor Vessels

The flow velocity can be calculated for the reactors of Figure 17 starting from Eq. (10).3,4 Flow velocities are dependent of the friction factor that is dependent on the construction data of each air lift and therefore no general relation can be derived. The riser velocity values found are in the range of 0.1 up to more than 1 m s1 for pilot plant (10 l) up to commercial scale (>1000 m3) reactors.14 From the circulation velocity, the circulation time could be calculated and the mixing time measured. Verlaan et al.4 reported that for an airlift reactor tm ¼ (4–7)tc. Downcomers with a relatively small cross section can hinder the circulation too much. In that case, the velocity in the riser becomes very low and the riser will start to behave as a bubble column. Nothing can be said in that case for velocities from calculations. A comparable effect may occur at high air entrainment into the downcomer.15

2.08.6

Comparison of the Reactor Types

Relations are given in the previous sections for the bulk liquid mixing time based on equations similar to that derived from theories and modified on the basis of experimental results. Mixing time values can differ from that given by the relations shown. As discussed, this is mainly due to tracer injection position and number, and type and position of the stirrers (in the case of multiple stirrers). These differences are limited when compared to differences that occur when different scales and different types of reactors are compared. As an example, calculated mixing times are shown in Table 3. This simple table shows trends that go far beyond differences of 0.5 up to 2 times the value of the formula at given conditions. All rows in the table show the increase of mixing time with scale. At very large scales, times of several minutes are inevitable. For mixing time-sensitive fermentations, we always have to solve problems at scale-up. The first group of data at an aspect ratio of 10 shows that the airlift has a mixing time about half of that of the bubble column. However, the second group of data shows that already at an aspect ratio of 5 instead of 10, the mixing time in the bubble column becomes smaller when compared to the airlift. This difference increases further at an aspect ratio of 2. It is highly questionable to prefer the mechanically more complicated airlift for mixing reasons only. The group of data at an aspect ratio of 5 shows that the bubble column is to be preferred when compared to the stirred tank, even though the stirred tank has an additional energy dissipation of 2 W kg1 due to the stirrer. Bubble columns are much more efficient for mixing at these aspect ratios. The stirred tank is even more unacceptable at an aspect ratio of 10. This can be expected with a high aspect ratio reactor known from the chemical industry in mind, the rotating disc contactor that is used in those cases where plug flow is desired. The third group of HL ¼ 2T shows that at low aspect ratios the mixing time of a bubble column and that of a stirred tank become more comparable, yet the bubble column is still better and uses much less power. The table also shows a stirred tank of 1000 m3 with a dissipated power of 2 W kg1, amounting to a total of 2 MW. Such vessels did not exist until now, because of mechanical limitations. From the viewpoint of mixing also, these reactors are not to be preferred. When one considers mixing, the conclusion is that bubble columns are by far to be preferred in most cases because of lower mixing time, lower power consumption, and simple mechanics. However, it is known that oxygen mass transfer in viscous systems is limited or even very low in bubble columns and airlift reactors.2 Stirred tanks can handle viscous broths better. The mixing time will not change very much at large scales and at not too high viscosity values for all type of reactors because of the high Rest number. Mass transfer is not Rest dependent but only viscosity dependent, and this explains largely why stirred tanks are preferably used for a number of fermentations up to a volume of about 300 m3. Also, the use of stirred tanks in multipurpose (pilot) plants can be explained by this fact as it provides flexibility.

Table 3

Mixing time values calculated for a stirred tank (Eq. 21), a bubble column (Eqs. 22 and 24), and an airlift reactor (based on Refs. 2 and 4) eu ¼ 2W kg1 VL (m3) VGsc (m1)

HL/T 10 10 10 5 5 5 2 2 2

Airlift BC STR Airlift BC STR Airlift BC STR

0.01

10

100

1000

0.01

0.10

0.01

0.10

0.01

0.10

0.01

0.10

49 78 270 47 23 79 46 11 15

19 36 270 18 11 79 18 5 15

220 363 1200 198 106 364 192 52 71

96 168 1200 78 50 364 72 24 71

420 604 2000 360 176 606 340 86 119

180 281 2000 150 82 606 140 40 119

730 1000 3500 610 294 1000 570 144 198

320 468 3500 260 136 1000 240 67 198

Mixing in Bioreactor Vessels

91

Another reason to use stirred tanks might be that the volumetric productivity of a stirred tank can be higher than that of a bubble column. Volumetric productivity for aerobic fermentations is nearly always limited by mass transfer rate. This is directly related to power consumption, which can be higher in a stirred tank.

2.08.7

Gas-Phase Mixing

Gas-phase mixing data are scarce. This is despite the fact that the measurements are relatively easy to do by means of residence time measurements. This is not a real problem because in most cases the gas-phase mixing is hardly relevant for the process in the bioreactor. And when the gas phase is in plug flow, it is even an opportunity for mass transfer because of the optimal driving force, in contrast to the liquid mixing where bad mixing can easily lead to problems.

2.08.7.1

Stirred Tank

Gas-phase mixing in a stirred tank will be largely dependent on the flow regime. Only in the case of full recirculation and T ¼ HL, intensive mixing of the gas phase will occur. Recirculating gas will mix, immediately after entering the vessel, with the gas entering the tank by coalescence in the cavities behind the stirrer blade. This is also the case for the so-called noncoalescing systems. At these conditions, the gas phase can be regarded as ideally mixed.2 Without full recirculation, mixing will be much less. In addition, recirculation downward to a lower compartment will hardly occur at HL > T. At these conditions, the gas phase can be regarded as near to plug flow. Gas-phase recirculation will also be dependent on bubble size. Smaller bubbles will more easily recirculate than larger ones due to their lower rise velocity.

2.08.7.2

Bubble Column

Nearly all commercial-scale bubble columns are of the type with HL > T or HL >> T. Most of the published experiments are done at diameters smaller than those of commercial-scale bubble columns and at high aspect ratios. The results show that the gas phase can be regarded as near to plug flow. It can be expected that this will not be different for commercial-scale reactors. The relation in the review of Heijnen and Van’t Riet16 also should lead to the conclusion that this will be the case. In large-scale low aspect ratio bubble columns, circulation velocities can easily be much higher than the bubble rise velocity and air will be entrained downward. Nevertheless, most air will be in upflow and the effect is limited.

2.08.7.3

Airlift

The riser of an airlift reactor can be described with the number of mixers or Peclet number at least the same and probably much more than that of a bubble column. This means that the gas can be regarded as plug flow. However, in the case of an internal airlift reactor in the recirculating gas regime, a considerable amount of air may be recirculated.

2.08.8

The Meaning of Mixing

2.08.8.1

Characteristic Times

It is the purpose of the mixing process to get a homogeneous distribution of those molecules or conditions that are important for the functioning of the microorganisms. As such, mixing time is not interesting as long as the distribution is even or even enough. Microorganisms do not know about mixing time, their only experience is their own microenvironment, which they may prefer not to be too variable. However, in some cases where microorganisms produce a product due to stress conditions, we prefer to exert stress conditions. To get an insight into the consequences of mixing, the principle of characteristic times is a useful tool. Characteristic times are reactor parameters and can be determined for a number of processes. Here, we will determine the values for heat production, oxygen mass transfer, and C-substrate consumption. For heat production, a characteristic time can be defined as the time needed to heat up the vessel content by 1  C, adding up all heat sources. Most of the heat is due to the metabolic activity, and for an aerobic fermentation at an oxygen uptake rate (OUR) of 100 mol m3 h1 this time is of the order of 1 h. For oxygen mass transfer, the value of 1/kLa is the characteristic time for saturation by mass transfer. For a fully loaded commercial fermentation, this value can be around 10 or 20 s. Another characteristic time is that of oxygen depletion of the broth, which is the time value of the actual oxygen concentration in the liquid divided by the OUR. For C-substrate consumption, the characteristic time can be defined as the ratio between the concentration in the liquid divided by the rate of C-substrate consumption (C-substrate depletion). Assuming Monod kinetics, the concentration in the microenvironment at half of the maximum growth rate is equal to the Monod constant Km. The value of Km is different for each substrate and microorganism. It can be less than 1 up to 10 g m3. For instance for a glucose consumption rate of 1 kg m3 h1, the characteristic time becomes less than 1 s up to around 10 s.

92

Mixing in Bioreactor Vessels Table 4

Order of magnitude of characteristic time valuesa

Mixing time Oxygen depletion Mass transfer Heating C-substrate (1) C-substrate (2)

Small scale 0.1 m3

Large scale 100 m3

10 10 20 3600 400 kDa

200–400 g Shell 5109 Shell

III III

2.5109

I/II

2–4 1010 Shell

I I

100 kDa 100 kDa

200–230 g 6.5109 70–120 g Shell 1–21010

Shell

I I

402

Membrane Bioreactors for Bioartificial Organs

improving the oxygenation and by adding filters, for preventing risk of cancer cell migration.45 Notwithstanding concerns related to the use of hepatoma cell line, that exhibit decreased liver-specific functions (i.e., ureagenesis and drug metabolism), over 250 subjects were treated in clinical trials with ELAD. Among these, the phase III VTI-208 was the largest, randomized, controlled, open-label trial to date, which has begun in 2013 and completed in 2015, enrolling 208 subjects with alcohol-induced liver decompensation (AILD) and severe acute alcoholic hepatitis (sAAH). Although pre-specified subsets based on age and lesser disease severity showed promising trends toward efficacy, VTI-208 failed to achieve its primary and secondary endpoints. Therefore, a second phase III trial, VTI-210, for subjects with sAAH, and a phase II clinical trial, VTI-212, for subjects with acute liver failure (ALF), both begun in 2014, were discontinued. Currently, a new phase III clinical trial, VTL-308, is enrolling subjects from May 2016. The Company expects to enroll at least 150 subjects at about 40 sites in the United States and Europe, and to report top-line results in mid-2018.46 HepatAssistÔ Circe Biomedical utilizes polysulfone (PSF) HF membranes with pore size of 0.2 mm (MWCO 3000 kDa) and 5  109 cryopreserved primary porcine hepatocytes loaded into the extracapillary space and attached to collagen-coated dextran microcarriers. The blood plasma passes through a charcoal adsorber and membrane oxygenator before entering the bioreactor, into the lumen of the HF membranes. This device was the first to be tested on a large clinical scale, with more than 200 subjects treated. In particular, in a phase III randomized controlled clinical trial, enrolling patients with fulminant and sub-fulminant liver failures from 20 sites in the United States and Europe, HepatAssistÔ demonstrated safety and improved survival in a post hoc subgroup analysis,47 but failed to demonstrate improved survival after 30 days in the overall study population.48 A more complex configuration was developed in the liver support system (LSS) by Gerlach et al.,49 in which four independent HF membranes are interwoven: polyamide (PA) for plasma inflow; PSF or PES for plasma outflow; hydrophobic PP or silicone membrane for oxygen supply and carbon dioxide removal; and hydrophilic PP for sinusoidal endothelial co-culture.49 In this device, primary hepatocyte aggregates are cultured on and between the fibers creating a three-dimensional environment for decentralized cell perfusion and gas exchanges with low gradients. The device underwent to phase I/II clinical trials.50 Thereafter, it was integrated into a modular extracorporeal liver support system (MELS), and combined with DetoxModule for albumin dialysis.51 MELS underwent to phase I clinical trials loaded with porcine52 as well as human hepatocytes harvested from donor livers found to be unsuitable for transplantation.53 Notwithstanding first encouraging results reporting on eight patients successfully bridged to liver transplantation, the device never progressed in controlled, randomized clinical trial required for regulatory approval. A different configuration was used for the design of a radial flow bioreactor (RFB), developed at the University of Ferrara, in which hepatocytes adhere on a polyester mesh between two sheets of polyester layers and are perfused by the patient’s plasma that flows from the center to the periphery of the device. RFB was tested in phase I clinical trials by using primary porcine hepatocytes54 and human hepatocytes. The Excorp Medical bioartificial liver support system (BLSS), developed at the University of Pittsburg, utilizes 70–120 g of primary porcine hepatocytes embedded in a collagen matrix in the extracapillary space of CA hollow fibers. This device was involved in phase I trials.55,56 TECA hybrid artificial liver support system (TECA-HALSS) is another BAL tested in phase I clinical trials.57,58 In this device, 1–2  1010 primary porcine hepatocytes are loaded in the extracapillary space of PSF HF membranes with MWCO 100 kDa. Spheroid reservoir bioartificial liver (SRBAL), developed by Nyberg’s group at Mayo Clinic, represents the second generation of BALs in which pig hepatocytes are cultured as spheroidal aggregates in a tank reservoir, instead of adhesion on supporting membranes. Spheroids are formed by a rocking technique allowing the multicellular aggregation of 3D polarized hepatocytes with bile canaliculi and preventing apoptosis. Two hollow fiber cartridges, containing membranes with MWCO ranging from 400 to 65 kDa, are integrated in the extracorporeal circuit, working as dialyzer for blood separation and as ultrafiltration to maximize removal of wastes, respectively. The device has been used in a pivotal preclinical trial.59 Human trials are planned but not yet underway so far. Recently, another clinical-scale BAL, using a fluidized bed bioreactor containing encapsulated three-dimensional organoids from a hepatoblastoma cell line, has been developed for GMP and tested in a preclinical test on a porcine model.60 Recent advances in stem cell technology enable for differentiating hepatocyte-like cells (HLCs) exhibiting highly specific liver functions. The promise on the potential use of HLCs as a feasible alternative for the treatment of liver failures seems to be in the near future.61,62 A radial flow bioreactor using HLCs induced from human fibroblasts is the first BAL system that has been tested in a preclinical trial on a pig model. To date, no stem cell-based BAL system has undergone clinical trial. Currently, BAL devices exhibit a limited clinical efficacy and improvements must be implemented in the device and trial. The progress achieved in the development of in vitro functional liver platforms can provide benefits for the advancement of BAL. The need of predictive biomarkers, the evaluation of the effect of shear stress and oxygen tension on hepatocyte function as well as the study of neuroprotective functions of the device in pathological state may be explored in vitro achieving information that may guide future research and development of a bioartificial liver as an approved therapy for liver failure.

2.28.4

Membrane Bioartificial Pancreas

There are over 400 million diabetics in the world today. Diabetes is one of the primary causes of death in high-income countries. Diabetes contributes to the death of 231,404 Americans annually. Diabetes mellitus is a disease in which high levels of sugar occur in the blood and urine. The cause of the raised sugar levels is insufficient secretion of the hormone insulin by the pancreas. In the absence of this hormone, the body’s cells are not able to absorb sugar from the blood stream in normal fashion, and the excess sugar

Membrane Bioreactors for Bioartificial Organs

403

is excreted in the urine. While therapeutic solutions such as injectable insulin allow diabetics to live longer, diabetes remains the third major killer, after heart disease and cancer. Diabetes is a very disabling disease, because the usual therapies do not control blood sugar levels constant preventing the swing between high and low blood sugar levels, which cause damages to other organs as kidneys, eyes, and blood vessels. Diabetes is classified into two main types: Type 1 diabetes or insulin dependent diabetes is usually associated with a complete lack of insulin brought about by autoimmune destruction of the insulin-producing beta cells.63 The events that cause Type 1 diabetes are unknown, but possibly there are viral or environmental triggers that act upon a genetically susceptible population. Type 2 diabetes or non-insulin dependent arises from peripheral resistance to insulin and a relative insufficiency of insulin, resulting in an initial attempt by the beta cells to compensate with release of higher than normal amounts of insulin. As Type 2 diabetes progresses, b cells become desensitized to persistently high glucose concentrations and normal responses to glucose signaling are lost.64 The treatment of diabetes mellitus was limited to dietary manipulation prior to the discovery of insulin by Banting and Best (1921). The discovery of insulin allowed converting an often rapidly fatal disease to a chronic condition requiring life-long treatment. Current treatment for diabetes, both Type 1 and Type 2, includes exogenous insulin therapy and endocrine replacement by transplantation. Both of these clinical approaches have considerable inherent drawbacks. Exogenous insulin treatment implies a poor control of blood glucose levels that leads to severe secondary complications such as retinopathy, neuropathy, nephropathy, and cardiovascular diseases.65 Alternative therapy is the pancreas transplantation. Since 1966, more than 30,000 pancreas transplants have been performed worldwide. Transplantation, however, requires major surgery and dependence on lifelong immunosuppression to prevent rejection. Because of the limited availability of human pancreases and the need for immunosuppression, relatively few pancreas transplants are done compared to the entire diabetic population. Improvements in surgical technique or immunotherapy are unlikely to make whole organ pancreas transplantation available to the majority of patients with diabetes. Islet transplantation promises to be a cure at least as effective as pancreas transplantation, while being much less invasive. The efficiency of islet recovery from the whole organ pancreas and the susceptibility of allogeneic islet to immune attack are the two major barriers to successful islet transplantation. There are approximately 1 million islets in an adult human pancreas thus islet transplantation usually requires islets isolated from two or more donor pancreases. Because islet isolation requires manipulation of human tissue, the process must be carried out in a good manufacturing process (GMP) facility, which adds to the expense of the procedure. Insulin-dependent diabetes mellitus is a chronic disease characterized by high blood glucose levels and long-term complications, including micro- and macroangiopathic lesions leading to retinopathy, neuropathy, and nephropathy. It is an autoimmune disease that results from the destruction by the patient’s own immune system of the islets of Langerhans, which are responsible for insulin production and secretion. Endogenous insulin is almost nonexistent; therefore, exogenous insulin must be administered. This approach in the long term results in the so-called ‘diabetic complications’ such as kidney failure, neuropathy, which lead to the patient’s severe incapacity or death. Transplantation of the whole pancreas, pancreatic tissue fragments, or islets of Langerhans would assure the required glucose-related insulin secretion, but the rejection of implants and scarce availability of donor organs limit this therapeutic approach. An alternative approach is the development of membrane bioartificial pancreas (BAP) using isolated islets of Langerhans or single beta cells, which are capable of sensing the plasmatic glucose concentration and produce insulin amounts related to the actual glycemia, entrapped by means of membranes.

2.28.4.1

Requirements of a Membrane Bioartificial Pancreas

The idea to develop a membrane bioartificial pancreas encapsulating cells in a semipermeable membrane dates back to 1933. In 1975, W. L. Chick and colleagues transplanted isolated islets protected by hollow-fiber UF membrane (an acrylonitrile-vinyl chloride copolymer) into dogs made diabetic by surgically removing the pancreas. The device consists of a chamber through which passes a copolymer membrane connected to standard vascular grafts. Islets are placed inside the chamber, through ports in the housing into the cavity, but are outside of the bloodstream. Nominal molecular porosity of 80,000 Da permits free diffusion of nutrients and insulin across the membrane but inhibits the entry of immunoglobulins and immunocytes from the blood stream into the chamber66 (Fig. 2A). Since 1975, research efforts were devoted to the development of a hybrid bioartificial membrane pancreas based on encapsulation system. In particular, the research has been focused on the approaches to reduce immunogenicity, to optimize capsules dimension and implant site and to choose appropriate cell source. The requirements of an encapsulation system involve (i) the maintenance of viability and function of encapsulated pancreatic cells, (ii) the sensitivity to stimulus and rapidity in reaction of cells that rapidly detect the increase of blood glucose concentration and promptly release insulin, (iii) immunoisolation, to avoid the encapsulated cells coming into contact with immunocompetent species present in the patient’s blood, (iv) biocompatibility of the materials used for the capsule preparation in order to activate inflammatory processes; (v) ease of implantation through minimally invasive surgical procedures; (vi) retrievability, which should be removed in an easy and safer manner in case of failure. The choice of polymer to be used for encapsulation is a critical issue in the development of systems. Optimal pore size and a proper membrane thickness are required to provide immunoprotection and a correct nutrient diffusion. The approaches undertaken in the preparation of membranes are focused to obtain highly selective membranes with high diffusivity of the nutrients with low molecular weight and low diffusivity of species with high molecular weight. The molecular weight cut-off of membranes used for cell encapsulation ranges from 50 to 150 kDa. Indeed some membranes retain also IgG preventing only the contact with cells of immune system. Three different strategies have been developed to encapsulate islet cells within the semipermeable membrane based on the dimensions of the implantable capsules: macro-, micro- and nano-encapsulation. Macrocapsules of the order of centimeters can

404

Membrane Bioreactors for Bioartificial Organs

Figure 2 Schematic of bioartificial pancreas configurations: (A) device developed by Chick et al., pancreatic cells are loaded outside the tubular membrane; (B) pancreatic cells are loaded inside of hollow fiber in lumen compartment; (C) microencapsulated cells are loaded in the lumen compartment of hollow fiber membrane; (D) pancreatic cells loaded outside of hollow fibers in extracapillary compartment; (E) cells loaded outside of tubular membrane; (F) microencapsulated pancreatic cells; (G) pancreatic cells cultured in islet sheets between flat-sheet membranes; and (H) pancreatic cells inside coating. Reprinted by Ref. 4.

encapsulate a large number of pancreatic cells. Depending on the implant site, macroencapsulation systems can be distinguished into extravascular, typically placed in the peritoneal cavity or subcutaneously, or intravascular, which are connected to the patient’s cardiovascular system via a shunt. Extravascular macrocapsules can be made of different materials: one for the ECM enabling islet viability,67 another for the immunoisolating membrane, and an external surface coating to promote neovascularization. Their main drawbacks involve low surface-to-volume ratios, large diffusion distances and constraints related to the eligible implant site. Intravascular macrocapsules allow implanted pancreatic islets to be in direct contact with blood vessels, providing a direct access to nutrients and oxygen as well as to track blood glucose level and to reduce insulin absorption delay.68 Microcapsules having diameter of 250–1000 mm consist of a semipermeable membrane enclosing a small number of islets of Langerhans (usually 1–3), suspended in a polymeric gel matrix. Most common polymers are alginate, chitosan, agarose, polyethylene glycol (PEG), copolymer of acrylonitrile. The smaller dimensions and the higher surface-to-volume ratios enhance the molecules diffusion and enable the implantation by using minimally invasive procedures in locations such as the peritoneal cavity, the kidney capsule or subcutaneous sites.69,70 Disadvantages of microencapsulation include the limited vascularization and difficulty to retrieve due to the capsule size in case of failure. Nanoencapsulation devices have been investigated in order to exploit narrower implant sites and overcome diffusion related issues. Nanocapsules having a diameter less than 100 mm can envelop single pancreatic b cells in a semipermeable membrane, directly bound to the cell membrane with a thickness varying between some nanometers to some micrometers. Nanocapsules are particularly advantageous in terms of nutrients diffusion and insulin release, thanks to their really small dimensions. Different polyelectrolytes have been successfully investigated to this aim.71,72 Considering that the diameter of the encapsulated islets must be much smaller than that currently attained to allow for transplantation, several efforts have been made in order to develop coating techniques that permit the coating of islets with very thin membrane. Islets can be covered with a thin polyion complex membrane using layer-by-layer method or with an alginate/PLL/alginate multilayer coating or with PEG. The advantage to use coating technique or conformal coating is to reduce the diameter of microcapsules.73

Membrane Bioreactors for Bioartificial Organs

405

Surface modification is a form of immunoisolation that does not rely on microencapsulating islets, but rather on altering the surface of the islets themselves to form an immune barrier. A PEG complex may be covalently bound to the surface of the islets to provide a thin barrier to macrophages and reduce the release of cytokines. For example, a succinimidyl ester-functionalized PEG can react with amine groups present upon the cell surface of pancreatic islets to conceal host immunogenic surface antigens.82 As an alternative approach, an electrostatic adsorption has been proposed, in which poly(L-lysine)-graft-poly(ethylene glycol) (PLLg-PEG) copolymers were physically coated onto islet interfaces. Although this approach can attain noncovalent surface modification of pancreatic islets, the inherent cytotoxicity of PLL polymers hampers the safety of this approach. To create a functional coating for improving response to glucose, a layer-by-layer (LbL) self-assembly technique has been studied.

2.28.4.2

Membrane Devices Developed as BAP

Different BAP were designed in four physical types: hollow fibers, capsule, coatings, and sheet. Table 4 reports the different devices developed and the membrane features.74–82 The many different types of prosthetic devices proposed can be grouped into three main categories: extravascular devices, intravascular devices, and microencapsulated islets of Langerhans. In the first case, the tissue is enclosed between membranes, if in a flat-sheet configuration, or in the lumen of HFMs, and then implanted in an extravascular site (see Fig. 2). Hollow-fiber devices have been proposed as extravascular BAP. This device consists of HFMs containing in the lumen islets of Langerhans and microencapsulated cells (Fig. 2B–C). The extravascular systems generally suffer from an intrinsically slow insulin response following changes of blood glucose concentration, limited by the purely diffusive mass transport and by the fibroblastic response of the host. Vascular hollow-fiber devices as BAP by using polysulfone membranes with MWCO of 100,000 Da and islets of Langerhans have been developed and established as good systems, for in vitro testing.80 Intravascular membrane devices are designed so that the membrane separates the graft directly from blood stream of the host. In Fig. 2D, cells are cultured outside HFMs arranged in a housing in a shell-and-tube configuration. These devices suffer from blood clotting at the interface between blood and the synthetic material of the membranes or the point of access, but they are extremely attractive in terms of flexibility of design and use. Additionally, the implant site can be chosen on the basis of reducing the response time of the prosthesis following an increase of blood glucose concentration. A device in preclinical development licensed from Circe Biomedical is the PancreAssist Bioartificial Pancreas System.83 This device consists of a single tubular membrane surrounded by insulin-producing porcine islets, which are, in turn, enclosed within a disk-shaped housing (Fig. 2E). The porous tubular membrane permits the transport of nutrients and glucose to cells and the transport of insulin from cells to blood. Membrane prevents also the contact between immunological species present in the patient’s blood and islets. This device should be implanted near the kidney and surgically connected directly to circulatory’s system using vascular graft. In the case of microencapsulated islets, the membrane in the form of alginate gel is formed around the islets of Langerhans, thus obtaining microcapsules with diameters of 300–400 mm (Fig. 2F). Microcapsules may make better implants than hollow fibers because they offer better conditions for diffusion of nutrients to the insulin-producing cells and waste products from them. Normoglycemia has been reported after intraperitoneal transplantation of alginate–polylysine microencapsulated allogenic

Table 4

Membrane bioartificial pancreas devices

Bioreactor configuration

Membrane material

Membrane configuration

MW cut-off

Culture technique

Tubular

50,000 Da

Immobilization in lumen space Microencapsulated islets between flat-sheet membrane Immobilization in lumen space

Microcapsules76

Poly-L-ornithine

Microcapsules77 Hollow fiber78

Alginate-polylysine Polysulfone

Microcapsule79 Hollow fiber80

Amynopropyl-silicate Polysulfone

Hollow-ribbon nitinol scaffold81

Amphiphilic conetwork

Flat-sheet membrane thickness¼ 25–150 mm Hollow fiber D¼ 0.5 mm Circular D¼ 200 mm Circular Capillary ID¼ 600–800 mm; wall thickness¼100 mm Circular Hollow fiber ID¼1 mm Thickness 280 mm Fiber Thickness 5–10 mm

80,000 Da

Hollow fiber75

Acrylonitrile-vinyl chloride copolymer Cellulose ester filter alginate film Cuprophan

Hollow fiber66 Planar islet sheets74

50,000 Da 50,000– 70,000 Da 50,000 Da 50,000 Da 60,000 Da 100,000Da

Encapsulation Encapsulation Macroencapsulated in intracapillary space Coated alginate beads Sepharose microspheres

Pore size 11 nm Immobilization

406

Membrane Bioreactors for Bioartificial Organs

and xenogenic islets in diabetic animal models and also in men. However, graft survival is always limited to several weeks. Graft failure is interpreted as nonspecific immune response, that is, foreign body reaction against the microcapsules resulting in a progressive overgrowth of the capsule and subsequent necrosis of islets. Researchers focused study on highly purified alginate and other biochemicals more biocompatible and on novel membranes able to prevent permeation of low-molecular-weight humoral molecules released by xenogeneic islets. To this purpose, new capsules with poly-L-ornithine membranes76 and membranes composed of aminopropylsilicate with an MWCO of approximately of 60,000 Da80 and of polysulfonepolyvinylpyrrolidone with an MWCO of 50,000 Da79 have been developed. One major limitation to the encapsulation device is that they are incapable of efficiently encapsulating large numbers of islets in a reasonable amount of time. This may result in hypoxic stress and loss of functionality to islets in larger scaled-up experiments.77 A proposed alternative microencapsulation method, which has the advantage of rapidly encapsulating large numbers of islets into microcapsules, utilizes multichannel air jacket microfluidic devices.84 A reduction in capsule size would benefit the islet and also exponentially decrease the total transplant volume. The smaller the diameter of the capsules, the better the diffusion of nutrients to the islets, and Omer et al. demonstrated that capsules with a diameter of 600100 mm showed improved stability in vivo over larger capsules with diameters of 1000100 mm.85 Furthermore islets are clinically transplanted into the liver through the portal veins and microcapsules with large diameter are expected to plug vessels.

2.28.4.3

Membrane BAP in Clinical Trials

The most recent and successful micro-encapsulation systems in clinical trials is represented by DIABECELL consisting of neonatal pig islets encapsulated in an alginate-based matrix and implanted in the peritoneal cavity that revealed a 70% reduction in hypoglycemic events and a 20% reduction in insulin dose.86 Other systems are represented by macrocapsule systems that can be distinguished in extravascular and intravascular based on their transplant location. Intravascular systems contain islets which are seeded enclosed within a larger tube and implanted into the vessels of the host. The hollow fibers are perfused by blood flow. This device has been successful in inducing normoglycemia in various diabetic animal models including rats, dogs and monkeys,87 although it requires intense systemic anticoagulation due to direct contact of the material with blood. Extravascular devices in contrast have the advantage that biocompatibility issues do not pose a serious risk to patient. They have been designed in both flat-sheet membrane and hollow fiber forms. A semipermeable membrane around the sheet allows diffusion of nutrients and secreted hormones but not macrophages. They are usually coated with hydrogels to achieve a smooth outer surface to improve biocompatibility. Initial studies with extravascular macrocapsules have encapsulated multiple islets in one or several large capsules. Islets aggregated in large clumps were not successful, due to necrosis at the center of the clumps.88 Later, this problem was addressed by immobilization of islets in a matrix before encapsulation. However, there are a number of products under development or on the market mainly extravascular macrocapsule devices (Table 5). bAir device is a subcutaneous insulin bioreactor containing a multilayer immunoprotective membrane of alginates and a PTFE membrane. Separated by gas-permeable membranes, two compartments surround the central gas cavity that houses alginate-immobilized pancreatic islets.89 TheraCyteÔ system produced by Baxter Healthcare consisting of two PTFE membranes bound to a polymeric structure: the exterior membrane has tighter pores promoting neovascularization, whereas the inner one provides for pancreatic cell immunoisolation. The membranes have different pore size: the inner 30-mm thick membrane with 0.4 mm pore size for selectivity and immunoisolation and its outer 15-mm thick membrane with 5 mm pore size for angiogenesis through an alternative foreign body response.90 ViaCyte, Inc. (San Diego, CA) is using a modified TheraCyte membrane (Encaptra) as an immune barrier in order to protect stem cells-derived b cells from the host immune system: VC-01. The cells are expected to differentiate in mature pancreatic cells.91 Cell Pouch SystemÔ is a polymeric macrocapsule devised for subcutaneous implantation that can reproduce a natural environment in the body for the long-term survival and function of therapeutic pancreatic cells. Table 5

Membrane BAP in clinical trials

Bioreactor configuration DIABECELL®86 bAir® device89 TheraCyte™82 VC-01TM 91 Cell Pouch System™ 92 Islet sheet device67

Membrane material

Membrane configuration Culture technique

Clinical trial

Alginate–poly-lysine–alginate microcapsules Multilayer membrane of alginates polytetrafluoroethylene Polytetrafluoroethylene (pore size of 0.4 mm) laminated to an outer membrane and covered by a loose polyester mesh Undisclosed polymers Polymeric membranes Cellulose ester filter alginate film

Microcapsules Flat sheet Flat sheet

Microencapsulation Phase III Macroencapsulation Phase I Macroencapsulation Preclinical phase

Flat sheet Circular Flat sheet

Macroencapsulation Phase I/II Macroencapsulation Phase I/II Macroencapsulation Phase I/II

Membrane Bioreactors for Bioartificial Organs

407

Sernova’s Cell PouchÔ consists of a multi-channel sheet inserted with an array of rods.92 The pouch creates a favourable pre-vascularized environment but it does not offer an immune-barrier to protect cells from the host immune system. This device is first placed under the skin for better vascular integration with the surrounding tissues for a month. Islet sheet is a thin planar BAP, licensed from Islet Sheet Medical Company, and contains live, functional islets in an artificial polymer matrix93 (Fig. 2G). Each sheet is several centimeters in diameter and contains 2–3 million cells microencapsulated within a mesh to increase the physical strength between two layers of semipermeable alginate membrane. A semipermeable membrane such as 0.2-mm cellulose ester filter membrane is saturated with cross-linking solution. The 4- to 6-sheet cells contain enough islet tissue to cure diabetes in an adult. The sheet is so thin (the overall thickness is 250 mm) that diffusion alone allows sufficient nutrients to reach the center of the sheet. A coat on the exterior of the sheet prevents contact between the cells inside and immune effector cells of the host besides inhibiting diffusion of antibody and complement. The alginate membranes show high permeability to different solutes and excluded immunocompetent species. No immune suppression drugs are needed. The sheet may be removed or replaced at any time. Large animal studies give encouraging results. The implantation of islet sheet in omentum of pancreatectomized dog permitted to return blood sugar to normal and at 60 days the blood sugar was lower at every measurement time.93 Human fetal pancreatic islet-like cell clusters have been enclosed in devices made of PTFE with pore size of 0.4 mm and were transplanted at a subcutaneous site. Ten weeks after transplantation in a non-obese diabetic mice, differentiated beta-cell progenitors were found into device and the glucose level was normalized indicating efficacy.

2.28.5

Membrane Bioartificial Kidney

The kidney was the first solid organ whose function was substituted by an artificial device. It is estimated that more than 10% of the worldwide population suffers from a more or less severe form of kidney disease. With the increased prevalence in risk factors, such as hypertension, cardiovascular disease and diabetes mellitus in aging population, the prevalence of chronic kidney disease (CKD) is rising. It has been recorded that 16% of the US general population of people aged 20 and older are suffering from CKD in 2014. According to this report, this prevalence of CKD in the United States has been continuously increasing from 13% in 1999 to 16% in 2014. The kidney function of CKD patients may progressively and irreversibly decline until total loss, called early-stage kidney disease (ESKD), which leads to the accumulation of a variety of endogenous metabolites with life-threatening consequences. Current therapy for ischemic or toxic acute kidney injury (AKI) or acute tubular necrosis (ATN) is predominantly supportive by haemodialysis (HD) or peritoneal dialysis (PD), which represents the only successful long-term ex vivo organ substitution therapy. Patients with AKI still have high mortality rate of greater than 50%, due to the propensity of these patients to develop systemic inflammatory response syndrome. In particular, CVD (cardiovascular disease) is the leading cause of morbidity and mortality in end-stage renal disease (ESRD): approximately 50% of ESRD patients die from CVD, and cardiovascular mortality is 15–30 times higher than in the age-adjusted general population. In ESRD, in addition to traditional Framingham risk factors, a considerable number of non-classical factors are known to play a role in the CVD progression, such as inflammation, vascular calcification and left ventricular hypertrophy. Although haemodialysis or peritoneal dialysis has dramatically changed the prognosis of renal failure, it cannot be considered a complete replacement therapy, because it provides only filtration function and does not replace the homeostatic, regulatory, metabolic, and endocrine functions of the kidney. Therefore dialysis should be considered as partial substitution rather than renal replacement therapy. The addition of metabolic activity, such as ammoniagenesis and glutathione reclamation; endocrine activity, such as activation of vitamin D3 (low levels of which seem to correlate with high mortality rates in hospitalized patients); immunoregulatory support and cytokine homeostasis may provide additional physiologic replacement activities to the current history of the disease. The development of an implantable bioartificial kidney composed of both biologic and synthetic components could result in substantial benefits for patients by increasing life expectancy, mobility, and quality of life, with less risk of infection and reduced costs.

2.28.5.1

Requirements of Bioartificial Kidney Devices

A bioartificial kidney like human kidney requires two main units, the glomerulus and the tubule, to replace excretory and metabolic functions of the kidney94,95 (Fig. 3). Currently the haemodialysis and peritoneal dialysis are the methods for the replacement of kidney functions. These methodologies replace the excretory functions of the kidney but they do not provide the lost metabolic function. In order to develop a more complete and functional device, the last decade has been focused on engineering of a bioartificial kidney (BAK). Initially, several efforts have been focused on the development of an extracorporeal BAK constituted of a conventional synthetic hemofilter with a renal tubule cell assist device (RAD) in an acute extracorporeal blood circuit. A bioartificial tubule has been constructed utilizing renal tubule progenitor cells96,97 cultured on semipermeable hollow-fiber membranes coated with extracellular matrix to enhance the attachment and growth of epithelial cells. The hollow fiber synthetic membranes provide immunoprotection and also an architectural scaffold for cells in the long-term implantation in a xenogeneic host.98

408

Membrane Bioreactors for Bioartificial Organs

Figure 3 Schematic of bioartificial kidney configurations: (A) BAK; (B) Biokid; (C) implantable bioartificial kidney; (D) WEBAK; (E) lab-on-a-chip bioreactor system. Reproduced with permission by De Bartolo, L.; Curcio, E.; Drioli, E. Membrane Systems: For Bioartificial Organs and Regenerative Medicine; De Gruyter: Berlin/Boston, 2017; pp 1–264, Fig. 7.7.

Humes et al. scaled up the single hollow-fiber device in a multifiber bioartificial RAD utilizing porcine renal proximal tubule cells grown along the lumen of polysulfone hollow fiber membranes99 (Fig. 3). The device consists of hollow fibers with membrane surface areas as large as 0.7 m2, containing up to 108 renal tubule cells into the lumen. This device demonstrated in in vitro studies the retention of differentiated active transport of sodium, bicarbonate, glucose and organic anions as well as important differentiated metabolic processes of the kidney such as ammoniagenesis, glutathione metabolism and synthesis of 1,25-dihydroxyvitamin D3. The RAD was used in series with a hemofilter in an extracorporeal hemoperfusion circuit in an acutely uremic dog. The blood was pumped out and entered the fibers of a hemofilter, where it is ultrafiltrated. Then the ultrafiltrate was delivered into the fibers of the tubule lumens within the RAD and is collected and discarded as “urine.” The processed blood exiting by RAD was delivered back to the animal. The tubule unit was able to maintain differentiated renal functional performance because metabolic substrates and low-molecular-weight growth factors are delivered to the cells from the hemofilter and the blood in the extracapillary space. Membranes of suitable MWCO protect cells seeded into the lumen from immunoglobulins and immunocompetent cells present into the blood. Clinical experience has been made with RAD in patients with acute kidney injury. Human kidney cells were used in the RAD and the initial results in the first 10 treated patients in the phase I/II trial demonstrated that this device is efficient when used in conjunction with hemoperfusion.100 Cardiovascular stability of the patients was maintained and increased native renal function. Of these 10 patients, 6 survived past 28 days with renal function recovery. In the phase II clinical study, of 58 patients with AKI requiring continuous veno-venous hemofiltration, 40 patients received hemoperfusion together with RAD and 18 patients received only hemoperfusion. RAD treatment for up to 72 h promoted a statistically significant survival advantage over 180 days of follow-up. The research related to the development of BAK has been focused on the use of renal proximal tubule-derived cells (Table 6). Human primary renal proximal tubule cells (HPTCs) have been used in the clinical trials.100,101 Differently, most of the experimental work and the animal have been performed with porcine primary renal proximal tubule cell.102–105 Also, the proximal tubule-like porcine cell line LLC-PK1106,107 and other animal-derived cell lines such as Madin–Darby canine kidney (MDCK) cells have been used in BAK system. It is very critical to form a confluent differentiated epithelium sealed by tight junctions by the renal cells on the porous membranes of the device in order to compromise the cellular functions.

Membrane Bioreactors for Bioartificial Organs Table 6

409

Characteristics of Membrane Bioartificial kidney

Type of bioartificial system

Membrane configuration Membrane material

Cell capacity Cell type

Cell compartment

Renal tubule assist device (RAD)96,97 BAK101 Bioartificial renal epithelial cell system (BRECS)108 Bioartificial renal tubule devices (BTD)116

Hollow fiber

Polysulfone-PVA

108

Intraluminal space

Hollow fiber Flat

PSF/PVP ECM coated 108 Niobium-coated carbon 108 disk Ethylene vinyl alcohol 3–7108 (EVAL)

2.28.5.2

Hollow fiber

Proximal tubule renal derived cells HPTCs Renal epithelial cells

Intraluminal space Disk surface

Human renal proximal Intraluminal space tubular cells

Membranes for BAK

The membranes used for the development of a bioartificial kidney are mainly commercial hollow fiber membranes of PSF/PVP, which are designed for haemodialysis/hemofiltration. These membranes seem to be not appropriate for cell adhesion probably due to the PVP component that generates problems in HPTC growth and survival. For this reason the membranes are coated with ECM in order to improve cell adhesion.108 Membranes that are currently employed in haemodialysis/hemofiltration are optimized to be in contact with blood, and usually hydrophobic membranes such as membranes consisting of pure PSF are modified with hydrophilic additives in order to prevent protein adhesion. On the other hand the protein adsorption favors the cell adhesion. For this reason Ueda et al. suggested the use of asymmetric membranes with one hemocompatible and one cytocompatible surface.109 These authors described a membrane consisting of PSF blended with a phospholipid polymer, which was asymmetrically distributed between the skin and the sponge layer of the membrane. In general, it would be preferable to use the relatively rough sponge layer for cell growth and to expose the smooth skin layer to the blood. Alternatively asymmetric membranes can be realized by using coating of the surfaces with antifouling agents such as polyethylene glycol on the blood exposed side and adhesive coatings on the cell-exposed side. Dual-layered membranes with each layer composed of a different material would be interesting. Recent findings revealed problems with HPTC survival and differentiation on a variety of commercially available membrane materials and surface treatments and ECM coatings lead improvements. HPTCs formed confluent epithelia on membranes consisting of PSF blended with FullCure (FC) under bioreactor conditions. Single or double coating did not further improve cell performance on such PSF-FC membranes. Growth and differentiation of primary human cortical tubular epithelial cells were also observed on collagen IV-coated thin film and nanostructured materials.110 A living membrane was developed by Jansen et al. that cultured human conditionally immortalized proximal tubule epithelial cell (ciPTEC) monolayers on biofunctionalized MicroPES (polyethersulfone) hollow fiber membranes.111

2.28.5.3

Preclinical Bioartificial Kidney Devices

Research about the bioartificial kidney has been performed by two different groups: Humes et al.94 and Saito et al.112,113 Clinical trials with BAKs have been performed by Humes et al.94 In this device the patient’s blood (red) first enters the hemofiltration unit (left), which contains hollow fiber membranes for ultrafiltration (Fig. 3A). The blood and the ultrafiltrate (yellow) leaving the hemofiltration unit then flow into the bioreactor unit (right), which contains hollow fiber membranes with an epithelium of renal cells (green) on the inner surfaces. The cells secrete molecules, which become enriched in the ultrafiltrate during processing and in the blood flowing on the outside of the hollow fiber membranes. The blood enriched flows back into the patient, and the processed ultrafiltrate (orange) is discarded. An enlarged cross-section of a hollow fiber membrane from the bioreactor unit is shown in the lower right corner. The ultrafiltrate flows in the lumen of the hollow fiber membrane, and the blood flows on the outside. The inner surface of the hollow fiber membrane is covered with secreting renal cells, and enriched ultrafiltrate and the blood in the bioreactor unit. BMP-7 in the ultrafiltrate would regulate HPTC performance, whereas BMP-7 in the bloodstream would be delivered to the patient. One of the bioartificial kidney developed in the last years is the device realized by the group of the University Medical Center Groningen called BioKid114 (Fig. 3B). The bioreactor consists of hollow fiber membranes that are coated in the lumen side where nephron cells are cultured. The bioreactor mimics the function and operation of nephrons ensuring the removal of toxins that remain after haemodialysis. Thus, the device improves the quality of haemodialysis treatments reducing the risk of complications such as cardiovascular problems resulting from the accumulation of toxic waste products. Vanderbilt and the University of California at San Francisco are the lead institutions working to develop an implantable bioartificial kidney115 (Fig. 3C). The bioartificial kidney is made of two parts – a filter side and a cellular side. On the filter side, silicone membranes with microscopic pores separate toxins from the blood, much as dialysis machines do. The body’s own blood pressure forces blood through the filter, so no pumps is needed. The key to the filtration side is the silicone membrane, which can be made fairly inexpensively and precisely, much as computer chips are. On the cellular side, the filtered blood is pumped over a bed of cells taken from either the patient’s own failing kidneys or from a donor. The cells sense the chemical makeup of the filtered blood and trigger the body to maintain appropriate levels of salt, sugar and water. After entering the device the patient’s blood passes through components. First, silicone membrane filters the blood removing toxins, and in the second stage, a bed of cells regulates the

410

Membrane Bioreactors for Bioartificial Organs

chemical balance of the filtrated blood reabsorbing much water, sugar and salts. The toxins and excess of water are passed into the waste outlet connected to the bladder. In Table 6 are reported the main characteristics of membrane bioartificial kidney. Wearable bioartificial kidney (WEBAK) combining peritoneal dialysis with a bioartificial renal epithelial cell system (BRECS) has been developed108 (Fig. 3D). The BRECS is a perfusion bioreactor that utilizes primary renal epithelial cells derived from the kidney, expanded from progenitor cells during in vitro culture. Cells are seeded on porous disks, which are placed within a media flow path within the BRECS. WEBAK utilizes peritoneal fluid to maintain cell viability and functionality and comprises the use of sorbentbased technologies to replace the excretory function of the kidney and the compact BRECS described above to replace the metabolic function of the kidney. Lab-on-a-chip bioreactor system has been developed by using human renal proximal tubule epithelial cells (RPTEC) cultured in the luminal space of the single hollow fiber of PES-PVP precoated with fibrin109 (Fig. 3E). To engineer the tubule, a tuneable hollow fiber membrane with an exterior skin layer was used in order to provide immunoprotection for the cells from extracapillary blood flow and a coarse inner surface that facilitates a hydrogel coating for cell attachment. The hollow fiber membrane is located inside polydimethylsiloxane (PDMS) body and a glass substrate. The inner surface of hollow fiber is featured with pores of up to 0.3 mm in size, facilitating the attachment of extracellular matrix coating on the membrane. Together with a fibrin coating, a confluent monolayer of human kidney renal proximal tubule epithelial cells is successfully formed on the fiber inner surface under flow conditions in a “lab-on-a-chip” bioreactor system.

2.28.6

Conclusions

Extensive progress has been made in the field of the organ bioengineering in particular liver, pancreas and kidney. The greatest impact is being made by combining the biomaterials such as membranes and bioreactors in a manner which emulates enough of the environment present during development to promote the development of new tissues and organs with the necessary functionality. However, many issues must be addressed to achieve the development of an organ that recapitulates the in vivo organ microarchitecture and functions. Various improvements are needed in terms of biofabrication of more physiologically relevant cellular microenvironments, including the mass transport issues related to the delivery of nutrients and oxygen. Vascularization remains a critical issue. One of the greatest challenges encountered in the field has been the need to have a functional vascular network for complex tissues and organs. In this chapter, membrane technology offers several advantages over more conventional approaches including the capability to create a permissive environment for the development of a complex organ. The modulation of the membrane structural, physico-chemical, mechanical and transport properties can lead the achievement of a biomimetic cell environment, which is essential for the development of a functionally active organ and tissue.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

United Network for Organ Sharing on 15 January 2019. De Bartolo, L.; Curcio, E.; Drioli, E. Membrane Systems : For Bioartificial Organs and Regenerative Medicine ; De Gruyter: Berlin/Boston, 2017; pp 1–264. Freed, L. E.; Vunjak-Novakovic, G. Methods of Tissue Engineering ; Academic Press: San Diego, 2002; pp 97–110. Giorno, L.; De Bartolo, L.; Drioli, E. Membrane Bioreactors. In Comprehensive Biotechnology, Vol 2; Murray Moo-Young, Ed.; 2nd ed; Elsevier, 2011; pp 263–288. Curcio, E.; De Bartolo, L.; Barbieri, G.; Rende, M.; Giorno, L.; Morelli, S.; Drioli, E. Diffusive and Convective Transport through Hollow Fiber Membranes for Liver Cell Culture. J. Biotechnol. 2005, 117, 309–321. Khakpour, S.; Di Renzo, A.; Curcio, E.; Di Maio, F. P.; Giorno, L.; De Bartolo, L. Oxygen Transport in Hollow Fiber Membrane Bioreactors for Hepatic 3D Cell Culture: a Parametric Study. J. Membr. Sci. 2017, 544, 312–322. Curcio, E.; Piscioneri, A.; Morelli, S.; Salerno, S.; Macchiarini, P.; De Bartolo, L. Kinetics of Oxygen Uptake by Cells Potentially Used in a Tissue Engineered Trachea. Biomaterials 2014, 35, 6829–6837. Morelli, S.; Salerno, S.; Rende, M.; Lopez, L. C.; Favia, P.; Procino, A.; Memoli, B.; Andreucci, V. E.; d’Agostino, R.; Drioli, E.; De Bartolo, L. Human Hepatocyte Functions in Galactosylated Membrane Bioreactor. J. Membr. Sci. 2007, 302, 27–35. Salerno, S.; Piscioneri, A.; Laera, S.; Morelli, S.; Favia, P.; Bader, A.; Drioli, E.; De Bartolo, L. Improved Functions of Human Hepatocytes on NH3 Plasma–Grafted PEEK-WCPU Membranes. Biomaterials 2009, 30, 4348–4356. De Bartolo, L.; Morelli, S.; Lopez, L.; Giorno, L.; Campana, C.; Salerno, S.; Rende, M.; Favia, P.; Detomaso, L.; Gristina, R.; d’Agostino, R.; Drioli, E. Biotransformation and Liver Specific Functions of Human Hepatocytes in Culture on RGD-Immobilised Plasma-Processed Membranes. Biomaterials 2005, 26 (21), 4432–4441. Demetriou, A. A.; Arnaout, W. S.; Backfish, G.; Moscioni, A. D. In Artificial Liver Support; Brunner, G., Mito, M., Eds., 2nd ed.; Springer: Berlin, 1993; pp 283–295. De Bartolo, L.; Morelli, S.; Bader, A.; Drioli, E. Evaluation of Cell Behaviour Related to Physico-Chemical Properties of Polymeric Membranes to Be Used in Bioartificial Organs. Biomaterials 2002, 23 (12), 2485–2497. De Bartolo, L.; Catapano, G.; Della Volpe, C.; Drioli, E. The Effect of Surface Roughness of Microporous Membranes on the Kinetics of Oxygen Consumption and Ammonia Elimination by Adherent Hepatocytes. J. Biomater. Sci. Polym. Ed. 1999, 10, 641–655. Matsumura, K. N.; Robles Guevara, G.; Huston, H.; et al. Hybrid Bioartificial Liver in Hepatic Failure: Preliminary Report. Surgery 1987, 101 (1), 99–103. Hu, W. S.; Friend, J. R.; Wu, F. J.; et al. Development of a Bioartificial Liver Employing Xenogeneic Hepatocytes. Cytotechnology 1997, 23, 29–38. Nyberg, S. L.; Remmel, R. P.; Mann, H. J.; et al. Primary Outperform Hep G2 Cells as Source of Biotransformation Functions in a Bioartificial Liver. Ann. Surg. 1994, 220, 59–67. Shiraha, H.; Koide, N.; Hada, H.; Ujike, K.; Nakamura, M.; Shinji, T.; Gotoh, S.; Tsuji, T. Improvement of Serum Amino Acid Profile in Hepatic Failure with the Bioartificial Liver Using Multicellular Hepatocyte Spheroids. Biotechnol. Bioeng. 1996, 50, 416–421. Naka, S.; Takeshita, K.; Yamamoto, T.; Tani, T.; Kodama, M. Bioartificial Liver Support System Using Porcine Hepatocytes in a Three Dimensional Hollow Fibre Module with Collagen Gel: an Evaluation in the Swine Acute Liver Failure Model. Artif. Organs 1999, 23, 822–828.

Membrane Bioreactors for Bioartificial Organs

411

19. De Bartolo, L.; Jarosch-Von Schweder, G.; Haverich, A.; Bader, A. A Novel Full-Scale Flat Membrane Bioreactor Utilizing Porcine Hepatocytes: Cell Viability and Tissue Specific Functions. Biotechnol. Prog. 2000, 16 (1), 102–108. 20. Nagaki, M.; Miki, K.; Kim, Y.; et al. Development and Characterization of a Hybrid Bioartificial Liver Using Primary Hepatocytes Entrapped in a Basement Membrane Matrix. Dig. Dis. Sci. 2001, 46, 1046–1056. 21. Jasmund, L.; Langsch, A.; Simmoteit, R.; Bader, A. Cultivation of Primary Porcine Hepatocytes in an OXY-HFB for Use as a Bioartificial Liver Device. Biotechnol. Prog. 2002, 18, 839–846. 22. De Bartolo, L.; Salerno, S.; Curcio, E.; et al. Human Hepatocyte Functions in a Crossed Hollow Fiber Membrane Bioreactor. Biomaterials 2009, 30, 2531–2543. 23. Ahmed, H. M. M.; Salerno, S.; Morelli, S.; Giorno, L.; De Bartolo, L. 3D Liver Membrane System by Co-culturing Human Hepatocytes, Sinusoidal Endothelial and Stellate Cells. Biofabrication 2017, 9 (2), 025022. 24. Ahmed, H. M. M.; Salerno, S.; Piscioneri, A.; Khakpour, S.; Giorno, L.; De Bartolo, L. Human Liver Microtissue Spheroids in Hollow Fiber Membrane Bioreactor. Colloids Surfaces B Biointerfaces 2017, 160, 272–280. 25. Gerlach, J. C.; Brayfield, C.; Puhl, G.; Borneman, R.; Muller, C.; Schmelzer, E.; Zeilinger, K. Lidocaine/monoethylglycinexylidide Test, Galactose Elimination Test, and Sorbitol Elimination Test for Metabolic Assessment of Liver Cell Bioreactors. Artif. Organs 2010, 34, 462–472. 26. Mueller, D.; Tascher, G.; Muller-Vieira, U.; Knobeloch, D.; Nuessler, A. K.; Zeilinger, K.; Heinzle, E.; Noor, F. In-depth Physiological Characterization of Primary Human Hepatocytes in a 3D Hollow-fiber Bioreactor. J. Tissue Eng. Regenerat. Med. 2011, 5, e207–e218. 27. Hoffmann, S. A.; Muller-Vieira, U.; Biemel, K.; Knobeloch, D.; Heydel, S.; Lubberstedt, M.; Nussler, A. K.; Andersson, T. B.; Gerlach, J. C.; Zeilinger, K. Analysis of Drug Metabolism Activities in a Miniaturized Liver Cell Bioreactor for Use in Pharmacological Studies. Biotechnol. Bioeng. 2012, 109, 3172–3181. 28. Schmitmeier, S.; Langsch, A.; Jasmund, I.; Bader, A. Development and Characterization of a Small-scale Bioreactor Based on a Bioartificial Hepatic Culture Model for Predictive Pharmacological In Vitro Screenings. Biotechnol. Bioeng. 2006, 95, 1198–1206. 29. Mizumoto, H.; Funatsu, K. Liver Regeneration Using a Hybrid Artificial Liver Support System. Artif. Organs 2004, 28, 53–57. 30. De Bartolo, L.; Morelli, S.; Rende, M.; et al. Human Hepatocyte Morphology and Functions in a Multibore Fiber Bioreactor. Macromol. Biosci. 2007, 7, 671–680. 31. Lu, H. F.; Lim, W. S.; Zhang, P. C.; et al. Galactosylated Poly(vinylidene Difluoride) Hollow Fiber Bioreactor for Hepatocyte Culture. Tissue Eng. 2005, 11, 1667–1677. 32. Memoli, B.; De Bartolo, L.; Favia, P.; et al. Fetuin-A Gene Expression, Synthesis and Release in Primary Human Hepatocytes Cultured in a Galactosylated Membrane Bioreactor. Biomaterials 2007, 28, 4836–4844. 33. Curcio, E.; Salerno, S.; Barbieri, G.; De Bartolo, L.; Drioli, E.; Bader, A. Mass Transfer and Metabolic Reactions in Hepatocyte Spheroids Cultured in Rotating Wall GasPermeable Membrane System. Biomaterials 2007, 28, 5487–5497. 34. Ye, S. H.; Watanable, J.; Takai, M.; Iwasaki, Y.; Ishihara, K. High Functional Hollow Fiber Membrane Modified with Phospholipid Polymers for a Liver Assist Bioreactor. Biomaterials 2006, 27, 1955–1962. 35. Sauer, I. M.; Scwartlander, R.; Schmid, J.; Efimova, E.; Vondran, F. W. R.; Kehr, D.; Pless, G.; Spinelli, A.; Brandeburg, B.; Hildt, E.; Neuhaus, P. The SlideReactor - A Simple Hollow Fiber Based Bioreactor Suitable for Light Microscopy. Artif. Organs 2005, 29, 264–267. 36. Ostrovidov, S.; Jiang, J.; Sokai, Y.; Fujii, T. Membrane-based PDMS Microbioreactor for Perfused 3D Primary Rat Hepatocyte Cultures. Biomed. Microdevices 2004, 6, 279–287. 37. Roy, P.; Baskaran, H.; Tilles, A. W.; Yarmush, M.; Toner, M. Analysis of Oxygen Transport to Hepatocytes in a Flat-Plate Microchannel Bioreactor. Ann. Biomed. Eng. 2001, 29, 947–955. 38. Margulis, M. S.; Erukhimov, E. A.; Andreiman, L. A.; Viksna, L. M. Temporary Organ Substitution by Hemoperfusion through Suspension of Active Donor Hepatocytes in a Total Complex of Intensive Therapy in Patients with Acute Hepatic Insufficiency. Resuscitation 1989, 18, 85–94. 39. Flendrig, L. M.; La Soe, J. W.; Joerning, G. G. A.; et al. In Vitro evaluation of a Novel Bioreactor Based on an Integral Oxygenator and a Spirally Wound Nonwoven Polyester Matrix for Hepatocyte Culture as Small Aggregates. J. Hepatol. 1997, 26, 1379–1392. 40. van de Kerkhove, M. P.; Di Florio, E.; Scuderi, V.; Mancini, A.; et al. Phase I Clinical Trial with the AMC-Bioartificial Liver. Int. J. Artif. Organs 2002, 25, 950–959. 41. Nibourg, G. A.; Chamuleau, R. A.; van der Hoeven, T. V.; et al. Liver Progenitor Cell Line HepaRG Differentiated in a Bioartificial Liver Effectively Supplies Liver Support to Rats with Acute Liver Failure. PLoS One 2012, 7, e38778. 42. Sussman, N. L.; Chong, M. G.; Kousayer, T.; et al. Reversal of Fulminant Hepatic Failure Using an Extracorporeal Liver Assist Device. Hepatology 1992, 6, 60–65. 43. Demetriou, A. A.; Rozga, J.; Podesta, L.; et al. Early Clinical Experience with a Hybrid Bioartificial Liver. Scand. J. Gastroenterol. 1995, 208, 111–117. 44. Demetriou, A. A.; Whiting, J. F.; Levenson, A. M.; et al. New Method of Hepatocyte Transplantation and Extracorporeal Liver Support. Ann. Surg. 1986, 204 (3), 259–270. 45. Millis, J. M.; Cronin, D. C.; Johnson, R.; et al. Initial Experience with the Modified Extracorporeal Liver-Assist Device for Patients with Fulminant Hepatic Failure: System Modifications and Clinical Impact. Transplantation 2002, 74, 1735–1746. 46. Vital Therapies Targeting Liver Disease. ELAD® system clinical development. Available at: http://vitaltherapies.com/clinical-trials/. 47. Demetriou, A. A.; Brown, R. S., Jr.; Busuttil, R. W.; et al. Prospective, Randomized, Multicenter, Controlled Trial of a Bioartificial Liver in Treating Acute Liver Failure. Ann. Surg. 2004, 239, 660–667. 48. Nyberg, S. L. Bridging the Gap: Advances in Artificial Liver Support. Liver Transplant. 2012, 18 (Suppl. 2), S10–S14. 49. Gerlach, J. V.; Encke, J.; Hole, O.; et al. Bioreactor for Larger Scale Hepatocyte In Vitro Perfusion. Transplantation 1994, 58, 948–988. 50. Mundt, A.; Puhl, G.; Muller, A.; et al. A Method to Assess Biochemical Activity of Liver Cells during Clinical Application of Extracorporeal Hybrid Liver Support. Int. J. Artif. Organs 2002, 25, 542–548. 51. Sauer, I. M.; Gerlach, J. C. Modular Extracorporeal Liver Support. Artif. Organs 2002, 26, 703–736. 52. Sauer, I. M.; Kardassis, D.; Zeillinger, K.; et al. Clinical Extracorporeal Hybrid Liver Support–phase I Study with Primary Porcine Liver Cells. Xenotransplantation 2003, 10, 460–469. 53. Sauer, I. M.; Zeilinger, K.; Pless, G.; et al. Extracorporeal Liver Support Based on Primary Human Liver Cells and Albumin Dialysis–treatment of a patient with primary graft non-function. J. Hepatol. 2003, 39, 649–653. 54. Morsiani, E.; Brogli, M.; Galavotti, D.; et al. Long-term expression of highly differentiated functions by isolated porcine hepatocytes perfused in a radial-flow bioreactor. Artif. Organs 2001, 25, 740–748. 55. Patzer, J. F.; Mazariegos, G. V.; Lopez, R. Preclinical evaluation of the Excorp Medical, Inc, Bioartificial Liver Support System. J. Am. Coll. Surg. 2002, 195, 299–310. 56. Mazariegos, G. V.; Patzer, J. F., 2nd; Lopez, R. C.; et al. First clinical use of a novel bioartificial liver support system (BLSS). Am. J. Transplant. 2002, 2, 260–266. 57. Ding, Y. T.; Qiu, Y. D.; Chen, Z.; et al. The development of a new bioartificial liver and its application in 12 acute liver failure patients. World J. Gastroenterol. 2003, 9, 829–832. 58. Gan, J. H.; Zhou, X. Q.; Qin, A. L.; et al. Hybrid artificial liver support system for treatment of severe liver failure. World J. Gastroenterol. 2005, 11, 890–894. 59. Glorioso, J. M.; Mao, S. A.; Rodysill, B.; et al. Pivotal Preclinical Trial of the Spheroid Reservoir Bioartificial Liver. J. Hepatol. 2015, 63, 388–398. 60. Selden, C.; Bundy, J.; Erro, E.; et al. A clinical-scale BioArtificial Liver, developed for GMP, improved clinical parameters of liver function in porcine liver failure. Sci. Rep. 2017, 7, 14518. 61. Sakiyama, R.; Blau, B. J.; Miki, T. Clinical translation of bioartificial liver support systems with human pluripotent stem cell-derived hepatic cells. World J. Gastroenterol. 2017, 23, 1974–1979. 62. Iwamuro, M.; Shiraha, H.; Nakaji, S.; et al. A preliminary study for constructing a bioartificial liver device with induced pluripotent stem cell-derived hepatocytes. Biomed. Eng. Online 2012, 11, 93. 63. Mathis, D.; Vence, L.; Benoist, C. [beta]-Cell death during progression to diabetes. Nature 2001, 414 (6865), 792–798.

412 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111.

Membrane Bioreactors for Bioartificial Organs Costa, A.; Conget, I.; Gomis, R. Impaired glucose tolerance: is there a case for pharmacologic intervention? Treat. Endocrinol. 2002, 1, 205–210. Opara, E. C.; Mirmalek-Sani, S. H.; Khanna, O.; et al. Design of a bioartificial pancreas. J. Invest. Med. 2010, 58 (7), 831–837. Chick, W. L.; Like, A. A.; Lauris, V. Beta-cell culture on synthetic capillaries: An artificial endocrine pancreas. Science 1975, 187, 847–849. Uludag, H.; De Vos, P.; Tresco, P. A. Technology of mammalian cell encapsulation. Adv. Drug Deliv. Rev. 2000, 42, 29–64. de Vos, P.; Marchetti, P. Encapsulation of pancreatic islets for transplantation in diabetes: the untouchable islets. Trends Mol. Med. 2002, 8, 363–366. Wilson, J. T.; Chaikof, E. L. Challenges and emerging technologies in the immunoisolation of cells and tissues. Adv. Drug Deliv. Rev. 2008, 60, 124–145. Tuch, B. E.; Keogh, G. W.; Williams, L.; et al. Safety and viability of microencapsulated human islets transplanted into diabetic humans. Diabetes Care 2009, 32, 1887–1889. Teramura, Y.; Kaneda, Y.; Iwata, H. Islet-encapsulation in ultra-thin layer-by-layer membranes of poly (vinyl alcohol) anchored to poly (ethylene glycol)–lipids in the cell membrane. Biomaterials 2007, 28, 4818–4825. Zhi, Z. I.; Kerby, A.; King, A. J. F.; et al. Nano-scale encapsulation enhances allograft survival and function of islets transplanted in a mouse model of diabetes. Diabetologia 2012, 55, 1081–1090. Teramura, Y.; Iwata, H. Bioartificial pancreas microencapsulation and conformal coating of islets of Langerhans. Adv. Drug Deliv. Rev. 2010, 62, 827–840. Colton, C. K. Implantable biohybrid artificial organs. Cell Transplant. 1995, 4, 415–436. Chang, C.-C. Analysis of transport phenomena and kinetics in an extravascular bioartificial pancreas. AIChE J. 1996, 42, 2668–2682. Calafiore, R.; Luca, G.; Calvitti, M.; et al. Cellular support systems for alginate microcapsules containing isltes, as composite bioartificial pancreas. Ann. N. Y. Acad. Sci. 2001, 944, 240–252. De Vos, P.; Hillebrands, J. L.; De Haan, B. J.; et al. Efficacy of prevascularised expanded polytetrafluoroethylene solid support system as a transplantation site for pancreatic islets. Transplantation 1997, 63, 824–830. Petersen, P.; Lembert, N.; Wesche, J.; et al. Improved diffusion properties of a new polysulfone membrane for the development of a bioartificial pancreas. Transplant. Proc. 2001, 33, 1952–1953. Sakai, S.; Ono, T.; Ijima, H.; Kawakami, K. Newly developed aminopropyl-silicate immunoisolation membrane for a microcapsule-shaped bioartificial pancreas. Ann. N. Y. Acad. Sci. 2001, 944, 277–283. Silva, A. I.; Mateus, M. Development of a polysulfone hollow fiber vascular bioartificial pancreas device for In Vitro studies. J. Biotechnol. 2009, 139, 236–249. Kang, J.; Erdodi, G.; Kennedy, J. P.; et al. Toward a bioartificial pancreas: Diffusion of insulin and IgG across immunoprotective membranes with controlled hydrophilic channel diameters. Macromol. Biosci. 2010, 10, 369–377. Lee, S. H.; Hao, E.; Savinov, A. Y.; et al. Human beta cell precursors mature into functional insulin-producing cells in an immunoisolation device: Implications for diabetes cell therapies. Transplantation 2009, 87, 983–991. Circebio. http://www.circebio.com. Tendulkar, S.; McQuilling, J. P.; Childers, C.; et al. A scalable microfluidic device for the mass production of microencapsulated islets. Transplant. Proc. 2011, 4, 3184–3187. Omer, A.; Duvivier-Kali, V.; Fernandes, J.; et al. Long-term normoglycemia in rats receiving transplants with encapsulated islets. Transplantation 2005, 79, 52–58. https://clinicaltrials.gov/ct2/show/results/NCT00940173. Maki, T.; Lodge, J. P.; Carretta, M.; et al. Treatment of severe diabetes mellitus for more Than one year using a vascularized hybrid artificial pancreas. Transplantation 1993, 55, 713–717. Lacy, P. E.; Hegre, O. D.; Gerasimidi-Vazeou, A.; et al. Maintenance of normoglycemia in diabetic mice by subcutaneous xenografts of encapsulated islets. Science 1991, 254, 1782–1784. Neufeld, T.; Ludwig, B.; Barkai, U.; et al. The efficacy of an immunoisolating membrane system for islet xenotransplantation in minipigs. PLoS One 2013, 8, e70150. Brauker, J. H.; Carr-Brendel, V. E.; Martinson, L. A.; et al. Neovascularization of synthetic membranes directed by membrane microarchitecture. J. Biomed. Mater. Res. 1995, 29, 1517–1524. Agulnick, A. D.; Ambruzs, D. M.; Moorman, M. A.; et al. Insulin-producing endocrine cells differentiated In Vitro from human embryonic stem cells function in macroencapsulation devices In Vivo. Stem Cells Trans.l Med. 2015, 4, 1214–1222. Kriz, J.; Vilk, G.; Mazzuca, D. M.; et al. White, A novel technique for the transplantation of pancreatic islets within a vascularized device into the greater omentum to achieve insulin independence. Am. J. Surg. 2012, 203, 793–797. http://www.trademarkia.com/pancreassist-75261053.html. Humes, H. D.; Weitzel, W. F.; Bartlett, R. H.; et al. Initial clinical results of the bioartificial kidney containing human cells in ICU patients with acute renal failure. Kidney Int. 2004, 66, 1578–1588. Tumlin, J.; Wali, R.; Williams, W.; et al. Efficacy and safety of renal tubule cell therapy for acute renal failure. J. Am. Soc. Nephrol. 2008, 19, 1034–1040. Humes, H. D.; Buffington, D. A.; MacKay, S. M.; et al. Replacement of renal function in uremic animals with a tissue-engineered kidney. Nat. Biotechnol. 1999, 17, 451–455. Humes, H. D.; MacKay, S. M.; Funke, A. J.; Buffington, D. A. Tissue engineering of a bioartificial renal tubule assist device: In Vitro transport and metabolic characteristics. Kidney Int. 1999, 55, 2502–2514. Fissell, W. H.; Lou, L.; Abrishami, S.; Buffington, D. A.; Humes, H. D. Bioartificial kidney ameliorates gram-negative bacteria-induced septic shock in uremic animals. J. Am. Soc. Nephrol. 2003, 14, 454–461. Fissell, W. H.; Dyke, D. B.; Weitzel, W. F.; et al. Bioartificial kidney alters cytokine response and hemodynamics in endotoxin challenged uremic animals. Blood Purif. 2002, 20, 55–60. Fujita, Y.; Terashima, M.; Kakuta, T.; et al. Transcellular water transport and stability of expression in aquaporin 1- transfected LLC-PK1 cells in the development of a portable bioartificial renal tubule device. Tissue Eng. 2004, 10, 711–772. Inagaki, M.; Yokoyama, T. A.; Sawada, K.; et al. Prevention of LLC-PK1 cell overgrowth in a bioartificial renal tubule device using a MEK inhibitor, U0126. J. Biotechnol. 2007, 132, 57–64. Zhang, H.; Tasnim, F.; Ying, J. Y.; Zink, D. The impact of extracellular matrix coatings on the performance of human renal cells applied in bioartificial kidneys. Biomaterials 2009, 30, 2899–2911. Vesey, D. A.; Qi, W.; Chen, X.; et al. Isolation and primary culture of human proximal tubule cells. Methods Mol. Biol. 2009, 466, 19–24. Verhulst, A.; Sayer, R.; De Broe, M. E.; et al. Human proximal tubular epithelium actively secretes but does not retain rosuvastatin. Mol. Pharmacol. 2008, 74, 1084–1091. Zhang, H.; Lau, S. F.; Heng, B. F.; et al. Generation of easily accessible human kidney tubules on two dimensional surfaces in vitro. J. Cell Mol. Med. 2011, 15, 1287–1298. Kim, D.; Dressler, G. R. Nephrogenic factors promote differentiation of mouse embryonic stem cells into renal epithelia. J. Am. Soc. Nephrol. 2005, 16, 3527–3534. Baer, P. C.; Bereiter-Hahn, J.; Missler, C.; et al. Conditioned medium from renal tubular epithelial cells initiates differentiation of human mesenchymal stem cells. Cell Prolif 2009, 42, 29–37. Buffington, D. A.; Pino, C. J.; Hageman, G. M.; Chen, L.; Humes, H. D. Bioartificial renal epithelial cell systems (BRECS): a compact, cryopreservable extracorporeal renal replacement device. Cell Med. 2012, 4, 33–43. Ng, C. P.; Zhuang, Y.; Lin, A. W. H.; Teo, J. C. M. A Fibrin-Based Tissue-Engineered Renal Proximal Tubule for Bioartificial Kidney Devices: Development, Characterization and In Vitro Transport Study. Int. J. Tissue Eng. 2013;, 319476. Fissell, W. H.; Manley, S.; Westover, A.; et al. Differentiated growth of human renal tubule cells on thin-film and nanostructured materials. Am. Soc. Artif. Intern. Organs J. 2006, 52, 221–227. Jansen, J.; De Napoli, I. E.; Fedecostante.; et al. Human proximal tubule epithelial cells cultured on hollow fibers: living membranes that actively transport organic cations. Sci. Rep. 2015, 15, 16702.

Membrane Bioreactors for Bioartificial Organs

413

112. Saito, A. Research into the development of a wearable bioartificial kidney with a continuous hemofilter and a bioartificial tubule device using tubular epithelial cells. Artif. Organs 2004, 28, 58–63. 113. Saito, A.; Aung, T.; Sekiguchi, K.; Sato, Y. Present status and perspective of the development of a bioartificial kidney for chronic renal failure patients. Ther. Apher. Dial. 2006, 10, 342–347. 114. https://www.bmm-program.nl/SITE/PUBLIC/GO/article.aspx?id¼87&title¼Bio-artificialþkidney. 115. Ronco, C.; Davenport, A.; Gura, V. A wearable artificial kidney: dream or reality? Nat. Clin. Pract. Nephrol. 2008, 4, 604–605. 116. Sanechika, N.; Sawada, K.; Usui, Y.; et al. Development of bioartificial renal tubule devices with lifespan-extended human renal proximal tubular epithelial cells. Nephrol. Dial. Transplant. 2011, 26, 2761–2769.

Microbioreactorsq

2.29

Polona Znidarsic-Plazl and Igor Plazl, University of Ljubljana, Ljubljana, Slovenia © 2017 Elsevier B.V. All rights reserved. This is a reprint of Polona Znidarsiǒ-Plazl, Igor Plazl, Microbioreactors, Reference Module in Life Sciences, Elsevier, 2017.

2.29.1 Introduction 2.29.2 Microfluidic Devices 2.29.2.1 Properties of Microfluidic Devices 2.29.2.1.1 Decrease in physical size 2.29.2.1.2 Increase in the number of units 2.29.2.2 Fluid Flow at the Microscale 2.29.2.3 Fabrication Technologies 2.29.3 Microbioreactors for Cell Culturing 2.29.3.1 Micro-Petri Dishes 2.29.3.2 Microtiter Plates 2.29.3.3 Microfluidic Chips With Culture Chambers 2.29.3.4 Miniature Bioreactors 2.29.3.4.1 Microchemostats 2.29.4 Enzymatic Microreactors 2.29.4.1 Enzymatic Microreactors in Chemical Analysis 2.29.4.2 Applications for Screening and Kinetic Studies 2.29.4.3 Biotransformations in Microreactors 2.29.5 Future Perspectives on Bioreactor Miniaturization References Relevant Website

414 415 415 415 416 416 418 418 419 420 420 421 422 423 424 424 425 425 426 427

Glossary Continuous flow chemical processing (CFCP) Methodology based on a combination of microunit operations such as mixing and reaction, molecular capture, solid extraction, phase separation, heating, concentration, cell culture, and a multiphase flow network. Electrophoretically mediated microanalysis (EMMA) Homogeneous enzyme assay utilizing the different electrophoretic motilities of an enzyme and substrate(s) to initiate a reaction inside the capillary. International Conference on Microreaction Technology (IMRET) An international scientific conference series organized since 1997 that is devoted to the field of microprocess engineering and the microreactors. Lab-on-a-chip (LOC) A device that integrates one or several laboratory functions on a single chip while transporting and manipulating nano to microliter amounts of fluids. Microelectromechanical system (MEMS) A device of micrometer to millimeter scale, usually consisting of a central unit that processes data, the microprocessor, and several components that interact with the outside, such as microsensors. Microfluidics A research field that develops methods and devices to control, manipulate, and analyze flows on a nano to microliter scale. Microreactor technology (MRT) Technology based on processes and unit operations within microstructured devices. Micro-total analysis system (mTAS) Chip-based microchannel system that is used for complete analytics, usually integrating several consecutive operations.

2.29.1

Introduction

Nature uses the micrometer scale to its advantage in many processes. Microorganisms such as Escherichia coli, with a size of approximately 2 mm, move more slowly than diffusing nutrients and waste. E. coli can forage just as efficiently by simply waiting for food to

q

 Change History: May 2016. Polona Znidar sic-Plazl updated the Sections 3–3.4, 3.4.1, 4, 4.1, 4.2, and 5; text and introduced small edit throughout the article including citations; and also updated the “References” section. Igor Plazl updated the text and Section 4.3 and a new Fig. 6 has been added in this section.  Polona Znidar sic-Plazl and Igor Plazl updated their biographies.

414

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.09071-3

Microbioreactors

415

diffuse past rather than actively searching for food [1]. Living cells are biochemical microreactors, and multicellular life forms use the concept of numbering-up in a multitude of variants. Unlike nature, science and technology are only just beginning to harness microscale phenomena for practical use. However, with the availability of technologies for building microstructured devices in a controlled and repeatable manner, applying knowledge regarding the behavior of particles and fluids at the microscale has become more practical and creative [1]. Scientists and engineers are searching for new ideas and solutions for commercialization of micro–nano systems. The main advantages of miniaturization include savings in time, space, materials, and/or cost, together with increased performance. Microfluidics is expected to revolutionize chemistry, biology, and biotechnology, just as microelectronics has revolutionized information technology in the past century. In the past two decades, microreactor technology (MRT) has impressively demonstrated advantages in many chemical and biochemical processes; therefore, it is gaining importance in a broad range of areas from fuel production and processing to the biotechnology, chemical, pharmaceutical, and electronics industries, personal chemistry, and environmental protection. Initially, microfluidics research focused on analytical applications and the fundamental understanding of fluid flow in microchannels. The development of microfluidics started in the 1970s in the area of analytical chemistry; many applications are related to this field. An original patent from the former East Germany in 1986 could be considered the first important step in microreactor development. In 1989, the Forschungszentrum Karlsruhe built the first microheat exchanger and published its potential for microchemical engineering. Similar pioneering work also began in 1993 at the Pacific Northwest National Laboratory (United States); just 2 years later, the first workshop on Microsystems Technology for Chemical and Biological Microreactors was held in Germany. Many researchers refer to 1995 as the birth year of microreaction technology as a new discipline and as the starting point for its worldwide development. In 1997, the first International Conference on Microreaction Technology (IMRET 1) took place and has been held almost every year since [2]. At that time, fundamental microfluidic studies were reported using micro-total analysis systems (mTAS), which were later extended to lab-on-a-chip (LOC). For these types of systems, all necessary components, such as mixers, valves, reaction chambers, and detectors, are integrated. Many different microfluidic platforms have been developed for LOC. All these initiatives and increasing industrial interest, in addition to nationally initiated funding, have led to the recent exponential growth of microfluidics studies. A wide range of commercial microstructured and even complete chemical process plants already exist, as do companies that provide microstructured devices and services. The number of institutes that have devoted themselves to this area is markedly expanding; universities offer lectures and practical classes, and many excellent reviews and specialized books are available (e.g., Ref. [1–3]). Currently, microfluidic systems are being used to study diverse processes such as bioprocessing, chemical synthesis, enzymecatalyzed reactions, clinical diagnostics, and genetic engineering. These systems have been applied to essentially all major classical biological methods such as polymerase chain reaction (PCR), cloning, separations, in vitro protein synthesis, drug screening, cell-based screens, small molecule synthesis, DNA/peptide synthesis, enzyme screens, protein interaction screening, and crystallography [3]. Several microfluidic devices have also been developed for cell analysis. Microfluidic systems that are combined with physical microstructures/nanostructures or a topographically patterned substrate for cell research give rise to notable changes in mechanical forces in various ways, allowing for a low-expertise route for manipulation of cells inside a channel under dynamic or static flow conditions. The current methods commonly used in biological laboratories for manipulation, concentration, and separation of bioparticles include optical tweezers, fluorescence-activated or magnetic-activated cell sorting, centrifugation, filtration, and electric field-based manipulations and separations [4]. Microbioreactor technology also offers the potential to improve bioprocess development, which is currently performed in shake flasks and in conventional stirred-tank bioreactors with typical volumes of 0.5–10 L. However, the use of these conventional tools has the following limitations: laborious and time-consuming procedures, higher equipment maintenance costs, systems not able to perform experiments in parallel, risk of sample contamination during sampling procedures, use of large amounts of reagents, generation of large volumes of waste, and low throughput. Microbioreactors provide unique environments that are ideal for process control, waste reduction, high-throughput, rapid testing, parallel investigations, and culture optimization, as well as for rapid and efficient biosynthesis. In clinical diagnostics, patients samples are often subjected to biochemical analyses performed in clinical laboratories; therefore, a reliable diagnosis in many cases cannot be performed within the consultation time. One of the long-term goals in the field of microfluidics is to create integrated, portable, clinical diagnostic devices for home and bedside use, thereby eliminating time-consuming laboratory purification procedures [3,5].

2.29.2

Microfluidic Devices

2.29.2.1

Properties of Microfluidic Devices

A microfluidic device is used for performing processes designed or selected to create microflow phenomena, that is, flow guidance and flow processing with characteristic dimensions much smaller than those of conventional apparati, typically within the submillimeter range [6]. A microscale reactor is a device whose operation depends on precisely controlled design features with characteristic dimensions from submillimeter to submicrometer. The fundamental advantages of microstructured reactors compared to conventional reactors are associated with the decreases in physical size and increases in the number of units.

2.29.2.1.1

Decrease in physical size

An obvious effect of shrinking a system to the micrometer scale is the large increase in surface area relative to volume (often several orders of magnitude). Specific surface areas of microstructured devices lie between 10,000 and 50,000 m2 m
3, whereas those of

416

Microbioreactors

traditional reactors are generally approximately 100 m2 m
3 and, in rare cases, reach 1000 m2 m
3. Decreases in linear dimensions increase the heat transfer coefficient (up to 25 kW m
2 K
1); typical fluid layer thicknesses can be set to a few tens of millimeters and mixing times in micromixers amount to milliseconds. High heat-exchange efficiency, significantly higher than for traditional heat exchangers, allows for fast heating and cooling in reaction mixtures within the microstructured devices. Development of hot spots or accumulation of reaction heat within microstructures is suppressed; therefore, the outcome in many cases is higher selectivity, yield, and product quality. Film thicknesses of approximately 25 mm have been measured in falling films running through microchannels, which is approximately 10-fold smaller than for macroscopic falling film reactors. Even thinner films can exist between Taylor bubbles and channel walls in a gas–liquid slug flow, as well as in the case of an annular flow. Decreases in volume, which typically amount to a few microliters, replace batch with continuous flow processes. Process parameters such as pressure, temperature, residence time, and flow rate are more easily controlled in reactions that take place in small volumes. The hazard potential of strongly exothermic or explosive reactions can also be drastically reduced so that the small reaction volume of microreactors is also responsible for advantages in technical safety [2]. For fluids, large specific surface areas allow for more efficient heat and mass transfer in microsystems due to thegreater interface available for transfer to occur and less total mass or energy needing to be transferred to reach the final state. Therefore, both the creation and homogenization of solute or temperature gradients become faster as the system size is reduced.

2.29.2.1.2

Increase in the number of units

The fundamental advantages of microstructured devices due to the increased number of units include fast and inexpensive screening, production flexibility in terms of capacity, and process robustness. Fast, cost-saving screening of materials such as inorganic materials, catalysts, and polymers are feasible, and the continuous flow operation enables rapid serial synthesis. An increase in throughput in microreactors is achieved by a numbering-up approach rather than by a scaling-up one approach. In addition, the installation and removal of microstructured units may be sufficiently fast and flexible due to the numbering-up assembly. Integrated micro(bio)chemical systems have been developed by microunit operations and continuous flow chemical processing.

2.29.2.2

Fluid Flow at the Microscale

The defining feature of microfluidic systems is continuous flow experimentation, whereby very rapid heat and mass transfer can be achieved because of the reduction of the flow path to dimensions approaching the continuous phase hydrodynamic boundary layer. Fluid behavior in reduced dimensions is increasingly influenced by viscosity rather than inertia, and the fluid flow in microstructured devices with simple geometries is typically laminar. The laminar flows cause velocity profiles in the channel to appear parabolic in shape, which can lead to a relatively broad residence-time distribution. The ratio of inertial to viscous forces is indicated by the Reynolds number (Re), which is calculated as Re ¼ av=ðnÞ, where v is the velocity scale of the fluid, a is the characteristic distance of the system (in the case of flow through a pipe, a would be the pipe diameter), and ðnÞ is the kinematic viscosity of the fluid. Viscosity, the internal friction of a fluid, produces resistance to shear, and there is a tendency for the fluid to move in parallel layers known as laminar flow; inertia, the tendency of a body in motion to remain in motion, counters laminar flow and can ultimately result in turbulent flow [1]. There have been many studies concerning the transition from the laminar to the turbulent flow regime in microchannels, which suggest a significant lowering of the critical Reynolds number, i.e. the transition to turbulence. Because microchannels in microreactors are usually not circular due to the different fabrication processes, evaluation of the flow characteristics with regard to the shape of the microchannels should be considered, as should the effect of the relative surface roughness and the channel aspect ratio (aspect ratio¼width/depth of the microchannel). In laminar flow, diffusion can be effective for moving and mixing solutes on micrometer-length scales. The relative importance of diffusion and convective bulk flow for transporting solutes and solvent molecules is provided by the Peclet number and can be readily adjusted through the choice of flow velocity and the dimensions of the system used. The Peclet number provides an indication of the relative importance of diffusion (the random thermal motion of molecules within their surrounding environment) and convection (the transport as a result of bulk motion of a fluid). The Peclet number is defined as Pe¼uadL/D, where ua is the average velocity of the flow, dL is the characteristic length of the system perpendicular to the direction of the flow, and D is the diffusion coefficient of the particle or molecule of interest [1]. Surface tension can profoundly influence fluid behavior because of the large surface area-to-volume ratio typical for microfluidic systems. The most significant surface tension phenomenon is probably capillarity, which is driven by capillary forces that become more significant relative to other forces such as gravity as the system size is reduced. Surface properties can be selected to influence the competition between viscous forces and capillary forces to control the generation, breakup, and coalescence of droplets. Capillarity is the rise or fall of a liquid in a small passage, such as a thin tube. The ratio of viscous and capillary forces is given by the capillary number (Ca). The capillary number is calculated as Ca¼um/g, where u is the velocity of the flow, m is the viscosity of the fluid, and g is the surface tension [1]. Multiphase flows in microchannels of microstructured devices have been documented for liquid–liquid and gas–liquid mixtures. The Weber number, We, which is the ratio of inertia to surface forces, is suitable for their description. The Weber number is calculated as We¼DHusr/s, where DH is the hydraulic diameter (DH¼2HW /(HþW), where H and W are channel depth and width, respectively), us is the superficial velocity, r is the mass density of the fluid, and s is the interfacial tension. Weber numbers in microchannels are low (typically less than 1). In the case of two immiscible fluids, the interfacial tension dominates the effect of the dynamics of the free surface and the effects of interfacial tension are significantly greater than those of gravitational force. The Bond number provides an indication of the relative importance of gravitational forces and interfacial tension and is defined as Bo¼rgDH2/s, where r is the mass density of the fluid, DH is the hydraulic diameter, g is the

Microbioreactors

417

gravitational constant, and s is the interfacial tension [7]. Typically, the values of the Bond number in microchannels are very low, so the effects of gravity can usually be neglected. Recently, the flow of immiscible fluids in microchannels has attracted significant attention. The interest in flow patterns arose because microfluidic technology offers new prospects for emulsion science. The practical use of emulsions (droplets of one liquid dispersed in another) has long been based on their bulk properties. With the advent of microfluidics, we can now manipulate individual droplets and precisely control their properties. However, the progress of micro(bio)chemical engineering requires us to have knowledge of the flow patterns in microchannels if we want to optimize the design and operation of multiphase processes. In microscale flows with small-diameter flow channels, the flow pattern is mainly a function of the interfacial tension, the wall friction force, and the viscosity of the fluid. Molecular effects also become more significant in the microchannel when the characteristic length decreases to the point where the continuum assumption becomes invalid. Use of the classical Navier–Stokes equations with the no-slip boundary condition appears to be appropriate for the description of microchannel flows of liquids as long as the hydraulic diameter of the system is greater than 100 mm for conduits filled with Newtonian fluids, such as water, under standard conditions. However, the Navier–Stokes equations must be modified or augmented with auxiliary equations whenever dealing with nonNewtonian fluids and when surface phenomena, such as near-wall forces and relative roughness, become more important as the microchannel size becomes smaller. Non-Newtonian fluid effects are expected to be important for polymeric liquids and particle suspension flows. Wall slip effects are negligible for liquid flows in microconduits, and viscous dissipation effects and the friction factor are negligible in smooth microchannels [8]. The slip flow of gases, which occurs at high temperatures, low pressures, and/or narrow channels, is a notable exception. Slip flow is an important consideration because it may allow significant reductions in the friction pressure drop. It is known that an apparent slip occurs more readily on a surface that has opposite wet-ability as the fluid, on rough surfaces, and at high shear rates. The validity of slip flow boundary conditions can be proven by the Knudsen number (Kn). The Knudsen number is the ratio of the mean free path l to DH, where l is the average distance traveled by the molecule between two consecutive collisions. With higher Kn, the continuum assumption does not hold true; therefore, microfluidic flows cannot be described by the direct application of the Navier–Stokes equations with their customarily used no-slip boundary conditions. As Kn/0, the Navier–Stokes equations are reduced by neglecting heat conduction and viscous diffusion and dissipation. Most researchers agree on classifying the flow as in continuum (no-slip boundary conditions) for Kn 95%, at H2S loading rates of 11 g m
3.h, without any addition of nutrients and pH buffering substances. H2S is probably the most common odor generating problems. A comparison of H2S removal efficiencies achieved in biofilters, under a wide range of operating conditions, is summarized in Table 1.

2.31.7

Factors Affecting the Biofiltration Process

In general, bioreactors are applicable for the removal of any biodegradable pollutant, irrespective of whether the pollutant is a VOC, VIC or a mixture of VOC and VIC. However, the efficiency of any bioreactor depends on other operational parameters.5 In this section, parameters that influence a biofilter’s performance are described.

2.31.7.1

Humidity

A suitable moisture content or high enough relative humidity of the waste gas is required for optimum microbial activity in any biofilter. Hence, humidification is considered as a prerequisite for effective biofilter operation. Biodegradation processes are exothermic in nature, i.e., they generate heat. Besides microbial oxidation, humidity fluctuations are also responsible for water content changes in the filter bed. Gostomski et al.18 provided the following scientific justification for the requirement of maintaining a proper water balance in biofilters: (i) the water content affects both the physics and biology of the system, (ii) the presence of excess water will lead to high pressure drops as water displaces air in the interparticle spaces, leading to variations in flow pattern, (iii) changes in moisture content are time and position dependent and are associated with multiple mechanisms of water loss and gain depending on the thermodynamic characteristics of the biofilter, (iv) if the waste gas is not humidified, water loss will occur at the inlet, and condensation will add water at the outlet, and (v) excess water will also cause the bed material to consolidate, leading to an increase in the pressure drop and often causing channelling problems and a decline in the performance of the biofilter. Besides, temperature usually rises at higher pollutant loading rates. In full-scale biofilters, normally the filter bed is not saturated because there is no free-flowing water phase. Therefore, in practice, sprinklers are regularly used to maintain the optimal water content of the filter bed.19

2.31.7.2

Biocatalyst

The biocatalyst also plays a vital role in the biofiltration processes. The inoculation of a specific biocatalyst may be required for efficient biofilter operation, depending upon the type of packing material (i.e., natural, inert or synthetic) and characteristics of the biodegradable substrate (i.e., easily biodegradable pollutant or recalcitrant pollutant).5,19,20 As shown in Table 1, mixed microbial cultures are frequently used to inoculate the filter bed, and the inoculum source can be different, e.g., activated sludge from a wastewater treatment plant, depending on its availability and types of pollutants.20,21 Such type of natural inoculum generally contains microbial populations able to metabolize easily biodegradable pollutants, such as hydrogen sulphide. In some biofilters, defined co-cultures of different bacterial species have sometimes also shown to be more efficient than single pure organisms for the biodegradation of recalcitrant compounds because of the synergistic interactions between different pure strains. Besides, the characteristics of the microbial community structure also plays an important role in maintaining the functional stability of the biofilter depending on the pollutant loading characteristics, i.e., steady or transient loading patterns. The microbial species richness, evenness, dynamics, functional redundancy, microbial composition and microbial interactions within the filter bed are important factors that affect the performance of biofilters.22 When working with defined microbial populations, most of the biofilters are inoculated with bacteria. However, several studies have also investigated the efficiency of biofilters using fungi or a mixture of bacteria and fungi because of the following reasons: (i) fungi have the ability to degrade complex organic pollutants and they exhibit superior performance compared to bacteria,23 (ii) biofilters containing only bacteria show poor performance for the treatment of highly hydrophobic VOC because of the limited mass transfer between gas and liquid phases,24 (iii) fungi-inoculated biofilters are able to show high performance for the removal of hydrophobic VOC, under low pH and low relative humidity conditions,25 and (iv) the kinetics and the biodegradation rate may be faster and more efficient using fungal-bacterial cocultures because of the higher microbial biodiversity in the filter bed.26

2.31.7.3

Nutrient Availability

Additional nutrients may be required and added to the biofilters for maintaining the biofilm growth, especially during longer periods of filter bed operation because availability of both micronutrients (e.g., trace metals and vitamins) and macronutrients (e.g., N, P, K, S) may be limited in the filter bed.3,19 In a study that was aimed at testing the effect of different amounts of nutrient solution addition to the filter bed of a methane treating biofilter (250–1500 mL day
1), the authors reported clogging problems at the top section of the filter bed.27 The biofilter was operated in an upflow mode, i.e., methane gas was passed to the biofilter from the bottom, while the nutrient solution was added from the top. Furthermore, the authors also reported that biomass stratification occurred along the filter bed height, i.e., higher amount of biomass at the top section. This phenomenon led to the uptake of all the nutrients, and the middle and the bottom sections of the filter bed were deprived of such nutrients.

Table 1

Comparison of biofilters performance for the removal of H2S

OT, days

IS and/or remarks

FBM

Condition

IC, ppmv

213

Activated sludge (WWTP)

Expanded schist Cellular concrete

Anoxic

100

Activated sludge (WWTP)

Cellular concrete waste

168

Activated sludge (WWTP)

Polyethylene

Aerobic Biotic Biotic Abiotic Abiotic Aerobic

1100 450 900 450 133 360 100 350

84

Effluent (anaerobic reactor of CRLF) Activated sludge from WWTP Mixed pollutants (H2S and VFAs)

Coconut fiber

16 190

82 60 206

Latex wastewater, cell-immobilized GAC Compost, filter bed mixing once every 2 days

Compost and perlite (3:1) Lava rock, vermiculite and glass rings Granular activated carbon (GAC) Compost

 95  200 380 Aerobic  2300 Biogas:air (1:4)  8750 0.04b Aerobica Aerobic 1500 1500 1500 3000 Aerobic 100–4000 200 Aerobic 100 Bed mixing 100

ILR, g m
3 h
1

EC, g m
3 h
1

MILR, g m
3 h
1

37.8 37.8 10.5

8  16 >33

150 144 400

28.8

 130 >140 142  225

MEC, g m
3 h
1

RE, %

30.3 25.2

100  67  93  55 100  65  70  35

17.8

220 400

142 232 125

7 7

>95 95 85 98 >98 >98 >95 75 >97 100 80 100

pH

2.5

1.0 4.5

2.0 2.1 4.5

EBRT, s

References

300 60 240 60 63 63 63 63 63 62 62 62 78

[13]

50 85 31 31 31

[16] [8]

49 49

[17]

[14]

[15] [9]

[10]

Gas-Phase Bioreactors

Note: CRLF - concentrated rubber latex factory; EBRT - empty bed residence time; EC - elimination capacity; FBM - filter bed materials; IC - inlet concentration; ILR - inlet loading rate; IS - inoculum source; MEC - maximum elimination capacity; MILR - maximum inlet loading rate; OT - operating time; RE - removal efficiency; VFAs - volatile fatty acids; VOC - volatile organic compounds; WWTP - wastewater treatment plant a aerobic process for mixed pollutants (H2S and VOC). b unit was g m
3.

451

452

Gas-Phase Bioreactors Table 2

Mineral compositions of the nutrient solutions that were used in the biofilters

Media composition, g L
1

Pollutant

Packing materials

References

K2HPO4 - 0.80 KH2PO4 - 0.20 CaSO4$2H2O - 0.05 MgSO4$7H2O - 0.50 (NH4)2SO4 - 1.00 FeSO4 - 0.01 K2HPO4 - 1.50 KH2PO4 - 4.00 NH4Cl - 0.10 MgCl2$6H2O - 0.20 Trace metals (Na, Fe, Mn, Zn, Cu, Co, Mo) solution KH2PO4 - 0.025 Na2CO3 - 0.40 NH4Cl - 0.02 FeSO4$7H2O - 0.014

Toluene and xylene

Compost

[21]

H2 S

Polyethylene

[15]

H2 S

Expanded schist and cellular concrete

[13]

Nutrient availability and the presence of micro- and macro nutrients in the biofilter are extremely important due to the following reasons28: (i) to maintain the desired microbial growth rate or the biomass yield, (ii) in the case of organic packing materials, the nutrient will be deprived after long-term operation, (iii) in the case of biofilter shutdown, the biomass activity can be sustained by just adding nutrients periodically and providing air supply to maintain the aerobic conditions, and (iv) the availability of intrinsic nutrients in the packing material will also provide good buffering capacity to the filter bed and help maintain the desired microflora within the filter bed. Based on the literature review, the mineral compositions of the nutrient solutions that can be used in biofilters are summarized in Table 2, mainly based on lab-scale studies. At full-scale, simpler or cheaper nutrient solutions may be preferred, if possible.

2.31.7.4

pH

The pH can be different in the biofilter, even for the treatment of the same pollutant, depending on the operating conditions (e.g., aerobic or anaerobic, bed mixing) or the packing material used in the biofilter. For example, pH values of  1.0,9 2.0,10 or 4.517 have been reported in biofilters treating H2S (Table 1). pH regulation in biofilters may be vital for optimum microbial growth.29 A specific range of pH and temperature is required for optimum microbial activity based on the dominant microorganisms present in the biofilm and the composition of pollutants.5,19 Most of the microorganisms thrive at neutral pH, and high removal capacities can be achieved in a pH range of 6.6–8.0. Bacteria are usually considered to be less tolerant to pH fluctuations than fungi. In the case of fungi dominant biofilters, the microorganisms can generally tolerate low pH values in the range of 2.0–5.0. Different bed materials have different pH buffering capacities. For example, soil has higher pH buffering capacity compared to compost and wood chips. During the treatment of VIC such as ammonia and hydrogen sulphide in biofilters, sulfate (SO4 2
), hydrochloric acid (HCl), sulphuric acid (H2SO4), and other acidic metabolites can be generated.30 Their presence will further lower the pH of the filter bed. Adjustment of the desired pH range can be achieved by enriching the filter bed with buffering materials (e.g., calcium carbonate) and by intermittently adding a nutrient solution containing the pH buffer.3,29,30

2.31.7.5

Temperature

Temperature influences the biofilter performance. For example, in a mesophilic system, the removal rate could reduce to 98%), at initial H2S concentrations of 105, 510 and 1020 ppmv, wherein the pH was  8.0 and the moisture content was in the range of 80%– 85%.39 Agricultural or forest waste such as rice hull, bamboo, and camphor tree can be used as valuable sources to produce biochar because of their high H2S adsorption capacity. However, the adsorption capacity depends on the pyrolysis temperature and pH; for example, breakthrough time of an adsorbent increases with an increase of the pyrolysis temperature.38 As seen from the available literature, most of the previous studies on gas-phase pollutant removal using biochar were tested only in adsorption columns. There are only very few studies that have applied biochar in biofilters for waste gas treatment. Das40 tested the performance of a compost þ biochar-packed biofilter under the influence of different H2S concentrations and gas flow rates, at steady and transient conditions. After nearly 100 days of biofilter operation, the author reported that the critical loading rate of the biofilter was 80 g m
3.h at inlet loading rates exceeding 150 g m
3.h.

2.31.10 Biotrickling Filters Biotrickling filters can be considered as an alternate to biofilters. Biotrickling filtration has demonstrated to be a suitable technology to treat hydrophilic pollutants present in waste gases.1,5 Biotrickling filters can replace conventional physico-chemical technologies because the direct cost incurred while treating air emissions is more than three times lower than that of advanced technologies such as combined UV/H2O2 oxidation, catalytic oxidation, condensation, and pressure or temperature swing adsorption.41,42 Concerning the working mechanism, it is similar to the biofilter except for the fact that there is a continuously trickling liquid phase flowing over the filter bed. The well-acclimated microbes are immobilized on inorganic or synthetic packing materials such as pall rings, perlite, lava rock, ceramic balls or pellets, polymer-based packing, activated carbon, and sea shells. Such packing materials offer high porosity in comparison to the organic or natural materials used in biofilters, and they avoid operational problems such as clogging or increasing pressure drop. As the inorganic or synthetic packing material does not provide the necessary nutrients for microbial growth, a trickling liquid medium rich in essential micro- and macronutrients must be continuously fed to the biotrickling filter (Fig. 3). Usually, the polluted air enters counter- or co-currently to the liquid flow, depending on the specific operational requirements. As the liquid phase is circulated, control of pH and nutrient level can easily be done by adding acids/bases and fresh medium in order to maintain the optimal condition for pollutant removal. Due to the presence of a continuously trickling water phase, biotrickling filters are a good option in certain industrial locations because of their reduced footprint and the capability to deliver nutrients and simultaneously control pH, as well as eliminate by-products accumulation from the filter bed. The operational advantages of biotrickling filters can be summarized as follows30,42,43: (i) it can be used to treat off gases that are at high temperatures, such as flue gas emissions, (ii) the pH of the trickling water phase can be controlled to handle acid producing compounds, (iii) better control of the temperature, nutrient concentration and biomass growth, and (iv) it can be customized to handle hydrophobic VOC with the addition of a second liquid phase such as silicone oil in order to improve the mass transfer of the pollutant from the gas to the liquid phase. In biotrickling filters, the liquid trickling rate will affect their performance because an increase in the liquid flow rate will improve wetting of the packing materials thereby facilitating the absorption of more hydrophilic pollutants in the trickling liquid phase. However, for practical applications, very high liquid to gas flow ratios might result in weeping of the liquid phase, interfering with the gas flow and increase the pressure drop. Sometimes, low liquid to gas flow ratios will result in flooding conditions over the filter bed, resulting in poor mass transfer characteristics as well as low removal efficiencies of the biotrickling filter.44

2.31.11 Bioscrubbers A bioscrubber comprises two operational units namely the absorption tower and the bioreactor (Fig. 4). In the absorption unit, i.e., the first stage, the contaminants present in the waste gas are transferred from the gas phase to a liquid phase. In this stage, the gas and

Gas-Phase Bioreactors

Figure 3

Schematic of a biotrickling filter.

Figure 4

Schematic of a bioscrubber

455

liquid phases usually flow counter-currently within a packed-bed unit. Meanwhile, the contaminants that are solubilized in the liquid phase are treated in the second stage bioreactor, which is usually a suspended growth bioreactor that resembles the conventional activated sludge process.1,20 From a process engineering viewpoint, the first stage absorption unit is flooded with a liquid phase in order to increase the mass transfer efficiency. Flooding the bed increases the ability of the liquid phase to absorb the pollutants because as the gas phase impacts the bed media, it forms tiny bubbles that greatly increase the surface area of the interface between the gas and liquid phases. Increasing the interfacial area within the absorption unit improves the liquid phase’s ability to absorb pollutants that have a Henry’s coefficient < 0.1.5,45

456

Gas-Phase Bioreactors

Concerning the second stage bioreactor, the residence time of the pollutant (in liquid phase) often varies between 20 and 40 days, i.e., the hydraulic retention time (HRT), depending on the concentration of the pollutants and the composition of the waste gas, its physico-chemical characteristics, activity of the microorganisms, flow rate of the liquid phase and the size of the reactor. In order to achieve high biodegradation rates, a nutrient solution should be periodically added to supply the levels of microand macronutrients required for optimal microbial activity. In some situations, recycling of the sludge is done within the bioreactor unit, and/or a part of the treated solution from the bioreactor unit is recycled to the absorption unit for the absorption of volatile compounds. The advantages of a bioscrubber can be stated as follows46,47: (i) the waste gas does not require a pre-humidification step, (ii) good long-term operational stability and good control of the operating parameters such as pH, temperature, and nutrient availability, (ii) elimination of a wide range of water-soluble compounds such as H2S, SO2, alcohols, aldehydes and fatty acids, (iv) lower pressure drops, i.e., no clogging problems compared to packed-bed bioreactors, and (v) smaller footprints than most biofilters and less maintenance costs.

2.31.12 Airlift Bioreactors In order to overcome the commonly reported operational problems of biofilters and biotrickling filters such as excessive biomass growth, flooding, high pressure drops, preferential flow channels, formation of anaerobic zones and drying of support, airlift bioreactors were developed.48 The airlift bioreactors are used when the elimination of water-soluble compounds is desired and when there is no mass transfer limitation for the transfer of contaminants from the gas to the liquid phase.49 An airlift bioreactor serves both as an absorption unit as well as a bioreactor unit (Fig. 5). The reactor comprises two concentric tubes, wherein the inner tube is shorter than the outer tube. The combined medium of air and water in the riser section has a lower density than the water in the downcomer section, and this density difference will facilitate continuous fluid circulation within the reactor. The driving force for the liquid recirculation within the bioreactor is the hydrostatic pressure difference between the riser and the downcomer section. The resisting force is the frictional pressure drop around the flow circuit. The biomass is suspended in the liquid medium and the water-soluble contaminants are degraded in the liquid phase.1 In this bioreactor type, less energy is required for aeration and mixing because the agitator or stirrer that are commonly used in mechanically stirred tank bioreactors is eliminated pneumatically by the incoming waste gas that contains oxygen. The operational advantages of an airlift bioreactor can be stated as follows50: (i) simple construction and the absence of moving parts, (ii) well-defined fluid flow patterns using an internal or external recirculating loop, (iii) good mass transfer properties for water-soluble gas phase pollutants, (iv) high thermal stability, (v) low energy consumption, and (vi) low construction and operation costs. The important design parameters that should be considered are the gas to liquid interfacial area, gas hold-up or gas to liquid mass transfer coefficient, the reactor design and its hydrodynamics, the gas distributor, and the gas and liquid flow rates.51 For biological waste gas treatment, that is governed by chemical and biochemical factors, the gas to liquid oxygen transfer rate should also be evaluated because the rate of oxygen consumption by microorganisms is usually high, compared to the rate of oxygen transfer from the gas to the liquid media.52 Mechanical aeration in airlift bioreactors might be required sometimes if the viscosity of the liquid medium increases due to excess biomass growth. This approach is expensive, and it might also cause volatilization and re-dispersion of the volatile pollutants present in the aqueous phase. For practical reasons, it is advisable to minimize excess biomass growth or the volatile solids (VS) content by removing a part of the biomass periodically during the maintenance step. In some instance, when there is fluctuation in the gas flow rate, i.e., an increase or decrease of the pollutant loading rate, the performance of airlift bioreactors can drastically decrease. In order to overcome such situations, it is recommended to control the biomass concentration in the bioreactor or to construct the bioreactor in the form of several modules that can be added or be eliminated according to the pollutant load.

Figure 5

Schematic of an airlift bioreactor.

Gas-Phase Bioreactors

Figure 6

457

Schematic of a fluidized bed bioreactor.

2.31.13 Fluidized Bed Bioreactors The fluidized bed bioreactor is, in some aspects, an improved version of the conventional biological systems such as biofilters or biotrickling filters. The schematic of a fluidized bed bioreactor is shown in Fig. 6. According to Kunii and Levenspiel,53 in a fluidized bed, fluidization promotes homogeneous conditions due to the rapid and uniform mixing of particles. The polluted air enters the reaction space with the help of a nozzle type distributor or sparger placed near the tapering section of the bioreactor. These bioreactors are generally constructed as hollow cylinders with a perforated distribution plate that is usually placed just above the sparger. An important design consideration, for biofilm development, is the selection of an appropriate particle type to use in the bed, and that the particle can be easily fluidized. The fluidized state is influenced by the particle type as well as the particle properties such as size (usually 0.15–0.3 mm), size distribution, density, shape, and its moisture holding capacity. Besides, the efficiency of fluidization is also affected by the mass and heat transfer characteristics, reactor geometry, the concentration of pollutants and the composition of the waste gas.54 Some of the typical examples for the particle type include sand, carbon, fly ash, anthracite, glass and calcinated clay. These fine particles can be easily fluidized by the upward flow of waste gas entering the bioreactor. The advantages of a fluidized bed bioreactor can be summarized as follows55: (i) ability to achieve complete mixing of the particles, (ii) appropriate distribution of temperature and minimum temperature gradient across the height of the bioreactor, (iii) good performance achieved in a smaller reactor volume, (iv) the accumulation of toxic end products can be eliminated by a simple replacement of the aqueous phase, (v) low pressure drop, and (vi) offers high surface area compared to a biofilter or a biotrickling filter. On the other hand, the poor performance of fluidized bed bioreactors could be due to the following reasons56: (i) the inability of microbial cells to attach to the particles during fluidization, (ii) surface properties of the particle, (iii) excess biomass growth in suspended state, and (iv) highly fluctuating gas flow rates that affect fluidization and cause channelling. Therefore, it is recommended to use highly porous matrices as the fluidized particles, which will facilitate better biomass attachment within the porous internal structure.

2.31.14 Membrane Bioreactors Membrane bioreactors are conventionally used in water and wastewater treatment. However, this reactor configuration (Fig. 7) was also applied for studies on biological waste gas treatment in order to overcome the mass transfer limitations reported in biofilters and biotrickling filters.1 The high permeability and affinity of some specific membranes for hydrophobic pollutants endorse its application in waste gas treatment. The driving force behind the membrane separation process is the difference in partial pressure between the different pollutants present in the waste gas stream.57 Membrane reactors can be applied in the chemical and petrochemical industries, various biotechnological applications, environmental protection, energy conversion and hydrogen production.58 This multifunctional reactor configuration offers the following advantages: (i) good process intensification, i.e., it can be used for resource recovery from waste treatment, (ii) compact design, (iii) low energy consumption, and (iv) the pollutant concentration gradient facilitates better mass transfer. Furthermore, the membrane bioreactor allows the selective permeation of the pollutant, which is often not achieved in other bioreactor configurations used for

458

Figure 7

Gas-Phase Bioreactors

Schematic of a membrane bioreactor.

waste gas treatment. The membrane modules serve as a support for the growth of the microbial population responsible for the biodegradation of the pollutants. The performance of membrane bioreactors depends on the following factors: (i) concentration and composition of pollutants present in the waste gas, (ii) biofilm growth on the membrane surface, (iii) the type of membrane material (polyvinylidene difluoride, polyethylene, polydimethylsiloxane, polypropylene), and (iv) the type of membrane configuration (hollow fiber, tubular, spiral wound, plate and frame, capillary).59,60 Nevertheless, the performance of membrane bioreactors can be affected if the waste gas contains a complex mixture of sulphur and nitrogen derived compounds, corrosive solvents or excess biomass growth that would lead to membrane fouling and even rupture of membranes. The performance of membrane bioreactors is determined in terms of the pollutant removal efficiency (RE, %), membrane fouling resistance (Rf) and flux decline, corresponding to different pollutant loading rates. At a given transmembrane pressure (TMP), the permeate flux, total membrane resistance, and the fouling resistance are calculated according61 to Eqs. (6)–(8):  
Permeate flux J; L m2 h ¼

V At

Total membrane resistance ðR t Þ ¼

(6)

TMP mJ

(7)

Fouling resistance ðR f Þ ¼ R t
R m

(8) 2

where V is the volume of the aqueous phase (L), A is the effective area of the membrane module (m ), m is the viscosity of the aqueous phase (Pa.S), and Rm is the intrinsic membrane resistance (m
1). From a practical viewpoint, when compared to biofilters and biotrickling filters, membrane bioreactors are advantageous due to the following reasons: (i) the water phase allows optimal humidification of the biofilm and it also facilitates the removal of biodegradation by-products that are acidic in nature, (ii) it prevents biomass inactivation, and (iii) its simple design allows the possibility to modify or retrofit the bioreactor independently without disturbing the main bioreactor unit. Although the performance of different membrane bioreactors has been demonstrated in laboratory conditions, their long-term stability has not yet been demonstrated at the industrial scale.

2.31.15 Hybrid Bioreactor Configurations In industrial situations, the operation of waste gas treatment systems can be interrupted for a few hours daily, sometimes for a few days continuously during weekends/holidays, or the bioreactor could experience unexpected shock loads, i.e., a severalfold increase in the pollutant concentration due to a process change or failure of unit processes. Such conditions are termed transient or shock loading conditions. Such conditions can be briefly stated as follows: (i) no pollutant is supplied to the bioreactor, and (ii) an unexpected increase or decrease in the gas flow rate and/or the inlet concentration. These types of fluctuating loading patterns will affect the performance of any bioreactor configuration that is designed to tolerate a certain critical loading rate of the pollutant as well as the microbial activity of the biofilm.

Gas-Phase Bioreactors

Figure 8

459

Schematic of a hybrid bioreactor comprising of a first stage biotrickling filter and a second stage biofilter.

Hybrid or two stage reactors have been reported to withstand such harsh operating conditions because these reactors can be connected in series or in a modified reactor configuration having different pollutant removal mechanisms.62 According to Rene et al.,20 the following reactor configurations can be integrated to achieve high removal efficiencies of the pollutants present in the waste gas: biotrickling filter followed by a biofilter (Fig. 8) and an adsorption column followed by a biofilter (Fig. 9). These hybrid reactor configurations can be applied in industries for the following situations19,20: (i) when the waste gas contains a mixture of inorganic and organic pollutants, and some of their degradation end products are highly acidic, (ii) when it is desired to maintain different microbial species in the bioreactors for the removal of a mixture of pollutants, and (iii) for the treatment of exceedingly high loading rates of the pollutants, which could inhibit microbial activity in the biological reactor. From a practical viewpoint of applying hybrid or two stage reactors, during the design and decision-making stage, factors such as the power requirements, available land area, and operational simplicity should be considered.

Figure 9

Schematic of a hybrid reactor comprising of a first stage adsorption column and a second stage biofilter.

460

Gas-Phase Bioreactors

2.31.16 Bioprocesses for the Conversion of C1 Gases (CO, CO2) in Bioreactors Greenhouse gases such as CO2 have become a major concern nowadays. Emission of such gases to the atmosphere needs to be reduced. One promising alternative consists in converting them into commercially useful products, in bioreactors, rather than simply eliminating those pollutants as done in the systems described above. Interestingly, gases such as CO and CO2 are also common compounds in syngas. Production of metabolites like alcohols by direct biosynthesis from syngas (CO, CO2, H2) or from C1-rich waste gases by anaerobic bacteria is considered a versatile production process and is attracting attention as an alternative to thermochemical processes. Compared to the thermochemical process, this bioconversion process, known as syngas or waste gas fermentation, is highly specific in terms of the product spectrum. The possibility to operate the bioreactor at near ambient temperature and pressure along with its great flexibility in terms of the gas composition are some other advantages compared to the thermochemical approach. The bioprocess based on gas fermentation employs acetogenic bacteria and is also an alternative to second generation bioethanol production in bioreactors, from lignocellulosic feedstocks.63 Off gases of certain industries, such as steel mills, are rich in carbon monoxide and do also contain carbon dioxide and hydrogen. Thus, those waste gases can be captured and processed applying the syngas fermentation technology to generate value-added products with simultaneous reduction in waste gas and greenhouse gas emissions. However, one of the main bottlenecks to be addressed in suspended-growth bioreactors is the limitation of the mass transfer rate of poorly soluble gases into the liquid medium for microbial access.64 Most utilized wild-type or genetically modified acetogens include Clostridium ljungdahlii, C. autoethanogenum, C.carboxidivorans and C. ragsdalei. These acetogens follow the Wood–Ljungdahl pathway for using the gaseous substrates as their carbon and energy sources, with generation of valuable products and commodities such as ethanol, butanol, hexanol, 2,3-butanediol, acetic acid, butyric acid and higher fatty acids, as well as biopolymers. The stoichiometry of metabolites production from syngas is given in Eqs. (9)–(15) for the case of ethanol and acetic acid. During the non-availability of hydrogen or at a state of inhibition of hydrogenase enzyme, the reducing equivalent is obtained from CO, thereby a decrease in concentration of products may occur.1 6CO þ 3H2O / C2H5OH þ 4CO2

(9)

2CO2 þ 6H2 / C2H5OH þ 3H2O

(10)

3CO þ 3H2 / C2H5OH þ CO2

(11)

2CO þ 4H2 / C2H5OH þ H2O

(12)

4CO þ 2H2O / CH3COOH þ 2CO2

(13)

2CO2 þ 4H2 / CH3COOH þ 2H2O

(14)

2CO þ 2H2 / CH3COOH

(15)

However, the product spectrum when fermenting syngas is greatly influenced by several operating conditions such as pH or medium composition, among others. In bioconversion processes involving both acidogenic (production of acids) and solventogenic (production of solvents) phases, acidogenesis together with biomass growth will usually take place at high pH, either neutral or slightly acidic. Lowering the fermentation pH below its optimum growth value results in shifting from acidogenesis to solventogenesis. This is attributed to the stress environment created by lipophilic weak acids by lowering the cytoplasmic pH, thereby disrupting the pH homeostasis. One way adopted by bacteria to overcome this situation is by converting these acids into the corresponding alcohols.65 Like many enzymes involved in this metabolic pathway, this conversion is being facilitated by the action of a metal-containing enzyme, acetaldehyde: Fd oxidoreductase (AFOR). The presence of tungsten is known to stimulate the activity of this enzyme. In addition, other enzymes such as carbon monoxide dehydrogenase (CODH) and formate dehydrogenase (FDH) do also contain metals at their active sites. Thus, the presence of specific trace metals in the fermentation broth may greatly influence the outcome of the process.63

2.31.17 Bioreactors for Syngas Fermentation Different bioreactors have been considered for syngas/waste gas fermentation, with ethanol being one of the most studied metabolites (Table 5). The mass transfer between the gas-liquid phase and kinetic limitations are major challenges in syngas fermentation in bioreactors. A mass transfer limited situation occurs when the system receives sufficient amount of cells, but the poor mass transfer of sparingly soluble gaseous substrate causes the cells to starve. In the case of kinetic limited conditions, the system receives sufficient amount of gaseous substrates; however, there are not enough cells to consume it at the same rate as the gas transfer. This may lead to substrate saturation in the system and eventually cause inhibition of the cells, as high CO concentrations are toxic to various enzymes involved in the metabolic pathway of acetogens. Both situations could happen at some point during syngas

Gas-Phase Bioreactors Table 5

461

Different bioreactor configurations for syngas fermentation into ethanol by acetogenic bacteria

Reactor

Syngas composition (%)

Microorganism

Ethanol production (g L
1)

Ethanol/acetic acid

HFM-BR

CO¼20, H2¼5, CO2¼15, N2¼60 CO¼20, H2¼5, CO2¼15, N2¼60 CO¼55, H2¼20, CO2¼10, Ar¼15 CO¼100 CO¼20, H2¼5, CO2¼15, N2¼60 CO¼60, H2¼35, CO2¼5 CO¼55, H2¼20, CO2¼10, Ar¼15 CO¼13, H2¼14, CO2¼5, N2¼68

C. carboxidivorans P7

23.93

4.79

C. carboxidivorans P7

4.89

2.1

C. ljungdahlii

6.5

1.53

C. autoethanogenum C. ragsdalei P11

0.867 25.26

NA 6.8

C. ljungdahlii ERI-2

19.73

3

C. ljungdahlii

48

21

C. ljungdahlii ERI-2

2.74

0.64

MBR STR STR STR CSTR-BC CSTR (Cell recycle) ICR

Abbreviations: HFM-BR, Hollow fiber membrane biofilm reactor; MBR, Monolith biofilm reactor; CSTR; Continuous stirred tank reactor with continuous liquid and gas flow; STR, Stirred tank reactor with liquid batch; CSTR-BC, CSTR-bubble column; ICR, Immobilized cell reactor; NA, Not applicable. Ethanol/acetic acid ratio is given in molar concentrations. Adapted from Ref. 63

fermentation, and thus, properly addressing these challenges is required for successful implementation of the process and its commercialization. Several strategies have been adopted to improve the mass transfer of gaseous substrates and to allow them to easily reach the cells, considering that the resistance is largely contributed by the liquid film across the gas-liquid interface. The overall mass transfer rate (R) of sparingly soluble syngas to the fermentation medium through the liquid film is determined using Eq. (16): R ¼ KL a ðC
CL Þ

(16)

where KL is the diffusion coefficient per unit area of liquid film (ms
1), a is the interfacial area per unit volume (m2m
3), C is the saturation concentration in the liquid phase and CL is the dissolved gas concentration. To improve the overall performance of the process, increasing the volumetric mass transfer coefficient ðKL aÞ is one way which is influenced by many factors such as the geometry and operating characteristics of the reactor, the rheology of the fermentation medium, the syngas flow rate/pressure or the morphology of the biocatalyst.66 A general approach to improve the KL a is by increasing the impeller speed in stirred tank reactors (STRs), a reactor configuration often used in lab-scale syngas fermentation studies. However, this can cause damage to the microbes and will cause an increment in overall power requirements. A limited number of studies have been reported on bubble column reactors (BCRs) and gas-lift reactors (GLR) which offer several advantages over STR such as non-requirement of mechanical agitator, low shear rate and ease to scale-up. The syngas is fed through a sparger in the bottom side of the bioreactor at the center, which causes a pressure difference and thus leads to circulation of fluid through the system. Higher KL a can be achieved by increasing the flow rate; however, this will cause a heterogeneous flow inside the system which eventually leads to back mixing. In addition, the power requirements will linearly increase with increasing flow rate. Recently, membrane bioreactors (MBR) have been used as well, with hollow fiber membranes acting as a gas diffuser by creating microbubbles as well as a support for the biofilm. Hydrophobic membranes such as polypropylene, polyethylene, and polyvinylidene fluoride membranes are preferred over hydrophilic ones as the former are characterized by no filling of the membrane pore with liquid (pore wetting), and they prevent the accumulation of dead cells inside or over membrane pores (biofouling). In a study using an external type hollow fiber membrane bioreactor, a maximum KL a of 1096.2 h
1 was achieved using polypropylene hollow fiber membranes, which is much higher than obtained with a CSTR (35.5 h
1).67 In another study, a monolith biofilm bioreactor exhibited a higher KL a compared to a BCR. This might be due to the even distribution of the liquid inside the monolithic channels and the short diffusion path within the uniform parallel channels.68 Other bioreactor configurations such as the biotrickling filter and moving bed biofilm bioreactor (MBBR) have also successfully been tested for syngas fermentation. The MBBR employs inert biomass carriers for supporting microbial growth. The carriers are suspended in the fermentation broth inside a vessel. Syngas is fed from the bottom of the reactor and rises through the liquid medium while being used up by the microbes to convert it into end products.

2.31.18 Multistage Syngas Fermentation When considering the production of solvents such as alcohols, acetogens are characterized by two metabolic phases. The first one is an acidogenic phase which occurs when the bioreactor system is supplied with sufficient nutrients (e.g., macronutrients, vitamins, minerals) at optimal growth pH and temperature. In such case, high growth is observed associated with high production of organic acids (e.g., acetic acid). The solventogenic phase starts when the bioreactor creates a stress environment for the cells such as a lower pH or nutrient limitation. The low pH can be the result of the production of acids in the acidogenic phase. This metabolic phase is characterized by a higher production of solvents (alcohols) with little or no growth. Acetogens dispose the abundant reducing

462

Gas-Phase Bioreactors

equivalents by converting acids to more reduced alcohols. In this respect, some studies have also been performed by coupling two STRs or an STR-BCR system incorporating a cell recycle unit which allows acetic acid production in the first bioreactor, providing optimum conditions for growth, while conditions which are not optimal for growth are used in the second bioreactor where the produced acids (e.g., acetic acid) from the first reactor are used up for the production of alcohols (e.g., ethanol).1

2.31.19 Conclusions The conventional gas cleaning techniques based on physical and chemical methods are known for their high cost and energyintensive process that destroys most of the energy containing molecules (chemicals) present in the waste gas. Although limited information is available in full-scale systems for the conversion of waste gases to valuable resources such as liquid fuels, enzymes, and chemicals, there is sufficient evidence from laboratory scale research to prove the feasibility of such conversion using bioprocesses. Novel process applications, bioreactor designs and a sound financial investment for resource recovery from waste gases are important for process intensification, scale up and commercializing the developed biotechnologies. Besides, the surplus availability of natural gas and syngas has recently motivated several researchers to convert gases such as CH4, CO, CO2 and H2 to liquid transportation fuels or other platform chemicals.

Acknowledgments HNA, MCV and CK belong to the BIOENGIN group which is financially supported by the Xunta de Galicia as Competitive Reference Research Group (GRC) (ED431C 2017/66). HNA thanks the Xunta de Galicia, Spain, for his postdoctoral funding (reference ED481B 2016/195-0). JD thanks Nuffic, the Dutch organization for internationalisation in education, for providing a fellowship to complete his MSc degree at IHE-Delft, The Netherlands (2016–18). ERR and EVH thank IHE-Delft for infrastructural and staff time support.

References 1. Kennes, C., Veiga, M. C., Eds. Air Pollution Prevention and Control: Bioreactors and Bioenergy, Wiley: Chichester, UK, 2013. 2. Fernández-Naveira, Á.; Veiga, M. C.; Kennes, C. H-B-E (Hexanol-butanol-ethanol) Fermentation for the Production of Higher Alcohols from Syngas/waste Gas. J. Chem. Technol. Biotechnol. 2017, 92, 712–731. 3. Barbusinski, K.; Kalemba, K.; Kasperczyk, D.; et al. Biological Methods for Odor Treatment - a Review. J. Clean. Prod. 2017, 152, 223–241. 4. Yang, C.; Qian, H.; Li, X.; et al. Simultaneous Removal of Multicomponent VOCs in Biofilters. Trends Biotechnol. 2018, 36, 673–685. 5. Kennes, C.; Rene, E. R.; Veiga, M. C. Bioprocesses for Air Pollution Control. J. Chem. Technol. Biotechnol. 2009, 84, 1419–1436. 6. Vikrant, K.; Kim, K. H.; Szulejko, J. E.; et al. Bio-filters for the Treatment of VOCs and Odors - a Review. Asian J. Atmos. Environ. 2017, 11, 139–152. 7. Prado, O. J.; Gabriel, D.; Lafuente, J. Economical Assessment of the Design, Construction and Operation of Open-Bed Biofilters for Waste Gas Treatment. J. Environ. Manag. 2009, 90, 2515–2523. 8. Ramírez-Sáenz, D.; Zarate-Segura, P. B.; Guerrero-Barajas, C.; et al. H2S and Volatile Fatty Acids Elimination by Biofiltration: Clean-Up Process for Biogas Potential Use. J. Hazard Mater. 2009, 163 (2–3), 1272–1281. 9. Chaiprapat, S.; Mardthing, R.; Kantachote, D.; et al. Removal of Hydrogen Sulfide by Complete Aerobic Oxidation in Acidic Biofiltration. Process Biochem. 2011, 46, 344–352. 10. Rattanapan, C.; Boonsawang, P.; Kantachote, D. Removal of H2S in Down-flow GAC Biofiltration Using Sulfide Oxidizing Bacteria From Concentrated Latex Wastewater. Bioresour. Technol. 2009, 100, 125–130. 11. Dumont, E.; Guzman, L. M. A.; Susa, M. S. R.; et al. H2S Biofiltration Using Expanded Schist as Packing Material: Performance Evaluation and Packed-bed Tortuosity Assessment. J. Chem. Technol. Biotechnol. 2012, 87, 725–731. 12. Zhu, R.; Li, S.; Wu, Z.; et al. Performance Evaluation of a Slow-release Packing Material-embedded Functional Microorganisms for Biofiltration. Environ. Technol. 2016, 28, 1–11. 13. Jaber, M. B.; Couvert, A.; Amrane, A.; et al. Hydrogen Sulfide Removal from a Biogas Mimic by Biofiltration Under Anoxic Conditions. J. Environ. Chem. Eng. 2017, 5, 5617–5623. 14. Jaber, M. B.; Couvert, A.; Amrane, A.; et al. Removal of Hydrogen Sulfide in Air Using Cellular Concrete Waste: Biotic and Abiotic Filtrations. Chem. Eng. J. 2017, 319, 268–278. 15. Rabbani, K. A.; Charles, W.; Kayaalp, A.; et al. Biofilter for Generation of Concentrated Sulphuric Acid from H2S. Environ. Sci. Pollut. Control Ser. 2016, 23, 16781–16789. 16. Lebrero, R.; Rodríguez, E.; Martin, M.; et al. H2S and VOCs Abatement Robustness in Biofilters and Air Diffusion Bioreactors: a Comparative Study. Water Res. 2010, 44, 3905–3914. 17. Morgan-Sagastume, J. M.; Noyola, A. Hydrogen Sulfide Removal by Compost Biofiltration: Effect of Mixing the Filter Media on Operational Factors. Bioresour. Technol. 2006, 97, 1546–1553. 18. Gostomski, P. A.; Sisson, J. B.; Cherry, R. S. Water Content Dynamics in Biofiltration: The Role of Humidity and Microbial Heat Generation. J. Air Waste Manag. Assoc. 1997, 47, 936–944. 19. Rene, E. R.; Veiga, M. C.; Kennes, C. Biofilters. In Air Pollution Prevention and Control: Bioreactors and Bioenergy; Kennes, C., Veiga, M. C., Eds.; Wiley: Chichester, UK, 2013; pp 57–119. 20. Rene, E. R.; Veiga, M. C.; Kennes, C. Innovative Bioreactors and Two-stage Systems. In Air Pollution Prevention and Control: Bioreactors and Bioenergy; Kennes, C., Veiga, M. C., Eds.; Wiley: Chichester, UK, 2013; pp 221–246. 21. Rene, E. R.; Sergienko, N.; Goswami, T.; et al. Effects of Concentration and Gas Flow Rate on the Removal of Gas-phase Toluene and Xylene Mixture in a Compost Biofilter. Bioresour. Technol. 2018, 248, 28–35. 22. Rodríguez, E.; García-Encina, P. A.; Muñoz, R.; et al. Microbial Community Changes during Different Empty Bed Residence Times and Operational Fluctuations in an Air Diffusion Reactor for Odor Abatement. Sci. Total Environ. 2017, 590–591, 352–360. 23. Cheng, Z.; Lu, L.; Kennes, C.; et al. Treatment of Gaseous Toluene in Three Biofilters Inoculated with Fungi/bacteria: Microbial Analysis, Performance and Starvation Response. J. Hazard Mater. 2016, 303, 83–93. 24. Dorado, A. D.; Baquerizo, G.; Maestre, J. P.; et al. Modeling of a Bacterial and Fungal Biofilter Applied to Toluene Abatement: Kinetic Parameters Estimation and Model Validation. Chem. Eng. J. 2008, 140, 52–61. 25. Kennes, C.; Veiga, M. C. Fungal Biocatalysts in the Biofiltration of VOC-polluted Air. J. Biotechnol. 2004, 113, 305–319.

Gas-Phase Bioreactors

463

26. Cheng, Z.; Li, C.; Kennes, C.; et al. Improved Biodegradation Potential of Chlorobenzene by a Mixed Fungal-bacterial Consortium. Int. Biodeterior. Biodegrad. 2017, 123, 276–285. 27. Menard, C.; Ramirez, A. A.; Nikiema, J.; et al. Biofiltration of Methane: Effect of Temperature and Nutrient Solution. WIT Trans. Ecol. Environ. 2009, 127, 239–245. 28. Malakar, S.; Saha, P. D.; Baskaran, D.; et al. Comparative Study of Biofiltration Process for Treatment of VOCs Emission from Petroleum Refinery Wastewater - a Review. Environ. Technol. Innov. 2017, 8, 441–461. 29. López, M. E.; Rene, E. R.; Boger, Z.; et al. Modelling the Removal of Volatile Pollutants under Transient Conditions in a Two-stage Bioreactor Using Artificial Neural Networks. J. Hazard Mater. 2017, 324, 100–109. 30. López, M. E.; Rene, E. R.; Malhautier, L.; et al. One-stage Biotrickling Filter for the Removal of a Mixture of Volatile Pollutants from Air: Performance and Microbial Community Analysis. Bioresour. Technol. 2013, 138, 245–252. 31. Datta, I.; Fulthorpe, R. R.; Sharma, S.; et al. High-temperature Biotrickling Filtration of Hydrogen Sulphide. Appl. Microbiol. Biotechnol. 2007, 74, 708–716. 32. Ryu, H. W.; Yoo, S. K.; Choi, J. M.; et al. Thermophilic Biofiltration of H2S and Isolation of a Thermophilic and Heterotrophic H2S-degrading Bacterium, Bacillus sp. TSO3. J. Hazard Mater. 2009, 168, 501–506. 33. Xue, S.; Chen, W.; Deng, M.; et al. Effects of Moisture Content on the Performance of a Two-stage Thermophilic Biofilter and Choice of Irrigation Rate. Process Saf. Environ. Protect. 2018, 113, 164–173. 34. Zhang, J.; Li, L.; Liu, J. Effects of Irrigation and Water Content of Packing Materials on a Thermophilic Biofilter for SO2 Removal: Performance, Oxygen Distribution and Microbial Population. Biochem. Eng. J. 2017, 118, 105–112. 35. Lehmann, J.; Czimczik, C.; Laird, D.; et al. Stability of Biochar in Soil. In Biochar for Environmental Management: Science and Technology; Lehmann, J., Joseph, S., Eds., Earthscan: London, 2009; pp 183–206. 36. Oliveira, F. R.; Patel, A. K.; Jaisi, D. P.; et al. Environmental Application of Biochar: Current Status and Perspectives. Bioresour. Technol. 2017, 246, 110–122. 37. Xu, X.; Cao, X.; Zhao, L.; et al. Comparison of Sewage Sludge- and Pig Manure-derived Biochars for Hydrogen Sulfide Removal. Chemosphere 2014, 111, 296–303. 38. Shang, G.; Li, Q.; Liu, L.; et al. Adsorption of Hydrogen Sulfide by Biochars Derived from Pyrolysis of Different Agricultural/forestry Wastes. J. Air Waste Manag. Assoc. 2016, 66, 8–16. 39. Kanjanarong, J.; Giri, B. S.; Jaisi, D. P.; et al. Removal of Hydrogen Sulfide Generated during Anaerobic Treatment of Sulfate-laden Wastewater Using Biochar: Evaluation of Efficiency and Mechanisms. Bioresour. Technol. 2017, 234, 115–121. 40. Das, J. Removal of Hydrogen Sulphide in a Biofilter Packed with Compost and Biochar (M.Sc. thesis); Delft, the Netherlands: IHE-Delft Institute for Water Education, 2018. 41. San-Valero, P.; Dorado, A. D.; Martínez-Soria, V.; et al. Biotrickling Filter Modeling for Styrene Abatement. Part 1: Model Development, Calibration and Validation on an Industrial Scale. Chemosphere 2018, 191, 1066–1074. 42. Álvarez-Hornos, F.-J.; Martínez-Soria, V.; Marzal, P.; et al. Performance and Feasibility of Biotrickling Filtration in the Control of Styrene Industrial Air Emissions. Int. Biodeterior. Biodegrad. 2017, 119, 329–335. 43. Parnian, P.; Zamir, S. M.; Shojaosadati, S. A. Styrene Vapor Mass Transfer in a Biotrickling Filter: Effects of Silicone Oil Volume Fraction, Gas-to-liquid Flow Ratio, and Operating Temperature. Chem. Eng. J. 2016, 284, 926–933. 44. Treybal, R. E., Ed.; Mass Transfer Operations, McGraw Hill: New York, 1980. 45. van Groenestijn, J. W.; Hesselink, P. G. Biotechniques for Air Pollution Control. Biodegradation 1993, 4, 283–301. 46. Mudliar, S.; Giri, B.; Padoley, K.; et al. Bioreactors for Treatment of VOCs and Odours - a Review. J. Environ. Manag. 2010, 91, 1039–1054. 47. Liu, F.; Fiencke, C.; Guo, J.; et al. Performance Evaluation and Optimization of Field-scale Bioscrubbers for Intensive Pig House Exhaust Air Treatment in Northern Germany. Sci. Total Environ. 2017, 579, 694–701. 48. Vergara-Fernández, A. O.; Quiroz, E. F.; Aroca, G. E.; et al. Biological Treatment of Contaminated Air with Toluene in an Airlift Reactor. Electron. J. Biotechnol. 2008, 11, 1–7. 49. Lo, C. S.; Hwang, S. J. Degradation of Waste Gas Containing Toluene in an Airlift Bioreactor. Environ. Sci. Technol. 2004, 38, 2271–2280. 50. Drandev, S.; Penev, K. I.; Karamanev, D. Study of the Hydrodynamics and Mass Transfer in a Rectangular Air-lift Bioreactor. Chem. Eng. Sci. 2016, 146, 180–188. 51. Vial, C.; Poncin, S.; Wild, G.; et al. A Simple Method for Regime Identification and Flow Characterisation in Bubble Columns and Airlift Reactors. Chem. Eng. Process. 2001, 40, 135–151. 52. Jurascík, M.; Blazej, M.; Annus, J.; et al. Experimental Measurements of Volumetric Mass Transfer Coefficient by the Dynamic Pressure-step Method in Internal Loop Airlift Reactors of Different Scale. Chem. Eng. J. 2006, 125, 81–87. 53. Kunii, D.; Levenspiel, O. Fluidization Engineering, 2nd ed.; Butterworth-Heinemann Series in Chemical Engineering: Newton, MA, USA, 1991. 54. Clarke, K.; Hill, G. A.; Pugsley, T. Improved VOC Bioremediation Using a Fluidized Bed Peat Bioreactor. Process Saf. Environ. Protect. 2008, 86, 283–290. 55. Mofidi, A.; Asilian, H.; Jonidi, J. A. Adsorption of Volatile Organic Compounds on Fluidized Activated Carbon Bed. Health Scope 2012, 2, 84–89. 56. Clarke, K.; Hill, G. A.; Pugsley, T. S. Direct Comparison of Fluidized and Packed Bed Bioreactors for Bioremediation of an Air Pollutant. Int. J. Chem. React. Eng. 2007, 5, 1–12. 57. López, M. E.; Rene, E. R.; Veiga, M. C.; Kennes, C. Biogas Technologies and Cleaning Techniques. In Environmental Chemistry for a Sustainable World; Lichtfouse, E., Schwarzbauer, J., Robert, D., Eds.; Springer: Dordrecht, The Netherlands, 2012; pp 347–377. 58. Brunetti, A.; Zito, P. F.; Giorno, L.; et al. Membrane Reactors for Low Temperature Applications: an Overview. Chem. Eng. Process. 2018, 124, 282–307. 59. Kumar, A.; Dewulf, J.; Van Langenhove, H. Membrane-based Biological Waste Gas Treatment. Chem. Eng. J. 2008, 136, 82–91. 60. Lebrero, R.; Volckaert, D.; Pérez, R.; et al. A Membrane Bioreactor for the Simultaneous Treatment of Acetone, Toluene, Limonene and Hexane at Trace Level Concentrations. Water Res. 2013, 47, 2199–2212. 61. Priya, V. S.; Philip, L. Membrane Bioreactor for the Treatment of VOC Laden Pharmaceutical Wastewater: Effect of Biological Treatment Systems on Membrane Performance. J. Water Process Eng. 2015, 7, 61–73. 62. Rene, E. R.; López, M. E.; Veiga, M. C.; et al. Steady- and Transient-state Operation of a Two-stage Bioreactor for the Treatment of a Gaseous Mixture of Hydrogen Sulphide, Methanol and a-pinene. J. Chem. Technol. Biotechnol. 2010, 85, 336–348. 63. Kennes, D.; Abubackar, H. N.; Diaz, M.; et al. Bioethanol Production from Biomass: Carbohydrate vs Syngas Fermentation. J. Chem. Technol. Biotechnol. 2016, 91, 304–317. 64. Yasin, M.; Jeong, Y.; Park, S.; et al. Microbial Synthesis Gas Utilization and Ways to Resolve Kinetic and Mass-transfer Limitations. Bioresour. Technol. 2015, 177, 361–374. 65. Abubackar, H. N.; Veiga, M. C.; Kennes, C. Production of Acids and Alcohols from Syngas in a Two-stage Continuous Fermentation Process. Bioresour. Technol. 2018, 253, 227–234. 66. Wainaina, S.; Horváth, I. S.; Taherzadeh, M. J. Biochemicals from Food Waste and Recalcitrant Biomass via Syngas Fermentation: A Review. Bioresour. Technol. 2018, 248, 113–121. 67. Shen, Y.; Brown, R.; Wen, Z. Syngas Fermentation of Clostridium carboxidivorans P7 in a Hollow Fiber Membrane Biofilm Reactor: Evaluating the Mass Transfer Coefficient and Ethanol Production Performance. Biochem. Eng. J. 2014, 85, 21–29. 68. Shen, Y.; Brown, R.; Wen, Z. Enhancing Mass Transfer and Ethanol Production in Syngas Fermentation of Clostridium carboxidivorans P7 through a Monolithic Biofilm Reactor. Appl. Energy 2014, 136, 68–76.

2.32

Fuel-Cell Bioreactors

J Vilas Boas, VB Oliveira, and AMFR Pinto, CEFT, Departamento de Eng. Química, Universidade do Porto, Faculdade de Engenharia, Porto, Portugal M Simo˜es, LEPABE, Departamento de Eng. Química, Universidade do Porto, Faculdade de Engenharia, Porto, Portugal © 2019 Elsevier B.V. All rights reserved.

2.32.1 Fuel Cell Bioreactors 2.32.1.1 Introduction 2.32.1.2 Operating Principle 2.32.1.3 Microorganisms, Substrates and Electron Transfer Mechanisms 2.32.1.4 Reactor Design 2.32.1.5 Operating Conditions 2.32.1.5.1 Electrode Material 2.32.1.5.2 Membrane 2.32.1.5.3 pH 2.32.1.5.4 Temperature 2.32.1.5.5 Hydrodynamic Stress 2.32.1.5.6 Organic Load 2.32.1.6 Models and Modeling Approaches 2.32.1.7 Diagnostic Techniques 2.32.1.8 Applications 2.32.1.8.1 Wastewater Treatment 2.32.1.8.2 Hydrogen Production 2.32.1.8.3 Bioremediation 2.32.1.8.4 Biosensors 2.32.1.8.5 Desalination 2.32.1.8.6 Other Applications 2.32.1.9 Economic Evaluation and Life Cycle Assessment 2.32.1.10 Challenges and Perspectives 2.32.2 Concluding Remarks Acknowledgments References

2.32.1

Fuel Cell Bioreactors

2.32.1.1

Introduction

464 464 464 465 467 468 468 469 469 470 470 470 471 471 472 472 472 473 473 473 473 473 474 475 476 476

The global strategy for the next decades targets a smart and sustainable economy based on knowledge, innovation and efficient use of resources, especially the environmental-friendly ones. Among the different cutting-edge technologies emerged in the last decade, the Fuel Cell Bioreactors (FCB) became part of the portfolio of technologies, which aim the development of low-carbon, sustainable and secure energy supply. Therefore, FCB have been the focus of many studies and research due to their potential to avoid two global problems: wastes and energy demand. In addition, these systems can be adaptable to a wide variety of applications, being used for energy, hydrogen and chemicals of high value production and wastewater treatment. The main problems of FCB are the lower performance levels, higher costs and having to compete with the well-established conventional technologies. Therefore, for its successful implementation, the FCB benefits (energy, hydrogen and chemicals production and wastewater treatment) should be higher than its costs (of implementation and operational). Additionally, until now, only a few pilot studies have been performed in real conditions, so more pilot and durability studies and demonstration projects are needed to prove the reliability of these systems. This chapter, starts by covering the fundamentals of the FCB, including a description of the operating principle, microorganisms, substrates and electron transfer mechanisms, reactor design, and presenting the role of the operating conditions. Then, an overview on the models and modeling approaches and the diagnostic techniques used to analyze and optimize these systems is provided, followed by its main applications. This chapter, also, includes an economic evaluation and the life cycle assessment of this technology, as well as its main challenges and perspectives.

2.32.1.2

Operating Principle

The FCB are bioelectrochemical devices that share similarities with biological reactors and chemical fuel cells. In fact, the FCB work under the concept of a biological reactor, in which the microorganisms grow (anaerobically) oxidizing organic substrates and

464

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00143-9

Fuel-Cell Bioreactors

Figure 1

Schematic diagram of an FCB.

Figure 2

Operating principle of an FCB.

465

producing electrons and protons and a chemical fuel cell, by being able to produce energy through the transfer of these electrons and protons to the opposite side, usually the cathode side, where a reduction reaction occurs. It also produces water, a small amount of carbon dioxide (CO2) and heat (Fig. 1). An FCB includes two electrodes, anode and cathode, that are, normally, separated by a proton exchange membrane (PEM) and connected by an external conductive wire that conducts the electrons from one side to the other (Fig. 2). The PEM allows the proton diffusion toward the cathode and should prevent the diffusion of other components from reaching the cathode, as well as the undesirable oxygen diffusion toward the anode. The electrochemical reactions that take place at each electrode are responsible for a continuous flow of electrons and protons, producing energy, due to the difference between the anodic and cathodic redox potentials. Therefore, the power output of an FCB depends on the electrochemical reactions occurring in both electrodes and on the different losses that naturally affect these systems. These include the activation loss, due to the activation energy that must be overcome by the reacting species, the Ohmic loss, due to ionic and electronic conduction, and the concentration loss, due to a decrease of the substrate concentration and to mass transport limitations. Polarization curves and electrochemical impedance spectroscopy (EIS) measurements indicate these losses and the extent of each one pointing out possible solutions to minimize them in order to achieve the ideal performance. Fig. 2 shows the operating principle of an FCB. The microorganisms grow at the anode, oxidizing the organic matter and releasing electrons and protons, which, at the cathode, combine with oxygen (or other electron acceptor) to produce water (or other reduced compounds). In the anodic chamber, the microorganisms can be in the anodic environment in their planktonic form or more desirably attached at the electrode surface forming a biofilm (Fig. 2). A biofilm can be defined as an aggregate of single and/or multi species microorganisms, conventionally adhered on a surface, that naturally produces a matrix of extracellular polymeric substances (EPS) that can represent more than 90% of the biofilm dry mass.1 The EPS matrix is essentially constituted by polysaccharides, proteins, DNA and lipids and has an important role on providing the biofilm attachment on the electrode surface, allowing mechanical stability, enhancing microbial cells communication, accumulation of nutrients and even protection in unfavorable environments.1,2 The described operating principle can present some deviations based on the different FCB applications and design, as will be seen further ahead in this chapter.

2.32.1.3

Microorganisms, Substrates and Electron Transfer Mechanisms

The role of microorganisms in FCB and their interactions with the electrode is of major importance. Several microorganisms, mostly bacteria, have been used in single cultures and identified in biofilms formed on the FCB anode electrode.3 However, not all of them

466

Figure 3

Fuel-Cell Bioreactors

Schematic representation of the different electron transfer mechanisms.

have the ability to transfer the electrons to the electrode, and only few species were identified as having this ability, being known as electrochemically active bacteria, electroactive bacteria, anode respiring bacteria, electricigens or electrogens among other designations.4–6 These include the Gram-negative ones of the genus Geobacter and Shewanella that are by far the most studied and commonly used in FCB.7–10 In fact, metal-reducing bacteria belonging to the phylum Proteobacteria are attractive for FCB due to their ability to use soluble and insoluble compounds as terminal electron acceptors, such as Fe(III), having potential to use the electrode (anode) for the same purpose.11 Additionally, these species completely oxidize the substrates to carbon dioxide and develop biofilms on the anode electrode, allowing an increase on the FCB energy production. Based on each microbial ability to transfer electrons, the electron transfer mechanism may be categorized in two groups: the direct electron transfer (DET) and the mediated electron transfer (MET) (Fig. 3). As the name suggests, the DET implies the direct contact between the microorganism and the anode surface, via cytochromes and conductive pili or nanowires.5,6 For DET, the development of a biofilm is essential to attach/sustain the microorganisms on the electrode surface, increasing the energy production.5,6 The strain Shewanella oneidensis MR-1 is well known for its ability to transfer electrons via DET, but the mechanism is a little different than the usual one, since this strain produces outer membrane and periplasmic extensions to provide the contact with the anode.10 This strain forms thin biofilms, which may limit FCB energy production and in their planktonic form is also able to deliver electrons by MET.9,10 The MET is possible due to electron shuttles, or mediators, that are added to the system or self-produced by the microbial cell. A good mediator should cross the cell membrane easily, have a high electrode reaction rate, be non-toxic to microbes and have low cost. These include organic molecules that can be reversibly reduced, inorganic compounds, such as hydrogen or sulfurous compounds, and chemical ones, such as methylene blue or neutral red. However, the last ones have a high toxicity and cost, instability and low efficiency limiting its application in FCB.4 S. oneidensis uses self-produced flavins as shuttles to transfer the electrons to the anode electrode. The flavins showed also positive effects on electron transfer in non-electroactive microorganisms.12 The biofilm matrix is also an important parameter that affects the electron transfer mechanisms as it provides cell-to-cell communication, increasing the electron transfer from the cells that are not in physical contact with the anode and hydrolyze the substrate producing additional protons and electrons.5 Different organic compounds, such as acetate, lactate, glucose or pyruvate, were used as substrate on FCB. Among them, acetate is the most used one, since it has a simple structure, being easily degraded by the microorganisms and providing a higher energy output.13 However, according to the FCB applications, more complex substrates, such as wastewater, soil, marine sediments or leachates, are used. It was found that, to oxidize them, a consortium, where different microorganisms have specific metabolism functions, is preferable than a single culture, since besides increasing the oxidation rate it produces more electricity.14,15 Activated sludges or marine sediments are commonly used as inoculum for FCB, enabling electroactive bacteria to be present on the biofilms developed at the FCB electrodes.16,17 For practical applications, a system with a non-defined consortium is desirable as it is easier to operate and requires lower operational control. For lab-scale studies, single cultures are preferred and commonly used, since they allow studying electron transfer mechanisms, with some details, allowing an easier microbial characterization and FCB optimization. Microorganisms in single or multiple cultures can also be used to perform the cathodic reaction by accepting the electrons from the cathode electrode,18 renaming the cathode as biocathode.18–20 According to the different FCB applications (Section 2.32.1.8.), the biocathode can work aerobically, where the microorganisms reduce the oxygen to produce water, or anaerobically, where the microorganisms use diverse electron acceptors, for example, CO2, to produce compounds such as acetate, butyrate, ethanol or butanol.19,20 This is a relatively new research field, so the corresponding electron transfer mechanisms are poorly understood. Based on the already established mechanisms for the anode reaction, the most acceptable mechanism for the cathode rely on a mediated electron transport between the electrode and the microorganism by hydrogen (or other compound) that is produced chemically or biologically on the cathode during the process.19–21 On the presence of a multispecies biofilm, the most acceptable mechanism is the interspecies electron transfer, suggesting that the protons are reduced to H2 by electroactive bacteria and the H2 is further oxidized by other microorganisms, commonly acetogenic, receiving the electrons.4,19,22

Fuel-Cell Bioreactors

467

Understanding and controlling the electron transfer mechanisms is crucial to increase and sustain the energy production by FCB. However, until now, only the mechanism of few microorganisms was clearly identified and characterized, remaining a lack of knowledge on complex consortia and how the environmental conditions affect the biofilm development/growth.

2.32.1.4

Reactor Design

FCB designs commonly used are schematically depicted in Fig. 4 and depend on the number of compartments, application, costeffective construction and scale-up viability. The conventional design is a dual-chamber and, contains two different compartments or chambers, one the anode and the other the cathode, physically separated by a separator or PEM (Section 2.32.1.2). The FCB employing this design works according to the operating principle described in Section 2.32.1.2 and Fig. 2. The dual-chamber FCB can have an “H-shape” or a more compact layout, by shortening the electrodes distance (Fig. 4A).4,7,9,21 As cathodic electron acceptor was initially used as a solution of ferricyanide, this solution has been replaced by oxygen dissolved in water, through an inlet of air on this compartment, a more economically and environmental friendly solution.23 However, as this airflow lead to an extra energy consumption, more cost-effective designs emerged to suppress this disadvantage. Among them, the most well-known is the single-chamber design or air-cathode configuration (Fig. 4B). As the name suggests, this design is characterized for having only one compartment, the anodic one, so the cathode is opened to the surroundings and its electrode is in closer contact with the membrane.24,25 This design is very attractive, since the need of an airflow is eliminated, reducing the energy requirements, and due to its simplicity is more suitable for scale-up, toward the use of FCB in real applications. Additionally, it has been proven that operation with this design can avoid the use of the membrane, which is responsible for a major portion on these systems costs, and allow decreasing the distances between the anode and cathode electrodes, decreasing the cell internal resistances.24,25 The anodic chamber can have different shapes, such as cubic or cylindrical (tubular). Another design used is the tubular up-flow, based on the design of the up-flow anaerobic sludge blanket, and comprises two chambers, anodic and cathodic. The cathodic chamber may be placed on the top of the anodic chamber (Fig. 4C) or inside of it.26,27 This design leads to a high cell density and a high mass transfer efficiency.28 The up-flow design can, also, have a single chamber, where the cathode electrode is placed at the inner or outer side of the tubular anode (Fig. 4D).29 Despite the advantages and disadvantages of each of these designs, until now all failed on the scale-up process through the increase of the two compartment’s volume, since the scaled-up system suffers from higher ohmic, activation and concentration losses. Therefore, this strategy becomes less desirable, and it has been seen that the best solution is using small reactors connected in series or parallel, staked systems, through tubular or flat designs (Fig. 4E).30 Another alternative, easy to implement and scale-up, is using several electrodes on the anodic and cathodic compartments.31 However, stacking FCB is still the most promising alternative, allowing a more efficient system with increased energy rates at both lab and pilot scales.32–34

Figure 4 FCB designs: (A) dual-chamber, H-shape and cubic; (B) Single-chamber; (C) dual-chamber tubular up-flow (D) tubular single-chamber; (E) stack (all schemes meant to be an illustration, so the electrodes (A – Anode; C – Cathode) may not be flat as presented here, the distance between them and their orientation may be different and the PEM can be removed).

468

Fuel-Cell Bioreactors

2.32.1.5

Operating Conditions

As explained, an FBC is a multiphase system involving simultaneous biological and electrochemical processes, mass, charge and energy transfer. As all of them are intimately coupled, there is a need to access the effect of operating conditions, such as the electrode material, membrane, pH, temperature, hydrodynamic stress and organic load on the cell performance toward its optimization. In fact, some experimental work had been done in order to achieve that; however, the cell performance is still below the desired. Furthermore, a prompter way to understand this interdependence and its effect on the cell power output is through both modeling and experimental studies.

2.32.1.5.1

Electrode Material

The layers where the electrochemical reactions occur in an FCB are known as electrodes, anode electrode and cathode electrode. The properties of these two electrodes are similar, but the cathode electrode allows gases to move through and is able to remove water, while the anode electrode allows liquid transport, microbial attachment and electron collection. Therefore, besides the similarities of the materials used in both electrodes, its overall characteristics and design are quite different. Regarding the anode electrode, the required properties, toward a proper electrons transfer and biofilm formation, are a good electrical conductivity, high corrosion resistance and mechanical strength, high surface area and porosity, chemical stability, biocompatibility and low cost. The materials that best meet these requirements and that have been used as anode electrode in FCB are the carbon-based and the metallic-based ones. However, these materials have different characteristics, such as different porosities, surface area and conductivity, which directly affect the microbes adhesion, biofilm development and electron production/transfer.35–37 For example, it is known that a higher surface area leads to a higher area for microorganisms to adhere on it and for electrons exchange, improving the FCB performance. These can be, easily, achieved using three-dimensional (3D) electrodes, such as brushes.38,39 Among the carbon-based materials, carbon cloth, carbon brush, carbon veil and carbon felt are the most used.35–39 Carbon cloth provides a high surface area, porosity, electrical conductivity, flexibility and mechanical strength; however, it has a high cost. Carbon brush is a 3D material that has a titanium core in which carbon fibers are twisted, having for that reason a high surface area, an optimal area to volume ratio and a high electrical conductivity due to the titanium core. However, this core is also responsible for its high cost. So research on this topic has been devoted toward cost reduction. Carbon veil has high electrical conductivity and porosity and low cost, but has a low mechanical strength. Similarly to carbon veil, carbon felt has a high porosity and electrical conductivity and low cost, but its mechanical strength can be higher than the one of the carbon veil, since it depends on the material thickness. Other carbon-based materials like carbon paper, carbon mesh, carbonized cardboard, graphite plate/sheet, granular activated carbon and granular graphite have also been used as anode electrode in FCB,35–37 but in a less extent. Carbon paper is a planar material, relatively porous, but has a high cost and low mechanical strength. Carbon mesh has a relatively low cost and a low electrical conductivity and mechanical strength, leading to a low durability under hydrodynamic conditions, such as higher shear stress and flow-rate.40 Carbonized cardboard is a 3D material and, has a very low cost, high electrical conductivity and porosity.41 Graphite plate/sheet is a material with a very simple structure and, has a high electrical conductivity and mechanical strength and relative low cost, but has a relatively low surface area and surface/volume ratio, leading to power outputs lower than the ones obtained with porous or more structured materials. Therefore, it is used as support for modified structures. Granular activated carbon and granular graphite have a higher biocompatibility, low cost and high porosity, but the last one has a lower surface area. Due to their characteristics, these two materials are mainly used as packing material rather than a single electrode. When packed into a bed, they offer an increased surface area compared to the other materials, leading to higher power outputs.42 Among the metal-based materials, stainless steel plates and meshes, silver, nickel, copper and gold sheets and titanium plates have been used and are commercially available.35,36,43 These materials have been used mainly due to its high mechanical strength, ductility, good electrical conductivity and for some of these (stainless steel) due to the very low cost. However, caution should be taken when using these materials to replace the carbon-based ones, since after a long-term of operation they suffer from corrosion. This will increase the contact resistances and will lead to the presence and accumulation of corrosion products on the anode compartment, poisoning it.43 Besides the electrode surface area, its surface characteristic, such as charge and hydrophilicity/hydrophobicity, is another important parameter that affects the microorganism attachment and its electrical connections with the electrode surface.35,36,44 Therefore, different surface modifications of the anode electrode, using different methods, have been successfully performed to facilitate these phenomena and improve the FCB power output. The most common ones include chemical treatment, such as acid treatment and ammonium treatment, electrochemical treatments, heat treatment and addition of highly conductive or electroactive coatings, such as gold and palladium nanoparticles and iron oxide.35,36 In cathode reaction, protons, electrons and, usually, oxygen are combined to produce water. Thus, the cathode electrode materials and its characteristics will impact the FCB performance. Most of the materials used as the anode electrode can also be employed as cathode electrode, but in this case, a catalyst is usually added to promote the electrochemical reaction. Carbon cloth is the most used material for the cathode electrode, but it is very expensive.35,45 To overcome such limitation, lower cost carbon-based materials have been used as alternatives. These include carbon mesh and metal-based materials, such as stainless steel mesh, copper mesh, nickel mesh and foam. Regarding the cathode catalyst, platinum-based (PGM-based) materials have been by far the most employed to promote the oxygen reduction reaction (ORR). However, they present a high cost and their durability is strongly

Fuel-Cell Bioreactors

469

compromised by the presence of some anions, especially sulfur anions that are naturally present in a wastewater.46 To avoid that, metal-free and platinum-free (PGM-free) materials have been proposed and studied as alternatives.35,45–50 These materials demonstrated enhanced performances and a better stability in hostile environments than the PGM-based catalysts. The metal-free or carbonaceous-based materials, such as graphene, activated carbon, carbon nanotubes and carbon nanofibers, present a high surface area and low pore sizes, enhancing the ORR. Among them activated carbon is the most used,35,45–47 and it was found that better performances were achieved with the addition of a small amount of carbon black.48 Moreover, durability tests with this material as cathode electrode exhibited a stable operation and a decrease of only 20% on the power output.49 PGM-free catalysts include a transition metal, M, such as manganese, iron, cobalt and nickel and carbon and nitrogen, denominated as M-N-C, and different studies with very promising and interesting results can be found using these materials as cathode catalyst in FCB. Among the different metals used, iron appear as to be the best one followed by cobalt, nickel and manganese.50

2.32.1.5.2

Membrane

One important component of an FCB is the membrane (or separator), since it affects both the cell performance, through its effect on the cell internal resistance and concentration losses, and its practical implementation, through its effect on the system overall costs. Moreover, the membrane prevents the contact of both electrodes (anode and cathode), avoiding short circuits and facilitating the protons transfer from one side to the other. An ideal membrane should have a high proton conductivity, low cost, and must be chemically, mechanically and electronically stable in the fuel cell environment. Typically, the membranes for FCB are made of a perfluorocarbon-sulfonic acid ionomer, which results from the combination of tetrafluorethylene with perfluorosulfonate monomers. The perfluorinated sulfonic acid membranes were developed by Dupont and are sold and known by their commercial name, Nafion.51 Nafion is a sulfonated tetrafluoroethylene copolymer made up of a fluorocarbon backbone attached to sulfonated groups ðSO3  Þ, which are responsible for creating hydrophilic regions, thus absorbing large amounts of water. The protons movement within the well-hydrated regions makes these materials proton conductive. However, Nafion membranes have high costs, limiting the FCB large-scale applications. As these membranes are not selective to protons, the fact that other cationic species cross them toward the opposite site leads to an accumulation of cations on the cathode side, increasing its pH and decreasing the reaction rate. To avoid these drawbacks, major efforts are made to find alternative membranes/materials that allow the optimum combination between a high performance and a low cost and that can be viable for the different FCB target markets. These include materials such as agar, salt bridge, ultrex, sulfonated polyimides, polyethylene and styrene and its derivatives, chitosan and ionic liquids.25,51–56 The salt bridge is the simplest material and consists on an ionic salt, such as KCl or NaOH, melted with agar. After solidification, the salt bridge is placed between the anode and cathode acting as a membrane.52 Other materials, such as j-cloth, nylon fibers, glass fibers, ceramics and biodegradable shopping bags, and unconventional ones, such as natural rubber or laboratory gloves, were also tested as an alternative membrane/separator and some of them show advantages over membrane fouling.57–60 Another approach commonly adopted to reduce the cell internal resistance and costs and the biofouling that affects the membrane is the development and optimization of FCB without using a membrane/separator, known as membrane-less FCB.24,51 However, these systems suffer from oxygen diffusion limitations from the cathode to the anode, creating a competition in the anode for electrons and from the crossover of substrate and microorganisms toward the cathode, which lead to the development of a biofilm on the cathode electrode, limiting the oxygen mass transfer rate and consequently decreasing the fuel cell performance.24,51 Therefore, these systems present a high deterioration rate, being unsuitable for long-term operation.

2.32.1.5.3

pH

The pH value of both FCB sides is crucial toward its proper operation and optimization. The pH of the anode side strongly affects microbial metabolic activity and consequently the electrons and protons production, while the one of the cathode side affects the ORR according to the Nernst equation (the ORR rate decreases with an increase of the pH). Microorganisms require a pH close to neutral for their optimal growth, and depending on the microorganism used and its growth conditions, variations in the anodic pH can cause modifications on its primary physiological parameters, such as its ability to produce and transfer electrons and biofilm development. Therefore, for an ideal operation, a neutral pH at the anode side and a lower pH at the cathode side should be ensured. The main problem of that emerge from the fact that the anode reaction produce protons, which can accumulate on this side due to a slow and incomplete proton diffusion through the membrane, causing a decrease of the anodic pH and consequently a decrease of the microorganisms’ activity and electrons production and transfer. In addition, the continuous proton consumption in the ORR, at the cathode, results in a pH increase on this side and, therefore, a decrease on the cell power output. These conditions lead to a higher pH gradient between the anode and cathode and a decrease on the cell efficiency and power generation.61–65 Therefore, research on this field has been focused on maintaining the ideal pH on both sides, and it was found that this can be more easily achieved in a traditional dual-chamber MFC, since it allows maintaining two different pH conditions, the optimal one on each side. This can be achieved with the addition of chemical buffers, such as phosphate (the most used one) to the system.66 However, they affect the FCB power output by interfering in the electrochemical reactions and microorganisms’ physiology. Additionally, its use is not practical for large-scale applications, since it leads to an extra cost. To avoid these drawbacks and provide a cheaper buffer, carbon dioxide and saline solutions such as sodium chloride have been used in FCB.65–68 Carbon dioxide is used on the cathode side since it is an acid that combines with the hydroxide ions producing bicarbonate and carbonate, which act as buffers.67 Additionally, carbon dioxide is available as a waste gas making it a low-cost buffer. The saline solutions besides being effective on controlling the pH also lead to an increase of the solution conductivity, a decrease of the internal resistance and an increase of the power density. However, caution should be taken when using these high saline solutions at the anode

470

Fuel-Cell Bioreactors

side as the anodic microbial communities can be affected by them.65,68 An innovative approach that has been used to control the pH, under buffer-less conditions, is the anolyte recirculation.69–71 Despite the very promising results, this approach demands an energy supply system, consuming energy.

2.32.1.5.4

Temperature

According to the Arrhenius equation, high temperatures lead to an increase of the electrochemical reactions rates and open-circuit voltage, a decrease of the FCB activation losses and also lead to an increase of the conductivity with a consequent decrease on the ohmic losses. However, high temperatures decrease the membrane stability and the oxygen partial pressure. Additionally, the microbial growth and activity are also affected by the operating temperature, and the optimal temperature depends on the microorganism used, on its ability to adjust to the environment changes and on the FCB operation time. An increase on temperature leads to an increase on the microorganisms’s metabolism rate and growth rate, but in the long-term operation, its proteins, nucleic acids or other temperature-sensitive parameters may suffer an irreversible damage, which will lead to a decline of the microorganisms’ cell function. Therefore, the microorganisms’ growth rates will be higher and its cell function will not be affected only if the optimum temperature range is ensured.65,72 The microorganisms used in FCB may be divide in mesophilic (32–42  C) and thermophilic (48–55  C), while at the transitory range of 40–45  C, both mesophilic and thermophilic microorganisms function under not-optimized conditions. Studies, regarding the effect of the operating temperature on the FCB, have shown an optimal range between 30  C and 40  C, the optimal range for most of the electrogenic species, leading to higher power outputs and microbial and biofilms bioelectrocatalytic activity.65,72 Additionally, it was also found that once the biofilm is fully developed/formed, the microorganisms were able to regulate their metabolism for different operating temperatures without significant losses on their activity and growth rate.

2.32.1.5.5

Hydrodynamic Stress

The hydrodynamic conditions of an FCB affect not only its power output/performance but also the microbial activity, adhesion and biofilm formation. Therefore, the flow-rates and the shear stress are key operating parameters that need to be addressed when developing an FCB system. Due to that, some efforts have been made to analyse their impact on the cell’s overall performance, such as its power output, microbial activity and biofilm characteristics.24,29,40,63,73–75 Usually, an increase on the shear stress rate leads to higher biomass concentrations and the development of thicker biofilms, which enhance the electrons production and collection and therefore leads to higher power outputs.73–75 The development of these thicker biofilms, under adverse conditions, may be due to an increase of the biofilm cohesion as a response to the high detachment forces induced by the high shear stress and/or to an increase of the biomass concentration, resulting from an increase of the mass transfer rate with the shear stress rate. Additionally, a higher shear stress also lead to biofilms with a lower microbial diversity, resulting in biofilms with a dominated species, usually the more active one, and with a more uniform and younger age.74 This more active species, with its higher metabolic activity, expelled the mature in age cells and formed a more active biofilm that improves the electron conduction via direct contact and enhances the electron transfer, since more cells are involved on that and more electron shuttles can be produced. However, it was also verified that when the shear stress is too high, cell detachment occurs and prevails over the biofilm cohesion, reducing the biofilm thickness, microbial growth and energy production.73,75 Toward the optimization of an FCB, the time required for the development of an active and efficient biofilm is crucial. Therefore, a compromise between the flow-rate or the FCB hydraulic retention time and the biofilm development should be achieved. Different studies at lab-scale showed an increase of the FCB power output with an increase of the flow rate, until a point where a further increase on the flow-rate conducts to a lower power outputs.24,29,40,63,73 This is due to a microbial wash out, leading to a decrease of its concentration, a decrease on the electrons production rate and consequently a decrease of the power production. Additionally, these extreme conditions also affect the biofilm stability and growth, due to the lower available time for microorganisms to oxidize the organic matter, resulting in thinner and less active biofilms attached to the electrode.24,29,40,63,73

2.32.1.5.6

Organic Load

As explained in Sectionto 2.32.1.2, the anode of an FCB works under the principle that microorganisms oxidize the organic matter, producing protons and electrons. So, the amount of organic matter presented on the anode will have a clear effect on the anode performance and consequently on the overall performance.24,63,76–78 This effect is evaluated through the organic loading rate (OLR), a parameter that relates the organic load conversion rate and the reactor volume. Studies regarding this issue showed an increase on the FCB power output and substrate oxidation with an increase of the OLR, due a higher availability of substrate to maintain the metabolic activity and an increase of the ionic strength of the anode solution. However, when the OLR is very high, the power output decreases, despite an increase on the substrate degradation rate.76–78 This additional substrate leads to a saturated environment, and the microorganisms responsible for producing energy will compete with microorganisms involved in other processes, such as methane production, hindering the performance of the energy production ones. This will lead to a higher organic matter removal not directly related with the energy generation and a decrease on the FCB power output.77 Therefore, the FCB should be operated with an organic load high enough to increase the power output, but not too high to avoid other parasitic process, such as methane production. This optimum value will depend on the microbial characteristics and specific requirements of the FCB.

Fuel-Cell Bioreactors 2.32.1.6

471

Models and Modeling Approaches

The FCB operating principle is simple to describe but understanding, predicting, and controlling the phenomena beneath its behavior is challenging. Experimental studies are essential to overcome this challenge, however, such approach can be expensive and some parameters are difficult or even impossible to obtain by experimental techniques. Therefore, the development of mathematical models describing the different phenomena occurring in an FCB (heat, charge and mass transfer, electrochemical reactions, microorganisms activity and biofilm characteristics) and their interactions allow acquiring a deep knowledge on them and predict their effect on the reactor power output toward its optimization. Additionally, this approach helps on reducing the amount of lab work, being for that reason a very economic alternative. Mathematical models can be categorized through their levels of complexity, as analytical, semi-empirical and mechanistic models.79–82 The analytical models are, usually, one-dimensional (1D) models developed using simpler and linear equations to predict the FCB power output and the effect of the different parameters on that, not requiring a very robust software to solve them.79,80 The semi-empirical models use simple empirical equations and relay on many assumptions and correlations, so the model parameters are estimated by fitting the model equations with the experimental date, being this the main drawback of these models.81 Its accuracy is very limited, not allowing to perform predictions for operation and design conditions different from the ones used on the model development. Mechanistic models are developed using different conservation equations, such as mass, momentum, energy, species and charge, based on the electrochemistry, biology and physics governing the phenomena taking place in the FCB. These models enable a more accurate prediction of the fuel cell performance and allow obtaining two-dimensional (2D) and three-dimensional (3D) information inside the fuel cell. However, solving these equations involve extensive calculations and thus, it is necessary to rely on computers and CFD (computational fluid dynamics) software packages.82,83 Mathematical models may also be classified according to their domain, such as full domain models,79–82 when it is intended to understand the system as an single unit (anode and cathode) or specific domain models, where the aim is to understand a specific process or component/side. Most of the FCB specific domain models were developed considering only the anode side, on an attempt to describe the biofilm development and the electron transfer mechanisms, since these are key issues to increase the energy production of these systems.84,85 These models are crucial to understand the phenomena at the biofilm/electrode interface with accuracy; however, modeling the full domain is not less important from a practical point of view. Regardless of the approach used, the model should be robust and accurate to predict the FCB performance in a wide range of operating conditions. However, its accuracy will depend on the equations and assumptions adopted and on its parameters, which should be selected based on the system characteristics and not by fitting with the experimental data. Beyond the heterogenicity of the different approaches, similarities on the FCB-specific equations and laws can be found. The Monod equation is used to describe the substrate consumption rate and microorganisms’ growth, the Nernst equation to describe the electrochemical reactions, the Tafel equation to describe anode and cathode reaction kinetics (usually combined with Monod equation), the Fick’s law to the diffusion and mass transfer phenomena, the Butler-Volmer to estimate the current density and Ohm’s law to calculate the ohmic losses.79–85 Additionally, the model development should be performed side by side with the experimental work to identify the main energy losses occurring in the system and give insights to these losses caused, allowing to design strategies to reduce and control them, providing a faster and more efficient improvement of the FCB performance.

2.32.1.7

Diagnostic Techniques

As already referred, FCB are complex systems, involving different phenomena, and an experimental investigation on the impact of each parameter, operating condition and/or phenomena occurring in the cell is time consuming and costly. As shown in the previous section, mathematical modeling is used as an additional tool to optimize these systems, however, all the models rely on a number of specific parameters that in most cases are not available in literature. Therefore, different diagnostic techniques are being widely applied to provide the data needed as input for the mathematical models and to track the influence of each parameter on the cell performance. These include electrochemical techniques, such as polarization measurements, EIS, cyclic voltammetry (CV) and chronoamperometry (CA), microscopic techniques such as scanning electron microscopy (SEM), confocal laser scanning microscopy (CLSM) and atomic force microscopy (AFM), molecular techniques such as denaturing gradient gel electrophoresis (DGGE), nextgeneration sequencing (NGS) and fluorescence in situ hybridization (FISH) and chemical characterization techniques such as chemical oxygen demand (COD) quantification. The electrochemical techniques use variables such as voltage, current and time to characterize the FCB performance under different operating conditions. The measurements can be performed in steady-state conditions, where the control parameter (voltage or current) is constant in time, such as polarization measurements, or in dynamic conditions, where the control parameter changes with time, such as EIS, CV and CA. Steady-state measurements are a simpler and helpful tool to obtain information about the steady-state properties of the system, while the dynamic measurements give information about the contribution of the different fuel cell components on its performance. As already referred, the polarization curves are plots where the cell voltage is represented against the current and can be used to describe the overall fuel cell performance and to identify the major losses affecting the system: activation, ohmic and mass transfer losses (Section 2.32.1.2). These curves can be obtained both in potentiostatic, set the voltage and the system will

472

Fuel-Cell Bioreactors

give the current, or galvanostatic, set the current and the system gives the voltage, modes, using an electrochemical workstation73 or applying external resistances.86 The polarization curves can be easily converted into power curves (power versus current) by multiplying the current by the cell voltage. This curve is very helpful on putting in evidence the maximum power output obtained with the system under study. CA is a technique where the voltage of the working electrode is stepped from an initial value to a final one and the resulting current is measured over the time. In FCB it is commonly used to estimate diffusion coefficients and obtain electrochemical polarization data, electrode active area and biofilm growth rate.87,88 In CV, the cell voltage is swept back and forth from a lower (LV) to a higher (HV) value and the current response to this change is recorded and plotted against the voltage, known as voltammogram. When a specific voltage is applied to a fuel cell system, a certain current starts flowing as contribution of two different phenomena: a constant capacitive charge current, which is a response to the linear variation of the voltage and a nonlinear response due to the electrochemical reactions occurring at the electrodes and characterized by current peaks on the voltammogram. Typically, an upward peak indicates an oxidation reaction and the downward picks a reduction reaction. Therefore, the characteristics of each peak (shape and size) give valuable information regarding the electrochemical reaction rates. Specifically, in FCB the presence of these peaks indicate the activity of the electron transfer and identify the DET or MET mechanisms. The peaks height provides information about the cytochromes or shuttles concentration. Therefore, the CV is applied to FCB to monitor the biofilm growth and its activity and determine the electron transfer mechanism.9 The EIS allows estimating the individual contributions of the different mechanisms (ohmic, activation and mass transport) and identification of the parameters that have a negative impact on the cell performance. The measurements are performed by imposing a sinusoidal perturbation to the system and measuring the impedance in a wide range of frequencies and can be conducted under both galvanostatic and potentiostatic modes. The EIS results are evaluated by modeling the impedance spectrum, Nyquist plot (imaginary part of the impedance plotted against the real one), with an equivalent electric circuit (comprised by different electrical elements and with different combinations between them) to obtain the meaningful properties of the system under study, such as the resistances that negatively affect the cell performance.87–89 Lower resistances lead to enhanced performances. The different losses are clearly identified on a Nyquist plot through the presence/absence of an arc at a specific frequency or region. The benefit of the EIS measurements can be increased by using a reference electrode, since it allows discriminating the different electrode losses (anode and cathode) from the overall cell response. The EIS technique have been widely used in FCB as a nondestructive technique, preventing biofilm damage and allowing to quantify the internal losses and estimate the kinetic parameters of the anodic and cathodic reactions, the biofilm behavior and the electron transfer mechanisms.86,90–93 Regarding the microscopic, molecular and chemical characterization techniques, SEM, CLSM and AFM are used to study the biofilm architecture and composition.17,92,93 AFM has the advantage of not requiring a previous biofilm preparation and allows the determination of the attachment force of the cell to the electrode. The biofilm microbial diversity is determined by the DGGE5,94 and NGS,17,95 and the FISH technique is, mainly, used to identify the electroactivity of the different microorganisms.95 Toward the evaluation of the ability of an FCB to treat an effluent, the COD content is measured and the corresponding COD removal rate is estimated.

2.32.1.8

Applications

The first target application for FCB was energy production. However, due its limitations, already addressed in this chapter, these systems have been used in a wide range of applications, where the major goal is not the energy production. As in most of them, some energy is produced; this energy can be used to offset the operational costs and achieve self-sufficient systems. The different FCB applications under study and development are described with some details in the following sub-sections.

2.32.1.8.1

Wastewater Treatment

The wastewater treatment is the most studied and developed application of FCB and, besides allowing treating a wastewater, produces energy that can be used on the process, toward a reduction of the system energy needs and offsetting some of the operational costs. Additionally, FCB produces approximately 50%–90% less solids/sludge than the conventional treatments, decreasing this treatment need, which also conduct to additional costs. However, the efficiency of these systems relies on different parameters, regarding its design, operating conditions and wastewater characteristics, which also have a clear impact on the microbial activity responsible to perform the treatment and the energy production. The parameters used to evaluate the system efficiency/performance are the COD removal rate, maximum power density and Coulombic efficiency.92,93,96 As different wastewaters have different characteristics that affect the treatment efficiency and energy production, domestic, municipal, agro-food industrial (brewery, dairy and fruit processing) and pharmaceutical and textile industrial wastewaters have been studied and applied, mostly in a lab-scale, to FCB with very promising and interesting results.97

2.32.1.8.2

Hydrogen Production

The FCB can produce hydrogen by modifying the operating principle described in Section 2.32.1.2, where microorganisms act as anode catalysts to oxidize the substrates presented on this side, producing electrons and protons that flow to the cathode to

Fuel-Cell Bioreactors

473

form water, eliminating any possibility of hydrogen production. However, if an energy input is applied to the system, it is possible to overcome the thermodynamic barrier and produce hydrogen at the cathode with the combination of electrons and protons. Despite these energy requirement, this method is very efficient since more than 90% of the protons and electrons generated by the microorganisms at the anode are turned into hydrogen. Conventional production of hydrogen requires 10 times the amount of energy than an FCB, making this technology an efficient and environmentally friendly way to generate hydrogen. The parameters used to evaluate the system efficiency/performance are the hydrogen production rate and the amount of energy recovered. The biggest challenge hiding this application is to find a low cost membrane, since the membrane-less system is not efficient on recovering the hydrogen produced.20,98

2.32.1.8.3

Bioremediation

The critical pollution problems arising from the complex organic wastewaters produced by our society are due to the fact that these wastewaters contain metals such as iron, copper and chromium and pollutants such as nitrates, sulphide and sulphates. Heavy metals especially chromium pose a serious risk to human, animals, and the environment. Therefore, it is necessary to treat these wastewaters prior to their discharge. Furthermore, there is an economic interest on its recovery since they are non-renewable and expensive raw materials. Several treatment techniques such as chemical precipitation, coagulation–flocculation, ion exchange, membrane filtration and biosorption have been applied to treat these wastewaters. Despite the effectiveness of these conventional techniques, they have high energy requirements, excessive chemical consumption and in some situations generation of a large quantity of toxic waste sludge. To avoid these drawbacks, FCB are being used in an effective way to remove and recover these compounds from wastewaters.99,100

2.32.1.8.4

Biosensors

A biosensor is an analytical device that converts a biological response into an electrical signal. A biosensor can be used to determine the concentration of substances and other parameters of biological interest even where they do not use directly a biological system. FCB can be used as biosensors for monitoring the biological oxygen demand (BOD) in water, identification of specific chemicals, such as volatile fatty acids, which have an important role on evaluating the performance of anaerobic and aerobic processes, biological contamination, evaluating the water toxicity and pollutants concentration.101–103 One of the major advantages of using an FCB in remote sensing rather than a traditional battery is that the microorganisms reproduce allowing a significantly longer operation time. Furthermore, the biosensor can be left alone in a remote area for many years without maintenance.

2.32.1.8.5

Desalination

Desalination of seawater and brackish water for use as drinking water has always presented significant problems due to the energy requirements needed to remove the dissolved salts from the water. By using an adapted FCB, this process could proceed without an external energy input. Toward that, a third chamber is placed between the two standard electrodes of an FCB (anode and cathode), which are separated by ion-selective membranes and linked by an external circuit. Forward osmosis and reverse electrolysis are the major mechanisms used in these systems to desalinize water. Filling the additional chamber with seawater, the cell positive and negative electrodes attract the positive and negative salt ions presented on that, removing them from the water and producing energy. As most of the convectional desalination technologies require a high-energy input and capital cost, and are not very efficient, FCB are seen as a very promising alternative technology, since are able to perform desalination with an energy self-sufficient system.104,105

2.32.1.8.6

Other Applications

FCB can also be used to capture the CO2 produced on the anode reaction by microorganisms present on the cathode side, such as microalgae and bacteria, and releasing O2 to the environment. Therefore, these systems became a zero-emission system.106,107 Another application of FCB is to produce compounds of industrial interest such as acetate, using the CO2 produced on the anode side and specific microorganisms that are introduced on the cathode compartment.19,20,108

2.32.1.9

Economic Evaluation and Life Cycle Assessment

The economic evaluation and the assessment of the environmental benefits and impacts should be well addressed in order to indorse the FCB as an alternative, sustainable and viable technology. Toward the implementation of FCB in real applications, their costs, which until now are prohibitive, need to be carefully addressed and reduced to acceptable levels. Therefore, for these systems to be competitive with their main opponents, their benefits should be maximized and should be higher than their total costs (manufacturing, implementation and operational costs), which should be minimized. Most of the studies regarding FCB economic prospects are based on their application on wastewater treatment, since wastewaters have a higher potential as energy source. However, most of the wastewater treatment facilities, instead of trying to recover this energy, spend energy using the conventional technologies to perform the treatment, such as anaerobic digestion and aerobic treatment.109 The anaerobic digestion can recover energy due to the ability of producing biogas, methane (CH4), but requires meso- and thermophilic temperatures to operate. The aerobic treatment requires huge amounts of energy for aeration, accounting for almost 60% of the

474

Fuel-Cell Bioreactors

total energy required in a wastewater treatment plant and half of its operational costs. Moreover, this treatment is known to produce large amounts of sludge, which require further treatment leading to an additional cost that can reach up to 500V/ton dry matter.110 Therefore, eliminating or reducing these costs, by replacing these technologies for more efficient ones, can lead to an appreciable reduction of the input of energy and the treatment costs. FCB emerged as an alternative and very interesting technology, offering many advantages over the conventional technologies:

• • • • • • •

direct conversion of organic matter into electricity, allowing higher conversion rates/efficiencies; operate at ambient or low temperatures, reducing the heat costs; do not require gas treatment since the only gas produced is carbon dioxide, which has no useful energy content; do not require an input of energy for aeration since the cathode can be passively aerated; can be used in remote areas, areas with a lack of electrical infrastructures/grid; can use a diversity of fuels, allowing to satisfy the energy requirements for each application; can reduce the solids/sludge production, reducing the sludge handling costs.

However, until now the FCB still have some performance limitations and costs above the desirable, being a step behind the conventional treatments. As already referred, the overall FCB costs include material, manufacturing, operating and maintenance costs. Material costs are responsible for the major fraction on the overall cost, while the manufacturing, operating and maintenance costs only have a slight effect on it. Regarding the material costs, it is known and accepted that the major contributions to that costs come from the Nafion membranes and the cathode electrodes, when platinum is used as cathode catalyst to promote the cathode electrochemical reaction, ORR, followed by current collector and other essential parts.111 Therefore, FCB can only became economically interesting if these costs are reduced, by changing the materials commonly used for less expensive ones (Sections 2.32.1.5.1 and 2.32.1.5.2).112 Current collectors are usually made from metallic materials, such as stainless steel. To avoid corrosion, decrease the contact resistance between the current collector and the membrane and increase the electrical conductivity, these plates are usually coated with precious metals such as platinum, gold and titanium. This coating is responsible for an increase on the plates costs. As this coating is needed, an optimization of the coating thickness is essential in order to maintain a proper balance between corrosion resistance and conductivity and coating costs. Life Cycle Assessment (LCA) is a tool for the assessment of potential environmental impacts of products or services along its entire life cycle. Analyzing the whole life cycle of an FCB system allows to assess which part of the process contributes to the most relevant environmental impacts, providing possible solutions for their improvement. The LCA, usually, follows four different steps, according to ISO 14040:

• • • •

definition of the goal and scope: reasons to perform the study, the proposed audience, geographic and temporal considerations, the system function and boundaries, data categories, competitive systems, impact assessment and interpretation methods and plans for critical review; inventory analysis: quantifies the material and energy use and waste produced by each process, so they are linked by economic, environmental, consume and production flows (flows in and out, such as the consumption of energy or carbon dioxide emissions); impact assessment: estimates the contribution of environmental flows to environmental benefits (such as habitat protection) and impacts (such as global warming); interpretation: identifies sensitive parameters and quantifies the uncertainties of the results, allowing to give recommendations or draw conclusions.

From the environmental point of view, FCB present benefits such as COD removal, energy production from wastewater streams, production of chemicals, such as hydrogen peroxide and hydrogen, nutrients and metals recovery and the use of CO2 to produce chemical and biofuels.111,113 The environmental impacts of these systems are related with the production of the carbon fiber and carbon brush electrodes, which represent 80% of the total electrode costs and have unwanted emissions on the production process, membranes, the Nafion production process is very expensive, stainless steel components and cathode electrodes, when PGM-based catalysts are used. PGM are produced as a by-product of Ni mining, which have considerable environmental effects due to the SO2 emissions during the production process. Despite the importance of performing an LCA to any new technology toward its commercialization and massive use, only few studies conducted LCA for FCB. A more detailed analysis with consistent assumptions is required to fully assess FCB benefits and impacts over the conventional technologies.

2.32.1.10 Challenges and Perspectives Currently, FCB can achieve power densities considerably higher than those achieved with the first devices of this type appeared, showing that the major efforts performed toward the optimization of these systems, through both experimental and modeling studies, are being profitable. However, despite the great progress already achieved, the FCB low power outputs and higher costs remain problematic. It is commonly accepted that to achieve the desirable levels of energy density and costs, an FCB system must overcome the following key challenges:

Fuel-Cell Bioreactors

• • • • •

475

Electrodes: there is a need for new and low cost materials that favor the biofilm adhesion on the anode electrode surface and increase conductivity promoting a faster electron transfer to the electrode and cathodes with higher electrochemical activities and lower costs; Membranes: it is important to increase the proton transfer from the anode to the cathode, while minimizing the risks of unwanted electron acceptor transfer from the cathode to the anode, using new and low cost materials; Microbiology: it is needed to find a microorganism or a consortium that grows on low cost substrates and is able to produce and transfer electrons directly and at a faster rate to the anode electrode; Applicability: it is important to verify if it is economically acceptable to develop FCB that are able to compete with the conventional technologies on efficiency, electricity production and costs; Operating conditions: the accurate effect of operating conditions on the FCB performance and biofilm formation, structure and composition is still unknown.

As mentioned previously, toward the use of FCB in real applications, cost effectiveness is essential. In general, high conversion rates, evaluated based on the power output, are a major condition for maximizing the benefits of these systems and to offset the higher costs. However, this rate is limited by the energy losses occurring in these systems: activation, ohmic and mass transfer losses. Therefore, it is extremely important to reduce these losses, through the selection of new microbes (that could transfer electrons directly to the electrode) and substrates and performing modifications in the FCB design and configuration. The ohmic losses can be reduced by shortening the distance between the two electrodes and by increasing the ionic conductivity of the electrolytes and membrane/ separator. Using 3D electrodes on the anode side and new electroactive catalyst to promote the cathode reaction will reduce the activation losses. The mass transport limitations can be minimized, increasing the substrate concentration and operating the system under different hydrodynamic conditions. In addition to these losses, on the anode side methanogens compete with the microorganisms that produce energy, to convert the organic matter in methane, reducing the electricity generation. Therefore, toward an increase of the FCB power output, it is mandatory to operate the cell in unfavorable conditions to methanogens in order to avoid its competition with the target microorganisms, the energy producers ones. The lifetime or durability of an FCB is evaluated by its ability to produce a constant power output over time. Although the performance loss during the fuel cell lifetime is unavoidable, its rate can be minimized through an understanding of the degradation and failure mechanisms. Therefore, performance degradation is another major challenge that hinders the application of FCB in real systems. It is then mandatory and recommended to perform durability tests before commercialization of any fuel cell, and the major goal of performing these tests is to extend the lifetime through the identification of the primary mechanisms leading to the performance losses and to find materials and design solutions to avoid or minimize them. The performance degradation can be permanent or temporary, which is reversible, therefore can be recovered. The permanent degradation is mainly due to the presence of foreign ions and contaminants, changes on the different materials microstructure and transport properties and a reduction on the catalyst activity. The temporary degradation is mainly due to pH and temperature changes, lower microbial activity and substrate concentration, being easily solved after stopping and restarting the system. Despite their major importance, only few studies were performed having in consideration the FCB lifetime.49,114,115

2.32.2

Concluding Remarks

FCB are a very promising technology, since it can be used in a wide range of applications, such as biosensors, desalination, hydrogen production, electricity generation, wastewater treatment among others. This is due to the ability to work under different electron donors and substrates, with different concentrations and at low and moderate temperatures, making the FCB a unique technology. However, as FCB performances at lab and pilot scale are still lower than the desirable, in the last decade, the amount of work on these systems, both on the microbiological issues and on the engineering ones, has been growing, which allowed to increase the power outputs by an order of magnitude. Nevertheless, these values have not reached the levels needed for real applications. In addition, the FCB lifetime and durability need to emerge as a future development area in these systems research. The performance of an FCB can be enhanced using more active microorganisms or consortia, more complex substrates and by optimization of the electron transfer mechanism and operating conditions. Such optimization includes the anode and cathode electrode and membrane materials, pH, temperature, hydrodynamic conditions (feed rate and shear stress) and organic load. Additionally, the use of mathematical models and different diagnostic techniques can help on achieving faster results and optimized performances, allowing the detailed understanding of the different phenomena that affect the FCB performance, the microbial activity and biofilm development. Moreover, the models help to predict the main losses that negatively affect these systems, pointing out possible measures to avoid or reduce them. The reduction of the FCB costs will be possible through the development of new low cost and with an enhanced performance materials and changing the current ones by them. However, as FCB are still in a younger stage than the well-established conventional technologies, estimating their cost is very difficult and challenging, since some of them are associated with the manufacturing technologies that are still not fully automated. In a more realistic point of view, even with further developments and performance and costs improvements, the FCB maybe not be able to compete directly with the conventional technologies in a large scale, but will be more suitable for a low/medium scale and decentralized plants.

476

Fuel-Cell Bioreactors

Acknowledgments The authors acknowledge the PhD scholarship (PD/BD/114314/2016) supported by the Portuguese “Fundação para a Ciência e Tecnologia” (FCT), by “Fundos Nacionais do Ministério da Ciência, Tecnologia e Ensino Superior” (MCTES), inserted in the program “Programa Operacional Capital Humano” (POCH), cofinanced by Portugal 2020 and European Social Fund (ESF). The authors also acknowledge the national funding of FCT/MEC (PIDDAC). POCI (FEDER) also supported this work via CEFT and LEPABE.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.

Flemming, H.-C.; Wingender, J. The Biofilm Matrix. Nat. Rev. Microbiol. 2010, 8, 623–633. Flemming, H.-C.; Wingender, J.; Szewzyk, U.; Steinberg, P.; Rice, S. A.; Kjelleberg, S. Biofilms: an Emergent Form of Bacterial Life. Nat. Rev. Microbiol. 2016, 14, 563–575. Mao, L.; Verwoerd, W. S. Selection of Organisms for Systems Biology Study of Microbial Electricity Generation: a Review. Int. J. Energy Environ. Eng. 2013, 4 (17), 1–18. Philips, J.; Verbeeck, K.; Rabaey, K.; Arends, J. B. A. Electron Transfer Mechanisms in Biofilms. In Microb. Electrochem. Fuel Cells, Elsevier, 2016; pp 67–113. Saratale, G. D.; Saratale, R. G.; Shahid, M. K.; Zhen, G.; Kumar, G.; Shin, H.; Choi, Y.; Kim, S. A Comprehensive Overview on Electro-Active Biofilms, Role of Exo-Electrogens and Their Microbial Niches in Microbial Fuel Cells (MFCs). Chemosphere 2017, 178, 534–547. Sydow, A.; Krieg, T.; Mayer, F.; Schrader, J.; Holtmann, D. Electroactive Bacteria-Molecular Mechanisms and Genetic Tools. Appl. Microbiol. Biotechnol. 2014, 98, 8481–8495. Chen, M.; Zhou, X.; Liu, X.; Zeng, R. J.; Zhang, F.; Ye, J.; Zhou, S. Facilitated Extracellular Electron Transfer of Geobacter sulfurreducens Biofilm with in situ Formed Gold Nanoparticles. Biosens. Bioelectron. 2018, 108, 20–26. Nevin, K. P.; Richter, H.; Covalla, S. F.; Johnson, J. P.; Woodard, T. L.; Orloff, A. L.; Jia, H.; Zhang, M.; Lovley, D. R. Power Output and Columbic Efficiencies from Biofilms of Geobacter sulfurreducens Comparable to Mixed Community Microbial Fuel Cells. Environ. Microbiol. 2008, 10, 2505–2514. Choi, S.; Kim, B.; Chang, I. S. Tracking of Shewanella Oneidensis MR-1 Biofilm Formation of a Microbial Electrochemical System via Differential Pulse Voltammetry. Bioresour. Technol. 2018, 254, 357–361. Pirbadian, S.; Barchinger, S. E.; Leung, K. M.; Byun, H. S.; Jangir, Y.; Bouhenni, R. A.; Reed, S. B.; Romine, M. F.; Saffarini, D. A.; Shi, L.; Gorby, Y. A.; Golbeck, J. H.; El-Naggar, M. Y. Shewanella Oneidensis MR-1 Nanowires Are Outer Membrane and Periplasmic Extensions of the Extracellular Electron Transport Components. Proc. Natl. Acad. Sci. USA 2014, 111, 12883–12888. Hubenova, Y.; Mitov, M. Extracellular Electron Transfer in Yeast-Based Biofuel Cells: A Review. Bioelectrochemistry 2015, 106, 177–185. You, L. X.; Liu, L. D.; Xiao, Y.; Dai, Y. F.; Chen, B. L.; Jiang, Y. X.; Zhao, F. Flavins Mediate Extracellular Electron Transfer in Gram-Positive Bacillus Megaterium Strain LLD-1. Bioelectrochemistry. 2018, 119, 196–202. Gude, V. G. Wastewater Treatment in Microbial Fuel Cells – an Overview. J. Clean. Prod. 2016, 122, 287–307. Hodgson, D. M.; Smith, A.; Dahale, S.; Stratford, J. P.; Li, J. V.; Grüning, A.; Bushell, M. E.; Marchesi, J. R.; Avignone Rossa, C. Segregation of the Anodic Microbial Communities in a Microbial Fuel Cell Cascade. Front. Microbiol. 2016, 7, 1–11. Park, Y.; Cho, H.; Yu, J.; Min, B.; Kim, H. S.; Kim, B. G.; Lee, T. Response of Microbial Community Structure to Pre-acclimation Strategies in Microbial Fuel Cells for Domestic Wastewater Treatment. Bioresour. Technol. 2017, 233, 176–183. Pisciotta, J. M.; Zaybak, Z.; Call, D. F.; Nam, J.-Y.; Logan, B. E. Enrichment of Microbial Electrolysis Cell Biocathodes from Sediment Microbial Fuel Cell Bioanodes. Appl. Environ. Microbiol. 2012, 78, 5212–5219. Chabert, N.; Amin Ali, O.; Achouak, W. All Ecosystems Potentially Host Electrogenic Bacteria. Bioelectrochemistry 2015, 106, 88–96. Kumar, R.; Singh, L.; Zularisam, A. W.; Exoelectrogens. Recent Advances in Molecular Drivers Involved in Extracellular Electron Transfer and Strategies Used to Improve it for Microbial Fuel Cell Applications. Renew. Sustain. Energy Rev. 2016, 56, 1322–1336. Puig, S.; Ganigué, R.; Batlle-Vilanova, P.; Balaguer, M. D.; Bañeras, L.; Colprim, J. Tracking Bio-Hydrogen-Mediated Production of Commodity Chemicals from Carbon Dioxide and Renewable Electricity. Bioresour. Technol. 2017, 228, 201–209. Jourdin, L.; Lu, Y.; Flexer, V.; Keller, J.; Freguia, S. Biologically Induced Hydrogen Production Drives High Rate/High Efficiency Microbial Electrosynthesis of Acetate from Carbon Dioxide. ChemElectroChem 2016, 3, 581–591. Ulusoy, I.; Dimoglo, A. Electricity Generation in Microbial Fuel Cell Systems with Thiobacillus Ferrooxidans as the Cathode Microorganism. Int. J. Hydrogen Energy 2018, 43, 117–1178. Kato, S.; Hashimoto, K.; Watanabe, K. Microbial Interspecies Electron Transfer via Electric Currents through Conductive Minerals. Proc. Natl. Acad. Sci. USA 2012, 109, 10042–10046. Parkash, A. Microbial Fuel Cells: A Source of Bioenergy. J. Microb. Biochem. Technol. 2016, 8, 247–255. Vicari, F.; Albamonte, M.; Galia, A.; Scialdone, O. Effect of Mode of Operation, Substrate and Final Electron Acceptor on Single-Chamber Membraneless Microbial Fuel Cell Operating with a Mixed Community. J. Electroanal. Chem. 2018, 814, 104–110. Hernández-Flores, G.; Poggi-Varaldo, H. M.; Solorza-Feria, O.; Ponce-Noyola, M. T.; Romero-Castañón, T.; Rinderknecht-Seijas, N.; Galíndez-Mayer, J. Characteristics of a Single Chamber Microbial Fuel Cell Equipped with a Low Cost Membrane. Int. J. Hydrogen Energy 2015, 40, 17380–17387. He, Z.; Minteer, S. D.; Angenent, L. T. Electricity Generation from Artificial Wastewater Using an Upflow Microbial Fuel Cell Electricity Generation from Artificial Wastewater Using an Upflow Microbial Fuel Cell. Environ. Sci. Technol. 2005, 39, 5262–5267. He, Z.; Wagner, N.; Minteer, S. D.; Angenent, L. T. An Upflow Microbial Fuel Cell with an Interior Cathode: Assessment of the Internal resistance by Impedance Spectroscopy. Environ. Sci. Technol. 2006, 40, 5212–5217. Lay, C.; Kokko, M. E.; Puhakka, J. A. Power Generation in Fed-Batch and Continuous Up-Flow Microbial Fuel Cell from Synthetic Wastewater. Energy 2015, 91, 235–241. Li, X.; Zhu, N.; Wang, Y.; Li, P.; Wu, P.; Wu, J. Animal Carcass Wastewater Treatment and Bioelectricity Generation in Up-Flow Tubular Microbial Fuel Cells: Effects of HRT and Non-precious Metallic Catalyst. Bioresour. Technol. 2013, 128, 454–460. Oliveira, V. B.; Marcon, L. R. C.; Vilas Boas, J.; Daniel, A.; Pinto, A. M. F. R. Microbial Fuel Cells for Wastewater Treatment. In Sew. Treat. Plants Econ. Eval. Innov. Technol. Energy Effic.; Stamatelatou, K., Tsagarakis, K. P., Eds. IWA Publishing, 2015; pp 213–231. Velvizhi, G.; Venkata Mohan, S. Multi-electrode Bioelectrochemical System for the Treatment of High Total Dissolved Solids Bearing Chemical Based Wastewater. Bioresour. Technol. 2017, 242, 77–86. Vilajeliu Pons, A.; Puig, S.; Salcedo-Dávila, I.; Balaguer, M. D.; Colprim, J. Long-term Assessment of the Six-Stacked Scaled-Up MFCs Treating Swine Manure with Different Electrode Materials. Environ. Sci. Water Res. Technol. 2017, 3, 947–959. Asensio, Y.; Mansilla, E.; Fernandez-Marchante, C. M.; Lobato, J.; Cañizares, P.; Rodrigo, M. A. Towards the Scale-Up of Bioelectrogenic Technology: Stacking Microbial Fuel Cells to Produce Larger Amounts of Electricity. J. Appl. Electrochem. 2017, 47, 1115–1125. Ieropoulos, I.; Greenman, J.; Melhuish, C. Microbial Fuel Cells Based on Carbon Veil Electrodes: Stack Configuration and Scalability. Int. J. Energy Res. 2008, 32, 1228–1240. Wei, J.; Liang, P.; Huang, X. Recent Progress in Electrodes for Microbial Fuel Cells. Bioresour. Technol. 2011, 102, 9335–9344. Sonawane, J. M.; Yadav, A.; Ghosha, P. C.; Adeloju, S. B. Recent Advances in the Development and Utilization of Modern Anode Materials for High Performance Microbial Fuel Cells. Biosens. Bioelectron. 2017, 90, 558–576.

Fuel-Cell Bioreactors

477

37. Liu, Y.; Harnisch, F.; Fricke, K.; Schröder, U.; Climent, V.; Feliu, J. M. The Study of Electrochemically Active Microbial Biofilms on Different Carbon-Based Anode Materials in Microbial Fuel Cells. Biosens. Bioelectron. 2010, 25, 2167–2171. 38. Lanas, V. A.; Ahn, Y.; Logan, B. E. Effects of Carbon Brush Anode Size and Loading on Microbial Fuel Cell Performance in Batch and Continuous Mode. J. Power Sources 2014, 247, 228–234. 39. Liao, Q.; Zhang, J.; Li, J.; Ye, D.; Zhu, X.; Zhang, B. Increased Performance of a Tubular Microbial Fuel Cell with a Rotating Carbon-Brush Anode. Biosens. Bioelectron. 2015, 63, 558–561. 40. Wu, S.; He, W.; Yang, W.; Ye, Y.; Huang, X.; Logan, B. E. Combined Carbon Mesh and Small Graphite Fiber Brush Anodes to Enhance and Stabilize Power Generation in Microbial Fuel Cells Treating Domestic Wastewater. J. Power Sources 2017, 356, 348–355. 41. Kretzschmar, J.; Riedl, S.; Brown, R. K.; Schröder, U.; Harnisch, F. eLatrine: Lessons Learned from the Development of a Low-Tech MFC Based on Cardboard Electrodes for the Treatment of Human Feces. J. Electrochem. Soc. 2017, 164, H3065–H3072. 42. Yasri, N. G.; Nakhla, G. The Performance of 3-D Graphite Doped Anodes in Microbial Electrolysis Cells. J. Power Sources 2017, 342, 579–588. 43. Baudler, A.; Schmidt, I.; Langner, M.; Greiner, A.; Schröder, U. Does it Have to Be Carbon? Metal Anodes in Microbial Fuel Cells and Related Bioelectrochemical Systems. Energy Environ. Sci. 2015, 8, 2048–2055. 44. Santoro, C.; Babanova, S.; Artyushkova, K.; Cornejo, J. A.; Ista, L. K.; Bretschger, O.; Marsili, E.; Atanassov, P.; Schuler, A. J. Influence of Anode Surface Chemistry on Microbial Fuel Cell Operation. Bioelectrochemistry 2015, 106, 141–149. 45. Wang, Z.; Mahadevan, G. D.; Wu, Y.; Zhao, F. Progress of Air-Breathing Cathode in Microbial Fuel Cells. J. Power Sources 2017, 356, 245–255. 46. Wang, Z.; Cao, C.; Zheng, Y.; Chen, S.; Zhao, F. Abiotic Oxygen Reduction Reaction Catalysts Used in Microbial Fuel Cells. ChemElectroChem 2014, 1, 1813–1821. 47. Dai, L.; Xue, Y.; Qu, L.; Choi, H.-J.; Baek, J.-B. Metal-Free Catalysts for Oxygen Reduction Reaction. Chem. Rev. 2015, 115, 4823–4892. 48. Zhang, X.; Xia, X.; Ivanov, I.; Huang, X.; Logan, B. E. Enhanced Activated Carbon Cathode Performance for Microbial Fuel Cell by Blending Carbon Black. Environ. Sci. Technol. 2014, 48, 2075–2081. 49. Zhang, X.; Pant, D.; Zhang, F.; Liu, J.; He, W.; Logan, B. E. Long-Term Performance of Chemically and Physically Modified Activated Carbons in Air Cathodes of Microbial Fuel Cells. ChemElectroChem 2014, 1, 1859–1866. 50. Kodali, M.; Santoro, C.; Serov, A.; Kabir, S.; Artyushkova, K.; Matanovic, I.; Atanassov, P. Air Breathing Cathodes for Microbial Fuel Cell Using Mn-, Fe-, Co- and Ni-Containing Platinum Group Metal-free Catalysts. Electrochim. Acta 2017, 231, 115–124. 51. Leong, J. X.; Daud, W. R. W.; Ghasemi, M.; Liew, K. B.; Ismail, M. Ion Exchange Membranes as Separators in Microbial Fuel Cells for Bioenergy Conversion: A Comprehensive Review. Renew. Sustain. Energy Rev. 2013, 28, 575–587. 52. Nair, R.; Renganathan, K.; Barathi, S.; Venkatraman, K. Performance of Salt Bridge Microbial Fuel Cell at Various Agarose Concentrations Using Hostel Sewage Waste as Substrate. Int. J. Advance Res. Technol. 2013, 2, 326–330. 53. Hernández-Fernández, F. J.; Pérez de los Ríos, A.; Mateo-Ramírez, F.; Juarez, M. D.; Lozano-Blanco, L. J.; Godínez, C. New Application of Polymer Inclusion Membrane Based on Ionic Liquids as Proton Exchange Membrane in Microbial Fuel Cell. Separ. Purif. Technol. 2016, 160, 51–58. 54. Yousefi, V.; Mohebbi-Kalhori, D.; Samimi, A. Ceramic-based Microbial Fuel Cells (MFCs): A Review. Int. J. Hydrogen Energy 2017, 42, 1672–1690. 55. Elangovan, M.; Dharmalingam, S. Application of Polysulphone Based Anion Exchange Membrane Electrolyte for Improved Electricity Generation in Microbial Fuel Cell. Mater. Chem. Phys. 2017, 199, 528–536. 56. Holder, S. L.; Lee, C.-H.; Popuri, S. R. Simultaneous Wastewater Treatment and Bioelectricity Production in Microbial Fuel Cells Using Cross-Linked Chitosan-Graphene Oxide Mixed-Matrix Membranes. Environ. Sci. Pollut. Res. 2017, 24, 13782–13796. 57. Zhang, X.; Cheng, S.; Wang, X.; Huang, X.; Logan, B. E. Separator Characteristics for Increasing Performance of Microbial Fuel Cells. Environ. Sci. Technol. 2009, 43, 8456–8461. 58. Zhang, X.; Cheng, S.; Wang, X.; Huang, X.; Logan, B. E. The Use of Nylon and Glass Fiber Filter Separators with Different Pore Sizes in Air-Cathode Single-Chamber Microbial Fuel Cells, Energy. Environ. Sci. 2010, 3, 659–664. 59. Winfield, J.; Ieropoulos, I.; Rossiter, J.; Greenman, J.; Patton, D. Biodegradation and Proton Exchange Using Natural Rubber in Microbial Fuel Cells. Biodegradation 2013, 24, 733–739. 60. Winfield, J.; Chambers, L. D.; Rossiter, J.; Greenman, J.; Ieropoulos, I. Towards Disposable Microbial Fuel Cells: Natural Rubber Glove Membranes. Int. J. Hydrogen Energy 2014, 39, 21803–21810. 61. Oliot, M.; Galier, S.; Balmann, H. R.; Bergel, A. Ion Transport in Microbial Fuel Cells: Key Roles, Theory and Critical Review. Appl. Energy 2016, 183, 1682–1704. 62. Margaria, V.; Tommasi, T.; Pentassuglia, S.; Agostino, V.; Sacco, A.; Armato, C.; Chiodoni, A.; Schiliro, T.; Quaglio, M. Effects of pH Variations on Anodic Marine Consortia in a Dual Chamber Microbial Fuel Cell. Int. J. Hydrogen Energy 2017, 42, 1820–1829. 63. Behera, M.; Ghangrekar, M. M. Optimization of Operating Conditions for Maximizing Power Generation and Organic Matter Removal in Microbial Fuel Cell. J. Environ. Eng. 2017, 143, 1–10. 64. Sasaki, D.; Sasaki, K.; Tsuge, Y.; Kondo, A. Comparative Metabolic State of Microflora on the Surface of the Anode Electrode in a Microbial Fuel Cell Operated at Different pH Conditions. Amb. Exp. 2016, 6 (125), 1–9. 65. Tremouli, A.; Martinos, M.; Lyberatos, G. The Effects of Salinity, pH and Temperature on the Performance of a Microbial Fuel Cell. Waste Biomass. Valor 2017, 8, 2037–2043. 66. Firdous, S.; Jin, W.; Shahid, N.; Bhatti, Z. A.; Iqbal, A.; Abbasi, U.; Mahmood, Q.; Ali, A. The Performance of Microbial Fuel Cells Treating Vegetable Oil Industrial Wastewater. Environ. Technol. Innov. 2018, 10, 143–151. 67. Fornero, J. J.; Rosenbaum, M.; Cotta, M. A.; Angenet, L. T. Carbon Dioxide Addition to Microbial Fuel Cell Cathodes Maintains Sustainable Catholyte pH and Improves Anolyte pH, Alkalinity, and Conductivity. Environ. Sci. Technol. 2010, 44, 2728–2734. 68. Merino-Jimenez, I.; Celorrio, V.; Fermin, D. J.; Greenman, J.; Ieropoulos, I. Enhanced MFC Power Production and Struvite Recovery by the Addition of Sea Salts to Urine. Water Res. 2017, 109, 46–53. 69. Zhang, L.; Zhu, X.; Li, J.; Kashima, H.; Liao, Q.; Regan, J. M. Step-feed Strategy Enhances Performance of Unbuffered Air-Cathode Microbial Fuel Cells. RSC Adv. 2017, 7, 33961–33966. 70. Wang, C.-T.; Huang, Y.-S.; Sangeetha, T.; Yan, W.-M. Assessment of Recirculation Batch Mode Operation in Bufferless Bio-Cathode Microbial Fuel Cells (MFCs). Appl. Energy 2018, 209, 120–126. 71. Ren, Y.; Chen, J.; Shi, Y.; Li, X.; Yang, N.; Wang, X. Anolyte Recycling Enhanced Bioelectricity Generation of the Buffer-free Single-Chamber Air-Cathode Microbial Fuel Cell. Bioresour. Technol. 2017, 244, 1183–1187. 72. Mei, X.; Xing, D.; Yang, Y.; Liu, Q.; Zhou, H.; Guo, C.; Ren, N. Adaptation of Microbial Community of the Anode Biofilm in Microbial Fuel Cells to Temperature. Bioelectrochemistry 2017, 117, 29–33. 73. Vilas Boas, J.; Oliveira, V. B.; Marcon, L. R. C.; Pinto, D. P.; Simões, M.; Pinto, A. M. F. R. Effect of Operating and Design Parameters on the Performance of a Microbial Fuel Cell with Lactobacillus pentosus. Biochem. Eng. J. 2015, 104, 34–40. 74. Rochex, A.; Godon, J. J.; Bernet, N.; Escudié, R. Role of Shear Stress on Composition, Diversity and Dynamics of Biofilm Bacterial Communities. Water Res. 2008, 42, 4915–4922. 75. Pham, H. T.; Boon, N.; Aelterman, P.; Clauwaert, P.; De Schamphelaire, L.; Van Oostveldt, P.; Verbeken, K.; Rabaey, K.; Verstraete, W. High Shear Enrichment Improves the Performance of Theanodophilic Microbial Consortium in a Microbial Fuel Cell. Microb. Biotechnol. 2008, 1, 487–496.

478

Fuel-Cell Bioreactors

76. Juang, D. F.; C Yang, P.; Chou, H. Y.; Chiu, L. J. Effects of Microbial Species, Organic Loading and Substrate Degradation Rate on the Power Generation Capability of Microbial Fuel Cells. Biotechnol. Lett. 2011, 33, 2147–2160. 77. Velvizhi, G.; Mohan, S. V. Electrogenic Activity and Electron Losses under Increasing Organic Load of Recalcitrant Pharmaceutical Wastewater. Int. J. Hydrogen Energy 2012, 37, 5969–5978. 78. Jia, J.; Liu, Y. T. B.; Wu, D.; Ren, N.; Xing, D. Electricity Generation from Food Wastes and Microbial Community Structure in Microbial Fuel Cells. Bioresour. Technol. 2013, 144, 94–99. 79. Oliveira, V. B.; Simões, M.; Melo, L. F.; Pinto, A. M. F. R. A 1D Mathematical Model for a Microbial Fuel Cell. Energy 2013, 61, 463–471. 80. Esfandyari, M.; Fanaei, M. A.; Gheshlaghi, R.; Akhavan Mahdavi, M. Mathematical Modeling of Two-Chamber Batch Microbial Fuel Cell with Pure Culture of Shewanella. Chem. Eng. Res. Des. 2017, 117, 34–42. 81. Wen, Q.; Wu, Y.; Cao, D.; Zhao, L.; Sun, Q. Electricity Generation and Modeling of Microbial Fuel Cell from Continuous Beer Brewery Wastewater. Bioresour. Technol. 2009, 100, 4171–4175. 82. Ismail, Z. Z.; Habeeb, A. A. Experimental and Modeling Study of Simultaneous Power Generation and Pharmaceutical Wastewater Treatment in Microbial Fuel Cell Based on Mobilized Biofilm Bearers. Renew. Energy 2017, 101, 1256–1265. 83. Yao, S.; He, Y.-L.; Song, B.-Y.; Li, X.-Y. A Two-Dimensional, Two-phase Mass Transport Model for Microbial Fuel Cells. Electrochim. Acta 2016, 212, 201–211. 84. Renslow, R.; Babauta, J.; Kuprat, A.; Schenk, J.; Ivory, C.; Fredrickson, J.; Beyenal, H. Modeling Biofilms with Dual Extracellular Electron Transfer Mechanisms. Phys. Chem. Chem. Phys. 2013, 15, 19262–19283. 85. Picioreanu, C.; Katuri, K. P.; Van Loosdrecht, M. C. M.; Head, I. M.; Scott, K. Modelling Microbial Fuel Cells with Suspended Cells and Added Electron Transfer Mediator. J. Appl. Electrochem. 2010, 40, 151–162. 86. Logan, B. E.; Hamelers, B.; Rozendal, R.; Schröder, U.; Keller, J.; Freguia, S.; Aelterman, P.; Verstraete, W.; Rabaey, K. Microbial Fuel Cells: Methodology and Technology. Environ. Sci. Technol. 2006, 40, 5181–5192. 87. Yoho, R. A.; Popat, S. C.; Fabregat-Santiago, F.; Giménez, S.; Heijne, A. t.; Torres, C. I. Electrochemical Impedance Spectroscopy as a Powerful Analytical Tool for the Study of Microbial Electrochemical Cells. In Biofilms Bioelectrochemical Syst.; Beyenal, H., Babauta, J., Eds; John Wiley & Sons, Inc: Hoboken, NJ, USA, 2015; pp 249–280. 88. Scott, K. Electrochemical principles and characterization of bioelectrochemical systems. In Microbial Electrochemical and Fuel Cells (Elsevier); Scott, K., Yu, E. H., Eds.; Woodhead Publishing, 2016; pp 29–66. 89. Park, J.-D.; Roane, T. M.; Ren, Z. J.; Alaraj, M. Dynamic Modeling of a Microbial Fuel Cell Considering Anodic Electron Flow and Electrical Charge Storage. Appl. Energy 2017, 193, 507–514. 90. Doyle, L. E.; Marsili, E. Methods for Enrichment of Novel Electrochemically-Active Microorganisms. Bioresour. Technol. 2015, 195, 273–282. 91. Rismani-Yazdi, H.; Carver, S. M.; Christy, A. D.; Tuovinen, O. H. Cathodic Limitations in Microbial Fuel Cells: An Overview. J. Power Sources 2008, 180, 683–694. 92. Luo, H.; Xu, G.; Lu, Y.; Liu, G.; Zhang, R.; Li, X.; Zheng, X.; Yu, M. Electricity Generation in a Microbial Fuel Cell Using Yogurt Wastewater under Alkaline Conditions. RSC Adv. 2017, 7, 32826–32832. 93. Cetinkaya, A. Y.; Ozdemir, O. K.; Demir, A.; Ozkaya, B. Electricity Production and Characterization of High-Strength Industrial Wastewaters in Microbial Fuel Cell. Appl. Biochem. Biotechnol. 2017, 182, 468–481. 94. Cetinkaya, A. Y.; Kaan, O.; Oguz, E.; Hasimoglu, A.; Ozkaya, B.; Ozdemir, O. K.; Koroglu, E. O.; Hasimoglu, A.; Ozkaya, B. The Development of Catalytic Performance by Coating Pt-Ni on CMI7000 Membrane as a Cathode of a Microbial Fuel Cell. Bioresour. Technol. 2015, 195, 188–193. 95. Ahn, Y.; Logan, B. E. Domestic Wastewater Treatment Using Multi-Electrode Continuous Flow MFCs with a Separator Electrode Assembly Design. Appl. Microbiol. Biotechnol. 2013, 97, 409–416. 96. Kim, H.; Kim, B.; Yu, J. Power Generation Response to Readily Biodegradable COD in Single-Chamber Microbial Fuel Cells. Bioresour. Technol. 2015, 186, 136–140. 97. Pandey, P.; Shinde, V. N.; Deopurkar, R. L.; Kale, S. P.; Patil, S. A.; Pant, D. Recent Advances in the Use of Different Substrates in Microbial Fuel Cells toward Wastewater Treatment and Simultaneous Energy Recovery. Appl. Energy 2016, 168, 706–723. 98. Escapa, A.; Mateos, R.; Martínez, E. J.; Blanes, J. Microbial Electrolysis Cells: An Emerging Technology for Wastewater Treatment and Energy Recovery. From Laboratory to Pilot Plant and beyond. Renew. Sustain. Energy Rev. 2016, 55, 942–956. 99. Wang, H.; Ren, Z. J. Bioelectrochemical Metal Recovery from Wastewater: A Review. Water Res. 2014, 66, 219–232. 100. Tao, Q.; Zhou, S.; Luo, J.; Yuan, J. Nutrient Removal and Electricity Production from Wastewater Using Microbial Fuel Cell Technique. Desalination 2015, 365, 92–98. 101. Feng, Y.; Barr, W.; Harper, W. F., Jr. Neural Network Processing of Microbial Fuel Cell Signals for the Identification of Chemicals Present in Water. J. Environ. Manag. 2013, 120, 84–92. 102. Commault, A. S.; Lear, G.; Bouvier, S.; Feiler, L.; Karacs, J.; Weld, R. J. Geobacter-dominated Biofilms Used as Amperometric BOD Sensors. Biochem. Eng. J. 2016, 109, 88–95. 103. Yu, D.; Bai, L.; Zhai, J.; Wang, Y.; Dong, S. Toxicity Detection in Water Containing Heavy Metal Ions with a Self-Powered Microbial Fuel Cell-Based Biosensor. Talanta 2017, 168, 210–216. 104. Saeed, H. M.; Husseini, G. A.; Yousef, S.; Saif, J.; Al-Asheh, S.; Abu Fara, A.; Azzam, S.; Khawaga, R.; Aidan, A. Microbial Desalination Cell Technology: A Review and a Case Study. Desalination 2015, 359, 1–13. 105. Zuo, K.; Chang, J.; Liu, F.; Zhang, X.; Liang, P.; Huang, X. Enhanced Organics Removal and Partial Desalination of High Strength Industrial Wastewater with a Multi-Stage Microbial Desalination Cell. Desalination 2017, 423, 104–110. 106. Pandit, S.; Nayak, B. K.; Das, D. Microbial Carbon Capture Cell Using Cyanobacteria for Simultaneous Power Generation, Carbon Dioxide Sequestration and Wastewater Treatment. Bioresour. Technol. 2012, 107, 97–102. 107. Wang, X.; Feng, Y. J.; Liu, J.; Lee, H.; Li, C.; Li, N.; Ren, N. Sequestation of CO2 Discharged from Anode by Algal Cathode in Microbial Carbon Capture Cells (MCCs). Biosens. Bioelectron. 2010, 25, 2639–2643. 108. Xiang, Y.; Liu, G.; Zhang, R.; Lu, Y.; Luo, H. High-efficient Acetate Production from Carbon Dioxide Using a Bioanode Microbial Electrosynthesis System with Bipolar Membrane. Bioresour. Technol. 2017, 233, 227–235. 109. Logan, B. E. Microbial Fuel Cells; Wiley Bicentennial: United States of America, 2008. 110. Mamais, D.; Noutsopoulos, C.; Dimopoulou, A.; Stasinakis, A.; Lekkas, T. D. Wastewater Treatment Process Impact on Energy Savings and Greenhouse Gas Emissions. Water Sci. Technol. 2015, 71, 303–308. 111. Shemfe, M.; Gadkari, S.; Yu, E.; Rasul, S.; Scott, K.; Head, I. M. Bioresource Technology Life Cycle, Techno-Economic and Dynamic Simulation Assessment of Bioelectrochemical Systems: A Case of Formic Acid Synthesis. Bioresour. Technol. 2018, 255, 39–49. 112. Stoll, Z. A.; Ma, Z.; Trivedi, C. B.; Spear, J. R.; Xu, P. Sacrificing Power for More Cost-Effective Treatment: A Techno-Economic Approach for Engineering Microbial Fuel Cells. Chemosphere 2016, 161, 10–18. 113. Foley, J. M.; Rozendal, R. A.; Hertle, C. K.; Lant, P. A.; Rabaey, K. Life Cycle Assessment of High-Rate Anaerobic Treatment, Microbial Fuel Cells, and Microbial Electrolysis Cells. Environ. Sci. Technol. 2010, 44, 3629–3637. 114. Lu, M.; Chen, S.; Babanova, S.; Phadke, S.; Salvacion, M.; Mirhosseini, A.; Chan, S.; Carpenter, K.; Cortese, R.; Bretschger, O. Long-term Performance of a 20-L Continuous Flow Microbial Fuel Cell for Treatment of Brewery Wastewater. J. Power Sources 2017, 356, 274–287. 115. Jiang, D.; Curtis, M.; Troop, E.; Scheible, K.; McGrath, J.; Hu, B.; Suib, S.; Raymond, D.; Li, B. A Pilot-Scale Study on Utilizing Multi-Anode/cathode Microbial Fuel Cells (MAC MFCs) to Enhance the Power Production in Wastewater Treatment. Int. J. Hydrogen Energy 2011, 36, 876–884.

2.33

Enzyme Bioreactors

C Zhang and X-H Xing, Tsinghua University, Beijing, China © 2011 Elsevier B.V. All rights reserved. This is a reprint of C. Zhang, X.-H. Xing, 2.23 - Enzyme Bioreactors, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 319-329.

2.33.1 2.33.2 2.33.2.1 2.33.2.2 2.33.2.3 2.33.2.4 2.33.3 2.33.3.1 2.33.3.2 2.33.3.3 2.33.3.4 2.33.4 2.33.4.1 2.33.4.2 2.33.5 2.33.5.1 2.33.5.2 2.33.5.3 2.33.5.4 References

Introduction Forms of Enzymes Used in Enzyme Reactors Adsorption Covalent Binding Entrapment Membrane Confinement Enzyme Reactors Stirred-Tank Reactor Packed-Bed Reactor Fluidized-Bed Reactor Membrane Reactor Design and Choice of Enzyme Reactors Design of Enzyme Reactors Choice of Enzyme Reactors Novel Enzyme Reactors Enzyme Reactors With Cofactor Regeneration Enzyme Reactors With Nonaqueous Media Enzyme Reactors With Multienzyme Reactions Microreactors

479 480 480 481 482 482 483 483 483 484 485 486 486 486 487 487 487 488 488 488

Glossary Fluidized-bed reactor (FBR) A kind of enzyme reactor consisting of a bed of immobilized enzyme, which is fluidized by the rapid upward flow of the substrate stream alone. Immobilization Restricting the enzyme molecules in a fixed space. Membrane reactor (MR) A kind of enzyme reactor that consists of a semipermeable membrane that allows free passage of the product molecules but constrains the enzyme molecules. Packed-bed reactor (PBR) A kind of enzyme reactor packed with immobilized enzymes, in which the substrate enters at one end of a cylindrical tube and the product stream leaves at the other end. Stirred-tank reactor (STR) The most common type of enzyme reactors, in which the substrates and enzymes are introduced normally into a tank fitted with fixed baffles that improve stirring efficiency.

2.33.1

Introduction

An enzyme reactor is a device within which biochemical transformations or reactions are performed by the catalysis of enzymes to generate the expected products under mild conditions. When the transformations or reactions are carried out by living cells, the bioreactor is often called a ‘fermenter’. However, in this article, we call the bioreactor employing enzymes an ‘enzyme reactor‘ to distinguish it from the bioreactor that employs living cells, the fermenter.4 Enzyme reactors are extensively used for food processing, industrial biotransformations, pharmaceutical processing, biosensors, and so on. An enzyme reactor consists of a vessel, or a series of vessels, containing the reactants and the enzyme biocatalysts to perform the desired conversion. Generally, all enzyme reactors deal with heterogeneous systems, consisting of two or more phases, for example, liquid, gas, and solid. Therefore, optimal conditions for the enzyme-catalyzed reactions necessitate efficient transfer of mass, heat, and momentum from one phase to the other(s). Thus, an enzyme reactor should provide agitation (baffles); regulation of temperature (heating/cooling coils in the reactor), pH, and so on; and withdrawal of enzymes or reaction solutions. Enzyme reactors differ from conventional chemical reactors in that they support and control biological entities. Because of the nature of the enzymes, which may have time-varying properties and may show complex kinetic patterns, enzyme reactors may show special characteristics compared to conventional chemical reactors. The first key difference between chemical reactors

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00083-5

479

480

Enzyme Bioreactors

and enzyme bioreactors is selectivity. In enzyme bioreactors, selectivity for producing the preferred products is much higher. Second, deactivation of the enzymes often poses more severe consequences than those resulting from a chemical upset.2 On the basis of mode of operation, an enzyme reactor may be classified as batch or continuous.7 In the batch operation mode, the enzyme and the substrate molecules have identical residence times within the reactor. In the ideal case, batch reactors have a homogeneous continuous phase; the liquid phase is well mixed and of uniform temperature and composition. There are no spatial variations in reactant or product concentrations. Batch operation mode for enzyme reactors has a number of advantageous features: (1) simplicity and flexibility both in use and in bioprocess development; (2) closely controllable environment useful for slow reactions; (3) usability when continuous operation of a bioprocess proves difficult due to the viscous or intractable nature of the reaction mix; (4) lower capital investment compared to continuous processes for the same bioreactor volume. However, the disadvantages of batch mode are (1) operating costs higher than those for continuous processes due to the time required for filling, heating, sterilizing, cooling, emptying, and cleaning the reactor; (2) uneven demands on both labor and services for this nonstationary process; and (3) pronounced batch-to-batch variations, as the reaction conditions may change with time, which may make the scale-up and quality control of the products difficult. On the other hand, in the continuous operation mode, fresh substrate is continuously fed, while the reaction solution is continuously removed to keep the solution volume constant. These systems have a number of advantages, which include (1) increased potential for automating the bioprocess and thus reducing labor costs; (2) significantly increased productivity of the reactor as downtime is eliminated; and (3) consistent product quality due to nonvarying operating parameters and the capability to operate at steady state. Along with the strengths, there are certain inherent disadvantages of continuous enzyme reactors: (1) higher investment costs for control and automation equipment; (2) higher processing costs for continuous replenishment of nonsoluble, solid substrates; (3) minimal operation flexibility, as only slight variations in the bioprocess are allowable.2 On the basis of the form of the enzymes used, enzyme reactors can be categorized into two broad classes: (1) reactors that use free enzymes, and (2) reactors that use immobilized enzymes.7 Also, on the basis of the type of the reactor used, enzyme reactors can be classified as stirred-tank reactor (STR), packed-bed reactor (PBR), fluidized-bed reactor (FBR), and membrane reactor (MR). In this article we will first introduce the two different forms of the enzymes used in enzyme reactors, and then describe the four types of enzyme reactors. The design and choice of enzyme reactors will also be discussed. Finally, novel enzyme reactors including enzyme reactors with cofactor regeneration, with nonaqueous media, and with multienzyme reactions will be introduced.

2.33.2

Forms of Enzymes Used in Enzyme Reactors

Most industrial enzymes are crude hydrolytic enzymes that can degrade high-molecular-weight polymers. Although immobilized enzymes are more and more feasible from the economic and technological point of view, free enzymes, such as amylase and protease, are still used in many industrial processes. We can use free enzymes in solution in a number of cases; however, in many practical applications, it is advantageous to employ enzymes in an immobilized form. In general, ‘enzyme immobilization‘ is defined as restricting the enzyme molecules to a fixed space; it is achieved by fixing the enzyme to, or within, a macroscopic support matrix. Immobilized enzymes offer several potential advantages over soluble free enzymes: 1. As enzymes are expensive, enzyme reutilization is critical for many bioprocesses. When free enzymes are used in a soluble form, they retain some activity after the reaction, which cannot be economically recovered for reuse. A main advantage of immobilized enzymes is that they can be reused, as they typically are macroscopic catalysts that can be retained in the reactors. 2. Soluble enzymes can contaminate the product, and their removal may involve extra purification costs. The most important benefit derived from immobilization is the elimination of enzyme recovery and purification processes. Product purity is usually improved, and effluent-handling problems are minimized, by immobilization, particularly if the enzyme is noticeably toxic or antigenic. 3. Immobilized enzymes can be employed in a wide range of reactor configurations, and, because high concentrations of the biocatalysts can be obtained, correspondingly high volumetric productivities are possible, which leads to lower capital costs. It also allows continuous operation to be practicable, with a considerable saving in enzymes, labor, and overhead costs. 4. An immobilized enzyme may show selectively altered chemical or physical properties, and it may provide a better environment for the enzyme activity. Therefore, immobilized enzymes are often more stable than free enzymes in a solution. It is also important to note that the properties of the supports used for immobilizing enzymes can in some cases be exploited to modify the behavior of the enzymes. Enzymes can be immobilized on the surface of or inside water-insoluble supports by a wide variety of methods. In recent years hundreds of papers that describe the techniques for immobilizing enzymes have been published. We will consider only some of the general characteristics of the major classes of immobilization techniques for enzymes and compare their advantages and disadvantages. There are four principal methods available for immobilizing enzymes: adsorption, covalent binding, entrapment, and membrane confinement (Figure 1).

2.33.2.1

Adsorption

The simplest way to immobilize enzymes is attachment of the enzyme molecules to the surface of support particles by weak physical forces.5 The driving force causing this binding is usually due to a combination of hydrophobic effects and formation of several salt

Enzyme Bioreactors

B

E

E

E

E

E

E

C

E

E

E E

E

E

E

E

E

C

D

E

E

E E E

E E

E

E

E

E

EE

E E

E

E

E

E

E

E

E

E E

Figure 1

E

E

E

C

E

E

E

E

E

E

A

481

Different methods for enzyme immobilization: (A) adsorption; (B) covalent binding; (C) entrapment; (D) membrane confinement.

links per enzyme molecule.8 The immobilization protocol consists of exposing the enzyme in solution to a surface-active adsorbent under appropriate conditions of pH, ionic strength, temperature, and so on. Commonly employed adsorbents can be inorganic materials, such as alumina, carbon, calcium carbonate, or organic materials, such as celluloses, starch, anion-exchange resins, or cation-exchange resins. In recent years, many functional nanoscale materials were also widely developed for immobilizing enzymes by methods such as self-assembly of the protein molecules and nanomaterials. The advantages of the adsorption techniques for enzyme immobilization are the following: 1. 2. 3. 4.

Adsorption of enzymes to the matrices is very easy and capable of high enzyme loading. It is possible to separate and purify the enzymes while being immobilized. The active sites of the adsorbed enzymes are usually unaffected, and the enzymes are not usually deactivated by adsorption. Easy removal of the enzyme from the support is possible, as adsorption is a reversible process.

However, adsorption methods for enzyme immobilization also have several disadvantages: 1. Desorption of enzymes is a common problem that limits immobilization efficiency. 2. The state of immobilization is very sensitive to pH, ionic strength, and temperature of the solution. 3. The amount of enzymes loaded on a support particle is usually low due to the weak interactive force.

2.33.2.2

Covalent Binding

An alternative to physical adsorption, covalent binding, is the retention of the enzyme on support surfaces by covalent bond formation.5 As enzymes typically contain some combination of functional groups that are reactive to a wide variety of common reagents, the enzyme molecules can be covalently attached via these nonessential amino acid residues to insoluble matrices. A variation of direct covalent attachment is co-polymerization of the enzyme with a reactive monomer. In this approach, the enzyme molecules are extensively cross-linked to each other, either with or without an added support, thus producing a polyenzyme network. The functional groups of enzymes that bind to support materials should be nonessential amino acid residues, which must not be at the active sites. The groups that are suitable for the immobilization process can be free amino, carboxyl, hydroxyl, and sulfhydryl groups. Particularly, lysine residues are the most useful groups in covalent binding due to their wide surface exposure and high reactivity, especially in slightly alkaline solutions; also they only very rarely occur at the active sites of enzymes. Some common water-insoluble supports used for covalent immobilization of enzymes are synthetic supports such as acrylamide-based polymers, maleic anhydride-based polymers, methacrylic acid-based polymers, styrene-based polymers, and polypeptides, and natural supports such as agarose, cellulose, dextran, glass, and starch. Functional groups on the support materials are usually activated by using chemical reagents, such as cyanogen bromide, carbodiimide, and glutaraldehyde.5 The most commonly employed matrices for covalent enzyme immobilization include agaroses, celluloses, and polyacrylamides. The main advantage of the covalent binding method is the strong binding force, and hence there is no enzyme loss while using the immobilized enzymes. However, the covalent immobilization method also has several disadvantages: 1. Regeneration of the biocatalyst is not possible.

482

Enzyme Bioreactors

2. Covalent binding through multiple sites on the enzyme molecule may lead to a loss in enzyme activity either because of the significant changes in the active site of the enzyme or because of immobilization of the enzyme in a particular orientation, which either distorts the active site or renders it unavailable. 3. The reagents involved in covalent binding increase the cost of the immobilization procedure. 4. The loading of enzyme is generally low (
0.02 g g1 matrix).

2.33.2.3

Entrapment

Entrapment is the physical enclosure of enzymes in a small space.5 Enzymes can be entrapped within cross-linked polymers by forming a highly cross-linked network of polymers in the presence of the enzyme. In this approach, enzyme molecules are held or entrapped within suitable gels or fibers, and there may or may not be covalent bond formation between the enzyme molecules and the matrix. A noncovalent entrapment may be viewed as putting the enzyme molecules in a molecular cage just like a caged bird/animal.1 As there is no chemical modification of the enzyme, the intrinsic properties of the enzyme are never altered. Further, enzyme loading can be high. However, this immobilization method has the following disadvantages: 1. Diffusion of the substrates toward the enzymes and of the products away from the enzymes pose some difficulties. However, diffusion limitations can be eliminated by reducing the particle sizes of the matrices and/or capsules.5 2. The enzyme may be deactivated during gel formation. 3. Enzyme leakage is also a problem, such as enzyme leakage into solution due to the breakage of the gels or capsules. But such leakages can be overcome by reducing the molecular weight (MW) cutoff of the membranes or the pore size of the solid matrices.5 4. Lack of control of the microenvironmental conditions inside the solid matrices always results in reduced enzyme activity and stability. However, by using different matrices and chemical ingredients, by changing the processing conditions, and by reducing particle or capsule sizes, more favorable microenvironmental conditions can be obtained.5 5. The difficulty which large molecules face in approaching the catalytic sites of the entrapped enzymes precludes the use of entrapped enzymes for high-molecular-weight substrates. Matrices used for enzyme immobilization are usually polymeric materials such as Ca alginate, agar, polyacrylamide, and collagen. However, some solid matrices such as activated carbon, porous ceramic, and diatomaceous earth can also be used for this purpose. The matrix can be composed of particles, membranes, or fibers.5 The most commonly employed cross-linker polymer is the polyacrylamide gel system. When immobilizing in a polymer matrix, the enzyme solution is mixed with the polymer solution before polymerization takes place. Either the polymerized, gel-containing enzyme is extruded or a template is used to shape the particles from a liquid polymer–enzyme mixture.5 Entrapment and surface attachment may be used in combination in some cases for enzyme immobilization; membrane entrapment of enzymes is an example. Hollow fiber units have been used to entrap an enzyme solution between thin and semipermeable membranes. Membranes made of nylon, cellulose, polysulfone, or polyacrylate can be used. A semipermeable membrane is used to retain high-molecular-weight compounds (enzyme), while allowing low-molecular-weight compounds (substrates or products) to move toward or away from the enzymes.5 Entrapment procedures can utilize membranes with pores small enough to prohibit release of the enzymes but large enough to permit the passage of substrates and products through the membrane. Thus, the enzyme remains in solution, and any potential adverse effects resulting from adsorption or covalent attachment can be avoided by membrane entrapment.3

2.33.2.4

Membrane Confinement

Membrane confinement is a special form of membrane entrapment, by which enzyme molecules in an aqueous solution may be confined within a semipermeable membrane. In this approach, microscopic hollow spheres are formed. These spheres contain the enzyme solution, while they themselves are enclosed within a porous membrane. The membrane can be polymeric or an enriched interfacial phase formed around a microdrop.5 A number of strategies are employed for membrane confinement of enzymes, all of which depend on the semipermeable nature of the membrane. The simplest strategy is realized by partitioning the reactor into two chambers by a semipermeable membrane: One chamber contains the enzyme, whereas the other has the reactant and product streams. Hollow-fiber membrane units are commercially available with large surface areas relative to their contained volumes and permeable only to substances of molecular weight substantially less than that of the enzymes.8 Immobilization of enzymes often results in an additional expense and is only undertaken if there is an obvious economic or bioprocess advantage in the use of the immobilized enzyme. In general, the best immobilization method for a particular enzyme and application will depend on several interrelated parameters. The most suitable support material and immobilization method vary depending on the enzyme and the particular application. These include as follows: 1. Cost of both the enzyme and the support. 2. Useful lifetime of the immobilized enzyme. It would be better if the carrier can be regenerated after the useful lifetime of the immobilized enzyme. 3. Binding capacity of the support material. 4. Stability and retention of enzymatic activity, which is a function of the functional groups on support material and the microenvironmental conditions.5

Enzyme Bioreactors

483

5. The nature of the supports will also have a considerable effect on an enzyme‘s expressed activity and apparent kinetics. The form, shape, density, porosity, pore size distribution, operational stability, and particle size distribution of the supporting matrix will influence the reactor configuration in which the immobilized biocatalysts may be used.8

2.33.3

Enzyme Reactors

2.33.3.1

Stirred-Tank Reactor

The most common type of enzyme reactor in use today is STR (Figure 2). The operating principles of the stirred-tank bioreactors are relatively simple. The substrates and enzymes are introduced into a tank, which is normally fitted with fixed baffles that improve the stirring efficiency, thereby enhancing mass transfer. Temperature and pH control are always needed in STRs for ensuring the optimal conditions of the enzymes. The structure of STR is simple; and inside the reactor, the enzyme and the substrate can be mixed easily, which results in lower diffusion resistance. Also, the control of STR is easy. Thus, STR is ideal for industrial applications – especially for the treatment of substrates with high viscosity or poor solubility – and can offer manufacturers both low capital costs and low operating costs. On the other hand, STR has some disadvantages such as low reaction efficiency, high energy input needed for stirring, and the need to remove the enzymes from the products once the desired bioreaction has taken place.2 Based on operation mode, STR can be classified as batch stirred-tank reactor (BSTR) and continuous stirred-tank reactor (CSTR).2 BSTR is the simplest type of reactor. Actually, the simplest reactor configuration for any enzyme reaction is the batch mode. An ideal batch reactor is assumed to be well mixed so that the contents are uniform in composition at all times. BSTR is one such reactor in which the entire product is removed, as rapidly as is practically possible, after a fixed time. Both free enzymes and immobilized enzymes can be used in BSTR. BSTR is useful for substrate solutions with high viscosity and for immobilized enzymes with relatively low activity. However, a problem that always arises is that an immobilized enzyme tends to decompose upon physical stirring. The batch system is generally suitable for the production of only relatively small amounts of chemicals.6 CSTR is an ideal reactor with continuous feeding of reactants based on the assumption that the reactor contents are well mixed. The substrate stream is continuously pumped into the reactor and at the same time the product stream is removed. Therefore, the concentrations of the various components of the outlet stream are assumed to be the same as the concentrations of these components in the reactor. CSTR is an easily constructed, versatile, and cheap reactor, which allows simple biocatalyst charging and replacement. Compared to BSTR, CSTR is more efficient; but the equipment is relatively more complicated. The high cost of enzymes always prevents their continuous addition to the feed of a CSTR; hence the enzyme must be retained in the reactor. For this the most common method is to use a combined CSTR/UF reactor, which is a combination of a CSTR with an ultrafiltration unit. A continuous ultrafiltration membrane device is added at the exit of the CSTR; thus the product passes through the ultrafiltration unit, where the enzyme is removed and recycled back into the reactor. A hollow-fiber device can also be used, and its characteristics are essentially the same as those of an ultrafiltration membrane. The combination of CSTR and UF membrane is suitable for a substrate of high molecular weight and a product of low molecular weight.6 Alternatively, immobilizing the enzyme in a pellet held in the reactor is another solution. Sometimes magnetic particles are used to immobilize the enzymes, which facilitates the separation of enzymes from the reactor by a magnetic field.

2.33.3.2

Packed-Bed Reactor

PBR is a kind of reactor packed with immobilized enzymes, in which the substrate enters at one end of a cylindrical tube and the product stream leaves at the other end4 (Figure 3). The most important feature of a PBR is that the reaction solution flows through

A

Figure 2

Stirred-tank rector: (A) BSTR; (B) CSTR.

B

484

Figure 3

Enzyme Bioreactors

Packed-bed reactor.

the reactor as a plug-flow; so they are also called plug-flow reactors (PFR). The long tube and lack of a stirring device prevents complete mixing of the fluid in the reactor. Properties of the flowing stream will vary in both the longitudinal and the radial directions, but the variation in the radial direction is always much smaller than that in the longitudinal direction. Ideally all of the substrate stream flows at the same velocity, parallel to the reactor axis with no back-mixing.8 In order to produce ideal plug flow within PBRs, a turbulent flow regime is preferred to laminar flow, as this causes improved mixing and heat transfer normal to the flow and reduced axial back-mixing. Because of the type of flow inside the reactor, it is also called a tubular-flow enzyme reactor. Generally, immobilized enzymes are used in PBR. Continuous PBR (CPBR) is the most widely used reactor for immobilized enzymes, which has the following advantages over a batch PBR: ease of automatic control and operation; reduction of labor costs; stabilization of operating conditions, and ease of quality control of products.6 A significant advantage of PBR is that high concentrations of immobilized enzymes can be used. With the enzyme immobilized in the reactor bed, the solution of the substrate for conversion is passed through for conversion into the product. The product is continuously collected as effluent from the bioreactor. Heterogeneous systems enable product recovery at lower separation costs than possible in the corresponding homogeneous systems.2 Concerning the conversion efficiency of PBR, any required degree of reaction may be achieved by use of an ideal PBR with suitable length. Besides, as compared with CSTR, PBRs are the preferred reactors for processes involving product inhibition, substrate activation, and reversible reaction. The disadvantages of PBR are the following: 1. The design of PBRs allow neither for control of pH by the addition of acids or bases, nor for easy temperature control where there is an excessive heat output, a problem that may be particularly noticeable in large-scale reactors.8 2. The concentrations of the substrate and the product change along the length of the reactor. 3. Immobilized enzymes are easily fouled by colloidal formation or precipitates, and hence the question of cleaning or changing the immobilized enzymes arises. 4. A vicious circle of increased back pressure, particle deformation, and restricted flow may eventually result in no flow at all through the PBR.8 5. Channels may form in the reactor bed due to excessive pressure drop, irregular packing, or uneven loading of the substrate streams, causing flow rate differences across the bed.8

2.33.3.3

Fluidized-Bed Reactor

FBR consists of a bed of immobilized enzyme fluidized by the rapid upward flow of the substrate stream alone8 (Figure 4). The substrate is passed upward through the immobilized enzyme bed at a velocity high enough to lift the particles. The enzyme is immobilized in small particles that move with the fluid. There is a minimum fluidization velocity needed to achieve the bed expansion, which depends upon the size, shape, porosity, and density of the particles, and the density and viscosity of the solution.8 FBR is a combination of CSTR and PBR, which behaves in a manner intermediate between those of these two types of reactors. Thus, the kinetic performance of an FBR normally lies between those of a PBR and a CSTR. The actual design of an FBR will determine whether it behaves in a manner that is closer to that of a PBR or a CSTR. It can, for example, be made to behave in a manner very similar to that of a PBR if it is baffled in such a way that substantial back-mixing is avoided. FBRs are chosen when these intermediate characteristics are required, for example, where a high conversion is needed but the substrate stream is colloidal or the reaction produces a substantial pH change or heat output.8 Particle size of the immobilized enzymes is an important factor for the formation of a smooth fluidized bed. FBR is normally used with fairly small-sized immobilized enzyme particles in order to achieve a high biocatalytic surface area. These particles must be sufficiently dense, relative to the substrate stream, so that they are not swept out of the reactor. Less-dense particles must be

Enzyme Bioreactors

Figure 4

485

Fluidized-bed reactor.

somewhat larger. For efficient operation, the particles should be of nearly uniform size, otherwise a nonuniform biocatalytic concentration gradient will form up the reactor. FBRs are usually tapered outwards the exit to allow for a wide range of flow rates.4 The advantages of FBR are the following: 1. 2. 3. 4.

ease of control of pH, temperature, and gas supply; excellent heat and mass transfer characteristics, which make the reactor ideal for highly exothermic reactions; especially suitability when a high-viscosity substrate solution or powder substrate is used; and relatively low pressure drop even when using very small particles.

However, FBR still has some disadvantages8: 1. 2. 3. 4. 5. 6.

requirement of a large power input for fluidizing the bed; difficulty in scaling up (can only be scaled up by a factor of 10–100 each time); channeling and biocatalyst loss caused by the very high flow rates; susceptibility to easy destruction and decomposition of the immobilized enzymes due to the fluidization of the particles; suitable only for relatively low concentrations of the immobilized enzymes; and complex changes in the flow patterns within these reactors are caused by changes in the flow rate of the substrate streams, which may have consequent unexpected effects upon the conversion rate.

2.33.3.4

Membrane Reactor

MR (Figure 5) is a functional combination of enzyme catalysis and membrane separation. The main requirement of an MR is a semipermeable membrane that allows the free passage of the product molecules but constrains the enzyme molecules. MR may be used in either batch or continuous mode and allows easy separation of the enzyme from the product. If a substrate is able to diffuse through the membrane, it may be introduced on either side of the membrane with respect to the enzyme, otherwise it must be within the same compartment as the enzyme.8 The kinetics of MR is similar to that of BSTR, in batch mode, and CSTR, in continuous mode (discussed later). We have previously discussed a simple case of MR as the combination of CSTR and UF units, which is mostly suitable for free enzymes. Deviations from this model occur primarily in configurations where the substrate stream is on the side of the membrane opposite to that of the enzyme and the reaction is severely limited by its diffusion through the membrane and the products‘ diffusion in the reverse direction. Under these circumstances the reaction may be even more severely affected by product inhibition or the limitations of reversibility than is indicated by these models.8 The usual choice for MR is a hollow-fiber reactor consisting of a preformed module containing hundreds of thin tubular fibers each having an inner diameter of
200–300 mm and an outer diameter of
300–900 mm. Diffusion of the substrate through the tubular wall allows it to make contact with the gelled enzyme and be converted into the product. Subsequent diffusion of the product makes the separation easy for its recovery. Under the influence of the differential pressure along the tubular wall, the product flows inside the tubes, eventually to be collected at a multitube header.2 The advantages of MR are the following: (1) because of the ease with which MR systems may be established, they are often used for production on a small scale (grams to kilograms), especially where a multienzyme system or coenzyme regeneration is needed and (2) MR allows easy replacement of the enzymes in processes involving particularly labile enzymes and it can also be used for biphasic reactions. The major disadvantages of MR are the cost of the membranes, membrane fouling, and the need for membrane change at regular intervals.

486

Enzyme Bioreactors

Figure 5

Membrane reactor.

2.33.4

Design and Choice of Enzyme Reactors

2.33.4.1

Design of Enzyme Reactors

Enzyme reactor design is a relatively complex engineering task. The goal of an effective enzyme reactor is to control and optimize the desired biological reactions. Under the optimum conditions, the enzymes are able to perform their functions with 100% rate of success. To accomplish this, suitable reactor-operating parameters for the desired biological reactions should be taken into consideration. That is, the enzyme rector‘s environmental conditions such as substrate supply, product and byproduct removal, controlled temperature, pH, agitation speed, and water availability need be monitored and controlled.7 The design objectives for an enzyme reactor are to make full use of the advantages of the chosen type of reactor; to overcome its disadvantages; and finally to acquire the products with high productivity, high quality, and low cost. The basic rules for the designing of enzyme reactors are the following4: 1. kinetics of the enzyme-catalyzed reaction for a substrate, and the effects of operating parameters (temperature, pressure, pH, etc.) on the kinetics; 2. type of enzyme reactor and the flowing state/heat transfer characteristics of the fluids inside the reactor; and 3. yield of the product and the production process. Basically, the key idea for the designing of enzyme reactors is to achieve the highest concentration of the product at the lowest cost.

2.33.4.2

Choice of Enzyme Reactors

As shown in Section 2.33.3, several different types of enzyme reactors are available, like STR, PBR, FBR, and MR. The various types of reaction systems covered here include batch and continuous. As stated before, these enzyme reactors have quite different properties for enzymatic reactions. There are several important factors that determine the choice of an enzyme reactor for a particular process. 1. The form of the enzyme to be used. The choice of a reactor type depends on the form of the enzyme used (i.e., free or immobilized). For a free enzyme, as it is difficult to separate the enzyme from the reagent, only a batch reactor like BSTR is suitable. Otherwise, a UF unit should be added at the exit of CSTR preventing the loss of the free enzyme. However, the advantages of immobilized enzymes as processing catalysts are most markedly appreciated in continuous flow reactors. In these reactors, the average residence time of the substrate molecules is far shorter than that of the immobilized enzyme. Other contributing factors are the kinetics of the enzyme reaction, and the chemical and physical properties of the immobilization support including whether it is particulate, membranous, or fibrous, and its density, compressibility, robustness, particle size, and regenerability.8 For example, particulate enzymes are suitable for CSTR or PBR; membranous or fibrous enzymes are suitable for PBR; whereas small particulate enzymes are suitable for FBR. 2. The physical properties of substrates. In principle, if the substrate is soluble, any type of reactor can be used. However, if the substrate is of low solubility, reactors operated in batch mode are more suitable for the reaction of large volumes of the solution containing a low concentration of the substrate. 3. The requirements of the reaction operation. Attention must also be paid to the scale of operation, the possible need for pH and temperature control, the supply and removal of gases, and the stability of the enzymes, substrates, and products. For example, when an enzyme reaction needs to adjust the pH frequently, STR will be preferred, whereas when oxygen or gas supply is needed, FBR is always the good choice. 4. Stability of the enzymes. In an enzyme reactor, the highest specific enzyme activity is always desirable. The most possible reasons for loss of activity of the immobilized enzymes in the reactors are the following: denaturing of the enzyme itself, disconnection

Enzyme Bioreactors

487

of the enzyme from the carrier, and disruption of the carrier. Of all the different types of enzyme reactors introduced in this article, CSTR is the one that would result in this kind of inactivation. 5. The cost concern. In general, the choice of reactor depends on the cost of a predetermined productivity within the product‘s specifications. This must be inclusive of the costs associated with substrate(s), downstream processing, labor, depreciation, overheads, and process development, in addition to the more obvious costs concerned with building and running the enzyme reactor. As discussed above, there is no one simple or standard rule for the choice of enzyme reactors; one should balance all the different concerns regarding the performance of the reactor and of the whole process, and then make a decision.

2.33.5

Novel Enzyme Reactors

2.33.5.1

Enzyme Reactors With Cofactor Regeneration

Cofactors are required in approximately one-third of all enzymatic reactions.3 Many oxidoreductases and all ligases utilize cofactors (e.g., NADþ, NADPþ, NADH, NADPH), which must be regenerated as each product molecule is formed. Although these enzymes represent many of the most useful biocatalysts, their applications are severely limited by the high cost of the cofactors and the difficulties encountered in their regeneration.8 Cofactors are expensive; hence it is important for them to be recycled, that is, retained and regenerated, especially in large-scale enzymatic processes. Efficient cofactor regeneration in vitro remains a technological challenge, however, and the limited regeneration system currently represents an obstacle to the industrial application of immobilized enzymes for multistep reactions. There are several systems available for the regeneration of cofactors, by chemical, electrochemical, or enzymatic means. Enzymatic regeneration of cofactors is advantageous because of its high specificity, but electrochemical methods for regenerating oxidoreductase dinucleotides are competitive. A useful cofactor-regenerating system, using enzymatic processes, must utilize cheap substrates and readily available enzymes and give noninterfering and easily separated products. Formate dehydrogenase and acetate kinase represent useful examples, although the presently available commercial enzyme preparations are of low activity. In a single-step reaction catalyzed by immobilized enzymes, cofactor regeneration can be accomplished by several means: 1. Cofactors usually must be modified for adequate immobilization and regeneration. One of the more effective strategies involves coupling NAD(H) to a soluble polymer (e.g., polyethylene glycol (PEG) or polysaccharide dextran).3 When successfully applied, the cofactors attached to the immobilization support can keep their biological function. The macromolecular polymer-bound cofactors can then be retained, along with the enzymes of interest, in an ultrafiltration apparatus for continuous reuse. High-molecular-weight water-soluble derivatives are the most useful, as they cause less diffusion resistance than insoluble coenzyme-contained matrices. Dextrans, polyethyleneimine, and polyethylene glycols are widely used. However, the cost of such derivatives is always likely to remain high, and they will only be economically adopted for the production of very high-value products. 2. MRs may be used to immobilize cofactors, but the pore size must be less than the cofactors‘ diameters, which is extremely restrictive. In this case, it is basically required that the cofactors have sufficient size to be retained within the system.

2.33.5.2

Enzyme Reactors With Nonaqueous Media

Even though enzyme reactions are always performed in aqueous solutions, many reactants are more soluble in organic solvents than in water and some products may be quite labile in an aqueous environment. It would often be useful if enzymatic reactions could be performed in solvents other than water, as water is not the ideal medium for the majority of organic reactions. Enzyme catalysis in organic solvents has been of great importance in organic synthesis. For enzyme reactions in organic solvents, the characteristics of the enzymes can always be significantly altered in terms of stability, substrate specificity, and enantioselectivity. One major factor must first be addressed: the stability of the enzymes in these organic solvent systems. The active integrity and stability of hydrophilic enzymes appear to depend on the presence of a thin layer of water within the microenvironment. This amount of water is minuscule, and the enzymes may be handled in an almost anhydrous state. If the enzyme-bound water is stripped out or diluted by the use of more water-soluble, or miscible, organic solvents, then the enzyme is usually inactivated. However, under the conditions where this does not occur, the associated reduction in water activity will considerably reduce the rate of thermo-inactivation of enzymes.8 This has a stabilizing effect on most enzymes. In addition, freezing point of water in organic solvent systems is lower, which allows the use of particularly heat-labile enzymes at very low temperatures. The lowering of water activity tends to produce a more rigid enzyme molecule, which may affect the catalytic efficiency and properties of the enzymes. The most important factor that influences the stability of enzymes and the catalytic reactions in organic solvents is solvent polarity. Solvents of lower polarity have lesser ability to disrupt the structure of the enzymes with the tightly bound necessary water molecules. The best measure of solvent polarity is the logarithm of the partition coefficient (LogP) of the organic liquid between n-octanol and water, which can be expressed by the following equation: P¼

solubility of a solvent in n  octanol solubility of the solvent in water

488

Enzyme Bioreactors

Therefore, the higher the value of LogP, the more hydrophobic the solvent. Enzyme reactions capable of being conducted in organic solvents include hydrolysis of esters, synthesis of esters, and transesterification (acidolysis, alcoholysis, ester change, and aminolysis). Different forms of enzymes can be used in organic solvents, including free enzymes dissolved in glycerol and dimethyl sulfoxide (DMSO), enzyme complexes formed by combining with PEG or surfactants (forming complexes such as PEG enzymes, lipid-coated enzymes, surfactant-modified enzymes), enzyme powders dispersed in solvents or adsorbed onto solid carriers, enzymes enclosed into reversed micelles, enzymes dispersed in an aqueous phase within porous carriers, enzymes encapsulated in hydrophobic gels, and enzyme-containing whole cells. In operating enzyme reactors for nonaqueous enzyme catalysis, special attention should be paid to the control of water activity for maintaining the enzymatic activity and stability in the organic solvents, as many enzyme reactions always lead to formation of water. For controlling the water content during enzymatic reactions, the reactors can be designed to remove water molecules from the system by methods such as distillation, flowing of dry gases, and pervaporation using a membrane. For example, pervaporation membrane has been incorporated in an enzyme reactor for ester change reaction by lipase, and removing the water formed during the reaction allows the reactor to be well operated at a stable and optimal reaction rate. As alternatives of organic–water solvent systems, aqueous two-phase systems have been used for many years in biotechnological applications, which can be formed by mixing two polymers, one polymer and one kosmotropic salt, or one chaotropic salt and one kosmotropic salt, at appropriate concentrations and at a particular temperature. The two phases are mostly composed of water and nonvolatile components, thus eliminating volatile organic compounds. Typical examples of aqueous two-phase systems include PEG-sodium carbonate or PEG with phosphates, citrates, or sulfates, and PEG-dextran. Unlike the organic solvent systems, aqueous two-phase systems for enzymes offer the opportunity to shift reaction equilibrium toward product formation by ensuring that the enzyme and the substrate partition into one phase whilst the product enters, and may be removed from, the other. Such a system offers some of the advantages of the immobilized enzyme process. The enzymes are largely retained and are stabilized by the presence of the polymers, yet the catalysis is performed in a homogeneous solution; so no diffusion limitations to mass transfer exist. Drawbacks of aqueous two-phase systems include the need to separate the product from the upper-phase polymers and the gradual loss of enzymes that enter the upper phase. Enzyme loss could be reduced, without introducing diffusion limitations, by linking them to hydrophilic polymers so as to form soluble complexes.8

2.33.5.3

Enzyme Reactors With Multienzyme Reactions

Unlike the relatively simple conversion by a single enzyme, many biochemical transformations require several enzymatic reactions with simultaneous cofactor regeneration.3 For example, the conversion of glucose to hydrogen requires thirteen enzymatic reactions, three of which involve cofactors. Clearly, it would be very difficult to mimic such a complex system by isolating and immobilizing all of the enzymes involved and devising appropriate cofactor regeneration schemes.3 At present enzyme reactors with multienzyme reactions are still under research, but they have very promising potentials: (1) multienzyme reactors would have very high conversion efficiency with the cooperation of different enzymes and cofactor regeneration systems; (2) they can be used in the construction of novel synthesis routes for new compounds; and (3) they can substitute microbial fermentation; smart and compact reactors can be used instead of traditional fermenters. The key issue for multienzyme reactors will be the development of the various enzymes with high productivity and low cost by molecular biotechnologies and advanced bioprocess technologies.

2.33.5.4

Microreactors

Please refer to the article on microreactors.

See Also: 2.29 Microbioreactors.

References 1. 2. 3. 4. 5. 6. 7. 8. 9.

http://www.molecular-plant-biotechnology.info/index.htm. Williams, J. A. Keys to Bioreactor Selections. Chem. Eng. Prog. 2002, 98, 34–41. Harvey, W. B.; Douglas, S. C. Biochemical Engineering, Marcel Dekker: Berkeley, CA, 1997. Lee, J. M. Biochemical Engineering, Prentice Hall: Englewood Cliffs, NJ, 1992. Shuler, M. L.; Kargi, F. Bioprocess Engineering, Prentice-Hall Inc: Englewood Cliffs, NJ, 1992. http://www.rpi.edu/dept/chem-eng/Biotech-Environ/IMMOB/typesofreactors.htm. http://en.wikipedia.org/wiki/Continuous_stirred-tank_reactor. Martin, C.; Christopher, B. Enzyme Technology, Cambridge University Press: London, UK, 1990. Zhang, C.; Xing, X.-H. Research Progress of Coenzyme Regeneration Systems. Chin. J. Biotechnol. 2004, 20 (6), 811–816.

2.34

Immobilized Cell Bioreactors

M Nemati, University of Saskatchewan, Saskatoon, SK, Canada C Webb, The University of Manchester, Manchester, United Kingdom © 2011 Elsevier B.V. All rights reserved. This is a reprint of M. Nemati, C. Webb, 2.24 - Immobilized Cell Bioreactors, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 331-346.

2.34.1 2.34.2 2.34.2.1 2.34.2.1.1 2.34.2.1.2 2.34.2.1.3 2.34.2.1.4 2.34.3 2.34.3.1 2.34.3.2 2.34.3.3 2.34.3.4 2.34.3.5 2.34.4 2.34.5 2.34.5.1 2.34.5.2 2.34.5.2.1 2.34.5.2.2 2.34.5.2.3 2.34.5.3 2.34.5.4 2.34.5.4.1 2.34.5.4.2 2.34.5.4.3 2.34.5.5 2.34.6 2.34.6.1 2.34.6.2 2.34.6.3 2.34.6.4 2.34.7 References

Introduction Immobilization of Microbial Cells Immobilization Techniques Attachment Entrapment Aggregation Containment Immobilized Cell Bioreactors: Configuration and Design Characteristics Stirred-Tank Bioreactors Fixed-Bed Bioreactors Fluidized-Bed Bioreactors Gas-Agitated Bioreactors Membrane Bioreactors Mass Transfer and Biokinetics in Immobilized Cell Bioreactors Merits of Immobilized Cell Bioreactors Biological Stability Improved Biomass Hold-Up Support Matrix Cell Characteristics Environmental Conditions Improved Mass Transfer in Bulk Liquid Product and Process Improvements Improved Yield Partitioning Effect Downstream Processing Cell Proximity and Reaction Selectivity Potential Drawbacks Mass Transfer Limitations Mechanical Problems Substrate Limitation Product Inhibition Concluding Remarks

490 490 490 491 492 493 493 493 494 495 496 497 498 498 501 501 501 502 502 502 502 502 502 502 502 502 503 503 503 503 503 503 504

Glossary Biocatalysis Use of enzymes or microbial cells to catalyze a reaction. Bioreactor A reactor in which microbial cells, cell extracts, or enzymes are used to carry out biological reactions. Bioremediation A treatment process that utilizes microorganisms to remove pollutants from air, water, and soil. Cell aggregate A body of loosely associated cells. Cell immobilization Restriction of viable cell mobility to a defined region in the medium or bioreactor. Cryptic growth Growth of viable cells on compounds, which are released into the medium as a result of cell lysis. Downstream processing The unit operations that follow the biotransformation in a fermenter or a bioreactor, aiming at recovery and purification. Viable cells Cells that are alive and able to develop normally.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00084-7

489

490

2.34.1

Immobilized Cell Bioreactors

Introduction

Immobilization of viable cells as an alternative for enzyme immobilization has been the subject of numerous studies for the past four decades. Advancements resulting from multidisciplinary research on the fundamentals and applied aspects of cell immobilization have led to the development of numerous applications in the areas of environmental remediation and pollution control, production of pharmaceuticals and biochemicals, bioprocessing, and biosensors; also, in recent years, a variety of beneficial applications in the fields of biomedical engineering and medicine have been developed. Cell immobilization techniques, which are classified broadly as passive (based on natural processes) and active (induced by physicochemical means) techniques, fall in the specific categories of attachment, aggregation, entrapment, and confinement. Regardless of the nature of the applied technique, cell immobilization localizes the microbial cells into a defined region in a way that their catalytic activity can be maintained and used repeatedly and, if needed, continuously.1 Furthermore, the ability to design and manufacture immobilized cell particles with the desired shape, size, and density allows one to employ these particles in the bioreactor as a discrete phase with hydrodynamic characteristics which are independent of the other existing phases, especially the liquid. This in turn opens up the possibility for utilizing a whole host of bioreactor designs and configurations with immobilized cells including the stirred-tank, fixed-bed, fluidized-bed, gasagitated, and membrane bioreactors. Moreover, the decoupling of the biomass residence time from that of the liquid phase allows the operation of the bioreactor at short residence times and high loading rates without any concern regarding the cell washout. In addition to flexibility in design and operating conditions, immobilized cell bioreactors enjoy prolonged stability, enhanced biomass hold-up, improved mass transfer in the bulk liquid, reaction selectivity, increased product yield, and simplified downstream processing. Intraparticle diffusional resistances, which could potentially lead to substrate limitation and product inhibition, and mechanical instability are some of the drawbacks attributed to these systems. As indicated earlier, applications of cell immobilization and immobilized cell bioreactors are numerous and diverse and have proved beneficial in environmental, industrial, and medical fields. The use of immobilized cells in the treatment of conventional wastewaters is one of the earliest applications in the field of environmental engineering and bioremediation. Progress in this field has expanded the utilization of the immobilized cell bioreactors in the treatment of specific waste streams, either liquid or gaseous, contaminated with sulfate, sulfide, metals, and various organic compounds such as aromatic, polycyclic aromatic, and heavy petroleum hydrocarbons, specially the tailings produced during the processing of the oil sands. Immobilized cell systems have been extensively used in the production of a variety of alcoholic and nonalcoholic beverages, bioprocessing in the dairy and meat industries, and bioconversions aiming at the production of specialties such as sake, soy sauce, mead, amino acids, and organic acids, to name a few.2 Production of biochemicals including ethanol fuel, pharmaceuticals such as penicillin, cephalosporin C, and other antibiotics, as well as viral vaccines, antibodies, interferons, enzymes, insecticides, hormones, growth factors, and plant cell metabolites is another area in which immobilized cell bioreactors have beneficial applications. As far as biomedical engineering and medical applications are concerned, immobilized cell systems have played a central role in the field of tissue engineering, especially in the development of artificial organs and in the effective and long-term delivery of therapeutic agents.2 Various applications of immobilized cell systems are presented schematically in Figure 1. The varied and diverse nature of these applications implicates the need for the design and utilization of a variety of bioreactor configurations suitable for the intended purpose. This necessity has led to the modification of conventional designs and the development of new configurations to meet the criteria required for each application. Research in the field of cell immobilization technology and immobilized cell bioreactors has been extensive and as a result a vast amount of information exists on both the fundamentals and the applied aspects. The space devoted to this article serves only to provide the reader with an overview of the primary topics such as immobilization techniques, bioreactor design and configurations, mass transfer and biokinetics, merits of immobilized cell bioreactors, and associated drawbacks. The interested readers are referred to many excellent publications available in the public domain, some of which are listed at the end of this article.

2.34.2

Immobilization of Microbial Cells

The hydrodynamic behavior of a microbial population which is dispersed as individual cells within the continuum of a liquid is governed by the behavior of that liquid. This imposes a severe limitation on the biological performance of the systems in which microbial cells are utilized for a particular purpose, whether it is production of a biochemical or treatment of a hazardous waste. This limitation could lead to reduced productivity and efficiency and is caused by the removal of the cells from the system when the main stream is removed. This constraint can be circumvented if somehow the hydrodynamic behavior of the cells is decoupled from that of the liquid, for instance by immobilization of the cells. In that context cell immobilization can be simply defined as the “restriction of cell mobility within a defined space”3 or, as elaborated by Webb and Dervakos, as the “confinement or localization of viable microbial cells to a certain defined region of space in such a way as to exhibit hydrodynamic characteristics which differ from those of surrounding environments.“4

2.34.2.1

Immobilization Techniques

Over the years, a variety of techniques have been developed and used for immobilization of microbial cells, with many having been taken directly from the enzyme immobilization technology. Cell immobilization techniques can be broadly classified as passive and

Immobilized Cell Bioreactors

491

Applications of immobilized cell technology Tissue engineering

Bioartificial organs

Blood vessels and bone

Cartilage and myocardium

Stem cells

Treatment of diseases Gene therapy

Protein therapeutic delivery

Enzyme and cell therapies

Therapy of malignant gliomas

Food and beverages Beer, wine, and cider

Dairy industry and meat processing

Bioflavoring of foods and beverages

Food bioconversions

Production of biochemicals Ethanol

Biopharmaceuticals

Biologics from animal cells

Plant cells

Environmental bioengineering Treatment of wastewaters

Air pollution control

Soil bioremediation

Figure 1 Medical, industrial, and environmental applications of the immobilized cell systems. Based on the information extracted from Nedovic V and Willaert R (eds.) (2005). Focus on Biotechnology, Vol. 8B: Applications of Cell Immobilisation Biotechnology. Dordrecht: Springer.

active techniques. In the passive approach, the process of immobilization occurs naturally because of the growth of the microbial cells on the internal or external surfaces of a solid support (carrier matrix) resulting in the formation of biological films.3,4 Formation of biofilms is common in nature and has been utilized industrially in the biological treatment of wastewaters, production of vinegar and mold fermentation, as well as laboratory-scale experimental systems used for research purposes. In the active techniques, cell immobilization is induced by physical or chemical means. In an overarching approach which has been adapted widely, immobilization techniques are categorized according to the physical processes involved, regardless of the passive or active nature of the process. These include attachment, entrapment, aggregation, and containment as represented schematically in Figure 2.

2.34.2.1.1

Attachment

Attachment is the term used for any form of immobilization in which cells form a bind to the surface of a solid support, either naturally (adsorption and adhesion) or as a result of the treatment of the surface of the cells and support material (covalent binding). Natural adhesion (adsorption) of cells to surfaces is a widespread phenomenon and provides a simple and mild method for immobilization of the cells.4,5 The type of involved forces depends mainly on the surface properties of the support and cells. With positively charged surfaces such as ion exchange resins or gelatin, electrostatic forces are dominant, while with negatively charged surfaces attachment occurs through covalent binding or hydrogen bonding.3 Support materials with large surface area such as sand or porous materials like activated carbon are the preferred choice as they usually allow high cell loadings, although intraparticle diffusional resistances could become a major impediment at these high cell loadings. The other common support matrices include porous glass, alumina, ceramics, gelatin, chitosan, wood chips, ion exchange resins, and Sepharose.3,6 The major advantage that adsorption offers is the direct contact between the cells and the medium containing the nutrients. However, direct contact of the cells with the surrounding and their exposure to the existing shear forces in conjunction with the weak binding forces involved in adsorption could result in the detachment of the cells from the support. Cross-linking of the cells following the adsorption has been suggested as a means to strengthen the attachment. Covalent binding, as the name implies, involves the formation of covalent bonds between the cells and the support surface. The treatment of the binding surfaces with coupling agents such as glutaraldehyde (supports with
NH2 group), carbodiimide (supports with
COOH group), and CNBr (supports with
OH group) is an essential step in promoting covalent binding, as the functional groups on the cell and support surfaces in general are not suitable for covalent binding.3 The toxic nature of these binding promoters is a drawback, and as a result, a number of support materials with a suitable functional group have been developed and used for cell immobilization. These include carboxymethyl cellulose plus carbodiimide, aldehyde-containing carriers, amine, epoxy, Zr(IV) oxide, Ti(IV) oxide, and cellulose plus cyanuric.3

492

Immobilized Cell Bioreactors

Flat support

Spherical support

Attachment

Matrix formed in situ

Preformed matrix

Entrapment Figure 2

2.34.2.1.2

Natural

Cross-linked Aggregation

Membrane formed in situ

Preformed membrane

Confinement

Schematic representation of various immobilization techniques.

Entrapment

Immobilization of the cells by entrapment is achieved using preformed support materials with micro- or macroporous structures. Entrapment of the individual cells and their natural growth lead to colonization of the support material. Entrapment within the preformed structure could occur at the microscopic level using microporous particles such as ceramic, sintered glass, or silica gel or at the macro level through utilization of particles with relatively large pores such as stainless steel mesh or reticulated foam biomass support particles.3,6 To compare with their attached counterparts, entrapped cells are protected from the shear forces existing in the bioreactor but by no means are confined within the structure and the liquid medium is unlikely to remain cell free. The existing shear forces, however, act on the cells that grow beyond the boundaries of the support and in conjunction with particle– particle attrition control the biomass hold-up in the support material. Entrapment of the cells within the porous structures which are formed in situ around the cells is achieved by mixing the microbial cells, in the form of a slurry or paste, with a suitable compound, which is then gelled or polymerized to form the porous structure containing the cells. Entrapment in polymeric network formed in situ around the cells is indeed the simulation of a natural phenomenon under controlled conditions. Certain types of microorganisms with the ability to excrete extracellular polysaccharides normally exist attached to surfaces or even in the form of large flocs (e.g., activated sludge), while entrapped within the slime or gel secreted by themselves or the other members of the population. The diversity of the materials and methods which could be used for the construction of such polymeric networks allows the design of a system with maximum retention of the cells and effective transport of the substrate and products.5 However, the success of such a system depends on the selection of a polymeric precursor with the least toxicity and performing the polymerization process under mild environmental conditions (appropriate pH, temperature, and solvent) so that cell viability is preserved. The common polymeric matrices used for the entrapment of cells include polyacrylamide, epoxy resins, agar, alginates, k-carrageenan, carboxymethyl cellulose, and chitosan,5,7–9 which are synthesized using gelation, precipitation, ion exchange gelation, and polycondensation as described briefly in the following sections.3 2.34.2.1.2.1 Gelation This is a simple and gentle method which involves phase transition in a mixture of a polymer solution and cells because of a shift in temperature.5 To achieve this, harvested microbial cells are first mixed with the polymer solution (e.g., agar or agarose) at elevated temperature. The prepared mixture is then passed through a template while undergoing cooling, which results in the formation of a solid three-dimensional network in the form of cell-containing beads. The use of a spherical hard-core support covered with a layer of cell-containing gel could overcome the soft and fragile nature of the gel beads and provide the required mechanical strength.3 Although biomass hold-up of individual particles could be lower than a bead entirely made of polymer, the overall productivity per particle may not be that different, as diffusional limitations impeding the activity of the cells in the inner core of a polymeric bead are eliminated in a hard-core support covered by a layer of gel.

Immobilized Cell Bioreactors

493

2.34.2.1.2.2 Precipitation In this method the designated polymer, typically polystyrene, cellulose triacetate, or collagen, is dissolved in an organic solvent such as ethanol or acetone. The harvested cells are then added to this solution, and through manipulation of pH or addition of a second solvent the precipitation of the polymer is achieved. Direct contact of the solvents with cells may cause deactivation or death and should be minimized. 2.34.2.1.2.3 Ion Exchange Gelation The reaction of water-soluble polyelectrolytes (ionotropic gels) with a counterion or a cross-linking agent results in the formation of a solid polymeric network which could maintain the cells. The suitable polyelectrolytes, which in general are inexpensive and do not impose a severe toxic effect, include those with carboxyl or sulfonyl groups (alginate, pectin, carrageenan) or those with amino groups such as chitosan obtained from chitin by deacetylation.5 A variety of counterions, both anionic and cationic, could be used as cross-linking agents. The most common example is the immobilization of microbial cell in calcium alginate, which is formed as a result of the addition of Ca2þ ion (in the form of CaCl2 solution) to a suspension of cells in sodium alginate. One of the major problems with the immobilization of cells in alginate is the destabilization of calcium alginate in the presence of chelating agents such as phosphate, citrate, and ethylene diamine tetracetic acid (EDTA). 2.34.2.1.2.4 Polycondensation A stable polymeric network with chemical resistance and mechanical strength such as those in epoxy resins can be formed by polycondensation reactions.3,5 Epoxy resins, polyurethane, silica gel, gelatin glutaraldehyde, albumin glutaraldehyde, and collagen glutaraldehyde are typical examples of polymers obtained by polycondensation.3 The elaborate chemistry of polymerization, high temperature and low or high pH required for these reactions, and toxicity of the functional groups have been the major impediments in the widespread application of polycondensation for cell immobilization.

2.34.2.1.3

Aggregation

The aggregation and flocculation of microbial cells resulting in formation of large flocs is a natural phenomenon which may be observed in many types of microorganisms at some stage of their life cycle. Polymers at the cell surface and extracellular polymeric substances are major factors in the natural development of microbial flocs and their integrity.3,5 Cross-linking agents such as glutaraldehyde could be used to promote the formation of large agglomerates in species which are not normally regarded as flocculent. Formation of large flocs as a result of aggregation either occurs naturally or is promoted artificially; it is considered a form of immobilization in the sense that these large flocs can be maintained in the continuously operated bioreactors. Formation of large aggregates also facilitates the separation of cells from the effluent, which can then be returned to the bioreactor. Formation of microbial flocs with good settling properties is instrumental in the treatment of municipal and industrial wastewaters by the activated sludge process.

2.34.2.1.4

Containment

Immobilization of the microbial cell could also be achieved by containment of the cell behind a barrier. The suitable barrier which should allow the flow of the substrate and metabolic products, while maintaining the cells, could be preformed such as semi-permeable membranes used for micro- and ultrafiltrations or could be formed in situ around the cells. Encapsulation and microencapsulation are typical examples of immobilization by in situ formation of a barrier around the cells.8,9 Microencapsules are hollow spherical particles with boundaries made of a semi-permeable membrane, which maintains the cells within their hollow volume and allows the transport of the products and nutrients from and into the capsule. A variety of polymeric compounds, including nylon, collodion, polystyrene, acrylate, polylysine-alginate hydrogel, cellulose acetate, ethyl cellulose, and polyester, can be used as the membrane in encapsulation.3 To compare with polymeric beads with entrapped cells, higher biomass hold-ups can be achieved by microencapsulation. Furthermore, intraparticle diffusional resistances in a microcapsule are lower because cells are suspended in a liquid within a capsule as opposed to a polymeric network. The interface of two immiscible liquids could also serve as a barrier for the containment of cells. In this case harvested cells or a suspension of cells are emulsified in an organic liquid and resuspended as a droplet in an aqueous phase.3 Containment of cells behind the preformed semipermeable membrane is another approach for immobilization of the cells and the underlying feature of the membrane bioreactors.

2.34.3

Immobilized Cell Bioreactors: Configuration and Design Characteristics

The ability of maintaining the immobilized cells in the bioreactor as a discrete phase with a hydrodynamic behavior independent of that of the liquid and gas phases (i.e., potential for decoupling of biomass and hydraulic residence times) in conjunction with the flexibility regarding the shape, size, and arrangements of the immobilized cell particles allow the design and utilization of a whole host of bioreactor configurations. Furthermore, immobilized cell bioreactors can be operated in various modes, including batch, semi-continuous, and continuous, and with a range of flow regimes, such as well-mixed, dispersed plug, and ideal plug flows. The choice of the bioreactor is influenced by a number of factors such as the immobilization method and characteristics of the

494

Immobilized Cell Bioreactors

particles (e.g., shape, size, density, and mechanical strength), reaction kinetics, and economic considerations.3 In general a properly designed immobilized cell bioreactor should meet the following criteria: (1) the level of existing shear forces should be sufficiently low so that the integrity of the particles is preserved; (2) the bioreactor design should allow maximum particle hold-up, while an efficient level of mass and heat transfers is maintained; and (3) the bioreactor of choice should accommodate a flow regime suited to the kinetics of the reaction of interest.3 The latter is an important consideration where the inhibitory effects of substrate or product are significant. A mixed flow is suitable when the reaction is inhibited by the substrate, as the incoming substrate is diluted continually with the content of the bioreactor. For those reactions which suffer from the inhibitory effects of the product a plug flow regime would be satisfactory, as high concentrations of product are experienced only toward the end section of the bioreactor. Nonetheless, the bioreactor configurations used with immobilized cells are numerous and in general fall in one of the categories of stirred-tank (mechanical or gas agitation), fixed-bed, fluidized-bed, and membrane bioreactors. The following sections provide an overview of each group and their specific characteristics.

2.34.3.1

Stirred-Tank Bioreactors

Stirred-tank bioreactors are advantageous when cell activity is impacted by the inhibitory effects of substrate at high concentrations but they are not the proper choice when the product inhibits the reaction. One of the major problems associated with stirred-tank bioreactors, especially those with conventional impellers such as flat blade turbines or propeller, is the strong shear forces exerted on the particles. These forces could severely damage the particles, especially those which are gel based.3,4 The detrimental effect of the shear field could be circumvented by enclosing the immobilized cells within a porous mesh as shown in Figure 3. The use of helical ribbon screws or anchor impellers could also reduce this damaging effect, owing to their gentler mixing when compared with flat blade turbines or propellers.10 Injection and circulation of air or a gaseous stream is another approach to achieve efficient mixing and mass transfer in the bioreactor under a relatively mild shear regime. This is the main feature of the gas-agitated (air-lift) bioreactors, which will be discussed in a separate section. Rotating biological contactors (RBCs), which are commonly used for the treatment of municipal wastewaters and acid mine drainage, also fall in the category of well-mixed bioreactors. As simply shown in Figure 3, RBC consists of a large number of circular disks, which are attached on a rotating shaft and partially submerged in the

Feed

Effluent

Feed

Air (optional) Stirred-tank bioreactor with anchor impeller

Air (optional) Stirred-tank bioreactor with cells in a porous mesh

Effluent

Feed

Rotating biological contactor Figure 3

Stirred-tank immobilized cell bioreactors. Solid circles: immobilized cells; open circles: gas bubbles.

Immobilized Cell Bioreactors

495

waste stream under treatment. The disks, which are usually made of polystyrene or polyethylene, provide a large surface area for the passive immobilization of cells and the formation of biofilms. Rotation of the disks exposes the cells to the air and the waste stream alternately and provides the cells with nutrients (usually the contaminating compounds) and oxygen. The configuration of RBC in general provides an efficient level of mass transfer and mixing without imposing strong shear forces on the cells. Moreover, frequent exposure of the cells to air usually fulfills the oxygen demands of the process and eliminates the additional costs associated with aeration in a conventional stirred-tank bioreactor. However, the open nature of RBCs limits their applications to those processes in which maintenance of the aseptic conditions is not a necessity. It should be pointed out that RBC could also be used in anaerobic processes such as denitrification or sulfate reduction. However, in these cases, disks are fully submerged in the liquid inside a closed vessel.

2.34.3.2

Fixed-Bed Bioreactors

Packed-bed and trickle-bed configurations, as shown in Figure 4, both fall under the category of fixed-bed bioreactors and are used commonly with the immobilized cells. Packed-bed bioreactors operated under the plug flow regime (once-through basis) are advantageous when the reaction product imposes a strong inhibitory effect. They offer enhanced reaction rates on account of high substrate concentrations which maintain over a considerable length of the bioreactor as opposed to a stirred-tank bioreactor, in which the incoming substrate is diluted instantly by the contents of the bioreactor. Poor heat and mass transfers because of low fluid velocity and absence of mixing are two of the main drawbacks of the packed-bed bioreactors.3 Lack of efficient contact between the gas and liquid phases, which results in poor mass transfer from the gas to the liquid and vice versa, is a serious impediment in three-phase operations. Slow release of gaseous metabolites such as CO2 or NH3 causes the accumulation of gas in the form of stagnant slugs, and channeling of the flow.3 Partial recycling of the liquid effluent could alleviate the accumulation of the gas and improve the extent of heat and mass transfers. This would also serve to improve the performance of the system when substrate inhibition decreases the activity of the cells. Apart from influencing the internal diffusional resistances, the size of the immobilized cell particles is a determining factor on the extent of the pressure drop in packed-bed bioreactors. Using immobilized cell particles of a uniform size increases the bed voidage and decreases the pressure drop across the bed. Compression of the particles as a result of the bed static weight and pressure from the flowing fluid is significant when soft and flexible particles (e.g., gels or reticulated foams) are used. The compressive pressure decreases the bed voidage, leading to higher pressure drops and increase of pumping cost.3 With growth-associated processes and in the absence of attrition and shear forces, excessive accumulation of the biomass could lead to channeling and eventual plugging of the bioreactor. In trickle-bed bioreactors, which are generally used for three-phase operations, the liquid flows from the top down through the bioreactor and the gas phase flows countercurrent from the lower part of the bioreactor to the top.11 Contrary to the packed-bed bioreactors in which immobilized cell particles are submerged in the liquid phase, in a trickle-bed bioreactor the liquid phase passes over the particles usually as a thin film. Partial recycling of the effluent liquid provides uniform distribution of the liquid and improves the performance of the systems. Effective mass transfer between the gas and liquid is one of the main features of the trickle-bed bioreactors. Trickling biofilters operating on the same basis as that of trickle-bed bioreactors but with natural aeration are used extensively in the treatment of gaseous streams containing odorous and hazardous compounds such as H2S and NH3 and

Air

Air

Effluent

Feed

Feed

Feed

Effluent

Air Effluent

Air Packed-bed bioreactor Figure 4

Air Packed-bed bioreactor with external aeration

Trickle-bed bioreactor

Fixed-bed bioreactors used with immobilized cells. Dashed line indicates the optional possibility for aeration.

496

Immobilized Cell Bioreactors

for the treatment of domestic wastewater, either as pretreatment prior to the activated sludge process or as a polishing step for the effluent from the aeration basin of the activated sludge process.

2.34.3.3

Fluidized-Bed Bioreactors

In fluidized-bed bioreactors (Figure 5), which are suitable for two- and three-phase operations (solid–liquid and solid–liquid–gas, respectively), liquid or a mixture of gas and liquid is introduced into the bottom of the bioreactor. The upward flow of the injected fluid results in suspension of the immobilized cell particles and expansion of the bed.4,5 The fluid velocity resulting in the optimum operation usually falls between two limiting levels: the minimum fluidization velocity, which is required to maintain the particles in suspension, and the maximum allowable velocity, above which suspended particles are entrained by the moving fluid and leave the system.5,12 In the majority of cases the proper control and provision of the adequate velocity is achieved by partial recycling of the liquid phase. This allows external aeration of the liquid and simplifies the operation to a two-phase regime. Furthermore, recycling facilitates the control of the pH, and removal of the excess and detached biomass.5 It should be pointed out that the adequate velocity depends significantly on the size and density of the immobilized particles, which in turn are influenced by the growth of the cells and the dominant hydrodynamic conditions. However, stable operation could be expected with heavy support matrices and when measures are introduced to ensure relatively stable biomass hold-ups. Particle size and density, especially the density difference between the liquid and the particles, could also affect the extent of mass transfer, pressure drop, and pumping costs.4,12 As stated by Webb and Dervakos,4 “Fluidized-bed bioreactors combine some of the advantages of stirred-tank and packed-bed bioreactors and a few of their drawbacks.“ One of the main attractive features of the fluidized-bed bioreactors is the effective level of mixing and mass transfer, particularly the improved gas–liquid contacts which result in higher oxygen mass transfer rates. Furthermore, compared with the packed-bed counterparts, the release and removal of gaseous metabolites are facilitated in fluidized beds. This is an advantageous feature as it prevents the formation of gas slugs and flooding, especially in applications in which large quantities of gas are formed and should be recovered. Production of biogas during the process of anaerobic digestion is a typical example of such situations. The biomass hold-up per unit volume of the fluidized-bed bioreactor is lower than that in the packed-bed bioreactors but the overall performance in most cases is better because of favorable operating and hydrodynamic conditions. Finally, in fluidized-bed bioreactors, clogging and dead zones are minimized. This is due to existing shear fields, which could also prevent the biomass overgrowth. Tapered fluidized-bed, also referred to as spouted-bed, bioreactor is a modified form of a fluidized-bed bioreactor, in which the lower part of the bioreactor has a conical shape (tapered diameter). The higher fluid velocities in the narrow zone create intensive mixing and strong shear forces in this region with the dual impacts of increased mass transfer and steady control of biomass without the risk of particle entrainment and their escape from the top of the bioreactor.4,12,13 Upflow sludge blanket (USB), its anaerobic counterpart upflow anaerobic sludge blanket (UASB), and expanded granular sludge blanket (EGSB) bioreactors represent other fluidized-bed configurations with extensive applications in the treatment of waste streams.10 In USB and UASB, dense sludge granules are fluidized in the upward direction due to upflow of the wastewater. As the treatment of wastewater proceeds and organic

Effluent

Effluent

Feed

Feed

Air (Gas)

Air (Gas)

Feed Effluent Air (Gas) Fluidized-bed bioreactor

Air (Gas) Tapered fluidized-bed bioreactor (spouted-bed bioreactor)

Inverse fluidized-bed bioreactor

Figure 5 Fluidized-bed immobilized cell bioreactors. Dashed lines indicate the optional possibility for aeration. In case of anaerobic processes, partial recycling of the produced gas induces the fluidization of particle. Solid circles: immobilized cells; open circles: gas bubbles.

Immobilized Cell Bioreactors

497

content of the wastewater is converted to CO2 or biogas, gas-borne granules are transported to the top of the bioreactor, where degasification occurs. The disengagement of the gas increases the density of the particles and results in the downflow motion of the sludge toward the lower part of the bioreactor. The established particle-circulating pattern improves the extent of mixing when compared with a conventional fluidized-bed bioreactor.10 The EGSB bioreactor design is based on the same concept as that of the USB. However, the addition of a specially designed three-phase separator allows efficient control of the sludge holdup in the bioreactor and prevents accumulation of the excessive sludge, a major drawback in USB and UASB bioreactors. The problem of biomass overgrowth could also be alleviated by operating the bioreactor in the inverse fluidization regime, in which liquid is introduced from the top and flows downward, while gas is introduced into the lower part of the bioreactor and flows in the upward direction. This countercurrent flow regime results in extensive mixing and turbulence and controls the bed height and biomass hold-up.10

2.34.3.4

Gas-Agitated Bioreactors

Injection of a gas, especially air, is one of the convenient ways to circulate the contents of a bioreactor, either internally through a draft tube or externally by an extended loop. It also assists in achieving good mixing and efficient mass transfer, almost comparable to that of the stirred-tank bioreactors but in an environment with much milder and evenly distributed shear forces.4,10 Such gas-agitated (air-lift) bioreactors (Figure 6) are attractive in large-scale bioprocesses because of their simplicity, absence of moving parts, ease of operation, and lower power consumptions.4 Air-lift bioreactors, regardless of the circulation configuration (either internal or external loops), consist of a base, a riser, a down-comer, and a headspace on the top.10 Gas is introduced at the bottom of the bioreactor, flows upward, and disengages from the liquid in the headspace. The different gas hold-ups which are established in the riser and down-comer create different bulk densities in these two regions, which in turn cause the circulation of both liquid and solid phases. In aerobic processes the air required for the process induces the circulation, while in anaerobic processes partial recycling of the produced gas or an inert gas could serve this purpose. The top section accommodates disengagement of the gas from the particles and prevents the carryover of the immobilized particles. The sloughed biomass and any freely suspended cells, however, are washed out with the effluent.10 The flow regime in an internal-loop air-lift bioreactor depends on the liquid circulation rate and gas flow rate. Under a low gas velocity and low circulation rate of the liquid, all the gas is disengaged at the top of the bioreactor, while with high liquid circulation rates gas bubbles are partially entrained in the down-comer region and circulation of the gas is established. The latter flow regime is advantageous in aerobic processes as it provides higher gas hold-ups and efficient mass transfer between the gas and liquid.10,13 At moderate liquid circulation rates, some gas bubbles are present in the down-comer section of the bioreactor but gas circulation is not significant. The liquid flow in the riser and down-comer of an air-lift bioreactor is usually approximated as plug flow with axial dispersion but at the top a well-mixed regime is dominant. Operation of air-lift bioreactors in the inverse mode (downward flow of the liquid) has been suggested as a suitable approach when immobilized cell particles have a density lower than that of the liquid phase. The gas phase is introduced in the riser which causes the circulation of the liquid, which in turn maintains the immobilized cell particles in suspension in the down-comer. In an inverse fluidized-bed air-lift bioreactor, the direct contact of the gas bubbles and particles is eliminated. This is important for sensitive cells such as animal cells, as direct contact of the cells with gas bubbles and bursting of the bubbles could damage the cells. In

Air (Gas)

Air (Gas)

Effluent

Feed

Air (Gas)

Feed

Air (Gas) Air-lift bioreactor with draft tube

Effluent

Effluent

Feed

Air (Gas) External-loop air-lift bioreactor

Air (Gas) Circulating-bed bioreactor (off-center aeration)

Figure 6 Gas-agitated (air-lift) immobilized cell bioreactors. In case of anaerobic processes, partial recycling of the produced gas induces the fluidization of particle. Solid circles: immobilized cells; open circles: gas bubbles.

498

Immobilized Cell Bioreactors

a circulating-bed bioreactor, both gas and liquid flow in the upward direction. However, gas is introduced only into a portion of the bioreactor cross-sectional area and in an off-center fashion. This induces the circulation of the liquid and the particles in the bioreactor and improves the extent of mixing and mass transfer.4,13

2.34.3.5

Membrane Bioreactors

Membrane bioreactors, which are configured either as flat sheet or hollow fiber modules, are in general complex in structure and design, and more expensive (due to high costs of the membrane material) when compared with conventional bioreactors. However, they offer simultaneous bioconversion and separation of the product usually in form of a concentrated stream.10 Elimination of the costly purification steps, especially in case of high-value biomolecules, further justifies the use of membrane bioreactors. A membrane bioreactor may be the preferred choice in the treatment of specific waste streams containing toxic compounds or heavy metals as it prevents the direct exposure of the cells to these harmful compounds. Biotreatment of acid mine drainage (AMD) is a typical example of such situations. In membrane bioreactors cells may be immobilized on the membrane in the form of biofilms or within the membrane, or cells may be separated from the bioreaction medium by the membrane and maintained in a separate compartment. Regardless of the immobilization mode, the membrane protects the cells from the existing shear forces and bubble bursting, both of which are detrimental to mammalian and plant cells. A variety of commercial membranes are available for use in membrane bioreactors. Characteristics such as the pore size, structure, and material of construction are important in selection of the membrane for a particular application. On the basis of the structural characteristics, membranes are classified as isotropic (with a homogeneous composition) and anisotropic (consisting of a thin layer of membrane supported by a dense layer of porous understructure).3,10 The most widely used type of membrane bioreactors is the hollow fiber membrane bioreactor, as shown schematically in Figure 7. As stated by Shuler,3 “A hollow fiber bioreactor is a mass transfer analog of a shell and tube heat exchanger in which the tubes are made of semi-permeable membranes.“ In the majority of applications, cells are maintained on the shell side and grow in place, while substrate containing the stream flows through the tube bundle. The semi-permeable membrane allows the diffusion of substrate and nutrients to the shell side, where they are used by the cells. The metabolic products then diffuse back into the tube-side liquid, which essentially remains cell free. The growth of the cells and formation of the biological film on the membrane, which could create significant resistance to mass transfer and even lead to complete plugging of the membrane, is one of the main drawbacks of these systems. The flat module membrane bioreactors have a simpler configuration, which allows easy access to the compartments for the cleaning of the membrane and, where necessary, its replacement. Hollow fiber counterparts, however, provide a higher surface-to-volume ratio and eliminate the need for a rigid membrane support.

2.34.4

Mass Transfer and Biokinetics in Immobilized Cell Bioreactors

The performance of an immobilized cell bioreactor depends greatly on the intrinsic biokinetics, and as such the design and proper operation of these systems require a thorough understanding of the biokinetics.11 The interaction between the enzymatic system of the cells and substrate is not the only governing parameter and other factors such as external and internal mass transfers influence the biokinetics and play instrumental roles in the performance of an immobilized cell bioreactor.3,11,14 As shown in Figure 8, in an immobilized cell bioreactor, substrate molecules are first transferred from the bulk liquid to the interface of the bulk liquid and the stagnant film which is formed on the surface of the biofilm. This transport occurs mainly by convection with little or negligible mass transfer resistance, mainly due to effective mixing in the bulk liquid. The substrate molecule must then diffuse through the film to reach the reaction sites on the surface of the biofilm. The diffusion process through the liquid film is referred to as external mass transfer.14 With many types of immobilized cell systems (entrapment, encapsulation, thick biofilms, or bioflocs), further reaction sites are available within the porous carrier matrix or biofilm. Substrate transport from the external surface of the particle to the inner parts and through the porous and tortuous pathways is referred to as internal mass

Feed

Effluent (product)

Medium (in)

Medium (out)

Air Figure 7

Hollow fiber membrane bioreactor. Dashed lines indicate the optional possibility for aeration.

Immobilized Cell Bioreactors

Liquid film

Immobilized cells

Support Immobilized Liquid film cells

499

Bulk liquid S = Sb

Bulk liquid S = Sb

R L z=0 S = Sf

r=0 S = Sf

z=L S = Si

Cells immobilized inside a porous matrix

Cells immobilized on a flat surface Figure 8

r=R S = Si

Schematic representation of immobilized cell particles and corresponding substrate concentrations.

transfer.11,14 The presence and significance of diffusional resistances against the transport of substrate is usually assessed by the relative rates of mass transfer and bioreaction and is represented by the dimensionless Damköhler number3,11,14: Da ¼

maximum bioreation rate rmax ¼
De  maximum mass transfer rate d Sb

(1)

where Da is the dimensionless Damköhler number, rmax is the maximum bioreaction rate, De is the effective diffusivity of the substrate, d is the thickness of the diffusion layer, and Sb represents the substrate concentration in the bulk liquid. With Da > 1, the maximum mass transfer rate is much smaller than the maximum bioreaction rate or observed reaction rate (referred to as effective or global reaction rate) and the system is considered diffusion limited.11 For Da values close to 1, biokinetics and mass transfer occur at comparable rates and both influence the performance of the system. When microbial cells are immobilized as thick biofilms or when cells are entrapped or encapsulated within the porous structure of a matrix, diffusion and bioreaction occur simultaneously. The mass transfer and biokinetic characteristics of the immobilized cell systems have been discussed in details in a number of publications.1,3,11,14 The analysis of simultaneous mass transfer and reactions in immobilized cell systems is usually achieved by assuming a quasi-steady state for the system and a uniform physiological state for the cells within the matrix or in the biofilm. Although the biofilm thickness or size of the cell aggregates increases as the bioreaction proceeds, these changes over short periods are small, especially when compared with the bioreaction rates, which makes the assumption of a quasi-steady state plausible.3 Writing a steady-state material balance for the limiting substrate within the immobilized cell layer (Figure 8, left panel) results in a differential equation with the corresponding boundary conditions as given below: De

d2 S 1 mm S X ¼ dz2 YX Ks þ S z¼0

(2)

dS ¼0 dz

z ¼ LS ¼ Si where De is the effective diffusivity, z represents the distance, YX is the cell yield coefficient, mm is the maximum specific growth rate, KS is the saturation constant, L is the biofilm’s thickness, and X represents the biomass concentration. Assuming that diffusional resistance through the stagnant film is negligible (i.e., sufficient mixing in the liquid phase), Si z Sb. The maximum reaction rate can also be defined as3 rmax ¼

mm X YX

(3)

500

Immobilized Cell Bioreactors

Eq. (2) can be written in dimensionless form and solved numerically with the following boundary conditions: d2 S 42 S  ¼ 2 dz 1 þ bS z ¼ 0

(4)

dS ¼0 dz

z ¼ 1 S ¼ 1 where S ¼

S Sb

(5)

z ¼

z L

(6)

Sb kS

(7)

b¼

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffi mm X rmax ¼L 4¼L YX De KS De KS

(8)

The maximum substrate mass flux in the absence and presence of diffusional limitations are given by Eqs. (9) and (10), respectively:   rmax Sb N¼ L (9) KS þ Sb   rmax Sb L N¼h KS þ Sb

(10)

where h is the effectiveness factor and defined as the ratio of the substrate consumption rate in the presence of diffusional resistance to that in the absence of such resistance.3 The effectiveness factor is calculated from the following equations: h¼1
h¼

 tanh 4  u
1 u1 4 tanh u

 1 tanh 4  u

1 u 4 tanh u

u>1

(11) (12)

where u is the modified Thiele modulus given by the following equation: 4b u ¼ pffiffiffi ½b
lnð1 þ bÞ
1=2 2ð1 þ bÞ

(13)

As can be deducted from the above equations, the effectiveness factor (h) is a function of 4 and b. In the systems with low values of 4 (i.e., 4 < 1), diffusional limitations are negligible, and in the absence of diffusional limitations the effectiveness factor approaches unity.3 In many cases microbial cells are immobilized within the void spaces of spherical porous matrices or attached to the external surface of a nonporous spherical support. A spherical immobilized cell particle (Figure 8, right panel) can be represented and analyzed by the following dimensionless differential equation and boundary conditions: d2 S 2 dS j2 S þ  ¼ 2 r dr dr 1 þ bS r ¼ 0

(14)

dS ¼0 dr 

r  ¼ 1 S ¼ 1 where S ¼

S Sb

(15)

Immobilized Cell Bioreactors r R

(16)

Sb KS

(17)

r ¼ b¼

501

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffi mm X rmax 4¼R ¼R YX De KS De KS

(18)

For a spherical system with a first-order reaction rate (i.e., low substrate concentration), the effectiveness factor is given by the following equation:   1 1 1
(19) h¼ 4 tanh 3 4 34

2.34.5

Merits of Immobilized Cell Bioreactors

Extensive research on the immobilized cell systems and numerous practical applications in which immobilized cell bioreactors have been utilized identify many advantageous characteristics for these systems when compared with the bioreactors employing freely suspended cells. One of the most notable differences is the decoupling of the cell retention time from that of the hydraulic residence time, which allows the operation of the immobilized cell bioreactors at short residence times or high loading rates without any concern for the cell washout. This together with high biomass hold-ups results in high efficiency and productivity in these bioreactors. Furthermore, the application of immobilized cell systems eliminates the need for isolation and purification of the enzyme and makes it possible to conduct sequential or parallel reactions requiring multiple enzymes in the same bioreactor.3,5 Creation of favorable microenvironmental conditions, protection of the cells against damage by the existing shear forces, and improvement in the biological stability are some of the other advantages associated with immobilized cell bioreactors. The following subsections serve to provide a brief overview of these positive attributes and potential mechanisms involved.

2.34.5.1

Biological Stability

There is a general agreement that the operational stability of the immobilized cells and the bioreactors in which they are employed is higher than that of freely suspended cells. This is of significance, especially in the case of genetically engineered cells, where the instability of culture may impede the commercial application of such systems. The biological stability of immobilized cells has been attributed to a number of mechanisms, though the underlying mechanism may vary according to the metabolic and physiological states of the cells. For nonviable cells, the protective effect of the immobilization matrix against harsh environmental conditions such as extremes in pH and temperature, presence of toxic compounds (e.g., organic solvents, heavy metals, oxygen in the case of strictly anaerobic cells), and existing shear forces is the main contributing factor in biological stability. In the case of nongrowing or slow-growing cells, increases in biosynthesis of enzymes and cryptic growth are additional contributing factors. With actively growing cells, maintaining a balance between the cell growth and cell deactivation is a significant factor in achieving metabolic stability over an extended period.4 Although reproduction and cell growth are instrumental in maintaining the catalytic activity, the growth rate which results in the maximum stability might not necessarily lead to the optimum productivity. This is especially important when cells are used for the production of secondary metabolites. In an immobilized cell bioreactor, the periodic cycling of the cells between growth and no-growth condition through cyclic provision of an essential nutrient or energy source (e.g., light in case of phototrophic cells) is suggested as a successful strategy to achieve extended stability as well as high productivity. The stability of the genetically engineered immobilized cells has been attributed to the mechanical and structural properties of the matrix, which may allow only a limited number of cell divisions and prevent the reappearance or takeover of the system by plasmid-free (wild) cells.

2.34.5.2

Improved Biomass Hold-Up

The performance of a bioreactor, defined in terms of volumetric reaction rate or volumetric production rate, depends significantly on the biomass hold-up, regardless of the state of the cells (free vs. immobilized). It is well documented that the biomass hold-up in immobilized cell bioreactors is substantially higher than those bioreactors utilizing freely suspended cells. The biomass hold-up in an immobilized cell bioreactor is defined as the product of the number of support particles and the average biomass hold-up per particle, plus any freely suspended cells which may be present in the system.4 Although it is desirable to maximize the number of support particles to achieve the highest productivity or minimize the volume of the bioreactor, other factors such as the need for some level of motion, suspension, or circulation of the particles, especially in systems with moving bed, determine the number of particles per unit volume of the bioreactor. The biomass hold-up per particles is affected mainly by the technique used for cell immobilization, but growth rate and bioreactor hydrodynamic conditions are also influential. Apart from those discussed thus far, other parameters influence the biomass hold-up in the bioreactor. These are reviewed briefly in the following subsections.

502

Immobilized Cell Bioreactors

2.34.5.2.1

Support Matrix

The fraction of the particle occupied by the support material limits the level of the biomass within the particle, while the pore size of the support matrix determines the amount of the cells which can penetrate into the matrix, the depth of penetration (especially with passive entrapment), and the leakage of cells from the particle.4 The electrical charge of the support particle certainly influences the attachment of the cells. Finally, the biological interaction between the cells and the support material is of importance, especially when the carrier matrix is also the main energy source for the cells (i.e., organic-containing supports such as manure, compost, wood chips, and other waste cellulosic materials).

2.34.5.2.2

Cell Characteristics

Biological characteristics of the cells, including type, history, surface charge, size, shape, reproduction method, cell wall composition, oxygen requirements, as well as inoculation size, all influence the biomass hold-up and degree of immobilization.

2.34.5.2.3

Environmental Conditions

The biomass hold-up is affected by the rate of attrition, which depends mainly on the hydrodynamic conditions and the strength of the existing shear forces. Physicochemical conditions such as pH, temperature, ionic strength, and dissolved oxygen concentration are the other influential factors.

2.34.5.3

Improved Mass Transfer in Bulk Liquid

In a bioreactor, effective mass transfer in the bulk liquid is of great importance. For instance, the transfer of oxygen from the gas phase to the liquid phase is a crucial step in any aerobic process. Similarly, bioremediation of waste streams contaminated with organic compounds such as heavy petroleum hydrocarbons, polycyclic aromatic hydrocarbons, and naphthenic acids requires effective mass transfer between the organic and aqueous phases. High concentrations of freely suspended cells, cell morphology, and potential secretion of extracellular polymeric substances all influence the rheology of the fluid and could change the behavior of the fluid from Newtonian to non-Newtonian and cause substantial decreases in mass transfer between gas and liquid or aqueous and organic phases.4 By maintaining a relatively cell-free liquid phase and preventing an unwanted increase in the liquid viscosity, improved mass transfer can be expected in an immobilized cell bioreactor. It should be pointed out that the presence of particles in the bioreactor (particle hold-up) will certainly interfere with mass transfer processes, and hydrodynamic studies aiming to determine the optimum particle hold-up may be necessary.

2.34.5.4 2.34.5.4.1

Product and Process Improvements Improved Yield

Changes in metabolic activity and channeling the flow of the material through a particular pathway as well as extended use of the cell catalytic activity have been demonstrated to contribute in increasing the yield of certain metabolites as a result of cell immobilization.4 Diverting the flow of nutrients from cell synthesis to production of other metabolites not only prevents excessive growth of the biomass but could also maximize the yield of a desirable metabolite. Periodic addition of nutrients and proper cycling of nutrients in the bioreactor both have been identified as successful strategies to channel the flow of the mass inside the cell to a certain pathway.

2.34.5.4.2

Partitioning Effect

The partitioning effect due to chemical composition of the carrier matrix may provide a favorable microenvironment for the immobilized cells. In other words, by selecting a proper material for the matrix, it is possible to increase the concentration of the substrate in the vicinity of the cells.4 For instance, using hydrophobic gels as carrier could improve the bioreactions involving organic substrates with poor solubility in water. In a similar way, a proper matrix may decrease the concentration of an inhibitory product near the cell through the partitioning effect (e.g., hydrophilic gel used for production of highly insoluble pigments).

2.34.5.4.3

Downstream Processing

The application of an immobilized cell bioreactor could potentially eliminate the need for separation of the cells from the productcontaining stream. Moreover, maintaining cells as a discrete phase prevents the unwanted increase of the bulk liquid viscosity, which in turn facilitates transportation and processing of the liquid for the recovery of the product. High throughputs and short residence times, attractive operational features of the immobilized cell bioreactors, are also beneficial when the desired product is unstable or could be degraded by the enzymes which are produced by the present microbial population.

2.34.5.5

Cell Proximity and Reaction Selectivity

The proximity of the cells in an immobilized cell system is beneficial for cells with the ability to communicate. This proximity could also facilitate the transfer of plasmid DNA among the cells and provides an alternative technique for conjugation. The co-immobilization of different species within the particle eliminates the need for a series of bioreactors in case of multiple reactions.

Immobilized Cell Bioreactors

503

Finally, differences in the molecular size of various reactants and the ability to choose a carrier matrix with a desirable permeability (i.e., semi-permeable membranes) could be exploited to enhance the selectivity of a particular reaction.

2.34.6

Potential Drawbacks

2.34.6.1

Mass Transfer Limitations

Intraparticle diffusional resistance, which could limit the access of cells to the substrate, particularly oxygen in case of aerobic cells, and result in reduced bioreaction rates, is one of the widely debated topics and is considered as one of the major disadvantages of immobilized cell systems. Diffusional limitations may also hinder the transport of the product from the particle‘s inner core to the bulk liquid in the bioreactor. The extent and presence of these diffusional limitations, as discussed earlier, depends significantly on the relative rates of mass transfer and bioreaction and should be considered in that context.

2.34.6.2

Mechanical Problems

As discussed earlier, immobilization of the microbial cell often leads to significant biological stability when compared with free cells. However, the immobilized cell bioreactors may become susceptible to operational instability on account of a number of mechanical problems, which in most parts originate from the carrier matrix. For instance, the formation and accumulation of gaseous products within the immobilized cell particles may lead to flotation and under extreme circumstances rupture of the particles. The intraparticle accumulation of gas can be alleviated by using small, non-densely populated particles. This facilitates the transport of the gaseous metabolites from the particles into the bulk liquid. Excessive and uncontrolled growth of the cells not only creates significant resistance against mass transfer but could also result in a number of operational problems such as channeling and nonuniform distribution of the liquid phase, a severe pressure drop, and eventual plugging of the bioreactor, especially in case of adsorbed cell systems. As indicated by Webb and Dervakos,4 in hollow fiber membrane bioreactors uncontrolled cell growth may compress the fiber toward the inner core and reduce and eventually stop the flow of the substrate-containing liquid. The mechanical characteristics of the matrix are one of the determining factors in the operational stability and lifetime of an immobilized cell bioreactor. Among these, the ability of resisting excessive pressure because of cell growth and intraparticle gas formation, withstanding harsh hydrodynamic conditions, and coping with the forces that tend to compress the particle bed are of significance.4 The mechanical characteristics of the matrix could also restrict the choice of the bioreactor, as hydrodynamic conditions vary among different configurations. As a typical example, when cells are immobilized in a soft polymeric compound such as a gel, a gas-agitated bioreactor may be a better choice than a mechanically stirred tank. The mechanical stability of the carrier matrix could also influence the choice of the medium and its composition. For instance, the presence of chelating agents such as phosphate may jeopardize the integrity of the alginate particles.

2.34.6.3

Substrate Limitation

Substrate limitation as a result of intraparticle diffusional resistance is considered as a main drawback of immobilized cell systems. This may become an acute problem in case of aerobic processes, where the low solubility of oxygen and diffusional resistance could both contribute in oxygen limitation and poor performance of the bioreactor. Webb and Dervakos summarize a variety of options for improved oxygen mass transfer in immobilized cell bioreactors including increase of oxygen partial pressure, aeration in an external loop, in situ production of oxygen, use of oxygen carriers such as hemoglobin and organic solvents, and, finally, use of membrane oxygenators.4

2.34.6.4

Product Inhibition

For the product-inhibited reactions, cell immobilization could impose a negative impact. This is due to the establishment of a concentration profile within the particle, in which product concentration increases with the distance from the surface. Thus, the product inhibitory effect becomes more pronounced in the inner core and leads to reduced reaction rates.

2.34.7

Concluding Remarks

Cell immobilization is a multidisciplinary subject which bridges the pure and applied sciences. Cell immobilization technology has found numerous practical applications in the areas of environmental bioremediation and pollution control, production of biochemicals and pharmaceuticals, bioprocessing of food and food derivatives, biosensors, and, most importantly, biomedical engineering and medicine. The varied and diverse nature of these applications has led to the design, development, and utilization of a variety of bioreactor configurations suitable for the intended purpose. These include stirred-tank, fixed-bed, fluidized-bed, gasagitated, and membrane bioreactors. In addition to superior performance and flexibility in operating conditions, immobilized cell bioreactors enjoy prolonged stability, enhanced biomass hold-ups, reaction selectivity, increased product yield, and simplification of the downstream processing. Mass transfer resistances within the immobilized cell particles, which could potentially lead to

504

Immobilized Cell Bioreactors

substrate limitation and product inhibition as well as mechanical instability are some of the drawbacks attributed to these systems. Performance of the immobilized cell bioreactors depends greatly on the intrinsic biokinetics, external and internal mass transfers, and many other issues discussed earlier, and as such proper design and operation of these systems requires a thorough understanding of these issues.

See Also: 2.35 Bioreactors for Solid-State Fermentation.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

Bailey, J. E.; Ollis, D. F. Biochemical Engineering Fundamentals, 2nd ed.; McGraw-Hill Book Company: Singapore, 1986. Nedovic, V.; Willaert, R., Eds.; Focus on Biotechnology; Applications of Cell Immobilisation Biotechnologyvol. 8B; Springer: Dordrecht, 2005. Shuler, M. L.; Kargi, F. Bioprocess Engineering Basic Concepts, 2nd ed.; Prentice Hall PTR: Upper Saddle River, NJ, 2002. Webb, C.; Dervakos, G. A. Studies on Viable Cell Immobilization, Academic Press: New York, 1996. Buchholz, K.; Kasche, V.; Bornscheuer, U. T. Biocatalysts and Enzyme Technology, Wiley-VCH Verlag GMBH & Co. KGaA: Weinheim, 2005. Mavituna, F. Pre-formed Carriers for Cell Immobilization. In Focus on Biotechnology; Nedovic, V., Willaert, R., Eds.; Fundamentals of Cell Immobilisation Biotechnology, vol. 8A; Kluwer Academic Publishers: Dordrecht, 2004; pp 121–134. Brodelius, P. Immobilized Plant Cells. In Enzymes and Immobilized Cells in Biotechnology; Laskin, A. I., Ed., The Benjamin/Cummings Publishing Company Inc.: Menlo Park, CA, 1985; pp 109–148. Lacik, I. Polyelectrolyte Complexes for Microcapsule Formation. In Focus on Biotechnology; Nedovic, V., Willaert, R., Eds.; Fundamentals of Cell Immobilisation Biotechnology, vol. 8A; Kluwer Academic Publishers: Dordrecht, 2004; pp 103–117. Riddle, K. W.; Mooney, D. J. Biomaterials for Cell Immobilization, a Look at Carrier Design. In Focus on Biotechnology; Nedovic, V., Willaert, R., Eds.; Fundamentals of Cell Immobilisation Biotechnology, vol. 8A; Kluwer Academic Publishers: Dordrecht, 2004; pp 15–27. Obradovic, B.; Nedovic, V. A. Immobilised Cell Bioreactors. In Focus on Biotechnology; Nedovic, V., Willaert, R., Eds.; Fundamentals of Cell Immobilisation Biotechnology, vol. 8A; Kluwer Academic Publishers: Dordrecht, 2004; pp 411–431. Blanch, H. W.; Clark, D. S. Biochemical Engineering, Marcel Dekker, Inc.: New York, 1997. Baron, G. V.; Willaert, R. G.; De Backer, L. Immobilised Cell Reactors. In Immobilized Living Cell Systems: Modelling and Experimental Methods; Willaert, R. G., Baron, G. V., De Backer, L., Eds., John Wiley & Sons Ltd: New York, 1996; pp 67–95. Margaritis, A.; Kilonzo, P. M. Production of Ethanol Using Immobilized Cell Bioreactor Systems. In Focus on Biotechnology; Nedovic, V., Willaert, R., Eds.; Applications of Cell Immobilisation Biotechnology, vol. 8B; Springer: Dordrecht, 2005; pp 375–405. Dunn, I. J.; Heinzle, E.; Ingham, J.; Prenosil, J. E. Biological Reaction Engineering, 2nd ed.; Wiley-VCH Verlag GMBH & Co. KGaA: Weinheim, 2003.

2.35

Bioreactors for Solid-State Fermentation

DA Mitchell, LF de Lima Luz, and N Krieger, Universidade Federal do Paraná, Curitiba, Brazil M Berovic, University of Ljubljana, Ljubljana, Slovenia © 2011 Elsevier B.V. All rights reserved. This is a reprint of D.A. Mitchell, L.F. de Lima Luz, N. Krieger, M. Berovic, 2.25 - Bioreactors for Solid-State Fermentation, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 347-360.

2.35.1 2.35.1.1 2.35.1.2 2.35.1.3 2.35.1.4 2.35.2 2.35.2.1 2.35.2.2 2.35.3 2.35.3.1 2.35.3.2 2.35.3.3 2.35.4 2.35.4.1 2.35.4.2 2.35.4.2.1 2.35.4.2.2 2.35.4.2.3 2.35.4.3 2.35.5 2.35.5.1 2.35.5.2 2.35.5.2.1 2.35.5.2.2 2.35.5.3 2.35.6 2.35.6.1 2.35.6.2 2.35.6.2.1 2.35.6.2.2 2.35.6.3 2.35.7 2.35.8 2.35.8.1 2.35.8.2 2.35.9 References

Introduction Definition of Solid-State Fermentation Processes for Which We Need SSF Bioreactors: Past, Present, and Future Microorganisms Used in SSF Processes An Engineering-Based Approach Classification of SSF Bioreactors and Basic Principles of Operation Classification of Solid-State Fermentation Bioreactors General Considerations About Bioreactor Performance Tray Bioreactors Basic Features of Tray Bioreactors Design, Operation, and Scale-Up of Tray Bioreactors Current Challenges in Design, Operation, and Scale-Up of Tray Bioreactors Packed-Bed Bioreactors Basic Features of Packed-Bed Bioreactors Design, Operation, and Scale-Up of Packed-Bed Bioreactors Key Considerations in Designing and Operating Packed Beds Other Important Considerations in Operating Packed-Bed Bioreactors Strategies for Scale-Up of Packed-Bed Bioreactors Current Challenges in Design, Operation, and Scale-Up of Packed-Bed Bioreactors Rotating-Drum and Stirred-Drum Bioreactors Basic Features of Rotating-Drum and Stirred-Drum Bioreactors Design and Operation of Rotating and Stirred Drums Key Considerations in Designing and Operating Rotating and Stirred Drums Strategies for Scale-Up of Rotating-Drum and Stirred-Drum Bioreactors Current Challenges in Design, Operation, and Scale-Up of Rotating Drums and Stirred Drums Forcefully Aerated Agitated Bioreactors Basic Features of Forcefully Aerated Agitated Bioreactors Design, Operation, and Scale-Up of Forcefully Aerated Agitated Bioreactors Key Considerations in Designing Forcefully Aerated Agitated Bioreactors Strategies for Scale-Up of Forcefully Aerated Agitated Bioreactors Current Challenges in Design, Operation, and Scale-Up of Forcefully Aerated Agitated Bioreactors Challenges Related to Changes Provoked by Microbial Growth Other Considerations Batch Operation Versus Continuous Operation Automated Control Conclusion

506 506 506 507 508 508 508 508 509 509 510 511 511 511 512 512 512 512 513 513 513 514 514 514 515 515 515 515 515 516 516 516 517 517 517 517 518

Glossary Axial dispersion Mixing along the flow path of fluids during processing Bacteria Any of a large group of microscopic organisms having round, rodlike, spiral, or filamentous unicellular or noncellular bodies that are often aggregated into colonies are enclosed by a cell wall or membrane and lack fully differentiated nuclei. Bacteria may exist as free-living organisms in soil, water, and organic matter, or as parasites in the live bodies of plants and animals. Batch processing A method of processing in which a bioreactor, for example, is loaded with raw materials and microorganisms, and the process is run to completion, at which time the products are removed (see ‘continuous processing’). Bioreactor Vessel in which a bioprocess takes place.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00085-9

505

506

Bioreactors for Solid-State Fermentation

Concentration diffusion Molecular diffusion resulting from a nonhomogeneous distribution of concentrations of components of a mixture. Conductive heat transfer The transfer of heat through a substance via direct molecule-to-molecule contact. There is no perceptible movement of the conducting medium; the energy is transferred through the vibrations of the molecules in the medium. Continuous processing Method of processing in which raw materials are supplied and products are removed continuously, at volumetrically equal rates (see ‘batch processing’). Convective mass transfer Mass transfer produced by simultaneous convection and molecular diffusion. The term is usually used to describe mass transfer associated with fluid flow and involves the mass transfer between a moving fluid and a boundary surface or between two immiscible moving fluids. Convective transfer The transfer of mass, heat, or momentum in a medium with a nonhomogeneous distribution of velocity, temperature, or concentration; it is accompanied by the displacement of macroscopic elements through the medium. Fermentation A bioprocess. Fermentation is carried out in bioreactors and is used in various industrial processes for the manufacture of products such as antibiotics, alcohols, acids, and vaccines by the action of living organisms (strictly speaking, anaerobically). Fungus Any of a major group of saprophytic and parasitic plants that lack chlorophyll, including molds, rusts, mildews, smuts, and mushrooms. Heat transfer Spontaneous irreversible process of heat transmission in a space with a nonisothermal temperature field, as in the cooling or heating of bioreactors and ancillary equipment. Mass transfer Spontaneous irreversible process of transfer of mass of a given component in a space with a nonhomogeneous field of the chemical potential of the component. In the simplest case, the driving force is the difference in concentration (in liquids) or partial pressure (in gases) of the component. Other physical quantities, for example, temperature difference (thermal diffusion), can also induce mass transfer. Metabolic heat Energy released to the surroundings, in the form of thermal vibrations of molecules, during the metabolism of an organism. Metabolism The physical and chemical processes by which chemical components are synthesized into complex elements, complex substances are transformed into simpler ones, and energy is made available for use by an organism. Plug flow Flow of materials in which there is no mixing in the direction of flow (see ‘axial dispersion’). Scale-up The transition of a process from research laboratory bench scale to engineering pilot plant or industrial scale.

2.35.1

Introduction

2.35.1.1

Definition of Solid-State Fermentation

Solid-state fermentation (SSF) involves the growth of microorganisms on moist particles of solid materials in beds in which the spaces between the particles are filled with a continuous gas phase. It should be noted that the word ‘fermentation’ within the term ‘solid-state fermentation’ is usually used in the broader sense of ‘any controlled microbial process’ and does not imply that the microorganism is using fermentative metabolic pathways. The word ‘fermentation’ will be used in this article with this broader meaning. As shown in Figure 1, a typical bioreactor for SSF will involve three phases: (1) the body of the bioreactor itself; (2) a bulk gas phase, which, if it is above the bed, is typically referred to as the headspace; and (3) the substrate bed. The substrate bed itself may be thought of as consisting of two subphases, namely the particles of solid material, to which the growing microorganism is attached, and the interparticle gas phase. This article focuses specifically on those SSF processes in which, first, aerobic growth is desirable and, second, it is desirable to maintain the temperature at or near the optimal temperature for a single microorganism or group of microorganisms. These processes represent the majority of SSF processes currently being applied or studied. In systems with these characteristics, one of the major considerations guiding bioreactor design and operation is the provision of mechanisms for the adequate removal of waste metabolic heat, in order to prevent the temperature within the substrate bed from reaching such high values that it deleteriously affects microbial growth and product formation. Note that this situation is different from that which occurs in traditional composting processes: in this particular SSF process it is in fact desirable that the bed temperature increases significantly during the process, provoking a succession of microbial populations originating from the original microbial flora of the material being composted. Bioreactors designed specifically for composting processes are not considered here.

2.35.1.2

Processes for Which We Need SSF Bioreactors: Past, Present, and Future

SSF has been used for many centuries for the production of traditional fermented foods, especially in Asia. Until the 20th century, the bioreactors were very simple, typically consisting of pots or trays that contained the inoculated substrate, these being placed in a room that did not have mechanisms for exact control of the temperature and humidity. During the 20th century, the soy sauce industry invested in the development of bioreactors for the first step of soy sauce production (called the koji step), which involves

Bioreactors for Solid-State Fermentation

A

Bed of solid particles Headspace

Bioreactor wall

507

B

Moist solid particle Interparticle gas phase

C

D

Biofilm

Hyphae

Figure 1 The phases within a solid-state fermentation system. (A) Appearance of the system at the macroscale. At this scale it is possible to discern the substrate bed, the headspace above the bed, and the wall of the bioreactor. (B) Appearance of un-inoculated substrate at the microscale. At this scale it is possible to discern the individual particles and the gas spaces between them. (C) Growth of a biofilm of unicellular organisms (bacteria or yeasts) on the surfaces of the particles, as indicated by the thick black layer at the particle surface. (D) Growth of a network of fungal hyphae. This network grows across the surface while some hyphae penetrate into the substrate and others extend into the spaces between the particles.

SSF of soybeans by the fungus Aspergillus oryzae. Various large-scale bioreactors are reported to be used in this industry,1 although details of the considerations used in their design and their performance in soy sauce koji processes are not available in the published literature. Since the mid-1970s, there has been a surge of interest in using SSF technology to produce microbial products other than traditional fermented foods, including, but not limited to, enzymes, pigments, aroma and flavor compounds, antibiotics, biosurfactants, protein-enriched fermented feeds, and organic acids. Although many such processes have been investigated in the laboratory, there are currently relatively few examples of the successful establishment of large-scale SSF processes for the production of these newer products. One of the limiting factors has been the relatively poor knowledge base with respect to the design and operation of large-scale SSF bioreactors. In the future, this situation will need to change. As petroleum resources dwindle, there will be an ever-increasing pressure to move toward biorefineries for the production of industrial biochemicals. The large-scale cultivation of microorganisms will be an integral part of such biorefineries. However, as a result of the ever greater demands that population growth puts on resources, it will be essential to minimize the use of water in these cultivations, wherever possible. This will be especially imperative when the feedstock consists of an organic material that is originally available in solid form. Processing such a solid material by traditional submerged-liquid fermentation processes involves the addition of a large amount of water and leaves a large volume of wastewater after extraction of the product. This should be avoided since, even though adequate treatment of this wastewater might be possible, it will contribute significantly to operating costs. Water contents in SSF systems are considerably lower than in traditional submerged-liquid fermentation processes. For example, in a system involving 1 kg of dry matter, an SSF system would typically contain 1–5 l of water. In comparison, a traditional submerged fermentation system would need between 10 and 20 l of water in order to prevent the culture medium from being too viscous. The use of SSF technology therefore offers the possibility to minimize the addition of water and thereby optimize process economics in biorefineries. However, in order for SSF to fulfill this potential, it will be essential to optimize the performance of large-scale SSF bioreactors.

2.35.1.3

Microorganisms Used in SSF Processes

The majority of SSF processes that have been investigated to date involve filamentous fungi, although there are processes that involve either bacteria or yeasts. It is likely that processes involving filamentous fungi will continue to dominate. This has an important consequence for SSF bioreactors: agitation of a bed containing solid particles that have a filamentous fungus growing on their surfaces will damage the hyphae of the fungus. Depending on the sensibility of the fungus to this damage, restrictions might need to be placed on the frequency or intensity of mixing. Such restrictions will limit the strategies that can be used to improve heat removal from the bed.

508 2.35.1.4

Bioreactors for Solid-State Fermentation An Engineering-Based Approach

Several groups, since the mid-1980s, have investigated the kinetics of growth and the mass and heat transfer phenomena that occur within SSF systems. This article gives an overview of how the engineering principles established in these investigations can be used to guide the design and operation of large-scale SSF bioreactors. It also points out where future efforts are needed in order to advance the SSF bioreactor technology further.

2.35.2

Classification of SSF Bioreactors and Basic Principles of Operation

2.35.2.1

Classification of Solid-State Fermentation Bioreactors

It is convenient to classify SSF bioreactors into four groups, based on the aeration and agitation strategies that are used.2 Representative bioreactors of each of these four groups will be described in Sections 2.35.3 – 2.35.6, but it is useful to list the basic features of each group here:

• • • •

Group 1: Bioreactors in which the substrate bed is not forcefully aerated and either remains static during the whole fermentation period or is agitated very infrequently. The archetypal bioreactor of Group 1 is the tray bioreactor. Group 2: Bioreactors in which the substrate bed is forcefully aerated, but the bed either remains static during the whole fermentation period or is agitated very infrequently. The archetypal bioreactor of Group 2 is the packed-bed bioreactor. Group 3: Bioreactors in which the substrate bed is not forcefully aerated, with air being introduced into the bioreactor in the headspace above the bed, but the bed is agitated either continuously or very frequently. The archetypal bioreactors of Group 3 are the rotating-drum and stirred-drum bioreactors. Group 4: Bioreactors in which the bed is not only forcefully aerated but also agitated either continuously or very frequently. Various designs are possible for this type of bioreactor and it is difficult to identify a single archetypal design.3

Note that it is not possible to provide an exact definition of the dividing line between frequent and infrequent mixing. As a rough guide, one might consider six or fewer mixing events per day as infrequent, while mixing every 2 h (i.e., 12 times per day) might be considered as frequent. This article makes no attempt to describe all possible variations in the design and operation of bioreactors within the four groups. Our current state of knowledge about each of these bioreactor types and the various variations that are possible are described in much greater depth elsewhere.2,3

2.35.2.2

General Considerations About Bioreactor Performance

In analyzing the various bioreactor types, it is essential to keep in mind that, as with bioreactors for submerged-liquid fermentation, the performance of the bioreactor is controlled, first, by the interactions between the microorganism and its local environment and, second, by how effectively the design and operating strategies influence the conditions in the local environment of the microorganism (Figure 2). Strategies that promote bulk flow of gas through the interparticle spaces will promote convective heat and mass transfer in the bed and will be more effective in controlling the bed temperature than strategies that restrict heat and mass transfer within the bed to conduction and diffusion. Mixing the substrate bed can help to overcome the spatial temperature gradients that are associated with convective heat removal; however, the ability to use mixing will depend on just how sensitive the process microorganism is to the damage caused by agitation of particles within the bed. It is crucial to recognize that even if an SSF bioreactor is said to be well mixed, this only means that the particles themselves are well mixed within the bed. At the level of the individual particle, there is no mixing, and therefore mass and heat transfer processes within individual particles are limited to diffusion and conduction. Strategies for bioreactor design and operation can influence the efficiency of transport within the gas phase of the bioreactor and between the gas phase of the bioreactor and the particle surface, but they cannot make the transport phenomena within the particle more efficient. For example, although it may be possible to operate the bioreactor in such a manner as to maximize the O2 concentration in the interparticle spaces and even to ensure relatively high flow velocities past the particle surface in an attempt to maximize the heat and mass transfer coefficients, once O2 is transferred from the gas phase into the liquid phase within the particle, it can only move by diffusion. A consequence of this is that the interior of the particle, and also the innermost parts of any fungal mat or bacterial biofilm at the surface, will typically be anaerobic. Another consequence of intraparticle mass transfer being limited to diffusion is that there can be significant concentration gradients of nutrients within the particle, and the concentrations being experienced by the microorganism can be quite different from those that are measured when a sample containing many solid particles is homogenized and the nutrient concentrations are determined in the homogenate. Although intraparticle diffusion processes have received attention in an attempt to understand how they can limit the growth of the microorganism, not much effort is put into characterizing or describing them when the aim is to design and optimize bioreactors.2,4,5 Typically growth is simply treated in terms of the mass of dry biomass per mass of total dry solids without any attempt to describe the heterogeneous distribution of the biomass at the level of individual particles. Also, since the presence of intraparticle concentration gradients means that the nutrient concentration experienced directly by the microorganism is unknown, it is typical not to try to describe the growth rate as depending on the nutrient concentration. Instead, simple empirical equations, such as the logistic equation, are used to describe the growth kinetics. The parameters of these equations are usually, in turn,

Bioreactors for Solid-State Fermentation

Growing microorganism

Microbial growth causes changes in the local environment (through consumption of O2 and nutrients, release of metabolic heat, etc.)

Nutrients, temperature, and water activity in the local environment affect growth

Microscale: (level of individual particles)

Local environment: • Nutrient levels • Temperature • pH • [O2] • Water activity

Macroscale: (level of the whole bioreactor) Supply of O2 removal of CO2 removal of heat Bioreactor

Auxiliary equipment and surroundings

509

Mass and heat transfer phenomena within the substrate bed affect the efficiency with which changes in the operating variables can control the local environment

Operating variables: that can be controlled are as follows: • Inlet air temperature, humidity • Agitation frequency • Additions to the bed

Design features: will determine what operating variables are available and will affect the efficiency of mass and heat transfer within the bed Figure 2

The interactions between the various factors that affect the performance of a solid-state fermentation bioreactor.

expressed as empirical functions of the temperature and water activity. The main attention in designing bioreactors is then typically given, first, to providing adequate heat removal, second, to maintaining adequate water activities in the solids, and, third, to providing high O2 concentrations within the interparticle spaces. Of course, it is possible to influence the efficiency of intraparticle diffusion processes during the preparation of the substrate particles prior to the fermentation. The use of small particles minimizes the distances over which solutes (such as enzymes, nutrients, and O2) must diffuse, while high water contents can help to improve the diffusion coefficients. However, the particle size and water content cannot be adjusted with complete freedom: small particle sizes and high water contents tend to favor the formation of a compact mass of substrate that makes aeration difficult, or even excludes air.

2.35.3

Tray Bioreactors

2.35.3.1

Basic Features of Tray Bioreactors

A tray-type bioreactor consists of a chamber, which may be an incubator or a room, in which the temperature and humidity are controlled to some degree, and in which various recipients containing solid substrate are placed (Figure 3). Typically, these recipients will be trays that contain relatively thin layers of substrate, although various other recipients can be used, such as jars or microporous plastic bags. The air is simply circulated around the surfaces of the substrate layer rather than being blown forcefully through it. Although the bed typically remains static, it is also possible to mix it infrequently, usually by hand. The tray-type bioreactor is the most traditional SSF bioreactor, having been used for many centuries in the production of traditional fermented foods, such as tempe (an Indonesian fermented food based on the fermentation of soybeans with the filamentous fungus Rhizopus oligosporus) or the koji step of soy sauce production. It is still possible to find processes of this type in Asia. Tray bioreactors tend to be relatively labor intensive, since each individual bioreactor needs to be handled (loaded, mixed,

510

Bioreactors for Solid-State Fermentation

The positioning and spacing of trays and the positioning and design of the air circulation equipment will determine how efficiently the air is circulated around the surfaces of trays

Air (controlled temperature and humidity)

Fresh air

Exploded views Cover Thin layer of substrate Thin layer of substrate

Perforated bottom

Base

Figure 3 Basic features of the tray bioreactor system. The bioreactor itself is a chamber in which the temperature and humidity of the air are controlled. Trays, containing thin layers of substrate, are stacked within the bioreactor. On the left is a stack of uncovered trays. On the right is a stack of covered trays.

and unloaded) separately. However, if the product is of sufficiently high value, then tray bioreactors may be profitable at large scale, even in facilities that process several tons of substrate per day.

2.35.3.2

Design, Operation, and Scale-Up of Tray Bioreactors

In designing a tray bioreactor, the system needs to be considered at two different levels: first, the individual tray and second, the incubation chamber. In terms of the tray itself, the major consideration is that the lack of forced aeration and the lack of mixing (or very infrequent mixing) mean that heat and mass transfer within the substrate bed are restricted to conduction and diffusion. This leads to steep temperature gradients across the bed, as well as steep concentration gradients of O2 and CO2 in the interparticle gas spaces within the bed. As a result of these gradients, it is usually necessary to limit the bed height to no more than 5 or 10 cm, in order to prevent deleteriously high temperatures or O2 limitation in the center of the substrate layer, or both. The exact value of the allowable bed height will depend on how fast the microorganism grows, since this will determine the heat generation rates and O2 consumption rates. In terms of the incubation chamber, it will typically be important to ensure that conditions are well controlled throughout the chamber, without any stagnant regions that have poor air circulation. Ensuring a high airflow velocity past the surfaces of the trays will aid in heat removal, but at large scale it is unlikely to be practical to do this, and, in any case, this is not a particularly effective strategy since heat transfer within the bed will still be limited to conduction. It may be desirable to maintain a high humidity within the chamber, for example, by misting it with water. However, since this increases the risk of contamination, it will probably require the use of covered trays to prevent dripping water from wetting the bed surface. If trays are closed, then each tray will have a particular gas microenvironment in its headspace; the effectiveness of gas exchange with the bulk air of the chamber depends on how large the gap is between the tray and its cover.

Bioreactors for Solid-State Fermentation

511

Most of the experimental and modeling work that has been done to characterize the operation of tray-type bioreactors has focused on the tray itself. The basic approach has been to consider that if one can optimize the operation of a single tray at small scale in the laboratory; it will be possible to scale up the process by simply increasing the number of trays. However, as pointed out in the next subsection, other important issues still need to be investigated.

2.35.3.3

Current Challenges in Design, Operation, and Scale-Up of Tray Bioreactors

Little attention has been given to important issues that will affect the success of tray bioreactors, such as how to optimize the spacing between trays, the positioning of tray stacks, the geometry of the tray chamber, and the design of the air conditioning and circulation system. Attention also needs to be given to questions such as scheduling and automating of tray-handling steps such as loading, unloading, and cleaning. For covered trays, it is necessary to investigate what design of the tray cover will best prevent wetting of the bed while maximizing gas exchange between the headspace of the individual tray and the bulk gas phase of the tray chamber.

2.35.4

Packed-Bed Bioreactors

2.35.4.1

Basic Features of Packed-Bed Bioreactors

Traditionally, packed-bed bioreactors involve a static bed of substrate that sits on a perforated base through which air is blown forcefully (Figure 4). The air must flow through the interparticle spaces of the bed in order to leave the bioreactor. Various modifications to this basic design are possible, for example: (1) in addition to introduction of the air at the base, air can be introduced by hollow perforated tubes inserted into the bed; (2) the air does not necessarily have to be introduced at the bottom of the bed, rather it may be introduced at the top of the bed; (3) the bioreactor can be cylindrical with air being introduced through a perforated pipe at the central axis and removed through a perforated drum wall (a ‘radial packed bed’); (4) the bed can be divided into compartments by heat transfer plates that are oriented parallel to the airflow; and (5) the bed can be broken up into relatively shallow layers with heat transfer plates, oriented perpendicularly to the airflow, inserted under each layer to cool the incoming air.3 Small-scale packed-bed bioreactors have a special importance in laboratory-scale studies of SSF processes: the so-called Raimbault column bioreactor is a system that contains multiple packed-bed bioreactors, each typically around 4 cm in diameter and 20 cm high.3 This system is useful for characterizing O2 uptake rates, as an O2 analyzer can be set up to monitor the off-gases of the columns automatically. The respirometric data obtained can be used for estimating the rates of metabolic heat generation that can be expected at large scale. Purely static operation of packed-bed bioreactors is not common at large scale. However, in many large-scale processes with forced aeration, the frequency of the mixing is so low (of the order of once per day) that the bioreactor operates as a packed bed for the great majority of the time. Such infrequently mixed packed-bed bioreactors, with a capacity for several tonnes of substrate, are commonly used in the production of soy sauce koji.1

Uniform velocity profile across the bed

Perforated base plate

Air box

Exploded view

Air inlet: air with controlled temperature and humidity

Figure 4 Basic features of the packed-bed bioreactor. In this and the remaining bioreactor diagrams, the particle sizes are somewhat exaggerated in relation to the bioreactor.

512

Bioreactors for Solid-State Fermentation

2.35.4.2

Design, Operation, and Scale-Up of Packed-Bed Bioreactors

This section will consider the characteristics of static beds. If a packed bed is intermittently mixed, then the principles of operation during the periods of agitation will be similar to those discussed later (see Section 2.35.6.2).

2.35.4.2.1

Key Considerations in Designing and Operating Packed Beds

Typically, in packed-bed bioreactors, the air flows unidirectionally from the air inlet to the air outlet, with a uniform velocity profile across the bed (this flow regime being known as ‘plug flow’). The unidirectional flow of air, combined with the release of metabolic waste heat, leads unavoidably to temperature profiles in packed-bed bioreactors, with the temperature increasing between the air inlet and the air outlet. The temperature reached at the air outlet end of the bed will depend on the growth rate of the process microorganism, as this will determine the rate of generation of the waste metabolic heat. It will also be affected by the height of the substrate bed and the temperature and the flow rate of the inlet air. One of the major considerations in designing and operating large-scale packed-bed bioreactors is to avoid the temperature at the outlet air end of the bed reaching such high values that it will have unacceptably adverse effects on growth and product formation in this region. It is usually not a useful strategy to supply unsaturated air at the air inlet, since this will tend to dry out the regions of the bed near the air inlet quite rapidly. In fact, the bed will dry out even if saturated air is used: the increase in the temperature of the air as it flows through the bed increases its water-carrying capacity, thereby creating a driving force for evaporation. If it is necessary to avoid or to have very infrequent mixing, in order to avoid damage to the process microorganism, then it is necessary to minimize this evaporation, otherwise the water activity in the bed will fall to values that are so low that they restrict growth. If the bed does dry out so much as to need replenishment of water, it will be necessary to mix the bed, since there is no practical alternative to adding the water as a spray or mist while the bed is being agitated. The use of internal heat transfer plates (either parallel or perpendicular to the airflow direction) will minimize axial temperature gradients and therefore minimize evaporation. However, their use also complicates the design of the bioreactor and will increase operating costs, especially if the water used needs to be refrigerated.

2.35.4.2.2

Other Important Considerations in Operating Packed-Bed Bioreactors

Other important considerations in the operation of packed-bed bioreactors are the pressure drop through the bed and channeling. The air blown into the bed will lose pressure as it flows from the inlet to the outlet, due to frictional energy losses. The air pressure at the inlet end will therefore be equal to the pressure of the air at the outlet of the bioreactor plus the pressure drop across the bed. The aeration system must be capable of providing air at this pressure. At the low superficial air velocities (of the order of 10 cm s 1) that are most frequently used in packed-bed bioreactors, pressure drops are typically relatively small early on during the process. However, many SSF processes involve filamentous fungi, which, as they grow, form a network of hyphae that begin to fill up the interparticle spaces. Even at maximum density, this network occupies only around 35% of the space originally available between the particles, but this is sufficient for the pressure drop in the bed to become quite significant, in some cases reaching values of the order of 1 m of water (around 0.1 atm) per meter of bed. Although this phenomenon has been long known in packed-bed bioreactors and has received some attention, it has not been well characterized. It is an important phenomenon in many non-SSF applications of packed beds in chemical engineering, but these systems do not involve the development of interparticle filaments (i.e., fungal hyphae) during the process, so the transferability of this non-SSF knowledge about pressure drop is limited. Along with other phenomena, such as bed shrinkage, high pressure drops can help to promote the formation of channels, a phenomenon that greatly interferes with the proper operation of packed beds, as explained below. One strategy that can be used to prevent the pressure drop from reaching high values is to mix the bed once, early during the process, before the hyphae have grown significantly into the interparticle spaces.6 The damage caused to the hyphae by this early mixing event decreases their ability to grow into the interparticle spaces during the process. Channeling involves the formation of gaps, within the bed itself, or between the bed and the bioreactor wall. If such gaps appear, then they represent a path that offers less resistance to flow through the bed, so the air will flow preferentially through the channel, rather than through the interparticle spaces in the bed. If this occurs, steep temperature and gas concentration gradients, like those that occur in tray bioreactors, will start to form within the bed. Channeling can be caused by one or a combination of factors. If the process organism is a fungus, it can tend to bind the particles together. The contractive forces resulting from this binding, combined with a general shrinkage of the bed due to drying or partial consumption of the particle structure by the microorganism, can tend to pull the bed away from the walls or cause the formation of cracks within the bed. If channels appear within the bed, the only effective remedy is to agitate it, in an attempt to separate the individual particles and let them settle again to form a bed of uniformly distributed particles. Note that any agglomerates of substrate particles that are not broken up by this mixing will tend to be poorly aerated when normal packed-bed operation is resumed. An early agitation event, before these agglomerates have a chance to form, helps to minimize this problem.

2.35.4.2.3

Strategies for Scale-Up of Packed-Bed Bioreactors

It is most useful to express the aeration rate in terms of the superficial velocity, namely the volumetric airflow rate divided by the whole cross-sectional area of the bioreactor. A simple approach to deciding how to design and operate a large-scale packed-bed bioreactor would be to determine, in a laboratory-scale packed bed, the conditions in which the temperature at the outlet end of the bed remains within acceptable limits and then to scale up on the basis of maintaining the ratio of superficial velocity to

Bioreactors for Solid-State Fermentation

513

bed height constant.2 In this case, it is important that the sides of the laboratory-scale bioreactor be insulated, in order to mimic the fact that, in a large-scale bioreactor with a wide bed, the proportion of the waste metabolic heat removed through the side walls is quite limited. Mathematical models have been developed to describe the operation of packed-bed bioreactors and coded into computer programs that can be used as tools to guide the scale-up.2 Although these models are relatively simple, failing to describe some key phenomena, such as pressure drops and bed shrinkage, they can provide useful guidance. The easiest models to apply are those that assume equilibrium between the solids and gas phases, as it is not necessary to know the solids-to-gas heat and mass transfer coefficients within the bed. These coefficients will depend on the superficial velocity of the air and the size and packing of the particles and may change during the process because of microbial growth, so it is not an easy task to characterize them. However, the assumption of thermal and moisture equilibrium between the solids and gas phases is not always valid, especially when there are high heat generation rates within the bed.7 Models are available that recognize the solids and gas as separate subsystems. However, they estimate the heat and mass transfer coefficients using correlations determined in experiments investigating the drying of grains,8 which may not be appropriate once the microorganism starts growing within the bed.

2.35.4.3

Current Challenges in Design, Operation, and Scale-Up of Packed-Bed Bioreactors

Infrequently mixed packed beds are being currently used at industrial scale, especially in the production of soy sauce koji. In fact, various companies offer pilot or even full-scale bioreactors of this type for sale. However, there is still room for improvement. The challenge is to maximize the performance of this type of bioreactor, while minimizing increases in capital and operating costs, such that the cost per unit of final product decreases. For example, it would be desirable to increase bed heights while preventing excessively high temperatures at the air outlet. In order to minimize increases in operating costs, this would need to be done while minimizing increases in airflow rates and while avoiding excessively high pressure drops. In order to achieve these aims, it would be highly desirable, first, to increase the predictive power of mathematical models and, second, to use these mathematical models, in conjunction with experimental work at pilot scale, to develop innovative design and operating strategies. Several improvements need to be made to the mathematical models of packed-bed bioreactors. It would be desirable for future models to predict how the pressure drop across the bed changes during the fermentation and how this affects the airflow, something which current models do not do. Such models may need to be empirical (i.e., totally based on experimental results), since any attempt to develop a mechanistic model would require a description of the spatial distribution of biomass within the interparticle spaces. Little effort has been made to characterize the development of biomass above the particle surface, and even less effort to describe it mathematically. It would be useful if models were able to predict bed shrinkage, as such models could be used to optimize the timing of the infrequent agitation events, these events being used, in part, to avoid the appearance of channels. Again, this will probably need to be done empirically, as the tendency of the bed to shrink varies significantly, depending on the substrate used. Further, it would be useful to have a reliable method, through which online measurements of bed temperature and inlet and outlet gas-phase O2 concentrations could be used to estimate the water content of the bed as a function of position. Such a method could be useful in guiding decisions about the amount of water to be added to the bed during the infrequent mixing events. It may be possible to adapt current predictive models for this purpose. With respect to specific design and operating strategies, more attention should be given to optimizing the use of heat transfer plates. Modeling studies have been used to predict optimal spacing between heat transfer plates oriented parallel to the airflow.2 Similar analyses have not been done for the case in which the heat transfer plates are oriented perpendicularly to the airflow. In order to minimize operating costs, it would be necessary to minimize water usage, while avoiding the need to refrigerate water. It may be possible to use a closed circuit, in which water is circulated from regions of rapid growth (and therefore high rates of metabolic heat generation) to regions of slow growth, but this has not been investigated.

2.35.5

Rotating-Drum and Stirred-Drum Bioreactors

2.35.5.1

Basic Features of Rotating-Drum and Stirred-Drum Bioreactors

Rotating-drum bioreactors have two main characteristics. First, the bed of substrate is contained in a horizontal or inclined drum, and the bed is mixed through the rotation of the drum (Figure 5). Second, air is blown into the headspace above the solid bed. This means that it is not forced to flow through the bed itself. Rather, the bed is aerated by the exchange of gas between the bed and the headspace. Design variations include the use of an inclined central axis and the mounting of lifters (so-called ‘baffles’) on the inside of the drum wall. Agitation does not need to be continuous, but if it is not frequent, then during the static periods the bioreactor will operate like a tray bioreactor. Rotating drums have a relatively long history of use. As long ago as 1914, a rotating-drum bioreactor was used to produce amylases on wheat bran. Soon after World War II, an SSF facility for the production of penicillin on wheat bran was established in the United States, using rotating drums of over 10 m3 in volume, although advances in submerged fermentation technology for fungi meant that the SSF production method was soon abandoned. The use of rotating-drum bioreactors capable of holding up to 1500 kg of soybeans has also been reported in the production of soy sauce koji.1 In stirred-drum bioreactors, the drum body remains static and the bed is mixed by an internally mounted agitator, such as paddles mounted on a central axis (Figure 5). Again, air is introduced into the headspace above the solid bed, such that their operating principles are quite similar to those of rotating drums.

514

Bioreactors for Solid-State Fermentation

A

Air in

B

Air in

Figure 5 Basic features of (A) rotating-drum and (B) stirred-drum bioreactors. In both cases, the drums are closed at each end, although there is provision for the entry and exit of air.

2.35.5.2 2.35.5.2.1

Design and Operation of Rotating and Stirred Drums Key Considerations in Designing and Operating Rotating and Stirred Drums

The most important design and operating considerations for rotating- and stirred-drum bioreactors are related to the effectiveness of heat and mass transfer between the substrate bed and the headspace air. The effectiveness of this transfer depends mainly on the radial flow regime of the solids and the airflow pattern in the headspace. The radial flow regime of the solids determines how effectively the substrate particles are brought into contact with the headspace gases. This involves two aspects: first, the effectiveness with which the particles in the center of the bed are brought to the upper surface and, second, the effective area of contact between the upper surface of the bed and the headspace gases. The flow regime is affected by the velocity of rotation of the drum or agitator, the percentage filling of the drum, and the presence of lifters. At low rotational speeds in rotating drums without lifters, the bed follows a slumping flow regime in which the solids tend to rise up as a single mass with the rotating wall of the drum and then slide back down. This flow regime therefore neither mixes the bed well nor promotes heat and mass transfer at the bed surface, leading to dead zones in the bed that are poorly aerated and inefficiently cooled. This problem can be overcome by installing lifters on the inside of the drum wall; as they leave the bed they lift a portion of the substrate that then falls back to the bed surface as a curtain. In drums without lifters, slumping flow can be avoided by using high rotational rates. When the rotational rate is high enough, the bed enters into a cataracting flow regime in which not only is the bed well mixed across its cross section but also solid particles are thrown out into the airstream. Both the curtain effect and the cataracting flow regime ensure good heat and mass transfer with the headspace for at least some of the particles at the surface of the bed. In the case of stirred drums, the paddle design must ensure good radial mixing within the bed. With relation to the airflow patterns in the headspace, the main consideration is that there should not be any dead spaces. In fact, ideally, there should be good axial flow past the bed surface in order to maximize heat and mass transfer between the bed and the headspace. The airflow patterns within the headspace will be affected by the designs and positioning of the air inlet and the air outlet, the aeration rate, the rotation rate, and the presence of lifters. However, although these aspects have received some attention, they are not particularly well studied. The fact that a rotating drum or stirred drum is continuously or frequently mixed means that it is quite easy to add water to the bed, for example, using nozzles mounted on the central axis. This opens up the possibility of using unsaturated air at the inlet in an attempt to promote evaporative cooling of the bed.

2.35.5.2.2

Strategies for Scale-Up of Rotating-Drum and Stirred-Drum Bioreactors

Mathematical models can be used to guide the scale-up of rotating-drum and stirred-drum bioreactors. The simplest models assume that the substrate bed is one well-mixed phase while the headspace gas is a second well-mixed phase.2 The key bed-to-headspace heat and mass transfer coefficients used in these models are based on relatively simple correlations for the axial flow of air past

Bioreactors for Solid-State Fermentation

515

surfaces. The effect of lifters or rotational rate on these transfer processes is not described directly but rather by the use of empirical ‘fold increases’ in the coefficients. Simulations with these models suggest that at large scale the effectiveness of bed-to-headspace mass and heat transfer will be of critical importance in determining the bed temperature. In other words, it will be necessary to operate the bioreactor in such a manner as to ensure high transfer coefficients. This will probably involve high aeration rates with dry air.

2.35.5.3

Current Challenges in Design, Operation, and Scale-Up of Rotating Drums and Stirred Drums

Given the importance of heat and mass transfer between the bed and the headspace in rotating-drum and stirred-drum bioreactors, it is necessary to give more attention to characterizing the effect of design and operating variables on the airflow patterns within the drum headspace and the flow patterns of the solids within the bed. Airflow patterns have not received much attention, although it is possible to have a combination of a plug-flow region and stagnant regions, with the stagnant regions occupying over 60% of the overall headspace volume.2 An insight into how design and operating strategies affect the performance of the bioreactor is given by the work of Schutsyer,6 who used a ‘discrete particle model’ to investigate the effects of design and operating variables on mixing within the beds of rotating-drum bioreactors. In this type of model, the movement of each individual particle within the bed is calculated on the basis of Newtonian laws of motion. Although this approach is computationally very demanding, it can be used for an initial evaluation of the effectiveness of mixing that will be achieved with a particular drum geometry and set of operating variables. The best number, geometry, and placement of lifters or agitator blades can also be investigated using this method. Once a better understanding of headspace flow patterns and bed mixing patterns is obtained, it will be possible to obtain reasonable estimates of the bed-to-headspace heat and mass transfer coefficients. A relationship has been established between the bed-to-headspace mass transfer coefficient and the effective Peclet number.9 However, more work is required to confirm and extend these findings.

2.35.6

Forcefully Aerated Agitated Bioreactors

2.35.6.1

Basic Features of Forcefully Aerated Agitated Bioreactors

In forcefully aerated agitated bioreactors (FAABs), air is blown forcefully into the bed in such a manner that it must flow through the interparticle spaces in order to exit from the bioreactor, as is the case with the packed-bed bioreactor. The difference with packed beds is that the bed of an FAAB is continuously or frequently agitated. There are many different bioreactor designs that combine agitation and forced aeration and various large-scale bioreactors of this type have been built. As mentioned in Section 2.35.4.1, some packed-bed bioreactors used for the production of soy sauce koji have provision for intermittent mixing.1 Such bioreactors could be used with frequent mixing; in this mode of operation they would be classified as FAABs. A bioreactor with capacity for around 1 tonne of substrate was developed at the Institut National de la Recherche Agronomique or National Agronomic Research Institute (INRA) in Dijon, France.3 This bioreactor has a design typical of a packed-bed bioreactor, except that the bed is mixed by a set of screws mounted on a carriage that sits on rails above the bed and which moves continuously from one end of the bioreactor to the other (Figure 6). Although the screws operate continuously, any one part of the bed is agitated only intermittently. Although fluidized-bed SSF bioreactors have been reported, very little information is available and they will not be discussed here.

2.35.6.2 2.35.6.2.1

Design, Operation, and Scale-Up of Forcefully Aerated Agitated Bioreactors Key Considerations in Designing Forcefully Aerated Agitated Bioreactors

Various considerations regarding the design and operation of FAABs will be similar to those of packed beds. This is especially the case for the periods of static operation in intermittently mixed bioreactors. During these periods, axial temperature gradients will be established within the bed, and it is important to prevent the bed at the air outlet from reaching temperatures that are high enough to be deleterious to the process. As with packed-bed operation, the ratio of the superficial velocity of the air to the bed height will be of prime importance. One important difference from packed beds is that unsaturated air can be used at the air inlet in an attempt to promote evaporative cooling. Water can be sprayed on the bed surface while it is being mixed, in order to replenish lost water and therefore maintain the water activity at values that are sufficiently high so as not to restrict growth. The design and velocity of the agitator will be important in determining the effectiveness of mixing; however, this subject has received relatively little attention in SSF systems. An important consideration is whether continuous or intermittent agitation is to be used. If intermittent agitation is chosen, then the frequency of mixing events and the duration of each agitation event need to be decided. For intermittent agitation, the frequency of agitation will be determined by one or more of four factors: (1) the need to prevent high bed temperatures near the air outlet; (2) the need to replenish water so that growth will not be restricted by low water activities; (3) the need to prevent high pressure drops; (4) the need to ‘reseat’ the bed, due to the appearance of channels. Decisions about the frequency of mixing will also be affected by how sensitive the microorganism is to the damaging effects of mixing. If channeling or large pressure drops cause problems, then the first mixing event during the process may need to be timed to prevent the fungus from binding the substrate into agglomerates.6 If the aim is to use intermittent mixing to diminish the maximum temperature reached within the bed, it will be necessary to mix relatively frequently, as the profiles typical of packed-bed operation

516

Bioreactors for Solid-State Fermentation

Exploded view

Detail of the mixing blades

Perforated base plate

Air box

Figure 6

Air inlet: air of controlled temperature and humidity

A forcefully aerated, intermittently agitated bioreactor.

are reestablished within a relatively short time, typically 30 min to an hour after a mixing event. In this case, the organism must be able to tolerate frequent mixing, but, if it does, continuous agitation might be better.

2.35.6.2.2

Strategies for Scale-Up of Forcefully Aerated Agitated Bioreactors

Mathematical models have been developed that can be used to guide the design and operation of large-scale FAABs. A model is available that describes intermittent agitation, although it focuses on the period of static operation and treats the agitation event very simply.8 Models are also available that describe continuously agitated operation. These models suggest that, in order to control the temperature of the bed within acceptable limits, it will be necessary to use higher superficial air velocities than those typically used in laboratory-scale bioreactors, combined with the use of unsaturated air and inlet air temperatures that are several degrees below the optimum temperature for growth.2

2.35.6.3

Current Challenges in Design, Operation, and Scale-Up of Forcefully Aerated Agitated Bioreactors

Insufficient attention has been given to the question of how best to agitate a forcefully aerated bed of solids. Although screw mixers are often used, there are no studies of optimum screw design, rotational speed, or spacing between multiple screws. Schutsyer6 undertook studies of mixing patterns in an agitated SSF bioreactor by monitoring positron emission from substrate particles labeled with a positron-emitting isotope. Discrete particle modeling was used to describe the mixing patterns that were found. These experimental and modeling approaches represent a promising approach to identifying design and operating strategies that will optimize mixing in these types of bioreactors. For intermittently agitated beds, more effort is needed to investigate the effects of agitation-related variables, namely, the frequency and duration of agitation events. Various aspects need attention. First, the effectiveness of heat and mass transfer during the agitation period is likely to be higher than during the periods of static operation, but this has not been studied quantitatively. Second, it is necessary to understand better just what are the deleterious effects that mixing has on the microorganism. Third, attention needs to be given to the fact that agitation can affect the structure of the substrate particles, potentially grinding them into a paste or favoring the formation of cohesive agglomerates from which O2 is excluded. It is also important that the agitation method does not lead to preferential airflows during the period of static operation (e.g., in gaps left beside the mixing blades) such that some regions of the bed are not well aerated. However, the question as to how to agitate the bed in such a way as to avoid this problem has not been addressed.

2.35.7

Challenges Related to Changes Provoked by Microbial Growth

For all bioreactor types, the properties of the substrate bed are important. Attention needs to be given to the effect of microbial growth on these properties. In the development of many of the mathematical models that have been proposed as tools for guiding the design and operation of SSF bioreactors, the model parameters are usually based on the properties of substrate beds prepared as

Bioreactors for Solid-State Fermentation

517

for a fermentation but which have not been inoculated. However, microbial growth during the process can lead to significant changes in bed properties, and these changes may have significant effects on the performance of the system. For example, the isotherm of the solids changes significantly during the growth of the microorganism, since the microbial biomass and the residual solid substrate typically have very different isotherms. Since the water content of the solids affects the driving force for evaporation, it is important to have an isotherm that takes into account the biomass content of the solid material in the bed. Other effects of microbial growth also need investigation. It is quite possible that the growth of the microorganisms at the particle surface will affect the coefficients related to heat and mass transfer between the solids and air phases. However, this question simply has not been investigated. Microbial growth will also likely provoke changes in the volume, density, or surface properties of particles, while fungal hyphae can bind particles together and fill up interparticle spaces. Although it is known that these phenomena can lead to bed shrinkage, increased pressure drops, the opening up of channels, the formation of agglomerates, or compacted beds, the amount of information that is available is limited.

2.35.8

Other Considerations

There are many aspects related to SSF bioreactors other than those that have been covered above.2 In the following two subsections, brief comments are made about two key areas that are quite important, but which have received relatively little attention, namely the possibility of continuous operation and the automated control of processes in SSF bioreactors.

2.35.8.1

Batch Operation Versus Continuous Operation

Most of the bioreactors presented in the previous sections are designed for batch operation. Continuous operation is also possible. It should be noted that the performance of continuous SSF processes carried out in continuous stirred-tank reactor (CSTR) mode is different from that of CSTR processes in submerged-liquid fermentations, due to the fact that each substrate particle acts as an independent microbioreactor.2 Based on considerations of product uniformity and efficient utilization of substrate, when continuous operation is used in SSF processes, it is most likely to be of the plug-flow type. In fact, several continuous, or at least semi-continuous (i.e., with intermittent flow), plug-flow processes have been investigated over the last 30 years or so.

2.35.8.2

Automated Control

In any fermentation process it is essential to detect deviations of the conditions within the bioreactor from values that are optimal for growth and production formation and then to make necessary adjustments to the operating variables in an attempt to bring the conditions back to the optimal values. Automated control schemes for this purpose involve two aspects: monitoring of the process conditions and control algorithms. Key process variables that are useful to monitor include, amongst others, the amount of microbial biomass, the bed temperature, the bed moisture content, the pressure drop across the bed, and the off-gas composition. The equipment and methods that can be used to monitor them have been well reviewed.10 It should be noted that the solid nature of the bioreactor contents in SSF processes creates challenges that are usually not faced in submerged-liquid fermentations. For example, it is difficult to ensure proper contact between the glass bulb of a pH electrode and the solid medium. Further, the forces generated during mixing will damage fragile probes, so they must be removed from the bed before each mixing event. A key challenge in SSF systems is that it is difficult to measure directly the amount of biomass in the substrate bed because of the presence of the residual solid substrate. This is especially the case in processes that involve fungi, because many fungi bind tightly to the substrate particles. As a result, the growth process is usually followed indirectly, for example, on the basis of oxygen consumption. The application of process control algorithms in SSF systems creates special challenges. Many bioreactors are not continuously agitated, meaning that the conditions within the bed are heterogeneous. Therefore the control task in most SSF bioreactors is not to maintain the value of a particular variable (such as the bed temperature) at its optimal value, but rather to minimize the deviations from that optimal value across the bed. It may also not be practical to undertake continuous control actions: for example, the addition of water will usually need to be done relatively infrequently, so as not to have too many mixing events. Further, since the great majority of SSF processes are batch processes and not continuous processes, the task is not necessarily to maintain a given condition but rather to ensure that an optimal trajectory is followed by the process. These issues have been discussed previously, but our understanding of how to apply control theory to SSF processes is rudimentary.2

2.35.9

Conclusion

In this article we have described the basic principles of design and operation of the various types of SSF bioreactors. Since SSF processes will be an integral part of future biorefineries, it is imperative that we strengthen and extend our knowledge base with respect to the engineering principles of these bioreactors. We currently understand the basic heat and mass transfer processes that occur within them, since the principles underlying these phenomena are well established in the chemical and biochemical engineering literature. On the other hand, relatively little effort has been put into obtaining correlations that are specific for SSF systems and which describe how the mass and heat transfer coefficients depend on substrate characteristics and operating

518

Bioreactors for Solid-State Fermentation

conditions. The growth of microorganisms on solid substrates also brings special challenges: Not only are the kinetics of growth quite complex in this system, but it is also difficult to study these kinetics because of the difficulty of separating the biomass from the residual solid substrate. Finally, more effort is required applying automatic control strategies to SSF bioreactors.

References 1. Sato, K.; Sudo, S. Small-scale Solid-state Fermentations. In Manual of Industrial Microbiology and Biotechnology; Demain, A. L., Davies, J. E., Eds., 2nd ed.; ASM Press: Washington, DC, 1999; pp 61–79. 2. Mitchell, D. A.; Krieger, N.; Berovic, M. Solid-state Fermentation Bioreactors: Fundamentals of Design and Operation, Springer: Heidelberg, 2006. 3. Durand, A. Bioreactor Designs for Solid State Fermentation. Biochem. Eng. J. 2003, 13, 113–125. 4. Mitchell, D. A.; von Meien, O. F.; Krieger, N.; et al. A Review of Recent Developments in Modeling of Microbial Growth Kinetics and Intraparticle Phenomena in Solid-state Fermentation. Biochem. Eng. J. 2004, 17, 15–26. 5. Rahardjo, Y. S. P.; Tramper, J.; Rinzema, A. Modeling Conversion and Transport Phenomena in Solid-state Fermentation: A Review and Perspectives. Biotechnol. Adv. 2006, 24, 161–177. 6. Schutyser, M. Mixed Solid-state Fermentation: Numerical Modeling and Experimental Validation, Wageningen University, 2003. PhD Thesis. 7. Weber, F. J.; Oostra, J.; Tramper, J.; et al. Validation of a Model for Process Development and Scale-up of Packed-bed Solid-state Bioreactors. Biotechnol. Bioeng. 2002, 77, 381–393. 8. von Meien, O. F.; Mitchell, D. A. A Two-phase Model for Water and Heat Transfer within an Intermittently-mixed Solid-state Fermentation Bioreactor with Forced Aeration. Biotechnol. Bioeng. 2002, 79, 416–428. 9. Hardin, M. T.; Howes, T.; Mitchell, D. A. Mass Transfer Correlations for Rotating Drum Bioreactors. J. Biotechnol. 2002, 97, 89–101. 10. Bellon-Maurel, W.; Orliac, O.; Christen, P. Sensors and Measurements in Solid State Fermentation: A Review. Process Biochem. 2003, 38, 881–896.

2.36

Bioreactors for Plant Cell Cultureq

S Furusaki, The University of Tokyo, Tokyo, Japan T Takeda, Kanto Gakuin University, Yokohama, Japan © 2017 Elsevier Inc. All rights reserved. This is a reprint of S. Furusaki, T. Takeda, Bioreactors for Plant Cell Culture, Reference Module in Life Sciences, Elsevier, 2017.

2.36.1 2.36.2 2.36.2.1 2.36.2.2 2.36.2.3 2.36.2.4 2.36.3 2.36.3.1 2.36.3.2 2.36.3.3 2.36.3.4 2.36.3.5 2.36.3.6 2.36.4 2.36.4.1 2.36.4.2 2.36.4.3 2.36.4.4 2.36.5 2.36.5.1 2.36.5.2 2.36.6 References

Introduction General Aspects of Plant Cells Plant Cell Culture Medium Component Secondary Metabolites Genetic Transformation of Plant Cells Various Types of Reactors for Plant Cell Culture Tank Reactors Bubble Beds Rotary Reactors Reactors for Hairy Root Culture Reactors with Product Separation Reactors for Immobilized Cell Culture Operation of Plant Cell Reactors Preparation of Plant Cells Operating Condition of Reactors Optimization of Production Efficiency Recovery of Products Industrial Applications and Outlook for the Future Production of Useful Substances Somatic Embryo Production Summary

520 520 520 521 521 522 522 522 523 523 523 525 526 526 526 527 528 528 529 529 529 530 530

Glossary Aneuploidy An abnormality involving a chromosome number. Callus (plural form, calli) Aggregate of irregularly shaped dedifferentiated cells of plants. The induction of callus is the starting procedure of plant cell culture. The kinds and concentrations of phytohormones affect the formation of callus cells. Hairy roots can be induced from callus cells. Conditioning factor Compounds secreted by the cells in plant cell culture to enhance the activity of proliferation or productivity of secondary metabolites (see the glossary term ‘elicitor’). De novo synthesis Synthesis of biological compounds from precursors through a metabolic pathway. Elicitor Substance to enhance the cell activity of proliferation or production of secondary metabolites. Secretes of fungi often reveal this effect. Organic substances such as methyl or ethyl jasmonate give similar effects as well as some kinds of peptides, polysaccharides, or heavy metal ions. Explant A part of a plant used for cultivation. Hairy roots Irregularly shaped roots of plants induced by infection of Agrobacterium sp. The growth rate of hairy roots is much higher than that of calli. Therefore, the oxygen demand is much greater. Protocorm A spherical tissue formed from the embryo of an orchid when it germinates. Ri plasmid A plasmid of Agrobacterium rhizogenes. It is widely used in plant cell culture to induce hairy roots. Secondary metabolite Metabolites that are not necessary for living creatures to live and grow but are useful to protect from infection or insect attack. Secondary metabolites in plants have been used as pigments, fragrances, and pharmaceutical ingredients. Shoot primordium Plant tissue that will develop into shoot tips.

q

Change History: January 2016. S. Furusaki and T. Takeda. The small mistakes are corrected in the text and references. There are no changes in the figures and tables.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.09076-2

519

520

Bioreactors for Plant Cell Culture

Somatic embryo Embryo induced from callus or hairy root cells, which can be differentiated to tissues and organs to form intact plants. Zygotic embryo Plant embryo formed in a seed and generated from a fertilized ovum.

2.36.1

Introduction

Plant cell culture is a unique process in biotechnology, which has interested many researchers because it can produce products that bacteria or animal cells cannot produce. Plant cells are the sole producers of alkaloids and anthocyanins. The biotransformation of biological compounds such as terpenoids or steroids is possible using plant cells. Proteins can also be produced using plants, particularly using genetically engineered plants, and there is little risk of human infection of proteins by viruses or pathogens compared to other methods such as production using animal cells or microorganisms. In addition, plant cell culture will enable the preparation of artificial seeds for seedling production, increasing the potential of producing various kinds of useful plants in field cultivation in agriculture. Artificial seeds provide genetically homogeneous plants and help enable easily controlled agriculture or plantation. However, plant cell culture is costly because of the slow proliferation rate and small productivity. The features of plant cells have been described in published literature.1–3 The use of plant cell culture needs to compete with production by intact plants cultivated in fields. It is often cheaper and easier to produce useful substances by intact plants than by cell culture. Careful economical assessments are, therefore, always necessary for the application of plant cell culture. Another drawback of plant cell culture is gene instability during culture. Unless effective procedures to stabilize the genes of the cultured cells are developed, the commercial application of plant cell culture may still be limited. Bioreactor studies for plant cell culture have been reported since the mid-20th century. There are a very large number of publications on plant cell cultures toward the end of the century. The types of bioreactors and their behavior are summarized with outlook by Sarj et al.4 Because of the above-described serious drawbacks of plant cell culture, however, the number of articles on their application seems to have recently decreased. Several devices have been proposed to realize efficient production by the use of plant cell culture. Plant cells are sensitive to the environment compared with microorganisms and are easily damaged by stresses such as shear stress. Reactor design to avoid fluid dynamic damage is salient, and several devices have been proposed for this purpose. Often, feedback inhibition by products is observed and simultaneous cultivation with separation of products is desirable in this case. Therefore, culture with product separation has been studied by many researchers. Immobilization is also useful to avoid fluid dynamic stress if products can be obtained without destroying cells. Continuous operation with product separation is possible by using immobilized plant cells. In many cases, mass transfer is not a limiting process in the secondary metabolite production by callus cells, although it often has a strong effect on hairy roots. Immobilization is a useful method in the application of plant cell culture using callus cells. By immobilizing plant cells, continuous operation becomes possible and fluid dynamic stresses can be avoided. The features of the bioreactors and apparatuses that are used for plant cell culture are described in this article. Some characteristics different from normal bioreactors using microorganisms are described in the following sections. Approaches for the commercial application of plant cell cultures are also discussed.

2.36.2

General Aspects of Plant Cells

2.36.2.1

Plant Cell Culture

Plant tissue/cell culture is the technique with which the tissues or cells isolated from plants or their organs are cultivated in an artificial environment. It has been known that plants can regenerate from explants such as cuttings. Such explant techniques have been used in large areas of agriculture. The first trial of plant tissue/cell culture was the cultivation of isolated cells by G. Haberlandt in 1902.5 In this trial, the cells could not grow. After the discovery that auxin and cytokinin are required to induce cell division and growth, isolated tissues and cells were proliferated and differentiated in the artificial environment using these phytohormones in many studies conducted in the 1950s. These proliferated cells have the same genetic information as the donor plant. Thus, the technique has been applied for the propagation of clone plants and the production of useful metabolites contained in the donor plants. The application areas in the plant tissue/cell culture are shown in Fig. 1. The origins of the plant tissue/cell culture are classified in the shoot tips and somatic cells. The shoot tips are the tissues that are originally able to regenerate the plant. Shoot-tip culture isolated from plants can multiply in the form of a protocorm-like body, axillary bud, or shoot primordium and regenerate clone plants. The shoot-tip culture has been applied in modern agriculture for the production of clone plants. With the use of auxins and cytokinins, calli, adventitious buds, adventitious roots, and somatic embryos can be induced from somatic cells that cannot originally regenerate plants. A callus is the agglomerate of cells grown without tissue formation, so it is the main object of cell culture and metabolite production. We will mainly review cultivation using calli in the following sections. Adventitious buds are a type of artificially induced shoot tips, which have been investigated for the regeneration of plants. Somatic embryos are the tissues that generate from somatic cells or

Bioreactors for Plant Cell Culture

521

Regeneration

Plant

Protocorm-like body Shoot tips

Axillary bud Shoot primordium Regeneration

Adventitious buds Somatic cells

Adventitious roots Somatic embryos Callus

Fig. 1

Application areas in plant tissue/cell culture.

calli and develop into tissues such as zygotic embryos. The somatic embryos have a potential as a material that can be used to produce artificial seeds. Hairy roots are also applied to metabolite production in bioreactors as are calli. They are induced by infection with Agrobacterium rhizogenes. This induction is the result of transformation via the Ri plasmid of this bacterium. The transformation of plant cells will be described later.

2.36.2.2

Medium Component

Media containing inorganic salts, sugars, and trace elements such as vitamins are used for plant tissue/cell culture. Natural products such as casein hydrolysate are also used. Many studies are based on the Murashige and Skoog (MS) medium (1962) and Gamborg‘s B5 medium (1975).5 Many other media have been reported and used in various cultures, although most of the ingredients are common. Phytohormones are important factors in the cultivation of calli and adventitious organs. They play a function in metabolic regulation similar to that of hormones in animals. However, their structures differ from animal hormones. Phytohormones are generally called plant growth regulators. Auxins and cytokinins are mainly used in plant cell culture among the applicable phytohormones, i.e., auxin, cytokinin, gibberellin, abscisic acid, and ethylene. Calli are induced with a proper balance of auxins and cytokinins. Adventitious buds are generally induced with a higher concentration of cytokinin than the balanced concentration. Adventitious roots are induced with a higher concentration of auxin. Most auxins and cytokinins used in cultivation are synthetic compounds, which do not exist in nature and differ in chemical structure from natural compounds. They are used as phytohormones because they have a function similar to natural compounds. The natural auxin, 3-indoleacetic acid (IAA), is unstable in the cultivation environment, and synthetic auxins such as 3indolebutyric acid, 1-naphthaleneacetic acid (NAA), and 2,4-dichlorophenoxyacetic acid (2,4-D) are thus widely used in plant cell culture. These synthetic auxins are often more efficient in callus induction than the natural auxin, IAA. It has been reported that these synthetic auxins act differently from IAA in intracellular metabolism. In addition, as for cytokinins, synthetic compounds such as kinetin and 6-benzylaminopurine are widely used in cultivation instead of the natural cytokinin, zeatin.

2.36.2.3

Secondary Metabolites

Calli induced from intact plants are generally used for plant cell culture instead of established cell lines, which are mainly used in mammalian cell cultures. One of the reasons is that most of the target products of plant cell culture are secondary metabolites specific to each plant. Plants contain many secondary metabolites, which human beings have used as pigments, fragrances, and pharmaceutical elements. A framework for the metabolic pathways of plant secondary metabolism is shown in Fig. 2. Among these metabolites are alkaloids, terpenoids, quinones, and phenylpropanoids. These compounds have played the roles of protection from infection, suppression of feedings, and guidance of mating in the ecosystem. Therefore, the production of secondary metabolites often requires some signaling factors to bring out their roles. Although a secondary metabolite is produced in a specific tissue of a plant, cultured cells can produce such a metabolite independently from original tissues. Each cell keeps all genetic information of the original plant and can express it under the proper condition. An exogenous signal and also an endogenous metabolic state are the condition under which genetic information can be expressed. For example, the accumulation of products in cells causes feedback inhibition in production pathways.

522

Bioreactors for Plant Cell Culture

D-Glucose

Terpenoid Isopentenyl diphosphate

Glyceraldehyde Erythrose 4-phosphate Phenyl3-phospate propanoids Phosphoenolpyruvic acid

Mevalonic

Pyruvic acid

Malonyl

Acetyl-CoA

Polyketi

Shikimic Tryptophan Indole

Citric acid TCA cycle

Ornithine

Quinones

Phenylalanine Tyrosine Isoquinoline

Pyridine alkaloids Tropane alkaloids

α-Ketoglutaric acid Fig. 2

A framework for metabolic pathways of plant secondary metabolites.

Fluctuation of the metabolic state and mutation sometimes induce high productivity of the target metabolite during cultivation. Thus, selecting cells with high productivity is important for the efficient production of metabolites in plant cell culture. It is necessary to use cells with higher productivity in order to keep down the production costs because sterilizing processes in the plant cell cultivation process cost a lot in comparison with costs in the field cultivation of plants.

2.36.2.4

Genetic Transformation of Plant Cells

The genetic transformation of plants was established in the 1980s. The first report of recombinant protein production using transgenic plant cell-suspension cultures appeared in 1990, and more than 300 investigations have been reported to date.6 The genetic transformation of plant cells is carried out with Agrobacterium-mediated transformation, particle bombardment, or electroporation of protoplasts. Agrobacterium-mediated transformation is a technique unique to plants and has supported gene transformation of plants as the most useful technique. Agrobacterium is the bacterium that infects plants. When infection occurs, a part of plasmid contained in the microbe cell is transferred to the infected plant. A. tumefacience generates a tumor called crown gall and A. rhizogenes generates hairy roots in infected plants. They are used for transformation of plants by substituting part of the plasmid to the desired sequence. Tobacco and rice cells have been used as the host cells in many cases. Cauliflower mosaic virus 35S promoter is generally used for constituent expression. Induced promoters created by photo-irradiation or by chemical factors have also been investigated for application. Recombinant proteins, which have been investigated for production using plant cells, are mainly therapeutic proteins including antibodies, single-chain variable fragments (scFv), antigens, vaccines, hormones, growth factors, and human blood proteins. Plant cells are capable of posttranslational modification, and proper folding of the recombinant protein is possible. Moreover, plant cells are free from contaminant risks for human health due to virus infections. Therefore, practical application of this technique is expected in the future.

2.36.3

Various Types of Reactors for Plant Cell Culture

Calli are first induced on solid (gel) media in laboratory-scale development. The most popular media are the MS or B5 medium.5 After a sufficient amount of cells are grown on the solid media, the cells are transferred into liquid media. In small laboratory scale, cells are normally cultured in flasks mounted on shaking apparatuses. After a certain number of days of culture, the cells are again transferred to larger-scale liquid-phase bioreactors. Hereafter, several types of industrial bioreactors for plant cell culture will be described (Fig. 3). Comprehensive reviews were published by Su.2,3

2.36.3.1

Tank Reactors

Cells are cultured in vessels equipped with mixing stirrers of various types (Fig. 4). The mixing blades are designed so that the hydrodynamic damage to cells is controlled to a minimum. Due to hydrodynamic stress, cells often lose their proliferation capability, viability, and productivity of secondary metabolites (described in Section 2.36.4). Air supply is necessary for cells to grow. Air is supplied into reactors through spargers forming bubbles. The spargers as well as agitating blades are devised so that the hydrodynamic stress becomes as small as possible. Gas bubbles tend to grow; if the size of these bubbles becomes large, the local turbulence at the liquid surface becomes large and the impact of bubble rupture damages the cells. By contrast, if the size of the bubbles is smaller than the size of the cells, then shear stress on the cell surface or cell membrane

Bioreactors for Plant Cell Culture

A

523

B

Medium inlet

Air outlet Air outlet

Suspended cells in medium

Medium inlet

Air inlet

Air inlet

Medium outlet

Bubbles Suspended cells in medium Medium outlet

C

D

Air outlet Tube side (medium)

Draft tube

Medium inlet

Bubbles

Suspended cells in medium

Air inlet

Medium outlet Shell side (cells)

E

Air inlet

Air outlet

Roller

Suspended cells in medium

Fig. 3 Bioreactors for plant cell culture: (A) stirred tank reactor; (B) bubble bed; (C) airlift reactor (with draft tube); (D) membrane reactor; and (E) rotary drum reactor (batch method).

will have some effect on cell metabolism. Thus, bubble control in terms of sizes and flow patterns is often essential to the operation of plant cell bioreactors (see Section 2.36.4).

2.36.3.2

Bubble Beds

When air supply is considered to be salient for cell viability, bubble beds, that is, bubble columns, can be applied. If the aspect ratio, the ratio of height to the diameter, is large, it is referred to as a bubble column. The major difference between bubble beds and stirred tank reactors is that bubble beds or columns are not equipped with impellers. The liquid mixing is performed by bubble motion. Therefore, the superficial velocity of air in bubble beds is generally larger than that in tank reactors. Bubbles tend to accumulate in the center of the bed, and circulating flow prevails in the bed, upward in the central part and downward in the periphery. Often, draft tubes are installed in the beds in order to stabilize the circulating flow (Fig. 3(C)). In some cases, a side vessel is attached to the main reactor, and aeration is carried out using the side vessel. Hydrodynamics and mass transfer in plant cell cultures were studied by Sarj et al.7 using a four-phase external-loop air-lift bioreactor.

2.36.3.3

Rotary Reactors

As plant cells easily deteriorate due to hydrodynamic stresses, reactors with less dynamic impact have been developed. The lateral rotary reactor is a prospective device with less hydrodynamic stress, as described in an article.8 Cells are suspended inside a rotating drum as shown in Fig. 5. Scoop paddles are often installed on the inner surface of the drum for mixing cells and medium. In many cases, oxygen is normally supplied through the surface of the culture medium. As oxygen consumption by callus cells is not as large as that for microbial fermentation, oxygen supply through the surface will generally be sufficient to preserve cell viability. Rotary reactors are suited to continuous operation and the reactivity is increased by stage-wise operation. A multistage system is possible by placing perforated baffle plates inside the drum (Fig. 5(B)).

2.36.3.4

Reactors for Hairy Root Culture

Hairy roots are induced from plant cells by Agrobacterium sp. infection. They generally grow more rapidly and are higher in productivity than callus cells. They can grow in higher density than callus culture. In many cases, they do not require phytohormones. Hairy roots are, therefore, expected to be used for the production of secondary metabolites prepared in root cells. Examples of products

524

Bioreactors for Plant Cell Culture

A

(A) Flat blade (B) Tilted blade

B

C

D

Shaft

Hole Top view Fig. 4 Various types of stirrers for tank bioreactors: (A) turbine blade (flat and tilted); (B) ribbon-type mixer; (C) paddle-type mixer; and (D) perforate disk impeller.

Rotating drum

A

Medium surface

Medium + Air

Medium + Air

Stationary rod Roller

Scoop paddle

Mechanical seal

Inner cross section

Inner side view

Rotating drum

B

Air

Air

Medium surface Medium + Air

Medium + Air

Mechanical seal

Stationary rod Roller Fig. 5

Perforated plate

Rotary bioreactor: (A) single-stage rotary reactor and (B) four-stage cascade rotary reactor.

Bioreactors for Plant Cell Culture Table 1

525

Examples of hairy roots appeared in literatures for engineering study

Product

Plant

Betanin Horseradish peroxidase Hyoscyamine, cinnamoyl putrescine Interleukin Saponin, glycosides Shikonin Tropane alkaloids

Beta vulgaris Armoracia rusticana Hyoscyamus muticus Nicotiana tabacum Panax ginseng Lithospermum erythrorhizon Duboisia leichhardth

from hairy roots are given in Table 1. Besides the production of secondary metabolites shown in the table, there are a number of studies using the hairy roots of transgenic plants for the production of various proteins. Engineering studies for hairy-root cultures were reviewed by Toivonen.9 Hairy roots grow irregularly; therefore, it is very difficult to construct regular and homogeneous liquid flow inside the bioreactor, where large diversity in the densities of root cells is observed. Reactors for culturing hairy roots need to be carefully designed to avoid the above disorder of the cell-density profile and poor distribution/channeling of liquid flow. As the root cell density is much higher than the cell density in callus culture, careful consideration regarding oxygen mass transfer is necessary. Several devices to culture hairy roots are shown in Fig. 6. As oxygen demand is an important factor in hairy root culture,10,11 Kino-oka et al.10 proposed a radial flow bioreactor to avoid oxygen starvation at the exit of medium flow. The medium flows from the outer periphery to the center of the reactor, and the flow rate is larger toward the exit. This causes higher oxygen transport at the central root-dense area near the exit. Often, products can be recovered from the culture media by oxygen starvation.11 Transport phenomena in trickle beds applied to hairy roots have been studied in detail by Ramakrishnan and Curtis.12

2.36.3.5

Reactors with Product Separation

When the product is secreted into the medium and feedback inhibition exists in the metabolism, a production system with product separation will result in efficient production. The schematic process diagram is shown in Fig. 7. Here, the major separation methods include extraction and adsorption. For hydrophobic products, organic solvents are used to extract the products or hydrophobic adsorbents are used to adsorb the products. Yamamoto et al.14 recently proposed the separation of hydrophobic paclitaxel by floss flotation (foam separation). This is a prospective method to separate hydrophobic products without using organic solvents. Membrane bioreactors can be applied for the separation of products or wastes from media. Plant cells are normally contained in the shell side of tubular membranes and stay in the reactor for sequential use. Culture media flow in the tube side. Substrates and products penetrate through the membrane. The membrane reactor is usually operated in continuous mode (see Fig. 3(D)). A

B

Air outlet Medium inlet

Medium inlet Misted medium

Mesh Mesh Hairy roots Hairy roots

Air inlet Medium outlet

Medium outlet

C

D

Medium outlet

Medium inlet

Medium inlet

Medium inlet

Hairy roots

Hairy roots Mesh Medium outlet

Medium outlet

Fig. 6 Examples of bioreactors for hairy roots: (A) mesh reactor; (B) mist reactor; (C) radial flow reactor; and (D) trickle bed. (B) Modified from Ref. 13. (C) Modified from Ref. 10.

526

Bioreactors for Plant Cell Culture

A

Air

Fresh medium

Extract

B

Fresh medium

Air

Purge

Air Solvent

Air

Purge Reactor

C

Fresh medium

Air

Adsorbent

Extractor

Reactor

Adsorption column

Air outlet Floss Over flow Medium with the condensed product

Air Air inlet

Fig. 7 Schematic illustration of plant cell bioreactor with product separation: (A) separation by solvent extraction; (B) separation by adsorption; and (C) separation by floss flotation.14

2.36.3.6

Reactors for Immobilized Cell Culture

Plant cell immobilization15 is as useful as microbe immobilization in the case the products are secreted from cells. Most of the immobilization is done via entrapment by gels such as calcium alginate, carrageenan, agarose, or other sol–gel materials, although some of the immobilization methods are processed by attachment to supporting carriers such as chitosan and chitin. Plant cells can also be immobilized in porous synthetic materials such as polyurethane foams and nylon fibers, or in natural porous materials such as a loofah sponge. The characteristics of the immobilized cells in porous materials are similar to those in gel matrices. The cell-immobilizing particles are used in reactors, which are generally used for solid catalysts, such as tank reactors, bubble beds, and packed beds as shown in Fig. 8. When cells are immobilized in gel matrices, they are protected from fluid dynamic stresses and are often stable in metabolic activities. There is a possibility of mass-transfer limitations regarding oxygen, substrate, and product transports. In many cases, the proliferation of plant cells is slow. Therefore, oxygen and substrate demands are less than those for microbes, and mass-transfer limitations will not be salient for the metabolic activities of callus cells unless cell density becomes very high. As hairy roots grow dense, mass transfer is often a limiting step in hairy root culture. Immobilized cells can be used continuously or repetitively unless their metabolic activities are kept unchanged.

2.36.4

Operation of Plant Cell Reactors

2.36.4.1

Preparation of Plant Cells

The most important subject in plant cell culture is the efficient preparation of the cells producing the target metabolite. Callus cells can be induced from plant tissues within approximately 1 month of cultivation. The obtained callus cells are, however, a mixture of cells in various states of differentiation and/or have the metabolism of different activities to produce the metabolites. The productivity varies during a series of cultivation due to somatic mutation, aneuploidy, and variation of the metabolic state. It has been reported that synthetic phytohormones such as 2,4-D induce somatic mutation and aneuploidy. The selective cultivation of productive cells often cancels the requirement of a specific signal for production and enhances the productivity. Shikonin production in Lithospermum erythrorhizon cells was increased eight times by the selection of cells, and anthocyanin production by selected Vitis sp. cells cancels the requirement of photoirradiation for production. Although the fluctuation of productivity during the cultivation generates some portion of specific cells with high productivity, it sometimes causes an undesired decline in productivity. The regulation and maintenance of productivity are essential for realizing practical use of a plant cell-culture system. Genetic expressions and their regulation relating to secondary metabolism have been extensively investigated. Anthocyanin production pathways are one of the metabolisms about which considerable knowledge has accumulated. The metabolic pathways initiating from hydrophobic amino acid, phenylalanine, include many biological reactions, and, thus, many genes that code for the enzymes catalyzing the corresponding metabolic pathways. It has been reported that the expression of specific genes is regulated by

Bioreactors for Plant Cell Culture

A

B

Medium outlet

Medium inlet

527

Air outlet Immobilized plant cells

Immobilized plant cells

Bubbles Air inlet Medium outlet

Medium inlet

C

D

Air outlet Medium inlet

Immobilized plant cells Bubbles

Air inlet

Immobilized plant cells

Sparger Medium outlet

Fig. 8

Medium outlet

Medium inlet

Immobilized plant cell bioreactors: (A) packed bed; (B) stirred tank; (C) bubble bed; and (D) membrane reactor.

other genes (regulatory genes). Regulation of the regulatory genes will be useful for establishing cell lines with high productivity. An important challenge in the selection process has been to grow a small amount of specific cells with stable genes. There are generally a minority of useful cells in the culture. Thus, selection is an important process in plant biotechnology. Plant cells have small specific growth rate under the culture condition of a low cell concentration. This is caused by the deficiency of growth factors secreted from cells. To supply the growth factors, the addition of a nearby cultivation layer with a high density of cells (feeder layer), which are often inactivated by ultraviolet (UV) irradiation, or the addition of supernatant media used for other stable cultivation batches, is employed. This kind of operation is called conditioning culture and the growth factor is called conditioning factor. To cultivate a small amount of cells besides the conditioning culture, a small-volume culture, a so-called microculture, is carried out in order to achieve high-density cultivation. A detrimental decrease in culture volume is often caused in microculture because of a high specific surface area and the high evaporation rate of the medium, which is an inherent problem in microculture.

2.36.4.2

Operating Condition of Reactors

When large-scale culture in reactors is operated, the relative balance of hydrodynamic stress and mass transfer should be taken into consideration. Plant cells grow and consume substrates at lower rates than microbes. Thus, plant cells do not require a high masstransfer rate, but oxygen supply tends to be insufficient. Plant cells form aggregates and the local oxygen consumption rate around the aggregates is high. Therefore, an efficient supply of oxygen is required because the solubility of oxygen in aqueous solution is low. However, the enhancement of aeration and/or agitation increases hydrodynamic stress. Plant cells are suggested to be less tolerant to mechanical stress than microbes, similar to mammalian cells. Although plant cells are protected from mechanical stress by the cell walls, large particle size caused by both large cell size and aggregation results in considerable damage by hydrodynamic stress. This is illustrated as follows. A high mass-transfer rate is generally achieved by turbulent flow in a reactor. The energy of large vortices is transferred consecutively to smaller vortices and finally dissipated as thermal energy due to fluid friction in turbulent flow. The scale of minimum vortex varies depending on the cultivation system and operating condition. The smaller particles such as microbes rotate in the vortex and the hydrodynamic stress on the particles is reduced (Fig. 9(A)). On the other hand, larger particles such as aggregates of plant cells suffer directly from the hydrodynamic stress by shear (Fig. 9(B)). It has been reported that hydrodynamic stress decreases viability or causes a change in the metabolic activity of plant cells.16,17 The metabolic response in plant cells is caused by a smaller intensity of hydrodynamic stress than that which causes physical cell destruction. Therefore, reactors should be selected and the operating condition should be designed for necessary oxygen supply with minimum hydrodynamic stress.

528

Bioreactors for Plant Cell Culture

B A

Vortex Fig. 9

Cell

Cell

Vortex

Interaction between a cell and vortices: (A) a small cell being rotated by a vortex and (B) a large cell stressed by vortices.

Higher cell density increases the oxygen demand and decreases the oxygen mass-transfer rate due to higher apparent viscosity of the culture media. Thus, aeration and agitation must be enhanced. High cell-density culture is required to achieve high productivity, but it is difficult to avoid inactivation of cells caused by hydrodynamic stress.

2.36.4.3

Optimization of Production Efficiency

Controlling the signal factor is effective for increasing the yield of metabolites because secondary metabolites, which are the main targets of plant cell culture, often hold stress-responding characteristics. For example, many secondary metabolisms such as anthocyanin production require photo-irradiation. Plants have photoreceptors such as phytochrome in addition to the photosynthesis system, and the photo-receiving system regulates some metabolisms. Several photoreactors have been investigated for their photo-responding metabolism of cultured cells. Another possible approach is the selection of mutant cells, which do not require photo-irradiation. Although there is a small probability that this kind of mutation is generated, the obtained mutants are promising materials for practical use. Elicitors, which trigger infection response, activate many secondary metabolisms. This is related to the fact that many secondary metabolites were developed as the infection-protective factors in an ecosystem, as described above. There are two kinds of elicitors: exogenous and endogenous elicitors. The exogenous elicitors are the homogenates or the extracts derived from microbes. The endogenous elicitors are the compounds generated inside plant cells when the cells are infected by viruses or fungi. They include the components of cell walls. Alginic acid used for immobilization of cells is a compound similar to pectin in the plant cell wall and is reported to function as an elicitor. Jasmonic acid is a widely used elicitor. It is secreted from cells when they suffer some sort of damage; this damage induces metabolic responses. Methyl jasmonate is reported to be effective to enhance productivity of paclitaxel by Taxus sp.18 Enhancement of production by elicitor or by nutrient limitation often suppresses cell growth. This is one of the reasons why intermediate metabolites are used in primary metabolism in actively growing cells. Two-stage culture is an effective method in such cases. In this method, the production-culture stage follows the growth-culture stage. Temperature-swing culture is often effective in enhancing the production of metabolites.

2.36.4.4

Recovery of Products

In addition to production, recovery of products is an important subject. Many secondary metabolites accumulate mainly inside cells. Thus, recovery of the product requires the destruction of cells and extraction from the cell debris. However, destruction is an undesirable procedure, because plant cells grow slowly and it is costly to obtain a large amount of cells. Some permeation techniques for product separation have been investigated using organic solvents, surfactants, heat, and ultrasonic treatment. The common trend in these investigations was that the activity of cells decreased when enhanced permeation was achieved. The threshold level of release without inactivation of cells seems to correspond to the level of accumulation around the cell wall. Most of the releasable compounds are accumulated in cell walls and are not secreted spontaneously into the medium. The permeation process described above probably enhances the secretion of products from cell walls. Many secondary metabolites are also accumulated in vacuoles, and such compounds cannot be released by permeation treatment. Metabolite accumulation inside cells or in the cell wall possibly causes product inhibition. The release of products can enhance the production of secondary metabolites. Improvement of product release is an important issue. Elicitor treatment is reported to increase the release of some metabolites. The use of protoplasts without the cell wall is another possible method to release metabolites. The development of reactor systems with simultaneous recovery of released metabolites was described in the previous section.

Bioreactors for Plant Cell Culture

2.36.5

Industrial Applications and Outlook for the Future

2.36.5.1

Production of Useful Substances

529

Secondary metabolites are useful products that can be obtained from plant cells, for example, steroids and alkaloids to be used as pigments and/or pharmaceuticals. They are produced by de novo synthesis or transformation using precursors. Numerous substances have been studied in the past, some of which are listed in Table 2. However, the products that are industrially produced are limited because of the low growth rate of cells, low productivity, high cost of media components, and instability of genes. The most successful product produced to date is paclitaxel.19,20 It is an expensive medicine, which is clinically effective for some cancers. Paclitaxel or its precursor is solely produced by plants. Although the majority of the paclitaxel products are extracted from leaves of intact yew trees, that is, Taxus spp., Bristol Myers Squibb, in conjunction with Phyton Biotech, Ltd., is commercially supplying paclitaxel produced by using cultured cells of Taxus sp. Phyton has the world‘s largest tank reactors for the plant cell culture of 75 kl for paclitaxel production. The cultured Taxus cells produce both paclitaxel and its precursors, from which paclitaxel is chemically synthesized. Another product currently produced by plant cell culture is Asian ginseng, Panax ginseng, a cell containing a kind of saponin, ginsenoside. The whole cells are used to prepare supplements for human healthcare. In order to compete with products from intact plants, the cost of cell culture must be reduced and genes of cultured cells must be stable. In addition, the prices of the products must be high. Strategies for the production of secondary metabolites using plant cell culture were reviewed by Doernenburg and Knorr.21 A review on anthocyanin production can be found in Ref. 22. Protein production is another prosperous area of production in plant cell technology. Proteins are used for human healthcare, diagnosis, and therapy. Proteins used for this purpose should be free from viruses or pathogens. If they are produced with microorganisms or mammalian cells, there may be a risk of infection or immune problems. Therefore, the use of plant cells for protein production is expected to be safer than the use of mammalian cells or microbes. Huang and McDonald6 recently reviewed the research articles on this topic and discussed future aspects for the use of plant cells. Products contained in the roots of plants can be produced by the use of hairy-root culture. No commercial products have been reported using hairy roots. The productivity of hairy roots is said to be high, but the growth of hairy roots is very random, inhomogeneous in spatial direction, and hard to control. Thus, the distribution of liquid flow inside bioreactors is heterogeneous and difficult to regulate. This is probably the reason why hairy roots could not be applied in commercial production. There is a possibility of transforming biological compounds such as terpenes and steroids into their derivatives by plant cell culture. However, no commercial production has been reported in the literature.

2.36.5.2

Somatic Embryo Production

The purpose of plant cell culture is not only the production of secondary metabolites but also the production of somatic embryos to be prepared for micropropagation of useful plants. Micropropagation is important in developing specific useful plants such as vegetables and flowers. Genetically manipulated plants are also grown by this technique. These plants are often useful because they are germ resistive or antibiotic. They can produce proteins for health- or medical care as well. The medium contents, particularly the phytohormone concentration, affect the induction of somatic embryos. In general, a low concentration or hormone-free condition is applied to induce somatic embryos. Differentiation of cells and tissues from somatic embryos is also carried out by controlling the concentration of phytohormones. Somatic embryos change their shapes successively starting from a globular shape to an oblong shape, then to a heart shape, and finally to a torpedo shape. This pattern of development of embryos has been studied using image analysis by researchers. From the torpedo embryo, shoots emerge. Table 2

Examples of useful products from plant cell culture

Product

Plant species

Ajmalicine Anthocyanins Berberin Betanin Capsaicin Digoxin Ginsenosides Nicotine, recombinant proteins, etc. Paclitaxel Pigment, biotin Sanguinarine, codeine Shikonin Tocopherol

Catharanthus roseus Vitis vinifera, Perilla frutescens Berberis wilsoniae Beta vulgaris Capsicum sp. Digitalis lanata Panax ginseng Nicotiana tabacum Taxus spp. Lavandula vera Papaver somniferum Lithospermum erythrorhizon Carthamus tinctorius

530

Bioreactors for Plant Cell Culture

Somatic embryos are often immobilized in gel particles, for example, calcium alginate gel, for further application. The immobilized embryos are called artificial seeds. Somatic embryos originate from a single plant source through callus cells or hairy roots. Therefore, their genes are considered the same, that is, the plants differentiated from artificial seeds grow with the same characteristics and similar productivities. This will considerably reduce the labor and cost of field cultivation in agriculture. As somatic embryos are rather large and have an irregular shape, they are vulnerable to fluid mechanical stresses. Careful consideration to reducing the stress should be applied to reactors when culturing somatic embryos. Several reactors have been proposed for culturing somatic embryos, such as a tank reactor with a helical ribbon stirrer.23 Other possible reactors for growing somatic embryos are tapered bubble beds and rotary reactors.

2.36.6

Summary

The physiology of plant cells, types of reactors dealing with plant cells, operation of plant cell culture, and problems in commercial application were described in this article. Plant cells grow more slowly than microorganisms and are vulnerable to environmental stresses. Therefore, the culture condition of plant cells must be carefully considered. There are currently not many industrial applications of plant cell culture because of the gene instability of calli, slow proliferation rate, and small content of product substances. The most successful application is the production of paclitaxel, as the demand for the product is large and the price is high. Paclitaxel can only be produced from plants commercially. The stability of callus genes must be enhanced in order to further develop the commercial application of plant cell culture. The proper use of elicitors helps increase the productivity of secondary metabolites. Stimulation by stresses with light, temperature, or other culture conditions may increase the productivity if the stresses are within the durable range. Temperature swing or change of the phytohormone content sometimes results in high productivity of secondary metabolites. The addition of precursors and/or elicitors/conditioning factors also helps increase the productivity of secondary metabolites by plant cell culture. Controlled secretion is desirable to separate products from the culture media and to enable repetitive use of the cells. Immobilization is worth considering if the product is secreted during cultivation because immobilized cells are protected from fluid dynamic stresses and because continuous operation becomes possible by immobilization.

References 1. Ryu, D. D. Y., Furusaki, S., Eds. Advances in Plant Biotechnology, Elsevier: Amsterdam, 1994. 2. Su, W. W. Bioprocess Technology for Plant Cell Suspension Cultures. Appl. Biochem. Biotechnol. 1995, 50, 189–230. 3. Su, W. W.; Lee, K. T. Plant Cell and Hairy Root Cultures – Process Characteristics, Products and Application. In Bioprocessing for Value-added Products from Renewable Resources; Yang, S. T., Ed.; Elsevier: Amsterdam, 2006; pp 263–292. Chapter 10. 4. Sarj, L.; Grubisic, D.; Vunjak-Novakovic, G. Bioreactors for Plant Engineering: An Outlook for Further Research. Biochem. Eng. J. 2000, 4, 89–99. 5. Mantell, S. H., Smith, H., Eds. Plant Biotechnology, Cambridge University Press: Cambridge, 1983. 6. Huang, T. K.; McDonald, K. E. Bioreactor Engineering for Recombinant Protein Production in Plant Cell Suspension Cultures. Biochem. Eng. J. 2009, 45, 168–184. 7. Sarj, L.; Obradovic, B.; Vukovic, D.; Bugarski, B. Hydrodynamics and Mass Transfer in a Four-phase External Loop Air Lift Bioreactor. Biotechnol. Prog. 1995, 11, 420–428. 8. Tanaka, H.; Nishijima, F.; Suwa, M.; Iwamoto, T. Rotating Drum Fermentor for Plant Cell Suspension Culture. Biotechnol. Bioeng. 1983, 25, 2359–2370. 9. Toivonen, L. Utilization of Hairy Root Culture Fir Production of Secondary Metabolites. Biotechnol. Prog. 1993, 9, 12–20. 10. Kino-oka, M.; Hitaka, Y.; Taya, M.; Tone, S. High-density Culture of Red Beet Hairy Roots by Considering Medium Flow Condition in a Bioreactor. Chem. Eng. Sci. 1999, 54, 3179–3186. 11. Kino-oka, M.; Hongo, Y.; Taya, M.; Tone, S. Culture of Beet Hairy Root in Bioreactor and Recovery of Pigment from the Cells by Repeated Treatment of Oxygen Starvation. J. Chem. Eng. Jpn. 1996, 25, 490–495. 12. Ramakrishnan, D.; Curtis, W. R. Trickle-bed Root Culture Bioreactor Design and Scale-up: Growth, Fluid-dynamics, and Oxygen Mass Transfer. Biotechnol. Bioeng. 2004, 88, 248–260. 13. Liu, C.; Towler, M. J.; Medrano, G.; et al. Production of Mouse Interleukin-12 Is Greater in Tobacco Hairy Roots Grown in a Mist Reactor than in an Airlift Reactor. Biotechnol. Bioeng. 2009, 102, 1074–1086. 14. Yamamoto, S.; Nohara, K.; Goto, Y.; et al. Enhanced Production of Paclitaxel by Callus Culture in a New Bioreactor Equipped with a Unit of Foam Separation. In Abstract 6th European Symposium of Biochemical Engineering and Science, p. 25. Salzburg, Austria, 27–30 August; 2006. 2006. 15. Rhodes, M. J. C. Immobilized Plant Cell Cultures. Top. Enzyme Ferment. Biotechnol. 1985, 10, 51–87. 16. Kieran, P. M.; Malone, D. M.; MacLoughlin, P. F. Effects of Hydrodynamic and Interfacial Forces on Plant Cell Suspension Systems. Adv. Biochem. Eng. Biotechnol. 2000, 67, 139–177. 17. Takeda, T.; Tamura, M.; Ohtaki, M.; Matsuoka, H. Gene Expression in Cultured Strawberry Cells Subjected to Hydrodynamic Stress. Biochem. Eng. J. 2003, 15, 211–215. 18. Yukimine, Y.; Hara, Y.; Nomura, E.; et al. The Configuration of Methyl Jasmonate Affects Paclitaxel and Baccatin III Production in Taxus Cells. Phytochemistry 2000, 54, 13–17. 19. Zhong, J. J. Plant Cell Culture for Production of Paclitaxel and Other Taxanes. J. Biosci. Bioeng. 2002, 94, 591–599. 20. Kim, J. K.; Gibson, D. M.; Shuler, M. L. Effect of the Plant Peptide Regulator, Phytosulfokine-a, on the Growth and Taxol Production from Taxus Sp, Suspension Cultures. Biotechnol. Bioeng. 2006, 95, 8–14. 21. Doernenburg, H.; Knorr, D. Strategies for the Improvement of Secondary Metabolite Production in Plant Cell Cultures. Enzym. Microb. Technol. 1995, 17, 674–684. 22. Zhang, W.; Furusaki, S. Production of Anthocyanins by Plant Cell Culture. Biotechnol. Bioproc. Eng. 1999, 4, 231–252. 23. Takayama, S.; Akita, M. The Types of Bioreactors Used for Shoots and Embryos. Plant Cell Tissue Organ Cult. 1994, 39, 147–156.

2.37

Bioreactors for Animal Cell Cultures

M Taya and M Kino-oka, Osaka University, Osaka, Japan © 2011 Elsevier B.V. All rights reserved. This is a reprint of M. Taya, M. Kino-oka, 2.27 - Bioreactors for Animal Cell Cultures, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 373-382.

2.37.1 2.37.2 2.37.3 2.37.3.1 2.37.3.2 2.37.4 2.37.5 References

Introduction Bioreactor Design Bioreactors for High Cell Density Cultures Radial Flow Bioreactor Systems Bioreactor for Manufacturing Mechanical Substitutes Automation of Cell Processing Toward Clinical Application Concluding Remarks

531 532 532 533 536 537 539 539

Glossary Autocrine-derived growth factor Growth regulator secreted by a cell as a hormone or chemical agent that binds to receptors on the same cell. Autologous transplantation Transplantation of organs, tissues, or even proteins from one part of the body to another in the same individual. Baby hamster kidney (BHK) cells BHK fibroblasts are an adhesive cell line used in molecular biology to be useful for transformations and for stable and temporary transfections. Chinese hamster ovary (CHO) cells CHO cells, which are a cell line derived from the ovary of the Chinese hamster, meant for use in biological and medical research and commercially in the production of therapeutic proteins. Hybridoma cells Hybrid cell lines obtained by fusing specific antibody-producing B cells with myeloma cells that are selected for their ability to grow and for an absence of antibody chain synthesis.

2.37.1

Introduction

Bioreactor systems for animal cell cultures can be employed for many purposes on various scales of operation in pharmaceutical production, cell therapies, and tissue engineering.2 These range from simple, small-scale systems for basic research to sophisticated production-scale systems for commercial manufacture. The development of industrial-scale bioreactors was initiated in the mid-1950s to meet the demand for mass production of vaccines. Animal cell-culture bioreactors initially employed stirred tanks containing microcarriers with adherent cells, which were, in principle, an adaptation of homogeneous culture systems used for microbial culture to meet the requirements of mechanically sensitive animal cells. The advent of the monoclonal antibody era in the 1970s then gave rise to the development of various types of bioreactors and culture systems suitable for suspension cultures that realized high product yields per unit volume via improved nutrient supply and waste-product removal. In parallel, alternative specialized systems emerged, including hollow fiber, fluidized bed, and other types of compartmentalized bioreactors based on cell immobilization and perfusion of medium through cell-containing compartments. The fundamental idea was to overcome the major limitations of cell cultivations that caused slow cell growth and low attainable cell densities by providing an environment that allowed the cells to continuously produce the products of interest at high levels. Moreover, a large number of cell retention devices for stirred tank or airlift bioreactors were developed for performing the continuous exchange of media in homogeneous systems. In the 1980s, recombinant DNA technology enabled the production of a variety of protein therapeutics by mammalian cells, which required further advancement and optimization of bioreactors for anchorage-dependent and anchorageindependent cells on a large scale. In recent years, a new trend has emerged, that of tissue engineering, which will play a key role in moving away from conventional surgery by providing new solutions to tissue loss.8,10 In contrast to traditional approaches for treatment of lost tissues or damaged organ functions, tissue engineering enables the replacement of damaged tissues with regenerated tissues that are designed and constructed to meet the needs of individual patients. Tissue engineering, which includes a broad spectrum of technology platforms, is defined as an integrated field based on the principles and methodologies of engineering and life sciences that enable a fundamental understanding of structure–function relationships in normal and pathological mammalian tissues and the development of biological substitutes to restore, maintain, or improve tissue functions. This is an emerging interdisciplinary area of research and technology development that has the potential to revolutionize methods in medicine and leads to the requirement for small-scale design of patient-oriented bioreactors for clinical use.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00086-0

531

532

Bioreactors for Animal Cell Cultures

This article aims to outline the features of bioreactors for cultured animal cells and tissues, covering fundamental principles, engineering considerations, and scale-up or scale-down strategies in various types of bioreactors, including the recent developments in automated bioreactor systems for cell and tissue processing.

2.37.2

Bioreactor Design

A bioreactor system allows control over in vitro culture conditions. The manipulation of cells and tissues has the following requirements: (1) to control the physicochemical environment (e.g., dissolved oxygen (DO) concentration, pH, and shear rate); (2) to ensure aseptic feeding and sampling; (3) to ensure an effective supply of expensive growth factors and medium components; (4) to facilitate monitoring of cell and/or tissue quality (e.g., cell functions and tissue structure); (5) to use materials that are compatible with the cells and processing steps; and (6) to use automated processing steps to increase the reproducibility of bioreactor operations. Therefore, there seem to be two purposes for the functions of bioreactors, which are mechanical tools for realizing operations that are impracticable to carry out by hand (culture systems accompanied by agitation for oxygen supply as well as stimulation by pressure, fluid flow, etc.) and mechanical tools that assist in the manual operation of culture processing (automated cellprocessing systems). The culture strategy for animal cells can be classified into categories in terms of aqueous flow, and supports for cell adhesion and the environmental surroundings of cells, consisting of three phases: gas (O2 and CO2), liquid (medium), and solid (cells and carriers). Static conditions in Petri dishes or flasks can be used for cell proliferation and for seeding onto carriers at the laboratory scale. However, to achieve the spatial uniformity of cell distribution in a culture vessel, mixed (stirred) conditions are required. In addition, mass transfer plays a critical role in the survival and function of cells and tissues in vitro and in meeting the challenges of high cell density cultures, which require adequate supply of nutrients and oxygen, removal of toxic wastes and end products, and release of cytokines. Therefore, bioreactor design for cell growth is mainly required for suspension cultures with and without carriers, on which cells can adhere under culture medium-flow conditions. Bioreactors using agitation in flasks and rotation of vessels can meet the above-mentioned conditions. At laboratory scale, bioreactor systems are characterized by a relatively simple design and low level of instrumentation and control. Traditionally, roller bottles and spinner flasks have been used for small-scale suspension culture, although T-flasks, Petri dishes, multiwall plates, and other stationary culture systems are applicable for suspension cell propagation on a small scale. Spinner flasks are available in the range of 100 ml to several liters and are incubated in humidified CO2 incubators with magnetically driven stirring with either a bar on a central axis or a conical pendulum, achieving oxygen transfer coefficients in the range of 0.1–4 h
1. The general drawback of this design with respect to scale-up is very low oxygen transfer. Another simple culture system is the bioreactor with wave-induced agitation,11 made for suspension cell cultures up to 500 l in disposable plastic bags. Oxygen is transferred through the bags by gas-permeable walls. Due to the rocking motion of the bag on a rocker base, oxygen transfer is somewhat improved compared with spinner flasks. For further scale-up with a higher level of oxygen transfer, stirred tanks that attain homogeneous agitation are most widely used, mainly due to the broad experiences obtained for microbial fermentation, and these are well documented with respect to many of the design requirements. They have been used successfully for cultivating a wide variety of microcarrier-supported cells and cells adapted to growth in suspension, such as hybridoma cells, Chinese hamster ovary (CHO), baby hamster kidney (BHK21), human embryonic kidney (HEK239), and others, with working volumes up to 10 000 l. Stirred tanks are ideally suited to maintaining a uniform physicochemical environment, and for obtaining representative cell samples. Bioreactors for animal cells are designed under specific conditions due to their shear-sensitive natures, compared with those in cultures of bacteria and plant cells. To improve the supply of nutrients, including oxygen under low shear stress and high cell densities, many researchers have paid attention to the design of impellers and spargers for mixing and aeration, respectively. In particular, the introduction of air bubbles from the sparger is a well-known cause of foaming in the medium, which involves proteins, leading to cell rupture by surface tension of bubbles when the bubbles burst at the top surface of the medium in the bioreactor. The creation of foaming in the medium may also cause biological contamination because the bubbles carry fluid to the nonaseptic environment outside the exhaust of the bioreactor. Figure 1 shows bioreactors, which consist of modified impellers with lower shear rates for suspension culture and bubble-free spargers that introduce oxygen into stirred tanks via thin, gaspermeable silicone tubes, or microporous membranes, made of polytetrafluoroethylene (PTFE) or polypropylene. These systems have relatively low oxygen-transfer capacities as compared to those using direct sparging. The scale-up of membrane aeration devices is limited and the largest vessels constructed provide a working volume of several hundred liters. On a relatively small scale of several liters-working volume, an alternative cell culture system that has controllable shear stress as well as being free from bubbles is the circulating medium flow bioreactor, in which cells are retained. Systems that use carriers exhibit added complexity because cells in the inner parts of the carriers are exposed to DO concentrations lower than those at the carrier surfaces. There will therefore be gradients in DO concentration, metabolites, and growth factors across the carriers in the system. Therefore, the potential effects of DO concentration on proliferation are important in the design of bioreactors.

2.37.3

Bioreactors for High Cell Density Cultures

In general, the oxygen requirement of animal cells is not as high as that of microorganisms, and, in the suspension culture of animal cells, it is considered that the oxygen supply in the bioreactor is relatively easy to estimate because spatial homogeneity of oxygen

Bioreactors for Animal Cell Cultures

A

B

533

D

C

E

Figure 1 Photographs of bubble-free stirred tank bioreactors. A 250-ml laboratory-scale stirred tank (A), 1-l bench-scale stirred tank (B), PTFE tube sparger and paddle (C), 20-l production-scale stirred tank (D), and PTFE tube sparger and paddle (E). Photos: Courtesy of Able Co.

supply is feasible. In packed-cell cultures with carriers and tissue cultures, however, the spatial heterogeneity of animal cells in three-dimensional structures is an obstacle to homogeneous oxygen supply. Consequently, a reasonable design approach is that the cells are fixed on the carriers and medium is forced to flow into or around the carriers. The oxygen gradient and the distance over which oxygen can diffuse before decreasing to lethal levels depend upon the oxygen consumption rate per unit volume of packed cells and tissues. Theoretical analyses have been carried out concerning oxygen consumption and diffusion in rectilinear, cylindrical, and spherical geometries to determine how many cells can be supported under given conditions. Depending on the medium circulation pattern and flow rate, there may also be substantial gradients across the reactor. These gradients can be minimized by optimizing the flow distribution within the reactor chamber and by increasing the medium flow rate. The medium flow rate should be adjusted to provide moderate growth by considering the sufficient supply of nutrients. However, oxygen is normally a limiting nutrient due to its low solubility in the culture medium. Higher flow rates for sufficient oxygen supply results in an adverse condition of high shear rates. Thus, several designs for providing a sufficient supply of oxygen at low flow rates have been proposed. One promising system for high cell density cultures is the hollow-fiber bioreactor. Gloeckner and Lemke3 developed a miniaturized hollow-fiber bioreactor system for mammalian cell culture. Compartments holding cells and medium are separated by a semipermeable membrane installed in the bioreactor, and oxygenation of the cell compartment is accomplished using an oxygen-permeable membrane. By means of the geometry of the transparent housing, cells can be observed microscopically during culture. Leukemic cell lines were cultivated up to densities of 3.5  107 cells ml
1 without medium change or manipulation of the cells. This new, miniaturized hollow-fiber bioreactor offers advantages in tissue engineering because of the continuous nutrient supply to cells kept at high density and the retention of added or autocrine-derived growth factors, without disturbing long-term culture in a closed system.

2.37.3.1

Radial Flow Bioreactor Systems

A radial flow bioreactor for cultivating mammalian cells was developed to produce antibodies by the Kirin Brewery Co. in Japan.15 Figure 2 shows the radial flow bioreactor system (Able Co., Japan), which is commercialized for cell cultures. Using recombinant CHO and BHK cells, the performance of the bioreactor was presented. Because the medium flows in a radial direction across the carrier beads, the supply of nutrients and oxygen can be achieved with a low shear stress. The advantages of a fixed-bed bioreactor lie in the fact that the system can be made to be compact, simple, and easy to operate. The medium feed rate is controlled so that glucose is not depleted, and the circulation flow rate is controlled to keep the DO level above 1.0 ppm at the exit from the bioreactor. The flow rate can reach 530 l d
1, with most of the cells being retained in the bioreactor, with the cell concentration on the beads reaching 8  107 cells ml
1beads. In the production of interleukin-6 using BHK cells, the radial flow reactor was scaled up to 5 l in volume. By considering the necessity of high oxygen supply with low shear stress suitable for animal-cell-culture bioreactors, the construction strategy should be described as follows (Figure 3): (1) increase the specific cross-sectional area (AC ¼ A/Vg), which is the ratio of

534

Bioreactors for Animal Cell Cultures

A

B

Inlet of medium

DO probe Outlet of medium

Filter Beads with cells Glass column Filter

Gas mixer

C

Pressure gage PI

Mass-flow controller PH cont. FIC DO in cont.

CO2 Pressure regulator PI

FIC

PI

FIC

DPW controller RS232C

RS232C Selector

RS232C

RS232C

O2

Air

Air filter

Medium chamber

DO out

TEMP PH

Heater

Radial flow reactor 400 ml DO out cont.

Medium

DO In

Stirrer

Electronic balance

PH up Effluent

Incubator Alkali Figure 2 Schematics of radial flow bioreactors. A 250-ml laboratory scale radial flow reactor system (A), growth chamber (B), and outline of total system (C). Courtesy of Able Co.

cross-sectional area (A) to volume in the growth unit (Vg); and (2) increase the effective volume for growth (Ve ¼ Vg/V), which is the ratio of volume in the growth unit to working volume (V). Supplying the required amount of oxygen while avoiding cell damage is a critical factor in successful bioreactor design. Because the radial flow bioreactor has a large cross-sectional area and short flow path for the same bed volume as that of an axial packed-bed bioreactor, it has advantages in terms of a small pressure drop and a low

Bioreactors for Animal Cell Cultures

A

B

Height Z (m)

Cells

u Z (m)

u

535

u DO conc.

DO conc.

Cells

u

Single-column bioreactor u : superficial velocity of medium

Medium

Radial flow bioreactor Capacity of oxygen supply → large

(m s–1)

• Possible to do high density culture • Easy to scale up Figure 3

Conceptual drawings of flows and DO concentrations in column (A) and radial flow (B) bioreactors.

concentration gradient. The critical value of shear stress for cell damage of BHK cells is 1.0–5.0 N m
2. The shear stress loading on cells in a packed-bed bioreactor can be determined as shear stress per unit surface area of matrix, FD/S, given by the following equation: FD DP d ¼ L 6ð1
εÞ S

(1)

where DP, L, d, and ε are the pressure drop, length of flow path, diameter of beads, and the void fraction, respectively. For the condition of laminar flow in a packed-bed bioreactor, the following Kozeny–Carman equation may be derived: q¼

DP kS2 ð1
εÞ2 mL ε3

(2)

where k, q and m are Kozeny constant, the superficial velocity and viscosity of the growth medium, respectively. From Eqs. (1) and (2), the following equation is derived. q¼

ε3 d FD 6kð1
εÞm S

(3)

When the inlet and outlet DO concentrations in a packed-bed bioreactor are 20 and 2 ppm, respectively, and the oxygen uptake rate (QO2) is 5  10
10 mmol-O2 (cell$h)
1, the value of L is calculated to be 6.6 cm at a cell density of 3  108 cells ml
1. Therefore, a radial flow bioreactor with a bed radius of 6.6 cm can supply oxygen to cells without cell damage caused by shear stress less than FD/S ¼ 1.0 N m
2. Scale-up in the direction of the axis of rotation has no limit. Moreover, Bohmann et al.1 developed a radial flow bioreactor with membrane dialysis in which hybridoma cells were cultivated continuously over a period of 6 weeks. The dialysis membrane enabled the supply of low-molecular-weight nutrients and removal of toxic metabolites, while high-molecular-weight nutrients and products (e.g., monoclonal antibodies) were retained and accumulated in the system. The bioreactor performance for the production of antibodies achieved a high level of accumulation, on an average of 100 mg l
1, which was approximately 10 times higher than in fixed-bed cultures without a dialysis membrane. In a modified design of a small-scale bioreactor for applications in toxicology, Hongo et al.4 fabricated a miniaturized radial flow bioreactor to conduct high density culture of hepatocyte cells (Figure 4). This bioreactor comprised a cylindrical vessel with radial flow for cell growth and a medium-conditioning tank for oxygen enrichment of the medium. The medium was subjected to oxygen enrichment in the tank and circulated between the chamber and the tank using a tubing pump. The total volume of medium in the system was 30 ml (working volume). The DO concentration in the bulk liquid could be measured with a DO probe installed in the tank. The cells were cultivated in a compartment of the growth chamber partitioned by two cylindrical meshes with a concentric arrangement (volume of growth chamber: 5 ml). The cells were inoculated uniformly by anchoring on porous beads (hydroxyapatite; particle size 0.6–1.0 mm; porosity 0.8) set in the growth chamber. During the culture, the medium was introduced into the outer cylinder of the vessel in a radial flow manner and discharged from a port situated at the center of the top of the vessel. The superficial velocity in the chamber could be varied by the tube pump (7 ml min
1). This bioreactor system was applied for the culture of hepatocytes (Hep-G2) to evaluate their proliferative ability during growth and stationary phases by changing the inoculum size. Active growth was observed in the case of 2  106 cells ml
1 of inoculum size, giving a final cell density of over 108 cells ml
1 after 12 days. Thus, the radial flow bioreactor system was suitable for maintaining an artificial liver over a long term, suggesting broad applications in pharmacology as an alternative to animal experimentation.

536

Bioreactors for Animal Cell Cultures

A pH down

Controller

B

PC

S

CO2

DOout control S

O2

Air pH up

Media add Alkaline solution

Incubator C

Inoculation and sampling

Fresh media bottle

Inlet

Outlet

Carrier Medium flow

Waist media bottle

RFB

DO sensor

TE DO pH

Medium chamber

Figure 4 Schematics of radial flow bioreactors with immobilized cells. A 5-ml radial flow reactor system (A), photographs of total system (B), and growth chamber of radial flow reactor (RFR) (C). Courtesy of Able Co.

2.37.3.2

Bioreactor for Manufacturing Mechanical Substitutes

Tissue-engineering approaches have used implanted cells, scaffolds, DNA, and protein and/or protein fragments to replace or repair injured or diseased tissues and organs. Despite early successes, tissue engineers have faced challenges in repairing or replacing tissues such as cartilage, bone, blood vessels, and heart valves that serve predominantly biomechanical functions in the in vivo environment. An evolving discipline, called functional tissue engineering, seeks to address these challenges. One approach to tissue engineering is to develop in vitro conditions to ultimately fabricate functional cartilage structures prior to practical implantation.14 Cell-based therapies for damaged or degenerative articular cartilage and intervertebral discs promote tissue regeneration by optimizing cells in a biodegradable matrix. Cell culture is a crucial process for the construction of neo-tissue and is the only means to manipulate cell differentiation and metabolic function prior to implantation. The culture of articular cartilage includes the general principles of functional tissue engineering that may produce tissues for the replacement or repair of load-bearing structures in the body. Various physical and/or physicochemical factors influence the development, growth, and metabolic function of cartilage. Dynamic compressive loading of cartilage disks introduces deformations and other changes within the tissue, such as hydrostatic pressure, interstitial fluid flow, and streaming potential.7,13 The growth of chondrocytes on scaffolds has been augmented by the use of medium flow-type bioreactors to increase the transport of nutrients toward and wastes away from cells. Various designs have been aimed at increasing mass transfer rates, including shaker flasks and stirred bioreactors using magnetic impellers and rotating bioreactors. The limitations of such designs include unregulated and unknown fluid flows through the construct during culture. Fluid flows of 1 mm s
1 have been shown to accelerate cartilage extracellular matrix assembly. This phenomenon has not been exploited in the process of bioreactor culture. Pazzano et al.9 constructed a pressure/perfusion three-dimensional culture system that, while it is not regarded as a model for the evaluation of physiology and pathophysiology for mature cartilage tissues, has distinct advantages for evaluating the regulation of anabolic and catabolic functions of cells in a deformation-free pressure environment. Bioreactors are one of the key technologies that allow alteration of culture conditions to be programmed for maximizing tissue regeneration. Based on the latest biochemical, molecular, and clinical notions, bioengineering knowledge will provide secure treatments to restore functional tissue. As shown in Figure 5, Mizuno and co-workers7,13 have developed a novel hydrostatic pressure/ perfusion (HP/P) culture system, with externally applied hydrostatic pressure and oxygen tension, and interstitially generated osmotic and swelling pressures that influence the metabolic function of cartilage disks, and improve the quality of neo-cartilage, providing an automated affordable system for clinical applications. They evaluated the effects of HP/P on cellular properties, viability, and proliferation of human chondrocytes seeded in gel/sponge constructs, leading to beneficial results for building high-density three-dimensional tissues.

Bioreactors for Animal Cell Cultures

A

537

B

C

Figure 5 Photographs of a hydrostatic pressure/perfusion (HP/P) culture system. Total system (A), growth chamber (B), and indicated signal of hydrostatic pressure (C). Courtesy of Takagi Industrial Co.

2.37.4

Automation of Cell Processing Toward Clinical Application

Innovative techniques for cell and tissue processing, on the basis of tissue engineering, have been developed for therapeutic applications.6 Cell expansion through subculturing and development of functionality through ex vivo cultures are core processes used to produce cultured tissues. In manufacturing, strict management against contamination and human error is necessary, due to the unsterilizable nature of the products and the complexity of culture techniques, respectively. In addition, the development of processing systems for cell and tissue cultures is one of the critical issues for ensuring stable processes and quality of therapeutic products. In the case of the manufacturing process for cultured cells and tissues aiming at autologous transplantation, the bare minimum of biopsies is harvested from patients to prepare starting cell populations as raw materials. The cells isolated are then propagated in a series of subcultures, performed in a batchwise manner. After the quantity of cells is expanded sufficiently, the suspended cells are administered to the respective patients or used in tissue cultures for the reconstruction process to form biologically functional substitutes as final products. Thus, the products for therapeutic use are self-originated cells and tissues for patients. An alternative strategy for future cell- and tissue-based therapies is the application of embryonic stem (ES) cell-derived grafts. The possibility of cell therapy using ES cells offers the particular advantages of prolonged proliferative capacity and great versatility in the lineages that can be formed in cultures. However, their application has faced many challenges for realizing efficient differentiation into the desired cell types, maintaining genetic stability during long-term culture, and ensuring the absence of potentially tumorigenic cell populations in the final products. Many researchers have reviewed methodologies for embryoid body formation to initiate stable differentiation toward desired cell lines, recent Good Manufacturing Practice (GMP) regulatory issues for propagation toward therapeutic usage, and assessment of the advances made in developing scalable culture systems. In terms of culture systems, the variance in culture properties induced by environmental fluctuations during manually operated culture of target cells requires automated cell processing to reproducibly control cell density, fluid flow, centrifugal forces, pH, and temperature for the dissociation and inoculation of the cells. Some of the current automated culture systems were developed by adopting or modifying commercially available cell-processing systems. The processing systems are classified into two types, namely a ‘sealed-chamber culture system’ and a ‘sealed-vessel culture system‘, as shown in Figure 6. The sealed-chamber culture system consists of aseptic chambers, where culture vessels can be opened temporarily, while the sealed-vessel culture system is designed to perform cell and tissue processing without opening the aseptic culture vessels during operation. Here, the culture space, which maintains the aseptic atmosphere in contact with the cells and medium, is substantially the same as the chamber in the sealed-chamber culture system or the culture vessel in the sealed-vessel culture system, respectively. Furthermore, as shown in Figure 7, cell processing consists of several culture procedures, including seeding, incubation, monitoring, medium change, and packaging. For automation throughout the process, the continuity of handling in a closed system is required. However, most of the cultures for clinical trial applications require flexible and multipurpose procedures in a closed system, resulting in discontinuity in the procedures in a sealed-vessel culture system. William and coworkers12 performed automated cultures of human mesenchymal stem cells using a sealed-chamber culture system with a robot arm. This system performed parallel and sequential procedures in cultures using a robot arm, thereby underlining the importance of automation systems in improving the capability and cost-effective scalability of product manufacture. Moreover, the importance of process engineering for manufacturing cells and tissues should be mentioned, realizing not only a reduction in cost but also an increased process stability enabled by the development of an automated culture system.

538

Bioreactors for Animal Cell Cultures

Sealed-chamber culture system

Sealed-vessel culture system

Chamber 3 Chamber 2

Vessel 1

Chamber 1

Vessel 2

Vessel

Culture device

Figure 6

Categories of culture systems.

Procedure for culture Seeding (and Passage)

Automation

Cell harvesting from a prior culture Centrifugation Preparation of cell suspension Cell counting Seeding Incubation Monitoring Medium change Packaging Figure 7

Vessel 3

Robot arm automated for universal operations

Discontinuity of handlings in closed system (demand of remote handling) Sequential procedure automated in closed system Operational flexibility using one arm Multipurpose application

Automated culture procedure and features assisted by robot-arm system.

The term ‘automation‘ used in most of the culture systems developed to date implies that manipulation is directed by the operator or through preset operations, giving a passive or inflexible system in response to the constantly fluctuating nature of tissue culture processes. Thus, an intelligent culture system is required to realize autonomous operations by incorporating sophisticated tools for monitoring fluctuating culture properties. Three specific design problems exist: process components (cell source, cell signals, scaffolds, and bioreactors); process requirements (GMP, bioprocessing, preservation, transport, storage, and regulators); and process functions (organogenesis, functionality, host tissue integration, immunoacceptance, and longevity). These design problems must be overcome for manufacturing. The application of real-time online monitoring techniques and process control strategies is expected to enhance the ability to operate bioprocesses with desired reproducibility and high product quality. Therefore, in vitro culture information required for bioprocess control can be obtained by methods including process considerations, monitoring and analytical tools, and experimental design. The successful application of these tools will result in time- and cost-effective, intelligent culture technologies. The developments required to realize a fully automated culture system are based on control engineering for introducing a decision-making step for culture scheduling. Medinet Co. in Japan established a sealed-vessel culture system that enabled the automated scale-up of the culture volume. Figure 8 shows a conventional system with manual operations for lymphocyte isolation from a patient‘s blood plasma, cell seeding, activation by the antibodies displayed on the culture surface, expansion in a T-flask and bag, and recovery of the cells for shipment. An automated system was recently designed to perform the operations of cell activation and expansion. The automated scale-up was performed using a flexible culture bag, a mechanical stage that was able to alter the size of the culture volume by simply shifting the height of the stage, and a digital camera for capturing images of the cell suspension to monitor the culture conditions. In the independent experiments, culture monitoring was performed automatically by capturing images from the bottom of the vessel. The times for cell activation by the antibody displayed on the culture surface in the early culture phase and for aggregation during the exponential phase were determined by evaluating the variation in size of single cells and the frequency of aggregation, respectively. In the future, decision making regarding these times would allow stepwise increments of the culture volume to be accomplished by positional shifts of the mechanical stage and the administration of additional medium, thereby realizing the autonomous scale-up in a one-bag culture without transfers between vessels, as is typically required by conventional processing during the expansion of T cells. Further development of a cell-processing system was conducted in subcultures of human skeletal muscle myoblast cells.5 The myoblasts achieve myotube formation when cell–cell contact occurs at high cell density, leading to lower growth ability in the subsequent culture after passage. Thus, the time for passage is critical. The monitoring of the degree of confluence (the extent of occupied cell area on the culture surface) by image processing helped the decision to perform the passage intelligently, leading

Bioreactors for Animal Cell Cultures

Manual operation

539

Automated operation

Cell isolation Cell activation

Cell seeding and activation Medium addition Observation Cell expansion Passage Medium addition Observation

60 cm2 Cell activation and expansion Medium addition Cell expansion Scale-up in one bag 300 cm2 Observation

Cell expansion

Cell recovery

1300 cm2 Camera system

Products Figure 8

One-bag culture on mechanical stages with stepwise scale-up

Schematics of culture procedure and photographs of a bioreactor system for lymphocyte culture. Courtesy of Medinet Co.

to populational improvement of capable cell growth in the subsequent culture. Therefore, integrated tools for cell manipulation, monitoring, and simulation permit us to establish an intelligent system to perform serial subcultures, operated in an autonomous way. These challenges may contribute to technical improvements as well as management of process validation during the manufacture of cultured tissue products.

2.37.5

Concluding Remarks

Bioreactors are a core element whenever we aim to produce high-value materials in biological processes, and the case of animal cell cultures is no exception. For the production of soluble metabolites secreted by animal cells, animal-cell bioreactors have been designed based on the configurations developed for microbial processes. From a practical viewpoint, the major features to be considered are the mechanical fragility and low growth rate of animal cells. These limitations led to the strategy of high cell density culture by holding the cells within bioreactors with an external medium flow. In recent decades, on the other hand, technologies for cell and tissue therapies have emerged in the field of regenerative medicine. For such purposes, the targets are ex vivo cultivation and expansion of human cells in undifferentiated or differentiated states, where the products of interest are the cells and/or tissue themselves. Therefore, the bioreactor used in these cultures should be operated under strictly controlled conditions to produce cells or tissues with quality and functions that meet specific clinical requirements. Sophisticated bioreactor systems will offer reliable and reproducible processes for human cell and tissue cultures by incorporating the automated handling of cells and nutrients, as well as the monitoring of cellular and process state parameters.

References 1. Bohmann, A.; Pötner, R.; Schmieding, J.; et al. The Membrane Dialysis Bioreactor with Integrated Radial-flow Fixed Bed - a New Approach for Continuous Cultivation of Animal Cells. Cytotechnology 1992, 9, 51–57. 2. Fenge, C.; Lüllau, E. Cell Culture Bioreactors. In Cell Culture Technology for Pharmaceutical and Cell-Based Therapies; Ozturk, S. S., Hu, W.-S., Eds., Taylor and Francis: New York, NY, 2006; pp 155–224. 3. Gloeckner, H.; Lemke, H. D. New Miniaturized Hollow-fiber Bioreactor for in Vivo like Cell Culture, Cell Expansion, and Production of Cell-derived Products. Biotechnol. Prog. 2001, 17, 828–831. 4. Hongo, T.; Kajikawa, M.; Ishida, S.; et al. Three-dimensional High-density Culture of HepG2 Cells in a 5-ml Radial-flow Bioreactor for Construction of Artificial Liver. J. Biosci. Bioeng. 2005, 99, 237–244. 5. Kino-oka, M.; Chowdhury, S. R.; Muneyuki, Y.; et al. Automating the Expansion Process of Human Skeletal Muscle Myoblasts with Suppression of Myotube Formation. Tissue Eng. C 2009, 15, 717–728. 6. Kino-oka, M.; Taya, M. Recent Developments in Processing Systems for Cell and Tissue Cultures toward Therapeutic Application. J. Biosci. Bioeng. 2009, 108, 267–276. 7. Mizuno, S.; Tateishi, T.; Ushida, T.; Glowacki, J. Hydrostatic Fluid Pressure Enhances Matrix Synthesis and Accumulation by Bovin Chondrocytes in Three-dimensional Culture. J. Physiol. 2002, 193, 319–327. 8. Nerem, R. M.; Sambanis, A. Tissue Engineering: From Biology to Biological Substitutes. Tissue Eng. 1995, 1, 3–13.

540

Bioreactors for Animal Cell Cultures

Pazzano, D.; Mercier, K. A.; Moran, J. M.; et al. Comparison of Chondrogenesis in Static and Perfused Bioreactor Culture. Biotechnol. Prog. 2000, 16, 893–896. Pötner, R.; Nagel-Heyer, S.; Goepfert, C.; et al. Bioreactor Design for Tissue Engineering. J. Biosci. Bioeng. 2005, 100, 235–245. Singh, V. Disposable Bioreactor for Cell Culture Using Wave-induced Agitation. Cytotechnology 1999, 30, 149–158. Thomas, R. J.; Chandra, A.; Liu, Y.; et al. Manufacture of a Human Mesenchymal Stem Cell Population Using an Automated Cell Culture Platform. Cytotechnology 2007, 55, 31–39. 13. Watanabe, S.; Inagaki, S.; Kinouchi, I.; et al. Hydrostatic Pressure/perfusion Culture System Designed and Validated for Engineering Tissue. J. Biosci. Bioeng. 2005, 100, 105–111. 14. Wendt, D.; Jakob, M.; Martin, I. Bioreactor-based Engineering of Osteochondral Grafts: From Model Systems to Tissue Manufacturing. J. Biosci. Bioeng. 2005, 100, 489–494. 15. Yoshida, H.; Mizutani, S.; Ikenaga, H. Scale-up of Interleukin-6 Production by BHK Cells Using a Radial-flow Reactor Packed with Porous Glass Beads. J. Ferment. Bioeng. 1997, 84, 279–281. 9. 10. 11. 12.

2.38

Metabolic Regulation Analysis and Metabolic Engineering

K Shimizu, Kyushu Institute of Technology, Iizuka, Japan; and Institute of Advanced Biosciences Keio University, Tsuruoka, Japan © 2011 Elsevier B.V. All rights reserved. This is a reprint of K. Shimizu, 2.30 - Metabolic Regulation Analysis and Metabolic Engineering, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 407-420.

2.38.1 Introduction 2.38.2 Metabolic Engineering Practice 2.38.2.1 Reduction of the Formation of Acetate Byproduct 2.38.2.2 Biofuels and Related Alcohol Production 2.38.2.3 Lactic Acid Fermentation 2.38.2.4 Succinic Acid Fermentation 2.38.2.5 Amino Acid and Other Fermentations 2.38.3 The Effect of Single-Gene Knockouts on the Metabolism 2.38.4 Global Regulators in Relation to the Cultural Environment 2.38.5 The Systems Biology Approach 2.38.6 Conclusion Nomenclature Appendix A: Global Regulators and Their Regulated Genes: D Activate, L Repress References

541 543 543 543 544 545 546 546 548 550 551 552 554 554

Glossary Biofuels The energy sources generated by microorganisms from biomass. Catabolism and anabolism The energy (ATP) generation mechanism and cell synthesis mechanism. Global regulator Transcription factor which regulates metabolic pathway genes. Metabolic engineering The technology of modifying the metabolism by gene manipulation for efficient target metabolite production. Metabolic flux analysis The quantitative analysis for the network of metabolic reaction rates from substrate uptake to metabolite production.

2.38.1

Introduction

The main goal of metabolic engineering is to improve the metabolic phenotype through genetic modifications.1 Most of the recent approaches to metabolic engineering have aimed to improve a particular biosynthetic capacity through engineering of the target pathway based on rational assumptions for its improvement. The resulting phenotypes are, however, often suboptimal and unsatisfactory because of the distant effects of genetic modifications or unknown regulatory interactions. It is therefore strongly desirable to take into account the overall metabolic regulation mechanism for metabolic engineering. A wealth of information is available on local genetic regulation and the biochemistry and physiology of cellular metabolism, but surprisingly little is known about the overall regulation of metabolism. Moreover, high-throughput techniques for transcriptomics, proteomics, metabolomics, and fluxomics have the potential to disclose the mechanism, but most of them provide a snapshot of one stage, and the methodology for interpreting the different levels of information is not yet established.2 We are still far from understanding regulatory phenomena from the perspective of the whole cell system. It is becoming even more important to analyze the cell in vivo and treat it as an entire system for post-genomic research (Figure 1). To accomplish this goal, it is important to investigate the global regulation of metabolism, taking different levels of information into account. It should be noted that gene expression, protein expression, and the concentration of intracellular metabolites can be directly measured, while metabolic fluxes or reaction rates cannot be directly detected, and must be estimated. The metabolic pathways of a variety of organisms have been well documented in several Web-accessible databases. Some wellknown examples include the KEGG and MetaCyc databases. The EcoCyc database (a subset of MetaCyc) is available for use with Escherichia coli. The K12 strain of E. coli has been sequenced, and genome-wide mRNA expression profiling is also available for this strain. The profiling of gene expression is useful in the analysis of metabolic regulation at the genomic level. It can be used to compare the global changes in gene expression that occur in response to environmental stimuli and genetic changes. This analysis can provide important information about cell physiology and has the potential to identify connections between regulatory proteins and unknown metabolic pathway genes. Biotechnological research and applications require knowledge of how genes work at the genomic level. Therefore, microarrays have been extensively used to study global regulation.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00088-4

541

542

Metabolic Regulation Analysis and Metabolic Engineering

Metabolic regulation analysis Genetic perturbation Environment

Cell DNA

mRNA

Transcription

Metabolism

Protein Translation

Enzymatic reactions

Genomics

Figure 1

Transcriptomics

Proteomics

RT-PCR, DNA microarray

2DE + TOF-MS Enzyme activity

Metabolomics and fluxomics 13C-experiment

NMR, GC–MS CE–TOF/MS

Conceptual figure for the metabolic regulation of the cell.

Two-dimensional electrophoresis (2DE) was proposed as early as 1975 to be the most efficient method of separating complex protein mixtures, in order to analyze global patterns of gene expression at the protein level. One major outcome of proteomic studies has been the establishment of 2DE databases for many organisms. These databases simplify the analysis of gene expression in response to genetic or environmental alterations at the protein level. Recent progress in mass spectrometry (MS) technology provides shotgun proteomics, and protein chips are also available. These can be used to identify protein–protein interactions. The E. coli proteome has recently been reviewed thoroughly.3 Metabolomics is an emerging area of research that can be used to gain insight into cellular function. In particular, the combination of capillary electrophoresis and mass spectrometry (CE–MS or, with time-of-flight mass spectrometry, CE–TOF MS), which uses only a small amount of sample, has been shown to be a more powerful technique for high-throughput quantitative analysis than gas chromatography–mass spectrometry (GC–MS), liquid chromatography–mass spectrometry (LC–MS), matrix-assisted laser desorption/ionization–mass spectrometry (MALDI–MS), etc.4 The most important information for understanding the complex metabolic control mechanism of the whole cell may be the metabolic flux distribution (MFD) in the central metabolism, as this is the manifestation of gene and protein expression (or enzyme activities) and the concentrations of intracellular metabolites. In principle, metabolic flux analysis is based on the mass conservation of key metabolites. The intracellular fluxes are calculated from the measured fluxes by applying mass balances to these intracellular metabolites. The number of measurable extracellular fluxes is limited, and stoichiometric constraints often lead to an underdetermined algebraic system. Therefore, cofactor balances are sometimes required to be introduced into the stoichiometric model or an objective function for optimization has to be introduced. The central metabolic pathway has both anabolic and catabolic functions, as it provides cofactors and building blocks for macromolecular synthesis (anabolism) as well as energy generation (catabolism). The optimization of the metabolic fluxes may be made in terms of catabolism or anabolism, or both, under the constraint of the stoichiometric equations. Flux balance analysis (FBA) has been extensively used to predict steady-state metabolic fluxes in order to maximize the cell growth rate.5 However, the accuracy of the flux calculations depends on the validity of the cofactor assumptions and depends on an appropriate choice of the objective function. The presence of unknown reactions that generate or consume the cofactor might invalidate the assumption that the concentration of the cofactor remains in balance, and the selected objective functions may not be appropriate, or their validity may be limited to certain states of the cell. An alternative approach is to use isotopic tracing, where isotopically labeled substrates can be introduced to the cell, and the labeled carbon atoms will be distributed throughout its metabolic network. The final isotopic enrichment in the intracellular metabolite pools can then be measured. The amino acids in biomass hydrolysates are much more abundant than their precursors in the central metabolism. It is therefore easier to deduce the labeling patterns of the intracellular metabolites from the labeling patterns of the proteinogenic amino acids, based on precursor–amino acid relationships. The tracer experiments generally use either nuclear magnetic resonance (NMR) spectroscopy or GC–MS. Currently, these tracer techniques, in combination with direct extracellular flux measurements, are considered to be a powerful method for obtaining intracellular MFDs using only a few modeling assumptions. Alternatively, flux ratio analysis can be used to constrain the flux at important branch points. One limitation when using the isotopomer distribution of the proteinogenic amino acids is that the analysis can only be applied to the steady state. In order to extend the analysis to the industrially important batch culture or to dynamic analysis, the isotopomer distribution of the intracellular metabolites must be determined. This can be measured using CE–TOF MS, GC/MS–MALDI/TOF MS, LC–MS/MS, or LC–MS. Note that since the pool sizes of primary metabolites are orders of magnitude smaller than those of the proteinogenic amino acids, the isotopic steady state is attained more rapidly. Both computational FBA and experimental MFD methods for flux determination have been shown to be quite useful in understanding cell physiology. FBA can treat several hundred metabolic pathways, while experimental approaches are useful in the primary treatment of the main metabolic pathways. The flux result must be consistent with, or integrated with, other information such as the direction of metabolic fluxes, allosteric regulation of enzymes by metabolites, and transcriptional regulation, as the

Metabolic Regulation Analysis and Metabolic Engineering

543

functional behavior of a metabolic network is the result of the interactions between gene expression, protein expression, and intracellular metabolite concentrations. In the following sections, some metabolic engineering practices are reviewed and it is considered how a systematic approach may be practiced on the basis of metabolic regulation analysis.6

2.38.2

Metabolic Engineering Practice

Here, current metabolic engineering practices for the specific production of metabolites are briefly explained, where the name of genes, enzymes, and metabolites are given in the nomenclature section.

2.38.2.1

Reduction of the Formation of Acetate Byproduct

One of the main obstacles to producing particular metabolites is the formation of byproducts such as acetate in E. coli or ethanol in yeast. Rapid aerobic growth of E. coli on glucose, gluconate, pyruvate, lactate, glucuronate, and serine (but not on glycerol or fructose) and conditions in which there is an excess of glucose are characterized by the formation and excretion of acetate. This phenomenon is referred to as overflow metabolism or the bacterial Crabtree effect, and its mechanism has been investigated in an attempt to reduce the formation of acetate.7 Escherichia coli excretes 10–30% of its carbon flux from glucose to acetate in glucose-containing media even when the culture is fully aerated. The excretion of acetate is a major limitation for the high cell density that is required for culture to yield high product concentration. Under aerobic conditions, acetate is generated from pyruvate either by oxidative decarboxylation by the PDHc followed by the conversion of AcCoA to acetate by Pta and Ack, with concomitant formation of ATP at the Ack reaction, or by decarboxylation to acetate directly by pyruvate oxidase (Pox) (Figure 2). The latter reaction is utilized during the transition from exponential growth to the stationary phase, under the control of the global regulator RpoS. After glucose depletion or at low concentrations, acetate is assimilated by AcCoA synthetase (ACS) with concomitant conversion of ATP to AMP and pyrophosphate, where acs is also under the control of RpoS. The AcCoA thus formed is metabolized by both the TCA cycle and the glyoxylate pathway for energy generation and cell synthesis. Since acetate is produced mainly through the actions of Pta and Ack, pta mutant has been created, which showed decreased acetate production. However, pta gene knockout leads to high pyruvate production, which is also undesirable. It is shown that the Pta–Ack pathway dominates in the exponential growth phase, and the Pox pathway dominates for acetate production in the stationary phase, and that the former pathway is repressed under acidic conditions, whereas the Pox pathway is activated. Aerobic acetate production is significant in particular at higher growth rates, and is a manifestation of the imbalance between glucose uptake and the demands for both biosynthesis and energy production. The most common arguments are that the glucose uptake rate is improperly controlled and that the activity of the TCA cycle is limiting. One way to lower acetate production is to construct mutants with modified glucose uptake rates. Using this strategy, the TCA cycle can handle all the AcCoA produced by the glycolytic pathway, and thus can eliminate acetate formation. Acetate production may also be reduced through decreasing the glucose uptake rate by decreasing the glucose concentrations in the fermentor in the fed-batch cultivations or by decreasing the expression of ptsG, which encodes the glucose-specific enzyme Ⅱ (EⅡ CBglc) of the phosphotransferase system (PTS) through expression of mlc gene, where Mlc represses ptsG, or by ptsG mutation. Alternatively, enhancement of the TCA cycle and the glyoxylate shunt also apparently reduces acetate production. It has been reported that constitutive expression of glyoxylate pathway genes may reduce the acetate production. The glyoxylate shunt contains two enzymes, isocitrate lyase (Icl) (encoded by aceA) and malate synthase (encoded by aceB), of which genes are located on the aceBAK operon, with aceK coding for isocitrate dehydrogenase kinase/ phosphatase. The transcriptional regulation of this operon involves many factors including IclR, FadR, Cra, ArcAB, and HimAB. The expression of aceBAK is induced during growth on either acetate or fatty acids, but induction is repressed in the presence of glucose, glycerol, or pyruvate. Mutation in fadR results in transcriptionally increased expression of aceBAK even for such carbon sources and affects the metabolism (Figure 3). Another way of reducing acetate production is to enhance anaplerosis via increasing PEP carboxylase (Ppc) and/or glyoxylate pathway enzyme levels.

2.38.2.2

Biofuels and Related Alcohol Production

Because of the pressure on environmental protection and limited resources of fossil fuels, biofuel production from crops such as corn and sugarcane has been paid much attention all over the world. Moreover, owing to the competition to food in the case when crops and sugarcane are used, biofuel production from cellulose has also been paid much attention. Metabolic engineering will play significant roles for the efficient production of biofuels such as ethanol, butanol, propanol, and biodiesel. Traditionally, Saccharomyces cerevisiae and Zymomonas mobilis have been exclusively used for ethanol fermentation. However, because of recent high demand for efficient bioethanol production from biomass for fuels, other microorganisms such as E. coli, Klebsiella oxytoca, and Pichia stipitis have also been investigated for use in ethanol production.8 It may be considered to replace the native fermentation pathways in E. coli with the ethanol-forming pathway from Z. mobilis. However, this strain restricts carbon flow into the oxidative TCA cycle, thereby limiting the cell growth and the rate of ethanol production. Both the growth of cells and the production of ethanol can be improved by the expression of NADH-insensitive citrate synthase (citZ) from Bacillus subtilis. Although ethanol has been paid much attention with respect to its fuel potential, it may not be an ideal replacement for gasoline as it has a high water content and a low energy density relative to gasoline. Higher alcohols (C4 and C5) have energy densities

544

Metabolic Regulation Analysis and Metabolic Engineering

Glc ATP

PEP

(ptsI, H, crr, ptsG) Hxk

Glk (glk)

Pyr

Pgm

G1P

NADP+ NADPH

ADP

G6P

Pgl (pgl)

6PGL

G6PDH (zwf)

NADP+

R5P

GAP

Rpi (rpiA, B)

F6P ATP

Tal (talA,B)

ATP

(fbp) Fbp

Pfk (pfkA, B)

Ru5P

Tkt (tktB)

E4P

ADP

ADP

DHAP

Tkt (tktA)

S7P

2K3DPG

NAD+

GAPDH (gapA, C)

Edd (edd)

X5P

GAP

Tpi (Tpi)

6PGDH (gnd)

Rpe (rpe)

F1,6BP Fba (fba)

6PG

NADPH

Pgi (pgi)

Eda (eda) NADH

1,3BPG ADP

(pgk) Pgk ATP

3PG (pgm) Pgm

2PG (eno) Eno

PEP Ppc (ppc)

ADP

ADP

(ppsA) Pps

Pyk (pykF, A) ATP

ATP

PYR

Pox

NAD+

PDK (aceE, F, lpdA) NADH

ADP Pck (pckA) ATP

AcCoA

Pta (pta)

ADP

Ack (ackA)

NADP+

Mez (mez)

ATP

AcP

NADPH

Acetate

CS (gltA) ACS

NADH NAD+

OAA

CIT Acn (acnA, B)

MDH (mdh)

ICIT

MAL

NAD(P)+

GOX

MS (aceB)

Fum (fumABC)

FUM

FADH2

SDH (sdhABCD) FAD

NAD(P)H

(susAB, lpdA) α KGDH

FAD

NAD+

(susCD) SUC

SUC

GTP

Figure 2

ICDH (icdA)

KG

Frd (frdABCD) FADH2

IC1 (aceA)

SucCoA

NADH

GDP

Metabolic pathway for acetate formation.

similar to gasoline and are less volatile than ethanol, and thus may be considered to be the next generation of biofuels.9 1-Butanol is hydrophobic and its energy content is similar to that of gasoline. Moreover, its vapor pressure is lower than that of ethanol. Thus, 1butanol may be a substitute for, or supplement to, gasoline as a transportation fuel. Microbial production of 1-butanol has been achieved using Clostridium acetobutylicum, which is a gram-positive anaerobe and produces such byproducts as butylate, acetone, and ethanol. It is not suited to industrial application because of its slow growth rate, spore-forming life cycle, and the byproducts mentioned above. It may be considered to produce 1-butanol using recombinant E. coli, in which 1-butanol production pathway genes from AcCoA such as thl, hbd, crt, bcd, etfAB, and adhE2 were cloned and expressed in E. coli. Moreover, it may be considered to knock out such genes as adhE, ldhA, frdBC, fnr, and pta to increase the yield of 1-butanol. Isopropanol produced by engineered E. coli is another alternative biofuel.9

2.38.2.3

Lactic Acid Fermentation

Lactic acid is widely used in the food industry and has attracted interest recently for the production of biodegradable plastics from renewable resources. In the past, lactic acid bacteria such as Lactobacillus strains have been employed for lactic acid fermentation. However, lactic acid bacteria require complex nutrients and are poorly able to utilize pentoses. Metabolically engineered E. coli

Metabolic Regulation Analysis and Metabolic Engineering

545

Glucose Glucose transport proteins PEP, ATP Glk 1.07 PtsH 1.48, PtsI 1.76, Crr 1.37 PYR, ADP NADPH

NADP 0.68 G6P Pgi 1.05

6PGDH 1.10

ATP

CO2

Tal 1.19

Ribu5P

ADP S7P

F1,6P

X5P

G3P

Tpi 1.08 DHAP NAD

HIS

GAPDH 1.47

NADH

G1,3P ADP

TRY TYR PHE

ATP

SER GLY THR ILE

R5P

ED 0.88

Fba 1.15

Pgk 1.21

NADP NADPH

E4P

F6P Pfk 1.23

6PG 0.94

G6PDH 1.23

3PG VAL, ALA ADP PEP

CO2 ADP Ppc 0.83 ATP

PYR CO2 Pyk 0.93 ADP

Pck 1.54 ATP

2.75 OAA ASP ASN MDH 1.27 Mae 1.67 LYS NADPH NADP MET THR 4.33 MAL NAD NADH Fum 3.26

MS 1.87

Glyo

NAD NADH

ADP ATP Acetate Ack 0.90

0.25 AcCoA

CS 1.28

LEU Fatty acid

CIT Acn 1.44 Icl 3.72

ICIT 2.06

CO2 ICDH 0.88

FUM FADH FAD

0.65

CO2

Sfc 0.77

0.80

ATP

SUC Suc-CoA

1.20

2-KG

CO2

NAD NADH

Energy metabolism proteins: GapA 1.50, Pgk 1.23, SucA 1.98, AtpA 1.41, AtpC +, AtpD 1.19, CyoD 2.21 Amino acids biosynthesis proteins: TrpD 1.67, AsnB 1.40, AroG 1.28 Fatty acids biosynthesis: AccB 0.73, FabD 0.66

NADP NADPH GLU GLN ARG PRO

Cell process: GreA 1.18, RplI 1.34 Others: DnaK 1.54, UspA 1.22, AceK 1.70, OppA0.67, SgaH 0.21

Figure 3 Metabolic regulation in fadR mutant E. coli. Red arrow indicates up-regulations of the flux, while blue arrow indicates down-regulation as compared to wild type. Oval indicates up-regulation of metabolite concentration, while rectangular indicates the down-regulation of metanolite as compared to wild type.

has, therefore, attracted recent attention. Escherichia coli lacking ptsG, pfl, and ldhA and expressing a gene encoding L-LDH can assimilate hexoses and pentoses and produce L-lactate from a sugar mixture. Furthermore, metabolically engineered E. coli strains that lack the genes encoding Pfl, Frd, ADH, and Ack and possess either D-LDH or L-LDH activity may be considered.

2.38.2.4

Succinic Acid Fermentation

Succinate has a variety of applications, including use as a surfactant, detergent extender, foaming agent, ion chelator, and food additive, and it can be used as a precursor for a variety of chemicals such as tetrahydrofuran and 1,4-butanediol. Under microaerobic or anaerobic conditions, a mixed-acid fermentation occurs in E. coli, forming such metabolites as lactate, acetate, formate, ethanol,

546

Metabolic Regulation Analysis and Metabolic Engineering

CO2, and succinate. Since most of the metabolites are produced from pyruvate, while succinate is produced from PEP through OAA, one approach to the selective production of succinate is to reduce pyruvate production by disrupting ptsG, which encodes EⅡ BCglc of the PTS, pykF, A genes. Simultaneous overexpression of the ppc gene of Sorghum vulgare (which encodes PEP carboxylase that is resistant to feedback inhibition by malate) and of the pyc gene of Lactococcus lactis (which encodes PYR carboxylase) increased the succinate yield. Additional disruptions of such genes as ldhA, ackA, and pta increased the succinate yield. Further improvement can be made by deletion of the adhE and ldhA genes, and by overexpression of the pyc gene obtained from L. lactis, where it increased the yield. The result indicates that inactivation of adhE is effective in directing reducing equivalents toward the formation of succinate. The glyoxylate pathway may be utilized as an alternative route to succinate formation by deleting the iclR gene, where IclR represses the aceBAK operon. The utilization of the glyoxylate pathway together with the Ppc pathway for succinate production requires less reducing equivalents. The strain lacking the genes iclR, ldhA, adhE, and ackA, and pta genes, and which overexpressed the pyc gene may increase a succinate yield. In order to overcome the disadvantages of anaerobic succinate fermentation, such as low growth rates and the formation of a variety of byproducts, aerobic succinate fermentation using metabolically engineered E. coli can be considered. The strain that lacks the icd, sdhAB, and iclR genes inactivate the oxidative TCA cycle while retaining the glyoxylate pathway. In addition, some of the acetate-producing pathway genes such as pta-ackA and poxB can be deleted.

2.38.2.5

Amino Acid and Other Fermentations

Amino acid fermentations have been exclusively studied using metabolically engineered E. coli and Corynebacterium glutamicum. In particular, fermentations of glutamic acid, lysine, threonine, isoleucine, proline, tryptophan, phenylalanine, and tyrosine have been reviewed recently.9 Fermentations of urines and pyrimidine and their nucleosides and nucleotides have also been reviewed.9 Aromatic compounds can also be produced using pathway engineering in E. coli. The production of vitamins by fermentation has also been investigated. Among them, popular vitamins including vitamin B12, riboflavin, vitamin C, biotin, pantothenic acid (vitamin B5), thiamine (vitamin B1), pyridoxine (vitamin B6), and carotenoids10 have been produced.

2.38.3

The Effect of Single-Gene Knockouts on the Metabolism

In order to understand the role of each gene on the metabolism, it is quite useful to study the effect of specific gene knockouts or that of the change in the environment of the culture on the metabolism, based on data regarding the MFD obtained by 13C labeling experiments, gene expression, protein expression (enzyme activities), and the concentrations of intracellular metabolites.11 While some single-gene knockout mutations in central metabolism preclude the growth on glucose, a majority of such mutations seem to be potentially compensated by the use of alternative enzymes or by the rerouting of carbon fluxes through alternative pathways, resulting in a robust phenotype such as cell growth. For example, if either the ppc or pckA gene, which code for the anaplerotic and gluconeogenic reactions, respectively, were knocked out, the formation of OAA would be reduced, which in turn would activate the glyoxylate pathway. The regulation mechanism is as follows: The reversible phosphorylation/inactivation of ICDH is catalyzed by a bifunctional enzyme (ICDH-kinase/phosphatase), which in turn is regulated by a number of effectors including OAA. OAA inhibits the ICDH-kinase and stimulates phosphatase. Thus, the decrease in OAA concentration due to the knockout of the ppc or pckA gene causes the phosphorylation of ICDH and consequent inactivation of ICDH, which causes in turn an increase in the concentration of the isocitrate. The flux through Icl significantly increases as a result. Moreover, either ppc or pckA gene knockout causes the accumulation of PEP, which in turn inhibits Pfk activity, thus reducing the rate of glucose consumption. These mutants produce less acetate and CO2, resulting in a higher cell yield with a lower growth rate than the wild-type E. coli. It has been shown that in vivo regulation of the Pck flux occurs mainly by modulation of enzyme activity and by the changes in PEP and OAA concentrations, rather than by the ATP/ADP ratio. This indicates that the reaction catalyzed by Pck can respond very flexibly to the availability of PEP and OAA, and may form the metabolic cycle at low glucose concentrations. It is known that PEP is a very important intermediate in E. coli metabolism, since it alone directly regulates the phosphotransferase system (PTS). It also affects the activity of Pfk and Pyk. Since PEP can be formed gluconeogenetically through Pck from the TCA cycle, Pck maintains the relative balance between the OAA and PEP pools, draining off excess carbon from the TCA cycle and supplying PEP for cellular requirements. As another example, knockout of the pgi gene (which codes for the first enzyme of the EMP pathway after the branch point at G6P) resulted in glucose catabolism proceeding exclusively by the oxidative pentose phosphate (PP) pathway. NADPH, which is overproduced as a result, inhibits the activity of G6PDH, thereby reducing the glucose consumption rate and resulting in a low growth rate. The glyoxylate pathway was activated, and so too may be the Entner–Doudoroff (ED) pathway in this mutant. The activation of the ED pathway may be due to the reduced production of NADPH as compared to the situation where the 6PGDH pathway is used. The activation of the glyoxylate pathway is due to the feedback regulation that compensates for the lowered OAA concentration that is caused by the lowered flux of the EMP pathway and the lowered anaplerotic flux through Ppc to supply OAA. The overproduced NADPH can be converted to NADH by transhydrogenase Udh, or this NADPH can be utilized for NADPH consumption for heterologous PHB production, and so on. Thus, the cell growth can recover to some extent. Although the knockout of the zwf gene showed little influence on central metabolism under glucose-limited continuous culture by flux rerouting via the nonoxidative PP pathway, this mutant showed significant overflow metabolism and extremely low TCA cycle fluxes under conditions where ammonia is limited. The effect of gnd gene knockout on metabolism is a little different from the knockout of the zwf gene (Figure 4). The gnd gene knockout activates the ED pathway and decreases the flux through G6PDH, which reduces the

Metabolic Regulation Analysis and Metabolic Engineering

Figure 4 mutant.

547

Metabolic flux changes in zwf and gnd mutant E. coli. The values are from top for wild type, middle for gnd mutant, and bottom for zwf

production of NADPH through the oxidative PP pathway (Figure 4). This decrease in the production of NADPH is compensated by the activation of Mez using MAL and the transhydrogenase Pnt. The reduced levels of OAA resulting from the utilization of MAL for Mez causes Ppc to be upregulated and Pck to be downregulated. The shortage of R5P due to the decreased flux through the oxidative PP pathway is compensated by the reversal of the nonoxidative PP pathway as compared to the wild type. This backup system has little effect on the phenotype (e.g., the cell growth rate) while the glucose consumption rate is increased. It has been shown using different levels of information that knockout of the pykF gene causes the accumulation of PEP, activating Ppc, MDH, and Mez,

548

Metabolic Regulation Analysis and Metabolic Engineering

resulting in the compensation of lowered PYR concentrations that are caused by the blocking of the Pyk pathway. Moreover, the accumulation of PEP inhibits Pfk allosterically, causing the accumulation of G6P, thereby lowering the glucose uptake rate, with consequent activation of the oxidative PP pathway flux. Under anaerobic or microaerobic conditions, the regeneration of NADH and the formation of ATP control the metabolic fluxes at branch points such as PEP, PYR, and AcCoA. As a result, knockout of the pflA,B gene causes the overproduction of lactate, and knockout of the ldhA gene causes increased fluxes toward the production of acetate, formate, and ethanol. Under anaerobic or microaerobic conditions, the glycolytic flux (or glucose consumption rate) is controlled by the ATP demand and increases as the production of ATP is reduced (Figure 5).

2.38.4

Global Regulators in Relation to the Cultural Environment

In addition to the effect of genetic mutation on the metabolism, the effect of the cultural environment is also of practical importance. The central metabolic pathways of E. coli are controlled by a number of global regulators, depending on the carbon sources available and the growth environment12 (Figure 6). The catabolite repressor/activator protein (Cra, which was initially characterized as the fructose repressor, FruR) plays an important role in the control of carbon flow in E. coli. Central metabolic pathway genes such as ptsHI, pfkA, pykF, zwf, and edd-eda are reported to be repressed, while ppsA, fbp, pckA, icd, aceA, and aceB are activated by Cra protein (Appendix A). It is known that a mutant defective in the cra gene is unable to grow on gluconeogenic substrates such as pyruvate, acetate, and lactate. This appears to be due to deficiency in the gluconeogenic enzymes such as PEP synthase, PEP carboxykinase, some TCA cycle enzymes, the two glyoxylate shunt enzymes, and certain electron transport carriers. The gluconeogenic pathway is deactivated by knockout of the cra gene, and the carbon flow toward catabolism and the glucose consumption rate are expected to increase since glycolysis pathway genes such as ptsHI, pfkA, and pykF are activated by the cra gene knockout. Global regulators such as the Fnr protein and the ArcAB system are mainly responsible for the regulation of the expression of numerous proteins that respond to the availability of oxygen and other electron acceptors in the cultural environment. Fnr regulates the genes in anaerobic metabolism, while the ArcAB system regulates the expression of numerous genes under microaerobic conditions. The ArcAB system is a two-component regulatory system, in which ArcB is a membrane-bound sensor kinase and ArcA is the cognate response regulator. ArcB autophosphorylates and transphosphorylates ArcA when oxygen is limited. It was demonstrated that ArcA, when phosphorylated, represses the expression of genes encoding enzymes that are involved in the TCA cycle and the Glucose PEP PYR G6P

FBP

GA NAD+ NADH

NADH CO2 PEP

FADH FAD+

ADP NADH

ATP

NAD+ Succinate

Lactate

PY NADH

NAD+

Formate CO2 + H2 Acetate

Figure 5

Acetyl Ethanol CoA 2NADH 2NAD+ ADP ATP

Metabolic regulation mechanism of lactate production in E. coli pfl
mutant.

549

Metabolic Regulation Analysis and Metabolic Engineering

Environmental stimuli

Global regulators

Metabolic pathway regulation

Glc

Catabolite repression

PEP (pts I, H, crr, ptsG) Hxk Pyr Pgm

cAMP-Crp

ATP Glk (glk)

NADP+ NADPH

ADP

G6P

G1P

Carbon limitation (gluconeogenic substrates)

Cra (FadR/[c]R)

O2 Low

ArcA/B,Fnr

O2 High

Pgl (pgl)

6PGL

G6PDH (zwf)

Pgi (pgi)

NADP+ NADPH Rpi (rpiA, B)

F6P ATP (fbp) Fbp ADP

ATP Pfk (pfkA, B) ADP

Tal (talA,B)

E4P

Ru5P

Tkt (tktB)

Tkt (tktA)

Fba (fba) Tpi (Tpi) GAPDH

Edd (edd)

X5P

S7P

GAP

SoxR/S

6PGDH (gnd)

Rpe (rpe)

F1,6BP DHAP

6PG

R5P

GAP

2K3DPG NAD+ Eda (eda)

(gapA, C)

NADH

1,3BPG ADP (pgk) Pgk

Nutrient stress condition

RpoS

Nitrogen limitation

Ntr regulon

Phosphate limitation

Pho regulon (PhoR,PhB)

ATP

3PG (pgm) Pgm

2PG (eno) Eno

PEP Ppc (ppc)

ADP (ppsA) Pps ATP

PYR ADP ATP

Temperature (heat shock)

NAD+ PDK (aceE, F, lpdA) NADH Pta (pta) AcCoA

Signal transduction (σ-factors)

CS (gltA)

OAA

ADP

ATP

Acetate

NADH AdhE (adhE) NAD+

AcAld

CIT

NAD+

Acn (acnA, B)

Ethanol

MDH (mdh)

ICIT

MAL

FUM FADH2

SDH (sdhABCD) FAD

MS (aceB)

GOX

NAD(P)+

NAD(P)H

KG

Frd (frdABCD) FAD

SUC

ICDH (icdA)

IC1 (aceA)

FADH2 (susAB, lpdA) α KGDH

NAD+ (susCD) SUC

GTP

Figure 6

Formate

Ack (ackA)

NADPH

Fum

σ 38 (stress) σ 54 (N-limit) σ 32 (heat shock) σ 70 (house keeping)

Lactate

AcP

NADP+ Mez (mez)

NADH

pH

NAD+

NADH ADP Pyk (pykF, A) Ldh (ldhA) ATP Pfl (pfLABCD)

SucCoA

NADH

GDP

Metabolic regulation in relation to cultural environmental stimuli.

glyoxylate shunt, including gltA, acnAB, icdA, sucABCD, sdhCDAB, fumA, mdh, and aceB (Appendix A). The genes that encode the primary dehydrogenases such as glpD, lctPRD, and lpdA are also repressed by ArcA. Escherichia coli possesses two terminal quinol oxidases in the respiratory chain. The cyoABCDE genes that encode cytochrome o oxidase (which has a low oxygen affinity and functions mainly under aerobic conditions in E. coli) are repressed by ArcA under microaerobic condition. The cydAB genes (encoding cytochrome d oxidase that has a high oxygen affinity and that functions mainly under microaerobic conditions) and the foc-pfl genes (which encode pyruvate-formate-lyase) are activated by ArcA under microaerobic conditions. It was shown that the ArcAB system exerts more significant regulation on cell catabolism under microaerobic conditions than under aerobic or anaerobic conditions. As expected from the aforementioned regulation, the TCA cycle is activated when the arcA/B gene is knocked out, causing high NADH/ NAD ratios, which in turn represses the activity of the TCA cycle. It may be considered to express the heterologous nox gene to oxidize NADH, thereby activating the TCA cycle, and it may be also considered to use nicotinic acid and sodium nitrate for activating the TCA cycle. Escherichia coli responds to oxidative stress by modifying the expression of many genes. Early studies using two-dimensional gel electrophoresis to analyze variations in the expression of proteins have shown that the synthesis of more than 80 proteins is activated in response to oxidative stress. Some of these induced proteins possess fundamental antioxidant functions, for example, superoxide dismutase and catalase. The search for mutants with altered antioxidant defenses has led to the isolation

550

Metabolic Regulation Analysis and Metabolic Engineering

and characterization of pleiotropic regulators that operate as redox-regulated genetic switches. The best-characterized pleiotropic regulators of the antioxidant response are the OxyR and SoxR proteins. Both of these proteins have the remarkable ability to directly transduce oxidative signals to genetic regulation. Both of these proteins are expressed constitutively in an inactive state and are transiently activated in cells under specific types of oxidative stress. The activation of the OxyR and SoxR proteins results in the transcriptional enhancement of sets of genes (regulons) whose products relieve the stress by eliminating oxidants and by preventing or repairing oxidative damage. SoxR is a member of the MerR family of metal-binding transcriptional factors and exists in solution as a homodimer, with each subunit containing a [2Fe-2S] cluster. These clusters are in the reduced state in inactivated SoxR and their oxidation activates SoxR as a powerful transcriptional factor. The active form of SoxR activates transcription of the soxS gene. The soxS gene product, the SoxS protein, belongs to the AraC/XylS family of DNA-binding transcriptional factors. Analyses using proteomics and genetic approaches have shown that SoxS activates the expression of 17 genes or operons. Although many industrial fermentations are conducted in the batch mode, most of the studies to date have focused only on the cell growth phase, and very little attention has been paid to the late growth and stationary phases. Since the important metabolites are produced at the early stationary or stationary phases, it is quite important to clarify the metabolic regulation mechanisms that occur during these phases. During batch fermentation, the cultural conditions change from being glucose rich to being acetate rich, and change further to conditions of carbon starvation. The presence of several global regulatory proteins including RpoD, SoxRS, Cra, FadR, and IclR has been reported to help E. coli to cope with different kinds of metabolic stresses. Apart from these regulatory proteins, RpoS, the master regulator of the stationary phase or stress-induced genes in E. coli, has recently been reported to regulate the expression of several metabolic pathway genes. These genes are those for the carbohydrate PTS, crr, glycolytic pathway genes such as fbaB and pfkB, the acetate-forming gene poxB, the nonoxidative PP pathway genes such as talA and tktB, and TCA cycle genes such as acnA and fumC. In addition, some of the genes relating to the amino acid and fatty acid metabolic pathways such as argH, aroM, and yhgY and energy metabolism genes such as narY, appB, and ldcC have also been identified as being regulated in an rpoSdependent manner. The complexity of the metabolic system is exemplified by the fact that many metabolic pathway genes are regulated by more than one global regulator. For example, icd of the TCA cycle is regulated by RpoD, ArcA, and Cra; acnA and fumC of the TCA cycle are regulated by SoxRS, ArcA, and RpoS; and aceA and aceB of the glyoxylate pathways are regulated by Cra, ArcA, and IclR in E. coli. Moreover, the metabolic pathway of E. coli consists of many genes that possess iso-genes. These iso-genes are known to encode backup enzymes in response to certain environmental stimuli, and the expression of these enzymes is often regulated by one or more of the global regulatory proteins. Examples include the tktA and tktB genes as well as the talA and talB genes. Other examples include the fumA and fumC genes and the acnA and acnB genes of the TCA cycle. Note that the expressions of tktA, talB, acnB, and fumA are dominant in the cell growth phase, while tktB, talA, acnA, and fumC become dominant during the stationary phase. In this way, the total expression levels are kept nearly constant throughout the fermentation, where the latter genes are all under the control of RpoS (Figure 7). For glucose catabolism, several other global regulatory proteins such as Crp, Mlc, and Cya may also be important.

2.38.5

The Systems Biology Approach

When bacteria adapt to the cultural environment, the global regulators detect the change in cultural environment and regulate the metabolic pathway genes. The global regulators may be considered as the decision makers and the pathway genes may be

Total gene expressions

4

3 acnA

2

acnB

1

0 EP

ES

STA

EP

ES

STA

EP

ES

STA

Figure 7 Expressions of iso-genes during batch culture: left, wild type; middle, fumC mutant; right, acnA mutant. EP, exponential phase; ES, early stationary phase; STA, stationary phase.

Metabolic Regulation Analysis and Metabolic Engineering

551

thought of as the executors. It is quite important to understand the complex and highly interrelated cellular behavior quantitatively. This may be achieved with the help of informatics by integrating the different levels of the ever-increasing amount of data to gain a deeper insight into the available data using biological knowledge. The ultimate aim of systems biology is to develop in silico models of whole cells or cellular processes that can predict cellular phenotypes. FBA has successfully predicted the cell growth rate, substrate uptake rate, and byproduct secretion rate by maximizing the energy generation or cell growth rate.5 It is more challenging to predict lethality and the phenotypes of organisms after genetic perturbations using constraint-based models. Genome-scale models of metabolism may be used to plan gene deletion strategies for the overproduction of chemicals or biochemicals using the so-called OptKnock framework. This hybrid model, which incorporates the kinetic model into the stoichiometric model, has been proposed for the expression of the dynamics of specific metabolic regulation at some branch points. The critical step forward toward whole-cell modeling is to integrate the different levels of information. There are several approaches for the modeling of the joint processes of gene regulation and metabolic regulation. One approach is to model the relationship between metabolites and gene expression, where gene expression data are used to impose binary constraints, explaining mRNA and protein expression data using different carbon sources with protein interaction cascades, or incorporating links as functions in Boolean networks. The genome-scale methods were developed by incorporating transcriptional regulation using Boolean on/off rules. Another approach considered the different allocation of resources to optimize the cell growth, the combination of the flexibility of the flux modes and the efficiency of cell growth, and the co-regulation of the enzymes along the metabolic pathways. Although these approaches give insight into the design rules for the regulatory and metabolic systems, they lack information about the underlying mechanisms. Large-scale transcript data can be integrated, and potential regulation mechanisms may be identified. The genome-scale stoichiometric models give a consistent framework for metabolic regulation by testing model predictions using experimental flux data. Probabilistic graphical methods can be developed for connecting the metabolic reactions and transcription, where the regulatory and metabolic networks may be integrated by adding links specifying the feedback control from the substrates of metabolic reactions to the enzyme levels and to gene expression. It is important to identify the common evolution or design principles that underlie the structure, regulation, and operation of networks. The ‘bow-tie’ structure of metabolism describes the fact that many parallel sequential pathways for the degradation of nutrients merge into a set of reactions from which a large number of biosynthetic pathways fan out. Classical metabolic control analysis (MCA) may be applied to identify the limiting pathways. MCA-based regulation analysis is, however, limited to local fluxes and cannot be applied to the entire network. With the availability of complete gene-knockout mutant libraries for many model organisms, systematic metabolic regulation analysis is in progress13 and is necessary for the construction of the precise computer cell models that will be required for the development of cell factories. Recently, minimization of the genome has been attempted to construct a more useful E. coli cell by identifying the minimal gene set containing only those genes that are essential and sufficient to sustain a functioning cell.14 The eventual goal may be to improve the cellular growth rate and the production of specific metabolites. Most of the minimum genome design has been based on the genome engineering approach. Metabolic regulation analysis is critical and has to be used in evaluating the performance of such designed cells and the re-engineering of cell factories.

2.38.6

Conclusion

Metabolic processes are hierarchical from transcription, translation, and enzyme activities to metabolite or metabolic fluxes. The presence of feedback loops among these regulatory processes makes their organization and functioning very complicated. Consequently, accurately predicting the cellular response to genetic or environmental perturbations is a difficult procedure and should take into account as many regulatory constraints as possible. Global information from different stages of the metabolic hierarchy needs to be integrated using mathematical and statistical methods. As stated previously, we are still far from understanding the regulation mechanism of the whole cell. It should be noted that each level of information does not contain functional information. These different levels of information have to be integrated by comparing the different states caused by gene mutations and/or changes in the cultural environment with the aid of a systems biology approach. This issue is the most critical to an understanding of the overall cell physiology. Systematic analysis of cells using multiple levels of biological information may reveal how the cells respond to genetic and/or environmental perturbations. Gene and protein expression may be globally regulated by different mechanisms and uncharacterized metabolic pathways may be revealed by comparing the different levels of information. Therefore, multi-omics data of the mutants under the same growth conditions may yield important information, and it has been shown that cells seem to use complementary strategies to maintain a robust biomolecular state, depending on metabolic demands and stress.

552

Metabolic Regulation Analysis and Metabolic Engineering

Nomenclature

Central metabolic pathway Pathway ED pathway EMP pathway PP pathway TCA cycle

Formal name Entner–Doudoroff pathway Embden–Meyerhof–Parnas pathway Pentose phosphate pathway Tricarboxylic acid cycle Metabolite

Symbol EMP pathway Glc G6P F6P F1,6BP GAP DHAP 1,3BPG 3PG 2PG PEP PYR PP pathway 6PGL 6PG Ru5P R5P X5P S7P E4P ED pathway 2K3DPG TCA cycle AcCoA CIT ICIT aKG SucCoA SUC FUM MAL OAA Glyoxylate bypass GOX

Formal name glucose glucose-6-phosphate fructose-6-phosphate fructose-1,6-bisphosphate glyceraldehyde-3-phosphate dihydroxyacetone phosphate 1,3-bisphosphoglycerate 3-phosphoglycerate 2-phosphoglycerate phosphoenolpyruvate pyruvate 6-phosphogluconolactone 6-phosphogluconate ribulose-5-phosphate ribose-5-phosphate xylose-5-phosphate sedoheptulose-7-phosphate erythrose-4-phosphate 2-keto-3-deoxy-6-phosphogluconate acetyl coenzyme A citrate isocitrate a-ketoglutarate succinyl-CoA succinate fumarate malate oxaloacetate glyoxylate Gene and enzymes

Gene EMP pathway glk ptsⅠ pts H crr, ptsG pgi pfkA,B fbp fba tpi gap

Enzyme

Full name

Glk EI HPr EⅡglc Pgi Pfk Fbp Fba Tpi GAPDH

pgk

Pgk

glucokinase PTS enzymeⅠ histidine phosphorylatable protein PTS enzymeⅡ phosphoglucose isomerase phosphofructokinase fructose-1,6-bisphosphatase fructose-1,6-bisphosphate aldolase triose phosphate isomerase glyceraldehyde-3-phosphate dehydrogenase phosphoglucokinase (Continued)

Metabolic Regulation Analysis and Metabolic Engineering

Gene and enzymes pgm eno pykF,A pps PP pathway zwf pgl gnd rpiA,B rpe tktA,B talA,B ED pathway edd eda TCA cycle aceE, F, lpdA gltA acn icdA sucA, B, lpdA sucC, D sdhC, D, A, B frdA, B, C, D fumA, B, C mdh Fermentation pathway ldhA pflABficCD pta ackA acs poxB adhE Anaplerotic pathway aceA aceB maeA(sfcA) maeB ppc pckA Respiratory chain cydAB cyoABCDE ndh nuo atpA Global regulators Cra (fruR) arcA fnr crp mlc rpoS soxR/S Others udhA pntA,B

Pgm Eno Pyk Pps

phosphoglucomutase enolase pyruvate kinase phosphoenolpyruvate synthase

G6PDH Pgl 6PGDH Rpi Rpe Tkt Tal

glucose-6-phosphate dehydrogenase 6-phosphogluconolactonase 6-phosphogluconate dehydrogenase ribulose-5-phosphate isomerase ribulose-5-phosphate epimerase transketorase transaldrase

Edd Eda

6-phosphogluconate dehydratase 2-keto-3-deoxy-6-phosphogluconate aldolase

PDH CS Acn ICDH aKGDH SCS SDH Frd Fum MDH

pyruvate dehydrogenase citrate synthase aconitase isocitrate dehydrogenase a-ketoglutarate dehydrogenase succinyl-CoA synthetase succinate dehydrogenase fumarate reductase fumarase malate dehydrogenase

LDH Pfl Pta Ack Acs Pox AdhE

lactate dehydrogenase pyruvate formate lyase phospho tansacetylase acetate kinase acetyl-CoA synthetase pyruvate oxydase alcohol/acetaldehyde dehydrogenase

Icl MS Sfc Mez Ppc Pck

isocitrate lyase malate synthase malic enzyme (NAD-dependent) malic enzyme (NADP-dependent) phosphoenolpyruvate carboxylase phosphoenolpyruvate carboxykinase

CydAB CyoABCDE NDH II NDH I ATPase

cytochrome bd cytochrome bo NADH dehydrogenase II NADH dehydrogenase I ATP sysnthase (a-subunit)

Cra ArcA Fnr Crp Mlc RpoS SoxR/S

catabolite repressor/activator anoxic redox control protein fumarate-nitrate respiration cAMP receptor protein making large colonies stress regulator superoxide stress regulon

UdhA

pyridine nucleotide transhydrogenase (soluble) pyridine nucleotide transhydrogenase (membrane-bound)

Pnt

553

554

Metabolic Regulation Analysis and Metabolic Engineering

Appendix A: Global Regulators and Their Regulated Genes: þ Activate,
Repress ArcA/B: ＋ pflB, focA, cydAB, pdhR,
aceBAK, lpdA, aceEF, gltA, acnAB, icdA, sucABCD, fumAC, sdhCDAB, mdh, ptsG. cyoABCDE, nuoABCDEFGHIJKLMN, fadAB Fnr: ＋ pfl, frdABCD, acs, aspA, focA, fumB
acnA, fumAC, icdA, lpdA, ptsG, sdhCDAB, talA Cra: ＋ aceA, acnA, fbp, icdA, pckA, ppsA, cyd, crr
ptsHI, pfkA, gapA, eno, pykF, acnB, eda, edd, zwf Mlc: ＋
ptsHI, crr, ptsG, manXYZ, malT Crp/Cya: ＋ aceAB, aceE, acnAB, crr, fumA, gltA, mdh, pckA, ptsG, ptsHI, sdhABCD, sucABCD, tpiA
pdA SoxR/S: ＋ zwf, acnA, funC, acnA, fumC, soda, acs, adhE
soxS RpoS: ＋ tktB, talA, acnA, fumC, acs, poxB, sucA, crr, fbaB, pfkB, sodC, argH, aroM, yhgY, narY, appB, ldcC


References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

Stephanopoulos, G. N.; Aristidou, A. A.; Nielsen, J. Metabolic Engineering, Academic Press: San Diego, 1998. Vemuri, G. N.; Aristidou, A. Metabolic Engineering in The’-omics Era: Elucidating and Modulating Regulatory Networks. Microbiol. Mol. Biol. Rev. 2005, 69, 197–216. Han, M. J.; Lee, S. Y. The Escherichia coli Proteome: Past, Present, and Future Prospects. Microbiol. Mol. Biol. Rev. 2006, 70 (2), 362–439. Soga, T.; et al. Quantitative Metabolic Analysis Using Capillary Electrophoresis Mass Spectrometry. J. Proteome Res. 2003, 2, 488–494. Palsson, B. O. Systems Biology, Cambridge University Press: New York, 2006. Shimizu, K. Toward Systematic Metabolic Engineering Based on the Analysis of Metabolic Regulation by the Integration of Different Levels of Information. Biochem. Eng. J. 2009, 46, 235–251. Wolfe, A. J. The Acetate Switch. Microbiol. Mol. Biol. Rev. 2005, 69, 12–50. Lin, Y.; Tanaka, S. Ethanol Fermentation from Biomass Resources: Current State and Products. Appl. Microbiol. Biotechnol. 2006, 69, 627–642. Atsumi, S.; Cann, A. F.; Connor, M. R.; et al. Metabolic Engineering of Escherichia coli for 1-butanol Production. Metab. Eng. 2008, 10 (6), 312–320. https://doi.org/10.1016/ j.ymben. Sanchez, S.; Demain, A. L. Metabolic Regulation and Overproduction of Primary Metabolites. Microbial. Biotechnol. 2008, 1, 283–319. Shimizu, K. Metabolic Flux Analysis Based on 13C Labeling Experiments and Integration of the Information with Gene and Protein Expression Patterns. Adv. Biochem. Eng. Biotechnol. 2004, 91, 1–49. Escherichia coli and Salmonella. In Cellular and Molecular Biology; Neidhardt, F. C., Curtiss R. III, Ingram, J. L.; et al., Eds., 2nd ed.; ASM Press: Washington, DC, 1996. Ishii, N.; Nakahigashi, K.; Baba, T.; et al. Multiple High Throughput Analyses Monitor the Response of E. coli to Perturbations. Science 2007, 316 (5824), 593–597. in press. Mizoguchi, H.; Mori, H.; Fjio, T. Escherichia coli minimum Genome Factory. Biotechnol. Appl. Biochem. 2007, 46, 157–167.

2.39

Proteomics and Protein Engineering

AG Pereira-Medrano and PC Wright, The University of Sheffield, Sheffield, United Kingdom © 2011 Elsevier B.V. All rights reserved. This is a reprint of A.G. Pereira-Medrano, P.C. Wright, 2.31 - Proteomics, Protein Engineering, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 421-439.

2.39.1 2.39.1.1 2.39.1.2 2.39.2 2.39.2.1 2.39.2.2 2.39.2.2.1 2.39.2.2.2 2.39.2.3 2.39.2.3.1 2.39.2.3.2 2.39.2.3.3 2.39.2.4 2.39.2.4.1 2.39.2.4.2 2.39.2.4.3 2.39.2.5 2.39.3 2.39.3.1 2.39.3.2 2.39.3.2.1 2.39.3.2.2 2.39.3.2.3 2.39.3.2.4 2.39.4 References

Introduction What Is Proteomics? Why Proteomics? Applications and Benefits Mass Spectrometry-Based Proteome Profiling Techniques Sample Preparation Protein and Peptide Separation Techniques 1D SDS-PAGE and 2D-GE Gel-Free and Gel-Based Shotgun Methods Protein Digestion In-Gel and In-Solution Tryptic Digestion Toward Trypsin Digestion Optimization Immobilized Enzyme Reactors Using Trypsin Mass Spectrometric Protein Identification MS Sources: MALDI and ESI Acquisition of Mass Spectra: MS and Tandem MS Database Searching: PMF and Sequence Tag Searches Protein Quantitation Current Advances in Protein Identification: Online and Microfluidic Proteomic Systems Online Proteomic Systems Based on Capillary Columns Microfluidics in Proteomics: Chip Formats Protein and Peptide Fractionation on a Chip Immobilized Enzyme Microreactors Chip Integration With MS Toward a Total Proteomic Analytical System in a Chip Current Challenges in Proteomics

557 557 557 558 559 559 559 560 561 561 561 561 562 563 563 565 566 567 567 569 569 569 570 571 571 573

Glossary 2D-GE Two-dimensional gel electrophoresis is the separation of proteins using two orthogonal parameters, isoelectric point (charge) and relative molecular mass, which are both usually determined on the basis of protein mobility in a polyacrylamide gel matrix. Collision-induced dissociation (CID) A method of energetically activating ions to dissociate. Typically, a gas-phase collision cell that is filled with argon gas is used to subject ions to low energy collision (10–50 eV) to cause energetic excitation. As ions become energetically excited, covalent bonds dissociate to produce structurally informative fragment ions. Often the molecular structure of the ion can be postulated from the fragmentation pattern, or in the case of peptides, the amino acid sequence can be deduced. Dalton (Da) The unit of the mass scale, which is defined as one-twelfth of the mass of the monoisotopic form of carbon, 12C (1 Da ¼ 1.6605  1027 kg). Other commonly, but not necessarily correctly, used units of relevance to mass spectrometry are the amu (an atomic mass unit that is based on 16O), the Thomson (the proposed unit for the mass-to-charge (m/z) scale), and the u (‘unit’, which is the same as Da). De novo sequencing Deriving the amino acid sequence (primary structure) of a peptide solely from the mass spectrometry, peptide fragmentation data (i.e., without using databases). DNA microarrays A high-throughput differential screen on mRNA expression using complementary cDNA or oligonucleotide libraries that are printed in extremely high density on microchips; these microchips are probed with a mixture of fluorescently tagged cDNAs that are produced from two different cell populations and analyzed with a laser confocal scanner. ESI The electrospray ionization process is achieved by spraying a solution (such as the effluent of a high-performance liquid chromatography column (HPLC)) through a charged needle at atmospheric pressure toward the inlet of the mass spectrometer; the voltage applied to the needle tip in a pressure differential results in the formation of ions for mass analysis and their transfer into the mass spectrometer.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00089-6

555

556

Proteomics and Protein Engineering

ICAT Isotope-coded affinity tag reagent comprising a chemical modifying group linked to an affinity group through a massencoded linker. Ionization Process of adding charge onto an uncharged (neutral) analyte, in other words, the formation of an ion; either ionization is conducted in a vacuum or ions formed at atmospheric pressure are transferred into the vacuum system of the mass spectrometer. Ion source Mass spectrometer component designed to use the principles of an ionization method for generating ions (charged analytes) for mass analysis. Matrix-assisted laser desorption ionization (MALDI) A process by which ion formation is promoted by short laser pulses; the sample is deposited on a sample plate into the source (which is held under vacuum) and then embedded in a matrix that promotes lionization; a laser fired at the sample that is co-crystallized with the matrix results in the desorption of the analyte from the sample plate and its ionization. Mass analyzer Mass spectrometer component that can measure the mass-to-charge ratio of charged molecules (ions); ion-trap, quadrupole and time-of-flight (TOF) analyzer are used most often. Mass spectrometry Accurate mass measurement of charged analytes (ions); in the context of proteomics, analytes are usually peptides or less frequently protein ions; a mass spectrometer measures the mass-to-charge ratio of charged species under vacuum and comprises, broadly, an ionization source and mass analyzer. Mass-to-charge ratio (m/z) Mass spectrometers measure the mass-to-charge ratios of ions. In matrix-assisted laser desorption ionization (MALDI) and electrospray ionization (ESI), peptides are typically ionized by the addition of one or more protons. Thus, a peptide of molecular weight 1000 Da will have an m/z value of 1001 after ionization by the addition of one proton and 501 with the addition of two (M þ 2H)þ2. Microfluidics Transporting and manipulating microliter amounts of fluid through a microchannel. Microscale capillary HPLC column High-performance liquid chromatography (HPLC) columns have inner diameters of 50– 150 mm and a reversed-phase stationary phase. Reversed phase means that the surface is made using long hydrophobic alkyl chains, so they retain hydrophobic compounds better than hydrophilic ones. MS spectra Single-stage mass spectrometry spectra provide mass information on all ionizable components in a sample; these data are used, for example, for peptide mass fingerprinting. MS/MS spectra MS/MS spectra are generated in instruments equipped with a mass filter that can select a peptide ion from a mixture of peptide ions, a collision cell in which peptide ions are fragmented into a series of product ions (through collision of the selected precursor ion with a noble gas in a process referred to as collision-induced dissociation (CID)), and a second mass analyzer that records the fragment ion mass spectrum; the fragment ion spectra are referred to as either MS/MS or CID spectra. Protein identification Method to determine the sequence identity of a protein; two common mass spectrometry-based approaches used are peptide mass mapping and searching uninterpreted MS/MS spectra; in both methods, observed data are matched to theoretically derived peptide and/or fragment ion masses calculated from sequence databases. Quadrupole ‘ion traps’ In ion traps, the ions are first caught (trapped) in a dynamic electric field and are then sequentially – according to their mass-to-charge (m/z) value – ejected onto the detector with the help of another electric field. Trapped ions can also be isolated and fragmented within the trap. Quadrupole mass spectrometer A mass-selective ‘quadrupole section’ only allows the passage of ions that have a specific mass-to-charge (m/z) value by applying a particular sinusoidal potential. Stepping through the m/z range by applying different potentials and detecting the ions that pass through at each m/z value generates the mass spectrum. Shotgun proteomics A gel-free approach based on multidimensional liquid chromatography separation of complex peptide mixtures coupled to mass spectrometry (MS). Systems biology Study of biological system by the systematic and quantitative analysis of all components that constitute the system. Tandem mass spectrometer (MS/MS) A tandem mass spectrometer combines two mass analyzers with a device (e.g., gas-phase collision cell) or method to energetically activate ions. In this approach, a particular mass-to-charge (m/z) value can be isolated from all other ions that enter the mass analyzer at the same time, dissociated, and the m/z values of the dissociation products can be determined in the second mass analyzer. The dissociation process causes covalent bonds to fragment, leading to a collection of ions that are diagnostic of the molecular structure of the ion. In the case of peptide ions, fragmentation processes predominate at or around the amide bond, creating a ladder of ions that is indicative of an amino acid sequence (after careful deliberation). Time-of-flight (TOF) mass spectrometer A mass analyzer that measures mass-to-charge (m/z) values by pulsing ions from the ion source into a flight tube. The time required for ions to travel a set distance and strike a detector is determined and m/z values are calculated from the TOF measurements. TOF mass spectrometers can be used with matrix-assisted laser desorption ionization (MALDI) or electrospray ionization (ESI) sources. Total ion current (TIC) The sum of all the ion signals in a mass spectrum as a function of elution time.

Proteomics and Protein Engineering

2.39.1

557

Introduction

Proteomics research is increasingly important to complement well-established genomics research cemented by the completion of the Human Genome Project.1 In the post-genomic era, predictions that prokaryotes and eukaryotes produce numerous protein profiles with uncharacterized functions highlight the major challenge of identifying and annotating this vast number of proteins.2 This prompted the development of large-scale approaches to protein studies, and the development of new methods and technologies for rapid and congruent analysis, all collectively conforming the proteomics field.3 In the past decade, among the advances in proteomic technologies, mass spectrometry (MS) has been consolidated as the preferred analytical technique for analyzing the production and function of proteins.1,2 The main reason is the incomparable capability of MS to acquire high-content quantitative information of highly complex biological samples, allowing for significant advances to be made in understanding cellular and physiological processes.2 These technical advances have produced a wide range of new MS-based proteomic procedures and experimental strategies that are now in routine use.1,4 In this article, an overview of MS-based proteomics is presented, indicating the main applications of proteomics research, and introducing well-established tools and platforms, as well as microfluidic-based emerging tools.

2.39.1.1

What Is Proteomics?

The term proteome was first proposed by Wilkins in 1995, referring to the total set of proteins encoded by the genome. Proteomics is then defined by the characterization of the proteome and has evolved to a systematic identification/characterization and quantitation of proteins expressed in cells, at a specific time and conditions.5 Proteomics offer a new platform for studies of complex biological functions, allowing the study of these gene products and their network interactions. The significant number of expressed proteins found in an organism due to alternative splicing mRNA and posttranslational modifications (PTMs) results in enormous diversity, complexity, and heterogeneity of gene products, making their determination a significant challenge.1 Since the gene sequence and pattern of gene activity inside the cell provide an incomplete picture, proteomics plays an important role and has developed widely5: 1. Separation. Separating protein mixtures to identify and characterize more proteins. In two-dimensional gel electrophoresis (2D-GE), proteins are resolved in two dimensions (isoelectric point and molecular weight – pI and MW). Shotgun proteomics based on high-performance liquid chromatography (HPLC) has been applied for multidimensional separation.6 2. Identification. Employing Edman degradation, with low-throughput sequencing. Higher-throughput proteomic techniques employ MS.4 3. Quantification. Gel-based methods include densitometric comparison of protein abundances in 2D-GE arrays. Proteins are detected with silver stain, Coomassie blue, or fluorescent tags/dyes. Gel-free methods include metabolic isotope incorporation (stable isotope labeling by amino acids in cell culture (SILAC), 15N, and 13C in vivo labeling), enzyme-catalyzed isotope incorporation (labeling with O-18), and isotope labeling by chemical derivatization (isotope-coded affinity tags (ICATs) and isobaric tag for relative and absolute quantitation (iTRAQ)).7,8 4. Sequence analysis. Bioinformatics searches MS or tandem mass spectrometry/spectrometer (MS/MS) spectra against protein sequence databases (e.g., NCBI or SwissProt) for possible protein/peptide matches, and predict function from sequence and evolutionary relationships.4 5. Structural proteomics. High-throughput determination of protein structures in three-dimensional (3D) space. X-ray crystallography and NMR spectroscopy are used.5 6. Interaction proteomics. Protein interactions on the atomic, molecular, and cellular levels.5 7. Modification. Specialized methods have been developed to study phosphorylation (phosphoproteomics) and glycosylation (glycoproteomics),5 for example.

2.39.1.2

Why Proteomics? Applications and Benefits

In the early 1990s, DNA-chip or microarray technology was developed as a powerful tool for large-scale characterization of the transcriptome. However, a difference in the transcription profile does not give complete information on cellular regulations and does not necessarily mirror events in the proteome. This is because gene expression is regulated posttranscriptionally, and because posttranslational events significantly increase the number of proteins beyond those directly predicted by the genome. An investigation of the proteome is needed to complement the information, which can allow the understanding of the regulation on the translational level, protein degradation, and the activation or deactivation of proteins by modifications, protein location, or translocation. A change in the proteome can be caused by endogenous factors, such as gene mutations and posttranslational processes, or exogenous factors, such as the effect of compounds (i.e., drugs), effects of environment, or changes in the cell or organism, such as stress or illness. Proteomics has facilitated the study of the effect of some external factors resulting from illness and drug administration, through the discovery of biomarkers of early diagnosis, and has proven to be useful in monitoring disease progression and drug response. Such understandings have been possible employing proteomics, along with biochemistry, genomics, and metabolomics, utilizing a systems biology philosophy. This philosophy now underpins modern approaches to drug development by the pharmaceutical industry.

558

Proteomics and Protein Engineering

Systems biology studies an entire biological system and the relationships among the elements in response to genetic or environmental perturbations, with the goal of understanding the system and its emergent properties.3 Systems biology employs a systematic and quantitative analysis of all the constituent components, via functional genomics, structural genomics, transcriptomics, pharmacogenomics, metabolomics, and proteomics. The integration of this multilevel information in a quantitative mathematical model framework is key. This relatively new discipline has not only being applied to healthcare, such as in drug development, in recombinant therapeutic protein production by mammalian host cells, and in prostate cancer studies, but also in the understanding of biological systems aiming for other applications. For example, (1) studies on cyanobacteria to identify their potential for future metabolic engineering and their use as ‘cell factories’ to manufacture natural products and biofuels; (2) studies on microbial communities to understand the physiology, ecology, and evolution; and (3) studies to understand the potential of microorganisms for biodegradation purposes. Proteomics has been shown to be a key element in systems biology studies.

2.39.2

Mass Spectrometry-Based Proteome Profiling Techniques

MS has emerged as the main method for analyzing the regulation and function of proteins in biological systems, due to its performance and versatility for rapid and robust proteins identification and quantitation.1,2 MS-based protein profiling is mainly divided into two platforms: bottom-up and top-down. ‘Bottom-up’ proteomics seeks to convert a complex protein mixture into peptides, obtain sequences of these peptides, and then identify the corresponding proteins via database sequence matching5,9 (see Figure 1 for more detail). In contrast to ‘bottom-up’ proteomics that gives protein sequence coverage of 5–70%, ‘top-down’ proteomics may achieve higher protein sequence coverage for small proteins or large peptides (5 kDa) to identify large proteins or specific domains of interest.10 This article focuses on the ‘bottom-up’ proteomics, as it has shown greater potential in increasing analysis throughput and coverage. A

Real sample cells

Protein mixture

Protein separation

Proteins

Protein digestion

Protein digestion

MS/MS analysis Peptide identification

Peptide separation

Peptide mixture

Peptides

Protein identification

Database search algorithms

MS data

Protein sequence analysis

B

Sample preparation/ fractionation

Cell culture

SDS-PAGE 2D-GE HPLC

Protein digestion

Peptides

Trypsin Chymotrypsin Glu-C

Peptide separation RP-HPLC Ion exchange

Sample Ionization Electrospray ionization MALDI

Peptide ions

Mass spectrometry

Data analysis MASCOT Sequest X!Tandem

Spray needle

m/z

Quadrupole Time of flight Quadrupole ion traps FTICR

Figure 1 (A) General workflow for protein identification in proteomics analysis, following a conventional ‘bottom-up’ approach. (B) A conventional ‘bottom-up’ proteomic experiment, indicating the most common techniques used in each step. HPLC, high-performance liquid chromatography; MALDI, matrix-assisted laser desorption/ionization; MS, mass spectrometry; MS/MS, tandem mass spectrometry/spectrometer; RP, reverse phase.

Proteomics and Protein Engineering 2.39.2.1

559

Sample Preparation

Most analytical proteomic research begins with a protein mixture containing intact proteins of varying MWs, modifications, concentrations, and solubilities. Usually the protein mixture needs to be extracted from biological samples. First, lysis of the cells or tissues is carried out by either pulverization, homogenization, sonication, or disruption to yield a soup that contains cells, subcellular components, and other biological debris.5 To extract the proteins, the following reagents and techniques are used:

• • • • • •

Surfactants solubilize membrane proteins and aid their separation from lipids, such as sodium dodecyl sulfate (SDS), CHAPS, and Triton X-100. Reductants reduce proteins disulfide bonds, such as dithiothreitol (DTT) and b-mercaptoethanol. Denaturing agents disrupt protein interactions, secondary and tertiary structures by altering solution ionic strength, and pH, such as urea and guanidine. Enzymes digest contaminating nucleic acids, carbohydrates, and lipids, such as lysozyme. Pulverization aided by a surfactant/reductant/enzyme/denaturing reagent-based buffer, flash freezing the cells in liquid nitrogen, and homogenizing in mortar with a pestle. Sonication aided by a surfactant/reductant/enzyme/denaturing reagent-based buffer, the cells are homogenized by ultrasonic disintegrators.

Second, it is necessary to remove interfering substances (phenolic compounds, nucleic acids, and interfering ions or agents with downstream techniques) by precipitation, and insoluble components by high-speed centrifugation.11 There is no general sample preparation method, and the method for each tissue or cell needs to be optimized to minimize proteolysis and modification of proteins. This is important, as initial steps of sample preparation can affect the protein separation and subsequent MS analysis.

2.39.2.2

Protein and Peptide Separation Techniques

The most popular approaches for protein separation before digestion are one-dimensional (1D) SDS polyacrylamide gel electrophoresis (SDS-PAGE) and 2D-GE. Other methods for protein/peptide separation include shotgun gel-free and gel-based techniques, including HPLC (1D or multidimensional separations), affinity capillary electrophoresis (CE), and isoelectric focusing/isoelectric focusing (IEF/IEF).6

2.39.2.2.1

1D SDS-PAGE and 2D-GE

Gel electrophoresis separates molecules based on physical characteristics such as size, shape, or pI within a gel matrix. During SDS-PAGE, proteins are separated according to their electrophoretic mobility, as a function of length of polypeptide chain or MW.5 The gel is made with different concentrations of acrylamide and a cross-linker, producing different sized mesh polyacrylamide networks. Proteins are denatured with SDS that coats the proteins with a negative charge in direct proportion to its mass, such that the mass-to-charge (m/z) ratio is constant. Denatured proteins become long rods instead of a complex tertiary shape; hence, the rate at which the SDS-coated proteins migrate in the gel is relative to its size and not its charge or shape.5 The separated proteins are then in-gel digested for peptide separation and HPLC MS/MS analysis. In 2D-GE, the two properties that are used to separate the proteins are pI (pH at which a molecule carries no net electrical charge using IEF) and MW (using 1D SDS-PAGE; see Figure 2). IEF separates proteins based on their relative content of acidic and basic residues. Proteins are introduced into a gel, which has an established pH gradient (immobilized pH gradient or IPG strips). IEF can resolve proteins that differ in pI values by as little as 0.01, with excellent reproducibility and high protein load capacity, especially with IPG strips. IEF has been widely employed not only in the first dimension of separation in 2D-GE but also for preparative fractionation of proteins in liquid and gel formats.

3.0

Isoelectric point (pI) 6.5

10.0

Molecular mass (kDa)

150 75 50 25 15 10 Figure 2 Two-dimensional gel electrophoresis (2D-GE) of a 3.0–10.0 pH range isoelectric focusing (IEF). Proteins of the psychrophile Pedobacter cryoconitis resolved in a 12.5% acrylamide gel.

560

Proteomics and Protein Engineering

2D-GE is still considered the workhorse technique for proteomics, as it is well established and robust, and has the advantage that it enables the simultaneous visualization of thousands of protein spots, the quantification of their levels, and the detection of PTMs. However, it has well-characterized limitations, such as considerable workload, low resolution for highly hydrophobic proteins, and extremely acidic or basic, large or small proteins. Use of automated spot excision systems and in-gel digestion machinery, along with image analysis software, have enhanced its capability and throughput. An alternative to 2D-GE protocol for the separation and detection of the most hydrophobic thylakoid membrane proteins has been developed, employing 2D blue native (BN) gel electrophoresis, where the solubilized thylakoid membranes are loaded with a dye to charge the complexes for separation, followed by SDS-PAGE in the second dimension (2D: BN/SDS-PAGE). This protocol minimizes loss due to its protein transfer efficiency. Another method, the differential in-gel electrophoresis (DIGE) technique, allows for multiple sample analysis in one gel. This is by using multiple fluorescent dyes to label protein samples prior to 2D-GE, with the samples being simultaneously separated and visualized in the same gel, and reducing gel-to-gel technical variation.

2.39.2.2.2

Gel-Free and Gel-Based Shotgun Methods

With the aim of improving detection of the more acidic, basic, and hydrophobic proteins, interest in gel-free fractionation methods has increased.6 These so-called shotgun methods are based on the development of protein/peptide separation techniques from a mixture prior to MS analysis, based on a particular physical or chemical property. These advances have combined gel and non-gel-based protein and peptide separations, protein property, and enrichment methods, providing new tools for proteome profiling with increased sensitivity and throughput. Groups of proteins/peptides are separated according to their size, hydrophobicity, charge, isoelectric point, or affinity. A separation according to hydrophobicity with C18 reverse-phase liquid chromatography (RP-HPLC) is commonly employed. However, other shotgun multidimensional separation methods have been developed, including CE and IEF. HPLC separates analytes depending on their specific chemical or physical interactions with the stationary phase. The retention time depends on the nature of the analyte, stationary phase, and mobile phase composition. Different types of liquid chromatography include6 RP, where proteins/peptides are differentially eluted by changing the organic modifier with time, depending on hydrophobicity; size exclusion chromatography (SEC), where proteins/peptides are separated according to their size; strong cation exchange and strong/weak anion exchange (SCX and SAX/WAX, respectively), where proteins/peptides are separated based on peptide charge state; hydrophobic interaction liquid chromatography (HILIC), which separates proteins and peptides based on hydrophilicity; and immobilized metal affinity chromatography (IMAC), where proteins with affinity to metal ions can be bound to the column and subsequently eluted by a change of conditions. A commonly used approach for protein separation after digestion is microcapillary HPLC. CE is also based on the molecular ability of different compounds to migrate according to their charge and frictional forces. It is performed in narrow tubes and achieves separation of hundreds of different compounds by application of an electric field, based on molecular charge and frictional force. Capillary zone electrophoresis (CZE) is the simplest form of CE where each peptide is separated according to its apparent mobility or m/z ratio.6 Even though HPLC and CE have increased the proteome coverage and throughput, 2D-GE still presents some advantages over these shotgun techniques, such as lower costs (HPLC and CE require special instrumentation) and the ability to resolve thousands of proteins, and provide a rapid comparison of differences in protein expression between distinct samples. IEF is also considered a shotgun technique when SDS-PAGE is not used in series. Key advantages of employing IPG strips for IEF include simplicity, high resolution, and high sensitivity. Another form of IEF is the solution-based approach, also known as free flow electrophoresis (FFE). FFE focuses the proteins by mixing the sample with the desired pH range carrier ampholyte mixture, or other carrier buffer, and applying an electric potential to the focusing cell.6 An advantage of liquid-phase IEF is the ability to fractionate a complex mixture of proteins according to their pI in a non-gel medium. The fractions can be collected and further analyzed, if needed, by electrophoresis or chromatography. The disadvantages are that high concentrations of ‘neutral’ proteins (e.g., when focused at their pI) often precipitate, causing overlaps between fractions. Additionally, the ampholytes used to establish the pH gradient may interfere with subsequent electrospray ionization (ESI)-MS analysis. A variety of multidimensional separation approaches have successfully been employed for protein and peptide shotgun separations, mostly consisting of 2D separations.12 Multidimensional separations maximize the number of proteins/peptides for analysis by MS, as separation enhances the resolution of molecules and the detection of low-abundance peptides in the mixture.12 Some of the multidimensional shotgun separation approaches include multidimensional protein identification technology (MudPIT), combining SCX and RP columns sequentially packed in an off-line or online setup; SEC/RP-HPLC; CIEF/CRP-HPLC; RP-HPLC-CZE; HILIC-RP-HPLC; and IEF/IEF. MudPIT has proven to be a powerful high-throughput tool to shotgun entire small proteomes, giving a great amount of protein identification information from a limited amount of sample. Indeed, Washburn et al. (described in Ref. 12) reported the identification of 1484 proteins from Saccharomyces cerevisiae. Alternatively, Chen et al. (described in Ref. 6) reported the use of CIEF-CRPLC, from which the first dimension separated proteins/peptides on the basis of their differences in pI with a greater resolving power than that achieved in SCX. Later, an IEF/IEF approach gave the best proteome coverage (20%) when compared with five alternative shotgun approaches (IEF, 1D SDS-PAGE, and WAX to separate proteins, IEF and SCX for peptide separation) all prior to RP-HPLC-MS/MS. 3D peptide separation approaches have also been developed and allow the resolution of a larger number of peptides. Some of these developed techniques include IEF/SCX/RP; RP/SCX/RP; SEC/ RP/CZE; SEC/SCX/RP; and Affinity/SCX/RP, using IMAC for phosphopeptides.6 These 3D approaches resulted in the identification of more proteins than via 2D HPLC procedures; however, the analysis throughput decreased.

Proteomics and Protein Engineering

561

No protein/peptide pre-fractionation method (gel-based or gel-free) is capable of resolving the complex mixture of peptides that results from a total digest of a proteome, and these methods are complementary in providing increased proteome coverage. Above all, it is clear that a protein/peptide pre-separation is vital for adequate protein identification. Factors such as the complexity of the protein/peptide mixture, sample type and size, available time, MS mode, and the purpose of the proteome analysis must be weighed when deciding which shotgun approach to follow. In addition, an important sample preparation technique to consider is the reduction of concentration ratio of the highly abundant proteins that interfere with the identification of less abundant and usually potentially more interesting. Such is the case of the human plasma proteome, where classical ‘plasma proteins’, such as albumin, have an extraordinary dynamic range of >10 orders of magnitude in concentration from the rarest proteins, some of which are measured clinically as biomarkers.6 Methods for the removal of highly abundant proteins derive from classical approaches of affinity chromatography. Based on this simple mechanism, several selective beads for given proteins can be manufactured.

2.39.2.3

Protein Digestion

Protein digestion is used to yield fragments that are most compatible with MS analysis, specifically fragments of 6–20 amino acids in length. Proteases carry out the digestion. Proteases that are stable, well characterized, and with well-defined specificities are needed, and include11 trypsin, chymotrypsin, Glu-C, Lys-C, Lys-N, and Asp-N. Proteins can also be cleaved with chemicals such as cyanogen bromide. Trypsin is the most commonly used enzyme for proteome profiling. Trypsin cleaves proteins on the C-terminal side of the basic amino acid residues lysine and arginine, except when they are immediately followed by proline.11 Trypsin is a serine protease whose active site consists of three amino acids: His 57, Ser 195, and Asp 102. It is preferred for proteomics because it has a well-defined cleavage specificity, is inexpensive, and yields peptides with an average length of about 10 amino acid residues.11 Usually modified trypsin is used because it has a reduced autolysis and inhibits the chymotryptic activity of any trypsin autolysis peptides.

2.39.2.3.1

In-Gel and In-Solution Tryptic Digestion

The tryptic in-gel digestion method is the most commonly used approach. After a 2D-GE, protein digestion is carried out in-gel. This method is lengthy (9–27 h), and the digestion time (4–20 h) is the bottleneck. A typical protocol includes several sequential reactions, including11 reduction of disulfides, alkylation of thiols, enzymatic digestion, and peptide extraction. These steps must be repeated potentially hundreds of times to analyze each of the relevant excised 2D-GE spots or 1D SDS-PAGE bands. A variation of the standard protocol has been reported, which balances the digestion time (and hence protein identification throughput) with the peptide yield. The protocol can be carried out in 4–20 h, including a digestion time of 30 min–20 h. The reduction in time is an optimization of the timings of reduction, alkylation, and peptide extraction. The ‘on-strip’ digestion also takes place after IEF protein fractionation. To avoid peptide diffusion along and outside of the IPG gel matrix, a spray-based procedure may be used. Tryptic in-solution digestion is performed on proteins that are not in a gel matrix prior to their proteolysis, and is a standard procedure for gel-free shotgun proteomics workflows. Proteins are digested by exposure to proteolytic enzymes in-solution as part of a multistep process that lasts between 7 and 24 h. This process includes11 (1) mixing the sample with buffer and denaturant, (2) mixing and incubating the sample with the reducing agent to reduce disulfide bonds, (3) mixing and incubating the sample with the alkylating agent to prevent disulfide bonds from reforming, and (4) mixing and incubating with the proteolytic enzyme. Conventionally, proteins are digested by trypsin in a pH 8.0–8.5 buffer at 37  C for up to 20 h.11 To avoid autolytic fragments of trypsin that would complicate MS peptide identification due to strong background from these ions, the enzyme-to-substrate ratio is usually 1:20 or 1:50.11 Similar to in-gel digestion, it has extended digestion times (4–20 h).

2.39.2.3.2

Toward Trypsin Digestion Optimization

Protein digestion optimization is of interest to improve peptide yield, as more detectable peptides improve their detection by MS. The different tryptic digestion parameters include temperature, pH, enzyme concentrations, and digestion solvents, as well as the use of equipment for a high increment of temperatures (microwaves and high-intensity focused ultrasound (HIFU)). These studies have achieved increased digestion efficiencies and reduced times compared to conventional overnight tryptic digestions. Among these studies, an improvement in peptide yield was demonstrated after in-gel tryptic digestion was carried out between 50 and 65  C, and after in-solution digestion at pH between 7.8 and 8.1. A comparable peptide yield to an overnight 37  C in-gel digestion with unmodified enzyme was achieved after 30-min digestion at 58  C using a threefold higher enzyme concentration. Some studies have evaluated the use of different solvents to improve digestion efficiencies, including an acetonitrile-containing solvent that increases the digestion efficiency and sequence coverage after a 1-h in-solution tryptic digestion (80% acetonitrile solvent at pH 7.8 and 37  C), and when digestion occurs under microwave irradiation. The use of microwave technology for protein digestion enabled equivalent (in-solution) or better (in-gel) digestion efficiency compared with the standard overnight method, in 6 or 25 min, respectively. Another approach used HIFU for fast (1.2 m3 m
2 day
1 bar
1 (50 L m
2 h
1 bar
1)c

Flux TMP (bar), suction

MF: 0.025–20. 0.1–0.6 bar

a

K-MBR (Kumho Membrane Bioreactors, South Korea). http://www.gewater.com/products/equipment/mf_uf_mbr/p-uf/zbox/s.jsp: UF. c Moulder M (1996) Basic Principles of Membrane Technology, p. 17. London; Kluwer. b

Figure 21 Ultrafiltration membrane bioreactor (Zeeweed) submerged in activated sludge wastewater-treatment tank image courtesy of Zenon/http:// www.water-technology.net/projects/carnation/carnation3.html.

• • • •

COD and biological oxygen demand (BOD) > 90% (>98% for BOD) removal, with permeate typically BOD5 < 5 mg l
1; total suspended solids (TSSs) > 99% removal, permeate typically TSS 90% removal, typically permeate ammonia 90% removal, using appropriate chemical dosing.

GE ZeeWeed MBR technology offers a proven alternative to conventional approaches to the treatment of pharmaceutical wastewater. This wastewater poses particular problems for a conventional-treatment plant because of variations in feed-water strength, floc destabilization and COD loss, and shock loading. ZeeWeed UF membrane technology, in combination with appropriate aeration treatment, is not compromised in this manner and consistently generates an effluent permeate of high quality that is suitable for discharge into even sensitive environments. More information in detail can be found in the monograph ‘The MBR Book: Principles and Applications of Membrane Bioreactors in Water and Wastewater Treatment‘.52

Multistage Continuous High Cell Density Culture

645

Medium feed

Electrodialysis Microfiltration II

Chromatography Bioreactor Crystallization and drying Biomass Final product Osmotic down shock Microfiltration I Figure 22 Ectoine bioprocess (bitop, AG). Cells are grown in membrane bioreactor and lysed with osmotic pressure difference. Further processing is accomplished with electrodialysis, chromatography, and crystallization and drying.

2.43.4.5.2

Halobacteria (Bitop, Germany)

Lentzen and Schwarz (2006) introduced a production of extremolyte ectoine using Halobacteria elongata. Halobacteria is a class of the Euryarchaeota found in water saturated or nearly saturated with salt, and they are also called halophiles, though this name is also used for other organisms that live in somewhat less-concentrated salt water. They are common in most environments where large amounts of salt, moisture, and organic material are available. The use of this bacteria eliminates the chance of contamination during the culture because other common bacteria cannot live in this high salt condition, and downstream processing of any intracellular products from this bacteria is very easy because the cell walls can be partially or completely broken by adding distilled water to the bacterial paste and placing it in a very low osmotic condition. This process will be a good example of intracellular HCDC products. Organisms that are capable of growing at high salt concentration achieve osmotic balance across the cell membrane using lowmolecular-weight polar compounds. Extremolytes means compatible solutes from extremophilic microorganisms. The ability of extremolytes to compensate osmotic pressure and to stabilize macromolecules was studied extensively, and a model for their mode of action of macromolecule stabilization was proposed. A proprietary industrial bioprocess, termed ‘bacterial milking‘, was established for the industrial-scale production of the extremolyte ectoine. The cells are grown in MBRs to a high cell density and further processed to the process shown in Fig. 22. A moderately halophilic H. elongata ectoine producer is grown under high-salt conditions (15%–20% w/v NaCl), and the intracellularly accumulated ectoines are released by applying an osmotic down-shock, leading to the opening of mechanosensitive channels in the inner membrane of H. elongata. This is a quite interesting part of the process because so-called ‘biomilking‘ releases only ectonine from the cells. The biomass is returned to the fermenter for the next round of fermentation, while the product solution is further purified by electrodialysis, chromatography, filtration, evaporation, and crystallization. An ectoine bioprocess with an even higher productivity, based on the continuous fermentation of H. elongata, was developed recently. This continuous ‘permanent milking‘ process is now used by bitop AG for the production of ectoines in metric ton scale. For the production of hydroxyectoine, a bioprocess using the Marinococcus strain M52 was described. Alternatively, hydroxyectoine can be produced with H. elongata by changing the fermentation conditions in the bacterial milking process.

2.43.4.5.3

Animal Cell Culture

New developments. Fed-batch and perfusion culture are the two dominant modes of operation for mammalian cell culture processes, especially for the production of recombinant therapeutic proteins and antibodies required in large amounts.13,48,120 A fed-batch culture of hybridoma cells operated for over 550 h reached a total cell concentration of nearly 5  107 cells ml
l and a peak viable cell concentration of over 1.5  107 cells ml
1 and antibody accumulated to 2.4 g L
l. Also, the culture span was extended to 340–550 h. These exceedingly high concentrations of cells and product were attained by feeding the culture with concentrated nutrients in stoichiometric amounts. The specific production rates for ammonia and lactate were further reduced from 0.0045 to 0.0048 mmol L
cell
1 h
1 in our previous fed-batch experiments to 0.0028 and 0.0036 mmol L
cell
1 h
1, respectively. Only 3.4% of the total glucose consumption was converted into lactate, compared to 67% in a conventional batch culture.127 A novel bioreactor perfusion system was developed for highly efficient production of mAb. The perfusion system was operated at a dilution rate of 0.15 h
1 and the total mAb production by the perfusion system was 1406 mg and that of the batch was 182.4 mg. This means that the perfusion system can produce 7.7 times more mAb than the batch system.115 Chu and Robinson26 reviewed patterns in the technologies applied. Categorizing these products on the basis of their applications leads to therapeutic recombinants vaccines, first followed by tissue-culture products or diagnostic products. There are clearly

646

Multistage Continuous High Cell Density Culture

three principal mammalian cell lines of choice: CHO cells and the murine myeloma lines SP2/0 and NS0. These three cell lines have been used to produce 11 out of 21 approved products and nearly all recombinant therapeutics. First, recombinant therapeutics, which are mainly recombinant proteins or antibodies, typically require more significant production quantities; therefore, this category is the best representation of large-scale protein production. In this category, at least 70% of the licensed processes use stirred-tank bioreactors, and at least 50% use serum-free medium. Second, vaccines and tissue-replacement products are produced in many different cell lines that require specialized bioreactor systems and mediums. Third, although diagnostic products, which are mainly mAbs, are similar to recombinant therapeutics, the required quantities are significantly lower. This allows manufacturers to consider more specialized cell lines and bioreactor systems. Disposable perfusion bioreactors (Wave Biotech, GE health care systems). Wave single-use bioreactors were developed by WAVE Biotech in 1996, which is now a part of GE health care systems. This development led the revolution in single-use technologies that is redefining upstream and downstream processing of biopharmaceuticals today. The bioreactor requires no cleaning or sterilization, providing the ultimate ease of operation and protection against cross contamination. In the WAVE Bioreactor system, cell culture medium and cells contact only a presterile, disposable chamber, the Cellbag, that is placed on a special rocking platform. The rocking motion of this platform induces waves in the cell culture fluid. These waves provide mixing and oxygen transfer, resulting in a perfect environment for cell growth that can easily support over 10  106 cells m l
1. Disposable perfusion bioreactor using wave-induced agitation is a continuous version of wave single-use bioreactors using MF membranes of Food and Drug Administration gas–approved biocompatible PE (Fig. 23). Oxygen transfer and mixing are accomplished by wave-induced agitation caused by rocking the chamber back and forth. The rocking mechanism has been optimized in terms of rocking angle, rocking rate, aeration, and mechanical design to provide a kLa for oxygen transfer of 2–4 h
1. This kLa value is quite low as compared to those for microbial cell cultures up to more than 100. However, the OUR of animal cells are quite low, which can support high cell density cultures of 7  106 cells ml
1. The bioreactor system, unlike spinner flasks, is not limited by gas–liquid transfer surface, and scale-up to 500 L operating volume has been demonstrated. This simple, low-cost, wave bioreactor system can be used for animal, insect, and plant cell culture. The closed design is very suitable for virus production or other applications requiring high containment. The system requires minimal instrumentation and can be operated inside a laboratory incubator. All handling can be done in the open, eliminating the need for a laminar flow cabinet.111

2.43.5

Summary

Rapid development of biotechnology during the last years resulted in the quantitative expansion and diversity of many biotechnology products. Especially, biotechnology in the biorefinery is experiencing an inherent low productivity problem, while current petrochemical industry has been producing its products with the efficiency of high concentration and high productivity. In biotechnology, batch and fed-batch processes have been employed commercially for products that were in high titer but with low productivity.

Inlet air filter Pressure control valve

Exhaust vent filter

Exhaust

Inflated cell bag disposable cultivation chamber

Inlet air

Cell culture media in bag

Wave motion Rocking Motion

Cell culture media in bag Pivot Angle: 5−10° Rate: 5−40 rpm Air: 0.01−0.1 vvm Figure 23

Schematic diagram of wave bioreactor. Continuous operation is also possible.

Base

Multistage Continuous High Cell Density Culture

647

To cope with low productivity problem in industrial and environmental biotechnology processing, MSC-HCDC is proposed. Much work has been done in high cell density continuous culture and multistage continuous bioreactor since the 1970s. These past literature have been gathered in the order of industrial microorganisms, OUR, OTR; substrate, major producing countries, fermentation bulk products, major petroleum products, and performances index of fermentation were defined. Before we start single-stage HCDC or MSC-HCDC, it would be necessary to define whether our target products are extracellular or intracellular. If the product is extracellular, we have to concentrate on the maintenance of HCDC in each reactor and minimize substrate for cell mass formation and maximize its use in product formation. If it is intracellular, the former should be maximized because the cell mass itself is the product. As the result of our efforts of achieving high productivity with high product titer of any designed bioproducts, we can list several important issues and intermediate conclusions of follows: 1. High-cell density is defined as 10 times normal cell density that can be obtained in a culture. 2. MSC-HCDC concept has been proven in mAb production (simulation), but ethanol. 3. MSC-HCDC is defined as transient and steady-state problems: the transient period is unusually long such as 1000 h. However, a steady-state solution can be obtained easily with solving fermentation kinetics equations by setting unsteady state terms to zero. 4. To achieve higher cell density than normal cell density, it is necessary to separate SRT from HRT, which is the most important thing in HCDC. 5. Separating SRT from HRT is important only in the first bioreactor and becomes less important in second to nth reactors because of cell supply from the previous bioreactors. 6. Separation of SRT and HRT can be achieved by immobilization and free cell recycling. 7. Stability and contamination limits the applicability of MSC-HCDC for high productivity process. 8. Animal cell culture of mAb production and mixed culture system, lactic acid production, and some halobacteria process can be candidates of MSC-HCDC system. Reproduced from Kim IH and Chang HN (1983) Variable volume enzyme reactor with ultrafiltration swing: A theoretical study on CSTR Case. AIChE Journal 29: 645–651 and Kim IH and Chang HN (1983) Variable volume hollow fiber enzyme reactor with pulsatile flow. AIChE Journal 29: 910–914. Reproduced from Lee CW and Chang HN (1987) Kinetics of ethanol fermentation in membrane cell recycle fermentors. Biotechnology and Bioengineering 29: 1105–1112. Reproduction from Kourkoutasa Y, Bekatoroua A, Banat IM, et al. (2004) Immobilization technologies and support materials suitable in alcohol beverages production: A review. Food Microbiology 21: 377–397. Reproduced from Chung BH, Chang HN, and Kim IH (1987) Rifamycin B production by Nocardia mediterranei immobilized in a dual hollow fiber bioreactor. Enzyme and Microbial Technology 9: 345–349. Reproduced from Schmidt JE and Ahring BK (1996) Granular sludge formation in upflow anaerobic sludge blanket (UASB) reactors. Biotechnology and Bioengineering 49: 229–246.

References 1. Arden, E.; Lockett, W. T. Experiments on the Oxidation of Sewage without the Aid of Filters. J. Soc. Chem. Ind. 1914, 33, 523–539. 2. Asenjo, J. A.; Merchuk, J. C. In Bioreactor System Design; Asenjo, J. A., Merchuk, J. C., Eds., Dekker: New York, NY, 1995; p 168. 3. Atkinson, B.; Mavituna, F. Properties of Industrially Important Microorganisms. In Biochemical Engineering and Biotechnology Handbook, The Nature Press: New York, NY, 1983; pp 19–21. 4. Bakker, W. A. M.; Overdevest, P. E. M.; Beeftink, H. H.; et al. Serial Air-lift Bioreactors for the Approximation of Aerated Plug Flow. Trends Biotechnol. 1997, 15, 264–269. 5. Banik, G. G.; Heath, C. A. Partial and total Cell Retention in a Filtration-Based Homogeneous Perfusion Reactor. Biotechnol. Prog. 1995, 11, 584–588. 6. Banik, G. G.; Heath, C. A. High-density Hybridoma Perfusion Culture Limitation vs Inhibition. Appl. Biochem. Biotechnol. 1996, 61, 211–229. 7. Bayrock, D. P.; Ingledew, W. M. Ethanol Production in Multistage Continuous, Single Stage Continuous, Lactobacillus-contaminated Continuous, and Batch Fermentations. World J. Microbiol. Biotechnol. 2005, 21, 83–88. 8. Bulock, J. D.; Comberbach, D. M.; Ghommidh, C. Study of Continuous Ethanol Production Using a Highly Flocculent Yeast in the Gas Lift Tower Fermenter. Chem. Eng. J. 1984, 29, B9–B24. 9. Bungay, H. R.; Millsbaugh, M. P. Cross-flow Lamellar Settlers for Microbial Cells. Biotechnol. Bioeng. 1984, 20, 640–641. 10. Buntenmeyer, H.; Bohme, C.; Lehmann, J. Evaluation of Membranes for Use in On-line Cell Separation during Mammalian Cell Perfusion Processes. Cytotechnology 1994, 15, 243–251. 11. Calleja, G. B.; Atkinson, B.; Reichenbach, H.; et al. In Microbial Adhesion and Aggregation; Marshall, K. C., Ed., Springer: New York, NY, 1984; pp 303–321. 12. Calleja, G. B. In The Yeast; Rose, A. H., Harrison, J. S., Eds., 2nd edn.; Vol. 2; Academic Press: London, 1987; pp 165–238. 13. Castilho, L. R.; Medronho, R. A. Cell Retention Devices for Suspended-cell Perfusion. Adv. Biochem. Eng. Biotechnol. 2002, 74, 131–169. 14. Chaabane, F. B.; Alenore, A. S.; Cameleyre, X.; et al. Very High Ethanol Productivity in an Innovative Continuous Two-stage Bioreactor with Cell Recycle. Bioproc. Biosyst. Eng. 2006, 29, 49–57. 15. Chang, F.-Y.; Lin, C.-Y. Biohydrogen Production Using an Up-flow Anaerobic Sludge Reactor. Int. J. Hydrogen Energy 2004, 29, 33–39. 16. Chang, H. N. Biochemical Engineering, Daeyoungsa: Seoul, 1993, 319 pp. 17. Chang, H. N.; Chung, B. H.; Kim, I. H. Dual Hollow-fiber Bioreactor for Aerobic Whole-cell Immobilization. In Separation, Recovery and Purification in Biotechnology: Recent Advances and Mathematical Modelling. ACS Symp. Ser. (Am. Chem. Soc.) 1986, 314, 32–42. 18. Chang, H. N.; Ji, D. J.; Sim, S. J. Citric Acid Production and Scale-up in Dual Hollow Fiber Bioreactor. J. Membr. Sci. 1992, 2, 122–128.

648

Multistage Continuous High Cell Density Culture

19. Chang, H. N.; Kim, B. J.; Jeong, C. M.; et al. Multi-stage CSTR Bioreactor System Equipped with Cell Recycle Unit. PCT WO 2008/091113 A1, 2008. 20. Chang, H. N.; Kim, B. J.; Kang, J. W.; et al. High Cell Density Ethanol Fermentation in an Upflow Packed-bed Cell Recycle Bioreactor. Biotechnol. Bioproc. Eng. 2008, 13, 129–135. 21. Chang, H. N.; Kim, N. J.; Kang, J. W.; Jeong, C. M. Biomass-based Volatile Fatty Acid Platform for Fuels and Chemicals. Biotechnol. Bioproc. Eng. 2010, 15, 1–10. 22. Chang, H. N.; Kim, N. J.; Kang, J.; Jeong, C. M.; Choi, J. D.; Fei, Q.; Kim, B. J.; Kwon, S.; Lee, S. Y.; Kim, J. Multi-stage High Cell Continuous Fermentation for High Productivity and Titer. Bioproc. Biosyst. Eng. 2010, 34 (4), 419–431. 23. Chang, H. N.; Kyung, Y.-S.; Chung, B. H. Glucose Oxidation in a Dual Hollow Fiber Bioreactor with a Silicone Tube Oxygenator. Biotechnol. Bioeng. 1987, 29, 552–557. 24. Chang, H. N.; Oh, D. J. Continuous Production of HcG Monoclonal Antibody in Various Membrane Bioreactors. In Harnessing Biotechnology for the 21st Century, Proceedings of 9th International Biotechnology Symposium and Exposition; Ladisch, M. R., Bose, A., Eds., American Chemical Society: Washington, DC, 1992; pp 563–567. 25. Chang, H. N.; Yoo, I. K.; Kim, B. S. High-density Cell-culture by Membrane-based Cell Recycle. Biotechnol. Adv. 1994, 12 (3), 467–487. 26. Chen, F. High Cell Density Culture of Microalgae in Heterotrophic Growth. Trends Biotechnol. 1996, 14, 421–426. 27. Cheryan, M.; Mehaia, M. A. Ethanol Production in a Membrane Recycle Bioreactor: Conversion of Glucose Using Saccharomyces cerevisiae. Process Biochem. 1984, 15, 204–208. 28. Choi, S. K.; Chang, H. N.; Oh, D. J. Continuous Production of Tissue Plasminogen Activator from Recombinant CHO Cells in a Depth Filter Perfusion System. Biotechnol. Tech. 1995, 9, 567–572. 29. Chu, L.; Robinson, D. K. Industrial Choices for Protein Production by Large-scale Cell Cultures. Curr. Opin. Biotechnol. 2001, 12, 180–187. 30. Chung, B. H.; Chang, H. N. High Density Cell Culture of E. coli in a Dual Hollow Fiber Bioreactor. Korean J. Appl. Microbiol. Bioeng. 1985, 13, 209–212. 31. Chung, B. H.; Chang, H. N. Aerobic Fungal Cell Immobilization in a Dual Hollow-fiber Bioreactor: Continuous Production of Citric Acid. Biotechnol. Bioeng. 1988, 32, 205–212. 32. Chung, B. H.; Chang, H. N.; Goh, Y. H. Dual Hollow Fiber Membrane Bioreactor for Whole Cell Enzyme Immobilization of Streptomyces Griseus with Glucose Isomerase Activity. J. Ferment. Technol. 1987, 65, 575–581. 33. Chung, B. H.; Chang, H. N.; Kim, I. H. Rifamycin B Production by Nocardia Mediterranei Immobilized in a Dual Hollow Fibre Bioreactor. Enzym. Microb. Technol. 1987, 9, 345–349. 34. Cooney, C. L. Bioreactors: Design and Operation. Science 1983, 219, 728–733. 35. Cysewski, G. R.; Wilke, C. R. Utilization of Cellulosic Materials through Enzymatic Hydrolysis. I. Fermentation of Hydrolysate to Ethanol and Single-cell Protein. Biotechnol. Bioeng. 1976, 18, 1297–1313. 36. Cysewski, G. R.; Wilke, C. R. Rapid Ethanol Fermentations Using Vacuum and Cell Recycle. Biotechnol. Bioeng. 1977, 19, 1125–1143. 37. Cysewski, G. R.; Wilke, C. R. Process Design and Economic Studies of Alternative Form of Fermentation Methods for the Production of Ethanol. Biotechnol. Bioeng. 1978, 20, 1421–1444. 38. Deindoerfer, F. H.; Humphrey, A. E. Principles in the Design of Continuous Sterilizers. Appl. Environ. Microbiol. 1959, 7, 264–270. 39. Deshpande, R. R.; Heinzle, E. On-line Oxygen Uptake Rate and Culture Viability Measurement of Animal Cell Culture Using Microplates with Integrated Oxygen Sensors. Biotechnol. Lett. 2004, 26, 763–767. 40. Deutschmann, S. M.; Jager, V. Optimization of the Growth conditions of Sf21insect Cells for High-density Perfusion Culture in Stirred-tank Bioreactors. Enzym. Microb. Technol. 1994, 16, 506–512. 41. EBC Microbiologica. Method 2.5.2. Flocculation. J. Inst. Brew. 1981, 87, 319–320. 42. Furey, J. Scale-up of a Cell Culture Perfusion Process. Gen. Eng. Biotechnol. News 2002, 22, 62. 43. Ghose, T. K.; Tyagi, R. D. Rapid Ethanol Fermentation of Cellulose Hydrolysate. I. Batch versus Continuous Systems. Biotechnol. Bioeng. 1979, 28, 1387–1400. 44. Gooijer, C. D.; Beeftink, H. H.; Tramper, J. Optimum Design of a Series of Continuous Stirred Tank Reactors Containing Immobilized Growing Cells. Biotechnol. Lett. 1996, 18, 397–402. 45. Herbert, D. A Theoretical Analysis of Continuous Culture Systems. Soc. Chem. Ind. (Lond.) Monogr. 1961, 12, 21–53. 46. Hiller, G. W.; Clark, D. S.; Blanch, H. W. Cell Retention-chemostat Studies of Hybridoma Cells – Analysis of Hybridoma Growth and Metabolism in Continuous Suspension Culture on Serum-free Medium. Biotechnol. Bioeng. 1993, 42, 185–195. 47. Hoffmann, H.; Kuhlmann, W.; Meyer, H. D.; Schugerl, K. High Productivity Ethanol Fermentations with Cross-flow Membrane Separation. J. Membr. Sci. 1985, 22, 235–243. 48. Hu, W.-S.; Aunins, J. G. Large-scale Mammalian Cell Culture. Curr. Opin. Biotechnol. 1997, 8, 148–153. 49. Inloes, D. S.; Taylor, D. P.; Cohen, S. N.; et al. Ethanol Production by Saccharomyces cerevisiae Immobilized in Hollow-fiber Membrane Bioreactors. Appl. Environ. Microbiol. 1983, 46, 264–278. 50. Janssens, J. H.; Bernard, A.; Bailey, R. B. Ethanol from Whey: Continuous Fermentation with Cell Recycle. Biotechnol. Bioeng. 1984, 26, 1–5. 51. Jin, Y. L.; Speers, R. A. Flocculation of Saccharomyces cerevisiae. Food Res. Int. 1998, 31, 421–440. 52. Judd, S.; Judd, C. In The MBR Book: Principles and Applications of Membrane Bioreactors in Water and Wastewater Treatment, Elsevier: Oxford, 2006. 53. Jung, S. G.; Lee, W. G.; Chang, H. N. Ethanol Fermentation by Cell Recycle Fermentor with a Fabric Filter. Korean J. Biotechnol. Bioeng. 1990, 5, 159–165. 54. Kamm, B.; Gruber, P. R.; Kamm, M. In Vols. 1 and 2; Wiley: Weinheim, 2006. 55. Karel, S. F.; Libicki, S. B.; Robertson, C. R. The Immobilization of Whole Cells: Bioengineering Principles. Chem. Eng. Sci. 1985, 40, 1321–1354. 56. Kim, B. J.; Chang, H. N.; Oh, D. J. Application of a Cell-once-through Perfusion Strategy for Production of Recombinant Antibody from RCHO Cells in a Centritech Lab II Centrifuge System. Biotechnol. Prog. 2007, 23, 1186–1197. 57. Kim, B. J.; Chang, H. N.; Oh, D. J. Limited Use of a Centritech Lab II Centrifuge in Perfusion Culture of RCHO Cells for Production of Recombinant Antibody. Biotechnol. Prog. 2008, 24, 166–174. 58. Kim, B. S.; Lee, S. C.; Lee, S. Y.; et al. Production of Poly(3-polyhydroxybutyric Acid) by Fed-batch Culture of Alcaligenes Eutrophus with Glucose Concentration Control. Biotechnol. Bioeng. 1994, 43, 892–898. 59. Kim, D. J.; Chang, H. N.; Liu, J. R. Plant Cell Immobilization in a Dual Hollow Fiber Bioreactor. Biotechnol. Tech. 1989, 3, 139–144. 60. Kim, D. Y.; Lee, J. C.; Chang, H. N.; Oh, D. J. Development of Serum-free Media for a Recombinant CHO Cell Line Producing Recombinant Antibody. Enzym. Microb. Technol. 2006, 39, 426–433. 61. Kim, I. H.; Chang, H. N. Variable Volume Enzyme Reactor with Ultrafiltration Swing: A Theoretical Study on CSTR Case. AIChE J. 1983, 29, 645–651. 62. Kim, I. H.; Chang, H. N. Variable Volume Hollow Fiber Enzyme Reactor with Pulsatile Flow. AIChE J. 1983, 29, 910–914. 63. Kitai, A.; Tone, K. H.; et al. Method of Fermentation Utilizing a Multi-stage Fermenting Device. AU38 US Patent 3,743,582, 1973. 64. Koehler, G.; Milstein, C. Continuous Cultures of Fused Cells Secreting Antibody of Predefined Specificity. Nature 1975, 256, 495–497. 65. Kourkoutasa, Y.; Bekatoroua, A.; Banat, I. M.; et al. Immobilization Technologies and Support Materials Suitable in Alcohol Beverages Production: A Review. Food Microbiol. 2004, 21, 377–397. 66. Kroner, K. H.; Schutte, H.; Hussedt, H.; Kula, M. R. Crossflow Filtration in Downstream Processing of Enzymes. Process Biochem. 1984, 19, 67–74. 67. Kuriyama, J.; Ishibashi, H.; Miyagawa, H.; et al. Optimization of Two-stage Ethanol Fermentation Using Flocculating Yeast. Biotechnol. Lett. 1993, 15, 415–420. 68. Kwon, S. H.; Yoo, I.-K.; Lee, W. G.; et al. High-rate Continuous Production of Lactic Acid by Lactobacillus Rhamnosus in a Two-stage Membrane Cell-recycle Bioreactor. Biotechnol. Bioeng. 2001, 73, 25–34. 69. Kyung, Y.-S.; Peshwa, M. V.; Gryte, D. M.; Hu, W.-S. High Density Culture of Mammalian Cells with Dynamic Perfusion Based on On-line Oxygen Uptake Rate Measurements. Cytotechnology 1994, 14, 183–190. 70. Ladish, M. R. In Bioseparations Engineering: Principles, Practice, and Economics, Wiley: New York, NY, 2001; pp 16–19.

,

Multistage Continuous High Cell Density Culture

649

71. Le, M. S.; Atkinson, T. Cross-flow Microfiltration for Recovery of Intracellular Products. Process Biochem. 1985, 20, 26–31. 72. Lee, C. W.; Chang, H. N. Kinetics of Ethanol Fermentation in Membrane Cell Recycle Fermentors. Biotechnol. Bioeng. 1987, 29, 1105–1112. 73. Lee, J. C.; Chang, H. N.; Oh, D. J. Recombinant Antibody Production by Perfusion Cultures of RCHO Cells in a Depth Fitler Perfusion System. Biotechnol. Prog. 2005, 21, 134–139. 74. Lee, J. C.; Kim, D. Y.; Oh, D. J.; Chang, H. N. Long-term Operation of Depth Filter Perfusion Systems (DFPS) for Monoclonal Antibody Production Using Recombinant CHO Cells: Effect of Temperature, PH, and Dissolved Oxygen. Biotechnol. Bioproc. Eng. 2008a, 13, 401–409. 75. Lee, J. C.; Kim, D. Y.; Oh, D. J.; Chang, H. N. Two-stage Depth Filter Perfusion Culture for Recombinant Antibody Production by Recombinant Chinese Hamster Ovary Cell. Biotechnol. Bioproc. Eng. 2008b, 13, 560–565. 76. Lee, S. Y. High Cell-density Culture of Escherichia coli. Trends Biotechnol. 1996, 14, 98–105. 77. Lee, Y. L.; Chang, H. N. High Cell Density Culture of a Recombinant Escherichia coli Producing Penicillin Acylase in a Membrane Cell Recycle Fermenter. Biotechnol. Bioeng. 1990, 36, 330–337. 78. Lehmann, J.; Vorlop, J.; Buntemeyer, H. Bubble-free Reactors and Their Development for Continuous Culture with Cell Recycle. Anim. Cell Biotechnol. 1998, 3, 221–237. 79. Lentzen, G.; Schwarz, T. Extremolytes: Natural Compounds from Extremophiles for Versatile Applications. Appl. Microbiol. Biotechnol. 2006, 72 (4), 623–634. 80. Lettinga, G.; van Velsen, A. F. M.; de Zeeuw, W.; Hobma, S. W. Feasibility of the Upflow Anaerobic Sludge Blanket (UASB) – Process. In National Conference on Environmental Engineering, American Society of Civil Engineers: New York, NY, 1979; p 35. 81. Lettinga, G.; van Velsen, A. F. M.; Hobma, S. W.; et al. Use of the Upflow Sludge Blanket (USB) Reactor Concept for Biological Wastewater Treatment, Especially for Anaerobic Treatment. Biotechnol. Bioeng. 1980, 22, 699–734. 82. Maiorella, B. L.; Blanch, H. W.; Wilke, C. R. Economic Evaluation of Alternative Ethanol Fermentation Processes. Biotechnol. Bioeng. 1984, 26, 1003–1025. 83. Maiorella, B. L. In Comprehensive Biotechnology, Vol. 3, Moo-Young, M., Ed.; Pergamon: Oxford, 1985; pp 879–885. 84. Margaritis, A.; Wilke, C. R. The Rotorfermentor. II. Application to Ethanol Fermentation. Biotechnol. Bioeng. 1978, 20, 737–753. 85. McCarthy, J. L. In Alcoholic Fermentation of Sulfite Waste Liquor, Industrial Fermentations (2), 97, Underkofler, L. A., Hickey, R. J., Eds.; Chemical Publishing Co: New York, 1954. 86. Mehaia, M. A.; Cheryan, M. Ethanol Production in a Hollow Fiber Bioreactor Using Saccharomyces cerevisiae. Appl. Microbiol. Biotechnol. 1984, 20, 100–104. 87. Mehaia, M. A.; Cheryan, M. Immobilization of Lactobacillus Bulgaricus in a Hollow Fiber Bioreactor for Production of Lactic Acid from Acid Whey Concentrates. Process Biochem. 1987, 22, 185–188. 88. Mulder, M. Basic Principles of Membrane Technology, Kluwer: London, 1996; p 17. 89. Nagashima, M.; Azuna, M.; Noguchi, S.; et al. Continuous Ethanol Fermentation Using Immobilized Yeast Cells. Biotechnol. Bioeng. 1984, 26, 992–997. 90. Navarro, J. M.; Durand, G. Modification of Yeast Metabolism by Immobilization to Porous Glass. Eur. J. Appl. Microbiol. 1977, 4, 243–254. 91. Nishiwaki, A.; Dunn, I. J. Performance of a Multistage Fermentor with Cell Recycle for Continuous Ethanol Production. Chem. Eng. Commun. 1997, 162, 179–198. 92. Nishiwaki, A.; Dunn, I. J. Analysis of a Two-stage Fermentor with Recycle for Continuous Ethanol Production. Chem. Eng. Commun. 1998, 168, 207–227. 93. Nishiwaki, A.; Dunn, I. J. Analysis of the Performance of a Two-stage Fermentor with Cell Recycle for Continuous Ethanol Production Using Different Kinetic Models. Biochem. Eng. J. 1999, 4, 37–44. 94. Nishiwaki, A.; Dunn, I. J. Performance of a Two-stage Fermentor with Cell Recycle for Continuous Production of Lactic Acid. Bioproc. Biosyst. Eng. 1999, 21, 299–305. 95. Nishiwaki, A.; Dunn, I. J. Comparison of Lactic Acid Productivities at High Substrate Conversions in a Continuous Two-stage Fermentor with Cell Recycle Using Different Kinetic Models. Chem. Eng. Commun. 2005, 192, 219–236. 96. Nishiwaki, A.; Dunn, I. J. Operating Conditions for High Production Rates of Ethanol in a Continuous Fermentor with Cell Recycle. Can. J. Chem. Eng. 2005, 83, 904–908. 97. Noble, J. GE ZeeWeed MBR Technology for Pharmaceutical Wastewater Treatment. Membr. Technol. 2006, 2006, 7–9. 98. Norton, S.; D‘Amore, T. Physiological Effects of Yeast Cell Immobilization: Applications for Brewing. Enzym. Microb. Technol. 1994, 16, 365–375. 99. Oh, D. J.; Chang, H. N. High-density Continuous Cultures of Hybridoma Cells in a Dual Hollow Fiber Bioreactor. Biotechnol. Tech. 1992, 6, 77–82. 100. Oh, D. J.; Choi, S. K.; Chang, H. N. High-density Continuous Cultures of Hybridoma Cells in a Depth Filter Perfusion System. Biotechnol. Bioeng. 1994, 44, 895–901. 101. Park, J. K.; Chang, H. N. Microencapsulation of Microbial Cells. Biotechnol. Adv. 2000, 18, 303–319. 102. Pirt, S. J.; Kurowski, W. M. An Extension of the Theory of the Chemostat with Feedback of Organisms. Its Experimental Realization with a Yeast Culture. J. Gen. Microbiol. 1970, 63, 357–366. 103. Purwadi, R.; Taherzadeh, M. J. The Performance of Serial Bioreactors in Rapid Continuous Production of Ethanol from Dilute-acid Hydrolyzates Using Immobilized Cells. Bioresour. Technol. 2008, 99, 2226–2233. 104. Reynolds, T. D.; Richards, P. A. In Unit Operations and Processes in Environmental Engineering, PWS: Boston, 1996. 105. Rittmann, B. E.; McCarty, P. L. Environmental Biotechnology: Principles and Applications, McGraw-Hill: Singapore, 2001; pp 207–434. 106. Robertson, C. R.; Kim, I. H. Dual Aerobic Hollow-fiber Bioreactor for Cultivation of Streptomyces Aureofaciens. Biotechnol. Bioeng. 1985, 27, 1012–1020. 107. Rogers, P. L.; Lee, K. J.; Triber, D. E. High Productivity Ethanol Fermentations with Zymomonas Mobilis. Process Biochem. 1980, 15, 7–11. 108. Samson, R.; Leduy, A. Multistage Continuous Cultivation of Blue-green Alga Spirulina Maxima in the Flat Tank Photobioreactors with Recycle. Can. J. Chem. Eng. 1985, 63, 105–112. 109. Schmidt, J. E.; Ahring, B. K. Granular Sludge Formation in Upflow Anaerobic Sludge Blanket (UASB) Reactors. Biotechnol. Bioeng. 1996, 49, 229–246. 110. Shang, L.; Tian, P.-Y.; Kim, N.-J.; et al. Effects of Oxygen Supply Modes on the Production of Human Growth Hormone in Different Scale Bioreactors. Chem. Eng. Technol. 2009, 32, 600–605. 111. Singh, V. Disposable Bioreactor for Cell Culture Using Wave-induced Agitation. Cytotechnology 1999, 30, 149–158. 112. Stanbury, P. F.; Hall, S. J.; Whitaker, A. In Principles of Fermentation Technology, 2nd ed.; Pergamon: Oxford, 1999. 113. Stratford, M. Yeast Flocculation: Restructuring the Theories in Line with Recent Research. Cerevisia 1996, 21, 38–45. 114. Tanase, T.; Ikeda, Y.; Iwama, K.; Hashimoto, A.; Kataoka, T.; Tokushima, Y.; Kobayashi, T. Comparison of Micro-filtration Hollow-fiber Bioreactors for Mammalian-cell Culture. J. Ferment. Bioeng. 1997, 83, 499–501. 115. Tang, Y.-J.; Ohashi, R.; Hamel, J.-F. P. Perfusion Culture of Hybridoma Cells for Hyperproduction of IgG2a Monoclonal Antibody in a Wave Bioreactor-perfusion System. Biotechnol. Prog. 2007, 23, 255–264. 116. Taticek, R. A.; Lee, C. W. T.; Shuler, M. L. Large-scale Insect and Plant Cell Culture. Curr. Opin. Biotechnol. 1994, 5, 165–174. 117. Tejayadi, S.; Cheryan, M. Lactic-acid from Cheese Whey Permeate – Productivity and Economics of a Continuous Membrane Bioreactor. Appl. Microbiol. Biotechnol. 1995, 43, 242–248. 118. Vick Roy, T. B.; Blanch, H. W.; Wilke, C. R. Lactic Acid Production by Lactobacillus Delbreuckii in a Hollow Fiber Fermenter. Biotechnol. Lett. 1982, 4, 483–488. 119. Vick Roy, T. B.; Blanch, H. W.; Wilke, C. R. Microbial Hollow Fiber Bioreactors. Trends Biotechnol. 1983, 1, 135–139. 120. Voisard, D.; Meuwly, F.; Ruffieux, P.-A.; et al. Potential of Cell Retention Techniques for Large-scale High-density Perfusion Culture of Suspended Mammalian Cells. Biotechnol. Bioeng. 2003, 82, 751–765. 121. Waites, M. J.; Morgan, N. L.; Rockey, J. S.; Higton, G. In Industrial Microbiology: An Introduction, Wiley: London, 2001. 122. Walsh, T. J.; Bungay, H. R. Shallow-depth Sedimentation of Yeast Cells. Biotechnol. Bioeng. 1979, 21, 1081–1084. 123. Wilcocks, K.; Smart, K. A. The Importance of Surface Charge and Hydrophobicity for the Flocculation of Chain-forming Brewing Yeast Strains and Resistance of These Parameters to Acid Washing. FEMS (Fed. Eur. Microbiol. Soc.) Microbiol. Lett. 1994, 134, 293–297. 124. Williams, D.; Munnecke, D. M. The Production of Ethanol by Immobilized Yeast Cells. Biotechnol. Bioeng. 1981, 23, 1813–1825.

650

Multistage Continuous High Cell Density Culture

125. Wittlich, P.; Capan, E.; Schlieker, M.; et al. Entrapment in LentiKats. In Fundamentals of Cell Immobilization Biotechnology; Nedovic, V. A., Wallaert, R., Hofmann, M., Jozef, A., Eds.; Focus on Biotechnology, Vol. 8A; Kluwer: Dordrecht, 2004; pp 53–63. 126. Xavier, A. M.; Gonçalves, A. M. D.; Moreira, J. L.; Carrondo, M. J. T. Operational Patterns Affecting Lactic Acid Production in Ultrafiltration Cell Recycle Bioreactor. Biotechnol. Bioeng. 1995, 45, 320–327. 127. Xie, L.; Wang, D. I. C. High Cell Density and High Monoclonal Antibody Production through Medium Design and Rational Control in a Bioreactor. Biotechnol. Bioeng. 1996, 51, 725–729. 128. Xu, T. J.; Zhao, X. Q.; Bai, F. W. Continuous Ethanol Production Using Self-flocculating Yeast in a Cascade Fermentors. Enzym. Microb. Technol. 2005, 37, 634–640. 129. Yoo, I.-K.; Seong, G. H.; Chang, H. N.; Park, J. K. Encapsulation of Lactobacillus Casei Cells in Liquid-core Alginate Capsules for Lactic Acid Production. Enzym. Microb. Technol. 1996, 19, 428–433. 130. Yu, H.-Q.; Mu, Y. Biological Hydrogen Production in a UASB Reactor with Granules. II: Reactor Performance in 3-year Operation. Biotechnol. Bioeng. 2006, 94, 988–995. 131. Zabriskie, D. W. In Comprehensive Biotechnology, Vol. 2, Moo-Young, M., Ed.; Pergamon: Oxford, 1985; pp 175–190. 132. Zhang, S.; Handacorrigan, A.; Spier, R. E. A Comparison of Oxygenation Methods for High-density Perfusion Cultures of Animal-cells. Biotechnol. Bioeng. 1993, 41, 685–692.

Integrated Production and Separationq

2.44

Katsutoshi Hori, Department of Biomolecular Engineering, Graduate School of Engineering, Nagoya University, Nagoya, Japan Hajime Unno, Department of Biotechnology, Tokyo Institute of Technology (Retired), Yokohama, Japan © 2019 Elsevier B.V. All rights reserved. This is an update of K. Hori, H. Unno, Integrated Production and Separation, Reference Module in Life Sciences, Elsevier, 2017.

2.44.1 2.44.2 2.44.3 2.44.3.1 2.44.3.2 2.44.3.3 2.44.4 2.44.5 2.44.6 2.44.6.1 2.44.6.2 2.44.6.3 2.44.6.4 2.44.6.5 2.44.7 2.44.7.1 2.44.7.2 2.44.7.3 2.44.8 References

Introduction Integration Methodology for Reducing Process Step Cross-sectional Technologies Through Integration Methodology Continuous Operation Immobilization of Biocatalyst Membrane Separation Separation Techniques for the Integration in Terms of Product Characteristics Bioreactor Configuration for the Integration of Production and Separation Techniques for ISPR Vacuum and Pervaporation Two-Liquid-Phase Systems Size-Selective Permeation Complex Formation Product Adsorption Process Integration by Biotechnology Secretion of Intracellular Proteins Utilization of Affinity Separation for Proteinaceous Products Programming of Self-disruption Perspective for the Process Integration

651 653 654 654 654 655 655 656 656 656 657 659 659 659 659 659 660 660 661 662

Glossary Bioreactor Equipment (reactors) for performing biochemical reaction using biocatalysis such as enzymes and cells. Continuous operation A way of process operation. In continuous operation, raw materials are continuously fed into a reactor and products are also recovered from the reactor continuously. Fed-batch A biotechnological batch process which is based on feeding of a growth-limiting nutrient substrate to a culture. The fed-batch strategy is typically used in bio-industrial processes to reach a high cell density in the bioreactor. Immobilize In bioprocesses, to bind enzymes or cells with a solid or insoluble carrier or to entrap them in a curtain space like gel. Mass transfer Transfer of chemical species between two phases through an interface or diffusion through a phase. The driving force for mass transfer is a difference in concentration. For separation processes, mass transfer determines the rate at which the separation will occur.

2.44.1

Introduction

Bioprocess usually consists of three processes – upstream, production/bioreactor, and downstream processes – as shown in Fig. 1. In most cases, each process is followed by the treatment of various waste materials excreted from the respective process. In the production process, raw materials are converted into products by biochemical reactions employing biocatalyst such as enzymes, microbial cells, animal cells, and plant cells. In the upstream process, materials used for the production process (e.g., raw materials, biocatalyst, and medium) are prepared. In the downstream process, target products are separated from reaction mixture and purified and/or adjusted to a required level.

q

Change History: April 2018. Katsutoshi Hori made text revisions in sections. References have been updated to reflect more up-to-date material. This is an update of K. Hori, H. Unno, 1.32 - Integrated Production and Separation, Editor(s): Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 579–590.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00132-4

651

652

Integrated Production and Separation

Raw materials

Upstream process

Production process

Downstream process

Preparation biocatalyst medium Pretreatment raw materials

Biochemical reaction in bioreactor

Separation Purification Modification

Products

Waste-treatment process Figure 1

General process flow of bioproduction.

In view of the process economy, the cost for separation and purification process frequently accounts for more than half of the total process cost. Therefore, the downstream process is deeply assessed in order to construct a cost-competitive bioprocess. To attain a required level of purity, many steps for separation and/or purification have been applied sequentially, resulting in an increase in the downstream process cost. Many steps in the process result in not only increases in equipment costs but also decreases in the whole separation efficiency of the process, even though the recovery ratio of each step is high enough. For example, if the recovery ratio of each step is equally a, the whole recovery ratio of the process consisting of n steps is an. This implies that the total recovery ratio of five steps decreases to 0.77 even if that of each step (a) is high as 0.95. Therefore, besides improvement of the recovery ratio of each step, a decrease in the number of steps (n) by the process design is much effective for increasing the efficiency of whole downstream process. In the above case, the whole efficiency improves to 0.86 when the steps are decreased from five to three. To decrease the step n, utilization of bioaffinity such as antigen–antibody reaction will be a key technology. For example, the recovery ratio was demonstrated to greatly improve from 16% to 81% by changing a combination process of precipitation, several kinds of chromatography, and electrophoresis into a single step using a monoclonal-antibody column. Configuration of the downstream process is affected by the features of the upstream and/or production processes. The structure of the downstream process strongly depends on product concentrations from bioreactor. A higher concentration of target product from bioreactor can be treated by a simpler and faster downstream process, resulting in a lower-cost product formation. A remarkable correlation between the product concentrations and the total production costs has been shown. Therefore, bioreactor design and/or operation aiming at attaining higher product concentration and/or decreasing the steps of downstream process are another concern for the whole-process economy. One such approach is the integration of some process steps among upstream, bioreactor, and downstream processes, where chemical and/or biotechnological characteristics of the relevant materials are taken into consideration. Design of bioreactor itself will be under the concept of process integration. Accumulation of by-products in a bioreactor sometimes inhibits the bioreaction, preventing increased product concentrations, resulting in decreased separation efficiency. The form of biocatalyst, for example, enzyme or whole cells, immobilized or suspended form, also largely affects the efficiency of the downstream process. Immobilized cells or enzymes can be easily removed from the reaction mixture; however, immobilized catalysis frequently decreases reaction efficiency and total reaction rate due to conformational changes of enzymes and the mass transfer resistance. Integration of the downstream process into the bioreactor will bring us other various advantages such as prevention of product inhibition, inhibition of reverse reaction, prevention of product conversion into other chemicals by further reaction, and minimization of product loss due to uncontrolled removal from the system such as evaporation. As for extracellular products, in situ product removal (ISPR), which is the fast removal of product from a producing cell, thereby preventing its subsequent interference with cellular or medium components, has been extensively studied. The extensive reviews by Roffler et al.1 and Freeman2 summarized much of the research on ISPR carried out previously. ISPR primarily requires the identification of properties by which the product differs from the background medium, thus allowing easy removal. On the other hand, intracellular products, which are synthesized and accumulated in cells, require cell disruption, which not only increases the processing steps in the downstream process but also make the processes complex and difficult due to necessity of removing cell lysis materials containing various types of proteins, saccharides, and lipids. In addition, cells themselves are biocatalysts that are produced through cultivation by supplying nutrients including carbon and nitrogen sources. Hence, production of the cells is costly. Disruption of the cells implies discarding precious biocatalyst; it is impossible to realize final products that are less expensive than cells themselves. Therefore, methodologies for the secretion of intracellular products without cell disruption will be one approach for the process integration. In any process consideration mentioned above, it must always be taken into consideration that most bioproducts such as proteins are sensitive to process conditions such as temperature, pH, and ionic strength, easily losing their activity under inappropriate conditions.

Integrated Production and Separation

2.44.2

653

Integration Methodology for Reducing Process Step

The processing pathway of the bioproducts produced using whole microbial cells as biocatalyst from bioreactor is summarized in Fig. 2. The starting material for the downstream process is either in a liquid phase or in a solid phase (cells) for extracellular or intracellular products, respectively. The latter needs cell separation from the culture broth and product removal from the separated cells by methods such as cell disruption, after which a process similar to that for the former is followed. Process integration aiming at the decrease in processing step can be divided into four categories as shown in Fig. 2. The first is the implementation of the method for removal of bioproducts from culture broth directly by using various biomolecular characteristics, which is indicated as ISPR. The second is an invention of bioreactor itself by means of separation devices such as membrane or cell immobilization. The third is to modify the cell features by biotechnological methods such as genetic modification. The last is the implementation of cell treatment method into the bioreactor by utilizing genetic principles. According to the four categories, strategies for process integration can be discussed to determine which steps in the respective processes can be integrated and/or which steps can be omitted. In the first category shown in Fig. 2, extracellular products can be directly removed from culture broth containing cells in a bioreactor by one or more methods of ISPR. In the second category, the separation of cells and supernatant in the culture broth in a bioreactor is considered, where centrifugation is the most popular method. This step can be skipped by cell isolation techniques such as cell-immobilization or membrane separation. The immobilized cells are already in a form of separation from culture liquid. Cell immobilization allows simultaneous product separation during bioreaction from the bioreactor. Membranes can retain cells in a bioreactor during continuous recovery of the culture liquid from the bioreactor. In the third category, target products are intracellular. One effective methodology for this case is to alter them into extracellular in nature by fusing a secretion sequence to them. As another method, cell surface display is a potential technology for production extracellularly. This method allows performing bioconversion at the outer surface of the cells by displaying enzymes that are responsible for the relevant reaction on the cell surfaces. Thus secreted products or compounds produced extracellularly can be separated from culture broth in the same methods as ISPR. If it is impossible to use these methods, intracellular products must be separated from cell lysate after cell disruption. This is the fourth category shown in Fig. 2. To skip both the cell separation and cell disruption steps, the cell features are modified genetically

Bioreactor Biochemical reaction Culture broth

2

Separation 3

1

Supernatant Extracellular products

Cell Intracellular products

Separation

4

Disruption

Purification Cell lysate

Extracellular products

Separation Purification Intracellular products

Integration methods ISPR; immobilization of cells; membrane separation; programming self-disruption

secretion; cell surface display;

Figure 2 Strategy map for integration of production and downstream processes. Solid lines denote conventional process flow. Broken lines denote the integration process flows which skip the conventional steps.

654

Integrated Production and Separation

beforehand for self-disruption under a programmed condition. The condition will be that the programmed cell is auto-lyzed as soon as a sufficient amount of accumulation of the products in a cell is attained. This integration method decreases the number of separation steps and simultaneously minimizes the intracellular products to be degraded or further metabolized into other chemicals.

2.44.3

Cross-sectional Technologies Through Integration Methodology

Integrated production and separation conspicuously demonstrates its efficiency in continuous operation of a bioreactor. Immobilization of biocatalyst and utilization of membrane separation are themselves means of integrated production and separation to skip a cell separation step, but they also have advantages in the continuous operation through the retention of cells in a high concentration in the bioreactor. Continuous operation, immobilization of biocatalyst, and membrane separation are effective crosssectional technologies through all methods for the integrated production and separation.

2.44.3.1

Continuous Operation

Continuous operation is almost equal to integrated production and separation because it implies continuous bleeding of culture broth, from which products can be separated continuously by an external separation unit. In other words, continuous cell culture lends itself to online integration with the downstream processing. This also allows monitoring of changes as a function of product quality and therefore adjustment of cell conditions as necessary to maintain product consistency.3 Batch processes are difficult to monitor, since the effects of any changes that occur are integrated throughout the entire volume of the vessel. Long-term continuous bioreactions are more economical, and real-time process monitoring is possible, which becomes increasingly important for a number of reasons. For example, mammalian cell culture gives product concentrations of typically 1–100 mg L 1. This high background makes it difficult to monitor product quality in the cell culture medium. Monitoring the purified product from the integrated separation and purification process, instead of the broth, is extremely advantageous.

2.44.3.2

Immobilization of Biocatalyst

Immobilization of enzymes or cells is a widely used method for simplification of the separation process and is employed in various operation modes from batch to continuous operations in a continuous-flow stirred-tank bioreactor (CSTB), fixed-bed plug-flow bioreactor (PFB), or fluidized-bed bioreactor, with and without recycle of the medium. However, one of the greatest merits of immobilized biocatalyst is that it enables continuous or repeated use of precious biocatalyst. Immobilization is frequently coupled with one of the ISPR methods such as evaporation, extraction, and adsorption because it minimizes damage of the biocatalyst from direct contact of the biocatalyst with toxic solvent for extraction or adsorbent in an external unit.2 Enzymes can be immobilized to water-insoluble supports by a wide variety of methods, including physical adsorption, covalent binding or cross-linking, and entrapment. On the other hand, immobilized cells can carry out more complex reactions than immobilized enzymes due to the far more extensive and sophisticated biochemical capabilities of living cells compared with a single enzyme. Whole-cell biocatalysts can also be simpler to prepare. In general, the methods for immobilizing whole cells are similar to those for enzyme immobilization. They are categorized into (1) carrier-free immobilization through cross-linking or flocculation; (2) immobilization onto a preformed carrier through adsorption or covalent attachment; (3) immobilization in the course of carrier formation through entrapment or encapsulation; and (4) immobilization within a preformed semipermeable membrane. Recently, biofilms, which are accumulated biomasses of microbial cells and extracellular polymeric substances (EPS) on solid surfaces, have attracted attention as a form of natural immobilization4,5 or passive immobilization.6 Immobilization using biofilms are also classified into three categories, (1), (2), and (4) stated above: (1) granule formed by autoflocculation, (2) biofilms formed onto a carrier surface, and (4) biofilms formed onto one side of a membrane. Conventionally, biofilms have been used in the fields of wastewater and off-gas treatments and bioremediation. Now, they are studied for their application to the production of conventional fermentation products, biofuels, and chemicals.7–11 However, the use of biofilms for the production of a chemical has several drawbacks: difficulty in screening a special microbial strain that possesses both high capability to produce the target chemical and the ability to form a biofilm,5 a long startup period for biofilm formation,5 and difficulty in controlling ideal biofilms with high biocatalytic activity and appropriate thickness in the balance between cell concentration and mass transfer limitation.5,9,11 Very recently, a novel nanofibrous protein AtaA, which mediates notably tenacious adhesion of bacterial cells to various material surfaces ranging from hydrophobic plastics to hydrophilic glass and stainless steel, was discovered.12 A novel method for bacterial cell immobilization using AtaA was invented. Immobilization with AtaA enabled exclusive cell growth in the support space and only a few cells existed in the bulk medium. Immobilization of resting cells drastically increased cell concentration, depending on the support material; dry cells of approximately 110 g/L could be immobilized onto glass wool.13 Immobilized recombinant bacterial cells with AtaA greatly improved the productivity of a chemical through enhancing the tolerance of the cells to a toxic substrate14 and were efficiently used for repetitive reactions.13 This new method is expected to overcome drawbacks of conventional immobilization methods, such as the limitation of mass transfer in the inner part of a gel,15,16 fragility of gels, leakage of cells from a support matrix, adverse effect on cell viability and catalytic activity,4 and easy detachment from physical adsorption.

Integrated Production and Separation 2.44.3.3

655

Membrane Separation

Membrane separation is also combined with other integration methods of production and separation. The principle of the membrane separation is based on the difference in permeability of materials and substances. The driving force for the separation is given by the difference in the pressure, concentration, and electric potential. Fig. 3 shows pressure-driven membrane separation modes depending on the size of separation targets. Microfiltration (MF) is frequently used for separation of microbial cells from medium in CSTB and PFB in the combination with other ISPR techniques such as product adsorption, extraction, and evaporation. The major mode of application of MF in this context is for retention of microbial cells17 or mammalian cells and hybridoma cells.3 In addition, like immobilization of cells, cell separation by MF minimizes damage of the biocatalyst from direct contact of the biocatalyst with toxic solvent or adsorbent.2 Therefore, it is a growing trend to use MF as an intermediate component for cell separation in a cascade operation rather than the sole means for ISPR. Membranes can also be used for enhanced selective product removal, in which it is essential to maintain a concentration driving force by removing products on the downstream side. For example, membranes have been applied to extractive fermentation, in which organic solvents are used for ISPR, to separate the aqueous and organic phases.

2.44.4

Separation Techniques for the Integration in Terms of Product Characteristics

Separation methods integrated with the production process can be determined by the chemical characteristics of bioproducts. Therefore, products can be categorized for the integration methodology by their chemical characteristics such as volatility, molecular weight, hydrophobicity, charge, solubility, composing chemicals (protein, sugar, lipid, etc.), specific affinity to chemicals, and so on. Possible separation techniques depending on characteristics of products for the integration with production are listed in Table 1. One of the important and unique points of bioprocesses, compared with conventional chemical processes, is that they can produce protein that can be designed and modified in the cells by genetic manipulation. Proteinaceous products can be previously designed for simplifying the downstream process. For example, introduction of a His-tag at the N-terminus or C-terminus of the protein for simple separation by affinity chromatography is a widely used method in biotechnology. Because the proteinaceous products already contain the separation tool, this method can be considered one of the examples of the integration of production and separation. This method can also be considered to be a way for the integration of the upstream and downstream processes because the cells, which contain the DNA engineered for products integrated with the separation means, are designed before production. Thus, in the case of proteins, this effective method for the integration can be utilized. Introduction of a signal peptide at the N-terminus of target protein for secretion is also categorized in the same methodology for the integration. When products bind to specific chemicals, separation methods using bioaffinity such as affinity chromatography are some of the most efficient methods for integrated production and separation. Because this method using molecular recognition has high selectivity against a target product, purification of the products can be carried out with high efficiency without additional purification steps. Interactions between antigen and antibody, protein A and antibody (immunoglobulin G (IgG)), and lectin and sugar are famous examples for utilization of affinity separation. Introduction of the His-tag mentioned above is also a method for utilizing bioaffinity. RO (1–10 MPa)

UF (0.1–1 MPa) MF (10–200 kPa)

NF (1–10 MPa)

Salt

Saccharide

0.1 nm

1 nm

Protein

Virus

Bacteria

10 nm

100 nm

1 µm

Animal cells Yeast 10 µm

Figure 3 Pressure-driven membrane separation mode depending on the size of separation targets. RO, reverse osmosis; NF, nanofiltration; UF, ultrafiltration; MF, microfiltration.

Table 1

Separation techniques for integration in terms of product characteristics

Characteristics of products

Separation techniques for integration with production

Volatile Hydrophobic Hydrophilic Low molecular weight Proteinaceous Complexation

Vacuum, pervaporation Organic solvent extraction, hydrophobic adsorption Aqueous biphasic extraction, ion exchange Permeation (dialysis) Secretion, affinity separation Precipitation

656

2.44.5

Integrated Production and Separation

Bioreactor Configuration for the Integration of Production and Separation

Operation mode of a bioreactor, that is, batch, fed-batch, or continuous with or without cell retention or immobilization, and specific needs for aeration and sterile operation determine the ease of integration of separation into the production process.2 Of course, as mentioned above, whether target products are intracellular or extracellular determines how to design separation modules for subsequent steps after cell separation depending on whether the cell fraction or culture supernatant should be further subjected to purification steps to obtain a final product. Bioreactor configurations for integration of the separation process can be divided into two types in which the separation is performed within the bioreactor (internally) or outside the bioreactor (externally).2 The latter type has a separate loop through which a fraction of the medium is circulated for product separation. Although the internal type has an advantage for rapid equilibration and transfer of the product from each individual cell surface to the product pool, resulting in lower product concentrations in the reactor, which may be critical in the case of chemically reactive or highly toxic products, they also have engineering constraints such as limited solids concentration in stirred-tank bioreactors. Although most membrane modules used to be employed as an external unit, submerged membrane modules assembled by porous hollow-fiber membrane made from polymer materials are frequently employed as an internal unit in a bioreactor. Extraction with organic solvents is also carried out both internally and externally. In the internal case, a water-immiscible organic solvent is directly added into the bioreactor, and in the external case the medium is circulated through an external extraction unit. In the latter case, by using immobilized cells or membrane, the direct contact between the extracting solvent and the producing cells can be prevented, reducing the toxic effects of the solvent on cell activity. In contrast, extraction by means of an aqueous two-phase system is usually carried out within the bioreactor. Reversible complex formation for ISPR has been carried out with soluble reagents in the medium within the bioreactor.18 Adsorbents for product immobilization are sometimes directly introduced into a bioreactor.1 However, external adsorption columns, which are separated from a bioreactor by means of membrane modules19 or immobilization of either the producing cell20 or the adsorbent21 are used when problems arise from the direct contact of the adsorbent with the cells and the medium. Configurations of bioreactors integrated with separation module are depicted in Fig. 4. The integrated systems are categorized by the localization of products, extracellular (A to E) or intracellular (F and G) products. Once intercellular products are released into the medium by cell disruption or constrained secretion from cells, same methods with extracellular products can be employed for integration of production and separation. The operation modes are divided into two modes, batch (A, B, C, and F) and continuous modes (D, E, and G). Separation is carried out internally within a bioreactor (A, E, and F) or externally at the outside of the bioreactor (B to D and G). Cell capturing systems utilizing immobilized cells or membrane are also integrated into a bioreactor (C to E). For intracellular products, there is a strategy in which cell disruption system is integrated into the production process (F and G). Because cell disruption in the bioreactor implies the end of biochemical reaction, the process must be a batch system (F). For continuous operation, a fraction of cells should be disrupted externally (G).

2.44.6

Techniques for ISPR

Even intracellular products can be targets of ISPR once they are released into the medium by cell disruption or constrained secretion from cells. Thus, ISPR techniques are the center of the issue in the integration of production and separation. The important ISPR techniques are explained here. Five techniques for ISPR are listed as follows2: (1) evaporation, via vacuum fermentation, gas stripping, or pervaporation, which are effective for low-molecular-weight volatile product such as ethanol and butanol; (2) extraction into another phase, for example, water-immiscible organic solvent, supercritical fluid, or second aqueous phase; (3) size-selective permeation using membranes; (4) reversible complex formation based on chemical reaction with soluble reagents or biological recognition, leading to soluble or insoluble complexes and (5) product immobilization via adsorption or specific binding onto water-insoluble polymeric carriers. Affinity chromatography for the biochemical products is contained in this category.

2.44.6.1

Vacuum and Pervaporation

Volatile fermentation products can be separated during fermentation by maintaining the bioreactor under vacuum so that the products are boiled off at the normal temperature of the fermentation.1 Ethanol and butanol have been the sole targets of this method. However, a severe problem of this method is the accumulation of toxic nonvolatile byproducts. To solve this problem, continuous vacuum fermentation with liquid bleed has been carried out, where some of the broth from the bioreactor was continuously removed. This method enables feeding a concentrated sugar substrate, resulting in less water requirement in the production process and therefore reduction of wastewater-treatment costs. However, even with liquid bleed for removal of nonvolatile byproducts, large amounts of by-product carbon dioxide must be compressed from the bioreactor pressure up to atmospheric pressure using bulky expensive compressors. To overcome this problem, flash fermentation, in which the bioreactor remains at atmospheric pressure while broth is circulated to a vacuum chamber where products are continuously boiled off, was developed. In the gas stripping, the fermentation gas is bubbled through the fermentation broth and then passed through a condenser for solvent recovery.22 The stripped gas is then recycled back to the bioreactor and the process continues until all the substrate in the

Integrated Production and Separation

657

Batch system for extracellular products

A

B

C

Continuous operation system for extracellular products D

E

S

S

P

P

System for intracellular products G

F Batch

Continuous operation S

P Bioreactor

Internal cell disruption system

Separation module

External cell disruption system

Cell retention system

S Substrate

P Product

Figure 4 Conceptual schematics for integrated production and separation: (A) batch system integrated with internal separation; (B) batch system integrated with external separation; (C) batch system integrated with internal cell retention system and external separation; (D) continuous operation system integrated with internal cell retention and external separation; (E) continuous operation system integrated with cell retention, internal separation, and regeneration system of the separation carrier; (F) batch system integrated with internal cell disruption; and (G) continuous operation system integrated with external cell disruption and external separation. Modified from Ref. 2.

bioreactor is utilized. Fed-batch fermentation integrated with gas stripping successfully reduced substrate inhibition and increased cell mass in acetone–butanol–ethanol production.23 Pervaporation is a membrane-based process that allows selective removal of volatile compounds from fermentation broth. The membrane is placed in contact with the fermentation broth, and the volatile liquids or solvents diffuse through the membrane as a vapor which is recovered by condensation.22 Both liquid and solid pervaporation membranes have been used. Pervaporation was identified as the ISPR technique having the greatest potential for improving butanol production.24 Recently, by integration with a pervaporation system using an ionic liquid polydimethylsiloxane ultrafiltration membrane, the overall solvent productivity in continuous acetone–butanol–ethanol fermentation by Clostridium acetobutylicum was increased remarkably.25

2.44.6.2

Two-Liquid-Phase Systems

Two-phase liquid systems for integrated production and separation can be divided into the aqueous/organic biphasic system and the aqueous biphasic system. These liquid–liquid extractive bioconversion processes seem to have the greatest potential among the

658

Integrated Production and Separation

different approaches to integrating reaction and product recovery.26 In particular, the aqueous/organic biphasic system has been, so far, extensively studied by many researchers and has been at the center of the study about ISPR. Hydrophobic bioproducts can be extracted during fermentation by contacting the broth with a suitable organic solvent which is insoluble in the broth.1 Products dissolving into the solvent can later be recovered by distillation or back extraction. Product removal is performed internally or externally, as mentioned in the previous section. The concept of the two-phase partitioning bioreactor can also be applied to controlled delivery of a toxic substrate dissolved in an organic phase to a cell-containing aqueous phase.26 In this case, the solvent should be directly added into the bioreactor (internal removal) to which the toxic substrates are supplied (Fig. 5). This type of the bioreactor is called two-liquid-phase partitioning bioreactors (TPPBs). It is desirable to choose a solvent that has a high capacity, is selective for the fermentation product, and is relatively nontoxic to the fermenting microorganisms.1 However, many solvents with a high extraction capacity often have poor selectivity and also exhibit toxicity to the producing cells. In general, a positive correlation is found between hydrophobicity of solvents and nontoxicity for biocatalysts. A measure for hydrophobicity for characterization of solvents is the log P value which is the logarithmic of the partition coefficient of a solvent in a water–octanol two-phase system. Solvents with a log P value above four are very hydrophobic and generally show no toxic effects on biocatalysts. Organic solvents, for example, dodecanol, dibutyl phthalate, kerosene, dioctyl phthalate, octanol, cyclohexane, silicon oil, dodecane, hexadecane, and tetradecane, have been used in two-liquid phase systems. According to the solvent selection strategies presented by Bruce and Daugulis,27 it is possible to compose mixtures of solvents that show good extraction capacity as well as good biocompatibility even at relatively high concentrations of toxic solvents. Another problem of the direct addition of solvents to a bioreactor is emulsification. Frequently, the emulsion is so stable because of cell adhesion to oil surfaces that the organic phase containing the products is difficult to be separated. To avoid the emulsification problem, the external extraction in the combination of cell immobilization or use of membrane is preferable to the internal extraction. Utilization of solvent-tolerant microorganisms is another approach to overcome the problem of the toxicity of solvents added to a bioreactor. Solvent-tolerant bacteria are promising as biocatalysts for the production of highly toxic chemicals in TPPBs. Since Inoue and Horikoshi first (in 1989) reported Pseudomonas putida that is capable of growing in culture media containing more than 50% toluene,19 many bacterial strains such as Pseudomonas sp. and Rhodococcus sp. that are tolerant to solvents have been isolated. Because in TPPBs, the organic phase acts not only for product removal but also as a toxic substrate reservoir and keeps solventtolerant microorganisms in the aqueous phase separated from the substrates and the products, the substrate concentration can be increased with the expectation of the accelerating a chemical conversion rate. In this system, however, mass transfer of substrates dissolved in the solvent (i.e., substrate uptake by microorganisms) is a rate-limiting step in many cases. Employment of microorganisms that take up substrates by direct contact with the organic phase has an advantage in the rapid conversion in TPPBs. Bacterial strains showing monolayer adsorption to oil surfaces have also been reported to be effective for microbial conversion at oil–water interfaces.28 To overcome some of the problems with organic liquids in extractive fermentation, aqueous two-phase systems have been used for the integration of production and separation.1 Rather than using an organic solvent as the second phase, polymers are added to the broth until two separate phases form. The phases contain 85%–95% water and are normally biocompatible. Microbial cells often remain in one phase while low-molecular-weight products are distributed evenly between phases. Although the product is

Single-aqueous-phase system

aqueous phase hydrophobic and toxic substrates and products

Two-liquid-phase partitioning bioreactors (TPPBs)

organic phase • water-immiscible • non-bioavailable • nontoxic

bacteria

aqueous phase • Low solubility of hydrophobic substrates

Reservoir tank of hydrophobic, toxic substrates

• Toxic effect on bacterial cells • Increase in the substrate concentration • Reduction of the toxic effect Figure 5

Two-liquid-phase system for internal process integration.

Integrated Production and Separation

659

not concentrated by extraction, the phase volumes can be adjusted so that the phase containing the cells is much smaller than the other phase. Most of the product will then be in the phase without cells and can be drawn off and processed by distillation or other means. Polyethylene glycol and dextran are frequently used as polymers to form the incompatible two phases.

2.44.6.3

Size-Selective Permeation

The successful application of membrane modules for separation, up to industrial scale, serves as a sound basis for the integration of membranes with bioreactors.2 In dialysis fermentation, a selectively permeable membrane separates a bioreactor from a medium reservoir.1 Nutrients in the reservoir diffuse to the bioreactor while metabolite products diffuse to the medium reservoir. In general, dialysis fermentation not only relieves product inhibition but also retains cells so that high cell densities are achieved; thus, smaller, less expensive bioreactors could be used. The dialysis fermentation is generally categorized as membrane bioreactor (MBR). The detail was already explained above and also there are many reports about MBRs.

2.44.6.4

Complex Formation

Reversibly soluble polyelectrolytes, whose solubility in aqueous solutions is dependent on factors, such as pH, ionic strength, and temperature, are used in the integrated system as supports for enzymes or affinity ligands.18 Their dual nature takes advantage of the soluble form during enzymatic reactions and the insoluble form during separation of products. The ability cyclodextrins to form crystalline–insoluble complexes with organics can be used for a selective separation of dilute products in fermentation.29 It was demonstrated that butanol was selectively precipitated from the acetone–butanol–ethanol and butanol–iso-propanol systems and butyric acid from the butyric acid–acetic system. Employment of epichlorohydrin-cross-linked b-cyclodextrin in ISPR for product immobilization also showed high yield and easy pure product recovery in g-decalactone production from castor oil by a filamentous fungus.30

2.44.6.5

Product Adsorption

Product removal by immobilization via adsorption onto polymeric matrices has been demonstrated for a relatively large variety of products such as ethanol, salicylic acid, cycloheximide, anthraquinones, alkaloids, monoterpenes, and tissue plasminogen activator.2 Activated carbon, zeolites, polystyrene beads, cellulose, and ion exchanges served as the adsorbing matrix, mainly on the basis of hydrophobic interactions. A limitation of these materials may be the nonspecific adsorption of organic compounds from solution. This results in the removal of both products and substrates leading to decreased product removal efficiencies as well as increased substrate loadings to compensate for sorbed materials. However, hydrophobic polymeric resins can be effective for specific adhesion of hydrophobic products such as butanol from culture broth containing hydrophilic substrates.31 Thermoplastic polymers can be an alternative to both immiscible organic solvents and solid-phase adsorbents.32 Uptake of small molecules into thermoplastic polymers is analogous to that of uptake into organic solvents, which operate on an adsorption mechanism, while solid adsorbents operate on a surface adsorption phenomenon. These polymers have been successfully used partition and deliver toxic organic molecules in biphasic systems. For example, a polyether ester thermoplastic that consists of cylindrical-shaped pellets is used to recover a toxic product, 3-methylcatechol, from the cell broth.

2.44.7

Process Integration by Biotechnology

Biotechnological methods are powerful tools for the integration of production and separation processes. However, the methods are a pool of various ideas and still in the state-of-the-art meaning a bit far from systematic description of them. Here, some examples of such biotechnology applicable to the integration are discussed.

2.44.7.1

Secretion of Intracellular Proteins

Secreted production of proteins is one of the most important strategies for integrated production and separation of proteinaceous products, in particular, recombinant proteins. Intracellular production requires many steps in the downstream process from cell disruption, separation of inclusion body, solubilization, and refolding of a target protein product before purification. For the secretion of intracellular proteins, secretion pathways have been utilized. In bacteria, the general secretory pathway called Sec system is well known.33 The Sec system consists of membrane proteins SecY, SecE, and SecG, which form a channel on a cytoplasmic membrane for translocation of unfolding proteins, and the intracellular protein SecA, which mediates to conveyance of presecretory proteins into the channel depending on energy from adenosine triphosphate (ATP) hydrolysis. The pre-secretory protein possesses a signal peptide consisting of 20–40 amino acids at N-terminal region as a precursor. The pre-secretory protein passes through the channel pore in an unfolding extended string form.34,35 By this system, protein is secreted out of cells in the case of Gram-positive bacteria, or into the periplasm in the case of Gram-negative bacteria. Therefore, addition of a signal peptide sequence to the N-terminal of a target protein enables the secreted production of the protein that is originally an intracellular protein. Expression vectors designed for this purpose are commercially available. Although the Sec system has been studied for efficient

660

Integrated Production and Separation

protein-secreted production system for more than 20 years, there remains a limitation; some proteins that are important in industry cannot be secreted effectively by the Sec system. Recently, the secondary general secretory pathway, twin-argentine translocation pathway (Tat pathway), first found in a chloroplast of plant cells has been shown to be conserved in bacteria.36,37 Proteins secreted by this pathway also possess a signal peptide, which contains a specific motif of Arg–Arg–F–F (F denotes a hydrophobic amino acid) at its N-terminal. Unlike the Sec system, the Tat pathway transports folded proteins using proton motive force. In the Tat pathway, there is a hitchhiker mechanism, in which an enzyme complex is formed in the cells, and one of the constitutive subunits that are appended to the Tat signal peptide cotranslocates the other subunits in the complex.37 These unique properties of the Tat pathway are expected to expand the possibility of application in a protein bioprocess. This system has been applied to the secreted production of Fab antibody fragment in E. coli.38 Generally, disulfide bonds are not formed in the cytoplasm due to its reducing environment. The heavy chain fragment tagged with the Tat signal peptide and the light chain fragment were produced, and the disulfide bonds were formed in the cytoplasm of the mutant Escherichia coli strain, whose cytoplasm is in oxidative environment. By this strategy, correctly folded Fab fragment was successfully secreted.

2.44.7.2

Utilization of Affinity Separation for Proteinaceous Products

Proteins such as antibodies and enzymes are typical and important products from bioprocesses. Proteinaceous products are unstable and easily degraded by proteases released from cells. Integration of production and product separation minimizes the loss of proteinaceous products by disintegration. Generally speaking, proteins physiologically function through the specific interaction with other molecules. Affinity separation utilizes these interactions, that is, affinity between respective proteins and their ligands such as antigen–antibody, enzyme–substrate, enzyme–inhibitor, cytokine–receptor, and hormone–receptor. There have been many commercially available affinity separation systems for proteinaceous products (Table 2), and they can also be employed in the integrated production and separation by utilizing inline affinity chromatography. After all, affinity chromatography has much higher purification efficiency and higher recovery than other types of chromatography and also is capable of treating a large volume of samples so that it should be used in an early stage of the purification steps. Therefore, affinity chromatography is one of the efficient methods for integrated purification into production for proteinaceous products. For example, restriction endonucleases were produced as a model protein by continuous two-stage cultivation with a recombinant E. coli strain, which was integrated with continuous cell disintegration and purification by affinity chromatography.39 In this study, the enzymes were expressed as fusion proteins consisting of protein A from Staphylococcus aureus, and Fc-region of IgGsepharose-FF column was used for the affinity chromatography.

2.44.7.3

Programming of Self-disruption

As mentioned previously, the intracellular products have to be processed first by cell disruption after harvesting cells from broth in bioreactor, then transferred to the succeeding separation processes. On the contrary, extracellular products are unnecessary to harvest cells and their disintegration. If the products are modified to be excreted genetically as mentioned above, the process for cell treatment can be skipped. However, such modification is not necessarily effective for all kinds of products and cells, especially for nonproteinous products. In the case of difficulty in excretion modification, another idea comes out. One such idea is to genetically modify the cell features into self-disruptive after sufficient amount of product is accumulated in the cell. Fig. 6 shows an Table 2

Examples of commercially available carrier for affinity chromatography

Carrier

Target for separation

Nickel Glutathione Protein A IgG Heparin Concanavalin A Lectins Streptavidin Calmodulin Gelatin Cibacron blue 3 GA Procion Red HE-3B

His-tag-fused protein Glutathione S-transferase (GST)-fused protein IgG Protein A-fused protein Fibronectin, fibroblast growth factor, hepatocyte growth factor Glycoprotein, membrane protein Glycoprotein Biotinylated protein ATPase, protein kinase, phosphodiesterase Gelatinase, fibronectin NAD(P)-dependent protein, interferon NADP-dependent protein, matrix metalloproteinases, tissue inhibitor of metalloproteinases Plasminogen, plasminogen activator Prothrombin, plasminogen activator Serine protease

Lysine Arginine Benzamidine

Integrated Production and Separation

661

A

6

15

5 4

10

3 2

Glucose (g l–1)

PHB present in supernatant (g l–1)

20 7

5

1 0

0

10

20 30 Time (h)

40

0 50

20

100

6

80

15 5 4

10

3 2

5

Glucose (g l–1)

PHB present in supernatant (g l–1)

7

60

40

20

1 0 0

10

20 30 Time (h)

40

0 50

0

Relative efficiency of cell disruption (%)

B

Figure 6 Programmed self cell-disruption: (A) wild-type strain and (B) transformant. Cell disruption of transformant was initiated after complete exhaustion of glucose in medium followed by release of PHB accumulated in the cell (B), while wild-type strain showed scarce release of PHB due to no cell disruption (A). Modified from Hori K, Kaneko M, Tanji Y, Xing X-H, Unno H. 2002. Construction of self-disruptive Bacillus megaterium in response to substrate exhaustion for polyhydroxybutyrate production. App. Microbiol. Biotechnol. 59: 211–216.

example for such modification, where product, polyhydroxybutyrate (PHB), synthesized and accumulated in Bacillus megaterium intracellularly was successfully solubilized into the supernatant by a program in which lysis gene from Bacillus amyloliquefaciens phage was engineered into the cell expressing the lytic function by triggered by the exhaustion of glucose in the medium.40 Strategic viewpoints on cell disruption in bioprocess design, partially aiming at the process integration, are discussed in the review article by B. Balasundaram et al.41

2.44.8

Perspective for the Process Integration

Bioprocesses are now expected as alternative production processes to the chemical production process developed so far because they are considered to be energy-saving processes under mild conditions (ambient temperature and pressure), decrease emission of pollutants, employ carbon neutral raw materials, utilize organic waste as raw materials, and so on. However, many bioprocesses encompass the drawbacks of high production costs. To make it possible to adopt the bioprocess as an alternative process to conventional one, process integration will surely be one strategy. Research on integration so far is mostly based on specific ideas. Although most of them are a combination of conventional separation tools of hardware, soft approaches are also employed through biotechnological modification of cells like excretion of products and adhesiveness for immobilization given by a tenacious adhesin,12–14 modification of product characteristics appropriate for processing, effective utilization of raw materials and biocatalysts by simultaneous production of effective products aiming at the decrease of production cost.42 Approaches to process integration by biotechnological features will surely be accelerated in a variety of integration types by the development of new technologies.

662

Integrated Production and Separation

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42.

Roffler, S. R.; Blanch, H. W.; Wilke, C. R. In Situ Recovery of Fermentation Products. Trends Biotechnol. 1984, 2, 129–136. Freeman, A.; Woodley, J. M.; Lilly, M. D. In Situ Product Removal as a Tool for Bioprocessing. Nat. Biotechnol. 1993, 11, 1007–1012. Szathmary, S.; Grandics, P. Bioreactor Integration with Downstream Processing. Biotechnology 1990, 8, 924–925. Junter, G. A.; Jouenne, T. Immobilized Viable Microbial Cells: from the Process to the Proteome. or the Cart before the Horse. Biotechnol. Adv. 2004, 22, 633–658. Gross, R.; Hauer, B.; Otto, K.; Schmid, A. Microbial Biofilms: New Catalysts for Maximizing Productivity of Long Term Biotransformations. Biotechnol. Bioeng. 2007, 98, 1123–1134. Cheng, K. C.; Demirci, A.; Catchmark, J. M. Advances in Biofilm Reactors for Production of Value-added Products. Appl. Microbiol. Biotechnol. 2010, 87, 445–456. Urbance, S. E.; Pometto, A. L., 3rd; Dispirito, A. A.; Denli, Y. Evaluation of Succinic Acid Continuous and Repeat-batch Biofilm Fermentation by Actinobacillus succinogenes Using Plastic Composite Support Bioreactors. Appl. Microbiol. Biotechnol. 2004, 65, 664–670. Qureshi, N.; Annous, B. A.; Ezeji, T. C.; Karcher, P.; Maddox, I. S. Biofilm Reactors for Industrial Bioconversion Processes: Employing Potential of Enhanced Reaction Rates. Microb. Cell Factories 2005, 4, 24. Rosche, B.; Li, X. Z.; Hauer, B.; Schmid, A.; Buehler, K. Microbial Biofilms: a Concept for Industrial Catalysis? Trends Biotechnol. 2009, 27, 636–643. Cheng, K.-C.; Demirci, A.; Catchmark, J. M. Continuous Pullulan Fermentation in a Biofilm Reactor. Appl. Microbiol. Biotechnol. 2011, 90, 921–927. Halan, B.; Buehler, K.; Schmid, A. Biofilms as Living Catalysts in Continuous Chemical Syntheses. Trends Biotechnol. 2012, 30, 453–465. Ishikawa, M.; Nakatani, H.; Hori, K. AtaA, a New Member of the Trimeric Autotransporter Adhesins from Acinetobacter Sp. Tol 5 Mediating High Adhesiveness to Various Abiotic Surfaces. PLoS One 2012, 7, e48830. Hori, K.; Ohara, Y.; Ishikawa, M.; Nakatani, H. Effectiveness of Direct Immobilization of Bacterial Cells onto Material Surfaces Using the Bacterionanofiber Protein AtaA. Appl. Microbiol. Biotechnol. 2015. Ishikawa, M.; Shigemori, K.; Hori, K. Application of the Adhesive Bacterionanofiber AtaA to a Novel Microbial Immobilization Method for the Production of Indigo as a Model Chemical. Biotechnol. Bioeng. 2014, 111, 16–24. Cassidy, M.; Lee, H.; Trevors, J. Environmental Applications of Immobilized Microbial Cells: a Review. J. Ind. Microbiol. Biotechnol. 1996, 16, 79–101. Carballeira, J. D.; Quezada, M. A.; Hoyos, P.; Simeo, Y.; Hernaiz, M. J.; Alcantara, A. R.; Sinisterra, J. V. Microbial Cells as Catalysts for Stereoselective Red-ox Reactions. Biotechnol. Adv. 2009, 27, 686–714. Holst, O.; Mattiasson, B. Cultivation Using Membrane Filtration and Cell Recycling. Bioprocess Technol. 1991, 11, 11–26. Fujii, M.; Taniguchi, M. Application of Reversibly Soluble Polymers in Bioprocessing. Trends Biotechnol. 1991, 9, 191–196. Inoue, A.; Horikoshi, K. A Pseudomonas Thrives in High Concentrations of Toluene. Nature 1989, 338, 264–266. Takamatsu, S.; Ryu, D. D. Y. Recirculating Bioreactor-separator System for Simultaneous Biotransformation and Recovery of Product: Immobilized L-aspartate b-decarboxylase Reactor System. Biotechnol. Bioeng. 1988, 32, 184–191. Dykstra, K. H.; Li, X.-M.; Wang, H. Y. Computer Modeling of Antibiotic Fermentation with On-line Product Removal. Biotechnol. Bioeng. 1988, 32, 356–362. Lee, S. Y.; Park, J. H.; Jang, S. H.; Nielsen, L. K.; Kim, J.; Jung, K. S. Fermentative Butanol Production by Clostridia. Biotechnol. Bioeng. 2008, 101, 209–228. Ezeji, T. C.; Qureshi, N.; Blaschek, H. P. Acetone Butanol Ethanol (ABE) Production from Concentrated Substrate: Reduction in Substrate Inhibition by Fed-batch Technique and Product Inhibition by Gas Stripping. Appl. Microbiol. Biotechnol. 2004, 63, 653–658. Groot, W. J.; van der Lans, R. G. J. M.; Luyben, K. C. A. M. Technologies for Butanol Recovery Integrated with Fermentations. Process Biochem. 1992, 27, 61–75. Izák, P.; Schwarz, K.; Ruth, W.; Bahl, H.; Kragl, U. Increased Productivity of Clostridium Acetobutylicum Fermentation of Acetone, Butanol, and Ethanol by Pervaporation through Supported Ionic Liquid Membrane. Appl. Microbiol. Biotechnol. 2008, 78, 597–602. Malinowski, J. J. Two-phase Partitioning Bioreactors in Fermentation Technology. Biotechnol. Adv. 2001, 19, 525–538. Bruce, L. J.; Daugulis, A. J. Solvent Selection Strategies for Extractive Bioeatalysis. Biotechnol. Prog. 1991, 7, 116–124. Watanabe, H.; Tanji, Y.; Unno, H.; Hori, K. Rapid Conversion of Toluene by an Acinetobacter sp. Tol 5 Mutant Showing Monolayer Adsorption to Water-oil Interface. J. Biosci. Bioeng. 2008, 106, 226–230. Shity, H.; Bar, R. New Approach for Selective Separation of Dilute Products from Simulated Clostridial Fermentation Broths Using Cyclodextrins. Biotechnol. Bioeng. 1992, 39, 462–466. Dukler, A.; Freeman, A. Affinity-based in Situ Product Removal Coupled with Co-immobilization of Oily Substrate and Filamentous Fungus. J. Mol. Recogn. 1998, 11, 231–235. Nielsen, D. R.; Prather, K. J. In Situ Product Recovery of N-butanol Using Polymeric Resins. Biotechnol. Bioeng. 2009, 102, 811–821. Prpich, G. P.; Daugulis, A. J. A Novel Solid-liquid Two-phase Partitioning Bioreactor for the Enhanced Bioproduction of 3-methylcatechol. Biotechnol. Bioeng. 2007, 98, 1008–1016. Pugsley, A. P. The Complete General Secretory Pathway in Gram-negative Bacteria. Microbiol. Rev. 1993, 57, 50–108. Levy, R.; Weiss, R.; Chen, G.; Iverson, B. L.; Georgiou, G. Production of Correctly Folded Fab Antibody Fragment in the Cytoplasm of Escherichia coli trxB gor Mutants via the Coexpression of Molecular Chaperones. Protein Expr. Purif. 2001, 23, 338–347. Danese, P. N.; Silhavy, T. J. Targeting and Assembly of Periplasimic and Outer-membrane Proteins in Escherichia coli. Annu. Rev. Genet. 1998, 32, 59–94. Benanti, E. L.; Nguyen, C. M.; Welch, M. D. Virulent Burkholderia Species Mimic Host Actin Polymerases to Drive Actin-based Motility. Cell 2015, 161, 348–360. Berks, B. C.; Palmer, T.; Sargent, F. Protein Targeting by the Bacterial Twin-arginine Translocation (Tat) Pathway. Curr. Opin. Microbiol. 2005, 8, 174–181. Rodrigue, A.; Chanal, A.; Beck, K.; Müller, M.; Wu, L. F. Co-translocation of a Periplasmic Enzyme Complex by a Hitchhiker Mechanism through the Bacterial Tat Pathway. J. Biol. Chem. 1999, 274, 13223–13228. Beer, H. D.; Maschke, H. E.; Schugerl, K. Continuous Production of Restriction Endonucleases: Continuous Two-stage Cultivation with E. coli JM103; Continuous Cell Disintegration and Purification by Affinity Chromatography. Appl. Microbiol. Biotechnol. 1992, 38, 220–225. Hori, K.; Kaneko, M.; Tanji, Y.; Xing, X. H.; Unno, H. Construction of Self-disruptive Bacillus megaterium in Response to Substrate Exhaustion for Polyhydroxybutyrate Production. Appl. Microbiol. Biotechnol. 2002, 59, 211–216. Balasundaram, B.; Harrison, S.; Bracewell, D. G. Advances in Product Release Strategies and Impact on Bioprocess Design. Trends Biotechnol. 2009, 27, 477–485. Marsudi, S.; Unno, H.; Hori, K. Palm Oil Utilization for the Simultaneous Production of Polyhydroxyalkanoates and Rhamnolipids by Pseudomonas aeruginosa. Appl. Microbiol. Biotechnol. 2008, 78, 955–961.

2.45

Bioreactor Models and Modeling Approaches

Constantinos Theodoropoulos and Chenhao Sun, School of Chemical Engineering and Analytical Science, Biochemical and Bioprocess Engineering Group, University of Manchester, Manchester, United Kingdom © 2019 Elsevier B.V. All rights reserved.

2.45.1 Introduction 2.45.2 Black-Box Models 2.45.2.1 Macroscopic Kinetic Models 2.45.3 Other Enzyme Kinetics 2.45.3.1 Double Michaelis–Menten Kinetics 2.45.3.2 Enzyme Inhibition Kinetics 2.45.3.3 Hill Kinetics 2.45.4 Other Microbial Kinetics 2.45.4.1 Contois Kinetics 2.45.4.2 Powell Kinetics 2.45.4.3 Moser Kinetics 2.45.4.4 Tessier Kinetics 2.45.4.5 Logistic Equation 2.45.4.6 Haldane–Andrew Model 2.45.4.7 Diauxic Growth 2.45.4.8 Luedeking–Piret Equation 2.45.5 Summary 2.45.5.1 Classic Cybernetic Modeling 2.45.5.2 Hybrid Artificial Neural Network Approach 2.45.6 Grey-Box Models 2.45.6.1 Constraint-Based Stoichiometric Modeling 2.45.6.2 Flux Balance Analysis 2.45.7 Dynamic Flux Balance Analysis 2.45.7.1 Cybernetic Metabolic Modeling 2.45.8 White-Box Models 2.45.8.1 Current Challenges and Progress 2.45.9 Reactor Models 2.45.10 Conclusions References Relevant Websites

663 664 664 665 665 666 666 666 666 666 666 667 667 667 667 667 667 668 669 669 670 671 672 673 674 675 675 677 677 680

Glossary Industrial biotechnology A set of practices that utilize living cells or enzymes to generate industrial products and processes. Metabolic engineering The practice of optimizing genetic and regulatory processes within cells to increase the cells’ production of a certain substance. Computational fluid dynamics A tool which uses computer and applied mathematics to model fluid flow, heat transfer, mass and momentum transfer and related phenomena. Genome-scale metabolic networks Networks of intracellular biochemical reactions constructed based on information of genome sequencing. Metabolic modeling The practice of analyzing the properties and capabilities of metabolic networks.

2.45.1

Introduction

There is an imperative global need to pursue bioeconomy, which seeks to provide bio-based alternatives to conventional petrochemical products in the fundamental value chains in energy, chemical and material sectors through the deployment of biological manufacturing processes.1,2 Bioprocesses have increasingly employed microorganisms as mini-factories, due to their diverse metabolic pathways and their concomitant ability to convert a wide spectrum of renewable raw materials to value-adding compounds.3–5

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00144-0

663

664

Bioreactor Models and Modeling Approaches

Through the use of biotechnology, bioprocesses are able to directly or indirectly produce value-adding bio-based platform chemicals, biopolymers, fuels and biotherapeutic products.6–9 Although the bioprocess industry has not yet reached the same maturity as chemical industries,10 it is at a stage of rapid development. With sustained growth and increasing commercial opportunities in bioprocess sectors, the forward momentum looks promising for future development.11 Along with future promise, progress also brings about a host of new challenges. In regard to fermentation processes, the biggest challenges arise from four different aspects: economic (e.g., cost-effectiveness, productivity),12 scale-up (i.e., scalability of process),13 operational (i.e., process control and parameter monitoring)14 and strain development (e.g., strain robustness). To develop the bioprocesses of the future, ongoing interdisciplinary research continues to tackle current challenges, by optimizing process parameters, exploiting inexpensive and sustainable substrates, and creating genetically engineered strains with boosted production potency. Whichever aspect of research is pursued, it can significantly benefit from the comprehensive insights generated by mathematical modeling. Not only can modeling provide a clear picture of a biological system whose intrinsic complexity is beyond intuitive understanding, enabling intuition for further innovation, but also guide bioprocess optimization through the efficient use of predictive simulators. As an approximate representation of reality, mathematical models are indispensable for promoting basic and applied research. The greatest value of mathematical models lies in their ability to minimize unnecessary experiments which would otherwise be incurred by labor-intensive and time-consuming trial-and-error approaches. In chemical processes, design of complex catalytic reaction systems are based on mass and energy balance equations. The balance equations contain thermodynamic and kinetic models that describe the rate of reactions occurring along the process course at a molecular level. Intuitively, the same principle can also be applied to modeling dynamics of bioreactor cultures, which appear to be a result of interaction between the biocatalysts and their surrounding physical and chemical environments.15 Efficient use of mathematical models to explore the optimal solutions in these areas is, hence, the key to achieve successful process development.3 Yet, different from chemical systems, microbial systems are in a more complicated, hierarchical arrangement, as each individual cell in the bioreactor (i.e., the main system) can be further seen as a subsystem with metabolic and signaling networks, which in turn are composed of numerous molecular or enzyme-catalyzed reactions subject to reaction stoichiometry and gene expressions. The dynamics of the subsystems, expressed by their relaxation times, can span several orders of magnitude (e.g., from cell growth (103– 105 s) to enzymatic reactions under allosteric control (10
5–102 s)).16 It is true that a model could generate more valuable approximations and insights if it includes more details, such as reaction rate laws. Nonetheless, over-parameterized models can be difficult, if not impossible, to solve, validate or extrapolate with a reasonable degree of confidence. When constructing a bioreactor model, we should always consider the trade-offs between the model’s complexity and the ease of construction and validation, so the model can meet its intended purpose while still being computationally tractable. Given the inherent complexity of microbial cells, modeling of microbial behavior inside a bioreactor can take a top-down approach, by breaking down the original macroscopic system into segments. At the top-level, black-box models provide a macroscopic view of the dynamics in the bioreactor without specifying compositional and structural details of the segments. To obtain further insights into the functional significance and mechanisms of cellular metabolism, scales of modeling can be extended in the direction of increasing detail (e.g., metabolic networks, omics data), giving rise to grey-box and white-box models. On top of the models for microbial behaviors, scaling-up of bioreactor would require the assistance of spatially distributed models to elucidate the impact of mixing and fluid dynamics. In the following chapters, we do not intent to give an exhaustive review, but an overview on certain methods that has been or could be developed for bioreactor modeling at these four levels.

2.45.2

Black-Box Models

2.45.2.1

Macroscopic Kinetic Models

From a macroscopic point of view, fermentation in bioreactors can be related to chemical processes subject to physio-chemical environmental factors, such as temperature, pH, aeration and substrate concentrations. In the simplest scenario, S is the only limiting substrate and its concentration [S] is the only effective environmental factor. Conversion of S to a product P is catalyzed by bacterial pellets X, which are considered as black-boxes containing homogeneously distributed enzymes and chemical species. A further simplification of the process dynamics assumes the conversion of S to P is catalyzed by a single enzyme E, where E and S first form a substrate–enzyme complex C, which then disassociates into E and P: E þ S4 C ! E þ P

(1)

here k1, k-1 and k2 are the rate constants for the forward, reverse and catalytic steps, respectively. This theory led to the derivation of the well-known Michaelis–Menten kinetic expression of enzymatic reaction,17 shown in Eq. (2): d½P Vmax ½S ¼ dt Km þ ½S

(2)

here [P] is the concentration of the product; Vmax is the maximum specific rate of the enzymatic reaction; Km is the Michaelis constant, equivalent to the concentration of S when d[P]/dt equals to Vmax.

Bioreactor Models and Modeling Approaches

665

As a simple rate law, Michaelis–Menten kinetics is only valid under reactant stationary state, where the total enzyme concentration E0 is far smaller than (Km þ [S]).18 Despite this, Michaelis–Menten kinetic expression forms the basis of numerous black-box models. Notably, Monod developed an empirical kinetic expression in a similar manner to describe the specific microbial cell growth of bacterial cells, known as the Monod equation (Eq. 3)19: m ¼ mmax

½S Ks þ ½S

(3)

KS is the Monod constant (similar to Km in Eq. 2) and mmax is the maximum specific growth rate of a microorganism. It should be noted that biomass specific conversion rates like mmax are expressed per amount of biomass, so that they characterize the activity of cells.20 In order to construct the model, mmax and other specific rates, such as qp,max (i.e., the maximum specific rate of product formation), typically need to be derived experimentally. Based on the specific rate expressions, a simple black-box model describing the dynamics of a batch fermentation process can be formulated: Rate of cell growth

d½X ¼ m½X dt

Rate of product formation Rate of substrate consumption

d½P ¼ qp ½X dt



 d½S d½X 1 d½P 1 ¼

dt dt YXS dt YPS

(4) (5) (6)

here YXS and YPS are the yield factors of biomass and product with respect to substrate, respectively (given as mass of biomass or product/mass of substrate); and [X] is the biomass concentration. The model consists of three differential equations that define the rate of cell growth, the rate of product formation and the rate of substrate consumption. Solving the model leads to the prediction of the concentration profiles over a given period of time (Fig.1). The model, however, is highly limited in its predictive ability due to neglecting many other system variables and biological phenomena. Numerous variants were crafted with modified compartments and enhanced structure to study a wide spectrum of fermentation systems. In the following sections, some of the frequently used enzyme and growth kinetics are depicted.

2.45.3

Other Enzyme Kinetics

2.45.3.1

Double Michaelis–Menten Kinetics

When there are two substrates, S1 and S2, simultaneously used in the reaction, the original Michaelis–Menten equation (Eq. 2) can be modified as follows21: d½P Vmax ½S1 ½S2  ¼ dt ðKm1 þ ½S1 ÞðKm2 þ ½S2 Þ where Km1 and Km2 are the Michaelis constants with respect to S1 and S2

Figure 1

Prediction of concentration profiles by macroscopic kinetic models.

(7)

666 2.45.3.2

Bioreactor Models and Modeling Approaches Enzyme Inhibition Kinetics

Sometimes, an inhibitor I is also present in the media and can slow down the overall reaction. When I competes with substrate to bind on the same active site on the enzyme, I is a competitive inhibitor; when I binds to an allosteric site on the enzyme and changes the shape of its active site, thereby disabling its catalytic ability, I is an uncompetitive inhibitor; when I binds to another site on the enzyme and reduces its activity, I is a non-competitive inhibitor. The enzyme kinetics expressions under the three different inhibition mechanisms are shown in Eqs. (8–10).21 d½P V ½S  max  ¼ (8) Competitive dt Km 1 þ ½I þ S KI

Uncompetitive

d½P Vmax ½S   ¼ dt Km þ S 1 þ ½I

(9)

KI

Non-competitive

d½P Vmax ½S   ¼ dt ðKm þ SÞ 1 þ ½I

(10)

KI

Here KI is the Michaelis constant with respect to I. If I is actually a substrate or a product, then the above kinetics is also referred to as substrate and product inhibition.21

2.45.3.3

Hill Kinetics

Hill kinetics is specifically used to describe q, the fraction of the receptor protein saturated by a ligand, as a function of the ligand concentration [L].22 Hill kinetics takes the following form: q¼

½Ln ðKA Þn þ ½Ln

(11)

where KA is the ligand’s microscopic dissociation constant.

2.45.4

Other Microbial Kinetics

2.45.4.1

Contois Kinetics

Contois kinetics assume a linear correlation between the Monod constant KS and the initial substrate concentration [S]0 (i.e., KS ¼ b [S]0). The specific growth rate is thus given as23: ½S (12) m ¼ mmax b½S0 þ ½S In both batch and continuous fermentation, the concentration of biomass [X] is correlated to both [S]0 and [S]. Therefore, Eq. (12) can be rearranged to Eq. (13): ½S (13) m ¼ mmax 0 B½X þ ½S Both B and mmax0 are growth constants determined experimentally. Contois kinetics accounts for the inhibition effect of very high biomass concentration in bioreactors.

2.45.4.2

Powell Kinetics

Cells may need substrates for maintenance (e.g., metabolites turn-over) even though they do not grow. To account for this effect, Powell introduced a term of specific maintenance rate, m, into the original Monod equation24: ½S
m (14) m ¼ ðmmax þ mÞ Ks þ ½S

2.45.4.3

Moser Kinetics

Moser kinetics considers the case where growth dynamics is dominated by chemisorption of cells onto solid surfaces or structures, or when an enzyme has multiple binding sites for substrate S.25 m ¼ mmax

½Sn Ks þ ½Sn

(15)

Bioreactor Models and Modeling Approaches 2.45.4.4

667

Tessier Kinetics

Tessier kinetics mimics the behavior of Monod kinetics but is purely based on physical interpretations. When the substrate concentration is low, Tessier kinetics yields the same results as Monod kinetics.26   S (16) m ¼ mmax 1
e
k

2.45.4.5

Logistic Equation

Like Tessier kinetics, the logistic equation does not involve any biological interpretation. Its depiction of specific cell growth rate is only based on the assumption that there is an upper limit of biomass concentration in the culture broth, denoted as KX. As biomass concentration [X] increases and approaches KX, the specific growth rate drops.26
 ½X (17) m ¼ mmax 1
KX

2.45.4.6

Haldane–Andrew Model

Some substrates, such as nitrite, are toxic to cells and can inhibit cell growth at high concentrations. To account for the effect of toxicity on cell growth, Haldane–Andrew model includes an inhibition term [S]2/Ki, in the original Monod equation.27 ½S m ¼ mm (18) 2 KS þ ½S þ ½S Ki here Ki is the inhibition constant equal to the highest substrate concentration [S] when m ¼ 0.5mmax. In contrast, KS is the enzyme constant equal to the lowest substrate concentration [S], when m ¼ 0.5mmax.

2.45.4.7

Diauxic Growth

When cells are cultured in a medium with two carbon sources S1 and S2, they often show a strong preference for S1 (usually glucose) so that consumption of S2 will be inhibited until the depletion of S1. This phenomenon is known as diauxic growth.28 In this case, the specific growth rate m is given by21,29: ½S1  ½S2  þ mm2 m ¼ mm1 (19) ½S 2 K1 þ ½S1  K þ ½S  þ 2 2

2

K1

where mm1 and mm2 are the maximum specific growth rates of cells using S1 and S2 as the sole carbon source. Note that the second term on the RHS of Eq. (19) represents the specific cell growth sustained by inhibited utilization of the less favorable substrate, which takes a very similar form as the Haldane–Andrew equation (Eq. 18)

2.45.4.8

Luedeking–Piret Equation

The Luedeking–Piret equation is an empirical correlation linking the production rate of a product P with the microbial cell growth rate as well as with the cell density [X]. Initially, it was proposed by Luedeking and Piret while studying lactic acid fermentation30 Its application was later extended to studying numerous other fermentation processes, such as ethanol31,32 and citric acid33 production. It replaces Eq. (5) with the following general form: d½P d½X ¼a þ b½X (20) dt dt where a and b are growth-associated and non-growth-associated coefficients, respectively.

2.45.5

Summary

Macroscopic kinetic models describe the dynamics of the system by including kinetic expressions of essential extracellular species while neglecting intracellular activities. Model construction starts with a careful evaluation of the kinetics expressions to be included to ensure they can describe the real system with adequate coverage and accuracy, while still being simple so the resulting model remains solvable. This is followed by systematic estimation of process parameters, such as Ks and m, using appropriate experimental setups and parameter fitting strategies. Despite producing a rather simplified description of complex biological systems, kinetic models, if constructed properly, can closely predict the behavior of batch, fed-batch and continuous fermentation systems. More importantly, they can be used to extrapolate and make predictions on the growth and metabolite profiles in the bioreactors beyond current experimental conditions, and reveal the effect of a range of external variables (e.g., availability of carbon and nitrogen sources, level of dissolved oxygen) on process dynamics.34 The predictions serve as a guide for choosing optimal growth media,35 aeration schemes,36 and feeding strategies.37

668

Bioreactor Models and Modeling Approaches

Moreover, such kinetic models can serve as alternatives to hardware sensors which have a number of limitations (e.g., measurement delay), or for bioreactor control purposes capable of online monitoring, maintaining process variables at desired setpoints (e.g., pH, dissolved oxygen) and disturbance rejection.38–41

2.45.5.1

Classic Cybernetic Modeling

Models based on macroscopic kinetic expressions describe cell metabolism in an empirical manner. The gene regulatory mechanisms underlying certain notable biological phenomena, such as diauxic growth, are often accounted for by mere ‘inhibition terms’ (such as [S]2/KS) in the rate expressions, instead of a mechanistic description of gene actions or enzyme activities. Rather than following the same way of thinking, Ramkrishna and coworkers saw the necessity of linking the regulatory mechanisms to a survival goal of the microorganism. Hence, they proposed a more systematic modeling approach known as cybernetic modeling.42 The concept of cybernetic modeling is based on a reasonable hypothesis that eons of evolution have shaped the regulatory behavior of microorganisms and trained them into survival experts. Cells allocate limited resources to different activities and pathways in a way that optimizes some survival goals such as cell growth.43 Thus, in cybernetic modeling, the model completion becomes an optimization problem with a hypothesized objective function to account for the regulatory mechanisms. Let us consider a diauxic growth case with two carbon sources, S1 and S2, whose consumption relies on cell synthesizing key enzymes, E1 and E2. The entire reaction scheme is summarized below: d½Xi  ei ½Si ½X ¼ mi;max dt Ki þ ½Si 

(21)

d½X X ¼ rX;i vi dt

(22)

  mi;max þ b ½Si  dei dln½X ¼ ui
bei
ei dt dt Ki þ ½Si 

(23)

rX;i d½Si  ¼
vi dt YXS;i

(24)

rX;i ¼

Eq. (21) defines the rate of biomass synthesis through the consumption of substrate Si, using a Monod kinetic expression. i (i¼1,2) denotes the substrate; ei is the relative concentration of enzyme Ei that catalyzed the conversion of Si to biomass, i.e., ei ¼ [Ei]/[Ei]max. Eq. (22) thus defines the overall rate of biomass synthesis. Eq. (22) is the rate of change of ei contributed by three different components. The first term represents the specific rate of enzyme synthesis given by a Monod-type kinetic expression; the second term represents the first-order degradation rate of active enzyme, with b being the degradation constant; the third term represents the dilution of enzyme level due to cell growth. Eq. (24) is the rate expression of substrate consumption, with YXS,I defining the yield factor of biomass with respect to substrate Si. The kinetic parameters can be easily obtained from single substrate experiments. In the above equations, vi and ui are the so-called cybernetic variables, which serve as ‘resource allocators’ and controllers of enzyme production. According to the matching law,44 ui is the fractional allocation of the limited resources (i.e., the substrates) to the formation of enzyme Ei, that optimizes the total return (Eq. 25). rX;i ui ¼ P2 j¼1 rX;j

(25)

According to the proportional law, vi ensures that the overall return from the allocation of resources is maximized when the enzymes utilizing the best-suited substrate are given the maximum activity. For less-suited substrates, the corresponding enzymes are activated proportionally to their reaction rates44 (Eq. 26). vi ¼

r X;i  max r X;1 ; r X;2 

(26)

In general, the optimization problem is about allocating the limited resources between synthesizing E1 and E2 so that the overall return, namely the combined growth rate on S1 and S2, is optimized. Despite their empirical nature, cybernetic models provide a good framework for quantitative study of dynamic biological phenomena when the underlying mechanisms are unknown. Initially, the cybernetic principle was proposed to investigate batch or continuous fermentation involving utilization of mixed substrates.42,44–46 Further models have been developed with enhanced details to successfully describe the behavior of systems with extracellular enzymatic degradation47 and storage pathways.48 Later, the classic cybernetic principle based on the optimality hypothesis was extended to offer a higher degree of biological insights by providing a systematic representation of the influence of metabolic control.49 The classic cybernetic framework has since evolved into the realm of grey-box models suitable for analysis of metabolic networks, as will be discussed later.

Bioreactor Models and Modeling Approaches 2.45.5.2

669

Hybrid Artificial Neural Network Approach

The black-box models based on macroscopic kinetics often exhibit limited ability of extrapolation when the operating conditions are far from those under which the models are originally constructed. This limitation is mainly associated with the non-linearity of the system dynamics. To fully capture the dynamics of the fermentation process, such models will have to be constructed with more parameters and state variables. Constructing such models is not always a viable option as the computational and experimental complexity of parameter determination can be prohibitive. Fortunately, artificial neural networks (ANNs) can provide an appropriate solution to this problem.50 The ability of ANNs to learn from environments and perform massive multivariate data projection makes them very suitable for modeling nonlinear kinetics. As a result, a hybrid approach based on black-box models and artificial neural networks is developed for modeling fermentation processes. In the hybrid method, a black-box kinetic model is first constructed to interpret the change of the state variables in the fermentation process, such as d[S]/dt and d[X]/dt. During the iterative training phase, experimental measurements of the state variables are used as inputs for the ANN for parameter estimation. ANN models have three layers, namely input, hidden and output layer with interconnected artificial neurons. Each input state variable has a corresponding neuron in the input layer. The input neurons are connected to a number of hidden neurons via sigmoid functions to be able to learn the nonlinear features of process dynamics. The output layer then applies certain functions to the output of the hidden layer to compute the final outputs, namely the parameter functions.51 The black-box models subsequently use the parameters estimated by ANN to predict a new set of state variables. The whole model compares the predicted results with input data using an error term (i.e., the normalized squared error between the estimated and measured state variables). The weights in the hidden and output layers are then scaled according to the error term so that in the future runs the entire model will generate predictions closer to the experimental results. The main advantage of ANNs is that they are capable of capturing all possible nonlinear and complex interconnections between dependent and independent variables that govern the dynamics of the biological process. On the other hand, ANNs produce nonidentifiable models as they approximate the kinetic relations in a highly implicit way. That is, one cannot extract any biological insights from the machine-learning process. In a way, ANN-based models are ‘darker than black’, as even the Michaelis–Menten kinetics widely used in macroscopic kinetic models are derived from the mechanism of enzyme catalysis. Nevertheless, recent research has confirmed the competence of hybrid-ANN models in fermentation simulation. Given sufficient training data, hybrid-ANN can predict the process trajectory under varying operating conditions with even greater accuracy than typical blackbox models.51,52

2.45.6

Grey-Box Models

Model-based process design has facilitated the development of cost-effective, high-yielding industrial fermentation processes employing optimal operating conditions. Meanwhile, an important branch of industrial biotechnology is devoted to the development of enhanced microbial strains to achieve overproduction of bioproducts. Conventionally, strain improvement takes a trialand-error approach and relies heavily on random screening of advantageous mutants.53 This can be facilitated by more targeted, rational and controllable approaches of genetic engineering and synthetic biology. However, the previously described black-box models fall short of providing the fast-advancing fields with a constant input of molecular and metabolic insights.15 Under this context, extended details of intracellular structure and activities are taken into consideration to bring transparency to the ‘blackboxes’. This has led to the generation of metabolic network models that retain the black-box model’s ability to describe macroscopic phenomena in bioreactors while exposing potentially important biological insights.54,55 Whether we choose to engineer an existing strain for desired functions or to develop a whole new genetic system with unique capabilities, deep understanding of metabolism and the relevant metabolic pathways is needed in the first place. Metabolism is an umbrella term for all biochemical reactions in living cells. The reactions, catalyzed by their respective enzymes in sequence, form groups of coordinated metabolic pathways to convert raw materials to metabolites essential for cell growth, proliferation and maintenance. The rate at which a metabolic pathway carries out metabolite interconversion is known as metabolic flux. The metabolic flux is a fundamental determinant of the metabolic state of a cell and its physiology, and in addition, a pivotal parameter for the quantification of the control of metabolic fluxes.56 Like a material balance equation in the context of the black-box model, metabolic fluxes define the movement of material through the cell’s metabolic network comprised of multiple metabolic pathways. Fluxes hence provide a quantitative measure of the contribution of different pathways as well as pathway enzymes to the overall metabolic process under different sets of genetic or environmental conditions. Therefore, quantification of in vivo metabolic fluxes is crucial to understanding cellular metabolism and the practical application of metabolic engineering.57 Metabolic modeling lies in the core of the approaches for quantification of metabolic fluxes. Thanks to the advancement in highthroughput DNA sequencing technology, a growing number of complete metabolic maps with annotated genome sequences, such as that of Saccharomyces cerevisiae and Escherichia coli, have been retrieved and made public via databases.58 The main metabolic routes for most industrially important strains have also been determined, even though in many cases the complete pathway topology remains unknown.59 The available information allows for reconstruction of both full genome-scale metabolic models and reduced metabolic models including only the reactions required to perform a given bioconversion. Metabolic models can predict the phenotypic outcome of particular genetic manipulations (e.g., gene deletion or addition) in the form of metabolic flux distributions and even concentration profiles of extracellular metabolites, thus providing a rational basis for metabolic engineering efforts.

670

Bioreactor Models and Modeling Approaches

From the expression of genes within cells to the operation of metabolic pathways, cellular metabolism encompasses many layers of subsystems. Accordingly, metabolism can be modeled with different degrees of biological details to gain structural and dynamic insights into the characteristics of metabolic networks. Depending on the nature of the information, metabolic network models can be further categorized into grey-box models and white-box models. This section focuses on the grey-box modeling approaches, which are primarily based on the knowledge of network stoichiometry.

2.45.6.1

Constraint-Based Stoichiometric Modeling

Constraint-based models (CBMs) have constituted a notable proportion of foregoing metabolic modeling effort in studying biological systems.60 The underlying principle of CBM is that, cellular metabolism is subject to various constraints, including substrate and enzyme availability, mass balance, thermodynamic feasibility, metabolic regulation and reaction kinetics.61 All feasible solutions of flux distributions that can explain the metabolic behaviors of cells exist in a multi-dimensional solution space bounded by these constraints. At any time instance, there exists a single solution of flux distributions which is the closest reflection of the metabolic situation in the cell, and this solution is theoretically obtainable when all constraints are employed. Though in real life, this solution is never obtainable due to lack of available constraints, we can still create a reduced, delimited solution space with the available ones to produce meaningful insights (Fig.2).61 The most basic steps of the model building process are selection of proper biochemical reactions to include in a metabolic network, identification of the reactions’ stoichiometry, and then rational representation of the metabolic network in its mathematical form, namely in terms of a stoichiometric matrix. For a metabolic network S with m metabolites and n reactions, its stoichiometric matrix is as shown in Fig.3: Where Sm,n refers to the stoichiometric coefficient of metabolite n in the reaction m. Then, molar balances regarding intracellular metabolite, intracellular fluxes and metabolic output can be established as shown in Eq. (27): dC ¼ S$v
m$C dt

(27)

where C (C ¼ C1, C2, . Cn) is the vector of intracellular metabolite concentrations, S is the stoichiometric matrix, v (v ¼ v1, v2, . vm) is the vector of metabolic fluxes, m is the specific growth rate of cells. Therefore, dC/dt is the vector of the rate of change of concentration with respect to time for all intracellular metabolites, and m$C is the vector of dilution due to biomass growth at a specific growth rate m.62 In the case of kinetic modeling, solving the system of the first order ordinary differential equations (Eq. 27) would yield the time-dependent vector of metabolite concentrations, C(t).

Figure 2

In constraint-based modeling, the imposition of various biological constraints defines the associated solution space.

Figure 3 A stoichiometric matrix with m metabolites and n reactions, where Sm,n refers to the stoichiometric coefficient of metabolite n in the reaction m.

Bioreactor Models and Modeling Approaches

671

In the context of grey-box modeling, however, intracellular kinetics is put on the back burner. Study of the metabolic network is mostly based on stoichiometric modeling, which is performed under steady-state conditions where the time-dependent behavior of the system is neglected. The concentration vector of metabolite pool is constant (i.e., no net accumulation or consumption of metabolites). Hence, the time derivative (dC/dt) equals to zero (Eq. 28). Steady-state data can be obtained from chemostat experiments. 0 ¼ S$v
m$C

(28)

Furthermore, cellular processes can occur on very different timescales, whose orders of magnitude are quantified by their respective characteristic times (i.e., the reciprocals of the rate constants of the processes which are approximated as being first-order).63 When studying a system whose characteristic time is more than three times larger than that of a certain process, the process can be considered to be at pseudo-steady state.16 Enzymatic reactions are relatively fast processes, with typical characteristic times ranging from milliseconds to seconds. Thanks to the fast kinetics, the intracellular metabolite pool can be adjusted to new levels under new environmental conditions rapidly, as has been confirmed by perturbation experiments in S. cerevisiae and E. coli.64–66 In contrast, typical fermentation processes operate on significantly larger timescales. For example, the characteristic time of cell growth can be in the order of hours to days.57 Therefore, the impact of metabolite pool adjustment can be neglected when the focus is placed on the dynamics of the entire fermentation process. This leads to the important assumption of pseudo-steady state (PSSA), which ignores the kinetics of enzymatic reactions whose characteristic times are not within the time frame of interest, thereby assuming that cells simply carry out the reactions at many different steady states (dC/dt ¼ 0) in environments which are externally dynamic (note that for a homogeneous system operating at constant volume, PSSA is also referred to as quasisteady state assumption).67 PSSA significantly simplifies the metabolic models, with the important implication that batch or fed-batch experimental data can also be used as the inputs for metabolic models. Since the level of most intracellular metabolites is very low, it is sensible to assume the dilution term m$C to be negligible compared to the fluxes affecting the metabolite.68 Therefore, Eq. (28) can be further simplified into the following linear system (Eq. (29)): S$v ¼ 0

(29)

Eq. (29) representing the steady-state mass (mole) balance constraint within the system forms the basis of constraint-based stoichiometric modeling. The metabolic network, which is underdetermined with m-n degrees of freedom, confines solutions in the subspace of a hyperplane. When we also limit the capacities of metabolic reactions by specifying respective lower and upper bounds for fluxes, the solution space can be reduced into a bounded convex cone. The solution space can be further narrowed down with additional constraints, such as thermodynamic69–71 and regulatory constraints72,73 to generate more explicit insights. Stoichiometric models serve different purposes when different sets of mathematical constraints are employed. Metabolic flux analysis (MFA) employs flux measurements from chemostat and 13C isotope tracer experiments to determine the actual flux distribution under physiological conditions of interest.74 Elementary mode analysis (EM)75 and extreme pathways (EP)76 examine the structure and topology of reconstructed metabolic networks via decomposition of flux distributions into multiple sets of elementary modes (EMs). Flux balance analysis (FBA) uses convex analysis and optimization criteria to predict a specific phenotype of the metabolism after genetic or environmental perturbations.77 Although stoichiometric models can be of various different forms, we concentrate on FBA which can be suitably adapted to depict the time evolution of macroscopic culture environments.

2.45.6.2

Flux Balance Analysis

As already introduced in the black-box model section, optimization (mostly using heuristic approaches) has also been applied in cybernetic models for fermentation processes. The rationale for this assumption is that, cellular metabolism is in a constant state of optimization as its has been shaped by millennia of evolution to respond to varying environmental conditions to meet the optimal physiological needs of the cells. Meanwhile, cellular behaviours are subject to multiple constraints, such as mass balance, thermodynamic feasibility, metabolic regulation and reaction kinetics. Thus, metabolism can be effectively represented as a constraintbased optimization problem, which forms the fundamental concept of FBA. By solving the optimization problem, FBA is able to compute intracellular flux distributions which represent steady-state cellular phenotypes expressed under specific optimality criteria, such as optimal cell growth or ATP yield. The optimization framework also allows for identifying the parts of metabolic pathways whose modification may be beneficial to achieving desirable phenotypic properties.77,78 An objective function provides information as to which biochemical reactions(s) are targeted for flux optimization, as well as their relative contribution to a specific phenotype. Mathematically, the objective function, denoted as O, is a 13n sparse vector of weights, where n is the total number of metabolic fluxes. If, for example, we are to calculate the flux distribution when only v1 and v2, namely the flux of the 1st and the 2nd reaction (as specified in the stoichiometric matrix), are maximised: O ¼ ½W1 W 2 0 0 0...0

(30)

As we can see from Eq. (30), the values of the element O(1,1) and O(1,2) are set equal to weighting factors W1 and W2, respectively, which assign the relative importance of v1 and v2 when running the optimization. Since the other fluxes (v3 to vn) are not

672

Bioreactor Models and Modeling Approaches

considered for optimization, the remaining elements in O should be set to zero. Therefore, the objective of this example optimization problem is: maxðW 1 v1 þ W 2 v2 Þ

(31)

The objective function used in FBA must be appropriately formulated in order to predict flux distributions that are biologically meaningful. Objectives such as maximizing biomass growth or maximizing ATP production per unit flux are commonly used as they, to some extent, demonstrate the intrinsic survival requirement of bacteria.79–81 However, the universality of these optimality criteria is questionable as microorganisms can have more sophisticated physiological needs that do not explicitly depend on biomass synthesis. C. necator fermentation in glycerol is one such example, where the focus of the metabolism gradually shifts from cell growth to energy reservation as the availability of nitrogen source decreases. In this scenario, optimizing cell growth or energy production is almost certainly not an applicable cellular objective. Many biological hypotheses have been made and tested to rationalize the use of other objective functions. For instance, the assumption that cells carry out metabolic activities with minimal efforts suggests the use of ‘flux minimisation’ as the objective82; while the assumption that cells conserve energy or utilize energy as efficient as possible prompts the use of ‘minimization of redox potential’ as the objective.80 Moreover, multiple cellular objectives may be optimized simultaneously, such that FBA modeling generates a range of optimal solutions on a Pareto surface.60,83 Some studies also assume that cells behave in a slightly suboptimal manner in order to maintain a certain amount of leeway for metabolic adjustments,60 or in the case of a genetically engineered strain, to try approximating its original behavior before being genetically modified.84

2.45.7

Dynamic Flux Balance Analysis

As discussed earlier, PSSA allows the use of metabolic fluxes obtained from non-steady state experiments (i.e., batch and fed-batch). Thus, the original computation framework of FBA can be extended to cover the dynamic flux distributions over the entire duration of fermentation, bypassing the intricate intracellular kinetics (Fig.4). Depending on the sources of extracellular fluxes, dynamic flux balance analysis (DFBA) can be further categorised into kineticdriven and data-driven approaches. Data-driven DFBA directly employs fluxes derived from time-resolved concentration profiles.85 This provides an expeditious approach to monitor the transient physiological response in biological systems where metabolism is subjected to various types of environmental changes.66,86 Nevertheless, this approach is highly sensitive to extracellular inputs which could contain gross measurement errors. Hence, Leighty and Antoniewicz have developed an integrated approach that allows direct flux estimation from metabolite concentration profiles without data pre-processing.87 Such approach can enhance the quality of fluxes, consequently allowing for closer monitoring of cell physiology. Kinetic-driven DFBA is complemented by the established black-box models. It receives flux inputs generated by phenomenological expressions, such as Michaelis–Menten equations for substrate uptake, to account for missing regulatory mechanisms.88 In this regards, the kinetic approach provides a valuable predictive framework for simulating dynamics of fermentation processes, evaluating alternative operating conditions and identifying potentially useful genetic manipulations for improved process

Figure 4

Time series flux distributions can be obtained as a result of DFBA.

Bioreactor Models and Modeling Approaches

673

performance.85 Varma and Palsson89 first constructed a predictive model based on DFBA, which was constrained by pre-determined maximum rates and aimed to optimize cell growth. The entire experiment duration was first divided into many small time steps (Dt). With the initial concentrations of biomass and by-products specified, linear optimization was performed consecutively from the beginning to predict concentrations at the next time step. In a dynamic study of diauxic growth in E. coli, Mahadevan et al.90 further elucidated this DFBA formalism and referred to it as the static optimization approach (DFBA-SOA). In the same contribution, they also introduced the dynamic optimization approach (DFBA-DOA), which directly optimizes the biomass concentration over the entire time scale instead of biomass flux at each time step. Although optimization at concentration level doesn’t yield many mechanistic insights, the ability to reproduce the macroscopic observations under various conditions makes DFBA capable of replacing kinetic metabolic models for simulating and optimizing fermentation systems even beyond laboratory scales.91,92 Application of DFBA in larger genome-scale networks (with more than 200 fluxes) is facilitated by techniques such as compartmentalisation and charge balancing,93 simultaneous solution of LP-embedded extracellular mass balance equations,94 and reformulation of DFBA into hybrid differential algebraic equations.95 Moreover, DFBA can be made more competent by incorporating additional information, such as isotopomer data,96 enzyme production costs,97 metabolomic measurements,98 intracellular regulatory and signalling networks,92 transport constraints99 and macromolecular crowding.100,101

2.45.7.1

Cybernetic Metabolic Modeling

The structure of classic cybernetic framework and its application in black-box models have been discussed earlier. The classic approach assumes that the utilization of a limiting substrate for biomass synthesis is bottlenecked by a single enzyme, whose expression and activity are regulated according to the availability of other substrates and cellular need for an optimal nutritional state. Through lumping complex biochemical reaction networks into a few simple enzymatic reactions, cybernetic models have proved their ability to simulate mix-substrate fermentation systems. Therefore, the classical cybernetic approaches are also referred to as lumped cybernetic models (LCMs).102 However, such a simplification inevitably ignores the effect of the topology of metabolic networks and the potential importance of multiple key enzymes subject to complex regulatory controls. This has limited the application of the cybernetic framework for modeling processes of more sophisticated nature (e.g., growth in nitrogen-limited or carbonlimited conditions).103 Therefore, the cybernetic framework was extended to consider the distribution and control of material flow within metabolic networks. Given the complexity of the metabolic network, the extension of the cybernetic framework takes a modular approach, by which a global metabolic network is dissected into a number of basic pathways, namely elementary modes (EMs).104–106 Each EM is treated as a discrete cybernetic unit, having its own resource allocation problem subject to the constraints on the level of resources. The local objective functions specify how nutrients should be directed toward the synthesis of different competing enzymes in the elementary pathways, such that the global objective of maximizing the nutritional outcome is fulfilled.107 Although the regulatory influences within each EM are considered individually, the modes are not isolated pieces that form the overall metabolic network without interacting with each other. To facilitate the assembly of individual regulatory influences into a complete global regulatory machinery, metabolic regulation is decomposed into three hierarchical components: (1) elementary components exclusively related to the functionality of individual modes; (2) local components describing the interplay between elementary components; (3) global components describing higher nutritional and topological regulatory influences, such as catabolite repression and feedback inhibition. Subsequently, cybernetic variables are defined to represent metabolic control at the three different levels. For the jth enzyme in the network, the complete cybernetic variable governing the resource allocation for its synthesis is given as: g

uj ¼ uj ulj uj is the product of global cybernetic variable regulatory contributions, given as:

g uj

and local cybernetic variable

g

uj ¼

ulj ¼

nðjÞ Y q

Uj;q

mðjÞ Y k

ukj

(32) g ulj . uj

and ulj are in turn the products of the individual

(33)

(34)

where Uj;q is the global cybernetic variable reflecting the qth global signal influencing the expression of the jth enzyme; ukj is the elementary cybernetic variable of the kth control signal influencing the expression of the jth enzyme within the EM. The set of cybernetic variables regulating the activity of the jth enzyme at all three levels can be formulated in a similar manner: g

vj ¼ vj vlj

(35)

674

Bioreactor Models and Modeling Approaches

g

vj ¼

vlj ¼

xðjÞ Y q

V j;q

yðjÞ Y k

vkj

(36)

(37)

Like in the classical framework, the functional forms for the elementary cybernetic variables ukj and vkj are derived from the matching and proportional law, respectively (See Section 2.45.5.1 Classic cybernetic modeling).104 If a key enzyme E is shared between p linear EMs, then the elementary cybernetic variable governing the allocation of critical resources for synthesizing E from the pth pool is given by: rE p uE ¼ P (38) kðpÞ r k where

P

r k is the sum of the specific rates of all reactions in the pth pathway, used as a measure of the return on resource

kðpÞ

investment. The specific rates can be calculated from Monod-type kinetic expressions. Next, the elementary cybernetic variable governing the activity of E is given by: r p nE o (39) vE ¼ max r kðpÞ p

vE ensures that E possesses the maximum activity if it catalyzes the pathway reactions that maximize the end product among the other competing reactions in the pth EM. The above resource allocation problems are solved at elementary level, with the influence of global control acting as additional constraints and filter on the local and elementary metabolic activity. A global objective of the metabolic network (e.g., maximizing biomass synthesis) is then the superimposed goal of the individual allocation problems. Finally, a complete cybernetic model is established by linking the metabolic network model with a macroscopic growth model. This is achieved by lumping the dynamics of the key enzymes catalyzing the extracellular fluxes into a set of control variables, which are functions of directly measurable extracellular rates (e.g., biomass synthesis and substrate uptake rates).49 The resulting cybernetic model can provide a macroscopic description of the fermentation process, while promising deeper insights into metabolic regulation even when the actual control mechanisms remain elusive. Young et al. has used the cybernetic metabolic model to simulate the dynamics of reduced metabolic networks of E. coli strains comprising 12 biochemical reactions.108 The model proved to be useful in guiding metabolic engineering efforts, as metabolic responses to gene knockouts and additions were mostly predicted in a quantitative manner. The extension of the cybernetic framework to larger metabolic models has, however, encountered computational challenges stemming from non-linear intracellular kinetics. To reduce the demanding computational cost, Kim et al. formulated hybrid cybernetic models (HCMs) synthesizing the cybernetic models with the aforementioned FBA framework. Cybernetic models use global cybernetic variables to estimate extracellular fluxes (i.e., uptake fluxes), which are subsequently coupled with FBA to predict intracellular fluxes and fermentation by-products based on PSSA.109 By avoiding the need for accurate kinetics in intracellular flux estimation, the HCMs now have a complexity comparable to that of the classical cybernetic models and can be efficiently used as fermentation simulators. Another obstacle to the application of cybernetic modeling at larger scales comes from the exponential increase of the number of EMs required for model construction. In this aspect, Song and Ramkrishna proposed a method called yield analysis (YA) to extract a minimal set of EMs in yield space that are essential for representing metabolic behavior. This a priori reduction procedure facilitates the application of HCM for the growth of a recombinant Saccharomyces strain on mixed substrates by reducing 95% of the total EMs.110 In the case of very large networks where available data are too limited to support any form of detailed dynamic analysis, lumped HCM (L-HCM) can come into play. L-HCM simplifies the network by lumping EMs with mutual substrates or vital products into EM families. Each of the EM families can be effectively considered as a single EM, so that the original L-HCM is treated as an HCM. Although L-HCM is less suited for accurate dynamic simulation of metabolism than HCM due to skipping details of basic EMs, it allows for first-hand quantitative assessment of metabolic functions with considerably fewer data.102

2.45.8

White-Box Models

Construction of dynamic grey-box models is an important step toward simulation and quantitative prediction of cellular processes. The bulk of this modeling effort uses mainly extracellular dynamics to calculate the unknown fluxes responsible for macroscopic observations. Meanwhile, such models take a reductionist approach and assume the subsystems (i.e., intracellular dynamics) to be static, thereby avoiding prior knowledge of intracellular rate laws or metabolite concentrations to describe intracellular kinetics. Cybernetic models are able to account for internal kinetics to some extent, but for large-scale network applications they still rely heavily on simplifications offered by PSSA and lumped EMs. While grey-box models can lead to significant biological insights, they are still not in the position to provide full-featured, dynamic simulation of the entire cell. For instance, they do not generate complete mechanistic understanding of metabolic response to fine-tuning of genetic parameters and environmental conditions. To

Bioreactor Models and Modeling Approaches

675

this end, system biology ultimately seeks the development of genome-scale kinetic models capable of complete description and simulation of cellular systems with quantitative accuracy.111 Here, such models are generally referred to as white-box models, in contrast to their less detailed grey-box counterparts. Essentially, white-box models can be constructed based on kinetic formalisms in a similar manner as black-box models: rate expressions of particular mathematical formalisms are incorporated into target metabolic networks to describe the kinetics of all enzyme-catalyzed reactions subject to substrate-level, allosteric, transcriptional and post-transcriptional regulation.112 Mathematically, white-box models are formulated as systems of ODEs, which are further integrated with cell submodels describing multiple cellular functions. Rigorous white-box modeling is advantageous because it leads to quantitative prediction of dynamic metabolite states of cells and comprehensive identification of optimization potentials for microbial producers. Admittedly, white-box models currently are not as widely used in bioreactor predictive modeling as grey- or black-box models, which are simpler and more specific. Nevertheless, the fundamental framework they provide for examining the joint-effect of metabolic and process engineering is a valuable asset for industrial process development.113

2.45.8.1

Current Challenges and Progress

Indeed, development of whole-cell white-box models is a highly challenging task. First, such models would require a formidable number of heterogeneous parameters and rate laws, which are still undetermined for most microorganisms, barring a few exceptions for model microorganisms.114 Furthermore, as with all kinetic models, training and test datasets are required to identify best-fit parameters and to validate model predictions, respectively. Despite laudable progress in high-throughput omics technologies and development of simulation platforms,115,116 the model’s data-hungry nature remains such a burden that only very few whole-cell models have been constructed.117,118 The metabolic network that can be modeled dynamically is usually limited to a smaller system composed of a few well-characterized pathways, such as glycolysis.119,120 In addition, the non-linearity of explicit kinetic relationships makes it exceedingly difficult to obtain analytical solutions for the models. These hurdles may be partly overcome by tractable approximation of enzyme kinetics with standard kinetic formalisms, such as mass-action kinetics,121 lin-log kinetics,117 or ‘convenience kinetics’,122 and the unknown parameters need to be estimated via integration of kinetic, metabolic, and proteomic data.123 Nonetheless, coarse-grained approximation for kinetics are only valid when the fluxes are near the chosen reference state (i.e., locally valid). In face of the uncertainties in the approximated rate-laws and structural uncertainties of regulatory mechanisms, the ensemble modeling approach (EM) emerged as a promising alternative.124,125 Instead of creating a single kinetic model, the EM approach creates a large ensemble of models, containing all possible kinetic models obtained through sampling of kinetic parameters. The ensemble is analyzed statistically to reject any invalid models whose predictions are inconsistent with flux measurements, leaving one or several accurate sets of kinetic parameters in the end.126 The optimization-based EM approach has been used to construct medium- and large-scale kinetic models for E. coli; the flux prediction reaches a relatively good agreement with experimental measurements.127,128 Another challenge is associated with the validity of the rate laws and enzyme kinetic parameters, which are often determined through in vitro experiments.126 In vitro experimental measurements performed on cell extracts are prone to errors/inaccuracies from neglecting intracellular metabolites–protein interactions, molecular crowding, as well as organism-specific regulatory functions. Consequently, in vitro derived parameters do not necessarily reflect the actual in vivo scenario, leading to discrepancies between model predictions and experimental data. Hence, at the final stage of model construction, the parameters need to be calibrated with in vivo data. In vivo data can be inferred from perturbation experiments (i.e., stimulus–response experiments),129 or derived from metabolite concentrations and flux values measured in vivo from chemostat cultures.130 Recently, computational approaches combining metabolic modeling and molecular crowding have been developed.100,101 Finally, detailed white-box models are computationally expensive as they are stiff and high-dimensional. Sometimes it is useful to consider model reduction when microbial metabolism is modeled under a controlled environment, such as in bioreactors where only a few parameters are altered. There is a wealth of model reduction technologies applicable to metabolic networks. In many cases not all portions of metabolism are important and the scope of the model can be refocused to more crucial and wellcharacterized subsets of networks, such as the central carbon metabolism.130 For instance, Snowden et al. have proposed a combined model reduction strategy for E. coli modes based on proper lumping and empirical balanced truncation.131 Mannan et al. integrated kinetic models of the central carbon metabolism with steady-state data to account for the contribution of non-central pathways and cell growth.132 Model reduction improves the kinetic models’ computational efficiency while maintaining good input-output prediction accuracy.133 A wide variety of other useful model reduction technologies ranging from data-driven approaches, such as Proper Orthogonal Decomposition (e.g. Ref. 134,) to input-output equation-free techniques135 that have already found extended applicability to modeling optimization and control of chemical processes136 can prove beneficial towards making white-box biochemical models computationally tractable.

2.45.9

Reactor Models

Black-, grey- and white-box kinetic models have been extensively used to explore optimal genetic engineering strategies and operating conditions for bioreactors to achieve efficient biological production of chemicals. Model-based design and optimization always start with processes at lab-scales to assess their productivity and feasibility and to establish proof-of-concept before making

676

Bioreactor Models and Modeling Approaches

Figure 5 Homogeneous mixing (left) is common in small-scale bioreactor, whereas in large-scale bioreactor (right) the mixing is often heterogeneous.

any further investment. This is followed by scaling-up of fermentation processes to bring the biotechnology innovations to commercialisation. Nevertheless, switching fermentation process developed at lab-scale to manufacturing scale is not a simple matter of increasing bioreactor’s working volume. Due to insufficient mixing, chemical and physical heterogeneity regularly occur in large-scale bioreactors (>50 L). More specifically, it has been established that spatial and temporal gradients of substrates, oxygen, temperature and pH exist in large-volume cultures (Fig.5). The gradients, in turn, lead to phenotypic homogeneity in bioreactors. Furthermore, intensive agitation in bioreactors results in excessive turbulence, which may induce shear damage to cells and reduce their viability.137 These factors can compromise the optimization settings evaluated based on small-scale reactors and render the cultivation conditions suboptimal. This would not only harm the efficiency of bioreactors, but also bring pitfalls regarding process monitoring and product quality control. Building a fermentation process at manufacturing scale demands a substantial amount of capital investment, and the entire procedure can be complex and risky. To minimize setbacks that might occur, it is important to perform detailed process characterization to understand the scale effects and how they may impact on the adaptation of the optimization strategies established in small-scale reactors to large-scale ones. To this end, a scale-down model, which is representative of the proposed industry process in terms of configuration and performance, can be built. Such a model is generally composed of two parts: a cell model and a reactor model. The cell model is essentially a black-box or grey-box model describing the dynamics of cellular process under given bioreactor conditions. The reactor model determines the growth conditions for cells based on fluid dynamic simulation, which systematically considers the impact of fluid rheology, mass transfer and culture heterogeneity. The integrated scale-down models can be used to evaluate operating strategies and optimize key process parameters, and eventually to forecast the outcome of scale-up.138 The most common practice in reactor scaling up is maintaining geometric similarity to ensure similar mixing characteristics. Conventionally, this is done by holding one of several relevant system parameters constant, so the agitation speed and aeration rate can be estimated accordingly. The parameters, such as agitation power input power per unit volume, impeller tip speed, superficial gas velocity and gas(oxygen) mass transfer coefficients, are determined based on empirical correlations. However, the conventional approaches are plagued by lack of description of the hydrodynamic behavior across multi-reactor scales and its interaction with microbial activities. This can be complemented by computational fluid dynamics (CFD). Essentially, CFD solves the NavierStokes equations which describe the conservation of mass, momentum, energy and species in the fermentation broth. CFD can be used to simulate flow in the bioreactor and elucidate how turbulence occurs and alters the local environments of cells and how it can eventually impact cellular performance.139 One popular approach to study the environmental fluctuations experienced by cells is the Reynold-average Navier–Stokes (RANS) approach. It generates approximate solutions to the Navier–Stokes equations by considering the Reynolds decomposition procedure.140 Euler–Lagrange large-eddy simulation (LES) emerged as a more accurate alternative that tracks the motion and environments of individual cells by calculating their time-dependent positions in the main flow fields. The small scales in the flow fields, meanwhile, can be modeled with the eddy viscosity model or other turbulent models.140 Euler–Euler is another CFD-based approach used to model bioreactors, especially ones with larger sizes and higher volume fractions of solids. Euler–Euler approach treats the solid (cells) and liquid phases as coexisting interpenetrating continua. It is a less computationally expensive method than LES as the continuity and momentum equations only need to be solved separately for the phases involved.141 To further reduce the computational cost of reactor modeling, the so-called compartment models (CM) (also referred to as network-of-zones models) are sometimes used.142 As the name suggests, the models are based on the division of the large-scale

Bioreactor Models and Modeling Approaches

677

reactors into interconnected compartments. For example, a fed-batch reactor may be separated into an addition zone where major concentration gradients exist, represented by a plug flow reactor, and a bulk region represented by a stirred tank reactor. Flow properties within each compartment are considered homogeneous, and inter-compartment flux is represented by mean and turbulent flow rates.142,143 On the other hand, the turbulent flow rates derived from global quantities such as flow numbers are unable to capture the flow complexity within the bioreactor faithfully. Hence, CM frequently yields inaccurate predictions. To overcome this drawback, a combined CFD/CM approach has been proposed. In this approach, the bioreactor is compartmentalized based on CFD simulation results of turbulent flows, while the fluxes between different compartments are specifically derived from CFD velocity fields.144 Meanwhile, the distribution of cell particles between compartments is simulated using a stochastic model. This hybrid method incurs much less computational expense than the ordinary CFD approach while maintaining accuracy.

2.45.10 Conclusions The bioreactor is the heart of any bioprocess, as it is therein that bioconversion of raw materials to biological products occurs. On the road to designing optimal bioreactor systems for industrial biotechnology, there is a multitude of challenges on four different aspects: economic, scale-up, operational and microbial strain. Due to the highly multi-hierarchical nature of bioreactor systems, these challenges can extend from macroscopic to microscopic levels. For instance, the issue of scaling-up of bioreactors is essentially associated with the heterogeneous flow fields which occur at the system (i.e., bioreactor) level; whereas designing optimal microbial producers by means of genetic engineering requires in-depth knowledge of the nature of the subsystems (i.e., cellular metabolism). Fortunately, an immense body of information is available nowadays to construct bioreactor models to guide the development of strategies to overcome these challenges. Depending on the intended purposes and the level of biological details involved, bioreactor models fall into four categories. Black-box models are the first type of models proposed to simulate the biological dynamics in bioreactors. The word ‘black-box’ comes from the fact that they rely on phenomenological laws to explain empirical observations, meanwhile treating cells as ‘blackboxes’. In addition, based on black-box kinetics, grey-box models further incorporate the knowledge about the topological properties of metabolic networks, and sometimes limited information of intracellular kinetics to offer the first glimpse into the cellular metabolism. Grey-box bioreactor models allow for simulation of the fermentation process while being able to consider the impact of genetic and environmental perturbations on the overall performance. As further transparency is introduced by including explicit kinetic laws and thorough understanding of cell functions into genome-scale metabolic models, the so-called white-box models are generated. White-box models give a more complete representation of complex cellular factories than grey-box models, thus offering quantitative and more detailed predictions to facilitate the design of bioreactors. However, due to the existing challenges associated with modeling complexity and data shortage, white-box models have not reached the degree of maturity necessary for simulation applications in industrial processes. In general, dynamic models were classified into three classes in a hierarchical manner for ease of explanation. It should, however, be noted that, there are no strict class boundaries between the three types of models, as there exists significant flexibility for less ‘transparent’, higher-level models to incorporate biological information in a piecewise manner to upgrade the functionality of the models and obtain finer insights. Finally, the dynamic models can be combined with reactor models describing the mixing and fluid heterogeneity in the system to predict the effect of process scaling-up. With the extremely rapid development of biotechnologies and constant progress made in the mathematical, algorithmic and computational techniques, the opportunities of bioreactor modeling seem limitless.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

BECOTEPS. The European Bioeconomy in 2030: Delivering Sustainable Growth by Addressing the Grand Societal Challenges, 2011. EC. Innovating for Sustainable Growth: A Bioeconomy for Europe. Brussels, 2012. Volesky, B.; Votruba, J. Modeling and Optimization of Fermentation Processes, 1st ed.; Elsevier Science, 1992. National Research Council (US). Putting Biotechnology to Work: Bioprocess Engineering, National Academies Press, 1992. Du, J.; Shao, Z.; Zhao, H. Engineering Microbial Factories for Synthesis of Value-Added Products. J. Ind. Microbiol. Biotechnol. 2012, 40, 1301–1315. Scarlat, N.; Dallemand, J. F.; Monforti-Ferrario, F.; Nita, V. The Role of Biomass and Bioenergy in a Future Bioeconomy: Policies and Facts. Environ. Dev. 2015, 15, 3–34. IEA Bioenergy. Bio-based Chemicals, Value Added Products From Biorefineries, 2011. Choi, S.; Song, C. W.; Shin, J. H.; Lee, S. Y. Biorefineries for the Production of Top Building Block Chemicals and Their Derivatives. Metab. Eng. 2015, 28, 223–239. Shen, L.; Haufe, J.; Patel, M. K. Product Overview and Market Projection of Emerging Bio-based Plastics. Group Sci. Technol. Soc. 2009, 41. Krull, R.; Bley, T. Filaments in Bioprocesses, Springer International Publishing, 2015. Giovannetti, G. T.; Spence, P. Biotechnology Report 2017, 2017. Alves de Oliveira, R.; Komesu, A.; Vaz Rossell, C. E.; Maciel Filho, R. Challenges and Opportunities in Lactic Acid Bioprocess DesigndFrom Economic to Production Aspects. Biochem. Eng. J. 2018, 133, 219–239. Alford, J. S. Bioprocess Control: Advances and Challenges. Comput. Chem. Eng. 2006, 30, 1464–1475. Neubauer, P. Biochemical Engineering Taking the Challenge: Advanced Methods for Development of Industrial Scale Bioprocesses. Biofuels 2010, 1, 1–4. Bailey, J. E. Mathematical Modeling and Analysis in Biochemical Engineering: Past Accomplishments and Future Opportunities. Biotechnol. Prog. 1998, 14, 8–20. Stephanopoulos, G. N.; Aristidou, A. A.; Nielsen, J. Review of Cellular Metabolism. In Metabolic Engineering; 1998; pp 21–79. Michaelis, L.; Menten, M. L. Die Kinetik der Invertinwirkung. Biochem. Z. 1913, 333–369. Schnell, S. Validity of the Michaelis-Menten Equation - Steady-State or Reactant Stationary Assumption: That is the Question. FEBS J. 2014, 281, 464–472. Monod, J. The Growth of Bacterial Cultures. Annu. Rev. Microbiol. 1949, 3, 371–394.

678

Bioreactor Models and Modeling Approaches

20. Heijnen, J. J. Balances and Reaction Models. In The Metabolic Pathway Engineering Handbook; Smolke, C. D., Ed.; 2010; pp 1–22. 21. Dunn, I. J.; Heinzle, E.; Ingham, J.; Prenosil, J. E. Biological Kinetics. In Biological Reaction Engineering: Dynamic Modelling Fundamentals with Simulation Examples, 2nd ed.;; 2005; pp 67–100. 22. Cox, M.; Nelson, D. L. Lehninger Principle of Biochemistry, 2008. 23. Contois, D. E. Kinetics of Bacterial Growth: Relationship Between Population Density and Specific Growth Rate of Continuous Cultures. J. Gen. Microbiol. 1959, 21, 40–50. 24. Powell, E. O. The Growth Rate of Microorganisms as a Function of Substrate Concentration. In Microbial Physiology and Continuous Culture; 1967; pp 34–56. 25. Moser, H. The Dynamics of Bacterial Populations Maintained in the Chemostat, Carnegie Institution of Washington, 1958. 26. Villadsen, J.; Nielsen, J.; Lidén, G. Growth Kinetics of Cell Cultures. In Bioreaction Engineering Principles, 2nd ed.;; 2011; pp 271–357. 27. Andrews, J. F. A Mathematical Model for the Continuous Culture of Microorganisms Utilizing Inhibitory Substrates. Biotechnol. Bioeng. 1968, 10, 707–723. 28. Hermsen, R.; Okano, H.; You, C.; Werner, N.; Hwa, T. A Growth-Rate Composition Formula for the Growth of E. coli on Co-utilized Carbon Substrates. Mol. Syst. Biol. 2015, 11, 801. 29. VAN Dedem, G.; Moo-Young, M. A Model for Diauxic Growth. Biotechnol. Bioeng. 1975, 17, 1301–1312. 30. Luedeking, R.; Piret, E. L. A Kinetic Study of the Lactic Acid Fermentation. Batch Process at Controlled pH. Biotechnol. Bioeng. 1959, 1, 393–412. 31. Ahmad, F.; Jameel, A. T.; Kamarudin, M. H.; Mel, M. Study of Growth Kinetic and Modeling of Ethanol Production by Saccharomyces Cerevisae. Afr. J. Biotechnol. 2011, 10, 18842–18846. 32. Martynova, J.; Mednis, M.; Vigants, A.; Zikmanis, P. Kinetic Modeling of Ethanol Fermentation by Yeast Kluyveromyces Marxianus from Lactose- and Inulin-containing Substrates. Eng. Rural Dev. 2017, 16, 88–97. 33. Baei, M. S.; Mahmoudi, M.; Yunesi, H. A Kinetic Model for Citric Acid Production From Apple Pomac by Aspergillus niger. Afr. J. Biotechnol. 2008, 7, 3487–3489. 34. Vlysidis, A.; Binns, M.; Webb, C.; Theodoropoulos, C. Glycerol Utilisation for the Production of Chemicals: Conversion to Succinic Acid, a Combined Experimental and Computational Study. Biochem. Eng. J. 2011, 58 (59), 1–11. 35. Pérez Rivero, C.; Sun, C.; Theodoropoulos, C.; Webb, C. Building a Predictive Model for PHB Production From Glycerol. Biochem. Eng. J. 2016, 116, 113–121. 36. Tohyama, M.; Patarinska, T.; Qiang, Z.; Shimizu, K. Modeling of the Mixed Culture and Periodic Control for PHB Production. Biochem. Eng. J. 2002, 10, 157–173. 37. Khanna, S.; Srivastava, A. K. Computer Simulated Fed-batch Cultivation for over Production of PHB: A Comparison of Simultaneous and Alternate Feeding of Carbon and Nitrogen. Biochem. Eng. J. 2006, 27, 197–203. 38. Pons, M. N.; Halme, A.; Heinzle, E.; Karim, M. N.; Lee, K. S.; Lim, H. C.; Saner, U. Bioprocess Monitoring and Control, 1991. 39. Dunn, I. J.; Ingham, J.; Prenosil, J. E.; Snape, J. B. Dynamics of Environmental Bioprocesses: Modelling & Simulation, 1995. 40. Bogaerts, P.; Hanus, R. Macroscopic Modelling of Bioprocesses with a View to Engineering Applications. In Engineering and Manufacturing for Biotechnology; Hofman, M., Thonart, P., Eds.; 2001; pp 77–109. 41. Simutis, R.; Lubbert, A. Bioreactor Control Improves Bioprocess Performance. Biotechnol. J. 2015, 10, 1115–1130. 42. Kompala, D. S.; Ramkrishna, D.; Tsao, G. T. Cybernetic Modeling of Microbial Growth on Multiple Substrates. Biotechnol. Bioeng. 1984, 26, 1272–1281. 43. Ramkrishna, D.; Kompala, D. S.; Tsao, G. T. Are Microbes Optimal Strategists? Biotechnol. Prog. 1987, 3, 121–126. 44. Kompala, D. S.; Ramkrishna, D.; Jansen, N. B.; Tsao, G. T. Investigation of Bacterial Growth on Mixed Substrates: Experimental Evaluation of Cybernetic Models. Biotechnol. Bioeng. 1986, 28, 1044–1055. 45. Baloo, S.; Ramkrishna, D. Metabolic Regulation in Bacterial Continuous Cultures: II. Biotechnol. Bioeng. 1991, 38, 1353–1363. 46. Ramakrishna, R.; Ramkrishna, D. Cybernetic Modeling of Growth in Mixed, Substitutable Substrate Environments: Preferential and Simultaneous Utilisation. Biotechnol. Bioeng. 1996, 52, 141–151. 47. Gadgil, C. J.; Bhat, P. J.; V Venkatesh, K. Cybernetic Model for the Growth of Saccharomyces cerevisiae on Melibiose. Biotechnol. Prog. 1996, 12, 744–750. 48. Varner, J.; Ramkrishna, D. Application of Cybernetic Models to Metabolic Engineering: Investigation of Storage Pathways. Biotechnol. Bioeng. 1998, 58, 282–291. 49. Varner, J.; Ramkrishna, D. Metabolic Engineering From a Cybernetic Perspective: Aspartate Family of Amino Acids. Metab. Eng. 1999, 1, 83–116. 50. Roubos, J. A.; Krabben, P.; Setnes, M.; Babuska, R.; Heijnen, J. J.; Verbruggen, H. B.; Hybrid Model Development for Fed-Batch Bioprocesses; Combining Physical Equations with the Metabolic Network and Black-Box Kinetics 2000, Vol. 41; pp 17–23. 51. Pappu, J. S. M.; Gummadi, S. N. Modeling and Simulation of Xylitol Production in Bioreactor by Debaryomyces Nepalensis NCYC 3413 Using Unstructured and Artificial Neural Network Models. Bioresour. Technol. 2016, 220, 490–499. 52. Geethalakshmi, S.; Narendran, S.; Pappa, N.; Ramalingam, S. Development of a Hybrid Neural Network Model to Predict Feeding Method in Fed-Batch Cultivation for Enhanced Recombinant Streptokinase Productivity in Escherichia coli. J. Chem. Technol. Biotechnol. 2012, 87, 280–285. 53. Parekh, S.; Vinci, V. A.; Strobel, R. J. Improvement of Microbial Strains and Fermentation Processes. Appl. Microbiol. Biotechnol. 2000, 54, 287–301. 54. Zheng, Y.; Sriram, G. Mathematical Modeling: Bridging the Gap between Concept and Realization in Synthetic Biology. J. Biomed. Biotechnol. 2010. https://doi.org/10.1155/ 2010/541609. 55. Cvijovic, M.; Bordel, S.; Nielsen, J. Mathematical Models of Cell Factories: Moving Towards the Core of Industrial Biotechnology. Microbial Biotechnology 2011, 4, 572–584. 56. Stephanopoulos, G. Metabolic Fluxes and Metabolic Engineering. Metab. Eng. 1999, 1, 1–11. 57. Stephanopoulos, G. N.; Aristidou, A. A.; Nielsen, J. The Essence of Metabolic Engineering. In Metabolic Engineering, 1st ed.; Academic Press, 1998; pp 1–20. 58. Price, N. D.; Papin, J. A.; Schilling, C. H.; Palsson, B. O. Genome-Scale Microbial in Silico Models: The Constraints-based Approach. Trends Biotechnol. 2003, 21, 162–169. 59. Nielsen, J. Biochemical Reaction Network. In Bioreaction Engineering Principles; Nielsen, J., Villadsen, J., Lidén, G., Eds., 2nd ed.;; 2002; pp 119–188. 60. Bordbar, A.; Monk, J. M.; King, Z. A.; Palsson, B.Ø. Constraint-based Models Predict Metabolic and Associated Cellular Functions. Nat. Rev. Genet. 2014, 15, 107–120. 61. Llaneras, F.; Picó, J. Stoichiometric Modelling of Cell Metabolism. J. Biosci. Bioeng. 2008, 105, 1–11. 62. Heijnen, J. J. Metabolic Control Analysis. In The Metabolic Pathway Engineering Handbook; Christina, D. S., Ed., CRC Press: New York, 2010; pp 1–30. 63. John Villadsen, G. L. J. N. Scale-Up of Bioprocesses. In Bioreaction Engineering Principles, 2nd ed.; Springer US., 2003; pp 477–518. 64. Mashego, M. R.; van Gulik, W. M.; Vinke, J. L.; Visser, D.; Heijnen, J. J. In Vivo Kinetics with Rapid Perturbation Experiments in Saccharomyces cerevisiae Using a Secondgeneration BioScope. Metab. Eng. 2006, 8, 370–383. 65. Kresnowati, M. T. a P.; Suarez-Mendez, C. M.; a van Winden, W.; van Gulik, W. M.; Heijnen, J. J. Quantitative Physiological Study of the Fast Dynamics in the Intracellular pH of Saccharomyces cerevisiae in Response to Glucose and Ethanol Pulses. Metab. Eng. 2008, 10, 39–54. 66. Lequeux, G.; Beauprez, J.; Maertens, J.; Van Horen, E.; Soetaert, W.; Vandamme, E.; Vanrolleghem, P. A. Dynamic Metabolic Flux Analysis Demonstrated on Cultures where the Limiting Substrate is Changed from Carbon to Nitrogen and Vice Versa. J. Biomed. Biotechnol. 2010. https://doi.org/10.1155/2010/621645. 67. Soustelle, M. Pseudo- and Quasi-Steady State Modes. In An Introduction to Chemical Kinetics; 2013; pp 135–164. 68. Stephanopoulos, G. N.; Aristidou, A. A.; Nielsen, J. Metabolic Flux Analysis. In Metabolic Engineering; 1998. 69. Beard, D. A.; Babson, E.; Curtis, E.; Qian, H. Thermodynamic Constraints for Biochemical Networks. J. Theor. Biol. 2004, 228, 327–333. 70. Kümmel, A.; Panke, S.; Heinemann, M. Systematic Assignment of Thermodynamic Constraints in Metabolic Network Models. BMC Bioinf. 2006, 7, 512. 71. Angeles-Martinez, L.; Theodoropoulos, C. Estimation of Flux Distribution in Metabolic Networks Accounting for Thermodynamic Constraints: The Effect of Equilibrium vs. Blocked Reactions. Biochem. Eng. J. 2016, 105, 347–357. 72. Covert, W.; Markus; Schilling, C.; Palson, B. Regulation of Gene Expression in Flux Balance Models of Metabolism. J. Theor. Biol. 2001, 213, 73–88. 73. Stelling, J.; Klamt, S.; Bettenbrock, K.; Schuster, S.; Gilles, E. Metabolic Network Structure Determines Key Aspects of Functionality and Regulation. Nature 2002, 420, 190–193. 74. Stephanopoulos, G. N.; Aristidou, A. A.; Nielsen, J. Application of Metabolic Flux Analysis. In Metabolic Engineering; 1998.

Bioreactor Models and Modeling Approaches

679

75. Pey, J.; Theodoropoulos, C.; Rezola, A.; Rubio, A.; Cascante, M.; Planes, F. J. Do Elementary Flux Modes Combine Linearly at the “Atomic” Level? Integrating Tracer-based Metabolomics Data and Elementary Flux Modes. Biosystems 2011, 105, 140–146. 76. Papin, J. A.; Stelling, J.; Price, N. D.; Klamt, S.; Schuster, S.; Palsson, B. O. Comparison of Network-based Pathway Analysis Methods. Trends Biotechnol. 2004, 22, 400–405. 77. Orth, J. D.; Thiele, I.; Palsson, B.Ø. What is Flux Balance Analysis? Nat. Biotechnol. 2010, 28, 245–248. 78. Tajparast, M.; Frigon, D. Genome-scale Metabolic Model of Rhodococcus jostii RHA1 (IMT1174) to Study the Accumulation of Storage Compounds During Nitrogen-Limited Condition. BMC Syst. Biol. 2015, 9, 43. 79. Varma, A.; Palsson, B. O. Metabolic Flux Balancing: Basic Concepts, Scientific and Practical Use. Bio Technol 1994, 12, 994–998. 80. Knorr, A. L.; Jain, R.; Srivastava, R. Bayesian-based Selection of Metabolic Objective Functions. Bioinformatics 2007, 23, 351–357. 81. Schuetz, R.; Kuepfer, L.; Sauer, U. Systematic Evaluation of Objective Functions for Predicting Intracellular Fluxes in Escherichia coli. Mol. Syst. Biol. 2007, 3, 119. 82. Holzhütter, H. G. The Principle of Flux Minimization and its Application to Estimate Stationary Fluxes in Metabolic Networks. Eur. J. Biochem. 2004, 271, 2905–2922. 83. Costanza, J.; Carapezza, G.; Angione, C.; Liò, P.; Nicosia, G. Multi-objective Optimisation, Sensitivity and Robustness Analysis in FBA Modelling. In Computational Methods in Systems Biology; 2012; pp 127–147. 84. Wintermute, E. H.; Lieberman, T. D.; Silver, P. A. An Objective Function Exploiting Suboptimal Solutions in Metabolic Networks. BMC Syst. Biol. 2013, 7, 98. 85. Antoniewicz, M. R. Dynamic Metabolic Flux Analysis–Tools for Probing Transient States of Metabolic Networks. Curr. Opin. Biotechnol. 2013, 24, 973–978. 86. Niklas, J.; Schräder, E.; Sandig, V.; Noll, T.; Heinzle, E. Quantitative Characterization of Metabolism and Metabolic Shifts During Growth of the New Human Cell Line AGE1.HN Using Time Resolved Metabolic Flux Analysis. Bioproc. Biosyst. Eng. 2011, 34, 533–545. 87. Leighty, R. W.; Antoniewicz, M. R. Dynamic Metabolic Flux Analysis (DMFA): a Framework for Determining Fluxes at Metabolic Non-Steady State. Metab. Eng. 2011, 13, 745–755. 88. Henson, M. A.; Hanly, T. J. Dynamic Flux Balance Analysis for Synthetic Microbial Communities. IET Syst. Biol. 2014, 8, 214–229. 89. Varma, A.; Palsson, B. O. Stoichiometric Flux Balance Models Quantitatively Predict Growth and Metabolic By-Product Secretion in Wild-Type Escherichia coli W3110. Appl. Environ. Microbiol. 1994, 60, 3724–3731. 90. Mahadevan, R.; Edwards, J. S.; Doyle, F. J. Dynamic Flux Balance Analysis of Diauxic Growth in Escherichia coli. Biophys. J. 2002, 83, 1331–1340. 91. Meadows, A. L.; Karnik, R.; Lam, H.; Forestell, S.; Snedecor, B. Application of Dynamic Flux Balance Analysis to an Industrial Escherichia coli Fermentation. Metab. Eng. 2010, 12, 150–160. 92. Flassig, R. J.; Fachet, M.; Höffner, K.; Barton, P. I.; Sundmacher, K. Dynamic Flux Balance Modeling to Increase the Production of High-Value Compounds in Green Microalgae. Biotechnol. Biofuels 2016, 9, 165. 93. Hjersted, J. L.; Henson, M. A. Steady-state and Dynamic Flux Balance Analysis of Ethanol Production by Saccharomyces cerevisiae. IET Syst. Biol. 2009, 3, 167–179. 94. Hanly, T. J.; Henson, M. A. Dynamic Flux Balance Modeling of Microbial Co-cultures for Efficient Batch Fermentation of Glucose and Xylose Mixtures. Biotechnol. Bioeng. 2011, 108, 376–385. 95. Höffner, K.; Harwood, S. M.; Barton, P. I. A Reliable Simulator for Dynamic Flux Balance Analysis. Biotechnol. Bioeng. 2013, 110, 792–802. 96. Feng, X.; Xu, Y.; Chen, Y.; Tang, Y. J. Integrating Flux Balance Analysis into Kinetic Models to Decipher the Dynamic Metabolism of Shewanella oneidensis MR-1. PLoS Comput. Biol. 2012, 8. 97. Waldherr, S.; Oyarzún, D. A.; Bockmayr, A. Dynamic Optimization of Metabolic Networks Coupled with Gene Expression. J. Theor. Biol. 2015, 365, 469–485. 98. Willemsen, A. M.; Hendrickx, D. M.; Hoefsloot, H. C. J.; Hendriks, M. M. W. B.; Wahl, S. A.; Teusink, B.; Smilde, A. K.; van Kampen, A. H. C. MetDFBA: Incorporating Timeresolved Metabolomics Measurements into Dynamic Flux Balance Analysis. Mol. Biosyst. 2015, 11, 137–145. 99. Hohenschuh, W.; Hector, R.; Murthy, G. S. A Dynamic Flux Balance Model and Bottleneck Identification of Glucose, Xylose, Xylulose Co-Fermentation in Saccharomyces cerevisiae. Bioresour. Technol. 2015, 188, 153–160. 100. Angeles-Martinez, L.; Theodoropoulos, C. A Lattice-Boltzmann Scheme for the Simulation of Diffusion in Intracellular Crowded Systems. BMC Bioinf. 2015, 16, 1–15. 101. Angeles-Martinez, L.; Theodoropoulos, C. The Influence of Crowding Conditions on the Thermodynamic Feasibility of Metabolic Pathways. Biophys. J. 2015, 109, 2394–2405. 102. Song, H. S.; Ramkrishna, D. Prediction of Metabolic Function from Limited Data: Lumped Hybrid Cybernetic Modeling (L-HCM). Biotechnol. Bioeng. 2010, 106, 271–284. 103. Straight, J. V.; Ramkrishna, D. Cybernetic Modeling and Regulation of Metabolic Pathways. Growth on Complementary Nutrients. Biotechnol. Prog. 1994, 10, 574–587. 104. Varner, J.; Ramkrishna, D. Metabolic Engineering From a Cybernetic Perspective. 1. Theoretical Preliminaries. Biotechnol. Prog. 1999, 15, 407–425. 105. Varner, J.; Ramkrishna, D. Metabolic Engineering From a Cybernetic Perspective. 2. Qualitative Investigation of Nodal Architectures and Their Response to Genetic Perturbation. Biotechnol. Prog. 1999, 15, 426–438. 106. Varner, J.; Ramkrishna, D. The Non-linear Analysis of Cybernetic Models. Guidelines for Model Formulation. J. Biotechnol. 1999, 71, 67–104. 107. Young, J. D.; Ramkrishna, D. On the Matching and Proportional Laws of Cybernetic Models. Biotechnol. Prog. 2007, 23, 83–99. 108. Young, J. D.; Henne, K. L.; Morgan, J. A.; Konopka, A. E.; Ramkrishna, D. Integrating Cybernetic Modeling with Pathway Analysis Provides a Dynamic, Systems-Level Description of Metabolic Control. Biotechnol. Bioeng. 2008, 100, 542–559. 109. Il Kim, J.; Varner, J. D.; Ramkrishna, D. A Hybrid Model of Anaerobic E. coli GJT001: Combination of Elementary Flux Modes and Cybernetic Variables. Biotechnol. Prog. 2008, 24, 993–1006. 110. Song, H. S.; Ramkrishna, D. Reduction of a Set of Elementary Modes Using Yield Analysis. Biotechnol. Bioeng. 2009, 102, 554–568. 111. Wittmann, C. Analysis and Engineering of Metabolic Pathway Fluxes in Corynebacterium Glutamicum. Adv Biochem Engin/Biotechnol 2010, 120, 21–49. 112. Saa, P. A.; Nielsen, L. K. Formulation, Construction and Analysis of Kinetic Models of Metabolism: A Review of Modelling Frameworks. Biotechnol. Adv. 2017, 35, 981–1003. 113. Almquist, J.; Cvijovic, M.; Hatzimanikatis, V.; Nielsen, J.; Jirstrand, M. Kinetic Models in Industrial Biotechnology - Improving Cell Factory Performance. Metab. Eng. 2014, 24, 38–60. 114. Stanford, N. J.; Lubitz, T.; Smallbone, K.; Klipp, E.; Mendes, P.; Liebermeister, W. Systematic Construction of Kinetic Models from Genome-Scale Metabolic Networks. PLoS One 2013, 8. 115. Ishii, N.; Robert, M.; Nakayama, Y.; Kanai, A.; Tomita, M. Toward Large-Scale Modeling of the Microbial Cell for Computer Simulation. J. Biotechnol. 2004, 113, 281–294. 116. Jamshidi, N.; Palsson, B.Ø. Formulating Genome-Scale Kinetic Models in the Post-Genome Era. Mol. Syst. Biol. 2008, 4. 117. Smallbone, K.; Simeonidis, E.; Swainston, N.; Mendes, P. Towards a Genome-Scale Kinetic Model of Cellular Metabolism. BMC Syst. Biol. 2010, 4, 6. 118. Karr, J. R.; Sanghvi, J. C.; MacKlin, D. N.; Gutschow, M. V.; Jacobs, J. M.; Bolival, B.; Assad-Garcia, N.; Glass, J. I.; Covert, M. W. A Whole-Cell Computational Model Predicts Phenotype From Genotype. Cell 2012, 150, 389–401. 119. Teusink, B.; Passarge, J.; Reijenga, C. A.; Esgalhado, E.; Van Der Weijden, C. C.; Schepper, M.; Walsh, M. C.; Bakker, B. M.; Van Dam, K.; Westerhoff, H. V.; Snoep, J. L. Can Yeast Glycolysis Be Understood Terms of Vitro Kinetics of the Constituent Enzymes? Testing Biochemistry. Eur. J. Biochem. 2000, 267, 5313–5329. 120. Yugi, K.; Tomita, M. A General Computational Model of Mitochondrial Metabolism in a Whole Organelle Scale. Bioinformatics 2004, 20, 1795–1796. 121. Chakrabarti, A.; Miskovic, L.; Soh, K. C.; Hatzimanikatis, V. Towards Kinetic Modeling of Genome-Scale Metabolic Networks Without Sacrificing Stoichiometric, Thermodynamic and Physiological Constraints. Biotechnol. J. 2013, 8, 1043–1057. 122. Liebermeister, W.; Klipp, E. Bringing Metabolic Networks to Life: Convenience Rate Law and Thermodynamic Constraints. Theor. Biol. Med. Model. 2006, 3, 41. 123. Liebermeister, W.; Klipp, E. Bringing Metabolic Networks to Life: Integration of Kinetic, Metabolic, and Proteomic Data. Theor. Biol. Med. Model. 2006, 3, 42. 124. Steuer, R.; Gross, T.; Selbig, J.; Blasius, B. Structural Kinetic Modeling of Metabolic Networks. Proc. Natl. Acad. Sci. USA 2006, 103, 11868–11873. 125. Link, H.; Christodoulou, D.; Sauer, U. Advancing Metabolic Models with Kinetic Information. Curr. Opin. Biotechnol. 2014, 29, 8–14. 126. Saa, P. A.; Nielsen, L. K. Construction of Feasible and Accurate Kinetic Models of Metabolism: A Bayesian Approach. Sci. Rep. 2016, 6, 1–13.

680

Bioreactor Models and Modeling Approaches

127. Khodayari, A.; Zomorrodi, A. R.; Liao, J. C.; Maranas, C. D. A Kinetic Model of Escherichia coli Core Metabolism Satisfying Multiple Sets of Mutant Flux Data. Metab. Eng. 2014, 25, 50–62. 128. Khodayari, A.; Maranas, C. D. A Genome-scale Escherichia coli Kinetic Metabolic Model K-ecoli457 Satisfying Flux Data for Multiple Mutant Strains. Nat. Commun. 2016, 7, 13806. 129. Vasilakou, E.; Machado, D.; Theorell, A.; Rocha, I.; Nöh, K.; Oldiges, M.; Wahl, S. A. Current State and Challenges for Dynamic Metabolic Modeling. Curr. Opin. Microbiol. 2016, 33, 97–104. 130. Kadir, T.; Mannan, A. A.; Kierzek, A. M.; McFadden, J.; Shimizu, K. Modeling and Simulation of the Main Metabolism in Escherichia coli and its Several Single-Gene Knockout Mutants with Experimental Verification. Microb. Cell Factories 2010, 9, 88. 131. Snowden, T. J.; van der Graaf, P. H.; Tindall, M. J. A Combined Model Reduction Algorithm for Controlled Biochemical Systems. BMC Syst. Biol. 2017, 11, 1–18. 132. Mannan, A. A.; Toya, Y.; Shimizu, K.; McFadden, J.; Kierzek, A. M.; Rocco, A. Integrating Kinetic Model of E. coli with Genome Scale Metabolic Fluxes Overcomes its Open System Problem and Reveals Bistability in Central Metabolism. PLoS One 2015, 10, 1–36. 133. Wiechert, W.; Noack, S. Mechanistic Pathway Modeling for Industrial Biotechnology: Challenging but Worthwhile. Curr. Opin. Biotechnol. 2011, 22, 604–610. 134. Xie, W.; Bonis, I.; Theodoropoulos, C. Data-driven Model Reduction-based Nonlinear MPC for Large-Scale Distributed Parameter Systems. J. Process Contr. 2015, 35, 50–58. 135. Bonis, I.; Theodoropoulos, C. Model Reduction-based Optimization Using Large-Scale Steady-State Simulators. Chem. Eng. Sci. 2012, 69, 69–80. 136. Theodoropoulos, C. Optimisation and Linear Control of Large Scale Nonlinear Systems: A Review and a Suite of Model Reduction-Based Techniques. In Coping with Complexity: Model Reduction and Data Analysis; Gorban, A. N., Roose, D., Eds., 1st ed.; Springer-Verlag Berlin Heidelberg, 2011; pp 37–61. 137. Hewitt, C. J.; Nienow, A. W. The Scale-Up of Microbial Batch and Fed-batch Fermentation Processes. Comprehens. Bioprocess Eng. 2010, 295–320. 138. Yang, H.; Allen, D. G. Model-based Scale-Up Strategy for Mycelial Fermentation Processes. Can. J. Chem. Eng. 1999, 77, 844–854. 139. Delvigne, F.; Takors, R.; Mudde, R.; van Gulik, W.; Noorman, H. Bioprocess Scale-Up/Down as Integrative Enabling Technology: from Fluid Mechanics to Systems Biology and beyond. Microb. Biotechnol. 2017, 10, 1267–1274. 140. Koerich, D. M.; Rosa, L. M. Numerical Evaluation of the Low Reynolds Turbulent Flow Behaviour in a Bioreactor. Int. J. Simul. Process Model. 2016, 11, 66–75. 141. Guha, D.; Ramachandran, P. A.; Dudukovic, M. P. Evaluation of Large Eddy Simulation and Euler-Euler CFD Models for Solids Flow Dynamics in a Stirred Tank Reactor; 54; VTT Publications, 2008; pp 766–778. 142. Zheng, X.; Smith, R.; Theodoropoulos, C. Modelling and Optimisation of Distributed-Parameter Batch and Semi-Batch Reactor Systems. In European Symposium on Computer Aided Chemical Engineering; 2005; pp 1087–1092. 143. Delafosse, A.; Collignon, M. L.; Calvo, S.; Delvigne, F.; Crine, M.; Thonart, P.; Toye, D. CFD-Based Compartment Model for Description of Mixing in Bioreactors. Chem. Eng. Sci. 2014, 106, 76–85. 144. Delafosse, A.; Calvo, S.; Collignon, M. L.; Delvigne, F.; Crine, M.; Toye, D. Euler-Lagrange Approach to Model Heterogeneities in Stirred Tank Bioreactors - Comparison to Experimental Flow Characterization and Particle Tracking. Chem. Eng. Sci. 2015, 134, 457–466.

Relevant Websites https://www.brenda-enzymes.org/. https://www.kegg.jp/. https://ecocyc.org/. http://www.genedb.org/. http://eawag-bbd.ethz.ch/.

2.46

Product Recovery

R Wohlgemuth, Sigma-Aldrich, Buchs, Switzerland © 2011 Elsevier B.V. All rights reserved. This is a reprint of R. Wohlgemuth, 2.42 - Product Recovery, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 591-601.

2.46.1 2.46.2 2.46.3 2.46.3.1 2.46.3.2 2.46.3.3 2.46.3.4 2.46.3.5 2.46.3.6 2.46.3.7 2.46.3.8 2.46.3.9 2.46.3.10 2.46.4 2.46.4.1 2.46.4.2 2.46.4.3 2.46.4.4 2.46.4.5 2.46.4.6 2.46.4.7 2.46.4.8 2.46.5 2.46.5.1 2.46.5.2 2.46.5.3 2.46.5.4 2.46.5.5 2.46.6 2.46.7 References

Introduction Historical Background Modular Unit Operations in Downstream Processing Recovery of Solids and Liquids Cell Treatment Solvent Extraction Liquid–Liquid Phase Separation Crystallization and Precipitation Adsorption Distillation Chromatography Membrane Filtration Other Unit Operations Integrated Unit Operations in Downstream Processing Solid–Liquid Separation and Product Recovery Reaction and Solvent Extraction Reaction and Liquid–Liquid Phase Separation Reaction and Crystallization/Precipitation Reaction and Adsorption Reaction and Distillation Reaction and Chromatography Reaction and Membrane Filtration Product Purification Metabolite Purification Lipid Purification Protein Purification Nucleic Acid Purification Carbohydrate Biopolymer Purification Product Formulation and Stabilization Conclusion

682 682 682 683 684 684 684 684 684 685 685 685 685 686 686 686 686 687 687 687 687 688 688 688 689 689 689 689 689 690 690

Glossary Adsorption Favored product binding to an adsorber/resin compared to the aqueous bulk phase as a way of concentrating the product. Chromatography Workhorse unit operation for product separation from byproducts by passing a mixture over a suitable column and collecting the separated pure product fraction. Crystallization/precipitation Popular and often used downstream processing step making use of decreased product solubility by changing the medium/solvent composition and/or conditions. Distillation Separation of more volatile products from less volatile products by evaporation and subsequent condensation. Phase separation Primary recovery step making use of the separation of nonmiscible phases such as solid–liquid, organic– aqueous, and aqueous–aqueous phases. Membrane filtration Spatial separation of different molecules or particles by polymeric or ceramic membranes of defined pore sizes, which retain the products larger than the pore size and pass the impurities or vice versa.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00096-3

681

682

2.46.1

Product Recovery

Introduction

The knowledge of how to obtain products after biotechnological processes from a mixture of many ingredients has not only been essential in the historical development of biotechnology from small manual procedures to large industrial technologies but is also a key factor for industrial biotechnology and biotransformations today.1,2 The significant amount of work, energy, and equipment that is needed for the separation of products, reusable materials, and wastes depend to a large extent on the type of product (product mixture or highly purified product, low-molecular-weight product or high-molecular-weight product, and stable product or highly labile product) and also on the type of biological source. Whether the source of the biomass, crude extract, broth, or fluid comes from biotransformation, microbial fermentation or cell culture, plant or animal tissue or body fluids, processes of separation, refinement, purification, or transformation are required to recover the desired product. The product recovery operations, generally, require the isolation and purification of the product from dilute aqueous media and often constitute the major cost contribution of a production process. It is therefore important to develop high-yield processes with a reduced amount of work and energy, with a minimum number of steps. Each operation needed depends both on the preceding bioprocess design and on the form and concentration of the final stable product required. The variety of product types has led to different approaches and methodologies for these processes, and the number of publications on product recovery has steadily increased over the last century (Figure 1).

2.46.2

Historical Background

Product recovery operations from complex mixtures, obtained after bioprocesses, have accompanied human history since the time of producing beer, wine, vinegar, and other food preparations to the present industrial processes for making a variety of products such as organic acids and other chemicals, antibiotics, amino acids, vitamins, proteins and enzymes, solvents, and liquid fuels. As waste has accumulated with the industrial scale-up of processes, the waste treatment and utilization aspects of such bioprocesses have become more important. The increasing know-how in the selection of the best product recovery operations3–5 has been a key to the industrial large-scale production of these products, which have increased the quality of life over the last two centuries.

2.46.3

Modular Unit Operations in Downstream Processing

The great number of individual downstream processes in biotechnology can be perceived as an entity of steps or unit operations, which are based on common technologies and the same fundamental sciences as the unit operations in chemical engineering.6 Each unit operation is based on exploiting unique physico-chemical properties of the product of interest compared with the other species in the mixture. The challenges in the downstream processes of biotechnology are not diminished compared with chemical engineering but are different and are an important research area of biochemical engineering. The great product diversity makes separate considerations of low-molecular-weight and high-molecular-weight products useful. In the area of low-molecular-weight products, many unit operations in biotechnology such as extraction, phase separation, crystallization, adsorption, or distillation have the same physicochemical principles as in chemistry (Figure 2). Important physicochemical properties that can be utilized for the recovery of small-molecular-weight compounds are the charge, molecular Number of publications per year 90 80

Publications per year

70 60 50

Number of publications per year

40 30 20 10

Year Figure 1

Publications in product recovery over the last century.

2005

1997

1989

1981

1973

1965

1957

1949

1941

1933

1925

1917

1909

0

Product Recovery

Recovery of solids and liquids

Membrane filtration

Chromatography

Cell treatment

Biomass, crude extract, or broth from bioprocess

Solvent extraction

Liquid–liquid phase separation

Distillation

Adsorption

Figure 2

683

Crystallization and precipitation

Overview of unit operations in product recovery.

weight, hydrophobicity, volatility, and solubility behavior. The fundamental chemical and biochemical engineering of the underlying complex processes are essential for industrial production processes. Purification steps such as the separation of regioisomers, cis/trans isomers, diastereomers, or enantiomers, and protection of labile functional groups and their final deprotection also benefit from the interaction between chemistry and biotechnology. Unit operations such as cell harvesting and cell separation, cell preservation and cell immobilization, product secretion, or cell homogenization and cell lysis can build on the vast biochemical experience and are specific for biotechnology operations (Figure 2). This is also true for largemolecular-weight products such as proteins, nucleic acids, polysaccharides, and other biopolymers, where additional factors like increased viscosity, degradation by enzymes, stability under process conditions, and folding play a role. The development of the optimal product recovery process may require substantial work, and the final engineering of this process is often having a big influence on cost, equipment size, and waste. It is therefore of much interest to use conventional unit operations and to employ currently established and available techniques and materials like resins and membranes. This reduces the risks in scale-up, process robustness, technology transfer, and good manufacturing practice (GMP) issues for equipment, methods, and processes.

2.46.3.1

Recovery of Solids and Liquids

The separation of solid particles, such as adsorbers, cells, inclusion bodies, virus-like particles, or crystals, from a solution depends both on particle properties such as size and density and on the properties of liquid media such as viscosity, density, and gel, emulsion, or foam formation tendency. The selection of a suitable separation method is mainly determined by size, size differences, and density. Particle–particle separation of coarse particles from fine particles may be useful to separate first macroscopic particles of nonutilized solid substrates or high concentrations of products on crystals or on solid adsorbers. Larger polymeric adsorber particles can be well separated from smaller biological cells by a high-performance sieving step prior to solid–liquid separation. Initial harvesting, medium exchange, or clarification requires suspensions of solids to be separated from their liquid media. Primary recovery processing steps such as decanting, sedimentation or settling, centrifugation, depth filtration, or tangential flow microfiltration7 are in widespread use. Different types of centrifugal separators and continuous centrifuges provide robust and broadly applicable equipment for handling various suspensions with different methods of discharge. The pore size of the filtration unit can thereby be adjusted to the size of the particles to be separated. The advantage of using microfiltration membranes with 0.2 mm pore size compared with sedimentation or centrifugation is the generation of a particle-free solution, which can be directly used in the next downstream processing steps and does not need further clarification.8 Depending on whether the product is contained in the solids, which can consist of cells and intracellular products, cell debris, inclusion bodies, or crystals, or is dissolved in the liquid medium, the best-suited techniques for collecting or removing the solids are selected. The size and size distribution, morphology, and concentration of the solid particles have a key influence on the technique that is selected for the recovery of the solid particles. Therefore, it is already in the bioprocess design phase that these key parameters of the solid particles need to be considered in order to minimize the effort going into solid–liquid separations. The separation of liquids from a solution can be straightforward if the liquid is nonmiscible, the formation of foams, soaps, and emulsions can be avoided, and liquid droplets coalesce fast to form a separate phase. Since the presence of biological cells and nonmiscible product in the liquid aqueous medium can lead to complex phase behavior, the phase separation time can in certain cases be very long. Addition of antifoam reagents, salts, buffers, or other reagents to break emulsions can be used to reduce phase separation times to reasonable values.

684 2.46.3.2

Product Recovery Cell Treatment

The nature of the product (whole cell, extracellular product, or intracellular product) also determines the type of cell treatment. From the gentle recovery of viable whole cells, mechanical dewatering or water removal by drying or lyophilization to the complete homogenization and lysis of cells by enzymatic, chemical, and physical methods, each cell treatment procedure needs to be optimized for the specific target products. Various methods of flocculation can be used to aggregate the cells and therefore make sedimentation, centrifugation, or filtration easier. For intracellular products, selective product release strategies, considering key factors for manufacturability such as viscosity reduction, removal of product-related contaminants, and elimination of enzymes that reduce product quality, are crucial for successful downstream processing.9 One particular interesting selective product release strategy is product secretion, making use of natural transport mechanisms or engineering cells capable of secreting the product into the medium. This is essential in large-scale protein production, where the work-up is simplified significantly if the product can be recovered from the medium. In the case where the product of interest cannot be secreted, cell disruption is the key unit operation, which liberates intracellular products. A case-by-case assessment is required to find whether mechanical, physical, chemical, or biochemical method is the best disruption technology.

2.46.3.3

Solvent Extraction

Organic solvents that are nearly immiscible with water and rapidly form two liquid phases have been the classic way of extraction. High solubility of products in organic solvents or components of an extractant phase can be utilized to extract the product from the aqueous phase, if the distribution coefficient of the product between the aqueous and the organic phase is favorable.10 The product solubility of ionic molecules in the organic solvent may be increased by neutralization. The extraction method depends on the use of final product. This liquid–liquid extraction is a product concentration step and requires a good and fast separation of the organic and aqueous phases. If products are localized in cells or cellular compartments, medium and water removal can be useful for high-yield solvent extraction of the products. The liquid–liquid extraction of organic acids11–13 and alcohols14,15 has focused on the extractive recovery of neutral undissociated molecules and has improved the process technologies based on extraction. Phase separations of aqueous and organic phases into two phases with clear phase boundaries are scalable low-cost unit operations, which can be developed rapidly and are therefore used extensively in industry. As the time for phase separation can vary depending on the influence of additional components in the reaction mixture, small additions of antifoam compounds or salts help to accelerate phase separation processes. Although solvent extraction has been widely adopted by industry, some final products may preclude the use of solvent extraction and the large solvent consumption is a disadvantage.

2.46.3.4

Liquid–Liquid Phase Separation

While solvent extraction with organic–aqueous two-phase systems has been mainly the domain of small molecules product recovery, aqueous–aqueous two-phase system has been advantageous for the recovery of proteins and enzymes without any damage.16 The situation is different for the practical applications of aqueous two-phase partitioning and phase separations, where, in general, the costs of phase-forming polymers and process development times have limited this operation in industrial processes.17 A direct comparison of ion-exchange chromatography and aqueous two-phase separation (ATPS) has shown the superior overall process yield of ATPS at lower costs than ion-exchange chromatography.

2.46.3.5

Crystallization and Precipitation

The simplicity and low cost of these liquid–solid phase separations have made crystallization and precipitation one of the most often used and popular downstream processing steps. The ideal case for this product recovery operation is a preceding bioprocess whereby the product is formed at a concentration well above the product solubility in the reaction medium. If the product is soluble even at high concentrations in the reaction medium, changes in pH, temperature, solvent composition, and ionic strength can offer opportunities for product crystallization or precipitation. The development of new product precipitation and crystallization procedures is still challenging for both small and large molecules, but rewarding if successful. Even if the best conditions for the crystallization of pure molecules have been developed, their extension to the crystallization of the same molecules in their more complex media is not trivial. Despite its extensive applications, crystallization at production scale can be difficult to characterize and improve by process analytical technology.18 Precipitation and crystallization of products as insoluble barium or calcium salts have been commonly employed for the recovery of organic acids. This technique has been used for a long time in the small molecule field as well as in the protein field, although in the latter field mostly at small scale. Ammonium sulfate is among the commonly used precipitation reagents due to its high solubility and low cost and the availability of extensive data on its saturation concentrations under various conditions. Crystallization and precipitation are both separation and purification processes and do not need expensive equipment. Therefore, they are experiencing a renaissance as simple and low-cost purification methods for small biomolecules such as metabolites and large biomolecules such as proteins.

2.46.3.6

Adsorption

The distribution of products between the bulk aqueous phase and an adsorber is a good tool for product recovery, if organic solvents have damaging effects. The use of polymeric adsorbers for the recovery of small molecules from dilute media is well established.19 A

Product Recovery

685

variety of hydrophobic polymers, ion exchangers, and other functionalized polymers, which are available at large scale, can be used for designing the best product distribution between the adsorber and liquid medium. As the adsorption characteristics such as capacity, selectivity, or kinetics are not known, in general, for the product to be recovered, these data have to be determined experimentally. The assessment of adsorbent resins for their capabilities of product recovery has been useful for selecting the most suitable adsorbers for small molecule recovery (antibiotics and chiral building blocks). After separation of the productloaded adsorber from the reaction mixture, the product can be easily recovered from the adsorber by solvent elution or extraction, and the adsorber can be recycled.

2.46.3.7

Distillation

Volatile products can be separated from nonvolatile products by evaporation, gas stripping, and subsequent condensation or adsorption. Besides traditional alcoholic drinks and biofuels, a variety of products such as solvents, flavors and fragrances, terpenes, and oils from bio-based processes are preferentially recovered by distillation. Fractionation according to boiling points depends on the boiling point differences of product and other volatile components of the reaction mixture. If simple distillations do not achieve the required product purity, advanced fine distillation techniques have proven useful in the purification of terpenes from challenging natural oil mixtures with components of similar boiling points. The effect of small impurities on possible reactions of the product during distillation at increased temperatures needs to be checked.

2.46.3.8

Chromatography

Passing a product mixture over a chromatography column is not only an important standard unit operation but also a workhorse of downstream processes for product recovery. Differences in the partitioning of the individual components between the solvent and the chromatographic material in the column and the selection of the best conditions for enlarging these differences enable high-resolution bioseparations. Highly selective chromatographic capture steps under mild product recovery conditions are key elements of a flexible, but generic downstream process platform. Adsorption chromatography, affinity chromatography, hydrophobic and hydrophilic interaction chromatography, ion-exchange chromatography, and gel filtration chromatography can provide unit operation platforms for certain product classes with biochemical similarities. These major chromatographic methods use charge, affinity, polarity, and size as the basis of separation. Although the best purifications can be obtained with affinity chromatography, the most common chromatographic method is ion-exchange chromatography. The optimization of chromatographic separations relies heavily on experimental data, and the combination of high-throughput screening with genetic algorithms has provided powerful tools for rapid process development.20 Different small molecule product classes such as metabolites and lipids and large molecule classes such as proteins, nucleic acids, and polysaccharides can be separated by a wealth of specialized chromatographic experience accumulated over the years. Although chromatographic separations have a high resolving power for many mixtures, there are limitations from the industrial and large-scale perspective. Alternatives such as membrane chromatography can increase throughput and overcome traditional bottlenecks in column chromatography.21

2.46.3.9

Membrane Filtration

Nature provides the role model processes for product recovery by highly selective as well as nonselective membrane filtrations through the natural lipid membranes of biological cells. As the macroscopic engineering counterpart of the natural cell membranes, polymeric membranes for the spatial separation of different molecules have not yet reached their high performance but are nevertheless a key tool for the spatial separation of different molecules in biotechnological downstream processes. Most of the membrane separations are performed in an aqueous environment and are based on size differences of the components to be separated. The pore sizes of the membranes vary from micrometers (microfiltration) over pore sizes characterized by the molecular weight cutoff (500 000–1000) of the molecules which no longer pass to the filtrate (ultrafiltration) to pore sizes with a molecular weight cutoff below 1000 (nanofiltration). Tangential flow membrane filtration is a state-of-the-art operation with membranes of selected molecular weight cutoffs and is employed almost everywhere. It is used in batch or continuous mode as a major technique for both product concentration and product purification. Microfiltration is applied, for example, for the concentration of cells, crystals, and precipitates, whereas ultrafiltration is applied for the concentration of high-molecular-weight products and the separation of high-molecular-weight biopolymers from unwanted low-molecular-weight byproducts, media components, and salts.8,22 Nanofiltration is used in the downstream processing for the desalination and concentration of small-molecular-weight compounds23 and reverse osmosis membranes can be used for a subsequent concentration step. A variety of micro-, ultra-, and nanofiltration membranes made of different polymers and pore sizes are in widespread use. Other membrane materials of interest include highly porous, acid- and solvent-resistant ceramic membranes with pore sizes from 1 to >1000 nm. Membrane filtration equipment from small to large scale is available for the selection of the best product recovery yields and minimal adsorption.

2.46.3.10 Other Unit Operations Electroseparation processes such as electrophoresis and electrodialysis are based on charge and mobility differences in an electric field and are orthogonal to the molecular weight-based separation methods; however, they are used in more specialized preparative

686

Product Recovery

applications. Free-flow electrophoresis equipment has been applied for a variety of high-performance separations of cells, proteins, and small molecules over the last decades. Electrodialysis has been a suitable separation method with respect to energy requirements in the large-scale production of amino acids. The use of ion-exchange membranes in electrodialysis-based separation technologies is of much interest for process integration and scale-up.24 The removal of buffers from small molecules at the end of chromatographic separations can be achieved by lyophilization if volatile buffers are used in the preceding chromatographic steps. The best volatile buffer can be selected from various compounds according to the completeness of removal at the desired pH value, temperature, and pressure.

2.46.4

Integrated Unit Operations in Downstream Processing

The scalability, yield per step, and number of unit operations in downstream processing are key factors to the economics of product recovery. The replacement of multiple downstream processing steps by single-stage processes can improve overall operational efficiency by optimizing product quality, space–time yield, timing, and cost issues. In situ product removal (ISPR) techniques remove product from the vicinity of the reaction space thereby preventing its interference.25,26 The integration of product removal unit operations with bioreactions has been a successful way of process intensification for making the success of the bioprocess clearly visible from Pasteur’s biocatalytic tartaric acid resolution process in 1858 until the large-scale industrial bioprocesses today (Figure 3).

2.46.4.1

Solid–Liquid Separation and Product Recovery

The integration of these two unit operations can principally be done in one of two different ways, depending on whether the product is in the solid or liquid phase. In the first case, the solid particles can be the pure product in the form of crystals or the product in association with other components such as inclusion bodies, adsorbers or cell organelles, viruses, and virus-like particles. The liquid phase contains the rest of the impurities, side products, and additional reagents, which have to be separated from the product. In the second case, the liquid can be the pure product itself or the product can be dissolved in the liquid phase; the solid phase contains the rest of the impurities. Expanded-bed adsorption is such an integrated primary downstream process where the solid phase usually contains the product. This technique combines solid–liquid separation with product recovery into a single operation.27

2.46.4.2

Reaction and Solvent Extraction

A simultaneous reaction and extraction system involves organic–aqueous two-phase systems that require the organic solvents used to be compatible with the bioprocess. Such organic–aqueous two-phase systems are attractive for increasing the yield of hydrophobic products toxic to the biocatalysts or cells and that are highly soluble in the organic phase. A scheme that integrates a homogeneous phase consisting of organic–aqueous tunable solvents with a carbon dioxide-induced phase separation allows simultaneous product recovery and recycling of the biocatalyst.28

2.46.4.3

Reaction and Liquid–Liquid Phase Separation

The applications of ATPSs for extractive bioconversions have not yet resulted in wide industrial applications, despite attractive scientific exploitations for many years.17 Further work on reducing the costs of the phase-forming polymers and on reducing the

Membrane filtration

Adsorption

Bioreaction Bioprocess Biocatalysis

Crystal or precipitate formation

Figure 3

Examples of integrated operations in product recovery.

Solvent extraction

Product Recovery

687

complexity of the ATPSs involved will stimulate industrial applications of this research area. The use of two aqueous phases in extractive fermentation has been an attractive approach for overcoming low product yield or for obtaining the product in a cell-free phase. The increases in productivity have been achieved by overcoming existing bottlenecks such as product inhibition and degradation through phase separation of the product as it is formed in the reaction. Both small-molecular-weight and largemolecular-weight extracellular products have been recovered by this methodology. In the case of intracellular products, cell disintegration or cell treatment is required and therefore different integration strategies for product recovery need to be developed. Despite the current lack of translation of research applications into industrial processes, environmentally friendly ATPSs in view of the organic solvent waste contribution of solvent extraction processes have tremendous potential for new integrated bioprocesses.

2.46.4.4

Reaction and Crystallization/Precipitation

The combination of these two unit operations is useful for crystalline product formation by bioprocesses where the product is inhibiting the bioprocess or degrading under process conditions. The removal of the product by crystallization as soon as it is formed in the bioprocess can circumvent these limitations of product inhibition and degradation. As the thermodynamics of the integrated bioprocess can be changed compared with the nonintegrated process, unfavorable equilibriums in the nonintegrated bioprocess can be pulled to completion by the integration of the equilibrium bioprocess with the product crystallization. An example of such an integrated process is the fermentation and enzymatic deacetylation of adipoyl-7-aminodeacetoxy-cephalosporanic acid in one reactor.29 By using the liberated adipic acid again in the fermentation, avoiding the use of acids and bases for pH shifts, a reduced number of downstream processing units, and a reduction in waste salts production, the integrated process leads to economic advantages such as lower manufacturing costs and lower capital investments. The integration of product formation and crystallization has been shown to lead to significant advantages over the nonintegrated case.30

2.46.4.5

Reaction and Adsorption

The integration of the reaction part (biocatalytic reaction, fermentation, or cell culture) with an adsorption operation can serve the purpose of easier product recovery or of removing reaction components that inhibit the reaction at higher concentrations. Such inhibitions prevent the process to be run at higher concentrations and can be caused by substrates, products, and byproducts. Preferential adsorption of fermentation inhibitors from biomass hydrolysates can be used to improve the ethanol yield of the process.31 Adsorbent resins are gaining significant applications in antibiotic and natural product recovery, and the characterization of different adsorbers under the given conditions is important.32 If both substrate and product inhibit the reaction, the adsorption of both substrate and product is a way of keeping the free substrate and product concentration below the inhibitory concentrations, and this substrate feed and product recovery (SFPR) technique has been useful for improving space–time yields. The most suitable adsorbers can thereby be selected by a straightforward experimental determination of substrate and product adsorption as a function of concentration.33 The scale-up and further product recovery of a reaction in an SFPR mode benefit from complete conversion and scalable simple operations at large scale.

2.46.4.6

Reaction and Distillation

Alcoholic beverages with higher ethanol contents such as whiskey, gin, rum, brandy, and vodka have been traditionally recovered by distillation for centuries following fermentation of various grains, fruits, or molasses. The production of biofuels and solvents by fermentation and traditional distillation has been of varying interest since the 19th century, and the improvement of the thermal product recovery techniques after the fermentation with respect to cost reduction, energy reduction, and simple integration is a key factor for the whole production processes. Simple product recovery methods such as gas stripping, condensing the volatile product, and recycling the stripped gas to the fermentation are efficient alternatives. The combination of fermentation or a biocatalytic reaction with the removal of the volatile product via the gas phase can improve conversion yields by reducing product inhibition in the liquid phase. Selective removal of volatile products such as flavors and fragrances is highly interesting; however, the specialized fine distillation technology makes a division between the fermentation/reaction step and the distillation step necessary.

2.46.4.7

Reaction and Chromatography

This coupling is of interest for nonvolatile and temperature-sensitive components and for intensifying bioprocesses with unfavorable equilibriums by product removal and feeding back unreacted substrate to the bioreaction. Annular reactive chromatography has been found less efficient, but convenient for collecting multiple products, whereas simulated moving bed (SMB) reactors are more efficient but allow the separation of only two products.34 A high-fructose syrup process using immobilized glucose isomerase has been improved by simulating continuous countercurrent contact of the liquid stream with the solid adsorbent. Adsorption columns have been advanced against the fixed inlets and outlet of liquid streams without actual movement of the solid adsorbent, while the immobilized enzyme reactors have been stationary.35 Since the coupling adds constraints to the chromatographic separation, a robust product–substrate separation in the presence of additional components from the bioreaction is of key importance. In the biocatalytic condensation of glycine and acetaldehyde to L-allo-threonine, the separation performance with respect to the two amino acids has been shown to be only slightly reduced by coupling the SMB separation to a continuously

688

Product Recovery

operated enzyme membrane reactor (EMR) whose efflux contained, in addition to the amino acids, acetaldehyde and the cofactor pyridoxal-5-phosphate.36

2.46.4.8

Reaction and Membrane Filtration

The combination of a bioreactor with a membrane filtration system can serve different purposes of either mass recycling into the bioreactor or product recovery from the bioreactor.3 A large range of microfiltration membranes for the recycling of biological cells and ultrafiltration membranes for the recycling of large-molecular-weight compounds such as enzymes or large coenzymes are available. The membranes can have a symmetric or asymmetric structure and can thereby have a dual function of immobilizing the biocatalyst and of product separation or purely a product recovery function. In the first case, where the membrane serves as biocatalyst support and separation unit, the biocatalyst can be immobilized within the porous membrane or at the membrane surface by different methods such as covalent attachment, ionic interactions, adsorption, entrapment, gel formation, or crosslinking. A variety of different geometries such as flat-sheet, spiral-wound, and tubular structures are available for the membranes, which can then be assembled into the appropriate membrane modules and units for standardized couplings to the feed, retentate, and permeate lines. The concept of the EMR has been developed and successfully applied by Kula and Wandrey37 and others. The classical work on the enzymatic reductive amination of a-ketoisocaproate to L-leucine with L-leucine dehydrogenase and simultaneous cofactor regeneration with formate dehydrogenase in a continuously operated membrane reactor38 has pioneered this integration mode for a large number of industrial bioprocesses. The enzymatic production of amino acids in membrane reactors with simultaneous regeneration of reduced b-nicotinamide adenine dinucleotide (NADH) has been developed to industrial scale39 at Evonik in Germany. The EMR has been successfully applied in numerous routine productions by other industries (e.g., Tanabe Seiyaku, Sepracor, and Sigma-Aldrich), demonstrating the industrial relevance of this integration of reaction and membrane filtration.1 The membrane reactors have the advantage of using soluble components (enzymes, substrates, and products), which can be easily replenished, for example, for substrate supply or if more enzyme is needed to keep the bioconversion rate constant because of enzyme deactivation effects. Therefore, attention has to be given to the preparation of enzyme so that the operational stability of the free enzyme is good enough for its use in the EMR. Extensions of the classical EMR concept include the charged ultrafiltration membrane enzyme reactor, where negatively charged ultrafiltration membranes are used to retain the native cofactor in the reactor, or the use of nanofiltration membranes. For the bioconversion of poorly soluble substrates, an emulsion membrane reactor consisting of a separate chamber with a hydrophilic ultrafiltration membrane for emulsification, an EMR loop with a normal ultrafiltration module, and a circulation pump can be used.1 Other examples include enantiomerically pure intermediates, anticancer drugs, vitamins, anti-inflammatory compounds, cyclodextrins, and antibiotics.40

2.46.5

Product Purification

Although the ideal downstream processing and purification scheme would be the direct separation, for example, by solid–liquid or another phase separation of the highly pure product, the reality at the end of the biocatalytic process is similar to the work-up in organic chemistry. Many downstream processes and purification operations are unit operations which are robust, well established, and scalable. Therefore, available methods are often chosen in the development of purification methods. Since the selectivity of biocatalytic reactions is high, less side products and auxiliary reagents have to be removed in general, but depending on the number of main products formed and the type of educts and auxiliary compounds used, advanced isolation and purification technologies may be of use. In the case of two very similar products, such as regioisomers, formed in an equimolar ratio, the SMB chromatography is a useful purification method,41 which is also available at large scale.

2.46.5.1

Metabolite Purification

Biological cells represent a valuable source for metabolites, because the chemical synthesis in many cases is not possible or economically viable. As metabolites are small-molecular-weight compounds, many product purification processes from chemistry and biochemistry can be transferred to biotechnology. The wide structural diversity of metabolites requires a large series of purification procedures, which can vary from simple one-step purifications, utilizing one or more unit operations described previously, to the most challenging separations of homologues with very similar molecular properties.10,42 The diversity of closely related metabolites created by biological cells in biocatalytic pathways can be both an opportunity for multiproduct bioprocesses and a challenge for the purification of a single product. Upstream and bioprocess developments can narrow down the metabolite diversity and therefore simplify the purification work of such metabolite mixtures to highly defined and enriched single metabolite products. The production of industrially important metabolites such as organic acids, amino acids, vitamins, and chiral intermediates at large scale focuses on the selection of high-yield mutant strains and avoids the design of extensive separation processes. Rapid and simple purification methods such as pH adjustments, extraction, adsorption, crystallization, and chromatographic separations are in high demand, with crystallization/precipitation and chromatographic separation being the most commonly used techniques. In addition, special cases of liquid metabolite mixtures can be checked for purification opportunities by fine distillation.

Product Recovery 2.46.5.2

689

Lipid Purification

The large-scale processing of fats, oils, and lipids involves rendering, screw-pressing, expelling, or solvent extraction-based methods compatible with the applications. Separation and isolation procedures for lipids include crystallization, urea fractionation, distillation, enzymatic procedures, and liquid chromatography.43 Adsorption, partition, ion-exchange, and supercritical fluid chromatography are versatile and useful tools for fractionating lipid mixtures on a preparative scale. Sephadex LH-20 and Sephadex G-25 have been commonly used for chromatography. Preparative high-performance liquid chromatography (HPLC) and supercritical fluid chromatography (SFC) have also been successfully established lipid purification techniques.

2.46.5.3

Protein Purification

Although the isolation and purification of proteins have been established long ago, today‘s challenges require faster ways to purify more proteins. More than a century of practice and know-how in protein purification have resulted in a great variety of methods and techniques,44 each with its own advantages, disadvantages, and limitations. The number of unit operations involved depends on the demands on purity and safety of the final product. The development of efficient and simple protein purification sequences is a major bottleneck and makes use of the best methods. Miniaturized systems for the major unit operations and for screening relevant downstream process parameters have been developed in order to reduce the development time significantly. The fundamental investigation of basic purification processes is crucial and can lead to extremely novel applications reducing this major part of production costs. As the preservation of fully functional proteins during protein purification and high yields are the goals, methods for avoiding protein degradation or modification have become important. However, in certain cases, the purification of inclusion bodies and the subsequent in vitro refolding into fully functional proteins can be the purification method of choice.

2.46.5.4

Nucleic Acid Purification

A variety of well-known laboratory methods for DNA and RNA purification are traditional molecular biology procedures. The transfer to large scale of many laboratory methods for DNA purification such as density gradient centrifugation with cesium salts, ion-exchange chromatography, and reversed-phase chromatography can, however, be difficult due to the use of toxic and mutagenic reagents, kits, or time-consuming procedures.45 Large-scale processes using only generally recognized as safe reagents and scalable methodologies such as precipitation, salting-out, and chromatography (ion exchange, hydrophobic interaction, and size exclusion) are available for the large-scale purification of DNA. Nonchromatographic procedures based on selective precipitation with cetyltrimethylammonium bromide46 or aqueous two-phase systems containing polyethylene glycol with potassium citrate or potassium phosphate have allowed the fast and simple recovery of plasmid DNA. The complete sequence of lysis, precipitation, clarification, and extraction can be performed in a single vessel.47 Suitable large-scale methods for RNA employ overproduction and purification of recombinant RNA48 and polyacrylamide-free size-exclusion chromatography.49

2.46.5.5

Carbohydrate Biopolymer Purification

The carbohydrate biopolymers are widely distributed in animals, plants, seaweeds, algae, mushrooms, fungi, yeasts, and bacteria, where they have different functions such as nutritional reserve or structure-forming compounds. Many polysaccharides are formed extracellularly in larger amounts, are water-soluble, and have a long history of applications in industry, medicine, and our daily life. As the carbohydrate biopolymers show the greatest structural diversity of all biopolymers, it is also clear that the purification and analysis pose the biggest challenges. Depending on the localization in the biological cells, the carbohydrate biopolymers may have to be deproteinized without destroying the carbohydrate structure. Carbohydrate biopolymers with cross-links, consisting of glycosidic bonds or peptide units, are insoluble in aqueous media and require selective removal of byproducts. Economical production of polysaccharides is usually achieved at very large scale. Soluble carbohydrate biopolymers have the advantage of being extractable, but a variety of related products are usually extracted as well. Fractionation is based on selective precipitation and solubilization, but in contrast to small molecules the increased viscosity even at low product concentrations is one of the main factors influencing not only purification but also formation and isolation of polysaccharides above a certain molecular weight. Different physicochemical parameters such as ionic strength, pH, cations, and viscosity, even at very low levels, can have a substantial influence on the purification scheme. The fractionation of structurally diverse carbohydrate polymers as well as the separation of different molecular weight fractions consisting of the same building blocks (homopolysaccharide) or of different building blocks (heteropolysaccharide) requires the whole toolbox of unit operations and case-specific troubleshooting, until finally purified carbohydrate biopolymers are obtained.50

2.46.6

Product Formulation and Stabilization

In our interdependent economy, it is crucial that bioproducts can be transported and stored worldwide in a biologically active state from the production site to the application site. The properties of the product and its application, on the one hand, and the minimum shelf-life requirements for products, which are kept permanently under controlled storage or transportation conditions, on the other hand, set the boundary conditions for the formulation and stabilization of the final product.

690

Product Recovery

The storage stability of products is important for research, biomedical, and industrial applications and is influenced by many of the preceding process steps in production. The last operations are, however, often the most important ones,5 because these are determining the form, standardized quality, and impurities of the product. A pure powder or liquid is the preferred form of the product; however, criteria such as ease of application, production costs, or stability issues often make other formulations such as dried powders as products, immobilized or stabilized solids, and buffered and stabilized suspensions or solutions the formulation of choice. The actual best practices for the formulation of stable products focus on preventing the relevant molecular mechanisms of the product degradation pathway. When the molecular structure of the product is given, the formulation design of the product can protect against degradation by microbial, chemical, osmotic, pH, and oxidative stress. Changes in molecular product integrity as well as more physical changes in solubility, adsorption, and aggregation can be prevented by the addition of protecting agents such as buffers, anti-oxidants, cryo- or lyoprotectants, and osmolytes. Although much experience has been gained for determining the best formulation and method to maintain and store, stabilize and homogenize, and apply and deliver bioproducts under given conditions in an active state, predictions have to be experimentally tested for each new product. Process analytical technologies that are capable of recognizing changes in the activity and molecular integrity of the product are thereby of prime importance. Biocompatible additives such as glycerol, polyethylene glycol, carbohydrates, or amino acids can be useful for optimizing stability and production issues of the product, but have to be checked for their compatibility with the intended applications. Since pH or oxygen can influence product stability tremendously, the addition of pH or redox buffers is often required in the quality design in order to have a robust production process. The operational stability of products under the conditions of research, biomedical or industrial applications, is of equal importance, and strategies for product stabilization are not restricted to the ones already mentioned for storage stability, but can aim at a tailor-made product design.

2.46.7

Conclusion

The future directions of product recovery technology are influenced by numerous scientific and technological, industrial, economic and commercial as well as environmental and social developments. Nevertheless, it is quite clear that process improvements in product recovery with respect to waste minimization, volatile organic solvent reduction, energy conservation, safety, health, and environment issues are beneficial not only for the ecoefficiency of a product recovery scheme but also for economical aspects such as production cost reductions, because the costs of product recovery are substantial determinants of the total production costs. In addition, these process improvements can lead to additional macro- and microeconomic benefits on the local and global levels. The waste minimization in the recovery of a particular bioproduct depends both on technology changes, for example, replacing an extraction step with organic solvents by crystallization, and on finding new opportunities for useful applications of byproducts and thereby turning these from waste into products. Besides the engineering aspects of waste minimization, there is also the key molecular aspect of selectivity and the parameters influencing it in all product recovery operations, which are important for optimizing the recovery scheme to a defined purity of the product of interest. Therefore, innovative downstream processing taking into account both the molecular and engineering aspects continues to be of great importance for the whole bioprocess. With the changing costs of energy and the limited resources of certain fossil energy, it becomes not only a cost-saving exercise to minimize the energy input into product recovery bioprocesses but also an opportunity to develop new efficient product recovery processes for the economic production of biofuels for cars and jet planes. The introduction of renewable resources into product recovery processes can occur in various contexts and offer, in addition, great opportunities as raw materials in bioprocesses. Many renewable resources can be utilized for the production of auxiliaries for product recovery such as cellulose and dextran, which have traditionally been used in the form of functionalized cellulose ion exchangers (e.g., DEAE-cellulose), cross-linked agarose and dextran (e.g., Sepharose and Sephadex) for chromatography, or as dextran components in ATPSs. As the carbon dioxide and volatile organic carbon compounds in the Earth‘s atmosphere need to be reduced, product recovery processes can be transformed into more sustainable processes that reduce the use of volatile organic solvents or reduce carbon waste burned to carbon dioxide.51–53

References 1. Ghisalba, O.; Meyer, H. P.; Wohlgemuth, R. Industrial Biotransformation. In Encyclopedia of Industrial Biotechnology; Flickinger, M. C., Ed.; 5; Wiley: Hoboken, NJ, 2010; pp 2971–2988. 2. Wohlgemuth, R. The Locks and Keys to Industrial Biotechnology. New Biotechnol. 2009, 25, 204–213. 3. Bailey, J. E.; Ollis, D. F. Biochemical Engineering Fundamentals, 2nd ed.; McGraw-Hill: Singapore, 1986. 4. Demain, A. L., Solomon, N. A., Eds.; Manual of Industrial Microbiology and Biotechnology, American Society for Microbiology: Washington, DC, 1986. 5. Präve, P.; Faust, U.; Sittig, W.; et al. Handbuch der Biotechnologie, 2. Auflage; Oldenbourg Verlag: München, 1984. 6. McCabe, W. L.; Smith, J. C.; Harriott, P. Unit Operations of Chemical Engineering, 7th ed.; McGraw-Hill: New York, 2005. 7. Roush, D. J.; Lu, Y. Advances in Primary Recovery: Centrifugation and Membrane Technology. Biotechnol. Prog. 2008, 24, 488–495. 8. Van Reis, R.; Zydney, A. Membrane Separations in Biotechnology. Curr. Opin. Biotechnol. 2001, 12, 208–211. 9. Balasundaram, B.; Harrison, S.; Bracewell, D. G. Advances in Product Release Strategies and Impact on Bioprocess Design. Trends Biotechnol. 2009, 27, 477–485. 10. Schügerl, K. Solvent Extraction in Biotechnology – Recovery of Primary and Secondary Metabolites, Springer-Verlag: Berlin, 1994. 11. Eggeman, T.; Verser, D. Recovery of Organic Acids From Fermentation Broths. Appl. Biochem. Biotechnol. 2005, 121, 605–618.

Product Recovery 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53.

691

Kertes, A. S.; King, C. J. Extraction Chemistry of Fermentation Product Carboxylic Acids. Biotechnol. Bioeng. 1986, 28, 269–282. Yabannavar, V. M.; Wang, D. I. C. Extractive Fermentation for Lactic Acid Production. Biotechnol. Bioeng. 1991, 37, 1095–1100. Kertes, A. S.; King, C. J. Extraction Chemistry of Low Molecular Weight Aliphatic Alcohols. Chem. Rev. 1987, 87, 687–710. Roffler, S.; Blanch, H. W.; Wilke, C. R. In Situ Extractive Fermentation of Acetone and Butanol. Biotechnol. Bioeng. 1988, 31, 135–143. Azevedo, A. M.; Rosa, P. A. J.; Ferreira, I. F.; et al. Chromatography-free Recovery of Biopharmaceuticals Through Aqueous Two-phase Partitioning. Trends Biotechnol. 2009, 27, 240–247. Rito-Palomares, M. Practical Applications of Aqueous Two-phase Partition to Process Development for the Recovery of Biological Products. J. Chromatogr. B 2004, 807, 3–11. Barrett, P.; Smith, B.; Worlitschek, J.; et al. A Review of the Use of Process Analytical Technology for the Understanding and Optimization of Production Batch Crystallization Processes. Org. Process Res. Dev. 2005, 9, 348–355. Garcia, A. A. Strategies for the Recovery of Chemicals From Fermentation: A Review of the Use of Polymeric Adsorbents. Biotechnol. Prog. 2008, 7, 33–42. Susanto, A.; Treier, K.; Knieps-Grünhagen, E.; et al. High-throughput Screening for the Design and Optimization of Chromatographic Processes: Automated Optimization of Chromatographic Phase Systems. Chem. Eng. Technol. 2009, 32, 140–154. Thömmes, J.; Etzel, M. Alternatives to Chromatographic Separations. Biotechnol. Prog. 2007, 23, 42–45. Van Reis, R.; Zydney, A. Bioprocess Membrane Technology. J. Membr. Sci. 2007, 297, 16–50. Dudziak, G.; Fey, S.; Hasbach, L.; Kragi, U. Nanofiltration for Purification of Nucleotide Sugars. J. Carbohydr. Chem. 1999, 18, 41–49. Xu, T.; Huang, C. Electrodialysis-based Separation Technologies: A Review. AIChE J. 2008, 54, 3147–3159. Freeman, A.; Woodley, J. M.; Lilly, M. D. In Situ Product Removal as a Tool for Bioprocessing. Biotechnology 1993, 11, 1007–1012. Stark, D.; von Stockar, U. In Situ Product Removal (ISPR) in Whole Cell Biotechnology During the Last Twenty Years. Adv. Biochem. Eng. Biotechnol. 2003, 80, 149–175. Hubbuch, J.; Thömmes, J.; Kula, M. R. Biochemical Engineering Aspects of Expanded Bed Adsorption. Adv. Biochem. Eng. Biotechnol. 2005, 92, 101–123. Broering, J. M.; Hill, E. M.; Hallett, J. P.; et al. Biocatalytic Reaction and Recycling by Using CO2-induced Organic–aqueous Tunable Solvents. Angew. Chem. 2006, 45, 4670–4673. Roa Engel, C. A.; Straathof, A. J. J.; van Gulik, W. M.; et al. Conceptual Process Design of Integrated Fermentation, Deacetylation and Crystallization in the Production of Beta-lactam Antibiotics. Ind. Eng. Chem. Res. 2009, 48, 4352–4364. Buque-Taboada, E. M.; Straathof, A. J. J.; Heijnen, J. J.; et al. In Situ Product Recovery (ISPR) by Crystallization: Basic Principles, Design and Potential Applications in Whole-cell Biocatalysis. Appl. Microbiol. Biotechnol. 2006, 71, 1–12. Ranjan, R.; Thust, S.; Gounaris, C. E.; et al. Adsorption of Fermentation Inhibitors From Lignocellulosic Biomass Hydrolyzates for Improved Ethanol Yield and Value-added Product Recovery. Microporous Mesoporous Mater. 2009, 122, 143–148. Casey, J. T.; Walsh, P. K.; O‘Shea, D. G. Characterization of Adsorbent Resins for the Recovery of Geldanamycin From Fermentation Broth. Separ. Purif. Technol. 2007, 53, 281–288. Hilker, I.; Alphand, V.; Wohlgemuth, R.; et al. Microbial Transformations, 56. Preparative Scale Asymmetric Baeyer–Villiger Oxidation Using a Highly Productive Two-in-one Resin-based in Situ SFPR Concept. Adv. Synth. Catal. 2004, 346, 203–214. Ströhlein, G.; Mazzotti, M.; Morbidelli, M. Simulated Moving-bed Reactors. In Integrated Chemical Processes; Sundmacher, K., Kienle, A., Seidel-Morgenstern, A., Eds., Wiley-VCH: Weinheim, 2005; pp 83–202. Hashimoto, K.; Adachi, S.; Noujima, S.; et al. A New Process Combining Adsorption and Enzyme Reaction for Producing Higher-fructose Syrup. Biotechnol. Bioeng. 1983, 25, 2371–2393. Makart, S.; Bechthold, M.; Panke, S. Separation of Amino Acids by Simulated Moving Bed under Solvent Constrained Conditions for the Integration of Continuous Chromatography and Biotransformation. Chem. Eng. Sci. 2008, 63, 5347–5355. Kula, M. R.; Wandrey, C. Continuous Enzymatic Transformation in an Enzyme-membrane Reactor With Simultaneous NADH Regeneration. Meth. Enzymol. 1987, 136, 9–21. Wichmann, R.; Wandrey, C.; Bückmann, A. F.; et al. Continuous Enzymatic Transformation in an Enzyme Membrane Reactor With Simultaneous NAD(H) Regeneration. Biotechnol. Bioeng. 1981, 23, 2789–2802. Wandrey, C. Biochemical Reaction Engineering for Redox Reactions. Chem. Rec. 2004, 4, 254–265. Drioli, E., Giorno, L., Eds.; Membrane Operations, Wiley-VCH: Weinheim, 2009. Kaiser, P.; Ottolina, G.; Carrea, G.; et al. Preparative-scale Separation by Simulated Moving Bed Chromatography of Biocatalytically Produced Regioisomeric Lactones. New Biotechnol. 2009, 26, 216–221. Wohlgemuth, R. Tools and Ingredients for the Biocatalytic Synthesis of Metabolites. Biotechnol. J. 2009, 9, 1253–1265. Gunstone, F. D.; Harwood, J. L.; Padley, F. B. The Lipid Handbook, 2nd ed.; Chapman & Hall: London, 1994. Janson, J. C., Rydén, L., Eds.; Protein Purification, 2nd ed.; John Wiley & Sons: New York, 1998. Ferreira, G. N. M.; Monteiro, G. A.; Prazerers, D. M. F.; et al. Downstream Processing of Plasmid DNA for Gene Therapy and DNA Vaccine Applications. Trends Biotechnol. 2000, 18, 380–388. Murphy, J. C.; Winters, M. A.; Sagar, S. L. Large-scale Nonchromatographic Purification of Plasmid DNA. Meth. Mol. Med. 2006, 127, 351–362. Frerix, A.; Müller, M.; Kula, M. R.; et al. Scalable Recovery of Plasmid DNA Based on Aqueous Two-phase Separation. Biotechnol. Appl. Biochem. 2005, 42, 57–66. Ponchon, L.; Beauvais, G.; Nonin-Lecomte, S.; et al. A Generic Protocol for the Expression and Purification of Recombinant RNA in Escherichia coli Using tRNA Scaffold. Nat. Protoc. 2009, 4, 947–959. Lukavsky, P. J.; Puglisi, J. D. Large-scale Preparation and Purification of Polyacrylamide-free RNA Oligonucleotides. RNA 2004, 10, 889–893. Kamerling, J. P.; Boons, G. J.; Lee, Y. C.; et al. Comprehensive Glycoscience; Vols. 1–4; Elsevier Ltd: Oxford, 2007. Hilker, I.; Wohlgemuth, R.; Alphand, V.; et al. Microbial Transformations 59: First Kilogram Scale Asymmetric Microbial Baeyer–villiger Oxidation With Optimized Productivity Using a Resin-based in Situ SFPR Strategy. Biotechnol. Bioeng. 2005, 92, 702–710. Schügerl, K.; Hubbuch, J. Integrated Bioprocesses. Curr. Opin. Microbiol. 2005, 8, 294–300. Straathof, A. J. J. Auxiliary Phase Guidelines for Microbial Biotransformations of Toxic Substrate into Toxic Product. Biotechnol. Prog. 2003, 19, 755–762.

2.47

Cell Disruption

STL Harrison, University of Cape Town, Rondebosch, South Africa © 2011 Elsevier B.V. All rights reserved. This is a reprint of S.T.L. Harrison, 2.44 - Cell Disruption, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 619-640.

2.47.1 2.47.2 2.47.2.1 2.47.2.2 2.47.2.3 2.47.2.4 2.47.3 2.47.4 2.47.4.1 2.47.4.2 2.47.5 2.47.5.1 2.47.5.2 2.47.5.2.1 2.47.5.2.2 2.47.5.3 2.47.5.4 2.47.5.5 2.47.5.6 2.47.6 2.47.7 2.47.8 2.47.9 2.47.10 References

Introduction Characteristics of the Microbial Cell Influencing Resistance to Disruption Bacterial Cell Envelope Yeast Cell Walls The Cell Envelope in Archaea Algal Cell Walls Approaches to Microbial Cell Disruption Large-Scale Cell Disruption Technologies High-Pressure Homogenization High-Speed Bead Mills Laboratory-Scale, and Developing, Cell Disruption Technologies Laboratory-Scale Mechanical Disruption Cavitation Ultrasonic Cavitation Hydrodynamic Cavitation Chemical Treatment Enzymatic Attack Osmotic Treatment Thermal Treatment Selective Product Release Pretreatment to Augment Product Release Integration of Biomass Formation and Product Release Integration of Product Release and Product Recovery and Purification Closing Remarks

692 693 693 694 696 696 696 697 697 701 702 702 702 702 703 703 704 704 705 705 706 707 709 710 710

Glossary Cavitation Formation of vapor cavities in the liquid phase at regions of reduced pressure where localized pressure is less than the vapor pressure of the liquid, or approaches it in the presence of dissolved gases. High-pressure homogenization (HPH) Use of rapid change in pressure following flow through a constriction to mediate cell disruption or dispersion of particles or droplets. High-speed bead mill Rapid agitation of a grinding medium, typically small glass beads, to facilitate microbial cell disruption or break up of particles. Pretreatment Exposure of microbial cells to a cell weakening treatment prior to mechanical disruption. Selective product release (SPR) The ratio of the product released to the contaminants released on a mass basis. Selectivity (S) The ratio of the product released to the total soluble protein released on a mass basis. Ultrasound Sound at frequency higher than 15–20 kHz.

2.47.1

Introduction

Microorganisms form a rich and diverse source of molecules, including chemical entities, biologic-based medicines, nutraceuticals, enzymes, fine chemicals, and commodity products, as well as a source of biomass as raw material to the commodity chemical and energy product sector. Biologically based pharmaceuticals, healthcare products, and nutraceuticals are of increasing importance, owing to their ability to extend the product framework and the efficiency of biosystems, over chemical synthesis, in producing complex molecules with chiral specificity. On the other hand, microbial biomass is sought as raw material for energy products, commodities, and single cell protein, while also playing a key role in the conversion of biomass resources to energy products.

692

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00134-8

Cell Disruption

693

Although microbial cultures may export a range of useful molecules into the suspending medium, the majority of the resources for these applications is intracellular. For heterologous protein production by recombinant microorganisms, the common host organisms Escherichia coli and Saccharomyces cerevisiae are recognized as poor exporters of proteins from the cell such that high titers of the product proteins in the extracellular environment are seldom achieved.12 Although this situation can be forced through genetic manipulation, it requires a long research lead time. Alternative host organisms such as Bacillus subtilis and Pichia pastoris have been demonstrated for excretion of the recombinant products. Specific modifications to enhance the extracellular location have been developed. For example, co-expression of the recombinant proteins with the ‘kil’ gene can be used to provide products into the suspending medium as the kil protein induces cell lysis. A recombinant Bacillus halodurans system has been developed to produce extracellular peptide products in place of the flagellum construct.30 Advantages of these product secretion systems are balanced by disadvantages including the increased exposure of the products to proteases. The extracellular environment in the bioreactor is not designed to maximize protein product stability or activity, and temperature and shear may contribute to protein degradation. Products may become entrapped in the periplasm or the cell wall. Hence, while specific products are designed for extracellular production, this is not the dominant mode of production of microbial products. For both wild-type and many recombinant production systems, cell disruption constitutes a key unit operation in realizing maximum potential from the microbial resource pool.2,6,12 Copious research and development in this field took place in the 1970s with many of the industrially used technologies stemming from that period. However, developments in the bioprocess sector have presented challenges that require further development of the cell disruption operations. Specifically, the ability to provide selective product release (SPR), to minimize the energy requirement of the cell disruption operation and to manipulate the characteristics of the disrupted cell suspension produced to optimize subsequent product recovery and purification are among the current challenges for ongoing development in intracellular product liberation. The need for these developments is intensified by the recent successes in maximizing product titers in the bioreactor, specifically with respect to biopharmaceutical products as well as increasing emphasis on white bioprocesses for renewable commodity and energy products. The application of the cell disruption step can be grouped into two major types of applications:

• •

the recovery and purification of high-value specialty products and the release of commodity products for purification or further processing.

The above-mentioned application groups provide different challenges. In the former, product liability and purity are of the essence. In the latter, product recovery and yield must be maximized while minimizing cost and energy inputs. Hence, application of the unit operation informs its selection and optimization. Cell disruption as a unit operation is closely linked to both the upstream and downstream processes. It is well recognized that the upstream biomass production process influences both the equipment selection and the breakage operation. In terms of the former, growth conditions influence the biomass and product concentration, microbial morphology, and suspension properties. In terms of the latter, resilience to disruption through varying microbial cell envelope strength is influenced by growth conditions as well as culture history.2,6 Further, the nature of the disruption step influences the introduction of contaminating compounds into the product-containing suspension, the release of intracellular contaminating compounds, the nature of the cell debris, and the physical characteristics of the cell suspension or cell lysate. In this chapter, the state of the art in microbial cell disruption is reviewed in terms of both large-scale industrially relevant processes and small-scale, developing or laboratory-based processes. Empirical performance of these unit operations and factors influencing this is described. Where possible, the mechanism of disruption is reviewed. To inform this, the nature of the microbial cell-wall structure is described. Thereafter, the challenges to be overcome for the improvement of disruption process are raised. Specifically, SPR and the integration of up- and downstream processes with the cell disruption operation are highlighted.

2.47.2

Characteristics of the Microbial Cell Influencing Resistance to Disruption

For an intracellular microbial product to be recovered from the microbial cell, the cell envelope must be damaged sufficiently for product release. For the design of cell disruption systems and the prediction of their performance, it is necessary to understand the location of the product relative to the cell envelope, the effect of fluid forces on the cell wall, physicochemical and enzymic challenges to the cell wall, as well as the material properties of the cell wall. The latter is considered in this section. Microbial cell envelopes are generally made up of one or more membranes and a structural cell wall. The cytoplasmic membrane forms the biological boundary to the cell, controlling concentration gradients between the intracellular and external environments. The cell wall is typically a cross-linked polysaccharide structure that provides structural strength.6 In addition to maintenance of cell shape and its protection, the cell wall stabilizes the osmotic pressure within the cell and governs the permeability of the cell with respect to macromolecules. On recovering microbial products, complete breakage of the cell envelope is not required for periplasmic or wall-associated products. Where the product is cytoplasmic, the cell envelope must be ruptured. For soluble cytoplasmic structures, point breaks suffice whereas the release of granular inclusions or large molecules such as DNA requires complete disruption of the cell structure. Where the product is located in an organelle, disruption of these substructures is also required.

2.47.2.1

Bacterial Cell Envelope

Bacteria fall into two categories based on their cell envelopes. Gram-positive bacteria (e.g., B. subtilis) consist of a dominant peptidoglycan cell wall as the outermost structure and a cytoplasmic phospholipid bilayer membrane. In Gram-negative bacteria

694

Cell Disruption

(e.g., E. coli), the peptidoglycan wall is less dominant and both an outer membrane and cytoplasmic membrane are present. These categories are readily identified through Gram staining and are illustrated in Figure 1. The outer membrane is a lipid bilayer comprised of phospholipid and lipopolysaccharide with embedded transmembrane proteins.6,12 Noncovalent bonding involving Ca2þ and Mg2þ ions stabilizes the lipopolysaccharide molecules. Lipoproteins substituted into the phospholipid layer connect the outer membrane to the peptidoglycan layer. The peptidoglycan wall provides the structural strength to the cell. Its basic structure is similar across all bacteria. Peptidoglycan is made up of parallel linear polysaccharides of alternating N-acetyl-muramic acid and N-acetyl-D-glucosamine monomers, linked through b-1,4-glycosidic bonds. These are cross-linked through tetrapeptide side chains between which peptide bonds form. The resultant structure acts like a macromolecular grid structure, providing shape, tensile strength, and protection against osmotic variation. The tensile strength is governed by the frequency of peptide cross-linking. In Gram-negative bacteria, the peptidoglycan layer typically varies from a single layer of some 1.5 nm to about five layers of about 9 nm thickness, but this is debated and variable. It is estimated to comprise 10–20% of the cell-wall mass. In Gram-positive bacteria, peptidoglycan accounts for 50–80% of the cell wall by mass and is further strengthened by associated teichoic acids.6,12 The cell membrane is predominantly a phospholipid bilayer of 4 nm thickness, with embedded membrane proteins. This structure is key to the maintenance of the intracellular environment, maintaining concentration gradients, housing the transport systems, and providing the seat of bacterial adenosine triphosphate generation. It is easily ruptured by osmotic pressures in the absence of the structural wall and offers little resistance to mechanical cell disruption.

2.47.2.2

Yeast Cell Walls

Although it is recognized that fungal cell walls are diverse and differences are found in yeast cell-wall structures (e.g., between Schizosaccharomyces pombe and S. cerevisiae), there is strong evidence for similarity between the cell-wall structure of S. cerevisiae and that of a number of yeasts including Candida albicans, Kluveromyces lactis, and Zygosaccharomyces rouxii. To this end, the cell envelope of S. cerevisiae is described here as an example of a commercially important yeast. This is further justified by the key role of S. cerevisiae in traditional biotechnology, its common use as a recombinant host organism, and its rigorous characterization. The yeast cell wall is located externally to the phospholipid bilayer-dominated cytoplasmic membrane. It provides the major physical protection through mechanical strength and elasticity and enables the cell to maintain its shape. The cell wall creates a framework for wall-associated proteins that control the cell-wall permeability and provide a microenvironment external to the cell membrane. It also stabilizes internal osmotic conditions through providing a counter-pressure to water influx. Owing to its importance in disruption, description of the wall forms the core of this section. The cell wall of S. cerevisiae, comprising some 30% of the yeast dry mass, consists mostly of polysaccharides (85%) and proteins (15%). Its composition and structural thickness vary as a function of the environmental conditions and the period of the cell cycle. Typical thickness of 70–100 nm is reported for laboratory yeast strains while cell walls of brewing yeast of 200 nm have been reported.10 The composition, structure, and assembly of the yeast cell wall are estimated to be controlled by 1200 genes.11 The primary components of the yeast cell wall are detailed in Table 1, and its structure is shown diagrammatically in Figure 2. The Lipotechoic acid A

Techoic acid Peptidoglycan Periplasmic space Cytoplasmic membrane Membrane protein Lipopolysaccharide B

Porin Outer membrane Lipoprotein Peptidoglycan Periplasmic space Cytoplasmic membrane Membrane protein Figure 1 Structure of the cell wall of (A) Gram-positive bacteria and (B) Gram-negative bacteria. Reproduced from Prescott LM, Harley J, and Klein DA (1999). Microbiology, 4th edn. Boston, MA: McGraw-Hill.

Cell Disruption Table 1

695

Major components of the cell wall of Saccharomyces cerevisiae and their properties

Cell wall component

Location in the cell wall

Site of synthesis

DP

Branching

Mean MM (kDa)

Fraction of cell wall by mass

b-1,3-Glucan b-1,6-Glucan Manno-protein

Inner wall Inner/outer wall Outer wall

1500 140–150 200

Moderate branching Highly branched Highly branched

240 24 Varies

0.30–0.55 0.05–0.10 0.30–0.50

Chitin

Bud scars, low amount in lateral cell wall

CM CM Secretory pathway CM

120–190

Linear

25

0.01–0.06

CM, cytoplasmic membrane; DP, degree of polymerization; MM, molecular mass. Modified from Klis FM, Mol P, Hellingwerf K, and Brul S (2002). Dynamics of cell wall structure in Saccharomyces cerevisiae. FEMS Microbiology Reviews 26: 239–256 and Klis FM, Boorsma A, and de Groot PWJ (2006). Cell wall construction in Saccharomyces cerevisiae. Yeast 23: 185–202.

Mannoprotein β-1,6-Glucan β-1,3-Glucan Chitin Periplasmic space Cytoplasmic membrane Figure 2 Structure of the cell wall of the yeast Saccharomyces cerevisiae. Reproduced from Klis FM, Boorsma A, and de Groot PWJ (2006). Cell wall construction in Saccharomyces cerevisiae. Yeast 23: 185–202 and Prescott LM, Harley J, and Klein DA (1999). Microbiology, 4th edn. Boston, MA: McGraw-Hill.

main mechanical strength of the wall is imparted by the H-bonded b-1,3-glucan component, comprising the inner cell wall. b-1,3Glucan forms a moderately branched structure in which the polysaccharide chains take on a helical structure of varying extension, giving the cell wall its elasticity and ability to alter cell volume in response to osmotic stress. The branching, achieved through some 3–4% of bonds being b-1,6 linkages, prevents crystallization. Late in the cell cycle, the wall is further strengthened by the addition of chitin to the matrix. Glycosylated mannoproteins are the dominant components of the outer cell wall and are comprised of some 90% carbohydrate. The mannoprotein layer is less permeable than the b-1,3-glucan layer. Carbohydrate side chains with phosphodiester bridges cause a negative charge at physiological pH. Covalent disulfide bridges are found. Clustering of serine and threonine residues causes steric interaction and rigid regions. This limits accessibility to the inner wall and membrane, protecting these from physicochemical and biological attack.10 A further role of the mannoprotein region is cell–cell interaction. The mannoprotein is linked to the b-1,3-glucan layer through b-1,6-glucan, an amorphous polymer of glucose comprising about 10% of the glucan in the wall. Chitin is a linear polymer of b-1,4-linked N-acetylglucosamine monomers, visible under fluorescence using Calcofluor white. Most chitin is present in the chitin ring at the bud neck of the bud scars formed on cell division and in the septum. The chitin content of the lateral cell wall is typically only 0.1–0.2% in the late cell cycle, showing that it is not essential for cell strength.11 In response to perturbation, chitin synthesis may be induced, resulting in as much as 20% chitin by mass as a salvage mechanism to enhance cell strength of weakened cell walls.10 The mechanical disruption of microbial cells is dependent on both the fluid flow environment and the mechanical properties of the cell. The latter have been investigated for yeast cells using micro-manipulation with compression of the yeast cell between two flat plates to generate force-deformation behavior.80,81 From this, cell-wall properties can be estimated using simplifying assumptions, for example, homogeneity of cell-wall structure, incompressibility. These analyses have been undertaken accounting for cell permeability, and hence volume changes in response to forces imposed. The bursting force and deformation at bursting are interrelated and influenced by osmotic pressure. By developing this understanding of cell-wall strength, it is expected to inform improved approaches to mechanical cell disruption.

696 2.47.2.3

Cell Disruption The Cell Envelope in Archaea

The Archaea are recognized as a diverse group of prokaryotes, typically found in extreme environments and gaining importance in biotechnology owing to their potential product range for operation under such extreme conditions (e.g., high- or low-temperature enzymes and salt tolerance). One of the differentiators of the Archaea from bacteria is their cell envelope structure.14,86 The cell membrane, while meeting the physical requirements of the phospholipid bilayer, differs molecularly. The basic structure is formed from ether linkages, yielding L-glycerol di-ether or di-L-glycerol tetraether molecules, not ester bonds, in which the hydrophobic chains are comprised of isoprenoid chains, allowing a branched structure.58 Peptidoglycan is typically absent from the archaeal cell walls (with few exceptions, including the methanogens containing pseudomurein and Mycoplasma). In the Gram-positive Archaea, a variety of polymers may provide the structural wall. In Gram-negative Archaea, the wall is usually limited to a proteinaceous or glycol–proteinaceous layer of 20–40 nm thickness as found in Sulfolobus, Thermoproteus, Methanococcus, and Desulfurococcus. No outer membrane is present. A variety of cell-wall structures are reported, including a glycoprotein structure, surface layer protein S-layers or pseudopeptidoglycan. In some cases, the cell wall is absent. This considerable diversity in the cell envelope structure is associated with variable resistance to cell disruption.

2.47.2.4

Algal Cell Walls

Similarly, algal cell envelopes show substantial diversity, with the wall being absent in some species (e.g., Dunaliella). Most algal cell walls contain glycoproteins (e.g., Chlamydomonas reinhardii and the Volvocales) or polysaccharides or both. Cellulose, the most common polysaccharide in algal walls, shows varying degrees on crystallinity associated with varying resistance to breakage. The green algae also contain mannans and xylans, while the brown algal walls contain alginic acid. Chlorella species are commonly reported in bioprocess applications and have been shown to have a cell wall containing some 27% protein, 9% lipid, 15% a-cellulose, 31% hemicellulose, and making up 13% of the cell mass.69 Diatoms have a hard outer wall composed of silicic acid (H4SiO4). This requires less energy for synthesis than the typical polysaccharide wall. As with the Archaea, this diversity in the cell envelope structure is associated with variable resistance to cell disruption. Further, the nutrient availability, especially that of N and P, may affect strength of the wall with nutrient-starved algae showing greater resilience.

2.47.3

Approaches to Microbial Cell Disruption

Unit operations for microbial cell disruption are broadly classified into mechanical and nonmechanical techniques, as illustrated in Figure 3. The mechanical operations of bead milling and high-pressure homogenization (HPH) are most commonly used on the large scale,2,6,12 owing to their rapid handling of cell suspensions, generic application across cell types, and ease of scalability. These operations exploit fluid flow, particle–particle interaction, and pressure drop to affect cell disruption and have been modified from equipment used for particle or droplet size reduction in other industries. They are energy-intensive approaches that result in non-SPR through complete rupture of cells. Most of the energy is dissipated as heat, so good heat management through efficient cooling systems is essential to prevent protein inactivation. Shear-sensitive products, such as DNA, are not typically recovered through these mechanical approaches. Cell disruption operations

Mechanical approaches

Forces acting in suspension

Forces acting through solid–solid interaction

Physical

Chemical

Biological

Temperature extremes

pH extremes

Add external cell wall-lytic enzymes Autolysis

Detergents

HPH

Bead mill

Osmotic shock

Solvents

Impingement

Mortar and pestle

Dessication

Antibiotics

Gas decompression Sonication

Chelating agents

Hydrodynamic cavitation

Bead beating

Mechanical agitation Figure 3

Nonmechanical approaches

Classification of unit operations for microbial cell disruption.

Hydrodynamic cavitation

Chaotropes

Induce lysis Wall inhibitors

Cell Disruption

697

Nonmechanical methods include physical methods of osmotic shock, gas decompression and sonication, as well as chemical and enzymic processes. Typical chemical treatments include the use of solvents to extract lipids and thereby remove the cell membrane, chaotropic agents to weaken hydrophobic interactions in the envelope, and reducing agents to permeabilize the wall through reduction of disulfide bridges. Conversely to mechanical methods, nonmechanical approaches such as chemical and enzymic methods have potential to enable SPR. They are microorganism specific in their mode of action. Disadvantages associated with chemical and enzymic lysis include process economics, product contamination through chemical addition, and extent of product release. In selecting the appropriate approach to microbial cell disruption, it is important to assess the nature of the product, its value, and required purity, as well as scale of operation in decision making. For example, a high-value product for which high purity is demanded and a small volume is to be processed may benefit from a selective release approach with limited recovery, thereby aiding ease of purification. Conversely, a commodity product with little cost margin requires high product recovery, low variable costs, and good potential for scalability.

2.47.4

Large-Scale Cell Disruption Technologies

2.47.4.1

High-Pressure Homogenization

In the high-pressure homogenizer, adapted from the device designed to create emulsions, the microbial suspension is first pressurized using a positive-displacement pump. The pressure is rapidly released by passage through a fine orifice or annular gap. Thereafter, the cell suspension typically impacts a solid surface to enhance cell breakage further. This process is illustrated in Figures 4 and 5, representative of the Manton–Gaulin system manufactured by APV and the system manufactured by Constant Systems Ltd. (Daventry, UK), respectively. In the Microfluidizer (Microfluidics, USA), the pressurized cell suspension flows as two streams that impact a stationary surface at high speed, followed by impact of the streams with each other.74 The design of the valve used in the high-pressure system has received considerable attention as, particularly for oil droplets and Gram-negative bacteria (low resistance to disruption), the stress fields created at the valve provide a key contribution to the disruption. Currently, sharp-edged orifices are preferred. Under operating pressures of 50–100 MPa, the annular gap has been estimated as 10–100 mm in typical equipment, while pressure drops are achieved across milliseconds. The importance of the distance between the annular gap and the point of impact with the solid surface on defining the extent of disruption as well as the nature of this impact have been demonstrated, especially of the more resilient yeast cells.61,62 Multiple mechanisms are at work to achieve cell disruption in the HPH, with contributing factors including rate and magnitude of pressure release, impingement on a solid surface, cavitation, turbulence, and shear stress.6,12,52 The first two are most typically viewed as dominant mechanisms while the contribution of cavitation has been shown.52,75 Through comparison of disruption of Gram-negative bacteria and yeast as well as the breakup of oil droplets, it is apparent that the dominant mechanism is influenced by the strength of the entity disrupted. For example, the impact distance between the valve outlet and impact ring affected the disruption of S. cerevisiae to a greater extent than that of E. coli.12 The key operating variables in the homogenizer are the operating pressure and the number of passes through the valve. Operating pressures in the range of 20–120 MPa are used for HPH. Typical breakage performance as a function of these operating 3

A

B

2 1

6

7

P

4 5

Key: 1. Inflow 2. Cell Distributor Valve 3. Outflow 4. Impact Ring 5. Cell Distributor Housing 6. Piston 7. Mechanism Controlling Pressure on Piston Figure 4 Passage of cell suspension through the high-pressure homogenizer. (A) The discharge valve assembly of the high-pressure homogenizer and (B) discharge valve design and fluid flow in the high-pressure homogenizer.

698

Cell Disruption

Hydraulic circuit and system components – T-series disrupters up to 1.1 kW

Cooling flow

Confidential hydraulic circuit (BT &T+) 4.pre Copyright Constant Systems Limited, 2002

4 Disrupter head 3

Sample inlet valve

Sample exit jet Sample flow 1

5 2

Lid on (NC) Proximity switch

One−way flow control valve (down stroke)

HP cylinder

Lid on probe

Boundary of hydraulic block

Up switch (NC)

Down switch (NC)

Solenoid valve

LP cylinder

Lower test point (2) (Permanently connected on Z+)

Pressure relief valve Oil temperature switch (NC)

Pilot only

Upper test point (1)

Nonreturn valve

S Pressure switch

Motor

Return filter

Tank Inlet filter

P Accumulator

From outlet to inlet if P1 ≥ P2 + 15 psi P2

Normal inlet

P1 Normally closed relief valve

Normal outlet

Valve opens to allow reverse flow

Figure 5 The high-pressure cell disruptor supplied by Constant Systems Ltd., illustrating both the valve at which pressure drop is created and associated equipment. Reproduced by courtesy of Constant Systems Ltd., Daventry, Northants NN1145D, England.

Cell Disruption

A

1

B

120

12

100

10

80

8

60

6

40

4

20

2

699

0.7 0.6 0.5 0.4 0.3 0.2

DNA release into supernatant (kg/kg × 1000)

0.8

Soluble protein release (kg/kg × 1000)

Fractional cell disruption

0.9

0.1 0

0

0

0

20 40 60 Operating pressure (MPa) Sol. Protein

DNA

80

Cell number

0

2

4 6 Number of passes

15.2 MPa

27.6 MPa

62.0 MPa

17.2 MPa

34.5 MPa

72.4 MPa

8

Figure 6 Cell disruption as a function of (A) operating pressure (single pass) and (B) number of passes. This is illustrated for the Gram-negative bacterium Cupriavidus necator (formerly Alcaligenes eutrophus) in stationary phase. In (A), comparison of disruption in terms of soluble protein release (,), DNA release (C), and ruptured cells as a function of cells visible (-) provides data on the extent of disruption achieved. In (B), open symbols represent soluble protein release while closed symbols represent DNA release. Reproduced from Harrison STL, Dennis JS, and Chase HA (1991). The disruption of Alcaligenes eutrophus by high-pressure homogenisation: Key factors involved in the process. Bioseparation 2: 155–166.

Table 2

Dependence of cell disruption on high-pressure homogenization on pressure, reported as a function of operating variables across a range of microbial systems

Microorganism Saccharomyces cerevisiae (Baker’s yeast) Saccharomyces cerevisiae (Brewer’s yeast) Escherichia coli Cupriavidus necator (formerly A. eutrophus)

Growth conditions

m ¼ 0.17 h1 m ¼ 0.3 h1 Exponential Stationary phase

Disruption rate constant (MPaa)

Operating pressure

Pressure exponent, a

49 MPa

2.9

7

1.87

40

95 MPa 28 MPa, three passes 69 MPa, one pass

1.77 1.40 3.08 1.64

56.7  103 62.9  103

References

74 24

variables is summarized in Figure 6 and Table 2. Microbial disruption has been shown to be first order with respect to the number of passages through the homogenizer and an exponential function of operating pressure (Ref. 7, reviewed in Refs. 6 and 12), as described in Eq. (1): 
Rmax ¼ k$P a $N ln (1) Rmax  R where N is the number of passes through homogenizer and a is the pressure exponent. Harrison et al.53 (cited in Ref. 12) illustrated that Eq. (1) relates to the release of soluble components. Large molecules such as the nucleic acid and particulate products require a threshold to be overcome in terms of cell disintegration before release, as indicated in Eq. (2): 
Rmax ¼ k$P a $ðN  0:75Þ ln (2) Rmax  R As illustrated in Table 2, the rate constant k and pressure exponent a are dependent on the properties of the microbial cell, including cell type, growth phase, and physiological status. Various approaches to expanding the predictive modeling of cell disruption in the homogenizer have been presented in which combinations of cell-wall strength, impact distance, velocity, and valve dimensions are considered.12 From these, it is evident that knowledge of both the hydrodynamic and cell strength

700

Cell Disruption

A

B

C

Figure 7 Disruption of the Gram-negative bacterium Cupriavidus necator on passage through the high-pressure homogenizer viewed through transmission electron microscopy: (A) 0 passes, (B) 1 pass at 72 MPa, and (C) 4 passes at 72 Mpa. From Harrison STL, Dennis JS, and Chase HA (1991). The disruption of Alcaligenes eutrophus by high-pressure homogenisation: Key factors involved in the process. Bioseparation 2: 155–166.

distributions is required. Middelberg and co-workers (cited in Ref. 12) proposed a new model combining these, illustrated for a single pass and multiple passes, respectively, through the homogenizer in Eq. (3): Z N D¼ fD ðSÞfS ðSÞdS 0

Z D¼1

N 0

½1  fD ðSÞN fS ðSÞdS

(3)

where N is the number of passes through the homogenizer, fS(S) is the cell strength distribution, and fD(S) is the distribution of stress in the homogenizer. Middelberg12 correlated the effective cell strength in terms of the degree of peptidoglycan cross-linking and the average cell length. Further, he correlated the stress distribution in terms of the impact on a small cylinder and included impact distance. Although he demonstrated that inclusion of these terms allowed the generalized prediction of cell disruption for a particular cell type (E. coli, Baker’s and Brewer’s yeasts), these data are not typically available to bioprocess engineers across a range of cell types. Hence, more rigorous understanding of cell-wall strength is required. This has been attempted through cell compression studies to determine the relationship between compression force, cell deformation, and bursting force for yeast cells.16,80 Through these studies, it is shown that the yeast cell wall is modeled appropriately as a permeable structure. The mean maximum von Mises strain at failure depends on cell size, but not compression rate at compression rates exceeding 45 mm s1 at which point the pseudoelastic modulus is constant. Stenson81 provides the following data for the stationary phase Baker’s yeast cell strength: elastic modulus of 185  15 MPa, initial stretch ratio 1.039  0.006, circumferential stress at failure 115  5 MPa, circumferential strain at failure of 0.46  0.3, and strain energy per unit volume at failure of 30  3 MPa. It is hoped that, in future, such information will inform the optimization of mechanical cell disruption. However, such inclusion of cell strength distribution will require recognition of varying cell strength within a population, as acknowledged in unpublished data (Harrison) and by Donsi et al.4 It is evident from the micrographs provided in Figure 7 that on the first pass through the homogenizer, a point break in the cell results. This is consistent with rupture resulting from a pressure release event or single impact. Release of soluble products through the point break ensues. Further passes through the homogenizer are required for the further disruption of the cell structure and the release of granular materials such as inclusion bodies and large molecules such as the genomic DNA. Although temperature of the cell suspension has been shown to have a small influence on disruption, no influence of biomass concentration has been observed across the range 75–150 g dry mass l1 for yeast systems and 95–260 g dry mass l1 for bacterial systems.6 The process is easily operated on a large scale. Most of the power input into the homogenizer is dissipated as heat. Typical heat loads resulting have been measured in the range 0.15–0.23  C MJ1 across a range of studies.3,9,31,52 Owing to the heat labile nature of many bioproducts, the provision of effective cooling is critical and is typically provided independent of the volume processed through heat exchange. The provision of external cooling limits the maximum pressure that can be utilized in the homogenizer. Key references providing further detail on microbial cell disruption by HPH include papers of Hetherington et al.,7 Follows et al.,5 Keshavarz-Moore et al.,61 Harrison,6 Kleinig and Middelberg,62 Donsi et al.,4 Harrison et al.,53 and Middelberg.12 More recent

Cell Disruption

701

literature on high-pressure systems is focused on the mechanism of disruption,4,16,43 combination applications,42 and the use of high pressure in the deactivation of microorganisms in food applications.

2.47.4.2

High-Speed Bead Mills

In the high-speed bead mill, the microbial suspension is agitated vigorously in the presence of a particulate solid phase, typically glass beads. Disruption is achieved through interparticle collision and solid shear. The bead mill consists of a horizontal or vertical cylinder fitted with a central drive shaft and several impellers (Figure 8). The cylinder is partly filled (typically 80%) with small glass or ceramic beads. Agitation of the microbial slurry in this system in the presence of the beads at impeller tip speeds of approximately 15 m s1 results in cell disruption. As with the homogenizer, the energy added to the system is partly used in cell breakage, but largely dissipated as heat, requiring well-designed heat transfer systems. The main operating variables in the bead mill are the particulate loading, particulate size, agitation intensity, and time of agitation. Cell disruption in the bead mill is first order with respect to the intact cell concentration and can be defined as:   Rmax ¼ kt (4) ln Rmax  R where Rmax is the maximum soluble protein available for release, R is the soluble protein released, k is the disruption rate constant, and t is the treatment time. The rate constant is influenced by the microorganism type, the nature and location of the product, the bead size and loading, impeller design and speed, and temperature. A bead size of 0.5- to 3-mm diameter is favored for yeast breakage with k increasing with decreasing bead size in this range. Bacterial disruption requires smaller beads (0.1 mm diameter) for optimum disruption. Typical breakage performance as a function of these operating variables is summarized in Table 3. Breakage in the bead mill and a slurry bioreactor (baffled stirred tank reactor (STR) containing 20–40% solids by volume) can be modeled by equations of the same form. The disruption rate constant is a function of both power input (related to agitation rate) and solids loading,77 as defined in Eq. (5):
a P fb (5) k¼A V where P/V is the power input per unit volume (fN3D5), 4 is the solids loading expressed as volume fraction, and a, b, and A are the constants. The scaleup of the bead mill is typically constrained by the removal of the heat generated on the dissipation of the large amounts of energy added to the process. Owing to the extended residence time of the suspension in the chamber of the mill, heat removal in situ is required (unlike the homogenizer where heat removal can be provided through external heat exchange). This places a limit on dimensions such that the available surface area for heat exchange can handle the heat generation that varies as a function of the system volume. Key references providing further detail on microbial cell disruption in the high-speed bead mill include papers of Currie et al.,37 Schutte et al.,78 Schutte and Kula,79 and Ramanan et al.70 and reviews of Harrison6 and Middelberg.12 Key: 1. Inlet 2. Outlet 3. Microbial Slurry with Grinding Agent 4. Agitator Discs 5. Agitator Shaft 6. Separator 7. Drive Belt 8. Motor

2

1

4

6

7 5

3 8

Figure 8

The horizontal bead mill.

702

Cell Disruption

Table 3

Typical operating conditions used for cell disruption in the high-speed bead mill, across a range of microbial systems

Microorganism

Agitation speed (rpm)

Initial cell concentration

Bead size (mm)

Bead loading (v/v)

Reference

Baker’s yeast Baker’s yeast Recombinant S. cerevisiae Baker’s yeast, E. coli, L. casei, L. confusus, B. cerus, C. boidinii Baker’s yeast, Brewer’s yeast, Candida utilis Pichia pastoris Baker’s yeast

1100, 1700, 2300, 3100 600–1800 1000–4000 150–1800

170, 250, 325 g l1 0–600 g l1 0–20 g l1 50–600 g l1

0.5 0.5–3 0.25–0.75 0.1–1.5

20%, 30%, 40% 3–10 kg (4.2 l tank) 70–85% 40–95%

73 37 47 79

1640, 2340, 2930

110 g l1 (dry wt)

0.5–0.75

50%

66

10 ms1 1900

350–450 g l1 0–30% w/w

0.5–0.75 0.75–1.00

85% 10–80%

34 84

2.47.5

Laboratory-Scale, and Developing, Cell Disruption Technologies

2.47.5.1

Laboratory-Scale Mechanical Disruption

The operating principles of both the high-pressure homogenizer and the high-speed bead mill may be used on a laboratory scale. The conventional device using rate of pressure release to mediate cell disruption is the French Press. This operates as a batch system handling suspension volumes of 5–25 ml. The cell suspension is pressurized using a hydraulic or manual system. Thereafter, the pressure is released through a capillary valve. An example is the French Press Cell Disrupter supplied by Rensselaer. Several flow-through systems have been designed for lab-scale use, including the Rannie homogenizer from APV and the Constant Systems homogenizer. Although these mimic the high-pressure homogenizer more directly, differences in valve design are found. Small-scale bead mills are provided as laboratory equipment, functioning on the same principle as the bead mill. Further, the principle of interparticle collision and solid shear can be implemented in the laboratory through bead beating. On a small scale, this is conducted in a small tube with vortexing or using a purpose built instrument such as the FastPrepÒ from Q.BioGene. On a larger scale, agitation in a stirred tank reactor at an equivalent bead loading ( 40% by volume) can be used.77

2.47.5.2

Cavitation

Vapor cavities form in a liquid at locations of reduced pressure where the localized pressure is less than the vapor pressure of the liquid, or approaches it in the presence of dissolved gases. The formation, growth, oscillation, and collapse of these cavities are known as cavitation. During oscillation of the cavities and their subsequent collapse on bulk pressure recovery, pressure fluctuations occur, with concomitant energy dissipation and localized heating. Extremely localized pressures and temperatures have been recorded on cavitation. Physical effects associated with cavitation include erosion, damage, and disintegration of solid surfaces, dispersion of fragmented solid particles or gas bubbles, and emulsion of liquid–liquid systems. Cavitation generated by sound waves, known as ultrasonic cavitation, and that generated by pressure reduction in a flowing system, known as hydrodynamic cavitation, have been used to generate microbial cell disruption. The mechanism of cell disruption by cavitation is not fully understood. Disruption by ultrasonic cavitation was attributed to dynamic pressure differences across the cell caused by turbulent eddies with dimensions of the magnitude of the cell, resulting in the largest stable cell size being a function of the energy dissipation rate and cell strength distribution.39 This understanding has been expanded32,35 to include the role of the liquid micro-jet formed on cavity collapse and the propagation of the shock wave. Further, radial bubble motion or oscillation of the bubble prior to collapse contributes to cell deformation and may invoke cell-wall fatigue. As an example, cavities have been shown to oscillate between a radius of 3 and 37 mm in 16 ms on exposure to ultrasound at 26.5 kHz. Cavitation has also associated chemical effects, due to the formation of free radicals and subsequent oxidation reactions.

2.47.5.2.1

Ultrasonic Cavitation

Ultrasound, the sound of frequency higher than 15–20 kHz, causes cavitation in liquids, that is, the formation of vapor cavities in low-pressure zones. This has long been recognized as a method of microbial cell disruption and is a common laboratory technique. A number of factors affect the disruption of microbial cells using ultrasonication. These include power input per volume and the temperature of the suspension. Typical acoustic power ranges used lie in the range of 20–250 W at a frequency of 20 kHz and above. Cell disruption decreases with increasing volume and increases with increasing power input.48 A slight increase in disruption is observed with increasing temperature over the range 17–30  C. No impact of cell concentration is observed across the range 3–20 g l1 E. coli and 40–150 g l1 yeast (dry mass). Cell lysis by ultrasound is described adequately by first-order release kinetics. Much of the ultrasonic energy is converted to heat, and therefore good temperature control is required to avoid denaturation of proteins. Micronization of cell debris can result. It is difficult to transmit sufficient power to a large volume of cell material. Sonication is most commonly used as a laboratory technique. Although most laboratory-scale cell disruption methods are applicable at the scale of 5–500 ml, the use of adaptive focused acoustics (AFA) for the lysis of very small quantities (1.5 ml per sample) within 30–600 s is reported17 to facilitate microscale

Cell Disruption

703

process development. AFA operates through a mechanism similar to ultrasound, however, at a higher frequency (102–105 kHz, compared with 101–102 kHz for ultrasound). The disruption of S. cerevisiae has been demonstrated by AFA using the Covaris E210 instrument.

2.47.5.2.2

Hydrodynamic Cavitation

A cavitation event similar to that generated by ultrasound can be induced through fluid-flow patterns. According to Bernoulli’s equation, on flow through an orifice, the increasing velocity required to satisfy the continuity equation is accompanied by a decreasing pressure in the fluid. Where the pressure decreases to the vapor pressure of the suspending medium or below, the formation of vapor cavities results in the phenomenon of cavitation, described above, with its associated cell damage or disruption on cavity oscillation and collapse (Figure 9). Hydrodynamic cavitation was originally recognized as a contributing mechanism in homogenization.6 It has also been shown to mediate effective disruption of both bacteria and yeast with concomitant enzyme release.22 It should be noted that the role of pressure in homogenization and cavitation is different. In the former, cells are equilibrated at high pressure, followed by a sudden release inducing cell envelope failure to release the high internal pressure. Under conditions of cavitation, the cells experience the imposition of an extremely high-localized external pressure, mediating cell envelope failure. The disruption of microorganisms by hydrodynamic cavitation was first reported for S. cerevisiae and Cupriavidus necator by Harrison and Pandit.52,55 Subsequently, the release of proteins and intracellular enzymes from S. cerevisiae and E. coli by hydrodynamic cavitation has been studied further and its potential for augmenting selectivity of product release is reported.22–25,75,76 The extent of cell disruption can be correlated with the cavitation number, the ratio between the forces tending to cavity collapse and those initializing cavity formation: Cv ¼

P3  Pv 1 rv2 2

(6)

where P3 is the fully recovered downstream pressure, Pv is the vapor pressure of the medium, r is the density of the medium, and v is the velocity at the orifice. Cavitation inception is typically recognized to occur below Cv of 1. Maximum cell disruption was reported at a Cv of 0.13 for S. cerevisiae and 0.17 for E. coli.24,25 Key operating variables are the operating pressure of the cavitation system, influencing the collapse pressure and number of cavities, the geometry of the orifice, the cell concentration, the number of passes across the orifice, and the operating temperature. Hydrodynamic cavitation can be operated on a large scale and is recognized as energy efficient. Its application has been demonstrated at pilot scale; however, it has yet to find commercial application.

2.47.5.3

Chemical Treatment

Microbial disruption by chemical methods is dependent on the cell structure to be disrupted. These structures are reviewed in Section 2.47.2. Comprehensive review of chemical cell disruption is provided by Middelberg,12 Naglak and Wang,13 and Harrison.6 Typical agents used are pH extremes, especially alkali conditions, solvents, detergents, chelating agents, reducing agents, and chaotropic agents. In this section, only the most commonly used agents, showing broad applicability, are discussed. Cavity formation region

Cavity collapse region

Orifice plate Vapor cavities

Intact cell

Damaged cell

Orifice plate Pressure P1 P3 Velocity P2 Figure 9 Cavity formation and pressure fluctuations on flow through an orifice. From Balasundaram B (2004). A Detailed Investigation of Microbial Cell Disruption by Hydrodynamic Cavitation for Selective Product Release. PhD Thesis, University of Cape Town.

704

Cell Disruption

Alkaline cell lysis has been demonstrated for a number of bacterial systems, using pH 10.5–12.5 over a time period of 30 s to 30 min. Examples tested include Erwinia carotovora, E. coli, and C. necator.6,15 For application, a product stable at high pH is required and the requirement for neutralization will affect material inventories. Solvents may be used to extract lipid components from the cell membrane, causing release of intracellular components.6,12,13 Care must be taken in their application owing to their flammability and potential to cause protein denaturation. Solvents used for the release of intracellular compounds include alcohols such as ethanol, isopropanol, and butanol (at concentrations of 10–80%), dimethyl sulfoxide, toluene (2%), and methyl ethyl ketone. Application across a broad range of microorganisms including E. coli, S. cerevisiae, and Kluveromyces species has been demonstrated. Although permeabilization occurs at ambient temperature, increased release results on elevating temperatures in the range of 25–45  C. Detergent treatment is used extensively on laboratory scale to lyse or permeabilize cells for release of soluble components through perturbing the protein–lipid interactions through interaction with the nonpolar hydrophobic tail and polar hydrophilic head of the detergent molecule. Stability of the product in these systems must be ensured. Detergents are classified according to the nature of the hydrophilic head as anionic, cationic, and nonionic. Anionic detergents (e.g., sodium dodecyl sulfate, SDS) disorganize the cell membrane. Cationic detergents are suggested to act on the lipopolysaccharide component of the cell envelope as well as interacting with the phospholipids. Cetyltrimethylammonium bromide (CTAB) has been used for permeabilization of both yeast and bacteria at a concentration in the range of 0.02–0.4%.29 Nonionic detergents such as Triton X-100 and Pluronic F-68 cause a partial solubilization of proteins in the inner membrane structure, resulting in permeabilization.15 The lipopolysaccharide component of the outer membrane provides resistance to the detergent unless a combination of chemical approaches is used. Triton X-100, at concentrations varying between 0.1% and 4%, has been demonstrated to aid permeabilization of E. coli, S. cerevisiae, P. pastoris, Nocardia rhodocrous, and Yarrowia lipolytica.13,15,46,64 Chaotropic agents mediate cell lysis through disrupting H-bonding and altering hydrophobic interactions, thereby reducing cross-linking within the cell wall. Typical agents include guanidine hydrochloride and urea. The effect of both chaotropic agents and detergents can be enhanced in combination with a chelator such as EDTA. This chelates divalent ions, destabilizing membranes and lipopolysaccharide layers.29,38

2.47.5.4

Enzymatic Attack

Enzymatic cell lysis is a controlled low-energy operation, requiring low capital investment. It can yield biological specificity and takes place under mild operating conditions. Harsh physical conditions such as the high shear stress of mechanical disruption can be avoided. Selection of an appropriate enzyme or enzyme system and the determination of specific reaction conditions for efficient lysis are required. Three approaches are reported: autolysis, addition of foreign lytic enzymes, and phage lysis. The latter is not favored owing to risk of unintentional infection, so is not considered here. Application of enzymic lysis on a large scale is currently constrained by enzyme availability and cost. Autolysis, reviewed by Middelberg,12 used to prepare yeast extracts, is poorly understood, and hence poorly controlled. The production of lytic enzymes by the yeast is typically induced by mild chemical or thermal shock; hence, its classification can be unclear. Three groups of bacteriolytic enzymes are available: glycosidases that hydrolyze polysaccharide chains; acetylmuramoylL-alanine amidases that cleave polysaccharide polypeptide linkages, and endopeptidases that lyse polypeptide chains. Each of these attacks the peptidoglycan wall, requiring the prior removal of the outer membrane of Gram-negative bacteria. The most important bacteriolytic enzyme and only enzyme commercially available on large scale is lysozyme, produced from hen egg white and available from other natural sources. It hydrolyzes b-1-4-glucosidic linkages of polysaccharide chains of peptidoglycan. Other bacteriolytic enzymes include those from Cytophaga sp., Staphylococcus sp., and Streptomyces sp., which can lyse Gram-positive bacteria as well as the zinc endopeptidase Lysostaphin and the b-N-acetyl-D-glucoaminidase labiase. A lytic protease from Micronospora sp. can lyse lyophilized cells of Gram-negative bacteria, and a lytic protease produced by B. subtilis is reported to lyse cells of the bacterium E. coli without the need for pretreatment.1 The enzyme system for yeast lysis requires a mixture of different enzymes including b-1-3-glucanase, protease, b-1-6-glucanase, mannanase, and chitinase, acting synergistically to lyse the cell wall. Essentially, two enzymes are required: a wall lytic protease to degrade the outer protein–mannan complex and a lytic b-1-3-glucanase to degrade the inner layer.15 Most yeast lysing enzyme preparations represent such enzyme mixtures, dominant in b-glucanase. Cytophaga sp. and Oerskovia xanthineolytic lysing enzymes and zymolase have shown yeast-lysing potential.20,28

2.47.5.5

Osmotic Treatment

Where osmotic pressure is altered gradually, microbial cells maintain a balanced osmotic pressure across the cell envelope by altering their cytoplasmic composition. To induce lysis through osmotic shock, the cells are equilibrated under conditions of high osmotic pressure, typically provided as either a mono- or disaccharide solution or a salt solution (1 M). Thereafter, a rapid exposure to a solution of low osmotic pressure results in water entering the cell rapidly to remove the osmotic gradient. The increase in internal pressure causes lysis of cells. Osmotic shock as a disruption technique is restricted to systems in which the cell wall is weakened or absent. It has been reported to facilitate the release of proteins from E. coli without cell rupture or reduction of cell viability. Further osmotic shock has been used in combination with mechanical disruption methods, discussed later in

Cell Disruption

705

Section 2.47.7. The application of osmotic shock on a large scale is limited by the cost of the osmo-regulator12 as well as the increase in water use and potential for undesirable dilution of the process.

2.47.5.6

Thermal Treatment

Thermal treatment can result in release of intracellular components from both yeast and bacteria. Protein release from E. coli has been reported across the temperature range 30–90  C over a period of 20–30 min. However, this has not found widespread application as a cell disruption method owing to the concomitant denaturation of proteins of interest. Specific applications of thermal product release reported include the use of heat shock at 150  C for the release of the biodegradable thermoplastics of the poly-b-hydroxyalkanoate class from C. necator (previously known as Alcaligenes eutrophus) as well as the preparation of protein extracts from yeast for the food industry.6,52 Recent studies have highlighted key potential applications for large-scale application of thermolysis for intracellular product recovery. In the first application, the recovery of thermostable proteins is considered. As a case study,72 the recovery of a thermostable esterase, sourced from the thermophilic Archaea Aeropyrum pernix K1, from recombinant E. coli is described. On investigating the effect of temperature, pH, and cell concentration on protein release, Ren et al. showed temperature to be the controlling variable. While enzyme release was demonstrated at temperatures of 60  C and above, 80  C was optimum. The benefit of this approach to the recovery of thermostable proteins is the concomitant protein purification achieved as the heat labile proteins, including the native proteases, from the host organism denature and precipitate on heat treatment. In the case of the thermostable esterase, a 91% recovery and 12-fold protein purification were reported for the heat treatment at 80  C for 120 min. Notably, 80% recovery was achieved in only 20 min at 80  C and purification efficiency was observed at early stages of heat treatment. The cell debris resulting is large, facilitating solid–liquid separation. These authors did not report the rate of heating used in this heat treatment. As this has been recognized as an important factor,42,49 it is recommended that attention be given to heating rate in future studies. Similar application of thermolysis for the liberation of thermostable enzymes has been reported for the thermostable amidase, sourced from G. pallidus, produced in recombinant E. coli.33 The second application of interest for large-scale thermolysis is the recovery of plasmid DNA, required for gene therapy and DNA vaccines. Heat-treatment approaches used routinely for laboratory-scale preparation of DNA fractions can be extended to large-scale plasmid preparations. Plasmid DNA can be prepared by heat treatment at 95  C for 5 min in the presence of 100 mM EDTA and 5% Triton X100.85 Comparing with the typically used alkaline lysis, the first-order rate constant of cell lysis was 2.2-fold higher on alkaline lysis, whereas the first-order degradation rate constant was 2.7-fold higher, increasing the plasmid DNA concentration achieved by more than 2-fold. Little thermal degradation of the plasmid DNA is observed and the reduced degradation of chromosomal DNA on heat treatment, owing to denaturation of the nucleases present, enables easy separation of the chromosomal and plasmid DNA. Further, the lysate from heat treatment is less viscous than that from alkaline treatment, aiding further processing. The impact of thermolysis for the destruction of microorganisms in food has illustrated the importance of heating rate on the viability of the microbial cell.49 On increasing the heating rate to elevated temperatures in the range 25–50  C from 0.4  C min1 to a heat shock (instantaneous dilution at the temperature of interest), the remaining viability of S. cerevisiae decreased from 40% to 1%. A decrease in cell volume of 22% was observed, compared with G6PDH ¼ 6PGDH > fumarase  alkaline phosphatase

E. coli

Alkaline phosphatase (periplasmic)

E. coli

Penicillin acylase

Purity increased by eightfold but recovery decreased to 60%

45, cited by 2

E. coli

Long-R3-IGF-I

41, cited by 2

E. coli

Penicillin acylase

46% purity and 81% recovery obtained cf. 41% purity and 88% recovery by conventional process 94% recovery at an activity of 3.9 U mg1 cf. 0.31 U mg1 by sonication

90% acid phosphatase release cf. 45% soluble protein 93% release

78, cited by 2 5

[24, 25] 65, cited by 2

44, cited by 2

Typically, pretreatment has been investigated through the combination of chemical methods for cell-wall weakening with mechanical cell disruption. These include the use of detergents, chaotropic agents, solvents, chelating agents, and pH extremes. The performance of combined methods across a range of systems is reported in Table 5. In recent studies, it is recognized that the introduction of chemicals into the system should be minimized to avoid additional cost and purification demands. The use of temperature shock as a pretreatment has been demonstrated to be most effective for yeast systems, whereas pH shock and osmotic shock provided little improvement in mechanical disruption. Following exposure to a rapid temperature shock to 50  C at 4.3  C s1, S. cerevisiae was disrupted following four passes at 41 MPa, equivalent to disruption achieved on four passes at 69 MPa without pretreatment and exceeding the disruption achieved on eight passes at 41 MPa without pretreatment.42

2.47.8

Integration of Biomass Formation and Product Release

It is well recognized that both growth phase and growth rate of the microorganism influence the ease of mechanical disruption. Cells in the stationary phase are more resistant to mechanical disruption than actively growing cells.6,24,40,74 This is illustrated through consideration of the composite rate constant kPa in Eq. (1), describing high-pressure homogenization, where the pressure exponents are given in Table 2. This composite rate constant is illustrated as a function of specific growth rate in Figure 11. In accordance with this observation, cells with a higher specific growth rate are easier to disrupt. It is postulated to result from rapidly growing cells not directing resources to the reinforcement of the cell wall.40 Where cells are cultivated under conditions of increased shear, increased resistance to cell breakage has been reported.60 Structural strength of microbial cells varies across cell type as it is a function of the cell-wall structure and cell size. Typically, the order of decreasing resistance is as follows: yeast > Gram-positive bacteria > Gram-negative bacteria. The morphology of fungi, particularly hyphal length, dictate their position in the order, while the variability in cell-wall structure of both algae and Archaea do not allow generalization. Small unicellular algae with cellulosic cell walls, such as Chlorella and Scenedesmus, are among the most resistant to mechanical cell disruption. In addition to the average cell resilience, it is recognized that a distribution of cell strength will be found across a cell population, thereby affecting the overall disruption achieved under standard operating conditions. This has been observed in unpublished data of the author and described by Donsi et al.4 Bioprocess conditions and stage of growth phase at harvest may influence product location. For example, the location of glucose oxidase is reported predominantly as intracellular in Aspergillus niger, but extracellular in Penicillium sp.; however, the ratios shift

708 Cell Disruption

Table 5

Influence of microbial pretreatment on the subsequent mechanical cell disruption, across a range of microorganisms

Microorganism

Chemical treatment

E. coli

1.5 M G-HCl and 1.5% Triton X-100 SDS

Cupriavidus necator (formerly Alcaligenes eutrophus) C. necator (formerly A. eutrophus) Alkaline treatment pH 11

Pretreatment

Mechanical

Combined

HPH

62% soluble protein

41 MPa, 2 passes

HPH

4–11% Rmax

HPH

26% of Rmax of soluble protein Negligible protein release 6% disruption

100% sol. protein release (330 g kg1), 62 MPa, 2 passes 100% sol. protein release (330 g kg1), 62 MPa, 2 passes 81% Rmax, 4 passes at 41.4 MPa

82% protein, 41 MPa, 1 pass 138% Rmax following 1 pass at 62 MPa 137% soluble protein

32% disruption, 95 MPa, 4 pass

3.5% disruption

65% disruption, 95 MPa, 4 pass

S. cerevisiae

Heat shock to 50  C at 4.3  C s1 Zymolase

HPH (4 passes at 41 MPa) Microfluidizer (HPH)

Candida utilis

Zymolase

E. coli

EDTA (0.04 M)

Microfluidizer (HPH) HPH

S. cerevisiae

Product release (% of available product) on individual and combined treatments

Mechanical disruption

Maximum release on 4 passes at 34.5 MPa

97% Rmax, 4 passes at 41.4 MPa 100% disruption, 95 MPa, 4 pass 95% disruption, 95 MPa, 4 pass Maximum release on 4 passes at 13.8 MPa

Energy reduction (%)

Reference

50

21

33.3

54

40

54 42

75

28

33 60

19

Cell Disruption

709

Relative disruption rate constant as a function of growth rate

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0

50 100 Operating pressure (MPa)

C. necator

E. coli

150

C. utilis

a

Figure 11 Ratio of the effective disruption rate constant K ¼ kP for high pressure homogenization of rapidly and slow growing microbial cultures. (A) C. necator in stationary phase relative to early exponential phase53; (,) C. utilis at a specific growth rate of 0.1 h1 relative to 0.5 h1 40; and (-) E. coli at a specific growth rate of 0.17 h1 relative to 0.35 h1.74 Relationship between physicochemical properties of the disrupted cell suspension and subsequent product recovery and purification

Table 6

Suspension property Particle size

DNA content

Manner in which influenced by cell disruption

Parameter affected

Cell debris is micronized by extensive Settling velocity mechanical disruption, e.g., increasing passes through the homogenizer or Viscosity milling time in the bead mill V and eqn On complete disruption of the microbial Viscosity cell, high-MW DNA is released into suspension h ¼ K Ma where h ¼ intrinsic viscosity M ¼ molecular mass a ¼ 0.6–0.8 K2 ¼ 0.5  104 to 5  104

Polymeric components Increased mechanical disruption may result in liberation of glycans from the yeast cell wall Soluble nonproduct The contaminant load released into Adsorption or other separation operations components solution increases with increasing disruption efficiency, decreasing selectivity and the addition of chemical treatment agents

Unit operations in the DSP affected

References

Solid liquid separation processes such 18,26,27,36,56 as centrifugation and filtration Chromatographic processes in both fixed and expanded bed configurations Solid liquid separation processes such 18 as centrifugation and filtration Chromatographic processes in both 51,71 fixed and expanded bed configurations

67 Precipitation processes Adsorption processes Two-phase partitioning

59

with growth phase, in turn influencing the process selected and the potential yield.57 This illustrates the importance of integrating decisions on bioprocess and downstream processing operations on the subsequent product yields.

2.47.9

Integration of Product Release and Product Recovery and Purification

The nature of the disruption process influences the physicochemical characteristics of the resulting microbial cell lysate. This, in turn, impacts on the unit operations downstream of the cell disruption step in which product recovery and purification are optimized. Key characteristics of the disrupted cell suspension impacting product recovery and purification are the resultant particle size of the cell debris, the release of long chain molecules affecting viscosity and the loading of soluble contaminant molecules requiring separation from the product molecule. The influence of these on subsequent processing is shown in Table 6. For example, on extending the disruption of E. coli by HPH from one to three passes, the increased release of protein product from 58% to 78% was accompanied by a decrease in the mean diameter of the cell debris particles from 0.5 to 0.2 mm. This, in turn, reduced the sedimentable solids on centrifugation by 20%, providing an increased particle load into downstream operations or requiring extended centrifugation. Similarly, the effect of increased micronization of cell debris and contaminating load on chromatographic separations using an expanded bed has been documented.26,27

710

Cell Disruption

2.47.10 Closing Remarks Release of products formed in microbial cell culture is a critical unit operation to enable exploitation of the full range of products of these systems. In some microbial systems, this is achieved by natural secretion of the product or molecular modification to enable secretion; however, intracellular products remain the dominant product group, requiring a cell disruption step for their recovery. To date, mechanical disruption processes such as HPH, microfluidics, and treatment in the high-speed bead mills dominate the large-scale application of microbial cell disruption. Disruption mechanisms in all these cases are complex with a variety of contributing components including pressure release, shear stress, turbulence, cavitation, impingement, and attrition. Modeling of the disruption process for design purposes remains empirical; however, improved mechanistic modeling is developing with our understanding. The application of mechanistic model approaches is limited owing to the limitations on data available of microbial cell strength. While physical approaches are typically inefficient at the large scale, hydrodynamic cavitation has been demonstrated at pilot scale to show potential as a cell disruption approach. Further, the scaleup of sonication is considered. Enzymatic treatments remain promising for selective and low-energy product release; however, their availability and expense for large-scale operation continues to restrict application. Current developments in the field of cell disruption are focused on the selective release of products, pretreatment to enhance product release, low-energy processes for product release, and process integration. In terms of the latter, the interaction between the bioproduction environment and cell disruption for product release, as well as the impact of the disruption step on the resultant suspension characteristics and subsequent product recovery and purification, is increasingly recognized as critical in process optimization.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31.

Asenjo, J. A. Separation Processes in Biotechnology, Dekker: New York and Basel, 1990. Balasundaram, B.; Harrison, S. T. L.; Bracewell, D. G. Developments in Product Release for Whole Bioprocess Design Strategies. Trends Biotechnol. 2009, 27, 477–485. Chisti, Y.; Moo-Young, M. Disruption of Microbial Cells for Intracellular Products. Enzym. Microb. Technol. 1986, 8, 194–204. Donsi, F.; Ferrari, G.; Lenza, E.; Maresca, P. Main Factors Regulating Inactivation by High Pressure Homogenisation: Operating Parameters and Scale of Operation. Chem. Eng. Sci. 2009, 64 (3), 520–532. Follows, M.; Hetherington, P. J.; Dunhill, P.; Lilly, M. D. Release of Protein From Baker’s Yeast by Disruption in an Industrial Homogenizer. Biotechnol. Bioeng. 1971, 13, 549–560. Harrison, S. T. L. Bacterial Cell Disruption: A Key Unit Operation in the Recovery of Intracellular Products. Biotechnol. Adv. 1991, 9, 217–240. Hetherington, P. J.; Follows, M.; Dunhill, P.; Lilly, M. D. Release of Protein From Baker’s Yeast (Saccharomyces cerevisiae) by Disruption in an Industrial Homogenizer. Trans. Inst. Chem. Eng. 1971, 49, 142–148. Huang, R. B.; Andrews, B. A.; Asenjo, J. A. Differential Product Release of Proteins From Yeast: A New Technique for Selective Product Recovery From Microbial Cells. Biotechnol. Bioeng. 1991, 38, 977–985. Keleman, M. V.; Sharpe, J. E. E. Controlled Cell Disruption: A Comparison of the Forces Required to Disrupt Different Microorganisms. J. Cell Sci. 1979, 35, 431–441. Klis, F. M.; Mol, P.; Hellingwerf, K.; Brul, S. Dynamics of Cell Wall Structure in Saccharomyces cerevisiae. FEMS Microbiol. Rev. 2002, 26, 239–256. Klis, F. M.; Boorsma, A.; de Groot, P. W. J. Cell Wall Construction in Saccharomyces cerevisiae. Yeast 2006, 23, 185–202. Middelberg, A. Process-scale Disruption of Microorganisms. Biotechnol. Adv. 1995, 13, 491–551. Naglak, T. J.; Wang, H. Y. Recovery of Foreign Protein From the Periplasm of Escherichia coli by Chemical Permeabilisation. Enzym. Microb. Technol. 1990, 12, 603–611. Prescott, L. M.; Harley, J.; Klein, D. A. Microbiology, 4th ed.; McGraw-Hill: Boston, MA, 1999. Schutte, H.; Kula, M.-R. Pilot and Process-scale Techniques for Cell Disruption. Biotechnol. Appl. Biochem. 1990, 12, 599–620. Stenson, J. D.; Thomas, C. R.; Hartley, P. Modelling the Mechanical Properties of Yeast Cells. Chem. Eng. Sci. 2009, 64, 1892–1903. Wenger, M. D.; de Phillips, P.; Bracewell, D. G. A Microscale Yeast Cell Disruption Technique for Integrated Process Development Strategies. Biotechnol. Prog. 2008, 24 (3), 606–614. Agerkvist, I.; Enfors, S.-O. Characterization of E. coli Cell Disintegrates From a Bead Mill and High Pressure Homogenizers. Biotechnol. Bioeng. 1990, 36, 1083–1089. Anand, H.; Pandit, A. B.; Balasundaram, B.; Harrison, S. T. L. The Effect of Chemical Pretreatment Combined With Mechanical Disruption on the Extent of Disruption and Release of Intracellular Protein From E. coli. Biochem. Eng. J. 2007, 35, 166–173. Asenjo, J. A.; Dunhill, P. The Isolation of Lytic Enzymes From Cytophaga and Their Application to the Rupture of Yeast Cells. Biotechnol. Bioeng. 1981, 23, 1045–1056. Bailey, S. M.; Blum, P. H.; Meagher, M. M. Improved Homogenization of Recombinant Escherichia coli Following Pre-treatment With Guanididne Hydrochloride. Biotechnol. Prog. 1995, 11, 533–539. Balasundaram, B. A Detailed Investigation of Microbial Cell Disruption by Hydrodynamic Cavitation for Selective Product Release (Ph.D. thesis), University of Cape Town, 2004. Balasundaram, B.; Pandit, A. B. Selective Release of Invertase by Hydrodynamic Cavitation. Biochem. Eng. J. 2001, 8, 251–256. Balasundaram, B.; Harrison, S. T. L. Study of Physical and Biological Factors Involved in the Disruption of E. coli by Hydrodynamic Cavitation. Biotechnol. Prog. 2006, 22, 907–913. Balasundaram, B.; Harrison, S. T. L. Disruption of Baker’s Yeast by Hydrodynamic Cavitation: Process Variables and Their Influence on Selective Release. Biotechnol. Bioeng. 2006, 94 (2), 303–311. Balasunsdaram, B.; Harrison, S. T. L.; Li, J.; Chase, H. A. A Study of the Influence of Yeast Cell Debris on Protein and a-Glucosidase Adsorption at Various Zones Within the Expanded Bed Using In-bed Sampling. Biotechnol. Bioeng. 2008, 99 (3), 614–624. Balasundaram, B.; Harrison, S. T. L. Influence of Method and Extent of Disruption of Bakers’ Yeast on Protein Adsorption in Expanded Beds. J. Biotechnol. 2008, 133, 360–369. Baldwin, C. V.; Robinson, C. W. Enhanced Disruption of Candida utilis Using Enzymatic Pretreatment and High-pressure Homogenization. Biotechnol. Bioeng. 1993, 43, 46–56. Bansal-Mutalik, R.; Gaikar, V. G. Cell Permeabilisation for Extraction of Penicillin Acylase From Escherichia coli by Reverse Micellar Solutions. Enzym. Microb. Technol. 2003, 32, 14–26. Berger, E.; du Plessis, E.; Gerber, I.; et al. Impact of Inactivated Extracellular Proteases on the Modified Flagellin Type III Secretion Pathway of Bacillus halodurans. Appl. Environ. Microbiol. 2009, 75 (1), 271–274. Brookman, J. S. G. Mechanism of Cell Disintegration in a High Pressure Homogeniser. Biotechnol. Bioeng. 1974, 16, 371–383.

Cell Disruption

711

32. Brujan, E. A. The Role of Cavitation Micro-jets in the Therapeutic Applications of Ultrasound. Ultrasound Med. Biol. 2004, 3, 381–387. 33. Cameron, R. A.; Sayed, M.; Cowan, D. A. Molecular Analysis of the Nitrile Catabolism Operon of the Thermophile Bacillus pallidus RAPc8. Biochim. Biophys. Acta 2005, 1725, 35–46. 34. Canales, M.; Buxado, J. A.; Heynngnezz, L.; Enruquez, A. Mechanical Disruption of Pichia pastoris Yeast to Recover the Recombinant Glycoprotein Bm86. Enzym. Microb. Technol. 1998, 23, 58–63. 35. Chahine, G. L. Cavitation Dynamics at Microscale Level. J. Heart Valve Dis. 1994, 3, S102–S116. 36. Clarkson, A. I.; Lefevre, P.; Titchener-Hooker, N. J. A Study of Process Interactions Between Cell Disruption and Debris Clarification Stages in the Recovery of Yeast Intracellular Products. Biotechnol. Prog. 1993, 9, 462–467. 37. Currie, J. A.; Dunnill, P.; Lilly, M. D. Release of Protein From Bakers Yeast (Saccharomyces cerevisiae) by Disruption in an Industrial Agitator Mill. Biotechnol. Bioeng. 1972, 14, 725–736. 38. De Smet, M. J.; Kingma, J.; Witholt, B. The Effect of Toluene on the Structure and Permeability of the Outer and Cytoplasmic Membranes of Escherichia coli. Biochim. Biophys. Acta 1978, 506, 64–80. 39. Doulah, M. S.; Hammond, T. H. A Hydrodynamic Mechanism for the Disintegration of Saccharomyces cerevisiae in an Industrial Homogeniser. Biotechnol. Bioeng. 1975, 17, 845–858. 40. Engler, C. R.; Robinson, C. W. Effects of Organism Type and Growth Conditions on Cell Disruption by Impingement. Biotechnol. Lett. 1981, 3, 83–88. 41. Falconer, R. J.; O’Neill, B. K.; Middelberg, A. P. J. Chemical Treatment of Escherichia coli: 3. Selective Extraction of a Recombinant Protein From Cytoplasmic Inclusion Bodies in Intact Cells. Biotechnol. Bioeng. 1999, 62, 455–460. 42. Fan, J.; Harrison, S. T. L. Release of Protein and Enzymes From Baker’s Yeast by Ultrasonication Following Pretreatment by Heat, PH and Osmotic Shock. J. Chem. Technol. Biotechnol. 2010 (submitted). 43. Floury, J.; Bellettre, J.; Legrand, J.; Desrumaux, A. Analysis of a New Type of High Pressure Homogenizer. A Study of Flow Pattern. Chem. Eng. Sci. 2004, 59, 843–853. 44. Fonseca, L. P.; Cabral, J. M. S. Penicillin-acylase Release From Escherichia coli Cells by Mechanical Cell Disruption and Permeabilization. J. Chem. Technol. Biotechnol. 2002, 77, 159–167. 45. Gaikar, V. G.; Kulkarni, M. S. Selective Reverse Micellar Extraction of Penicillin Acylase From E. coli. J. Chem. Technol. Biotechnol. 2001, 76, 729–736. 46. Galabova, D.; Tuleva, B.; Spasova, D. Permeabilization of Yarrowia lipolytica Cells by Triton X-100. Enzym. Microb. Technol. 1996, 18, 18–22. 47. Garrido, U.; Banerjee, U. C.; Chisti, Y.; Moo-Young, M. Disruption of a Recombinant Yeast for the Release of b-Galactosidase. Bioseparation 1994, 4, 319–328. 48. Geciova, J.; Bury, D.; Jelen, P. Methods for Disruption of Microbial Cells for Potential Use in the Dairy Industry – A Review. Int. Dairy J. 2002, 12, 541–553. 49. Gervais, P.; Martinez de Maranon, I. Effect of the Kinetics of Temperature Variation on Saccharomyces cerevisiae Viability and Permeability. Biochim. Biophys. Acta 1995, 1235, 52–56. 50. Gervais, P.; Martinez de Maranon, I.; Evrard, C.; et al. Cell Volume Changes during Rapid Temperature Shifts. J. Biotechnol. 2003, 102, 269–279. 51. Harrington, R. E. Intrinsic Viscosity of DNA: Salt Dependence and Correct Electrolyte Theory. Biopolymers 1980, 19, 449–451. 52. Harrison, S. T. L. The Extraction and Purification of PHB from Alcaligenes eutrophus (Ph.D. thesis), University of Cambridge, 1990. 53. Harrison, S. T. L.; Dennis, J. S.; Chase, H. A. The Disruption of Alcaligenes eutrophus by High-pressure Homogenisation: Key Factors Involved in the Process. Bioseparation 1991, 2, 155–166. 54. Harrison, S. T. L.; Chase, H. A.; Dennis, J. S. Combined Chemical and Mechanical Processes for the Disruption of Bacteria. Bioseparation 1991, 2, 95–105. 55. Harrison, S. T. L.; Pandit, A. B. The Disruption of Microbial Cells by Hydrodynamic Cavitation. In 9th International Biotechnology Symposium, Washington, DC; 1992. 1992. 56. Hubbuch, J. J.; Brixius, P. J.; Lin, D.-Q.; et al. The Influence of Homogenisation Conditions on Biomass–adsorbent Interactions during Ion-exchange Expanded Bed Adsorption. Biotechnol. Bioeng. 2006, 94, 543. 57. Johnstone-Robertson, M.; Clarke, K. G.; Harrison, S. T. L. Comparison of the Distribution of Glucose Oxidase in Penicillium sp. CBS 120262 and Aspergillus niger NRRL-3 Cultures. Biotechnol. Bioeng. 2008, 99 (4), 910–918. 58. Kandler, O.; König, H. Cell Wall Polymers in Archaea (Archaebacteria). Cell. Mol. Life Sci. 1998, 54, 305–308. 59. Kee, G. S.; Pujar, N. S.; Titchener-Hooker, N. J. Study of Detergent-mediated Liberation of Hepatitis B Virus-like Particles From S. cerevisiae Homogenate Identifying a Framework for the Design of Future Generation Lipoprotein Vaccine Processes. Biotechnol. Prog. 2008, 24, 623–631. 60. Keshavarz, E.; Bonnerjea, J.; Hoare, M.; Dunnill, P. Disruption of a Fungal Organism, Rhizopus nigricans in a High-pressure Homogenizer. Enzym. Microb. Technol. 1990, 12, 494–498. 61. Keshavarz-Moore, E.; Hoare, M.; Dunnill, P. Disruption of Bakers Yeast in a High-pressure Homogenizer: New Evidence on Mechanism. Enzym. Microb. Technol. 1990, 12, 764–770. 62. Kleinig, A.; Middelberg, A. P. J. On the Mechanism of Microbial Cell Disruption in High-pressure Homogenisation. Chem. Eng. Sci. 1998, 53, 891–898. 63. Kuboi, R.; Umakoshi, H.; Takagi, N.; Komasuwa, I. Optimal Disruption Methods for the Selective Recovery of b-Galactosidase From Escherichia coli. J. Ferment. Bioeng. 1995, 79, 335–341. 64. Laouar, L.; Lowe, K. C.; Mulligan, B. J. Yeast Responses to Non-ionic Surfactants. Enzym. Microb. Technol. 1996, 18, 433–438. 65. Malamy, M. H.; Horecker, B. L. Release of Alkaline Phosphatase From Cells of Escherichia coli Upon Lysozyme Spheroplast Formation. Biochemistry 1964, 3, 1889–1893. 66. Mogren, H.; Lindblom, M.; Hedenskog, G. Mechanical Disintegration of Microorganisms in an Industrial Homogenizer. Biotechnol. Bioeng. 1974, 16, 261–274. 67. Mosqueira, F. G.; Higgins, J. J.; Dunnill, P.; Lilly, M. D. Characteristics of Mechanically Disrupted Bakers’ Yeast in Relation to Its Separation in Industrial Centrifuges. Biotechnol. Bioeng. 1981, 23, 335–343. 68. Naglak, T. J.; Hettwer, D. J.; Wang, H. Y. Chemical Permeabilisation of Cells for Intracellular Product Release. In Separation Processes in Biotechnology; Asenjo, J. A., Ed., Dekker: New York, NY, 1990; pp 177–205. 69. Northcote, D. H.; Golding, K. J. The Chemical Composition and Structure of the Cell Wall of Chlorella pyrenoidosa. Biochem. J. 1958, 70, 391–397. 70. Ramanan, N. R.; Ling, T. C.; Ariff, A. B. The Performance of a Glass Bead Shaking Technique for the Disruption of Escherichia coli Cells. Biotechnol. Bioproc. Eng. 2008, 13, 613–623. 71. Reinert, K. E.; Strassburger, J.; Triebel, H. Molecular Weights and Hydrodynamic Properties of Homogeneous Native DNA, Derived From Diffusion, Sedimentation and Viscosity Measurements on Polydisperse Samples. Biopolymers 1971, 10, 285–307. 72. Ren, X.; Yu, D.; Yu, L.; et al. A New Study of Cell Disruption to Release Recombinant Thermostable Enzyme From Escherichia coli by Thermolysis. J. Biotechnol. 2007, 129, 668–673. 73. Ricci-Silva, M. E.; Vitolo, M.; Abrahao-Neto, J. Protein and Glucose-6-phosphate Dehydrogenase Releasing From Bakers’ Yeast Cells Disrupted by a Vertical Bead Mill. Process Biochem. 2000, 35, 831–835. 74. Sauer, T.; Robinson, C. W.; Glick, B. R. Disruption of Native and Recombinant Escherichia coli in a High Pressure Homogenizer. Biotechnol. Bioeng. 1989, 33, 1330–1342. 75. Save, S. S.; Pandit, A. B.; Joshi, J. B. Microbial Cell Disruption: Role of Cavitation. Chem. Eng. J. 1994, 55, 67–72. 76. Save, S. S.; Pandit, A. B.; Joshi, J. B. Use of Hydrodynamic Cavitation for Large-scale Microbial Cell Disruption. Trans. Inst. Chem. Eng. 1997, 75, 41–48. 77. Scholtz-Brown, N. J.; Pandit, A. B.; Harrison, S. T. L. The Effects of Solids Suspension on Microbial Cell Disruption. BHR Group Conf. Ser. Publ. 1997, 25, 199–215. 78. Schutte, H.; Kroner, K. H.; Hustedt, H.; Kula, M.-R. Experiences With a 20 Litre Industrial Bead Mill for the Disruption of Microorganisms. Enzym. Microb. Technol. 1983, 5, 143–148. 79. Schutte, H.; Kula, M.-R. Analytical Disruption of Microorganisms in a Mixer Mill. Enzym. Microb. Technol. 1988, 10, 552–558.

712

Cell Disruption

80. Smith, A. E.; Zhang, Z.; Thomas, C. R. Wall Material Properties of Yeast Cells: Part 1. Cell Measurements and Compression Measurements. Chem. Eng. Sci. 2000, 55, 2031–2041. 81. Stenson, J. Investigating the Mechanical Properties of Yeast Cells (Ph.D. thesis), University of Birmingham, 2008. 82. Torner, M. J.; Asenjo, J. A. Kinetics of Enzyme Release From Breadmaking Yeast Cells in a Bead Mill. Biotechnol. Technol. 1991, 5, 101–106. 83. Umakoshi, H.; Fukuta, Y.; Kuboi, R. Utilization of Cell Response Under Heat, Chemical and Combined Stresses for Selective Recovery of Cytoplasmic b-Galactosidase From Escherichia coli Cells. Biotechnol. Prog. 1998, 14, 210–217. 84. van Gaver, D.; Huyghebaert. Optimization of Yeast Cell Disruption With a Designed Bead Mill. Enzym. Microb. Technol. 1990, 13, 665–671. 85. Wang, Z.; Le, G.; Shi, Y.; Wegrzyn, G. Studies on Recovery Plasmid DNA From Escherichia coli by Heat Treatment. Process Biochem. 2002, 38, 199–206. 86. Woese, C. R.; Magnum, I. J.; Fox, G. E. Archaebacteria. J. Mol. Evol. 1978, 11, 245–252.

Supercritical Fluid Technology in Bioseperationq

2.48

Ece Yildiz-Ozturk and Ozlem Yesil-Celiktas, Faculty of Engineering, Ege University, Bornova-Izmir, Turkey © 2019 Elsevier B.V. All rights reserved. This chapter replaces E.D. Ramsey, W. Guo, J.Y. Liu, X.H. Wu, 2.74 - Supercritical Fluids, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 1007-1026.

2.48.1 2.48.2 2.48.2.1 2.48.2.2 2.48.2.3 2.48.2.4 2.48.3 2.48.4 2.48.4.1 2.48.4.2 2.48.4.3 2.48.5 2.48.5.1 2.48.5.2 2.48.5.3 2.48.5.4 2.48.5.5 2.48.5.6 2.48.5.6.1 2.48.5.7 2.48.5.8 2.48.6 2.48.7 References

2.48.1

Introduction to Bioseparation Techniques Properties and Fundamentals of Supercritical Fluids Theory of Supercritical Fluids Physicochemical Properties of Supercritical Fluids Commonly Used Supercritical Fluids The Role of Modifier on Supercritical Fluid Extraction Advantages and Disadvantages of Supercritical Fluids Instrumentation of Supercritical Fluid Process Components (Fluid Reservoir, Pump, Reaction/Extraction Vessel, Restrictor, Collector, Detector) The Modes of Supercritical Carbon Dioxide Extraction The Operation Conditions of Supercritical Carbon Dioxide Extraction Applications of Supercritical Carbon Dioxide Extraction Food Applications Extraction and Fractionation of Fats and Oils Decaffeination of Coffee and Tea Dealcoholisation of Alcoholic Beverages Enrichment of Vitamin E From Natural Sources Removal of Pesticides Nutraceutical and Pharmaceutical Applications Biologically Active Compounds From Plant Material Biologically Active Compounds From Algae Future Perspectives Conclusions

713 714 714 714 716 716 716 717 717 718 719 719 719 719 720 720 720 720 720 720 721 723 723 724

Introduction to Bioseparation Techniques

Bioseparation is the practice of purifying biological products using basic principles of engineering, science and technology. The final aim is to obtain purified fractions of molecules, proteins and cells, therefore decreasing the overall cost of the product. Biological products have unique properties and can be separated based on the following properties: size, shape, density, thermal stability, diffusivity, solubility, polarity, volatility and electrostatic charge. Purification of biological products involves consecutive steps requiring the use of various unit operations, such as filtration, membrane processes, extraction, evaporation and drying. The fundamental recovery steps are mostly related to separating the product from microorganism, cells or biomass. The intermediate recovery stages concentrate on the product via certain operational processes that depend on the nature of the product, whereas the final purification stages involve the use of some driving forces such as charge, size or solubility differences. Therefore, the decision on the sequence of different processing and operating conditions depends on the specifications of the final product, impurities, biomasses, microorganisms, cells and tissues. The tendency towards high quality and healthy foods and pharmaceutical compounds and awareness in environmental aspects have been the main driver for implementation of tight regulations regarding the allowable limits of pollutants, synthetic additives and toxins. Consequently, tremendous efforts have been devoted to the design and development of alternative products fulfilling consumer requirements and green technologies and cost effective processes to manufacture these value- added products. One approach was to utilize existing technologies such as microwave, ultrasound and high pressure to replace conventional processes depending on the use of organic solvents, which are toxic, flammable and expensive. Indeed, the last couple of decades witnessed the emergence of these technologies as green alternatives. Among them is supercritical fluid extraction (SFE) technology for the production of specialty oils, plant extracts, oleoresins, flavors, colorants, and fragrances utilized by various sectors such as food, pharma, feed and cosmetics. From numerous studies of supercritical fluids, more and more information is available about effects of temperature and pressure on solvation properties. Carbon dioxide (CO2) is the most commonly used gas for supercritical fluid processing. As carbon dioxide is neither a liquid nor a gas above its critical pressure and temperature,

q

Change History: March 2018. Ece Yildiz-Ozturk, Ozlem Yesil-Celiktas wrote a new chapter replacing the previous edition version.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00122-1

713

714

Supercritical Fluid Technology in Bioseperation

it has no surface tension, which qualifies to be an extremely good ‘nonpolar solvent’ and therefore applicable for extraction of oils, micronization of nonpolar pharmaceutical compounds, drying of scaffolds, hydrogels and several other applications. The superiority of the technique is that the processed compounds remain intact as there is no thermal decomposition and the final concentrate is free of any residual solvent due to carbon dioxide’s natural tendency being a gas, which becomes volatile at ambient temperature and pressure. During processing, supercritical CO2 carries the compounds of interest to the separator and leaves the vessel as a gas once reaching the separator at ambient pressure. Then the pressure is released, leaving the processed phase completely free of solvents. Therefore, no further purification steps are required, which reduces downstream processing costs dramatically. Although comparatively high initial investment costs, industrial scale supercritical fluid processing plants are in operation for different applications such as fractionation, sterilization, particle formation, impregnation, dyeing and leather tanning. In this chapter, the focus is on properties and fundamentals of supercritical fluids, instrumentation of supercritical fluid process and various applications, particularly the extraction of biologically active compounds from medical plants and microalgae is elaborated in more details as a bioseperation approach.

2.48.2

Properties and Fundamentals of Supercritical Fluids

2.48.2.1

Theory of Supercritical Fluids

A supercritical fluid is an intermediate state of a matter between a gas and liquid in terms of its physicochemical properties. Supercritical state is reached by subjecting a liquid or a gas to temperature and pressure exceeding the critical point specific to that matter, referred as the critical temperature and pressure. By definition, the temperature at which a gas cannot become liquid except for applying additional pressure is called the critical temperature, whereas the minimum pressure required to liquefy a gas is called as the critical temperature. Supercritical fluids possess superior properties of gas and liquid phases, as it can act like both a gas due to its diffusivity and a liquid because of its density and solvation capability.1 Table 1 depicts physicochemical properties of supercritical fluids in comparison to gases and liquids. The supercritical state is formed as a result of a dynamic equilibrium where molecules diffuse in and out of liquid and gas phases by gaining and loosing energy. Eventually, the phase curve between liquid and gas phases disappears and supercritical phase becomes apparent.2 An ideal phase diagram is depicted in Fig. 1. The supercritical region is above the critical temperature (Tc) and critical pressure (Pc), which is specific to each fluid.

2.48.2.2

Physicochemical Properties of Supercritical Fluids

The physicochemical properties of supercritical fluids lie between that of gases and liquids. The density of a supercritical fluid is closer to that of organic liquids and can be manipulated by altering the temperature and pressure near the critical point.3 The effects of pressure and temperature on density of supercritical CO2 are depicted in Fig. 2. The density of supercritical CO2 (SC–CO2) decreases with increased temperature at constant pressure, whereas increases with increasing pressure at constant temperature in Table 1

Supercritical fluid properties in comparison to gases and liquids Supercritical fluid

Liquid

0.6  10
3 – 2.0  10
3 0.1–0.4 1  10
4 – 3  10
4

0.2–0.5 10
3 – 10
4 1  10
4 – 3  10
4

0.6–2.0 0.2  10
5 – 2.0  10
5 0.2  10
2 – 3.0  10
2

Pressure

Density (g cm
3) Diffusivity (cm2 s
1) Viscosity (cm s
1)

Gas

solid phase

compressible liquid

critical pressure Pcr

Ptp

liquid phase

triple point

supercritical fluid

critical point

superheated vapour gaseous phase Ttp

critical temperature Tcr Temperature

Figure 1

The schematic representation of a phase diagram for supercritical fluids.

Supercritical Fluid Technology in Bioseperation

715

Density (kg/m3)

1000 900

283 K 293 K 303 K

800

313 K

700

323 K

600

333 K

500 400 300

403 K

200 100 0 20 Figure 2

40

60

80 100 Pressure (bar)

120

160

140

The variation in density of carbon dioxide based on pressure and temperature, “black dot” represents the critical point.

the supercritical region. It is also worth to mention that the solubility strength of a supercritical fluid is based on the density, making supercritical fluids superior to gases. If SC-CO2 is entrained with a co-solvent, then the density of CO2 and co-solvent mixture at supercritical conditions can be calculated using a specific software such as the phase equilibria programme developed by the group of Professor Brunner formerly from the Institute of Thermal Separation Processes, University of Hamburg-Harburg. The viscosities and diffusivities of supercritical fluids are closer to that of gases. Therefore, supercritical fluids can diffuse faster than liquids. The effects of temperature and pressure on viscosity of supercritical CO2 are depicted in Fig. 3. Viscosity increases sharply with increasing pressure at constant temperature, whereas the increasing trend becomes much lower above 313 K. On the other hand, the viscosity decreases with increasing temperature at constant pressure above critical point. Diffusivity increases with temperature and decreases with pressure as the space between molecules becomes tighter. Again if a co-solvent is to be used, then the viscosity of CO2 and co-solvent mixture can be calculated by emperical approaches such as Kendall-Monroe equation, which is as follows: 1

1

1

m3mix ¼ xCO2 m3CO2 ¼ xco
solvent m3co
solvent x is the mole fraction of mixture components. 120 20 Bar 50 Bar

100

Viscosity 10-6 (Pa.s)

80 Bar 110 Bar

80

140 Bar 60

40

20

0 280 Figure 3

300

320

340 360 Temperature (K)

The variation in viscosity of carbon dioxide based on temperature and pressure.

380

400

420

716

Supercritical Fluid Technology in Bioseperation

The high diffusivities achieved by supercritical fluids make them faster mediators for bioseparations such as chromatography and extraction.

2.48.2.3

Commonly Used Supercritical Fluids

Carbon dioxide is the most common fluid utilized in supercritical applications due to its low critical temperature (304 K) and pressure (73.8 bar). Besides, CO2 is inert, non-toxic, easily available, cheap and a high solvating fluid for nonpolar compounds. Apart from CO2, ammonia, n-butane, diethyl ether, dichlorodifluoromethane, ethane, nitrous oxide, and tetrahydrofuran are used for various applications. Table 2 demonstrates selected solvents and their critical properties. Most of the solvents other than CO2 have actually higher critical temperature or pressure that leave only few being potential choices for bio-based applications.

2.48.2.4

The Role of Modifier on Supercritical Fluid Extraction

Although carbon dioxide is the most preferred fluid in supercritical applications, processing of polar compounds is not feasible. Therefore, the strategy in supercritical CO2 (SC-CO2) extraction of high molecular weight and polar compounds is to add polar solvents which are known as modifiers or co-solvents. The solubility of these type of compounds can be increased by adding cosolvents at a concentration ranging between 1% and 10 % of the total flow rate, which in turn enhances the solvating power of SC-CO2 entrained with the co-solvent. Ethanol which is a “generally regarded as safe” solvent and water are the most appropriate co-solvents for bio-based applications. There are two approaches for addition of co-solvent to the system based on the technical capabilities of the high pressure equipment. One approach is pumping co-solvent and CO2 separately to a mixing unit allowing the merge of two fluids prior to entering the extraction vessel, while co-solvent can be directly added to the extraction vessel interacting with the ground material in the extraction vessel. This latter approach requires a static extraction step where the co-solvent can diffuse into solid particles, solubilize the analytes and diffuse out of the particles, releasing them into the supercritical fluid phase.5

2.48.3

Advantages and Disadvantages of Supercritical Fluids

Supercritical fluids (SCFs) possess favorable properties that make them suitable for bioseparation processes. Some of the major advantages are as follows: (1) SCFs have superiority over liquid solvents in terms of mass transfer characteristics such as high diffusion and low surface tension, enabling easy penetration into porous solid matrices and high solvation power, whereas low viscosities allow for better flow.6 (2) Manipulation of pressure and temperature above critical points affects physicochemical properties of SCFs such as density, viscosity, diffusivity, and dielectric constant and enhances the extractability of SCFs.7 The tunable solvation power of SCFs may lead to a remarkably high selectivity.8 (3) SC-CO2 has relatively low critical temperature and pressure, which makes it an ideal candidate for extraction of thermally labile and oxidizing compounds.9 (4) SC-CO2 dissolves nonpolar or slightly polar compounds. However, the selectivity of the process and solubility of polar compounds can be leveraged by adding co-solvents, which enhances yields and decreases processing times, while offering milder processing conditions.3 (5) The solvation power of SC-CO2 for low molecular weight compounds is high but decreases with increasing molecular weight. However, it is possible to separate less volatile and high molecular weight compounds by increasing pressure.2 (6) Products manufactured by SC-CO2 and sub-, supercritical water processes offer health benefits, as they are nontoxic, non-carcinogenic and stable. Indeed, it is approved by Food and Drug Administration in USA and European Food Safety Authority as a “generally recognized as safe” process.10 (7) Contrary to classical extraction methods, solvents can be completely removed from the extract phase by depressurization, and better stability of compounds can be maintained due to lack of organic solvent.11 The use of SC-CO2 for extraction allows isolation and fractionation of compounds from total extracts at mild operating temperatures.12 One of the drawbacks of supercritical fluid processing is related to the high investment and maintenance costs due to the high pressures applied. Consequently, the prices are relatively high in comparison to the ones manufactured by conventional processes. Table 2

Critical properties of various solvents commonly used as supercritical fluids4

Solvent

Molecular weight (g mol
1)

Critical temperature (K)

Critical pressure (bar)

Critical density (g cm
3)

Carbon dioxide (CO2) Water (H2O) Ethanol (C2H5OH) Methanol (CH3OH) Ethylene (C2H4) Ethane (C2H6) Methane (CH4) Propane (C3H8) Propylene (C3H6) Acetone (C3H6O)

44.01 18.02 46.07 32.04 28.05 30.07 16.04 44.09 42.08 58.08

304.1 647.3 513.9 512.6 282.4 305.3 190.4 369.8 364.9 508.1

73.8 221.2 61.4 80.9 50.4 48.7 46.0 42.5 46.0 47.0

0.469 0.348 0.276 0.272 0.215 0.203 0.162 0.217 0.232 0.278

Supercritical Fluid Technology in Bioseperation

717

Another disadvantage of SC-CO2 is the nonpolar nature, and there are circumstances that the addition of 10%–20 % co-solvent might not be sufficient for extraction of target compounds, which are highly polar. Therefore, a logical strategy can be to focus on alternative approaches to decrease the polarity of the compounds of interest in order to increase the solubility. For this purpose, the selectivity of the process has been improved by chemical in situ derivatization of a group of compounds.5

2.48.4

Instrumentation of Supercritical Fluid Process

2.48.4.1

Components (Fluid Reservoir, Pump, Reaction/Extraction Vessel, Restrictor, Collector, Detector)

As CO2 is the most widely used fluid both at lab and industrial scale supercritical applications, different components for instrumentation will be elaborated specifically for supercritical CO2 process. Holistically, the system is comprised of a CO2 tube, high pressure CO2 and modifier pumps, CO2 flow meter, a reaction/extraction vessel to place the sample, a collecting vessel, temperature controllers, Automated Back Pressure Regulator (ABPR), cooler and heater (Fig. 4). The liquid is pumped and heated to supercritical conditions at a heating zone. Liquefied CO2 is introduced into the extraction vessel and diffused into the solid matrix, subsequently dissolving the compounds to be extracted. Both the pressure and temperature of the cartridge automatically reach and maintain the set conditions by a control unit. The dissolved compounds are transported to the bulk phase from the extraction vessel into the collector at lower pressure, and the extract phase can be collected. At the end of the extraction process, CO2 can be cooled, recompressed and recycled, or discharged. Fluid reservoir: Carbon dioxide is stored in aluminum or stainless steel cylinders, which are equipped with a pressurized headspace, dip tube, and a cooling device to ensure regular fluid delivery. A high pressure pump delivers the supercritical fluid at a controllable pressure and flow rate. CO2 is usually pumped as a liquid subsequent to pre-cooling at 4 C, and an additional pump is required if a co-solvent is to be used. A wide range of pressure between 75 and 750 bar can be applied along with non-pulsating flow rates. Reciprocating CO2 pumps or syringe pumps are often employed for small-scale extractions, whereas diaphragm pumps are most commonly used for large scales. The pump heads and CO2 are usually cooled before entering the pump.13 Extraction vessel is used to hold the sample during the process with typical volumes ranging from 0.1 to 100 mL. The frit lids prevent SC-CO2 from sweeping the solid sample. The vessel is exposed to high pressure and is equipped with a safety valve for protection if the system malfunctions. The temperature of the chamber is controlled by a heating tube. Controller ensures that the flow rate of SC-CO2, temperature and pressure of the extraction column are maintained at the set values. More functions are available in commercial systems such as valve switching and extraction time setting, which provides automatic operation. Collector has a role in maintaining the required pressure in the extraction vessel. A restrictor is connected between the vessel outlet and the collector inlet. The obtained extracts are commonly collected using an appropriate co-solvent, a cryogenic trap or alternatively a cartridge packed with adsorbent material.14

Mixer

CO2 pump Extraction vessel

Co-solvent pump

CO2 cylinder Figure 4

Seperator

Co-solvent reservoir

A schematic depiction of a laboratory scale supercritical CO2 extraction system (Thar Instruments, Inc., UK).

718

Supercritical Fluid Technology in Bioseperation

Heating and cooling: Carbon dioxide is cooled to maintain liquid conditions before pumping, and heated after pressurization. Heat is provided to prevent excessive cooling as CO2 is expanded into the separator. It is sufficient to preheat CO2 in the tubing inside the oven containing the extraction column for small-scale extractions, as in the case for analytical applications.13

2.48.4.2

The Modes of Supercritical Carbon Dioxide Extraction

The operating mode of SC-CO2 extraction is dynamic (continuous) in regard to continuously supplied SC-CO2 but on the contrary static (batch) in regard to feed material placed in reaction/extraction vessel. When the process starts, fresh SC-CO2 is supplied to the extraction column, while the collector is supplied with the solute continually. On completion of the process, extract is collected from the collector outlet after releasing CO2 at laboratory scale, whereas CO2 is transferred to an equilibrium tank to be recycled and used in the process at industrial scale (Fig. 5A). Although recuperation of non-volatile compounds do not require specific approaches, slightly volatile compounds are suggested to be led by aerosol formation, whereas sorbent trapping is reported to be favorable for highly or moderately volatile compounds. An extruder based SC-CO2 extraction system was designed as a continuous process. The feed material and SC-CO2 are fed to this extruder based system concurrently. SC-CO2 dissolves the solute while interacting with the feed and leaves the system continuously as enriched by the solute, whereas the feed is deprived of the solute (Fig. 5B). This countercurrent process might be of great interest as continuous processes are much favored in industry. There are several approaches of recuperating the extract phase after depressurization of the supercritical fluid via thermal trapping, sorbent and solvent trapping. Among them, thermal trapping is the simplest method as the supercritical fluid is depressurized in a cooled recovery container. However, this method is limited for the recovery of non-volatile components.15

A

CO2 + Solute Collector

Pressure reduction CO2 Extraction column

Extract

Supercritical CO2

CO2 Low pressure

CO2 Pump CO2 equilibrium vessel

B

Supercritical CO2

Feed

CO2

Supercritical fluid extruder

Extract phase

Figure 5 Schematic diagrams of industrial scale supercritical CO2 extraction with continuous CO2 flow (A) and extruder-based supercritical CO2 extraction with continuous flows of CO2 and the feed material concurrently (B).

Supercritical Fluid Technology in Bioseperation 2.48.4.3

719

The Operation Conditions of Supercritical Carbon Dioxide Extraction

The extraction and fractionation of products with supercritical fluids can be implemented under two operation models: selective extraction and selective separation. The selective extraction includes the solvation capacity of the fluid used in the extraction through the manipulation of the operational conditions such as temperature and pressure, along with the modification of the chemical nature of the solvent through addition of a modifier. The selective separation is obtained through an adjustment in the depressurization, which can be single or serial, allowing a controlled fractionation of the extractable products.16 The choice of experimental method for the process of samples depends on mainly the physical-chemical nature of the substances, the type and structure of the raw material from where the desired compound will be extracted. The impact of extraction variables has been evaluated in regards to the recovery, yield, and composition of the extract phase from different sources. Independent variables for extraction are moisture content of feed material, particle size, temperature, pressure, CO2 flow rate, extraction time, and solvent-feed ratio, which should be optimized to maximize the yield. Preparation of plant materials is a critical step for SFE process. Fresh plant matrices have high moisture content, which can reduce the efficiency of the process, thus a drying step is required,17 followed by grinding to achieve an average particle size between 0.5 and 0.005 mm. If the size is smaller than 0.005 mm, then the extraction vessel can be filled with some inert materials such as glass wool or sand. The next step is to decide on process parameters among which pressure and temperature are of prime importance as the solvent properties of supercritical fluids can be tuned by changing the values of these parameters, directly influencing density. Extraction yield increases with pressure, which is a result of increase in lipid solubility due to an increase in CO2 density. The majority of the studies indicated that an increase in the pressure of SC-CO2 leads to an increase in extracted bioactive compounds. On the other hand, the density of CO2 is reduced with increasing temperature at constant pressure, leading to a reduced solvation power.18 The impact of temperature is dependent on competing parameters; CO2 density decreases with temperature while the vapor pressure of solutes increases. A high temperature would result in lower extraction yield for a non-volatile compound due to a decrease in solubility. In addition, temperature increase may also cause breakdown of cell structure and increase the diffusivity of the targeted compounds, therefore accelerating the extraction process. Thus, the impact of temperature on solubility is governed by whichever parameter is greater at a given pressure. Extraction time is another important parameter that needs to be optimized in order to maximize the amount of bioactive compounds. It is necessary to prolong the contact time of SC-CO2 with the sample material to increase the efficiency of SFE process.17 During the process, CO2 flow rate should ideally be measured in terms of mass flow rather than volume because of the fact that the density of CO2 changes based on the applied temperature both before entering high pressure pump and during compression. For maximization of extraction rate, high flow rates can be applied for the process to be completely diffusion limited, but excessive amounts of solvent will be used, which will not be feasible at all. On the other hand, the process should be completely solubility limited to minimize the amount of solvent used, which means prolonged process durations. Resultantly, the flow rate should be determined based on the competing factors of time and solvent costs.13 Solvent/feed (S/F) ratio is a function of flow rate and time and is an indicator of solvating power. S/F ratios up to 30 to 50 are meaningful for an efficient SFE process. Operating at these S/F ratios becomes very challenging for extraction of natural bioactive compounds due to polar nature. Under these circumstances, CO2 alone is not sufficient, thus modifiers are added at about 1% to 10 % during extraction. Ethanol is a widely used modifier, which is generally regarded as safe and expands the range of compounds attainable by increasing the polarity of the supercritical mixture.17

2.48.5

Applications of Supercritical Carbon Dioxide Extraction

2.48.5.1

Food Applications

The most widespread application of supercritical fluid extraction is in the food field extending the implementation of the technology to industrial scale. CO2 is the most commonly utilized supercritical fluid in the food industry. Thermally labile food components can be extracted owing to the non-toxic nature of CO2 along with its low critical temperature, and the end product is free from residual solvent. Further, color, composition, odor, and texture of the extract are controllable, and extraction by SC-CO2 maintains the aroma of the product. Common food applications are covered in the following section.

2.48.5.2

Extraction and Fractionation of Fats and Oils

Specialty oils have high concentrations of biologically active lipid components, such as polyunsaturated fatty acids, tocopherols, tocotrienols, squalene, phytosterols and carotenoids, which have proven to exhibit various health benefits. They are classified under nut oils, extracted from peanut, hazelnut, walnut, pecan, almond and pistachio; seed oils from cherry, grape, apricot, hiprose, pumpkin, sesame, borage, sea buckthorn, flax, and evening primrose; cereal oils from wheat germ, rice bran, oat, and amaranth; and oils of fruits and vegetables such as carrot, tomato, olive husk and cloudberry. Specialty oils are of high value but small volume in contrary to the commodity oils such as corn, soybean, sunflower and canola. SC-CO2 extraction preserves the unique aroma and flavor of these specialty oils, whereas volatile aroma compounds cannot be retained in conventional processes.18 Removal of fats from food substances has created another niche by producing fat free or reduced fat products as in the case of potato chips. The amount of remaining fat in the potato chips is reported to be easily controlled based on the expected taste.19 Furthermore, the use of crude vegetable oils are common in the food industry for various applications. Before consumption, fatty acids and oils have to be refined in order to remove undesirable compounds. Otherwise, valuable compounds present in the oils can also be lost during this refining process. SFE has been suggested as an alternative refining process, which can yield extracts enriched with

720

Supercritical Fluid Technology in Bioseperation

targeted compounds. For instance, processing of wheat germ oil, crude palm oil and rice bran oil are typical applications.20 Apart from these, polar phospholipids can be recovered from oil seeds by adding ethanol as a modifier.18 Industrial scale facilities are located in Dusseldorf, Germany, for SC-CO2 extraction of fatty acids from spent barley, one in Hubei, China, for phytosterol, fatty acid, methyl ester, ginger oil, and another one in Soke, Turkey, for wheat germ, almond and black cumin oils.

2.48.5.3

Decaffeination of Coffee and Tea

Extraction of caffeine from green tea using supercritical CO2 entrained with water provides selectivity and retains antioxidant compounds in the matrix.20 Industrial scale facilities are located in Houston, Texas, USA, Bremen, Germany and Pozzallo, Italy, for decaffeination of coffee and in Münchsmünster, Germany, for decaffeination of tea.

2.48.5.4

Dealcoholisation of Alcoholic Beverages

Alcoholic beverages such as wine and beer can be de-alcoholized by the removal of ethanol from water. Distillation is a common process but poses a disadvantage of losing aroma compounds. The application of SC-CO2 and pervaporation can overcome this drawback. The procedure starts with an aqueous solution at a concentration of about 10% (w/w) ethanol, which is then removed by SC-CO2 in a stripping column. Temperature is the major factor affecting the rate of ethanol removal. Flow rates can be optimized to achieve efficient mass transfer. For instance, reduction of alcohol content to 0.5% (w/w) was reported to require about 2.5 h at 45  C.19

2.48.5.5

Enrichment of Vitamin E From Natural Sources

Supercritical CO2 processing is ideal for enrichment of tocochromanols, owing to lower operating temperatures. Most commonly used feed materials are soybean oil deodorizer distillate, which has about 50% tocopherols and crude palm oil, which contains tocopherols and tocotrienols at a concentration of about 500 ppm. Both oils can be utilized for enrichment purposes. Although direct recovery of tocochromanols from crude palm oil is possible, producing esters of triglycerides is suggested in order to separate these compounds more efficiently from tocochromanols. Therefore, triglycerides are subjected to an esterification with methanol to form fatty acid methyl esters, which can be extracted with SC-CO2. Subsequently, tocochromanols and other unsaponifiable compounds such as squalene and sterol are enriched at the bottom phase of the extraction column.19

2.48.5.6

Removal of Pesticides

Strict regulations have been established to limit the presence of pollutants in the food for the ultimate aim of protecting the health of consumers. Therefore, the rapid and precise detection of food pollutants have become vital for regulatory agencies, administrations and control laboratories. Generally, the analysis of food pollutants requires long extraction durations and cleanup procedures employing large volumes of toxic solvents. Alternative techniques have been developed based on supercritical fluid technology enabling reduced sample preparation times and minimizing the use of organic solvents. SC-CO2 modified with 10% acetonitrile is employed as a single clean-up step for analyses of pesticides such as organochlorine, organonitrogen, organophosphorus and pyrethroid in potatoes, tomatoes, lettuces, apples and honey. Further studies highlighted the use of SC-CO2 as a clean-up step for analyses of pesticide residues in fish muscle, cereals, vegetables, vegetable canned soups, infant and diet foods.19 Currently, SC-CO2 is utilized for the extraction of pesticides from solid matrices such as contaminated soil. Additionally, the process can be tailored for the extraction of pesticides containing more polar groups by adding co-solvents.9

2.48.5.6.1

Nutraceutical and Pharmaceutical Applications

Natural supplements have become one of the most important trends in the nutritional industry today due to the increasing evidence correlating diet and chronic illnesses and the limitation imposed on the application of synthetic chemical based nutraceutical additives. A paradigm shift is observed among pharmaceutical as well as chemical industries leading to replacement of traditional practices with green technologies in order to manufacture products free from residual solvents. Apart from consumer awareness, stringent environmental regulations also dictate utilization of green technologies. Therefore, new processes have been developed using supercritical fluids with low environmental impact by reducing the volatile organic compounds in drug manufacturing and eliminating residues in the final product.16 Industrial scale facilities are located in Edmonton, Canada, Wisconsin, USA, Tsukaba, Japan, Münchsmünster and Rehlingen, Germany.

2.48.5.7

Biologically Active Compounds From Plant Material

Recently, there has been increasing interest in discovering new bioactive compounds, which possess therapeutic properties such as antioxidant, antifungal, antibacterial, anti-inflammatory, antiviral, anti-mutagenic, anticancer, antithrombotic, anticholinesterase and cardiovascular.20,21 Much of the efforts are directed to process optimization and incorporation of extracted compounds into food and feed for functionalization purposes. An example is the study in which SC-CO2 and solvent extracted rosemary (Rosmarinus officinalis) were added to wheat germ oil as natural antioxidants to retard oxidation, and subsequently sensory properties were

Supercritical Fluid Technology in Bioseperation

721

investigated as well. SC-CO2 extract was reported to perform better in terms of both aspects.22 Another study reported SC-CO2 extraction of strawberry (Arbutus unedo L.) fruits, which yielded total phenol content of 25.72 mg gallic acid equivalent (GAE)/g extract and 99.9% radical scavenging capacity, higher than the values obtained by conventional ethanol (15.12 mg g
1; 95.8%) and water (24.89 mg g
1; 83.8%) extractions.23 A similar study focused on extraction of peach leaves by SC-CO2 as a potential source of phenolic compounds. Due to the polar nature of phenolic compounds, SC-CO2 was entrained with 6% ethanol at 150 bar and 60  C, which yielded 79.92 mg GAE total phenols/g extract and an EC50 value of 232.20 mg mL
1.24 SC-CO2 extraction of unsaturated fatty acids from Pistacia terebinthus was optimized in order to maximize the content of oleic, linoleic and linolenic acids.25 It is worth to mention that a-linolenic and linoleic acids are not synthesized in the body, therefore the only way to acquire these essential fatty acids is through daily diets. From this perspective, the findings are significant and the optimized process conditions can be translated to industrial scale. Alkannin, an anticancer compound, was extracted using SC-CO226 and incorporated into yoghurt at a concentration of 100 mg in order to obtain a functional yoghurt. No significant changes were reported in pH values, and viable counts of L. delbrueckii ssp. bulgaricus were better than that of control, which indicate a decrease in bitter peptide generation.27 Flavonoids and anthocyanins are health beneficial compounds. SC-CO2 was employed to extract total flavonoids from Echinacea purpurea28 as well as anthocyanin contents from blueberries.29 Biological activities of SC-CO2 extracts were higher than those of solvent extracts in both studies. The demand for low-calorie natural substances has created a niche market for stevia leaves, and the use of rebaudioside A and stevioside has been approved in many countries around the world. Steviol glycoside composition of Stevia rebaudiana leaves was extracted using SC-CO2 where pressure, temperature and co-solvent ratio were optimized. Total glycoside content was close to that obtained by conventional water extraction and slightly higher than ethanol extraction.30 Various plant materials extracted with SC-CO2 are listed in Table 3, along with target compounds and therapeutic properties. In all cases, supercritical conditions were attained where the pressure was between 80 and 450 bar, whereas the temperature was in the range of 40–60  C with couple of exceptions at 80  C, particularly for extraction of high polarity compounds. Besides plant materials, utilization of renewable bio-wastes has been of great interest as well. Among these, forestry wastes are the least utilized bio-based materials. In an attempt to investigate the extractability of pine bark using SC-CO2, both lab and pilot scale studies were carried out where the major compounds were catechin, epicatechin, catechin gallate and taxifolin.32 As for marine bio-wastes, detached leaves of seaweeds reaching the beaches create nuisance. Zostera marina and Posidonia oceanica are the two widespread species in the Mediterranean basin. Phenolic compounds from Z. marina33 and P. oceanica34 residues were extracted by SCCO2. Caffeic, p-coumaric, chicoric, benzoic, ferulic and rosmarinic acids were the major phenolic compounds in the extract phase.

2.48.5.8

Biologically Active Compounds From Algae

Algae are a diverse group of micro- and macroorganisms in marine and fresh water containing biologically active compounds involved in processes of growth, development, and protection. The vast array of bioactive compounds in algae is the result of adaptation to unfavorable environmental conditions and synthesis of these compounds increases in the presence of environmental stress factors. As, bioactive compounds produced by microalgae can be intra- or extracellular, bioseparation strategies are formed based on this aspect. Dunaliella salina is one of the most studied microalgae containing high amounts of b-carotene of about 14% of its dry weight, followed by Chlorella vulgaris rich in carotenoids, lutein and fatty acids. Besides, SC-CO2 extraction of g-linolenic acid, tocopherols and vitamin E from Spirulina platensis was studied as well.20 Bioactive compounds extracted from several algae and microalgae sources using SC-CO2 extraction are summarized in Table 4. In the presented examples, SC-CO2 was chosen as a solvent, occasionally supported by a modifier such as ethanol or combinations of ethanol and water. Operational conditions ranged within 40–85  C and 78.6–500 bar. In general, SC-CO2 was particularly used for extraction of pigments, polyphenols, carotenoids like b-carotene and lutein, vitamins, and lipids including polyunsaturated fatty acids such as omega-3 fatty acids, docosahexaenoic acids and eicosapentaenoic along with omega-6 fatty acids: g-linolenic, linoleic and arachidonic acids. Besides bioactive compounds, proteins from algal sources are also drawing attention. A systematic optimization approach was reported for SC-CO2 extraction of phycocyanin from Sp. platensis, and the effects of process variables such as pressure, temperature, co-solvent ratio and time were investigated. Additionally, both SC-CO2 and solvent extracts were tested against lung cancer cell line (A549) yielding IC50 values of 26.82 mg mL
1 and 36.94 mg mL
1, respectively. SC-CO2 extract exhibited higher cytotoxicity in comparison to the solvent extract, which might be related to higher purity and possible conformational changes in protein structure of phycocyanin subjected to high pressure.35 Indeed, the change in pressure and temperature was reported to positively alter the activity and stability of proteins.36 In another study, an optimized SC-CO2 method was developed for extraction of hydrocarbons from Botryococcus braunii. Hydrocarbon yields were maximized at a pressure of 200 bar and a CO2 flow rate of 8.71 g min
1, which yielded 147.5 mg hydrocarbon/g dry microalgae.37 Optimized process conditions can be used for industrial scale applications as efficient production of hydrocarbons has a significant implication on conversion to green naphtha and further to ethylene and propylene by cracking processes. Although the breadth and intensity of research related to SC-CO2 processing of algae is at a certain level, the direct extraction of algal cultures has so far not been studied in-depth. The algae cultivation should be integrated with downstream processes for product recovery in order to have a sustainable production platform. Ideally, the algae should not be harvested, but rather the product should be isolated while cultivation and growth is continued particularly for those species having slow doubling time so that secondary metabolites of interest can be extracted while keeping the algae viable, being able to produce more. However, the application needs research and development activities to minimize the impact of extraction conditions on the living algal culture. In an attempt to fill this gap, the effect of pressure on the viability of B. braunii cells was investigated under SCCO2 conditions and hydrostatically. The results indicated that SC-CO2 was lethal, whereas cell viability of algae has not been altered

722

SC-CO2 extraction of biologically active compounds from various plant materials16,20,21,31

Plant material

Compounds of interest

Functional bioactivities

Extraction conditions

Bidens pilosa Apium graveolens

Polyacetylenes Sedanolide, sedanenolide, 3-n-butylphthalide

CO2, 40  C, 250 bar, 15 g min
1, 240 min CO2, 40  C, 100 bar, 0.30 kg h
1

Ceratonia siliqua

Cinnamic acid, ferulic acid, naringenin, chrysoeriol, tricetin-3‘,5‘-dimethyl ether

Mentha spicata

Flavonoids (catechin, epicatechin, luteolin, rutin, apigenin, myricetin, naringenin) Total phenols, antioxidants, total anthocyanins Proanthocyanidins, total phenolic content

Anticancer (MCF-7 breast cancer cells) Antimicrobial (S. aureus, L. monocytogenes and Listeria ivanovii strains) Anticancer (rat N1E-115 neuroblastoma cells, MCF-7 breast cancer cells, human HeLa cervical cancer cells) Antioxidant Antioxidant Antioxidant

CO2, 45  C, 160 bar, 2 mL min
1, 30 min, 6% ethanol CO2, 40  C, 80 bar, 4–6 kg h
1, 180 min, 7.5%–10% ethanol/water (57% v/v) CO2, 40  C, 300 bar, 2 mL min
1, 30 min (static) þ 40 min (dynamic) CO2, 50  C, 500 bar, 2 L min
1, 90 min CO2, 40  C, 150–300 bar, 60 g min
1, 60 min (static1) þ 120 min (static2), 0%–7% ethanol CO2, 40–60  C, 130–250 bar, 3 mL min
1, 30 min (static) þ 120 min (dynamic) CO2, 40  C, 100, 300 bar, 50 g min
1, 480 min CO2, 50  C, 300 bar, 0.3 kg h
1, 270 min CO2, 211 bar, 80  C, 60 min

Vitis labrusca Grape marc Piper nigrum L. Phormidium valderianum Rosemary Cyperus articulatus L. var. articulatus

b-Caryophyllene, sabinene, limonene, a-pinene, b-pinene Total phenolics, total carotenoids Carnosic acid, carnosol

Antioxidant Antioxidant Anticancer (HT-29 human adenocarcinoma)

Salvia officinalis Cordia verbenacea Stevia rebaudiana

Terpenes (corimbolone, mustacone, caryophyllene oxide, a-ciperone) Camphor, 1,8-cineole borneol, a-Caryophyllene, b-caryophyllene Glycosides

Cynanchum paniculatum

Paeonol

Cuscuta reflexa Black cumin (Nigella sativa) Ginger (Zingiber corallinum Hance) Anoectochilus roxburghii Baccharis dracunculifolia Chamomile (Matricaria chamomilla) Eugenia uniflora fruits Pinus sp. Rosehip (Rosa canina)

Coumarin Essential oil Essential oil Phytosterols Phenolics Essential oil Carotenoids Flavonoids Carotenoids, fatty acids

Antibacterial (Staphylococcus aureus) Antifungal (Cladosporium sphaerospermum) Anti-inflammatory (THP-1 atherosclerotic) Antitumor (COX-2 and MCF-7 cells) Anti-inflammatory, hypoglycemic, hypotensive, diuretic Anti-inflammatory, cardiovascular protective, antidiabetic Anti-fungal, anticancer Antimicrobial Antipyretic Phytosterols Antioxidant Anti-inflammatory, anti-spasmodic Antioxidant Antioxidant Antioxidant

Hibiscus cannabinus Bamboo shavings

Oil Triterpenoids

Antioxidant Antifatigue

CO2, 40  C, 220 bar, 0.29 kg h
1, 450 min CO2, 60  C, 200 bar, 15 g min
1, 60 min, 3 gethanol/min

CO2 þ methanol, 150 bar, 55  C, 20 min (static) þ 90 min (dynamic) CO2, 248 bar, 55  C CO2, 400 bar, 40  C, 35 min CO2 þ methanol, 100 bar, 30  C, 40 min CO2 þ ethanol, 250 bar, 45  C, 1 h (static) þ1h (dynamic) CO2, 400 bar, 60  C, 20 min CO2, 250 bar, 40  C, 90 min (dynamic) CO2, 250 bar, 60  C, 120 min (dynamic) CO2 þ ethanol (3%, v/v), 200 bar, 40  C CO2, 450 bar, 80  C, 150 min CO2, 450 bar, 80  C, 3h (dynamic) CO2, 200 bar, 80  C, 150 min CO2 þ co-solvents, 250–350 bar, 50–65  C

Supercritical Fluid Technology in Bioseperation

Table 3

Supercritical Fluid Technology in Bioseperation Table 4

723

SC-CO2 extraction of biologically active compounds from algae and microalgae21,31

Plant material

Compounds of interest

Functional bioactivities

Extraction conditions

Haematococcus pluvialis

Astaxanthin

Antioxidant

Nannochloropsis oculata

Zeaxanthin, carotenoids, triglycerides Total polyphenol and flavonoid content

Phagocytotic activity

CO2, 65  C, 435 bar, 166 mL min
1, 210 min, 2.3 mL ethanol/gsample CO2, 50  C, 350 bar, 10 mL min
1 - 16.7% ethanol CO2, 50  C, 310 bar, 6 L min
1, 20 min, 50 mL Ethanol:water (1:1) CO2 þ ethanol, 300 bar, 40  C CO2, 437 bar, 40  C, 10 min (static) þ 90 min (dynamic) CO2 þ ethanol (5%), 500 bar, 60  C, 3 h (dynamic) CO2 þ ethanol (5%), 500 bar, 60  C, 3 h (dynamic) CO2, 91 bar, 40  C, 30 min (dynamic) CO2 þ ethanol, 400 bar, 32  C, 3 h (dynamic)

Chlorella vulgaris

Anticancer (human lung cancer H1299, A549 and H1437) Antioxidant Antioxidant

C. vulgaris Dunaliella salina

Carotenoids, fatty acids b-Carotene isomers carotenoids, chlorophyll

Nannochloropsis gaditana

Carotenoids, chlorophyll

Antioxidant

Dictyopteris membranacea Chlorella pyrenoidosa

Volatiles Lutein

Spirulina platensis

Vitamin E g-linolenic acid Flavonoids, b-carotene, a-tocopherol, fatty acids (linolenic, linoleic and palmitic acid) g-Linolenic acid, lipids

Antimicrobial Prevention macular degeneration Antioxidant Antimicrobial Antioxidant Antimicrobial

Arthrospira platensis

Arthrospira maxima

Antioxidant Antimicrobial

CO2, 220 bar, 83.3  C (dynamic) CO2 þ ethanol, 400 bar, 40  C, 1 h (dynamic) CO2, 200 bar, 48  C, 4h

CO2 þ ethanol, 250 bar, 50–60  C

by subcritical water up to 150 bar and algal cells were successfully recultivated.38 Still, sufficient data of the phase behavior of key substances in the complex matrix water-algae-hydrocarbons need to be determined. The development of technologies to recover and separate hydrocarbons while being able to feed back the algae cells for recultivation can provide a significant competitive advantage in large-scale applications. Consequently, more research is expected in this field to provide translational data to the industry, which will enhance the growth of bio-based economy.

2.48.6

Future Perspectives

The use of biotechnology for the production of primary and secondary metabolites has much promise for sustainable manufacturing of biochemical, medical and dietary products. However, this poses the need for the development of an integrated production and recovery process allowing for in situ separation of cells, microorganisms, extraction and separation of target compounds. Furthermore, implementation of advanced bioseparation processes are required, which allow efficient separation of target compounds from dilute aqueous systems. Supercritical fluid processing is a promising technology, which can serve to fulfill these requirements and enable separation of a wide range of bioactive compounds. However, the application needs research and development activities to minimize the impact of extraction conditions on the living cultures of microorganisms, cells and microalgae. Particularly, the impact of pressurization and depressurization must be fundamentally investigated. In spite of the well-established phase behavior of chemical systems, more data are needed for biological systems for a better and in-depth understanding. Supercritical fluid processing needs to be investigated in all of these areas, while the experimental work shall prove the applicability and enable the scale-up.

2.48.7

Conclusions

There is an increased public concern related to the use of some organic solvents in food processing and contamination of final products with solvent residues. In addition to the consumer perspective, the development of green technologies is of paramount importance due to the tight environmental regulations along with requirements of food and pharmaceutical industries to manufacture high purity and value added products. Supercritical fluid technology fulfills these requirements and can be utilized for bioseparation of health beneficial compounds for pharmaceutical, nutraceutical and food purposes. However, the success and sustainability of supercritical fluid technology heavily rely on awareness of the industry, economic feasibility and development of fully continuous processes.

724

Supercritical Fluid Technology in Bioseperation

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38.

Hizir, M. S.; Barron, A. R. Basic Principles of Supercritical Fluid Chromatography and Supercritical Fluid Extraction. OpenStax-CNX Module: m46150 2013, 1–8. Brunner, G. Supercritical Fluids: Technology and Application to Food Processing. J. Food Eng. 2005, 67, 21–33. José, A. M.; Miguel, H.; Alejandro, C.; Elena, I. Use of Compressed Fluids for Sample Preparation. Food Applications Journal of Chromatography 2007, 1152, 234–246. Mukhopadhyay, M. Extraction and Processing with Supercritical Fluids. J. Chem. Technol. Biotechnol. 2009, 84, 6–12. Rozzi, N. L.; Singh, R. K. Supercritical Fluids and the Food Industry. J. Compr. Rev. Food Sci. Food. Saf. 2002, 1, 33–44. Montañés, F.; Corzo, N.; Olano, A.; Reglero, G.; Ibáñez, E.; Fornari, T. Selective Fractionation of Carbohydrate Complex Mixtures by Supercritical Extraction with CO2 and Different Co-solvents. J. Supercrit. Fluids 2008, 45, 189–194. Dunford, N. T.; Teel, J. A.; King, J. W. A Continuous Counter Current Supercritical Fluid Deacidification Process for Phytosterol Ester Fortification in Rice Bran Oil. Food Res. Int. 2003, 36, 175–181. Reverchon, E.; Volpe, M. C.; Caputo, G. Supercritical Fluid Processing of Polymers: Composite Particles and Porous Materials Elaboration. Curr. Opin. Solid State Mater. Sci. 2003, 7, 391–397. Abbas, K. A.; Mohamed, A.; Abdulamir, A. S.; Abas, H. A. A Review on Supercritical Fluid Extraction as New Analytical Method. Am. J. Biochem. Biotechnol. 2008, 4 (4), 345–353. Knez, Z.; Markocic, E.; Leitgeb, M.; Primozic, M.; Hrncic, M. K.; Skerget, M. Industrial Applications of Supercritical Fluids: A Review. Energy 2014, 77, 235–243. Capuzzo, A.; Maffei, M. E.; Occhipinti, A. Supercritical Fluid Extraction of Plant Flavors and Fragrances. Molecules 2013, 18, 7194–7238. Sovova, H. Steps of Supercritical Fluid Extraction of Natural Products and Their Characteristic Times. J. Supercritical Fluids 2012, 66, 73–79. Sapkale, G. N.; Patil, S. M.; Surwase, U. S.; Bhatbhage, P. K. Supercritical Fluid Extraction-A Review. Int. J. Chem. Sci. 2010, 8 (2), 729–743. Wang, S. M.; Ling, Y. C.; Giang, Y. S. Forensic Applications of Supercritical Fluid Extraction and Chromatograph. Forensic Sci. J. 2003, 2, 5–18. Ayre, A.; Ghude, K.; Mane, P.; Nemade, M.; Gosavi, S.; Pathare, A.; Lad, A. Supercritical Fluid Extraction-A Green Paradigm in the Area of Separation Science. Asian J. Biomed. Pharm. Sci. 2013, 3 (23), 1–7. Machado, B. A. S.; Pereira, C. G.; Nunes, S. B.; Padilha, F. F.; Umsza-Guez, M. A. Supercritical Fluid Extraction Using CO2: Main Applications and Future Perspectives. Separ. Sci. Technol. 2013, 48, 2741–2760. Bimakr, M.; Ganjloo, A. Supercritical Carbon Dioxide Extraction of Bioactive Compounds. Gavin Publ. Food Nutri. J. 2016, 48–56. Temelli, F. Perspectives on Supercritical Fluid Processing of Fats and Oils. J. Supercrit. Fluids 2009, 47 (3), 583–590. Nautiyal, O. H. Food Processing by Supercritical Carbon Dioxide-review. EC. Chem. 2016, 2.1, 111–135. Herrero, M.; Mendiola, J. A.; Cifuentes, A.; Ibanez, E. Supercritical Fluid Extraction: Recent Advances and Applications. J. Chromatogr. A 2010, 1217, 2495–2511. Da Silva, R. P. F. F.; Rocha-Santos, T. A. P.; Duarte, A. C. Supercritical Fluid Extraction of Bioactive Compounds. Trends Anal. Chem. 2016, 76, 40–51. Yesil-Celiktas, O.; Isleten, M.; Karagul-Yuceer, Y.; Bedir, E.; Vardar-Sukan, F. Influence of Supercritical CO2 and Methanolic Extracts of Rosemary on Sensory Properties and Shelf-life of Wheat Germ Oil. J. Food Qual. 2009, 32, 709–724. Akay, S.; Alpak, I.; Yesil-Celiktas, O. Effects of Process Parameters on Supercritical CO2 Extraction of Total Phenols from Strawberry (Arbutus unedo L.) Fruits: an Optimization Study. J. Separ. Sci. 2011, 34, 1925–1931. Kazan, A.; Koyu, H.; Turu, I. C.; Yesil-Celiktas, O. Supercritical Fluid Extraction of Prunus persica Leaves and Utilization Possibilities as a Source of Phenolic Compounds. J. Supercrit. Fluids 2014, 92, 55–59. Senyay-Oncel, D.; Ertas, H.; Yesil-Celiktas, O. Effects of Supercritical Fluid Extraction Parameters on Unsaturated Fatty Acid Yields of Pistacia terebinthus Berries. JAOCS (J. Am. Oil Chem. Soc.) 2011, 88, 1061–1069. Akgun, I. H.; Erkucuk, A.; Pilavtepe, M.; Yesil-Celiktas, O. Optimization of Total Alkannin Yields of Alkanna tinctoria by Using Sub- and Supercritical Carbon Dioxide Extraction. J. Supercrit. Fluids 2011, 57, 31–37. Pilavtepe, M.; Erkucuk, A.; Akgun, I. H.; Yesil-Celiktas, O. Supercritical CO2 Extraction of Alkanna Species and Investigating Functional Characteristics of Alkannin Enriched Yoghurt during Storage. Eur. Food Res. Technol. 2012, 234, 807–812. Yildiz, E.; Karabulut, D.; Yesil-Celiktas, O. A Bioactivity Based Comparison of Echinacea purpurea Extracts Obtained by Various Processes. J. Supercrit. Fluids 2014, 89, 8–15. Kazan, A.; Sevimli-Gur, C.; Yesil-Celiktas, O.; Turgut-Dunford, N. Investigating Anthocyanin Contents and in Vitro Tumor Suppression Properties of Blueberry Extracts Prepared by Various Processes. Eur. Food Res. Technol. 2016, 242, 693–701. Erkucuk, A.; Akgun, I. H.; Yesil-Celiktas, O. Supercritical CO2 Extraction of Glycosides from Stevia rebaudiana Leaves: Identification and Optimization. J. Supercrit. Fluids 2009, 51, 29–35. Khaw, K. Y.; Parat, M. O.; Shaw, P. N.; Falconer, J. R. Solvent Supercritical Fluid Technologies to Extract Bioactive Compounds from Natural Sources: A Review. Molecules 2017, 22, 1186. Yesil-Celiktas, O.; Otto, F.; Gruener, S.; Parlar, H. Determination of Extractability of Pine Bark Using Supercritical CO2 Extraction and Different Solvents - Optimization and Prediction. J. Agric. Food Chem. 2009, 57, 341–347. Pilavtepe, M.; Sargin, S.; Celiktas, M. S.; Yesil-Celiktas, O. An Integrated Process for Conversion of Zostera marina Residues to Bioethanol. J. Supercrit. Fluids 2012, 68, 117–122. Pilavtepe, M.; Sargin, S.; Celiktas, M. S.; Yesil-Celiktas, O. Transformation of Posidonia Oceanica Residues to Bioethanol. Ind. Crop. Prod. 2013, 51, 348–354. Deniz, I.; Ozen, M. O.; Yesil-Celiktas, O. Supercritical Fluid Extraction of Phycocyanin from Spirulina platensis and its In Vitro Cytotoxicity on Human Lung Cancer. J. Supercrit. Fluids 2016, 108, 13–18. Senyay-Oncel, D.; Yesil-Celiktas, O. Activity and Stability Enhancement of a-amylase Treated with Sub- and Supercritical Carbon Dioxide. J. Biosci. Bioeng. 2011, 112, 435–440. Yildiz-Ozturk, E.; Yesil-Celiktas, O. Supercritical CO2 Extraction of Hydrocarbons from Botryococcus braunii as a Promising Bioresource. J. Supercrit. Fluids 2017, 130, 261–266. Yildiz-Ozturk, E.; Ilhan-Ayisigi, E.; Togtema, A.; Gouveia, J.; Yesil-Celiktas, O. Effects of Hydrostatic Pressure and Supercritical Carbon Dioxide on the Viability of Botryococcus braunii Algae Cells. Bioresour. Technol. 2018. https://doi.org/10.1016/j.biortech.2018.02.041.

Precipitation and Crystallizationq

2.49

Pedro de Alcaˆntara Pessoˆa Filho, Departamento de Engenharia Química, Escola Politécnica da Universidade de São Paulo, São Paulo, SP, Brazil Gisele Atsuko Medeiros Hirata, Departamento de Engenharia Química, Universidade Federal de São Paulo, Diadema, SP, Brazil E´rika Ohta Watanabe, Faculdade de Engenharia Química, Universidade Federal de Uberlândia, Uberlândia, MG, Brazil Everson Alves Miranda, Faculdade de Engenharia Química, Universidade Estadual de Campinas, Campinas, SP, Brazil © 2019 Elsevier B.V. All rights reserved. This is an update of P.A. Alcântara Pessôa Filho, G.A. Medeiros Hirata, É.O. Watanabe, É.A. Miranda, 2.46 - Precipitation and Crystallization, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 651-663.

2.49.1 Introduction 2.49.2 Solid–Liquid Equilibrium: Phase Diagrams 2.49.2.1 Solubility: General Aspects and Precipitating Agents 2.49.2.2 Characterizing the Equilibrium: Ternary Phase Diagrams 2.49.3 Modeling of Solid–Liquid Equilibrium 2.49.4 Crystallization of Proteins 2.49.4.1 Solubility and Supersaturation 2.49.4.2 Nucleation 2.49.4.5 Crystal Growth 2.49.4.6 Polymorphism 2.49.4.7 Protein Crystallization in the Biotechnology Industry 2.49.5 Developing a Protein Crystallization Process Acknowledgments References Relevant Websites

725 726 726 728 729 731 732 733 734 735 735 736 737 737 738

Glossary Crystallization formation of a crystalline solid phase containing a biomolecule from a solution, specifically a solid phase formed by crystals larger than 10 mm. Equilibrium the state of a system that undergoes no macroscopic change and wherein there are no net fluxes of mass and energy. Model a mathematical relationship between properties of a system which cannot be deduced from an exact physical law. Nucleation the onset of a thermodynamically stable phase through the formation of small bubbles, droplets or particles in suspension. Precipitation formation of a solid phase containing a biomolecule from a solution, specifically a solid phase which comes from the formation of nuclei that grow into small primary crystals (0.1–10 mm). Solubility the concentration of a certain solute in solution that is in equilibrium with a solid phase containing this solute.

2.49.1

Introduction

There are two unit operations commonly found in downstream processing wherein a solid phase comprising a biomolecule is formed from an aqueous solution previously containing this biomolecule: precipitation and crystallization. This solid-phase formation is induced by changing the properties of the aqueous solutiondfor instance, by shifting pH or temperature or by adding salts, polymers (either neutral or polyelectrolytes) or organic solvents. They constitute two of the most important unit operations in protein recovery and purification. While these unit operations are similar in overall mechanism, they have distinct objectives and are conducted differently. The term “precipitation” is used to identify the formation of an amorphous solid phase. This operation occurs most commonly at the q

Change History: October 2017. Gisele Atsuko Medeiros Hirata updated section Crystallization of Proteins, specially the sub-sections Solubility and Supersaturation, Nucleation, Crystal Growth and Polymorphism. December 2017. Everson Alves Miranda updated sections Crystallization of Proteins (sub-section Protein Crystallization in the Biotechnology Industry) and section Developing a protein crystallization process. January 2018. Pedro de Alcantara Pessoa Filho updated the Introduction and the section Modeling of solid–liquid equilibrium, harmonized the final document (e.g., renumbering references), and submitted the revised version (February 2018).

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00098-7

725

726

Precipitation and Crystallization

beginning of a downstream process, and its main objective is to concentrate the biomolecule, thus reducing the size of the equipments. More rarely, precipitation may aim at separating nontarget biomolecules and keeping the target molecule in solution. It is the most common unit operation found in downstream processing: while it is difficult to give a definite figure, it is estimated that more than 80% of all downstream processes have at least one precipitation step. The term “crystallization” is reserved for the unit operation occurring most commonly at the end of the downstream process and aiming at the final purification and/or formulation of the target biomolecule. The solid phase is crystalline with particles that have a definite morphology and are usually large. Due to their distinct objectives, precipitation and crystallization are conducted in different ways. Amorphous precipitation is usually conducted without specific control of supersaturation (i.e., the difference between the actual concentration and the equilibrium concentration of the biomolecule), and hence without control of nucleation rate (i.e., the rate of formation of new crystal particles). On the other hand, crystallization is conducted with specific control of these two aspects; it is usually a slower process, which allows the growth of crystals with a minimum of vacant sites and impurities. From the point of view of the underlying phase equilibrium, the distinction between precipitation and crystallization is not so sharp. The solid phase formed through precipitation also results from the formation of nuclei that grow into small primary crystals (0.1–10 mm) which agglomerate (not always into large sizes) and therefore may contain impurities trapped in the final particles. The first attempt to systematize a theory on the solubility of proteins (and on the effect of salts upon it) was presented by Hofmeister in a series of seminal papers in the 1880s and 1890s. Since then, a comprehensive theory of precipitation has been pursued, with relative success: while many aspects of precipitation and crystallization have been elucidated, the need to account for the particularities of biomolecules and precipitating agents remains. No single general theory has been developeddand it is doubtful that one ever will. Even rather successful colloid-like theories are currently challenged, as they cannot explain many phenomena experimentally observed. In this article, the terms “phase equilibrium” and “solid–liquid equilibrium” will be applied indistinctly to equilibrium with an amorphous or a crystalline solid phase. The term “solubility” refers to the saturation linedi.e., the concentration of biomolecule that, given specific constraints (for instance, pH, temperature, ionic strength and cosolvent concentration) is in equilibrium with a solid phase. The term “salting-out” will refer to a decrease in solubility due to the addition of a cosolvent, even when this cosolvent is not a salt; conversely, “salting-in” is an increase in solubility that results from adding a cosolvent. Precipitation and crystallization are ultimately physical–chemical processes, and as such they are subject to the same constraints as any kind of solid–liquid equilibrium. On the other hand, the multitude of factors influencing this equilibrium often leads to the common belief that it can only be understood on empirical or semi-empirical grounds. A specific issue that will be addressed in this article is the composition of the solid phase. Even more than “simpler” organic substances, biomolecules often undergo many solid phase transitions. These transitions may include different hydration states, different crystalline habits and the presence or absence of coprecipitated solutes (such as salts), and the accessibility of these solid phases may depend on thermodynamic as well as kinetic hindrances. Since precipitation and crystallization are often conducted to separate proteins, this article will focus specifically on these biomolecules. However, a general theory will not, in principle, depend on which biomolecule is considered, and most comments also apply to other biomolecules. Particularities such as the dependence of net charge upon pH are not exclusive to proteins; the physical chemistry underlying the phase equilibrium remains the same.

2.49.2

Solid–Liquid Equilibrium: Phase Diagrams

A phase diagram is a map representing the equilibrium state of a system as a function of physical conditions (e.g., temperature and pressure) and concentration. For systems containing proteins, protein concentration, temperature and characteristics of the solvent (pH, ionic strength, buffer concentration and presence of additives) are the main variables to be considered. Phase diagrams entail information about molecular interactions among the components and allow prediction of the conditions under which a protein can be precipitated or crystallizeddor reduce the number of conditions that initially should be examined. However, it may be difficult to characterize the phase behavior of protein solutions completely due to the variety of dense phases which can be formed and conditions which can be adjusted and due to the uncertainty in the kinetics for reaching the equilibrium. The solid–liquid equilibrium of systems containing protein is usually described either through two-dimensional solubility diagrams (representing protein concentration in solution as a function of precipitating agent concentration) or ternary diagrams (showing the composition of all components in both equilibrium phases).

2.49.2.1

Solubility: General Aspects and Precipitating Agents

A protein crystal added to a solution (comprising a solvent, usually water, under specific conditions of pH, temperature, ionic strength, concentration of other species in solution, etc.) initially free of protein will dissolve to some extent: the process will occur until the concentration of protein in solution reaches a definite value, wherein equilibrium is established. This equilibrium protein concentration is the solubility under these specific conditions. As the solubility of a protein depends on various constraints, solubility diagrams are representations of solubility as a function of a single parameter while other parameters are kept constant. The solubility of a protein is strongly influenced by the characteristics of the solvent. The term “solvent” in this case means an aqueous solution containing other compounds that bring the solution to the desired values of pH (buffers) and ionic strength

Precipitation and Crystallization

727

(salts) as well as other additivesdsubstances that affect the solubility either by modifying the water structure or by interacting with different functional groups of the protein, changing its conformation or net charge. Usually, an increase in temperature results in an increase in solubility for nonbiological solutes; while the same tendency is observed for protein solubility, the opposite behavior, known as retrograde behavior, is not rare. Perturbations of solvent–protein interactions that result from shifts in temperature can modify protein conformation in aqueous solution: an increase in temperature weakens the hydrophobic effect, which may lower the solubility. Protein precipitation may be carried out solely by pH adjustment, which is called isoelectric precipitation: acids or bases are added to the protein solution until the pH reaches the protein isoelectric point (pI). At the pI, the average net charge of protein molecules is null and the fraction of neutral protein molecules in solution (considering that there is a charge distribution among the molecules) is maximal. Since the solid–liquid equilibrium is established with a neutral solid phase (the principle of phase electro-neutrality), the overall protein concentration must theoretically be at its minimum. This mechanism is often used for precipitation of bulk proteins, such as soy or milk proteins. The main perturbation of protein-solvent interactions that influences protein solubility is the addition of salts or organic solvents. Salts may establish direct electrostatic interactions with charged residues on the protein surface, and nonpolar interactions may occur between protein hydrophobic residues and the hydrophobic part of organic salts. Moreover, salts in solution do alter the medium properties (e.g., dielectric constant), specially at higher concentrations. At low salt concentrations, protein solubility usually increases with increasing salt concentration (salting-in) up to a maximum; higher concentrations of salt cause a decrease in solubility and precipitation of the protein (salting-out). It is not uncommon for experimental diagrams to not show both regions, either because the salting-in region is too small or because the salt solubility limits the salting-out region. The salting-in phenomenon is well known by experimentalists: proteins usually have larger solubility in salt solutions of low ionic strength than in deionized water. A common explanation for salting-in is that at low salt concentrations salt ions act as counter-ions for exposed protein charged residues. The condensation of counter-ions stabilizes the protein molecule in solution through screening surface regions from water molecules. The solubility of proteins in the salting-out region strongly depends on the actual salt concentration: usually, the decrease in the logarithm of the solubility is proportional to the ionic strength. The salting-out phenomenon is often related to the interaction between ions and water molecules, since at higher salt concentrations, there would be no more exposed charged residues without the corresponding counter-ion condensation. Hydrophobic patches on the protein surface are surrounded by a well-ordered hydration layer: this is the so-called hydrophobic effect. Some ions, such as the sulfate anion (called kosmotropic, which means that they “promote order”), rearrange the water structure, ordering water molecules around them and thus sequestrating water molecules from the protein hydration layer. This destabilizes the protein molecule, thereby lowering its solubility. While this is the usual explanation of the phenomenon, it is undoubtedly incomplete, and there is still much discussion on the whole mechanism of protein precipitation by salting-out. The addition of organic solvents (such as ethanol, methanol or acetone) results in a decrease in the dielectric constant of the solution. Due to this reduction, the electrostatic interaction between oppositely charged regions of protein molecules increases, causing molecule aggregation and subsequent precipitation. A secondary effect is that the ordered water structure around the hydrophobic patches on the protein surface is displaced by organic solvents, which exposes these regions and favors aggregation. The addition of neutral polymers (such as polyethylene glycol) lowers the solubility through a similar mechanism along with the excluded-volume phenomenon. In this case, an analogous dependence of solubility (a decrease in the logarithm of the solubility proportional to polymer concentration) is usually observed. Finally, polyelectrolytes such as carboxymethyl cellulose, polyacrylic acid, polymethacrylic acid and hydroxy-propyl methylcellulose can also be used to induce the protein precipitation. The mechanism of precipitation by polyelectrolytes involves a combination of polymer bridging (wherein particles are joined by the same polymer chain) and electrostatic interactions between oppositely charged molecules. The first principle is important when high molecular weight polyelectrolytes are used, whereas the second principle plays a more important role for polyelectrolyte molecules of low molecular mass. It must be recalled that proteins are polyelectrolytes whose charges are heterogeneously distributed and that some residues and surface regions have charges that are opposite to the protein net charge. Experimental solubility diagrams for proteins as a function of variables such as type and concentration of precipitating agents, pH and temperature have been extensively studied, and presenting a literature survey on this subject would be outside the scope of this article. Just to mention an early attempt on this subject (to situate it in a historical context), Green1 investigated the solubility of hemoglobin in concentrated solutions of different electrolytes (chlorides and sulfates) under varying conditions, using potassium phosphate buffers of different concentrations and at different pH values. Besides verifying that the solubility of the hemoglobin in electrolyte solutions decreases with increasing ionic strength of salts, the author observed that solubility reaches a maximum in sulfate solutions and that the pH of minimum solubility varies with salt concentration. A breakthrough on the study of protein solubility was presented by Shih et al.2 These authors presented a comprehensive study on the solubility of lysozyme, a-chymotrypsin and bovine serum albumin (BSA) in aqueous electrolyte solution as functions of ionic strength, pH value, the chemical nature of salt and initial protein concentration. This study partially corroborates previous results: protein solubility depends on pH, and the minimum solubility is observed around the isoelectric point of the protein (for example, the minimum solubility for lysozyme is observed around pH 10). Nevertheless, there was an unexpected finding: the solubility of a-chymotrypsin and BSA in aqueous salt solutions is approximately proportional to the initial protein concentration (for the same pH, temperature and salt concentration), while lysozyme solubility does not depend on the initial concentration of lysozyme. These results on the dependence of protein solubility on the initial protein concentration indicate that the measured

728

Precipitation and Crystallization

value does not correspond to the true solubility. The authors explain this behavior considering that the dense phase is a second liquid phase rather than a solid one. These results point out the limitations of knowing only the solubility and the need to determine the composition of the dense phase to achieve a complete comprehension of this phase equilibrium.

2.49.2.2

Characterizing the Equilibrium: Ternary Phase Diagrams

Liquid–liquid equilibrium of a ternary system is often presented in ternary diagrams, wherein all fractions can be shown simultaneously. Considering that protein precipitation appears to occur with the separation of a second liquid phase, determination of the composition of all coexisting phases may be necessary. However, very few examples of this approach can be found in the literature. This scarcity may be related to experimental problems that arise in such experiments: for example, slow precipitation kinetics may prevent the system from attaining equilibrium within a convenient time period. Moreover, kinetics may also depend on parameters such as pH, ionic strength, salt type and temperature. Besides a few early attempts to carry out this kind of experiment in the 1950s (with gelatin and serum albumin), the first important recent work on this subject was conducted by Moretti et al.,3 who investigated the phase equilibrium for the system lysozyme (from chicken egg white) þ ammonium sulfate þ water at pH values of 4.0 and 8.0. These authors presented the protein-rich precipitate and protein-lean supernatant compositions and included observations about the solid phase morphology (crystal, amorphous aggregate, gel and coacervate). An important and unexpected result was that the salt partitioned unevenly between the equilibrium phases when amorphous aggregates were formed. The authors also determined the water content of the crystal formed at low salt concentration: a salt-free solid containing around 85 mass percent of lysozyme. A specific issue that must be addressed is the composition of the dense phase (the protein-rich one). The protein-rich precipitate analyzed is not a single phase, but a heterogeneous mixture of two phasesdthe true precipitate (the solid phase) and the supernatant which is present in the interstices between solid particlesdas a complete separation of this precipitate and the protein-lean supernatant is not possible. While Moretti et al.3 were able to achieve a complete separation when crystals were formed, the presence of an amorphous phase makes a complete phase separation impossible. A specific analysis of the composition of coexisting phases can be found in Watanabe et al.4 This analysis is based on the observation that extensions of several tie-lines (i.e., the line that connects the coexisting phases) in ternary diagrams converge to a common point. It corresponds to an extension of Schreinemaker’s analysis applied for protein phase diagrams. Considering that the precipitate is a solid, the composition of the protein-rich phase lies somewhere on the tie-line between the supernatant composition and the solid-phase composition. Since the tie-lines converge to a point, this convergence point represents (within experimental uncertainty) the composition of the homogeneous solid phasedi.e., the true precipitate. An example of such applications can be seen in Fig. 1. Let us analyze this phase diagram of the system previously presented. It reveals two intersection points of tie-lines; at low salt concentrations, the intersection point indicates a salt-free phase (containing around 86 mass percent of lysozyme and 14 mass percent of water) and at higher salt concentrations the intersection point is a protein-salt-water complex containing about 37 mass percent of lysozyme and 14 mass percent of ammonium sulfate. The composition of the salt-free true precipitate agrees with that determined by Moretti et al.3 For the true precipitate observed at high salt concentrations, two hypotheses can be made: either the salt precipitated as pure salt crystals adsorbed on the lysozyme–water complex or the salt coprecipitated by generating a new complex containing salt, protein, and water. Since the same proportion was determined in experiments at different pH’s and temperatures, the most likely mechanism is the coprecipitation. Phase diagrams for systems containing sodium sulfate were similar, whereas only a salt-free precipitate was found with sodium chloride. The same authors also conducted enzymatic activity balances. They observed that activity losses occur more intensely for systems at high salt concentrationsdfor which the precipitate is the salt–protein–water complex. (NH4)2SO4 0.0

1.0 0.8

0.2 0.4

0.6 0.4

0.6

0.2

0.8 1.0 H2O 0.0

0.2

0.4

0.6

0.8

0.0 1.0 Lysozyme

Figure 1 Phase diagram of lysozyme þ ammonium sulfate þ water at pH 7.0 and 25.0  C: feed (,), protein-lean phase (O), protein-rich phase () connected by tie lines (). All concentrations are expressed in mass fraction.

Precipitation and Crystallization

729

This kind of analysis is promising, but some difficulties in its use can be foreseen: the overall uncertainty in phase composition data cannot be neglected (which may shed uncertainty on the exact composition of the true precipitate), and it cannot be applied to barely soluble proteins. Finally, there may be a relationship between the kind of true precipitate formed and the crystal structure, but reliable data on this aspect are scarce and any interpretation is hitherto speculative.

2.49.3

Modeling of Solid–Liquid Equilibrium

Experimental determination of solubility is obviously the first step in the design and operation of protein downstream processes. As it is unfeasible to obtain experimental equilibrium data for all situations, mathematical and thermodynamic models must be used: their parameters may be fitted to experimental data to make interpolation (and careful extrapolation) possible. Nevertheless, the modeling of solid–liquid equilibrium of proteins remains a subtle subject: there is no single model that describes the multitude of situations and influences on the solubility of a protein. Let us consider a solid phase formed by a single biomolecule in equilibrium with a liquid aqueous phase that contains this biomolecule. The thermodynamic equilibrium condition of phase equilibrium is given by mLp ðT; xÞ ¼ mSp ðTÞ

(1)

wherein T is the absolute temperature, x is the liquid phase composition and m is the chemical potential of the protein (p) in either the liquid (L) or solid (S) phase. This equation shows that the chemical potential on the saturation line is constant at the same temperaturedprovided there is only one solid phase. This equation is deceivingly simple and cannot be readily used to model protein solubility. Firstly, there is no experimentally accessible information on the values of chemical potential; secondly, there is no single model for the chemical potential of a protein in the liquid phase. To use this equation, one must relate the chemical potential to the activity, which can be done through the following equation:
 ref mLp ðT; xÞ ¼ mp T; x ref þ RT ln ap (2) There are many activity models: depending on their complexity, they may account for the size and shape of the molecule and for specific interactions. At the concentrations usually found in downstream processing, proteins can be considered in the so-called “infinite dilution” statedwhich means that the interaction between protein molecules themselves is less important than those between protein molecules and molecules of other compounds in solution. As an example, considering the simplified Flory–Huggins model, the following equation for the protein thermodynamic activity is obtained:    Vp
4 4 cpw 4w þ cpc 4c  ccw 4c 4w (3) ln ap ¼ 1 þ lnð4p Þ  Vp w þ c þ Vw Vc Vs wherein the subscript c represents a co-solvent, i.e., the precipitating agent (salt, polymer or organic solvent), V is the molar volume, f is the volume fraction and c is the Flory interaction parameter, which is related to specific interactions between the constituents. Coupling previous Eqs. (1)–(3), one concludes that the equilibrium concentration at constant chemical potential (solubility) is approximately related to the co-solvent concentration through an equation of the following kind: ln cp ¼ k1 þ k2 ,cc

(4)

While Eq. (4) is suggested by the form of the Flory-Huggins equation, considering other commonly used thermodynamic activity models would result in analogous mathematical relationships. Without resorting to any thermodynamic reasoning, Cohn5 developed a semi-empirical equation to model experimental data on protein solubility in salt solutions. Cohn’s equation is usually written: ln cp ¼ b  KS ,I

(5)

wherein I is the ionic strength, b is a constant that is theoretically dependent on the protein, pH and temperature (but not on the salt) and KS is a constant that is theoretically dependent on the salt, the protein and temperature (but not on pH). In practice, the dependence of b on salt type is commonly observed. Cohn’s equation was rederived by Melander and Horvath6 based on a more phenomenological reasoning. These authors were able to identify the major contributions to the salting-out constant KS: KS ¼ U,s  L

(6)

wherein U is a parameter proportional to the hydrophobic surface area of the protein; s is the derivative of surface tension to molal salt concentration and L is a salting-in parameter, proportional to the dipole moment of the protein. While Eq. (6) is seldom used to model protein behavior, it allows an understanding of the main factors affecting the dependence of the protein solubility on the salt concentration. The incremental surface tension (s) shows how increasingly difficult it is to form a cavity in the solvent to allocate the protein: the larger this parameter is, the steeper the slope of the solubility curve will be. The

730

Precipitation and Crystallization

parameter U is proportional to the hydrophobic surface area: the larger it is, the more susceptible to changes in salt concentration the protein solubility will be. Finally, the larger the protein moment dipole is, the less sensitive to the presence of salts the protein will be. A second important factor that affects the protein solubility is the pH. An early attempt to assess the effect of pH upon protein solubility was presented by Anders Grönwall in the 1940s.7 As the solid–liquid equilibrium is established with the neutral protein molecules in solution, this author derived the following equation: d ln cp ¼i d ln aHþ

(7)

wherein i is the protein mean charge. This equation is valid provided ionic strength and temperature remain constant. An integrated form of this equation was found to correlate reasonably well with the solubility curve for lactoglobulin. However, little attention has been directed to this equation in subsequent literature. A general solution of Eq. (7), relating the solubility to the pH and to the protein primary structure, was recently published.8 An implicit assumption of these approaches is that the protein-rich phase is solid and does not change along the phase diagram. However, this may not be exactly true for all systems, and the nature of the solid phase strongly affects protein solubility. This dependence is noticeable when the solid phase changes from an amorphous phase to a crystalline one. The solubility of protein crystals is usually lower (at the same temperature, ionic strength, and pH) than the solubility of the amorphous solid phase.3 Concerning Cohn’s equation, it means that, even though the value of KS may remain constant, the value of b shifts when the solid phase structure changes. Even considering only crystalline solid phases, a different crystal habit may lead to a different solubility for the same solution conditions. From the strict point of view of equilibrium thermodynamics, the solid phase structure with the lowest solubility is the most stable one. However, transitions between two solid protein structures are so slow that situations wherein only a metastable equilibrium is achieved in a reasonable timeframe do occur. A second important aspect is the very nature of the solid phase. In the already mentioned paper, Shih et al.2 showed that the solubility of two proteins (a-chymotrypsin and bovine serum albumin) in aqueous solutions of some salts (sodium chloride, sodium sulfate and sodium phosphate) depends on the initial protein concentration. To model those experimental data, the authors introduced a distribution coefficient to represent the partitioning of the protein between the two phases and found that this distribution coefficient is independent of the initial protein concentration. The natural conclusion is that the underlying equilibrium state is a liquid–liquid equilibrium instead of a solid–liquid one. Concerning especially the modeling of phase equilibrium, the paper by Shih et al.2 is very important, as approaches based on Cohn’s equation (or ultimately on the invariancy of the protein chemical potential in the liquid phase preconized by Eq. 1) cannot be used to model liquid–liquid equilibrium. To cope with this difficulty, osmotic equations of state have been applied. The osmotic pressure of a protein solution is not a novel concept: it is simply the pressure that must be applied to a protein solution to reach equilibrium with a cell containing the solvent without the protein, which is separated from the protein solution by a membrane impermeable to the protein but permeable to the solvent. The solvent itself may be a mixture (as an aqueous solution of salt or polymer). Specific analyses of osmotic pressure are conducted within the so-called McMillan and Mayer framework. According to this approach, the osmotic pressure of a macromolecule in solution is analogous to the pressure generated by a gas, and thus the same equations can be used to model both kinds of systems. In this case, the solvent has the same role as vacuum for gas molecules: it is just the medium wherein the interactions between molecules occur. Since the behavior is analogous, they can also be modeled by similar equations: based on this fact, osmotic equations of state have been developed and applied to the modeling of protein solutions. An osmotic equation of state is an equation which relates the osmotic pressure generated by a protein in solution to the protein concentration and temperature. These equations are explicit in the osmotic pressure (i.e., the osmotic pressure is written as function of protein concentration and temperature). While the mathematical treatment of this kind of equation may be cumbersome, some common characteristics can be highlighted. Expressions for the osmotic pressure are developed within the so-called perturbation theory. In short, osmotic pressure equations usually comprise two separate contributions: a reference term and a perturbation term. The reference term is commonly a “hard-sphere” (acronym HS) equation. It accounts only for the size of the molecules, which are considered rigid spheres, as suggested by the name. For proteins in solutions, the Carnahan-Starling expression is often used:

PHS rRT

¼

1 þ br þ ðbrÞ2  ðbrÞ3 ð1  brÞ3

(8)

wherein P is the osmotic pressure, r is the molar density (the concentration of protein expressed in mol per volume), R is the gas constant and b is the molar volume of spheres of diameter s: b¼

ps3 NAV 6

(9)

The perturbation term accounts for interactions between protein molecules. While for the reference term the Carnahan-Starling expression is widely used, for the perturbation term different approaches are used. These approaches differ by which contributions to the interaction potential are considered and how they are considered. The simplest way to account for these interactions is the

Precipitation and Crystallization

731

random-phase approximation. According to this approximation, protein molecules are uniformly dispersed in the solution. While it results in qualitatively poor predictions, it allows an understanding of factors affecting the protein behavior. The use of random phase approximation results in the following perturbation term:

Ppert rRT

ar ¼ RT

(10)

wherein the parameter a is related to the interaction between pairs of molecules through ZN a ¼ 2p

uðrÞr 2 dr

(11)

s

and u(r) is the interaction potential. Other forms of the perturbation part of the osmotic pressure (first- and second-order expressions) can be found in the literature, but their description is outside the scope of this article. The interaction potential depends on a sum of different contributions: the potential due to coulombic interactions (arising from the fact that proteins are macro-ions), the dispersion potential (related to the attractive van der Waals dispersion potential), the “osmotic” potential (arising from differences in local concentration of ions around protein molecules) and other specific potentials (for instance, due to hydrophobic interactions). The presence of co-solvents (salts or polymers) changes the medium wherein interactions take place, thus changing the expression for u(r). For the solid phase, while the overall theory remains the same, both terms must be changed using models that are appropriate for molecules in this phase: they are not randomly distributed, but occupy specific sites in a lattice. Different crystals have distinct spatial distributions, which change the calculated equilibrium. While they provide a more complete description of phase equilibrium, osmotic equations of state are not extensively employed in common engineering practice. The main reason for this seems to be the difficulty in assigning specific expressions to the intermolecular potential for each protein as a function of the many variables involved. An important parameter in the modeling of solid-phase transitions is the second virial coefficient. It is defined through the equation   d P (12) B ¼ lim r/0 dr rRT The equation for B as a function of interaction potential somehow resembles the equation for a considering the random phase approximation. There is well-established evidence that the formation of either an amorphous or a crystalline solid phase is related to the value of Bdcrystallization occurs only in a specific range of this parameter. Since it depends on temperature, the presence of cosolvents and specific interactions, the control of this parameter allows the formation of a crystalline phase. Although the expression is simple, experimental curves of P are not common in the literature and can be found only for some specific proteins. The value of B can be obtained likewise through different experimental techniques, such as light scattering or sedimentation equilibrium. However, values of B obtained through different experimental techniques for the same protein in the same environment may differ significantly. Even though the mathematical treatment of osmotic equations of state is intricate, from the physical point of view they represent a serious simplification of the actual molecular phenomena. Proteins are not hard spheres interacting through isotropic potentials. A recent series of papers put the very fundamentals of the use of colloid-like theories to describe protein solutions at stake. Results from small angle X-ray scattering9 and neutron scattering, coupled with circular dichroism and rheological measurements,10 show that such kind of theory is clearly insufficient to describe the experimental behavior of protein solutions. Finally, experimental results have also shown that protein solubility may depend on the protein concentration before the addition of a co-solvent. Following Shih et al.,2 this fact has been usually ascribed to the formation of a second liquid phase (gel-like), instead of a solid phase, in the precipitation operation. However, recent experimental results11 show that even in case a solid phase is formed, the solubility may depend on the protein initial concentration. With those results in mind, we face the question of which kind of model shall be used to describe the equilibrium in protein precipitation/crystallization processes. The experimental determination of solubility at the very conditions occurring in the precipitation/ crystallization process is always necessary. Experimental data found in literature may be used as a guide, but they cannot be used without considering the system particularities such as the initial protein concentration. For modeling these experimental data, expressions such as Cohn’s equation, with parameters that can be easily fitted to experimental data, can be used. Although they shed no light upon the precipitation mechanism, they are sufficient for calculating the solubility at the system conditions. Experimental results for the second virial coefficient B can be used (and modeled) likewise. The conclusion for this modeling part of the article can therefore be summarized as “keep the modeling as simple as possible, but at the same time as related to the process conditions as possible.”

2.49.4

Crystallization of Proteins

This section of the article will describe the basic phenomena that govern crystallization from solution, followed by a discussion of some practices of protein mass crystallization, which is the use of crystallization as a unit operation aiming to obtain a large amount

732

Precipitation and Crystallization

of protein crystal as the final product. The objective is not to treat these topics in depth, but to present a few principles with related equations regarding the formation and the growth of a crystal and the practical knowledge that makes protein crystallization possible on a large scale, despite all the scientific knowledge still lacking for more precise engineering design, operation, and control of protein mass crystallization. The crystallization process is basically defined by two phenomena: nucleation and crystal growth. Nucleation is a process that represents a phase transitiondmolecules move from a fully disordered state to an ordered statedand crystal growth is the addition of solute molecules to a nucleus surface until equilibrium is reached. However, these two crystallization steps are dependent on supersaturation, which in turn is a function of solubility and temperature. Therefore, the discussion starts with the concept of supersaturation. Then the presentation of the two basic phenomena of crystallizationdnucleation and growthdis followed by a brief discussion of polymorphism, i.e., the different crystal structures of a compound (different arrangements of molecules of the same substance in the crystal lattice). The role of protein crystallization in bioindustry and some common practices in protein mass crystallization are presented at the end. There are some secondary phenomena such as aggregation, agglomeration, breakage and aging, which although important in large-scale operation, will not be treated here for the sake of brevity.

2.49.4.1

Solubility and Supersaturation

Solubility is the most important single parameter in crystallization. The availability of solubility curves and the information on how these curves are affected by different factors (such as temperature, solution composition and pH) provide an understanding of crystallization behavior and thus allow process development and control. Solubility can be determined from a supersaturated solution, where the equilibrium occurs through nucleation and crystal growth, or from an undersaturated solution, where equilibrium occurs by dissolution of crystals. In these methods, the concentration of solute in solution should converge to the same value. However, achieving equilibrium from a supersaturated solution is more difficult. The reason is that with the growth of crystals, the surface may be poisoned by impurities and molecules may be inappropriately oriented in the crystal lattice. This poisoning may stop the growth before equilibrium between the crystals and the solution is established.12 In some cases, there is a difference of 5%–10% between the values found for the equilibrium concentration using the two methods. When time and quantities of material are limited, solubility should be determined by dissolution. The variation in solubility as a function of temperature, pH, pressure, or concentration of a precipitant is the basis for the design of a crystallization process. From the solubility it is possible to obtain the supersaturation, Dc, the driving force for growth and nucleation. Supersaturation is defined as the difference between the solute concentration (c) and its equilibrium concentration under the same conditions (c*), i.e., its solubility in solution. Sometimes supersaturation is referred to as supersaturation ratio, S, or relative supersaturation, s:

Dc ¼ c  c S¼ s¼

Dc c

c c

¼S1

(13) (14) (15)

According to Myerson,13 a supersaturated solution is metastable, i.e., the supersaturation of a solution doesn’t mean that crystals are formed. After setting up a condition under which the system is supersaturated, there is a period of time needed before the first crystallization event (the appearance of the first nucleus) takes place. This event may be the formation of an aggregate of sufficient size to create a viable nucleus or the formation of an ordered region in an aggregate that allows for its stable growth or both, owing to statistical fluctuations.14 As in any chemical reaction, it is necessary to overcome an energy barrier for a cluster to form into a nucleus and not to redissolve. Since the driving force for nucleation is supersaturation, the metastability of a solution is inversely proportional to the supersaturation, which means that the higher the supersaturation, the faster the appearance of a crystal from a clear solution. Fig. 2 represents a solubility diagram for a protein with retrograde solubility. In the solubility region crystallization does not take place and crystal dissolution occurs. In the supersaturation region, the probability of crystallization occurring depends on the degree of supersaturation, which is different in the metastable and labile regions. In the metastable region, spontaneous crystallization is not likely to occur (although a crystal immersed in such a solution will grow), but in the labile region spontaneous crystallization is probable but not inevitable.15 The importance of the metastable region is that crystallization processes are designed to operate within this area of the solubility diagram. The “lower” limit of the metastable regiondthe solubility curvedcan be determined with high precision and there is even a thermodynamically defined locus for the “upper limit” of the metastable diagram called the spinodal curve. However, in practice this “upper limit” is not well defined, since its determination is a function of factors such as the presence of dust, the rate at which supersaturation is imposed on the system (e.g., cooling rate) and agitation. Controlling these variables as much as possible, the metastable region limit is determined by slowly increasing supersaturation of a solution free of crystals until primary nucleation occurs. Regardless of the region (metastable or labile), once the energy barrier for the formation of stable agglomerates is overcome, the system achieves nucleation, the subject of the next topic.

Precipitation and Crystallization

733

Concentration

Supersaturation region

Labile region

Metastable region

Solubility region Solubility curve Temperature Figure 2 A schematic solubility diagram of a protein with retrograde solubility showing the metastable regiondthe area between the solubility curve and labile region.

2.49.4.2

Nucleation

Nucleation, the first stage of the crystallization process, is known as the “birth” of a crystal; the appearance of a new phase is classified as primary or secondary. Primary nucleation is divided into homogeneous, when crystallization occurs from a clear solution, and heterogeneous, when crystallization is induced by the presence of “foreign” surfaces (e.g., vessel wall, impeller, dust). Secondary nucleation occurs because of the presence of crystals when nuclei are produced, even at relatively low supersaturation, for many reasons. In industrial crystallizers there are two main reasons for secondary nucleation: collisions of crystals with other crystals or surfaces, specially with agitation devices, and fluid shear on crystal growing faces, events called contact and shear nucleation, respectively.16 The primary nucleation occurs independently of the presence of crystalline surfaces (although it may occur in the liquid phase of a suspension of crystals) and it depends on the supersaturation of the system. The primary nucleation rate (B0) can be obtained from the equation presented by García-Ruiz17: " # 16ps3 n2 B0 ¼ A,exp (16) 3k3 T 3 ðln SR Þ2 wherein A is a constant, s is the interfacial tension, n is the molecular volume, k is the Boltzmann constant, T is the solution temperature and SR is the relative supersaturation. The term A depends on the solution viscosity, molecular charge, molecular volume, and solution density; the exponential term is related to the activation barrier for nucleation. The term 16p/3 is the form factor, valid only when the cluster is considered spherical. Secondary nucleation, as already mentioned, requires the presence of crystals in the medium and their interaction with the crystallizer wall, impellers, etc. and varies with agitation, supersaturation, and suspension density. In many cases, crystals are added to the crystallizer (a practice called seeding) to reduce the supersaturation required for the crystallization process to a lower value than the one required for primary nucleation. Secondary nucleation is a complex and poorly understood phenomenon, whose rate (B) for industrial crystallizers is described by a power law given by B ¼ kN W i MT Dcn j

(17)

wherein kN is the nucleation constant, W is the agitation speed, MT is the suspension density, Dc is the supersaturation, n is the nucleation order and i and j are empirical exponents.13 The primary nucleation rate can be determined experimentally by observing the induction time (defined as the interval of time necessary for the appearance of a new phase after setting up the supersaturation condition), which is detected by changes in solution properties (e.g., turbidity and refractive index), by methods that use the width of the metastable zone (e.g., polythermic method in which the temperature of the system is changed gradually and slowly) or by the population balance. The estimate of nucleation kinetics is often obtained through the population balance, which is based on population density, np, derived from the number of particles per unit volume, N: np ¼ lim

DL/0

DN dN ¼ DL dL

(18)

In practice, the population density is calculated from the relationship: np ¼

Dm

3

kv $rc $L $DL

(19)

734

Precipitation and Crystallization

with

Dm ¼ f $MT

(20)

wherein f is the mass or volume fraction for a given size range DL, MT is the total mass per unit volume, kv is the volumetric shape factor, rc is the density of the solid and L is the average size. The datadf, DL and L–used in these equations may be obtained from analysis of particle size distribution. Other important properties of the crystal size distribution, such as the total number and total mass of particles per unit volume up to a size L (NT and MT, respectively) can be calculated using the population density approach13: ZL NT ¼

np ðLÞ,dL

(21)

0

ZL MT ¼ rc $kv $

L3 $np ðLÞ$dL

(22)

0

2.49.4.5

Crystal Growth

The availability of data for determining the crystal growth kinetics is very useful, since the kinetics parameter can facilitate the development, design, and operation of industrial processes. An understanding of the crystal growth rate is important because it affects product purity and crystal shape (depending on the application, large or small sizes are required). The growth may be described by a change in the characteristic dimension of the crystal (e.g., a single face of the crystal or the diameter of a circle that has the same projected area as this crystal) with time or by a change in the mass of the crystal. The relationship between mass and size can be described by the following equation: 1 dm kv dL kv ¼ 3 rc ¼ 3 rc G (23) A dt ka dt ka wherein ð1=AÞðdm=dtÞ is the increase in mass per unit time per unit surface area of the crystal (A), kv and ka are the volumetric and area shape factors, respectively, and G is the linear growth velocity: dL (24) G¼ dt Crystal growth can be divided into two stages: diffusion of the growth unit toward the crystal surface and integration of this unit onto the surface. Either of these steps can control the growth; solubility, supersaturation, agitation level and crystal size together will dictate which one is dominant. The basic expression used for crystal growth is related to supersaturation: G ¼ kg Dcg

(25)

wherein Dc is the supersaturation, the exponent g is the growth order (usually 1  g  2; g> 2 only for poorly soluble compounds18) and kg is the growth constant. There are many models to describe the crystal growth. Among them are the B þ S (birth and spread) model, the BCF (BurtonCabrera-Frank) model and the diffusion and integration model. The B þ S model, also known as “nuclei on nuclei” or “polynuclear growth,” occurs when nuclei are formed on the surface of the crystal and these grow and spread all over the surface. It is more likely to occur at high supersaturation levels. The BCF model, known as spiral growth, starts with screw dislocations that are responsible for crystal growth. Examples of crystals formed by this mechanism are silicon carbide crystals and C36 normal alkane crystals.15 In the diffusion-integration model it is difficult to separate the systems that are controlled by diffusion from those controlled by integration because the growth limitation depends on the supersaturation of the system. These two steps occur under the influence of different driving forces. One of them is diffusion, described by the equation: dm ¼ kd Aðc  ci Þ (26) dt and the other one is reaction: dm ¼ kT Aðci  c Þ dt

(27)

wherein kd is the coefficient of mass transfer by diffusion, kr is the rate constant for the surface reaction (integration) and ci is the solute concentration in the solution at the crystal–solution interface. When the two processes are important, the following equation can be applied: dm ¼ KG Aðc  cÞg (28) dt wherein KG is an overall crystal growth coefficient.

Precipitation and Crystallization

735

For crystallization processes that are controlled by diffusion, KG z kd. On the other hand, if the process is controlled by surface integration, KG z kr. However, the growth process is more complex, e.g., for an electrolyte crystallizing from aqueous solution several steps should be considered simultaneously: bulk diffusion of hydrated ions through the diffusion boundary and adsorption layer, integration of ions into the lattice, etc.15

2.49.4.6

Polymorphism

A substance is said to be polymorphic when it has several crystalline arrangements under different crystallization conditions. Polymorphism is complex and controlling the crystallization of polymorphs is a challenge. Polymorphic structures have distinct properties, including solubility, and the control of polymorphism may be very important in some applications, as for example in the pharmaceutical industry: if a drug can be crystallized as different polymorphs, these may have different chemical reactivity and dissolution rates, which can affect the performance stability, and safety of the drug. Since polymorphs can have different properties, they can behave as different products. Therefore, a crucial step in drug development is to ensure that no transition between polymorphic forms take place during the drug shelf life. Polymorphism is a common phenomenon in crystallization and only one polymorph is thermodynamically stable under certain conditions (Ostwald’s rule of stages). A well-known example of polymorphism is carbon, whose polymorphs are graphite and diamond: the latter is a metastable form with a half-life close to infinite. In some cases, the composition of the crystal can differ only by the number of solvent molecules. The crystal lattice may contain solvent molecules (in stoichiometric or non-stoichiometric ratio), forming crystals that are called solvates or pseudo-polymorphs (or hydrates, when the solvent is water). Polymorphs and solvates can be identified by various techniques such as X-ray diffraction, differential scanning calorimetry, thermogravimetry, infrared spectroscopy (FT-IR), Raman spectroscopy, optical and electron microscopy.

2.49.4.7

Protein Crystallization in the Biotechnology Industry

Since crystallization of a protein (hemoglobin of the earthworm) was first observed by Hünefeld in 1840, and since the pioneering studies in the X-ray crystal structures in the 1950s, crystallization of proteins has been steadily improving, though initially it was viewed as a kind of “alchemy” and until recently was considered to be “more an art than a science.” When protein purification was in its infancy, crystallization of a protein was employed as a valuable purification step to obtain purified material especially for research. Over time, the large number of proteins known to crystallize has facilitated the application and development of the diffraction techniques aiming at determining structure. Nowadays, the production of quality diffraction crystals is routine in laboratories around the world (thousands of protein structures have been determined with this approach), despite all the remaining uncertainty regarding how to do so effectively. There is a vast body of literature on the subject of protein crystallization, focusing primarily on the production of crystals for X-ray diffraction, such as the work of McPherson.14 However, most of this literature deals with a basic, bottleneck question: what are the conditions under which a specific protein crystallizes? Despite some understanding of the effect of each of the many variables that play important roles in protein crystallization, the large number and the possibility of the combined effects of variables make prediction of such conditions an impossible task. Experimentation is mandatory. Nowadays, crystallization has three roles in biotechnology: a) production of single crystals aiming at determining structure, b) unit operation in the downstream processing (DSP) of a protein and c) formulation of final products for high stability and facilitated manipulation (e.g., solid–liquid separation and drying) or controlled release. The first application requires a single diffraction-quality crystal, as already discussed; the other two (called bulk or mass crystallization) require processes with high yields and products with specified solubility, size distribution, morphology and impurity levels. Mass crystallization is not absent from the biotechnology scenario since small molecules like antibiotics are routinely crystallized commercially on a large scale. However, in the case of proteins mass crystallization is less common; its potential as a unit operation in the DSP of proteins is still significantly untapped. Insulin is a classic example, having crystallization steps in largescale DSP processes ever since the days when it was produced exclusively by extraction from animal pancreas. The advances in recombinant DNA technology and the sequencing of the human genome should increase the number of proteins crystallized on a commercial scale. Among the reasons that crystallization is not more prevalent in protein DSP are the physical characteristics of protein crystals that differ markedly from the crystals of small molecules (organic or inorganic) and the difficulty or even impossibility of using scale-up methods on a small scale to produce quality diffraction crystals. As just stated, due to their macromolecular structure, protein crystals differ significantly from crystals of small molecules. First, protein crystals are very fragile for several reasons which include the following: a) their tertiary structures are most commonly not symmetrical and have few points of contact between protein molecules, b) the bond energy in these contacts is small and c) there is frequently a high water (solvent) content in the crystal structure.14 In the case of proteins, crystal size is smaller and crystallization times are longer. The practice of protein crystallization deviates from the traditional techniques used for small molecule crystallization, since protein molecules are very fragile under the relatively harsh conditions used for small molecules such as high temperature, organic solvents and pH extremes. Therefore, conventional crystallization methods based on evaporation, high pressure, large changes in temperature and the use of organic solvents at relatively high temperature (e.g., room temperature) are virtually impossible to use.

736

Precipitation and Crystallization

Other variables or a range of these variables that affect supersaturation, and consequently nucleation and crystal growth, must be selected for protein crystallization. The methods practiced on a small scale for the production of quality diffraction crystals and mass crystallization actually share some basic principles and techniques, but if we focus on the first stage of mass crystallization process developmentdscreening of crystallization conditionsdthere are important differences to consider. The four crystallization methods most frequently used for quality diffraction crystals (batch crystallization, vapor diffusion, liquid–liquid diffusion and dialysis) are adequate for work on the small scale required by the low availability of protein (volumes as low as 10 mL). Of these methods, only batch crystallization is applicable on a large scale. Relatively small crystals are adequate in mass crystallization (10–50 mm), but larger sizes are needed for diffraction (150–500 mm). Moreover, some characteristics of precipitants and additives used for quality diffraction crystals are incompatible with mass crystallization, as will be discussed later. Mass crystallization has relatively low capital and operating costs and can produce protein of high purity (>99.9%). Therefore, it is a unit operation with the potential to substitute for the usually expensive column chromatography, the workhorse of protein purification. Although protein crystallization from basically clarified culture broth has been reported,16,19 mass crystallization of proteins is mostly employed as a final purification or polishing step in DSP, frequently for pharmaceutical proteins. The use of crystallization in the first stages of DSP is desirable because of the high level of concentration it provides, besides the possibility of purification. However, the presence of impurities usually completely hinders nucleation. Seeding the solution may overcome this problem. Also, when it is known that a high-purity preparation of a protein crystallizes, the approach to studying the crystallization of this solution, to which impurities removed at earlier stages of purification are added one by one in batch experiments, may allow the crystallization step to be moved closer to the fermentation step in the purification train. The diversity of protein characteristics (in terms of amino acid sequence, structure, post-translational modifications, and size) and solution composition (due to different cell systems in which they are produced and to different DSP strategies) can also make a crystallization process developed for a specific protein a unique process. Successful large-scale protein crystallization processes developed by companies are usually not disclosed, being considered proprietary work giving the company a competitive edge. The low availability of protein in the initial stages of process development also hinders the study of crystallization as a step in the DSP train. Therefore, mass protein crystallization process development is a difficult task that is conducted based largely on empirical results rather than on theory. A final and important comment is that one must also keep in mind the large volumes of waste that will be generated in the commercial-scale production of a biomolecule, which will need to be properly treated and discarded.

2.49.5

Developing a Protein Crystallization Process

The development of a large-scale crystallization process can be thought of as a four-step task: (1) screening to select suitable crystallization conditions and its optimization followed by (2) determination of phase diagram with characterization of the solubility curve and metastable zone, (3) conduction of batch crystallization experiments and (4) pilot plant tests for scale-up. The first step in the development of a protein mass crystallization process starts with techniques and chemicals used for obtaining quality diffraction crystals, despite their different objectives. Screening for crystallization conditions (pH, type and concentration of precipitant and additives, protein concentration, etc) is usually on a small scale (with volumes as low as 10 mL), for example, in conventional 24- or 96-well microplates or more recently, microplates for high-throughput screening. Hanging-drop and sittingdrop vapor diffusion are the techniques usually used. However, the selection of the range of conditions to be screened must take into account the stability and the final use of the protein, operating cost, safety and the type of waste stream to be generated. Proteins crystallize over a wide range of pH (e.g., at pH as low as 3.5 and as high as 10.0). This parameter has a dominant effect on systems of low ionic strength and a lesser effect as the ionic strength increases. Determination of the pH range for which the screening will be designed should never underestimate the stability of the protein with pH, especially due to the relatively slow growth that results in long batch crystallization times that can extensively denature the product. The idea that proteins may be more easily crystallized at their pI (due to a supposedly pronounced solubility minimum) is not verified for all proteins, since under these conditions amorphous precipitation can frequently take place. Many of the precipitants and additives frequently used in screening for the generation of diffraction crystals should not be part of a screening design aiming at process development, since they may be expensive, toxic or carcinogenic or may cause problems in effluent treatment, especially at high concentrations. Examples of such compounds are methanol, dioxane, polyethylene glycol and its derivatives, ammonium sulfate, 2-methyl-2,4-pentanediol and lithium salts. Also, at high concentrations of precipitant the system may be at a high supersaturation, which causes too much nucleation and the formation of amorphous particles. The same may happen if protein concentration (usually in the range from 5 to 100 mg/mL) is relatively too high. The set of conditions will determine not only if crystals will grow instead of precipitate form but also for the same precipitant, the morphology of the crystal, which may be of extreme relevance for the final product.14,20 Screening is carried out at a fixed temperature (4  C or ambient temperature). A detailed evaluation of this important variable in crystallization starts in the next step: determination of the solubility curves and the width of the metastable region. The solubility curve is of paramount importance in the study of the phenomenon of crystallization and for the design of a crystallization process, since it determines the supersaturation of the system which in turn defines the nucleation and crystal growth, as

Precipitation and Crystallization

737

already mentioned. Together with delimitation of the metastable zone, the solubility curve defines the phase diagram for protein crystallization. The batch crystallization experiments are then performed using stirred-tank crystallizers. These tanks are usually large enough to provide sufficient material to withdraw samples for analysis (tank volume as low as 1.0 L or as high as 20 L), although geometrically similar stirred-tanks of 5 and 100 mL have been used as intermediate scale-up steps.19 The batch operation is in fact the most appropriate for protein mass crystallization due to the usually low production scale (at a maximum level of kilograms per batch). The objectives of these experiments are to find operating conditions (mainly temperature and agitation) and to define a strategy to achieve supersaturation that produces particles according to specifications. These specifications for product and process may be yield, particle size distribution, purity, filtration rate, morphology and productivity. Since nucleation is strongly affected by supersaturationdwhich may result in excess nucleation and consequently small particle sizedthe seeding of a clear protein solution with crystals is always implemented to reduce process time and improve the quality of the crystals formed.14,20 The batch data can be analyzed to determine the growth kinetics by two methods: one based on the monitoring of crystal length with time or and the other based on mass balance.20 The method based on mass balance has the advantage of not requiring expensive equipment such as laser light scattering and digital microscopy. In the case the required product purity is not achieved, recrystallization should be tried if the impurity is more soluble in the liquid phase than the product. Once process conditions that satisfy all specifications are found, scale-up studies can be started at a pilot plant. In the case of pharmaceutical proteins that are produced in relatively small amounts, the scale-up ratio at this stage would vary from 10 to 50. If the beginning of the development of a large-scale crystallization process, the answer to the question “what are the conditions under which a specific protein crystallizes?” is not predictabledbeing heavily based on experimentationdthe last step, scale-up, will also depend more on experiments than on modeling and simulation, despite advances in this area of crystallization kinetics and crystallizer design. One basic criterion which is easy to attain for scale-up is the maintenance of the geometric similarity of the crystallizer vessels. Also, one would think that to scale up protein crystallization, conditions which assure that mixing on a larger scale is equivalent to mixing on a smaller scale should be determined in order to guarantee a homogeneous level of supersaturation in the crystallizer as well as the dispersion of the crystals throughout the liquid volume. However, scale-up criteria for agitated vessels such power per unit volume or impeller tip speed are not feasible due to protein sensitivity to shear, which causes denaturation, and the fragile nature of protein crystals, already mentioned. Therefore, minimizing protein denaturation and crystal breakage are priorities in scale-up. The minimum impeller rotation speed required for particle suspension is the mixing that should be tried first with impellers that generate low shear such those with configurations used in mammalian cell culture but in a scoping mode (upward flow near the axis and annular flow near the wall).19 As already discussed regarding the small volumes of proteins produced, the factor of scale-up from pilot to commercial scale falls within a range of 10–50.

Acknowledgments We would like to thank Dr. André Bernardo for reviewing the crystallization part of the article. Our acknowledgments also go to the Brazilian financial agencies Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).

References 1. Green, A. A. Studies in the Physical Chemistry of the Proteins: X. The Solubility of Hemoglobin in Solutions of Chlorides and Sulfates of Varying Concentration. J. Biol. Chem. 1932, 95, 47–66. 2. Shih, Y. C.; Prausnitz, J. M.; Blanch, H. W. Some Characteristics of Protein Precipitations by Salts. Biotechnol. Bioeng. 1992, 40, 1155–1164. 3. Moretti, J. J.; Sandler, S. I.; Lenhoff, A. M. Phase Equilibria in the Lysozyme-ammonium Sulfate-water System. Biotechnol. Bioeng. 2000, 70, 498–506. 4. Watanabe, E. O.; Popova, E.; Miranda, E. A.; Maurer, G.; Pessôa Filho, P. A. Phase Equilibria for Salt-induced Lysozyme Precipitation: Effect of Salt Type and Temperature. Fluid Phase Equil. 2009, 281, 32–39. 5. Cohn, E. T. The Physical Chemistry of Proteins. Physiol. Rev. 1925, 5, 349–437. 6. Melander, W.; Horvath, C. Salt Effects on Hydrophobic Interactions in Precipitation and Chromatography of Proteins: an Interpretation of the Lyotropic Series. Arch. Biochem. Biophys. 1977, 183, 200–215. 7. Grönwall, A. Studies on the Solubility of Lactoglobulin. I. The Solubility of Lactoglobulin in Dilute Solutions of Sodium Chloride at Varying Ionic Strength and Hydrogen Ion Activity. C. R. Trav. Lab. Carlsberg 1941, 24, 185–200. 8. Franco, L. F. M.; Pessôa Filho, P. A. On the Solubility of Proteins as a Function of pH: Mathematical Development and Application. Fluid Phase Equil. 2011, 306, 242–250. 9. Franco, L. F. M.; de Oliveira, C. L. P.; Pessoa Filho, P. A. Thermodynamics of Protein Aqueous Solutions: From the Structure Factor to the Osmotic Pressure. AIChE J. 2015, 61, 2871–2880. 10. Sarangapani, P. S.; Hudson, S. D.; Pathak, J. A. Critical Examination of the Colloidal Particle Model of Globular Proteins. Biophys. J. 2014, 108, 724–737. 11. Pinheiro, M. J.; Freitas, S.; Miranda, E. A.; Pessôa Filho, P. A. Solubility of Lysozyme in Aqueous Solution Containing Ethanol or Acetone: Unexpected Dependence on the Initial Protein Concentration. Fluid Phase Equil. 2016, 429, 9–13. 12. Asherie, N. Protein Crystallization and Phase Diagrams. Methods 2004, 34, 266–272. 13. Myerson, A. S. Handbook of Industrial Crystallization, 2nd ed.; Butterworth-Heinemann: Woburn, 2002. 14. McPherson, A. Crystallization of Biological Macromolecules, Cold Spring Harbor: New York, 1999. 15. Mullin, J. W. Crystallization, 4th ed.; Butterworth-Heinamann: Woburn, 2001. 16. Harrison, R. G.; Todd, P.; Rudge, S. R.; Petrides, D. P. Bioseparations Science and Engineering, Oxford University Press: Oxford, 2003. 17. García-Ruiz, J. M. Nucleation of Protein Crystals. J. Struct. Biol. 2003, 142, 22–31.

738

Precipitation and Crystallization

18. van Rosmalen, G. M.; Bermingham, S.; Bruinsma, D.; et al. TUDelft – IPT Lectures on Industrial Crystallization and Precipitation, Instituto de Pesquisas Tecnológicas: São Paulo, 2004. 19. Hekmat, D. Large-scale Crystallization of Proteins for Purification and Formulation. Bioproc. Biosyst. Eng. 2015, 38, 1209–1231. 20. Etzel, M. R. Bulk Protein Crystallization – Principles and Methods. In Process Scale Bioseparation for the Biopharmaceutical Industry; Shukla, A. A., Etzel, M. R., Gadan, S., Eds., Taylor & Francis: Boca Raton, 2007.

Relevant Websites Both precipitation and crystallization protocols are developed ad hoc for target proteins: as seen previously, there is no single recipe applying to the precipitation/crystallization of any protein from any medium. Therefore, reliable web resources with open access are still virtually inexistent for these operations. For readers interested in further information, the authors can only recommend a detailed survey of relevant databases, such as the Scopus™ and ISI Web of Knowledge™, as these are the only web sources they consult regularly.

2.50

Adsorption and Chromatography

Y Sun, Q-H Shi, L Zhang, G-F Zhao, and F-F Liu, Tianjin University, Tianjin, China © 2011 Elsevier B.V. All rights reserved. This is a reprint of Y. Sun, Q.-H. Shi, L. Zhang, G.-F. Zhao, F.-F. Liu, 2.47 - Adsorption and Chromatography, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 665-679.

2.50.1 2.50.2 2.50.2.1 2.50.2.2 2.50.2.3 2.50.2.4 2.50.2.5 2.50.2.6 2.50.2.7 2.50.3 2.50.3.1 2.50.3.1.1 2.50.3.1.2 2.50.3.1.3 2.50.3.1.4 2.50.3.1.5 2.50.3.2 2.50.3.3 2.50.3.4 2.50.4 2.50.4.1 2.50.4.2 2.50.4.3 2.50.5 2.50.5.1 2.50.5.1.1 2.50.5.1.2 2.50.5.1.3 2.50.5.2 2.50.5.2.1 2.50.5.2.2 2.50.5.2.3 2.50.5.3 2.50.5.4 2.50.5.5 2.50.5.5.1 2.50.5.5.2 2.50.5.5.3 2.50.6 References

Introduction Molecular Interactions in Adsorption Hydrogen Bond Hydrophobic Interaction Electrostatic Interaction van der Waals Interaction Coordination Bond Covalent Bond Conformational Entropy Chromatographic Methods Packed-Bed Chromatography Size-Exclusion Chromatography Ion-Exchange Chromatography Hydrophobic Interaction Chromatography Affinity Chromatography Displacement Chromatography Expanded-Bed Adsorption Chromatography Electrochromatography Radial-Flow Chromatography Theoretical Aspects of Adsorption and Chromatography Adsorption Equilibria Uptake Kinetics Theoretical Considerations of Chromatography Development of Adsorption and Chromatography Innovation of Chromatographic Matrices Flow-Through Media Membrane Monolith Selection and Design of Affinity Ligands Combinatorial Library Approach Rational Design Combination of Rational Design and Synthetic Combinatorial Library Mixed-Mode Ligands Displacer Screening and Design Molecular Insight Into Protein Adsorption Modeling and Visualization Adsorption Process Protein Conformational Transition Conclusions

740 740 740 740 741 741 741 741 741 741 741 742 742 743 744 744 745 745 746 746 746 747 748 748 748 748 749 749 749 749 750 750 750 750 751 752 752 752 752 753

Glossary Adsorption The adhesion or retention of molecules of gas, liquid, or dissolved solids to a surface resulting from the force field at the surface or the molecular interactions between the molecules and the ligand attached on the surface. Affinity chromatography A liquid chromatography that uses a stationary phase with immobilized biologically related groups (affinity ligands), in which biomolecules are separated based on a highly specific biological interaction such as that between antigen and antibody, enzyme and substrate, or receptor and ligand. Chromatography The separation technology of solutes dissolved in a mobile phase as they pass down a column due to the differential distribution of the solutes between the mobile phase and the stationary phase in the column.

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00099-9

739

740

Adsorption and Chromatography

Hydrophobic interaction chromatography A liquid chromatography with a stationary phase bonded with weakly hydrophobic ligands, by which molecules are separated in the column according to their differences in hydrophobicities. Ion-exchange chromatography A liquid chromatography in which ionic solutes are separated according to their charges by binding to the cationic or anionic sites of the stationary phase known as ion exchangers. Size-exclusion chromatography A type of liquid chromatography, also called gel-filtration chromatography, which uses porous particles as the stationary phase, in which solutes are separated by the difference in their molecular sizes.

2.50.1

Introduction

Adsorption is a phenomenon of solute attachment to a solid surface, and adsorption operations are widely applied in chemical and biochemical processes as well as in the living activities of human beings for the recovery or removal of specific substances. Solid materials, usually referred to as adsorbents, are used for adsorption operations. Because adsorption takes place at the solid surface, adsorbents must be fabricated to be of high specific surface area for high adsorption capacity. For this reason, adsorbents are usually porous materials. Chromatography is the primary mode of adsorption operations. It is a multistage separation technology and offers high-resolution separation of different solutes of high similarities. Bioseparations mainly involve liquid feedstocks, so adsorption operations in biotechnology usually occur at liquid–solid interfaces. Over the past three decades, biotechnology has developed rapidly as marked by the advances in genetic engineering and cell fusion techniques. In this process, chromatographic technology has played the most important role in the separation and purification of biomolecules. More importantly, chromatography is a separation methodology of great diversity; it is based on various interactions between target solutes and the ligands coupled to a solid surface, and also links to various other separation methods. Hence, chromatographic separations can be achieved on the basis of different separation mechanisms, equipment configurations, and operation modes. As a result, the technology is so powerful that various substances, including small molecules, biopolymers, and particulate materials such as viruses and whole cells, can be purified by the combination of different chromatographic steps. Hence, chromatography is the most widely used separation method in biotechnology. This article is devoted to an overview of the science and technology of chromatography. Various molecular interactions involved in adsorptions are briefly described first. This is followed by the introduction of various chromatographic methods for bioseparations. In this section, size-exclusion chromatography (SEC) is first mentioned as an important chromatographic technique, although it does not involve adsorptions. Only the fundamentals of the methods are outlined and readers can refer to other monographs on chromatography (e.g., Refs. 1–3) for more detailed information. Next, the theoretical aspects of adsorption and chromatography are summarized, including adsorption equilibria, uptake kinetics, and the fundamental theories of chromatography. Finally, the emphases are on the recent advances in studies of adsorption and chromatography to provide insight into the future development of biochromatography.

2.50.2

Molecular Interactions in Adsorption

Adsorption of a solute on a surface involves various interactions between the solute and the surface or the chemical groups attached to the surface. The interactions that contribute to the adsorption of biomolecules include hydrogen bonding, hydrophobic interaction, electrostatic interaction, van der Waals interaction, coordination bonding, covalent bonding, and conformational entropy.4

2.50.2.1

Hydrogen Bond

A hydrogen bond, described as D–H.A, is an interaction in which a hydrogen atom (H) is attracted simultaneously by two electronegative atoms (D and A). The electronegative atom (D) covalently bonded to H is named as donor, while the other (A) is named as acceptor. Hydrogen bonding is a driving force for adsorption with an energy range of 13–30 kJ mol
1 and increases with electronegativity of the participants, D and A. Hydrogen bonding energy decreases with increasing temperature and ion strength, as well as by the presence of chaotropic agents such as urea and guanidine hydrochloride.

2.50.2.2

Hydrophobic Interaction

Redistribution of ordered water molecules around single apolar solutes back into bulk solution causes the association of apolar molecules in water and the decrease of the Gibbs energy of the system, which is termed as hydrophobic interaction. This interaction exists between hydrophobic groups such as benzene rings or hydrocarbon chains and the hydrophobic region in biomolecules, with an energy range of 12–20 kJ mol
1. Both the salt type and concentration can affect hydrophobic interaction. Kosmotropic/lyotropic salts, for instance, (NH4)2SO4, Na2SO4, NaCl, KCl, and CH3COONH4, promote hydrophobic interaction, while chaotropic agents, for instance, KSCN, NaI, KClO4, and urea, reduce hydrophobic interaction. Stronger promotion or reduction effect by the agents is usually observed at higher concentrations. Hydrophobic interaction is enhanced by increasing temperature. Adsorption based on hydrophobic interaction is called hydrophobic adsorption.

Adsorption and Chromatography 2.50.2.3

741

Electrostatic Interaction

When a solute and a solid surface are both charged, electrostatic interaction, attractive or repulsive, can occur between them. The strength of electrostatic interaction depends on the charge numbers, so it is significantly affected by both pH and ion strength. The pH at which the charges on the solute and the surface just compensate each other gives rise to the maximum electrostatic adsorption, while the increase of ion strength will weaken or even completely screen the electrostatic interaction. The increase of temperature reduces electrostatic effect due to the enhanced thermal motion of molecules and atoms at elevated temperature. Adsorption based on electrostatic interaction is usually called electrostatic adsorption or ion exchange.

2.50.2.4

van der Waals Interaction

van der Waals interaction, originating from the interactions between fixed and/or induced dipoles, often contributes to adsorption. It is very sensitive to the separation distance (r) between the dipoles, diminishing as r
6. Therefore, it operates only over a limited range of intermolecular distance (around 0.2 nm). Moreover, van der Waals interaction is weaker than most of the other molecular interactions, usually with energy range of 4–8 kJ mol
1.

2.50.2.5

Coordination Bond

Transitional metal ions, such as Cu2þ, Zn2þ, Ni2þ, and Co2þ, can form coordination bonds with the imidazolyl group of histidine. Once a protein and a solid surface form coordination bonds with the same metal ion, a protein–metal ion-surface sandwich structure is observed and leads to the indirect binding (adsorption) of the protein to the surface. Increasing temperature can weaken coordination bonding, and the presence of chelating agents (e.g., ethylinediaminetetraacetic acid) can diminish the bonding due to the competitive binding of the chelating agent to the transition metal ions.

2.50.2.6

Covalent Bond

Reversible covalent bonds, for instance, disulfide bond, can be applied in the adsorption of biomolecules. The chromatographic method based on covalent bonding is called covalent chromatography.

2.50.2.7

Conformational Entropy

Adsorption results in the reduction of conformational entropy, so conformational entropy is thermodynamically unfavorable for adsorption. Therefore, adsorption takes place only if the loss in conformational entropy is compensated by sufficient attraction between the solute molecules and the surface. In general, solute adsorption involves one or more interactions described above. Moreover, in the adsorption of biomacromolecules such as proteins, molecular conformational transition is an important phenomenon for consideration. This is not only due to its relevance to the biological functioning of the molecules but also due to the significant role it plays in the adsorption process. The structural flexibility of an adsorbed protein molecule strongly affects the interactions between the protein and the solid surface, such as electrostatic and hydrophobic interactions, and then affects its adsorption phenomena. So, protein adsorption is a complex process that is controlled by a number of subprocesses at the synergistic and antagonistic effects of the interactions mentioned above.

2.50.3

Chromatographic Methods

2.50.3.1

Packed-Bed Chromatography

Packed bed is an essential mode of chromatographic operations.2,3 In this mode, a chromatographic column for preparative separations is usually packed with porous beads (adsorbents) that serve as the stationary phase. Most stationary phases consist of two functional parts – porous matrix and ligand attached to the pore surface. The porous matrix possesses sufficient mechanical strength to endure the pressure across the column at fast mobile phase flow and provides high specific surface area for the coupling of ligands and then the adsorption of target molecules. The ligands can bind solutes based on different interactions and discriminate the solutes in feedstock. Various chromatographic techniques related to the difference in ligands will be introduced in the following sections. Chromatographic matrices for bioseparations are usually hydrophilic materials that do not interact with biomolecules and related solutes; so a mild environment is provided for maintaining the native structure of biomolecules. Pore size and particle diameter are two important parameters for the matrices. The pore size usually ranges from 10 to 100 nm, which is large enough for the accessibility of biomolecules and small enough to provide high specific surface area. The use of small-sized particles can offer high column efficiency but gives rise to high pressure drop across the column. So, considering the tradeoff between column efficiency and operation pressure, the matrices for preparative chromatography are mostly in the range of 20–200 mm. Some commercial media are listed in Table 1. They are modified with different ligands for use as different adsorbents.

742

Adsorption and Chromatography Table 1

Some commercial matrices for biochromatography

Commercial name

Chemical composition

Particle diameter (mm)

Manufacturer

Sepharose Superose Sephadex Superdex Sephacryl Minibeads Monobeads Source HyperD Hypercel Trisacryl HA Ultrogel Ultrogel AcA Toyopearl Bio-Gel A Bio-Gel P UNOsphere Bio-Gel HT Fluoroapatite Poros

Agarose Agarose Dextran Dextran and agarose Allyl dextran and N,N0 -methylene bisacrylamide Polystyrene/divinyl benzene with hydrophilic coatings Polystyrene/divinyl benzene with hydrophilic coatings Polystyrene/divinyl benzene with hydrophilic coatings Polystyrene-mineral composite filled with hydrogel Cellulose N-acryloyl-2-amino-2-hydroxymethyl-1,3-propanediol Hydroxyapatite and agarose Acrylamide and agarose Hydroxylated methacrylic polymer Agarose Polyacrylamide Polyacrylamide Hydroxyapatite Fluoroapatite Polystyrene with hydrophilic coatings

34, 90, and 200 13 and 30 17–520 13 and 34 47 and 65 5 10 15 and 30 50 90 40–80 60–180 60–140 35, 65, 75, 100, and 200 38–75, 75–150, and 150–300 45–90, 90–180 80 and 120 20, 40, and 80 40 20 and 50

GE Healthcare GE Healthcare GE Healthcare GE Healthcare GE Healthcare GE Healthcare GE Healthcare GE Healthcare Pall Life Sciences Pall Life Sciences Pall Life Sciences Pall Life Sciences Pall Life Sciences Tosoh Bio-Rad Bio-Rad Bio-Rad Bio-Rad Bio-Rad Applied Biosystems

Separation by adsorptive chromatography is usually achieved in five steps: column equilibration, feed loading, washing, elution, and regeneration. First, the column is equilibrated with a loading buffer. Second, after equilibration, the feedstock for separation is loaded onto the column and the target molecules are adsorbed to the stationary phase. Third, a washing step with the loading buffer is followed to remove any unbound materials from the column. Fourth, the target molecules are eluted with an elution buffer. A regeneration solution is then applied to remove any strongly bound substances from the column. Finally, the column is re-equilibrated with the loading buffer for the next separation.

2.50.3.1.1

Size-Exclusion Chromatography

SEC, also described as gel-filtration, steric exclusion, or gel chromatography, is a partition chromatography that separates molecules according to their molecular sizes. The separation of protein mixtures according to their sizes was first reported in 1959 using cross-linked polydextran gels devoid of ionic groups for the fractionation of water-soluble substances. In SEC, molecules are eluted in the decreasing order of their sizes. Because the column is packed with a gel filtration matrix with a definite pore-size distribution, the molecules go through the column in different paths according to their sizes. A large molecule whose size is larger than the biggest pore goes through the interspaces of the gels and is eluted at the void volume of the column. Smaller molecules penetrate into the interior of the gels, leading to greater retention time. The smaller the molecular size, the more pores the molecule can penetrate into. Therefore, the largest molecules pass through the column first, while the smallest ones come last, leading to the size separation. Because SEC is a size-selective separation method, the pore-size distribution of an SEC medium is crucial for the separation performance. The difference of pore-size distributions of different SEC media is represented by the difference in their fractionation range. Initially, cross-linked gels of either dextran (Sephadex) or polyacrylamide (Bio-Gel P) were used. Furthermore, agarose gels (Sepharose) were used for the separation of solutes of even higher molecular mass (e.g., nucleic acids and viruses). Alternatively, porous glass beads provide an exclusion matrix that avoids the problems of column compaction often encountered with soft polysaccharide gels. At present, SEC media cover a fractionation range from 102 to 8  107. Most of the matrices listed in Table 1 have been made to SEC media for bioseparations. The SEC has advantageous of mild condition, simple operation, isocratic elution, and easy scaling up. It has been extensively applied in biotechnology, including the separation and analysis of proteins, peptides, lipids, antibiotics, sugars, nucleic acids, and viruses (50–400 nm); desalting of bioproduct solutions; molecular mass estimation; and characterization of molecular interactions.

2.50.3.1.2

Ion-Exchange Chromatography

Ion-exchange chromatography (IEC) is one of the most frequently used techniques for the purification of proteins and other biomolecules. It is based on the different degrees of electrostatic interactions between the stationary phase and solutes. Various cation- and anion-exchange chromatography media have been developed for protein purifications. Nucleic acids have low isoelectric point (pI) values and are usually purified by anion-exchange chromatography. Some of the frequently used ion-exchange ligands are summarized in Table 2. Based on the dissociation properties, these ligands are classified as ‘strong’ or ‘weak’ ligands. Strong ion-exchange ligands can retain their charges in a wide range of pH. By contrast, the

Adsorption and Chromatography Table 2

743

A list of ion-exchange ligands

Ligand

Structure

Comments

Sulfopropyl (SP) Methyl sulfonate (S) Carboxymethyl (CM) Quaternary ammonium (Q) Diethylaminoethyl (DEAE) Diethylaminopropyl (ANX)


O
CH2 CHOHCH2 OCH2 CH2 CH2 SO
3
O
CH2 CHOHCH2 OCH2 CHOHCH2 SO
3 –O–CH2COO
–O–CH2Nþ(CH3)3 –O–CH2CH2NH(CH2CH3)2 –O–CH2CHOHCH2NH(CH2CH3)2

Strong cation exchanger Strong cation exchanger Weak cation exchanger Strong anion exchanger Weak anion exchanger Weak anion exchanger

ionic states of weak ion-exchange ligands are pH dependent, which can in some cases offer extra selectivity. Coupling of the ligands to the matrices listed in Table 1 leads to the production of ion-exchange adsorbents. The adsorption in IEC is usually achieved at low ionic strengths (typically 20–50 mmol l
1). The pH values that are 0.5–1 unit away from the isoelectric point of the target molecule are preferable for sufficiently high capacity. Elution is often achieved by an increasing salt gradient. At high salt concentrations, the salt ions compete with biomolecules in binding with the ligands and thus the biomolecules are eluted. Elution by pH change is also optional but is less often used, because this may involve crossing the isoelectric points of proteins and lead to precipitation. Hydroxyapatite (Ca5(PO4)3OH) is a special medium of chromatography that involves both anion- and cation-exchange interactions. The Ca2þ functional groups can interact with carboxylate residues at the protein surface, while PO2
4 can interact with basic residues. Proteins are usually adsorbed on hydroxyapatite chromatography at low phosphate concentrations and eluted by increasing phosphate gradient. The NaCl and (NH4)2SO4 do not influence the adsorption of proteins on hydroxyapatite, so samples eluted from common ion-exchange columns can be directly applied to a hydroxyapatite column for further purification.

2.50.3.1.3

Hydrophobic Interaction Chromatography

Hydrophobic interaction chromatography (HIC) is a liquid chromatography to separate and purify biomolecules by their hydrophobic interaction with the hydrophobic ligands coupled to porous media. The HIC was proposed for the first time by Tiselius in 1948, using the term ‘salting-out chromatography’. The name hydrophobic interaction chromatography was introduced by Hjerten in 1973. The HIC exploits stationary phase with weakly hydrophobic ligands such as short chain alkyl and phenyl immobilized on a hydrophilic matrix. Usually, there are some exposed hydrophobic amino acids on biomolecule surface. Thus, adsorption occurs due to the hydrophobic interaction between the hydrophobic surface patches on a solute and the ligands at moderately high salt concentrations (ion strength), usually 1–2 mol l
1 ammonium sulfate or 3 mol l
1 NaCl. Because kosmotropic salts such as (NH4)2SO4 and Na2SO4 promote hydrophobic interactions, the adsorption increases with salt concentration in the mobile phase, and vice versa. Therefore, elution is usually performed via a gradient or stepwise reduction of salt concentration. Ligands are crucial for the bioseparations by HIC. Ligand chemistry can affect HIC selectivity for different proteins. Moreover, because hydrophobic interaction is proportional to ligand hydrophobicity and coupling density on the surface, ligand density should be varied according to the ligand hydrophobicity. Generally, immobilized ligand density in commercial HIC adsorbents is in the range of 10–40 mmol ml
1. Some of the commonly used hydrophobic adsorbents are provided in Table 3. The HIC can directly deal with a sample containing high salt concentration, so it is promising for the processing of samples obtained from salting-out precipitation or IEC elution. Because hydrophobic interaction strength can be readily adjusted by altering salt concentration in mobile phase, HIC is an important method in the bioseparations of therapeutic proteins, DNA vaccines, and hydrophobically tagged proteins. Table 3

Some commercially available hydrophobic adsorbents

Ligand

Adsorbent name

Manufacturer

Methyl Ether

Methyl HIC SOURCE ETH Toyopearl Ether-650 Toyopearl PPG-600 SOURCE ISO SOURCE PHE, Phenyl Sepharose Toyopearl Phenyl-600, Toyopearl Phenyl-650 Butyl Sepharose Toyopearl Butyl-600, Toyopearl Butyl-650, Toyopearl SuperButyl-550 t-Butyl HIC Toyopearl Hexyl-650 Octyl Sepharose

Bio-Rad GE Healthcare Tosoh Tosoh GE Healthcare GE Healthcare Tosoh GE Healthcare Tosoh

Polypropylene glycol Isopropyl Phenyl Butyl t-Butyl Hexyl Octyl

Bio-Rad Tosoh GE Healthcare

744

Adsorption and Chromatography

2.50.3.1.4

Affinity Chromatography

In biological systems, biomolecules can bind specifically and reversibly to their complementary substances. The specific and reversible binding effect is called affinity interactions and the complementary substances are termed affinity ligands. Affinity chromatography (AC) is an adsorptive chromatography based on the affinity interactions with immobilized affinity ligands on a solid matrix. Table 4 lists the affinity systems that are often used in AC. The binding constants of the affinity pairs for AC should be in the range 104–108 l mol
1, that is, stable enough for high recovery in affinity adsorption and not too tight for convenient desorption in the elution process. Moreover, the ligands should be immobilized to the matrix surface via a spacer to avoid steric hindrance for the accessibility of target molecules. The importance of spacing between low-molecular-mass ligand and the surface was recognized in the early development of AC. Furthermore, to maintain the affinity, macromolecular ligands (e.g., proteins) must not be deformed by immobilization. The bioaffinity ligands such as hormones and monoclonal antibodies bind complementary receptors and antigens, respectively, in a highly specific manner, so bioaffinity chromatography is advantageous because of its extremely high selectivity. However, bioaffinity chromatography can only be used to purify a specific product or a small group of related biomolecules. By contrast, AC based on metal ions and synthetic dyes are less specific ligands that can bind to a variety of proteins, so they can be widely used in bioseparations. The AC based on metal ions and dyes are respectively called immobilized metal AC (IMAC) and dye-ligand AC. Recombinant deoxyribonucleic acid (DNA) technology has made it easy to express fusion proteins with polyhistidine tags; so IMAC is an important and cost-effective technique for the purification and/or refolding of recombinant proteins. Bioaffinity interactions are highly specific because they often combine steric complementarities and different interactions, including electrostatic, hydrophobic, hydrogen bonding, coordination bonding, and van der Waals interactions. This makes bioaffinity chromatography the most selective technique for protein purification. For the AC with ligands of moderate specificities, additional separation selectivity can be achieved by a selective elution method, such as applying a gradient of ionic strength, organic co-solvents, or competitive ligands that dissociate the bound biomolecules by competitively binding to the immobilized ligands or the bound biomolecules. The latter is referred to as specific elution, which is an important feature of AC that can be employed for the improvement of separation performance.

2.50.3.1.5

Displacement Chromatography

Most of the adsorptive chromatographic separations described above are carried out in elution-mode operations in which the bound solutes are eluted by decreasing their binding strengths via adjustment of mobile phase compositions (e.g., ionic strength in IEC and HIC). Displacement chromatography (DC) is an operational mode different from the commonly used elution chromatography. In DC, following the feed loading, the column is flushed with the solution of a substance (displacer) that binds to the stationary phase more strongly than any of the components in the feedstock. Due to the competition of the displacer in binding with the stationary phase, the adsorbed solutes will be desorbed (displaced) and move toward the column outlet with the mobile phase. The solutes with higher affinity to the stationary phase will in turn serve as the displacer for the low-affinity solutes, and the final pattern will be a series of adjacent bands of different solutes moving at the velocity of the displacer. This is called isotachic displacement train. After all the solutes in feedstock are displaced, the displacer bound to the stationary phase will be washed off with a regenerant and the column re-equilibrated for the next operation. A distinct difference between DC and elution chromatography is that displaced solutes migrate in the column before the displacer, while the eluant penetrates all the solutes. Moreover, the binding strength of the solutes to the stationary phase does not change distinctly in DC. The DC has several advantages. The solute bands in DC are closely connected with each other; so the column is more effectively utilized. Moreover, solutes can be simultaneously concentrated and separated by DC, while in elution chromatography a compromise

Table 4

Affinity pairs for affinity chromatography

Ligand

Target molecule (receptor)

Antigen/antibody Hormones (vitamins) Enzyme inhibitors (substrates or cofactor analogs) Coenzyme (NAD, NADP, AMP, ADP) Protein A DNA Heparin Lectins Metal ions (Cu2þ, Ni2þ, Zn2þ, Co2þ) Dyes (Cibacron Blue 3GA, Procion Red HE-3B, etc.) Histidine

Antibody/antigen Receptor proteins, carrier proteins Enzymes Enzymes (dehydrogenases, kinases) Antibody Polynucleotide, polynucleotide-binding proteins Proteins Glycoprotein, polysaccharide Histidine-rich proteins, metal-binding proteins Proteins Proteins

ADP, adenosine diphosphate; AMP, adenosine monophosphate; NAD, nicotinamide adenine dinucleotide; NADP, nicotinamide adenine dinucleotide phosphate.

Adsorption and Chromatography Table 5

745

Displacers for bioseparations

Displacer

Application

Streptomycin A Neomycin B N-a-benzoyl-L-arginine ethyl ester Expell™ SP and Isolis™ SP* p-Toluene sulfonic acid sodium salt Ethyleneglycolbis(-aminoethylether)-N,N,N0 ,N0 -tetraacetic acid Amaranth Expell™ Q, Isolis™ Q, and Propel™ Q* Benzyl tributyl ammonium chloride N,N-bis-(3-D-glucoamidopropyl)cholamide (Big Chap)

Cation exchange Cation exchange Cation exchange Cation exchange Anion exchange Anion exchange Anion exchange Anion exchange Hydrophobic interaction Hydrophobic interaction

*ExpellÔ, IsolisÔ and PropelÔ displacers are products of Sachem Inc. (Austin, Texas, USA).

often has to be made between concentration and purity. Research has also shown that DC can offer much finer discrimination between similar substances. The DC can be performed in almost all kinds of adsorption chromatography, provided suitable displacers are available. Therefore, availability of displacers is essential for the application of DC. Some displacers, including several specially designed commercial products, are listed in Table 5.

2.50.3.2

Expanded-Bed Adsorption Chromatography

Expanded bed is a stable liquid–solid fluidized bed in which the stationary phase with controlled particle size and/or density distribution is fluidized in a liquid stream directed upward.5 The distribution of particle size and/or density within the expanded-bed system results in a distribution of terminal velocities (as calculated by Stokes’ equation), leading to a solid-phase classification within the expanded bed. The particles with larger settling velocities are found at the bottom of the bed while those with smaller settling velocities are at the top end. Thus, lower liquid dispersion level is obtained in expanded bed because this classification can reduce the mobility of the adsorbents. Hence, compared to conventional fluidized bed, there is a stable particle size and/or density classification in the axial direction, so expanded bed is a low-mixing fluidized bed with minimized solid-phase mobility and reduced axial mixing of liquid phase. As a result, the chromatographic performance of an expanded-bed adsorption (EBA) can be comparable to a packed-bed adsorption. The increase in the upward flow velocity leads to bed expansion and bed-voidage increase, thus allowing particulates in a liquid stream to pass through the bed, so EBA is particularly suitable for application in the primary isolation of bioproducts from crude feedstock-containing particulate materials such as whole cells and/or cell debris. Moreover, EBA can be integrated into cell disruption or batch fermentation processes for direct product sequestration. Hence, using the EBA technology, a reduction in the number of process steps is achieved with particular advantages in terms of processing time and product yield, thus facilitating the establishment of a cost-effective bioseparation process. The EBA has been extensively studied in various aspects such as media development, column design, as well as process fundamentals and applications.6 So, EBA for single-step purification of proteins has made great progresses, and the technology is expected to find more applications in other areas such as recovery of nanoparticles (e.g., plasmid DNA and viruses) and protein refolding. As the principal pillar supporting the development of the EBA technology, diversity of matrices is required to meet various needs in different applications. It is essential to design small-sized dense microspheres of appropriate size distribution, hopefully in a pellicular structure to reduce mass transfer resistance. Moreover, EBA matrices should be designed to minimize the interactions with particulate contaminants such as cells and cell debris in biological feedstreams. The efforts would offer more robust adsorbents for selection in the purification of different biomolecules in repeated use, making the integrative separation technology more sophisticated for widespread applications.

2.50.3.3

Electrochromatography

Electrochromatography is a liquid chromatography coupled with an external electric field (eEF). In an electric field, two electrokinetic phenomena occur, that is, electrophoresis of charged solutes and electroosmosis at a charged surface. In a typical electrochromatography, two electrodes are located at the two ends of a chromatographic column, so the eEF is applied at the longitudinal direction of the column. Therein, charged solutes flow through packed columns or open tubes via three possible modes, convection driven by pressure, electrophoresis, and electroosmosis. Both electrophoresis and electroosmosis in an eEF can promote mass transfer, thus lead to the increase of chromatographic performance. In addition, by the influence of eEF, mass transfer flux in bulk liquid phase is larger than that within particles due to the effect of diffusional resistance and size exclusion. Such a difference of fluxes can lead to more solute deposit on the surface of porous matrix, resulting in electrically induced concentration polarization (CP). The CP can change retention behavior of charged solutes in electrochromatography with porous media.

746

Adsorption and Chromatography

Various types of chromatography that are operated at low ionic strength, for instance, SEC, IEC, and AC, can be coupled with an eEF.6 However, difficulties are often encountered in the scale-up and application of electrochromatography in bioseparations due to Joule heating and electrolysis gases accompanied with the eEF. So, more detailed studies for its column design and scale-up capability for bioseparations are required.

2.50.3.4

Radial-Flow Chromatography

Normal chromatography belongs to axial-flow chromatography in which mobile phase flows through a packed column in the axial direction. In this normal column configuration, the scale-up by increasing the length of the column causes a significant increase of hydrostatic pressure needed to flow mobile phases along the column. Radial-flow chromatography (RFC) provides an efficient option for eliminating or minimizing this problem. An RFC column consists of two concentric porous cylindrical frits between which adsorbents are packed. In this configuration, liquid phase flows radially from the outer cylinder through the column and collects at the inner cylinder. As the flow path of a radial column can be much shorter than that in an axial column, the operating pressure of a radial column can be much lower, and higher flow rate can be utilized in chromatographic operations. In scale-up, the column radius can be kept unchanged, so the increase of column height gives rise to the increase of processing capacity without increasing the operating pressure. Therefore, RFC is particularly suitable for soft stationary phases, which are prone to collapse at higher hydrostatic pressure. Separation in RFC is achieved in the radial direction, so the chromatographic height is the radial thickness of the packing stationary phase. Accordingly, RFC offers fewer theoretical plates than an axial chromatography. Hence, RFC is useful for adsorptive chromatography such as IEC, HIC, and AC, but not for SEC that has low selectivity and needs larger column height for high-resolution separations.

2.50.4

Theoretical Aspects of Adsorption and Chromatography

2.50.4.1

Adsorption Equilibria

Adsorption equilibria on adsorbents are described by the relations between free solute concentration (C) and the adsorbed solute concentration (Q). In liquid–solid adsorption systems, the relations are usually determined at constant temperature, so they are called adsorption isotherms. Adsorption equilibrium data and isotherms are of importance for adsorbent evaluation, as well as process analysis, design, and optimization of adsorption and chromatography. There have been a great deal of efforts on the development of adsorption equilibrium theories, but empirical or semi-empirical equations are still largely employed to express adsorption equilibria of biomolecules.1,2 Of the various formulas, the Langmuir equation (Eq. 1) is the most widely used isotherm: Qm Ka C (1) Q¼ 1 þ Ka C where Qm is the adsorption capacity and Ka is the association constant. For n-component adsorption, the isotherm of component i is expressed by Qmi Kai Ci ði ¼ 1; 2; 3; .; nÞ (2) Qi ¼ n P 1þ Kaj Cj j¼1

At low solute concentrations, if one has

Pn

j¼1 Kaj Cj

 1, Eq. (2) can be reduced to a linear isotherm, Qi ¼ mi ci

(3)

where mi is a constant. Originally developed to represent gas adsorption, the Langmuir theory is based on three essential assumptions, that is, monolayer coverage, binding sites equivalence, and binding sites independence. In general, the assumptions do not hold for the adsorption of biomolecules such as proteins, so the Langmuir equation is regarded as an empirical expression when applied to liquid–solid adsorption systems. Nevertheless, the equation can be used to express the adsorption equilibria of a variety of solutes in a wide concentration range, including both small molecules and biomacromolecules. The Langmuir isotherm is advantageous because of its simplicity and wide applicability, but a distinct drawback of the expression is the lack of its link to the effect of liquid-phase modulators (e.g., salt concentration in ion exchange) on adsorption. Hence, research efforts have been made to develop sophisticated models taking into account the effect of liquid-phase modulators. Most of the research have focused on the effect of salt concentration on the ion-exchange equilibria of proteins. Compared with small molecules, proteins are typical of polyelectrolyte characteristics and their adsorption behaviors are more complex. Protein adsorption depends on several small regions of the protein surface (e.g., the regions rich of charges or hydrophobic patches) termed as contact regions, rather than the whole protein surface. These contact regions in liquid phase will associate with solvent or counterions by solvation and electrostatic interaction. In ion-exchange adsorption, these solvent molecules or counterions bound to the protein are displaced, which is assumed to obey a stoichiometric relationship. Some isotherm models have been developed in terms of the stoichiometric displacement law, of which the steric mass action (SMA) model7 has been

Adsorption and Chromatography

747

recognized for better description of protein adsorption equilibria. In addition to the stoichiometric displacement, the SMA model accounts for the steric shielding effect of binding sites by the bound protein, as described below: 0 1z i   B C Qi B Cs C ði ¼ 1; 2; 3; .; nÞ Ci ¼ (4) n @ A P Kai L
ðzi þ si ÞQi i¼1

where Cs is the salt concentration, zi is the characteristic charge of protein i, L is the ionic capacity of the ion exchanger, and si is the steric factor of protein i. The SMA model offers a concise form to express the effect of ionic strength on protein adsorption equilibria to ion-exchange adsorbent and has proved to well describe nonlinear adsorption chromatography of proteins at the condition of varying salt concentrations. Moreover, the model can be extended to other adsorption systems such as hydrophobic and affinity adsorptions. Ion exchange is really based on electrostatic interaction, and the long-range electrostatic interaction does not follow the stoichiometric law. Therefore, besides the empirical and semi-empirical equations described above, many efforts have been made to develop theoretical models for the ion-exchange equilibria of proteins since the 1980s (see Ref. 8 and references cited therein). In the models, both the protein and adsorbent are defined as charged bodies (e.g., a sphere for protein and a planar surface for adsorbent) surrounded by electrical double layers. By the theoretical approaches, the retention behavior or the nonlinear adsorption isotherm of protein in IEC can be predicted. Although these models offer a strictly theoretical framework to elucidate the adsorption equilibria, it is in general difficult to correctly estimate the model parameters, which limits the applicability of the models in protein chromatography.

2.50.4.2

Uptake Kinetics

Solute uptake to porous adsorbent beads experiences several sequential steps: (1) the solute migrates through the stagnant layer of liquid-film adjacent to the surface of adsorbent, (2) the solute penetrates into the pore and moves toward the adsorption site, and (3) the solute binds to the surface and bound solute keeps in equilibrium with the free solute in the pore at the same radial position. If the solute is not tightly bound on the surface, it may also migrate on the surface along the bound concentration gradient, which is called surface diffusion. Surface diffusion is in parallel to the pore diffusion. Solute binding to a surface is generally very fast as compared with the diffusive mass transfer processes, so the uptake rate usually depends on the mass transfer behaviors. Liquid-film mass transfer rate (TR) is related to the thickness of stagnant layer and properties of liquid phase, expressed by the product of a liquid-film mass transfer coefficient and a linear driving force as defined below: TR ¼ kf aðCb
Cs Þ (5) where kf is the liquid-film mass transfer coefficient, a is the specific outer surface area of adsorbent, and Cb and Cs denote the solute concentrations in bulk liquid phase and on the adsorbent surface, respectively. The liquid-film mass transfer and pore diffusion are sequential mass transfer processes, and the slower one determines the process rate. In general, the pore diffusion within porous adsorbents is the rate-limiting step in biochromatography. Hence, liquid-film resistance can be ignored particularly for protein adsorption in porous media. However, the importance of liquid-film mass transfer increases with decreases in adsorbent size and solute molecular mass (i.e., the decrease in intraparticle mass transfer resistance). Mass transfer inside porous materials is driven by pore diffusion, surface diffusion, and sometimes intraparticle convection. Because normal adsorbents have a pore diameter comparable to the mean free path or molecular size of a solute, the intraparticle diffusion is hindered by the porous structure of the matrix, especially for macromolecules. This is an important reason, for that, intraparticle diffusion dominates the uptake rate of proteins. Considering the liquid-film mass transfer, the general diffusional mass transfer model for spherical adsorbents is given by:    vC vQ 1 v 2 vC vQ ¼ 2 r εp Dp þ Ds (6) εp þ vt vt r vr vr vr t ¼ 0; Q ¼ 0; C ¼ 0 r ¼ 0; r ¼ rp ; εp Dp

vQ ¼0 vr

vC vQ þ Ds ¼ kf ðCb
CÞ vr vr

(6a) (6b) (6c)

where εp is the porosity of the adsorbent, C is solute concentration in the pore fluid, r is the radial direction, rp is the particle radius, and Dp and Ds are the diffusivities in pore fluid and adsorbed phase, respectively. Eq. (6) reduces to a pore diffusion model if Ds ¼ 0, and it reduces to a surface diffusion model if Dp ¼ 0. One can also lump the intraparticle diffusions to a single parameter, effective diffusivity, De. In this case, the differential equation is expressed as:   vQ Dc v 2 vQ ¼ 2 r (7) vt vr r vr

748

Adsorption and Chromatography

In combination with an adsorption isotherm and a mass conservation equation for an adsorption operation (e.g., well-mixed contactor or packed-bed chromatography), the kinetic equations can be solved by numerical techniques.

2.50.4.3

Theoretical Considerations of Chromatography

The performance of chromatography is dependent on the adsorption equilibria, mass transfer and adsorption (if any) kinetics, and dispersion behavior of the mobile phase. There are various theoretical descriptions of chromatography,1 and the most widely applied ones are the plate model and general rate model. The plate model is developed for linear chromatography in which the equilibrium isotherm is linear (Eq. 3). In linear chromatography, the influence of thermodynamics on chromatographic profiles vanishes. Namely, the linear isotherm of a solute controls only the position of its peak (retention time), while the kinetics of mass transfer and axial dispersion controls the peak shape (band width). The plate model depicts a continuous column by a discrete number of identical well-mixed cells. The well-mixed cells are called equilibrium stages, or theoretical plates, because the mobile and the stationary phases in each of these successive plates are in equilibrium. Thus, the kinetics of mass transfer and axial dispersion are lumped by the plate number, to which the band width or profile shape is directly related. By the plate model, column efficiency is characterized by the plate number or the height equivalent of a theoretical plate (HETP). The general rate model is a sophisticated chromatographic theory, which can simultaneously consider all the possible contributions to the chromatographic performance, including axial dispersion, liquid-film mass transfer, intraparticle diffusions, and the rate of adsorption–desorption. Certainly, some unimportant phenomena (such as adsorption kinetics and/or liquid-film mass transfer) can be ignored in the general rate model for simplicity. By an axial dispersion assumption, the mass balance equation in the mobile phase is written as vC vQ v2 C vC þP ¼ Dz 2
u vt vt vz vz

(8)

where F ¼ (1 – ε)/ε is the phase ratio (ε is the column voidage), u is the interstitial velocity, Dz is the axial dispersion coefficient, and Q is the average value of adsorbed solute concentration over the entire particle. Combining an equilibrium isotherm, a kinetic expression (e.g., Eq. 7), and proper initial and boundary conditions, the model can be solved by numerical techniques to obtain chromatographic profiles. For linear chromatography, analytical solutions can be derived. By considering the homogeneous diffusion of solute in the stationary phase (Eq. 7), the column efficiency is derived from the general rate model as ! 2Dz 2mFurP2 1 5m þ þ HETP ¼ (9) u 15ð1 þ mFÞ2 De rp kf Eq. (9) can be reduced to the classical van Deemter equation: HETP ¼ A þ

B þ Cu u

(10)

where A is the contribution of axial dispersion and dependent on the packing quality of the stationary phase as well as adsorbent shape and size distribution, B is the contribution of molecular diffusion, and C is the contribution of mass transfer resistances. Eq. (10) indicates that there is a flow rate that gives rise to a minimum value of HETP (highest column efficiency). If the axial dispersion is negligibly small and the rate of mass transfer kinetics is infinite, the general rate model is simplified to equilibrium model. The equilibrium model describes an ideal condition of chromatography (ideal chromatography), in which the free and adsorbed solute concentrations are constantly at equilibrium at any time and position in the column. Under the ideal condition, the plate number of a finite-length column is infinite, and the elution peaks in linear chromatography are identical to the injection profiles. This situation is certainly unrealistic and is usually of little importance. However, for nonlinear isotherms, ideal chromatography can qualitatively depict the influence of the isotherm shape on elution profiles.

2.50.5

Development of Adsorption and Chromatography

2.50.5.1

Innovation of Chromatographic Matrices

In liquid chromatography with a porous stationary phase, intraparticle mass transfer of macromolecules is significantly hindered; so the intraparticle diffusivity is much lower than that in bulk liquid phase and decreases more with increasing molecular size. Hence, it is recognized that intraparticle diffusive mass transport is the rate-limiting step in chromatographic processes of biomacromolecules. Therefore, chromatographic matrices need evolution to overcome this problem. In the past two decades, various efforts were made to reduce mass transfer limitations for realizing high-performance preparative biochromatography.6

2.50.5.1.1

Flow-Through Media

A direct way to the goal of elimination or alleviation of intraparticle diffusive mass transfer resistance is to open convective flow channels in size of submicron to microns in porous particles. Mobile phase can flow through the channels in chromatographic operations, so the intraparticle mass transport is greatly enhanced due to shortened diffusive path. This kind of chromatography

Adsorption and Chromatography

749

is called flow-through chromatography or perfusion chromatography and the materials with intraparticle convection are called flow-through or perfusion media. As compared with the diffusive pores of conventional media (usually, 10–200 nm), the flow-through pores are over 600 nm, one to two orders of magnitude larger than the diffusive pores. So, the wide pores through which mobile phase can flow are also called superpores or gigapores, and the corresponding particles are called superporous (or gigaporous) microspheres. Moreover, one characteristic of the material is the bimodal pore size distribution, so this kind of microsphere is also called biporous bead or bidisperse porous bead. The micropores (i.e., diffusive pores) offer large specific surface area for solute binding, so high adsorption capacity can be maintained for the adsorbents of biporous geometry. There are some superporous adsorbents commercially available, such as Poros, HyperD, and Source listed in Table 1. In recent years, continuous efforts have been made to develop different superporous media made by double emulsification and using solid granules as porogen. The convective flow of mobile phase through the superpores can lower the backpressure and the HETP value at a flow velocity up to 50 cm min
1, and the dynamic binding capacity of a biporous adsorbent can be much higher than that of a microporous one at high flow rates. So, the superporous adsorbents are promising for high-speed biochromatography.

2.50.5.1.2

Membrane

To maximize chromatographic throughput, mass transfer limitations need to be eliminated for the fast uptake of target substance, and the flow rate should be as high as possible at a given pressure drop across the bed. This leads to efforts to design short bed of large diameter and other column configurations such as RFC. In this perspective, it is obvious that an ideal chromatography column is a piece of filter because a porous membrane in thickness of 100-mm order is the shortest bed available in reality. Microfiltration membranes primarily contain flow-through pores, so the main feature of membrane-based chromatography is the absence of pore diffusion, which is the main transport resistance in conventional chromatography using porous particles. In membrane chromatography, the target solute binds to the ligands attached to the inner surface of the through pores when it flows through the pores with feedstock, so only the surface film diffusion resistance is left. For this reason, membrane chromatography can be operated at high flow rate and low pressure drop at maximum efficiency of ligand utilization. Hence, membrane chromatography offers high-speed purification of biomacromolecules such as proteins and plasmid DNA, and now it is particularly popular for antibody purification.

2.50.5.1.3

Monolith

A drawback of membrane chromatography is its low chromatographic efficiency and low binding capacity for proteins. A solution to these problems is to use monolithic columns. A monolith can be regarded as a piece of very ‘thick’ membrane. Besides larger plate number than membrane chromatography, monolithic column can offer higher binding capacity than membrane because it can be made to contain both flow-through pores and diffusive pores. Monolithic column has several advantages over packed bed of porous particles. With controlled pore structure, a monolith can offer lower mass transfer resistance. At present, monolithic materials have been widely studied for diverse applications, especially for the analytical and preparative chromatography of biomolecules such as proteins and DNA. Recently, an ideal chromatographic medium was suggested as a monolith with straightforward and evenly distributed uniform through pores (flow channels) separated by a skeleton full of diffusive pores of proper size that provide large surface area accessible for solute molecules.6 With this structural design, high-performance chromatography (high adsorption capacity/column efficiency and low back pressure at high flow rate) may be realized.

2.50.5.2

Selection and Design of Affinity Ligands

Due to the high selectivity and purification power, AC has become one of the best ways for purification of biomolecules. However, the difficulty in the use of AC is the lack of specific ligands for target molecules. So, the research focus on AC has recently been shifted toward selecting and designing ligands of high affinity and specificity. Currently, there are mainly three approaches to the discovery of or generating affinity ligands, as described below.

2.50.5.2.1

Combinatorial Library Approach

This approach focuses on the selection of ligands from large libraries constructed randomly by synthetic or biological display techniques. Using combinatorial synthesis, a huge number of structurally distinct organic molecules are synthesized at a time, which can provide a great many novel compounds for random screening. Several laboratories have recently reported the application of combinatorial synthesis methods to select affinity ligands from the libraries based on substituted triazine and peptides. Biological display is another widely used method to construct a biological combinatorial library (e.g., peptide, oligonucleotide, protein domain, and protein), and it has rapidly matured and evolved as a tool for discovering high-affinity ligands. It includes phage display, ribosome display, and systematic evolution of ligands by exponential enrichment (SELEX) methods. Phage display and ribosome display are used to construct many peptides, protein domains, or antibodies, whereas SELEX is used for oligonucleotide ligands. Phage display is basically achieved by inserting a randomized oligonucleotide sequence at an appropriate site in the structural gene of coat protein. Ribosome display utilizes a cell-free transcription, translation, and selection approach to display ligand. The SELEX exploits oligonucleotide libraries constructed by solid-phase oligonucleotide synthesis, and cell-independent enzyme-based in vitro selection approaches. When a large library is generated, it is screened and ligands are selected against the target

750

Adsorption and Chromatography

molecule in immobilized form. However, these approaches often suffer from the problems such as pseudo-positives, high cost, and long screening time.

2.50.5.2.2

Rational Design

The method uses the information of the structure of natural ligands or its target protein to upgrade or create a new ligand. Many molecular simulation techniques (e.g., docking, molecular surface analysis, and molecular dynamics (MD) simulation) have been developed to calculate, visualize, formulate, and hypothesize about the energy and orientation of candidate ligands in the pocket of a target protein. Especially, with the rapid advances in computational tools and the availability of more 3D structures of proteins obtained by X-ray crystallography, nuclear magnetic resonance (NMR) spectroscopy, and homology modeling technology, rational design of affinity ligands has become faster, more feasible, and more powerful. There are two distinct rational design methods, that is, structural template approach and functional approach. The first one is based on the knowledge of the target protein structure and the interactions between the target protein and its natural ligands or counterparts. The natural ligand-protein complex is investigated, and the conformation of the bound molecule is used as a template to design a new ligand. If there are no natural ligands or counterparts, and only the structure of the target protein is known, a candidate ligand library is first constructed and used for screening using docking softwares. Then, the affinity between candidate ligands and the target protein is evaluated using a suitable scoring function.6 The second approach is employed when no sufficient structural data are available for the target protein and a reliable protein model cannot be built by homology modeling. The method is based on preexisting knowledge for the interactions of the target protein with the functional groups, moieties, and molecular shapes of natural ligands, for example, substrates and/or inhibitors of enzymes. Ligands can be designed by: (1) exploiting recognizable molecular shape and properties, such as hydrophobicity and electrostatic potential, (2) introducing a specific functional group, and (3) combining the above two functional features on the same ligands.

2.50.5.2.3

Combination of Rational Design and Synthetic Combinatorial Library

Recently, the combination of rational design and synthetic combinatorial library emerged as a new and promising approach for ligand selection.9 It can integrate the strengths of the above two approaches. For example, this method not only enables the selection and discovery of lead compounds or groups using molecular modeling but also reflects the chemical, geometrical, and steric constraints imposed by the complex 3D solid support environment. The method involves the following steps: (1) selection of an appropriate site on the target protein, (2) design of a complementary ligand compatible with the candidate pocket using modeling techniques, (3) synthesis of a limited ligand library of structures resembling the rationally designed lead ligand, and (4) screening of the library against the target protein.

2.50.5.3

Mixed-Mode Ligands

As has been discussed in Section 2.50.3, most traditional chromatographic methods (except for AC) are dedicated to separate target molecules by the differences in a specific mode of interaction between the targets and the stationary phase. Mixed-mode interactions or the so-called nonspecific interactions are usually to be avoided, because they may counteract with each other or deteriorate resolution. However, recent researches have shown that a proper combination of different interactions can give rise to unique selectivity and facilitate the separation process. A number of mixed-mode ligands had been developed accordingly. Some of the commercially available mixed-mode media are listed in Table 6. Other mixed-mode ligands have been recently reviewed.10 Streamline Direct HST I is a mixed-mode adsorbent specially designed for EBA. The proper balance of hydrophobic interaction, electrostatic interaction, and hydrogen bonding enables high adsorption capacity of proteins in a wide range of ionic strength, whereas the capacity of traditional ion-exchange and hydrophobic adsorbents are strongly salt dependent. Therefore, crude cell extract can be directly applied to the Streamline Direct HST I column without prior adjustment of salt concentration. Elution can be achieved by simultaneously increasing the salt concentration and adjusting the pH. The ‘salt-tolerance’ of mixed-mode adsorbents can greatly simplify the purification process and reduce production cost. The 4-mercapto-ethyl-pyridine (MEP) Hypercel is a new-generation hydrophobic adsorbent with electrostatic interaction functionalities. The ligand of MEP Hypercel has a chargeable aromatic pyridine ring with a pKa of 4.85. At physiological conditions (pH  7), the ligand is uncharged and adsorption is achieved by hydrophobic interaction. When the pH is reduced to 4 or lower, the ligand will take on a proton and become positively charged. Most proteins are also positively charged at those pH values, so elution is achieved by charge repulsion. This process is denoted as hydrophobic charge induction chromatography (HCIC),11 in which proteins are adsorbed by hydrophobic interaction and desorbed by charge repulsion. This facile elution process enables easy recovery of proteins from adsorbents of high ligand densities. Moreover, because of the high ligand densities, high-capacity hydrophobic adsorption takes place at physiological salt concentrations, thus eliminating the need of high salt concentration usually employed in the adsorption stage of traditional HIC.

2.50.5.4

Displacer Screening and Design

Displacers are essential for the application and development of DC. In recent years, various approaches have been exploited to develop high-affinity displacers for protein purification, with a number of high-affinity displacers identified.

Adsorption and Chromatography Table 6

751

Some commercially available mixed-mode media

Medium

Ligand

Manufacturer

Modes of interaction

Capto™ MMC and Streamline Direct HST I

GE Healthcare

Cation exchange, hydrophobic interaction, and hydrogen bonding

Capto™ adhere

GE Healthcare

Anion exchange, hydrophobic interaction, and hydrogen bonding

MEP 4-mercapto-ethyl-pyridine Hypercel

Pall Life Sciences

Hydrophobic interaction and charge repulsion

HEA hexylamine Hypercel

Pall Life Sciences

Hydrophobic interaction and anion exchange

PPA phenylpropylamine Hypercel

Pall Life Sciences

Hydrophobic interaction and anion exchange

MBI 2-mercapto-5-benzimidazole sulfonic acid Hypercel

Pall Life Sciences

Hydrophobic interaction and cation exchange

High-throughput screening (HTS) is an efficient experimental approach to screen displacers from a number of existing compounds.12 In the screening, a known amount of adsorbent is equilibrated with a protein solution, and the amount of protein adsorbed is calculated by the equilibrium concentration in the supernatant. Then, the medium is divided into small aliquots, which are incubated with solutions of different displacers. Protein concentrations in the supernatants are then assayed, and the efficacy of displacers are denoted by the percent protein displaced or the displacer concentration needed to displace 50% of the adsorbed protein. This approach has enabled the evaluation of many displacers in parallel, thus improving the screening efficiency. Moreover, the efficacy data obtained in HTS can be used to establish quantitative structure–efficacy relationships (QSERs) with the assistance of molecular simulation software. The QSER models thus established can then be used for the virtual screening of displacers from a broad range of compounds. Another way to get high-affinity displacers is de novo design and synthesis. A common strategy of displacer design is to synthesize linear or dendrimeric polymers/oligomers from monomers that have affinity to the stationary phase. Monosaccharides can also be used as the base for high-affinity displacers, to which multiple affinitive moieties are attached via the hydroxyl groups. Recently, it was found that some displacers can selectively displace some proteins while leaving others with similar affinity on the column.13 Experimental studies and MD simulations show that the mechanism of this ‘chemically selective displacement’ is that these displacers can selectively bind the proteins and retain them on the stationary phase. Based on this mechanism, specific chemically selective displacers can be designed and synthesized by combination of a protein-binding group and a stationaryphase-binding group. The protein-binding group can be highly specific ligands used in AC or general protein-binding moieties (e.g., hydrophobic groups), and the stationary-phase-binding group is usually a known high-affinity displacer. Fluorescent hydrophobic groups have also been used as the protein-binding group for the synthesis of fluorescent chemically selective displacers for online monitoring of the displacement process.

2.50.5.5

Molecular Insight Into Protein Adsorption

Adsorption of biomolecules at liquid–solid interfaces is of fundamental importance in chromatographic separation process, so a comprehensive understanding of the adsorption phenomena and particularly the molecular mechanism is crucial for the research and development of biochromatography. Many microscopic experimental examination techniques, for instance, atomic force microscopy, NMR, X-ray crystallography, surface plasmon resonance, hydrogen–deuterium isotope exchange, and confocal laser scanning microscopy, have been used to explore the microscopic information of the process. However, none of these techniques can detect the dynamic process and protein conformational transition within adsorbent pores, which restricts not only the exploration of adsorption mechanism but also the ligand design and process optimization.

752

Adsorption and Chromatography

Molecular simulation has been used to explore the molecular insights into protein adsorption,14,15 including the description of adsorbed state, analysis of adsorption dynamics, and protein conformational transition at an interface. Molecular simulation is a powerful tool with sufficiently small scale in both time and space, thus it can offer clear microscopic information in a direct manner. It has been widely used to understand protein conformational transition at molecular-level resolution, and is becoming a fundamental technique complementary to experimental and theoretical studies. In molecular simulations, both the adsorbate and surface can be visualized using coarse-grained models or all-atom models, depending on the description precision required and the computational power provided. Furthermore, the actual chromatographic process, especially the adsorption can be monitored through the models using Monte Carlo (MC) or MD simulation. To date, molecular simulations have been successfully used to examine adsorption processes, including modeling and visualization, adsorption process, and protein conformational transition on ligand surface.

2.50.5.5.1

Modeling and Visualization

The adsorbent is usually modeled as a plate, cylinder, or sphere with immobilized ligands; the ligand conformation and surface morphology are visualized and examined through statistical mathematics. The changes caused by ligand parameters such as ligand length, composition (inclusion of embedded polar groups), and bonding density can be easily monitored and analyzed to explore the general rule, which is helpful for the rational design and fabrication of ligands. For example, a coarse-grained model was constructed to simulate porous dextran layers on the surface of a base matrix, using implicit flat and nonflat agarose surfaces, and the 3D porous structures were characterized using MD simulation. An all-atom model of ligand immobilized to agarose with a spacer arm was proposed, and the conformation of ligand–spacer–agarose was visualized using MD simulation to examine the influence of spacer arm and the interaction between agarose and ligands.

2.50.5.5.2

Adsorption Process

The adsorption behaviors, especially adsorbate-ligand interactions, adsorption and desorption processes, solvent partitioning and retention properties, can be described by molecular simulations. The effects of various chromatographic parameters can be investigated, including the composition of mobile phase, column pressure, and pore shape. For example, various chromatography behaviors in reversed-phase liquid chromatography have been examined using a constructed all-atom adsorbent model in contact with mobile phases of water/methanol mixtures. A 3D stochastic simulation was performed to provide a detailed understanding of the mass transfer processes in liquid chromatography, including the kinetics of partitioning mechanism in both homogeneous and heterogeneous systems.

2.50.5.5.3

Protein Conformational Transition

As discussed in Section 2.50.2, conformational transition is of importance in protein adsorption because it affects the recovery yield of the native product. Furthermore, the orientation and extension of protein on ligand surface strongly affects the protein-surface interactions. Thus, protein conformational transition on the ligand surface has been widely examined, focusing on the proteinligand interaction, protein orientation and conformation. For instance, the interaction between a polymer chain and planar surface as well as the adsorption and orientation of antibodies on charged surfaces have been examined by MC simulation. Moreover, MD simulations were used to study the interactions between lysozyme and the self-assembled monolayers in the presence of explicit water molecules and ions, and to investigate the initial stages of lysozyme adsorption at a charged solid interface through both all-atom model and simplified uniformly charged sphere model. The equilibrium and flow properties of polymer liquid between two brush-covered surfaces were also investigated through a coarse-grained bead-spring model. Furthermore, MD simulation studies on HCIC have been reported and molecular insights into protein conformational transition within adsorbent pores were explored. Based on the applications summarized above, it can be concluded that molecular simulations can yield practically exact results and significantly new insights at the molecular level, and thus are suitable for exploring the mechanisms of protein adsorption. It is expected that combination of computational quantum chemistry at quantum level, molecular simulations at atomistic level, and experiments at macroscopic level can get more comprehensive understanding of protein adsorption and chromatography.

2.50.6

Conclusions

Various chromatographic methods have been widely applied in the practice of downstream processing of peptides, proteins, nucleic acids, viruses, and many other biological substances because of their favorable characteristics in high resolution, wide availability, and mild operating conditions. At present, biochromatography is still a major area of bioseparation research activities, and new achievements are reported continuously. It is expected that more efforts will be denoted to the fundamentals of biomolecule adsorption and the innovation of chromatographic materials, including matrices, ligands, and displacers. In this process, molecular simulations will play an important role in understanding molecular interactions at liquid–solid interfaces, adsorption equilibrium, kinetics, and molecular transport phenomena in adsorbent pores, besides experimental and theoretical studies.

Adsorption and Chromatography

753

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

Guiochon, G.; Felinger, A.; Shirazi, D. G.; Katti, A. M. Fundamentals of Preparative and Nonlinear Chromatography, 2nd ed.; Elsevier: Amsterdam, 2006. Subramanian, G., Ed.; Bioseparation and Bioprocessing, Vol. I: Biochromatography, Membrane Separations, Modeling, Validation, Wiley-VCH: New York, NY, 1998. Janson, J.-C., Rydn, L., Eds.; Protein Purification: Principles, High-resolution Methods, and Applications, Wiley-Liss: New York, NY, 1998. Norde, W. Driving Forces for Protein Adsorption at Solid Surfaces. In Biopolymers at Interfaces; Malmsten, M., Ed., 2nd ed.; Dekker: New York, NY, 2003. Hubbuch, J.; Thömmes, J.; Kula, M. R. Biochemical Engineering Aspects of Expanded Bed Adsorption. Adv. Biochem. Eng. Biotechnol. 2005, 92, 101–123. Sun, Y.; Liu, F. F.; Shi, Q. H. Approaches to High-performance Preparative Chromatography of Proteins. Adv. Biochem. Eng. Biotechnol. 2009, 113, 217–254. Brooks, C. A.; Cramer, S. M. Steric Mass-action Ion Exchange: Displacement Profiles and Induced Salt Gradients. AIChE J. 1992, 38 (12), 1969–1978. Su, X. L.; Sun, Y. Thermodynamic Model for Nonlinear Electrostatic Adsorption Equilibrium of Protein. AIChE J. 2006, 52 (8), 2921–2930. Rupasinghe, C. N.; Spaller, M. R. The Interplay Between Structure-based Design and Combinatorial Chemistry. Curr. Opin. Chem. Biol. 2006, 10 (3), 188–193. Zhao, G. F.; Dong, X. Y.; Sun, Y. Ligands for Mixed-mode Protein Chromatography: Principles, Characteristics and Design. J. Biotechnol. 2009, 144, 3–11. Burton, S. C.; Harding, D. R. Hydrophobic Charge Induction Chromatography: Salt Independent Protein Adsorption and Facile Elution with Aqueous Buffers. J. Chromatogr. A 1998, 814 (1–2), 71–81. Mazza, C. B.; Rege, K.; Breneman, C. M.; et al. High-throughput Screening and Quantitative Structure–efficacy Relationship Models of Potential Displacer Molecules for Ion-exchange Systems. Biotechnol. Bioeng. 2002, 80 (1), 60–72. Rege, K.; Ladiwala, A.; Tugcu, N.; et al. Parallel Screening of Selective and High-affinity Displacers for Proteins in Ion-exchange Systems. J. Chromatogr. A 2004, 1033 (1), 19–28. Mungikar, A. A.; Forciniti, D. Computer Simulations and Neutron Reflectivity of Proteins at Interfaces. ChemPhysChem 2002, 3 (12), 993–999. Raffaini, G.; Ganazzoli, F. Understanding the Performance of Biomaterials Through Molecular Modeling: Crossing the Bridge Between Their Intrinsic Properties and the Surface Adsorption of Proteins. Macromol. Biosci. 2007, 7 (5), 552–566.

2.51

Modeling Chromatographic Separationq

AR O¨zdural, Hacettepe University, Ankara, Turkey © 2017 Elsevier B.V. All rights reserved. This is a reprint of A.R. Özdural, Modeling Chromatographic Separation, Reference Module in Life Sciences, Elsevier, 2017.

2.51.1 2.51.2 2.51.2.1 2.51.2.1.1 2.51.2.1.2 2.51.2.1.3 2.51.2.2 2.51.3 2.51.3.1 2.51.3.1.1 2.51.3.1.2 2.51.3.1.3 2.51.3.1.4 2.51.3.2 2.51.3.2.1 2.51.3.2.2 2.51.4 2.51.4.1 2.51.4.2 2.51.4.3 2.51.4.4 2.51.4.5 2.51.4.6 2.51.5 References

Introduction Theoretical Background Intraparticle Diffusion Pore diffusion Surface diffusion Homogeneous diffusion Particle Concentration Profile Development Models for Chromatography Formulation of the Models The ideal model The equilibrium dispersive model The transport dispersive model The general rate model Alternative Method for the Numerical Solution of the GR Model With Nonlinear Isotherms Single component solution Multi component solution Case Studies Case Study 1 Case Study 2 Case Study 3 Case Study 4 Case Study 5 Case Study 6 Summary

755 756 756 757 757 758 758 759 760 760 760 761 761 762 762 763 765 765 766 766 767 768 769 771 771

Nomenclature A, B Components a, b Parameters in Eq. (33) Bi Biot number kf/(Dsrp) c Solute concentration in the bulk eluent, mg mL
1 cinj Sample injection concentration, mg mL
1 cp Solute concentration of pore liquid, mg mL
1 c* Interphase solute concentration of eluent, mg mL
1 Da Axial dispersion coefficient, cm2 s
1 Dc Column inside diameter, cm Dp Particle diameter, cm Dpr Pore diffusivity, cm2 s
1 Ds Homogeneous solid diffusivity, cm2 s
1 Dsr Surface diffusivity, cm2 s
1 F Eluent volumetric flow rate, cm3 min
1 g1, g2 Parameters in Eq. (13) h Increment in distance cm i x panel index used in numerical solution j t panel index used in numerical solution

q

Change History: July 2016. A.R. Özdural updated Nomenclature, Section 3.2 "Alternative Method for the Numerical Solution of the GR Model with Nonlinear Isotherms," and Section 4 "Case Studies"; added Figs. 12 and 13, Table 1, and five new references to the References section.

754

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-12-809633-8.09098-1

Modeling Chromatographic Separation

755

K Constant in Langmuir isotherm, mg mL
1 k Increment in time, s kf Film mass transfer coefficient, cm s
1 km Lumped mass transfer coefficient used in LDF model, s
1 L Column length, cm n Number of components Q Solute concentration in the particle, whether free (ie, in the pore liquid) or in the skeleton, mg mL
1particle q Same as Q for homogeneous diffusion, mg mL
1particle qm Langmuir isotherm maximum adsorption capacity, mg mL
1particle qs Interphase solute concentration of particle, mg mL
1particle qsp Solute concentration in the particle, given by Eq. (3), mg mL
1particle qss Solute concentration in the skeleton, given in per skeleton volume, mg mL
1skeleton q Solute average concentration in the particle, mg mL
1particle q* Solute concentration in the particle that is in equilibrium with the local bulk mobile phase concentration, mg mL
1particle r Radial direction in the particle rp Particle radius, cm t Time, s Vinj Sample injection volume, mL z Column axial distance, cm

Greek Letters a Component index (-) b Component index (-) ε Void fraction in the bed (-) εp Particle porosity (-) zA, zB Proportionality constants in Eqs. (34) and (35) v Interstitial velocity, cm s
1 hA, hB Parameters defined in Eqs. (44) and (46)

2.51.1

Introduction

Chromatography is a technique for molecular partition in which a fluid (mobile phase) carries the material containing the mixture to be separated (sample) past or through a solid or gel (stationary phase) contained in a vessel. It is a highly selective process of separation which is often employed in the separation of complex mixtures such as sugars, proteins, pharmaceuticals, fine chemicals, flavorings, foods, enantiomers, isomers [1] and isotopes [2]. The stationary phase has characteristics that delay the passage of some molecular components of the sample more than the passage of others causing them to separate in the mobile phase emerging from the column. A significant number of chromatographic separation operations use porous materials as a stationary phase [3]. In batch column chromatography, the feed pulse is introduced on top of the stationary phase and eluted with the mobile phase along the column length. During a batch operation, several components may be separated from a mixture. But it is a discontinuous process and leads to highly diluted products, and furthermore it is generally expensive in large-scale separation processes. As opposed to conventional batch chromatography, continuous chromatographic separation processes, mainly based on simulated moving bed (SMB) process [4], have gained greater interest in the last decades due to their advantages in terms of productivity and eluent consumption. Lately, in the separation of fine chemicals, particularly chiral molecules, continuous chromatography is now an established technique [5]. Modeling and simulation studies indicate that SMB systems might also be applied in reactive chromatic separation processes [6]. Another continuous separation system is the continuous annular chromatography (CAC) that allows largescale continuous preparative chromatographic separation and purification [7] of multi-component systems. The interest in continuous chromatography has motivated a great deal of theoretical work to achieve a better understanding of SMB and CAC to devise useful simulation procedures for design and process development purposes. Both for batch and continuous chromatographic processes, the model-based understanding of chromatographic separation largely prevents the inconveniences of experimenting with the real process. The simulations by modeling gradient elution in non-linear ion-exchange chromatography provided a new insight into the phenomena involved in bio-chromatography [8]. Exploring new operating conditions to respond to changes in product purity demands, capacity loads and feed compositions can be carried out before committing to a fixed chromatography

756

Modeling Chromatographic Separation

separation unit. The formulation of batch chromatography mathematical models is also a crucial step for understanding continuous chromatographic separation processes modeling as well, since similar governing transport equations hold. Furthermore, batch chromatography is still a standard operating mode due to its flexibility. Therefore, here we focused our attention on modeling of batch column chromatographic separation. The common and distinct model parameters are obtained either from the experimental setup or from literature correlations. The optimization and scale-up of separation conditions for preparative column chromatography requires knowledge of isotherm model, and a proper understanding of the adsorption equilibrium is necessary. The equilibrium adsorption isotherm is essential to describe the interactions between the solute molecules in the mobile phase and the molecules adsorbed on the stationary phase at a constant temperature. In linear chromatography, the equilibrium isotherm is defined by a linear equation. However, nonlinear effects occur in most applications of chromatography, and many nonlinear adsorption isotherm models are available in the literature [1]. The isotherm parameters are often determined experimentally. In the classical method of the adsorption isotherm measurement, the calculated parameters do not depend on mass transfer resistances since the adsorbent and the fluid phase are brought into batch-wise contact in flasks and shaken for prolonged times in order to secure that the equilibrium is reached. Usually a local equilibrium assumption is made in less sophisticated chromatographic separation modeling, ie, at any instant, equilibrium conditions prevail throughout a column. The validity of the local equilibrium assumption is dubious in a dynamic state that exists in a chromatographic column. Thus, the incorporation of isotherm parameters, determined through classical method, to local equilibrium-based models, becomes cumbersome unless the mass transfer resistances are small and have a minor influence on the adsorption isotherm profiles. When local equilibrium conditions are falsely assumed, a dynamic isotherm parameter estimation technique, such as frontal analysis, usually helps to avoid the inconsistency between the experimental data and model predictions to a certain extent. But this does not alter the reality; ie, the accuracy of dynamic isotherm parameter determination techniques largely depends on the column efficiency. In the following sections, the diffusion models used in chromatography modeling are briefly described, and a comparative description of the general nature of modeling chromatographic separations is given.

2.51.2

Theoretical Background

There are two fundamental chromatography theories that deal with solute retention and solute dispersion, the plate theory and rate theory, respectively. The first effective theory to be developed was the plate theory. The original plate theory was first applied to chromatography by Martin and Synge [9] who borrowed the equilibrium plate concept from distillation towers and treated a chromatographic column as if it were made up of adjacent plates. The solute moves down the column by transfer of equilibrated mobile phase from one plate to the next. The plates serve as a way of measuring column efficiency, by stating the number of theoretical plates in a column. The plate theory has come under some criticism because the theory assumes that the solute is in continuous equilibrium with mobile and stationary phases, which in fact can hardly be realized in a chromatographic system. Skoog et al. [10] justify the retaining of well entrenched “plate number” term in the chromatographic literature for historic reasons only, and not because it has physical significance. A more realistic description of the chromatographic separation process inside a column takes account of the time taken for the solute to equilibrate between the stationary and mobile phase. The resulting band shape of a chromatographic peak is therefore influenced by the rate of elution. The rate theory provides better understanding of mass transfer effects. The main features of rate-based chromatographic column modeling studies are the effect of intraparticle and external mass transfer resistances during solute transfer into the solid phase and solute elution rate, axial dispersion within the column, adsorption equilibrium between the mobile and stationary phases and the adsorption rate. Their influences on simulation results differ in various mathematical models that are describing the chromatographic column separation process, due to the simplifying assumptions made during the model development studies. It should also be noted that simulation results may vary with the molecular properties of the compounds investigated, as well as the nature of the stationary and mobile phases. In addition to the mass transfer resistances and axial dispersion, the model complexity and number of model parameter increases when nonlinear isotherms and radial dispersion are incorporated into the modeling studies.

2.51.2.1

Intraparticle Diffusion

Intraparticle diffusion of solute molecules, as illustrated in Fig. 1 for a porous adsorbent particle, is usually explained with three models: pore diffusion, surface diffusion and homogeneous solid diffusion. Eqs. (1) and (2) give the adaptation of Fick’s second law of diffusion in a spherical porous solid where pores are filled with a stationary liquid.
 vQ 1 v 2 vQ ¼D 2 r (1) vt r vr vr   Q ¼ εp cp þ 1
εp qss

(2)

Modeling Chromatographic Separation

757

Intraparticle diffusion

rp r Skeleton

Pore Fig. 1

Schematic drawing of a porous adsorbent.

Q is the solute concentration in the particle (whether in the pore liquid or in the skeleton), cp is the solute concentration in the pore liquid, qss is the skeleton solute concentration, t is the time, r is the radial direction within the particle, D is the intraparticle diffusivity, and 3p is the particle porosity. Eq. (3) gives the definition of a new concentration term, qsp, where the amount of solute in the skeleton is divided with particle volume (poreþskeleton).   qsp ¼ 1
εp qss (3) Substitution of Eq. (3) into Eq. (2) gives Q ¼ εp cp þ qsp

2.51.2.1.1

(4)

Pore diffusion

In the pore diffusion model, diffusion is assumed to take place in the liquid-filled pores. The driving force for intraparticle mass transfer is solute concentration gradient in the pore phase of the particle. Taking the derivative of Eq. (4) with respect to r yields vcp vqsp vQ ¼ ep þ vr vr vr

(5)

In the pore diffusion model, the second term of the right-hand side of Eq. (5) can be neglected, since the mass transfer driving force is the pore–liquid concentration gradient. vcp vQ ¼ ep vr vr

(6)

vcp vqsp vQ ¼ ep þ vt vt vt

(7)

Taking the derivative of Eq. (4) with respect to t

Substitution of Eqs. (6) and (7) into Eq. (1) gives ep


 vcp vqsp 1 v 2 vcp r þ ¼ Dpr 2 ep r vr vt vt vr

(8)

where intraparticle diffusivity D is replaced with pore diffusivity Dpr. Eq. (8) is known as the pore diffusion equation in porous particles.

2.51.2.1.2

Surface diffusion

The surface diffusion model assumes that the driving force for intraparticle mass transfer is solute concentration gradient in the particle skeleton, ie, vqss/vr. Taking the derivative of Eq. (2) with respect to r. vcp vQ vqss ¼ ep þ ð1
ep Þ vr vr vr

(9)

In the surface diffusion model, the first term of the right-hand side of Eq. (9) can be neglected since the mass transfer driving force is the particle-skeleton concentration gradient. Then substituting Eq. (3) into Eq. (9) gives: vQ vqsp ¼ vr vr

(10)

Substitution of Eqs. (7) and (10) into Eq. (1) gives: ep


 vcp vqsp 1 v 2 vqsp r þ ¼ Dsr 2 r vr vt vt vr

(11)

758

Modeling Chromatographic Separation

where intraparticle diffusivity D is replaced with surface diffusivity Dsr. Eq. (11) is known as the surface diffusion equation in porous particles.

2.51.2.1.3

Homogeneous diffusion

The homogeneous diffusion model is often called the homogeneous solid diffusion model. Here all solute within the particle, whether it is in pore liquid or adsorbed by particle skeleton, is lumped into a single quantity, q. The driving force for intraparticle mass transfer is assumed to be governed by the gradient in the total solute concentration, ie, vq/vr. Thus Eq. (1) is directly applicable to the homogeneous diffusion model by replacing Q with q and D with homogeneous solid diffusivity, Ds. Eq. (12) is known as homogeneous diffusion equation.
 vq 1 v 2 vq ¼ Ds 2 r (12) vt r vr vr

2.51.2.2

Particle Concentration Profile Development

Local equilibrium assumption neglects the development of a solute concentration profile within adsorbent particles, ie, it is assumed that particle concentration is homogeneous at any instant. However, unless the internal mass transfer resistances are negligible, the local equilibrium assumption might lead to erroneous conclusions since slow kinetics strongly influence the band broadening effect. Only at near particle-saturation condition might the assumption of homogeneous concentration distribution within the particle become valid [11]. On the other hand, in elution chromatography processes, the stationary phase remains in an unsaturated condition due to the presence of the eluent stream in the column. Therefore, for chromatographic separation processes, a quantitative account of the concentration profile development within the particle is necessary. For a parabolic concentration profile, Eq. (13) gives the change of solid concentration with time, t and particle radius, r at a column axial distance, z, where g1(z,t) and g2(z,t) are column height- and time-dependent parameters. qðz; r; tÞ ¼ g1 ðz; tÞþ g2 ðz; tÞr 2

(13)

Using the homogeneous solid diffusion model, we can write the boundary condition at the external surface of particle in a chromatography column where rp is the particle radius, kf is the film mass transfer coefficient, c(z,t) and c*(z,t) are the bulk and interface liquid concentrations as shown in Fig. 2.    vqðz; r; tÞ Ds ¼ kf cðz; tÞ
c ðz; tÞ (14)  vr r¼rp Taking the derivative of Eq. (13) with respect to r and evaluating at r¼rp gives:  vqðz; r; tÞ ¼ 2g2 ðz; tÞrp  vt r¼rp From Eqs. (14) and (15)



1 g2 ðz; tÞ ¼ Bi cðz; tÞ
c ðz; tÞ 2

where Bi is the Biot number kf/(Dsrp). Substituting Eq. (16) into Eq. (13) and solving for g1(z, t) at particle surface gives:

Mobile phase c *(z,t ) qs(z,t ) c(z,t )

r

Film layer

q (z,t )

Solid phase rp

q(z,r,t ) = a1(z,t ) + a2(z,t )r2

Solid adsorbent particle Fig. 2

Solute adsorption in a spherical adsorbent particle.

(15)

(16)

Modeling Chromatographic Separation 1 g1 ðz; tÞ ¼ qðz; r; tÞjr¼rp
Bi½cðz; tÞ
c ðz; tÞ 2

759

(17)

Mean solid concentration, qðz; tÞ can be expressed in terms of volume average concentration. R rp 4pr 2 qðz; r; tÞdr qðz; tÞ ¼ 0 4pr 3 3

(18)

Substituting Eqs. (13), (16) and (17) into Eq. (18) and integrating the final expression in order to find the relationship between particle surface and average concentrations. 1 qs ðz; tÞ ¼ qðz; tÞþ Bi½cðz; tÞ
c ðz; tÞ 5

(19)

In Eq. (19) and the remaining part of this article, for the sake of simplicity, stationary phase concentration at particle surface, qðz; r; tÞjr¼rp is shown as qs(z,t). Similarly, one can derive the following expression for the particle center concentration in a chromatographic column. qðz; r; tÞjr¼0 ¼ qðz; tÞ


2.51.3

3 Bi½cðz; tÞ
c ðz; tÞ 10

(20)

Models for Chromatography

Several groups of researchers have proposed and solved chromatographic column models at different levels of complexity and accuracy [12–19] whether at the analytical scale or at the preparative/production scale. The use of sophisticated models allows the detailed investigation of separations for which the mass transfer kinetics is slow. In one-dimensional chromatographic simulation and modeling studies, the complexity of the problem increases if transient resistances are brought into play. By coupling energy balance equations that allows the study of non-isothermal behavior further increases the complexity of modeling of chromatographic separation. For modeling gas chromatographic separations, the control of retention by the fluid compressibility should also be taken into account. In this article, only the incompressible mobile phase case is investigated. Thus, the discussions apply to liquid chromatography where the assumption of isothermal operation is usually valid. The major chromatography models are the ideal model (IM), the equilibrium dispersive model (ED), the transport dispersive model (TD) and the general rate model (GR). Fig. 3 illustrates the important characteristics of these models. The ideal model, based on the equilibrium theory of chromatography, has been a powerful tool to analyze the dynamics of chromatographic columns for decades. IM assumes that the equilibration of a solute between the mobile and stationary phase is an infinitely rapid process. Any contributions from hydrodynamic effects or mass transfer phenomena are neglected. This model estimates the highest production and recovery rate allowed by the thermodynamics. Therefore, it may be useful for the analysis of the qualitative behavior of the process and the influence of some of the process parameters. In order to account for band

A

Sample

Stationary phase

Eluent Mobile phase

Instantaneous adsorbent equilibrium*

Convective mass transfer

B

Sample

Eluent

Stationary phase

Mobile phase

Instantaneous adsorbents equilibrium*

Convective mass transfer + lumped axial dispersion

C

Sample

Eluent

Stationary phase

Mobile phase

Noninstantaneous adsorbent equilibrium*

Convective mass transfer + axial dispersion Mobile and stationary phase mass transfer resistances lumped into a single mass transfer coefficeint

D

Sample Stationary phase

Eluent Mobile phase

Noninstantaneous adsorbents equilibrium*

Mass transfer resistance in the mobile phase + concentration profile within the particle

Convective mass transfer + axial dispersion + mass transfer resistance in the mobile phase

Fig. 3 Models for chromatography: (A) the ideal model, (B) the equilibrium dispersive model, (C) the transport dispersive model, and (D) the general rate model.

760

Modeling Chromatographic Separation

broadening, the ED and the TD models are employed by many authors [13]. In the ED model, similar to the IM model, instantaneous equilibrium is assumed between the stationary and the mobile phases, and an apparent axial dispersion term is used to describe and understand the band broadening effects of both axial dispersion and the finite rate of the mass transfer kinetics. In the TD model, individual mass transfer resistances are not directly taken into effect, instead a lumped mass transfer coefficient is used, where pseudo homogeneous concentration within the particle is assumed. For the mathematical modeling of processes with poor efficiency, neither the ED model nor the TD model is sufficiently accurate [12]. The GR model is widely acknowledged as being the most comprehensive among the chromatography models available in the literature as it accounts for axial dispersion and all the mass transfer resistances, for example, external mass transfer of solute molecules from bulk phase to the external surface of the adsorbent, diffusion of the solute molecules through the particle, and adsorption–desorption processes on the site of the particles. When the influence of mass transfer resistances is important, the result of smoother peaks compared to the ideal shape can successfully be explained by the GR model. The GR model involves two partial differential equations in space and time coordinates in its mass balances. The solution of the model involves simultaneous calculation of mobile phase concentration profiles that percolate through the column and the particles concentration. Therefore, it is more computationally demanding than the simpler chromatography models. The ED model and to a certain extent the TD model can be solved in a short time with a fair degree of accuracy, while the GR model captures the process dynamics more accurately, as it takes into account all the mass transfer resistances in the process. The application of either model, however, also requires knowledge of how to obtain the model parameters from experimental and literature data.

2.51.3.1

Formulation of the Models

It is not difficult to derive Eq. (21) by making a mass balance on a chromatography column, where axial molecular diffusivity term is replaced with axial dispersion term, Da, and it is assumed that dispersion mimics diffusion in the sense that the dispersive fluxes in z dimension appear to be driven by concentration gradients, and can be expressed using the same mathematical form as Fick’s law for diffusive flux. vcðz; tÞ vcðz; tÞ 1
ε vqðz; tÞ v2 cðz; tÞ þv þ ¼ Da vt vz ε vt vz2

(21)

where ε is the void fraction in the bed, v is the eluent interstitial velocity and Da is the axial dispersion coefficient. The third term in the left-hand side of Eq. (21) is the time change of stationary phase average concentration, ðvqðz;tÞ=vtÞ, and it can be estimated by various intraparticle particle diffusion equations. In order to obtain the eluent concentration, c(z,t), it is required to solve a coupled partial differential equation system composed of Eq. (21) and an intraparticle diffusion equation such as pore diffusion, surface diffusion or homogeneous diffusion, given by Eqs. (8), (11) and (12), respectively. Furthermore a relationship between c(z,t) and qðz; tÞ is required during the simultaneous solution of the coupled partial differential equation system of interest. An appropriate adsorption isotherm expression might be used for this purpose by keeping in the mind that under non-equilibrium conditions, instant equilibrium may hold only at the liquid–solid interphase. Eq. (21) can also be treated by making several simplifying assumptions, those leading to different chromatography models.

2.51.3.1.1

The ideal model

As shown in Fig. 3(A) in this model, it is assumed that the external and the intraparticle resistance to mass transfer are not taking place, and instant equilibrium is reached between the stationary and the mobile phases concentrations, ie, between qðz; tÞ and c(z,t), respectively. Consequently, the particle concentration profile, which is illustrated in Fig. 2, is disregarded. In the IM, by making use of equilibrium isotherm expression, one can calculate stationary phase average concentration, qðz; tÞ as long as the bulk liquid concentration c(z, t) is known and vice versa. Furthermore, the IM neglects axial dispersion effects. If the isotherm is linear, the elution profile is essentially a reflection of the solute injection profile shifted by a time equal to column hold-up time. For ideal model, Eq. (21) reduces to Eq. (22). vcðz; tÞ vcðz; tÞ 1
ε vqðz; tÞ þv þ ¼0 vt vz ε vt

(22)

For several isotherms, it is possible to solve Eq. (22) analytically by using the method of characteristics. The use of numerical solution techniques are required for complex isotherm expressions. In the case of binary mixtures where competitive Langmuir isotherm holds, the band profiles can be calculated using algebraic equations once retention time of the first component is known [13]. Nevertheless, due to the unrealistic assumptions of ideality, the accuracy of prediction of this modeling approach is not sufficient for design, control or optimization purposes.

2.51.3.1.2

The equilibrium dispersive model

The difference between the IM and the ED models is the incorporation of axial dispersion term into ED model. The particle concentration profile, which is illustrated in Fig. 2, is disregarded, but the drawback of instant equilibrium conditions, as it does not account for band broadening, is surmounted by a simple means. For this purpose, all the mass transfer resistances along with the true axial dispersion effects are included into a lumped axial dispersion term. The effect of particle size on chromatographic separation cannot be explained with this model, due to the fact that particle diameter is not incorporated into the model

Modeling Chromatographic Separation

761

development. Fig. 3(B) illustrates the ED model. Similar to the ideal model, if the equilibrium isotherm expression is available, stationary phase average concentration, qðz; tÞ can be obtained as long as the bulk liquid concentration c(z,t) is known and vice versa. By replacing Da in Eq. (21) with lumped axial dispersion coefficient, Dal, we get Eq. (23) for the ED model. vcðz; tÞ vcðz; tÞ 1
ε vqðz; tÞ v2 cðz; tÞ þv þ ¼ Dal vt vz ε vt vz2

(23)

The value of the lumped axial dispersion coefficient is assumed as constant. Model output eluent concentration at column exit versus time data are generated by substituting different values for lumped coefficient into the model until the predicted and experimental profiles agree. Analytical solution of the ED model is possible, and the simplest solution of this model assumes that at the inlet of a hypothetically very long column, an infinitely narrow pulse injection is made. Under nonlinear conditions, although approximate analytical solutions are available, there are no exact analytical solutions available for the ED model, and several numerical solution techniques have been proposed [13]. On the other hand, stationary phase external and internal mass transfer resistances reflect entirely different mass transfer mechanisms, as well as the axial dispersion. Especially for the separation of slowly diffusing molecules, lumping all the mass transfer resistances into an apparent dispersion coefficient might lead to erroneous conclusions, due to the fact that the internal and external mass transfer resistances and axial dispersion do not usually respond in harmony to the changes in column operation parameters. For example, the stationary phase internal mass transfer resistance is almost immune to that of the mobile phase interstitial column velocity, whereas the mobile phase velocities have a direct effect on the dispersion. Additionally, when interstitial column velocity increases so does the external film mass transfer coefficient. This phenomenon tends to lessen the band broadening. However interstitial column-velocity increase results increase axial dispersion, which has a positive effect on the band broadening. Therefore, the internal and external mass transfer resistances are not additive properties with the axial dispersion, and when column efficiency is poor, the ED model should not be applied for modeling chromatographic separation.

2.51.3.1.3

The transport dispersive model

In this model, boundary layer and intraparticle mass transport resistances are lumped into a mass transport coefficient. Solute uptake mechanism is based on linear driving force model (LDF) which can be described as: the adsorption rate of a single adsorbate into an adsorbent particle is essentially proportional to the amount of adsorbate still required to produce equilibrium in the adsorbent [20]. Fig. 3(C) illustrates the TD model. Eq. (24) gives the mathematical definition of LDF model for a chromatography column, where q*(z,t) is the solute concentration in the particle that is in equilibrium with the local bulk mobile phase concentration, c(z,t). vqðz; tÞ ¼ km ½q ðz; tÞ
qðz; tÞ vt

(24)

Substitution of Eq. (24) into Eq. (21) gives the TD model governing differential equation. vcðz; tÞ vcðz; tÞ 1
ε v2 cðz; tÞ þv þ km ½q ðz; tÞ
qðz; tÞ ¼ Da vt vz ε vz2

(25)

There are no closed-form analytical solutions to the TD model. However, several numerical solution techniques might be used due to the comparative simplicity of the model. During the numerical solution of Eq. (25), the drawback of the simplified pseudo homogeneous models brought into play, ie, although in the TD model c(z,t) is not in instant equilibrium with qðz;tÞ, which is actually the condition of the IM and ED models, it is assumed to be in instant equilibrium with q*(z,t). This assumption makes the calculation of q*(z,t) value possible, and its substitution into Eq. (24) allows the evaluation of vqðz; tÞ=vt. As in the ED model, here, concentration profile development within the particle is ignored. It is reported that in some cases km value shows an inexplicable dependency on the sample concentration [21].

2.51.3.1.4

The general rate model

The GR model accounts for external and intraparticle mass transfer resistances, as well as axial dispersion in the column. A schematical illustration of this model is given in Fig. 3(D). The effect of development of particle concentration profile on eluent concentration profiles is reflected in the model equation. The time change of average particle concentration is due to the rate of transfer of solute through the particle liquid film layer. Referring to Fig. 2, the driving force for film mass transfer rate is the difference between bulk eluent concentration, c(z,t) and eluent interphase concentration c*(z,t). For spherical particles, vqðz; tÞ 3kf ¼ ½cðz; tÞ
c ðz; tÞ vt rp

(26)

Substitution of Eq. (26) into Eq. (21) gives Eq. (27) which is the GR model governing differential equation, vcðz; tÞ vcðz; tÞ ð1
εÞ 3kf v2 cðz; tÞ þv þ ½cðz; tÞ
c ðz; tÞ ¼ Da vt vz ε vz2 rp

(27)

Regarding the mass transfer resistances, Eq. (27) contains the film mass transfer coefficient, kf only. But stationary phase internal resistance is inherently included in the c*(z,t) term which is susceptible to the particle concentration profile, and

762

Modeling Chromatographic Separation

thereby to the stationary phase internal resistance. Eq. (27) contains two dependent variables, c(z,t) and c*(z,t), and it cannot be solved in its present form. We can express the mobile phase interphase concentration, c*(z,t), in terms of the stationary phase interphase concentration, qs(z,t). As shown in Fig. 2, it is reasonable to assume that at the interphase c*(z,t) is in equilibrium with qs(z,t) since adsorption itself (transfer of solute at the interphase to adsorbed state) is generally very fast. In Eq. (27) the value of c*(z,t) can be stated through an appropriate isotherm expression by the employment of qs(z,t) value. Then, the new challenge is to express the stationary phase interphase concentration, qs(z,t) in terms of the bulk mobile phase concentration, c(z,t) so as to have only one dependent variable in Eq. (27). For this purpose the selected (either pore or surface or homogeneous) partial differential equation of intraparticle diffusion and Eq. (27) is simultaneously solved by bringing the isotherm expression, which relates interphase concentrations, into play. The GR model is complex and requires the employment of sophisticated numerical algorithms, such as orthogonal collocation on finite elements using a variety of initial and boundary conditions. Its numerical solution has been discussed extensively in the literature [12–16,19,22] in connection with the linear and nonlinear adsorption isotherms.

2.51.3.2

Alternative Method for the Numerical Solution of the GR Model With Nonlinear Isotherms

As explained above, the lengthy computational time requirements experienced in numerical solutions of the GR model with nonlinear isotherms stems from the necessity of a solution of coupled partial differential equations. For homogeneous diffusion, Özdural et al. [11] proposed a new algorithm for numerical solution of GR model, though it might well be applied to other stationary-phase diffusion mechanisms. The advantage of this methodology lies in the fact that it does not require the solution of coupled partial differential equations; instead, the stationary phase concentrations were evaluated through unsteady-state component mass balance expressions written in discretization schemes. Thus, the number of partial differential equations to be solved reduces to one.

2.51.3.2.1

Single component solution

The technique presented here is directly applicable to Langmuir-type nonlinear isotherms. For other nonlinear isotherm formulations, a simple routine should be integrated into the algorithm that calculates c*(z,t) value through isotherm expression once qs(z,t) value is known. Eq. (28) gives the Langmuir adsorption isotherm where qs(z,t) and c*(z,t) terms indicate that equilibrium between the mobile phase and the solid phase is valid only at the interphase, which is the fundamental condition of the GR model. The qm and K terms are the Langmuir constants. qs ðz; tÞ ¼

qm c ðz; tÞ K þ c ðz; tÞ

If Eq. (19) is substituted into Eq. (28) and the remaining equation is solved for positive values of c*(z,t), one gets rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi iffi h
M þ M2 þ 4K cðz; tÞþ 5qðz;tÞ Bi c ðz; tÞ ¼ 2

(28)

(29)

where M¼

5qm 5qðz; tÞ þ K
cðz; tÞ
Bi Bi

substitution of Eq. (29) into Eq. (27) gives 2

vcðz; tÞ vcðz; tÞ ð1
εÞ 3kf 4 þv þ cðz; tÞ
vt vz ε rp


M þ

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi iffi3 h M2 þ 4K cðz; tÞþ 5qðz;tÞ 2 Bi 5 ¼ Da v cðz; tÞ vz2 2

(30)

(31)

Comparison of Eq. (27) with Eq. (31) reveals that in the latter equation, c*(z,t) term disappears but a new dependent variable, namely stationary phase average concentration, qðz; tÞ appears. Referring to Fig. 4, mass balance for the control volume of 2Dz height gives stationary phase average concentration in discretization scheme as shown by Eq. (32), where h and k are the distance and time increments, respectively.  

  εk vcðz; tÞ Da ci
1;j
2ci;j þ ciþ1;j v ci
1;j
ciþ1;j
2h þ qi;j ¼ qi;j
1 þ (32) 2ð1
εÞh vt h For a certain distance panel of i, the unknown qðz; tÞ value at j time panel is evaluated from the known qðz; tÞ value at j
1 time panel. Further details are given elsewhere [11]. The so obtained stationary phase average concentrations are employed during the numerical solution of Eq. (31). We have one initial and two boundary conditions.

Modeling Chromatographic Separation

Mobile phase

763

Eluent in

Stationary phase

2Δz

i –1

Δz

i i +1

L

i = nz i = nz +1 Eluent out Fig. 4

Schematic display of a chromatography column for the evaluation of stationary phase average concentrations.

I:C: t ¼ 0 for all z values cðz; tÞ ¼ 0 B:C:1 z ¼ 0 for 0 < t < tinj then cðz; tÞ ¼ cinj ; otherwise cðz; tÞ ¼ 0 B:C:2 z ¼ L vcðz; tÞ=vz ¼ 0 The first boundary condition states that during the sample injection period, tinj mobile phase concentration at column inlet is equal to the sample injection concentration, cinj and becomes zero for the rest of the process. The second boundary condition is defined by the stop of mass transfer at the column outlet. Since Langmuir isotherm parameters are included in Eq. (31), its numerical solution will reflect the effect of adsorption isotherm. Thus, only one partial differential equation, ie, Eq. (31), is left to be taken care of during the calculation of eluent concentration, c(z,t). This alternative methodology for the numerical solution of GR model offers considerable machine time saving against coupled partial differential equations solution procedures, without loss of accuracy in the calculated values.

2.51.3.2.2

Multi component solution

Prediction of multi-component adsorption is one of the most challenging problems in the field of adsorption. Early applications of the Langmuir model to competitive adsorption phenomena were published by Butler and Ockrent [23]. Eq. (33) gives competitive Langmuir nonlinear isotherm expression [14]. As discussed in single component solution section, here again qs(z, t) and c ðz; tÞ terms indicate that equilibrium between the mobile phase and the solid phase is valid only at the interphase, where a and b are solute indexes. qsa ðz; tÞ ¼

a c ðz; tÞ Pan a 1 þ b¼1 bb cb ðz; tÞ

(33)

The methodology of multicomponent solution presented here is well applicable to more than two components. However, for the sake of simplicity, employment of Eq. (33) for two component solutions, namely A and B, will be illustrated here. Average concentration of a solute in an adsorbent particle, qðz; tÞ is proportional with the solid concentration at the liquid–solid interphase, qs(z,t) and its diffusivity in the solid, Ds. If the proportionality constant is designated as z then for components A and B, qA ðz; tÞ ¼ zA DsA qSA ðz; tÞ

(34)

qB ðz; tÞ ¼ zB DsB qSB ðz; tÞ

(35)

As a good approximation we can assume that zAzzB. From Eqs. (34) and (35) qsA ðz; tÞ DsB qA ðz; tÞ z qsB ðz; tÞ DsA qB ðz; tÞ

(36)

Regarding components A and B let’s write Eq. (33) individually qsA ðz; tÞ ¼

aA cA ðz; tÞ  1 þ bA cA ðz; tÞþ bB cB ðz; tÞ

(37)

qsB ðz; tÞ ¼

aB cB ðz; tÞ  bA cA ðz; tÞþ bB cB ðz; tÞ

(38)

1þ

764

Modeling Chromatographic Separation

If we divide Eqs. (37) and (38) side by side and substitute Eq. (36) into the resulting expression, DsB qA ðz; tÞ aA cA ðz; tÞ ¼ DsA qB ðz; tÞ aB cB ðz; tÞ

(39)

or cB ðz; tÞ ¼

aA DsA qB ðz; tÞ  c ðz; tÞ aB DsB qA ðz; tÞ A

(40)

In the literature, several notations are adopted for Langmuir equation parameters such as shown in Eqs. (28) and (33). The relationships between them are as follows: qm ¼ a=b

(41)

K ¼ 1=b

(42)

If we substitute Eq. (40) into Eq. (37) and make use of Eq. (41) we get: qsA ðz; tÞ ¼ where

aA cA ðz; tÞ 1 þ bA hA cA ðz; tÞ

DsA qmA qB ðz; tÞ hA ¼ 1 þ DsB qmB qA ðz; tÞ

(43)

(44)

Similarly qsB ðz; tÞ ¼ where

aB cB ðz; tÞ 1 þ bB hB cB ðz; tÞ

DsB qmB qA ðz; tÞ hB ¼ 1 þ DsA qmA qB ðz; tÞ

(45)

(46)

Now let’s write Eq. (19) for components A and B qsA ðz; tÞ ¼ qA ðz; tÞþ

 BiA  cA ðz; tÞ
cA ðz; tÞ 5

(47)

qsB ðz; tÞ ¼ qB ðz; tÞþ

 BiB  cB ðz; tÞ
cB ðz; tÞ 5

(48)

If Eq. (43) is substituted into Eq. (47), we get a quadratic equation. If it is solved for positive values of cA ðz; tÞ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h i 5qA ðz;tÞ
MA þ M2A þ 4 KhA cA ðz; tÞþ Bi A A cA ðz; tÞ ¼ 2

(49)

where MA ¼

5qmA 5q ðz; tÞ KA
A þ
cA ðz; tÞ BiA BiA hA hA

Similarly cB ðz; tÞ ¼


MB þ

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h iffi 5qB ðz;tÞ M2B þ 4 KhB cB ðz; tÞþ Bi B B

2

(50)

(51)

where MB ¼

5qmB 5q ðz; tÞ KB
B þ
cB ðz; tÞ BiB BiB hB hB

(52)

Now let’s write Eq. (31) for A and B components

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h i 5qA ðz;tÞ 3
MA þ M2A þ 4KA cA ðz; tÞþ Bi 2 A vcA ðz; tÞ vcA ðz; tÞ ð1
εÞ 3kfA 4 5 ¼ DaA v cA ðz; tÞ þv þ cA ðz; tÞ
vt vz ε vz2 2 rp 2

(53)

Modeling Chromatographic Separation

2

vcB ðz; tÞ vcB ðz; tÞ ð1
εÞ 3kfB 4 cB ðz; tÞ
þv þ vt vz ε rp


MB þ

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h iffi 5qB ðz;tÞ 3 M2B þ 4KB cB ðz; tÞþ Bi 2 B 5 ¼ DaB v cB ðz; tÞ vz2 2

765

(54)

Eq. (53) has two independent (t, z) and two dependent [cA(z, t), qA ðz;tÞ] variables as does Eq. (54) with the respective component terms. In order to carry out the simultaneous solution of partial differential equations of (53) and (54) numerically, the technique described in Section 2.51.3.2.1 for single component systems are followed. For the component in question, the unknown qðz; tÞ value at j time panel is evaluated from the known qðz; tÞ value at j
1 time panel. Thus, the number of dependent variables in Eqs. (53) and (54) reduces to one and makes it possible to evaluate time and distance changes of cA(z, t), and cB(z, t). Furthermore, Eqs. (49) and (51) enable the evaluation of time and distance changes of cA ðz; tÞ and cB ðz; tÞ values.

2.51.4

Case Studies

In the previous section, a systematic approach has been described where formulation of the main chromatographic models for separation processes is presented. In order to demonstrate the characteristic features of each model better, the simulation results of each modeling approach are illustrated for different case studies. Nonlinear isotherm is envisaged by the introduction of Langmuir adsorption isotherm expression into formulations. For the IM, ED and TD models, the implicit scheme finite difference technique has been employed to provide a numerical solution. For the GR model, using the same parameters, a linear isotherm system is solved both by the orthogonal collocation on finite elements method and by the alternative method explained in Section 2.51.3.2. The eluent concentration versus time profiles predicted by both methods showed a very close agreement, except that an observable machine time saving is accomplished in using alternative method and therefore it is employed during the numerical solution of GR model with nonlinear isotherms. The eluent concentration at the column outlet is approximated by linear extrapolation of the numerically calculated eluent concentrations at the vicinity of outlet [11]. Simulation results are made for both single and multi-component systems and based on descriptive data that are in proximity with the commonly reported values. For single component studies eluent flow rate kept constant as 1.0 mL min
1 and parameter values are indicated in the corresponding figures. It was assumed that homogeneous diffusion model, ie, Eq. (12), represents the diffusion of solutes in the stationery phase.

2.51.4.1

Case Study 1

The first illustration is a case study of the comparison of IM, ED and GR model simulated chromatograms, of which the same parameters are introduced into all models, where appropriate. Mass transfer parameters are only employed at the GR model, since the remaining models assume that the external and the intraparticle resistance to mass transfer are not taking place. Fig. 5 illustrates 0.08

Eluent concentration (mg ml–1)

Common parameters: qm = 5 mg ml–1solid K = 0.5 mg ml–1 Cinj = 10 mg ml–1 Vinj = 0.01 ml–1 Dc = 0.46 cm L = 20 cm ε = 0.5

Applicable to GR model only: Dp = 2 × 10–3 cm kf = 1 × 10–3 cm s–1 Ds = 5 × 10–9 cm2 s–1

0.07 0.06

Ideal model 0.05 0.04 0.03

ED model Da = 1 × 10–2 cm2 s–1

0.02

ED model Da = 5 × 10–2 cm2 s–1

GR model Da = 1 × 10–2 cm2 s–1

0.01 0

0

0.1

0.2

0.3

0.4

0.5

Time (h) Fig. 5 The effect of axial dispersion coefficient on the ED and the GR model generated chromatograms and their comparison with the IM generated chromatogram.

766

Modeling Chromatographic Separation

the effect of axial dispersion coefficient on the simulation results. Since IM neglects the mass transfer resistances as well as the axial dispersion, a very narrow band profile is obtained accordingly. The comparison of the ED and GR model-generated chromatograms clearly indicates that for the same Da value, the band broadening in the latter is more pronounced as the theory predicts. Fig. 5 shows that when higher Da values are employed in the ED model, the corresponding peaks flatten and eventually ED and GR model-generated chromatograms might nearly fit. This phenomenon is exactly what is intended to be achieved by the ED model, ie, by lumping all the mass transfer resistances into an artificially large axial dispersion term and thereby mimicking the real processes, but without including the proper mass transfer resistance terms in the model equation.

2.51.4.2

Case Study 2

The second case study considers the possibility of fitting the ED model-generated chromatogram to the GR model-generated chromatogram when mass transfer resistances are more evident than that of case study 1. IM is a special case of ED model where axial dispersion coefficient Da is zero; therefore, its behavior shall no longer be analyzed in the case studies since understanding the ED model for different Da values would give insight to IM. Fig. 6(A) and (B) illustrates the change of model-generated chromatograms with Da for small and large diameter particles, respectively, packed into a chromatography column. In the GR model simulated profiles, particle size plays an important role in band broadening, as particle internal resistance to diffusion of solute will become more evident in larger particles. In Fig. 6(B) the particle diameter is convincingly larger (6-fold) than that of Fig. 6(A). Therefore, the corresponding GR model-generated peak of the former is considerably flatter than the latter. It might also be worth noticing that the effect of the change of Da on the GR model-generated peak profile is small, since band broadening essentially originates from masstransfer resistances. This will be explained in detail in the coming case studies. Fig. 6 clearly illustrates that for the case of the ED model-generated peaks, the effect of Da on the model-generated peak profile is sizeable. Since mass transfer resistances are omitted in the ED model, the dominant factor in peak flattening is the axial dispersion coefficient. Fig. 6 also indicates that for larger particles, it is rather difficult to fit the ED model-generated peak to the GR model-generated peak. This is a clear indication of the inapplicability of the ED model to chromatographic separation processes where mass transfer resistances are noticeably high.

2.51.4.3

Case Study 3

The third case study considers Langmuir isotherm parameter, K, and its effect on the ED and GR model-generated chromatograms. In this simulation study, all the other parameters are kept constant so as to understand the sole effect of K value. Keeping qm constant, Langmuir isotherm becomes more favorable as the K value decreased. The reflection of this behavior into chromatograms is the observation of peak maximum at extended times, as the K value is lessened. This can be attributed to Langmuir adsorption isotherm equation that predicts higher stationary phase concentration for the same eluent phase concentration where qm is kept constant but K value is decreased. Therefore, the elution of solute from the stationery phase will be delayed. Fig. 7 illustrates this phenomenon for both ED and GR model-generated chromatograms. Since the mass-transfer resistances are included as separate identities in the GR model, the resulting peak band broadening is much larger. It might be interesting to note that the peak profiles of the two models are different in nature, and this becomes more noticeable for the small K value generated profiles. The GR modelgenerated peaks exhibit tailing, but the ED model-generated peaks exhibit fronting for the same data set used in the models, where 0.08

B 0.08 Applicable to GR model only: Dp = 1 × 10–3 cm kf = 1 × 10–3 cm/s Ds = 5 × 10–9 cm2/s

Eluent concentration (mg ml–1)

0.07 0.06

ED model Da = 1 × 10–3 cm2 /s

0.05 0.04

Common parameters: qm = 5 mg/mLsolid K = 0.5 mg/mL Cinj = 10 mg/mL Vinj = 0.01 mL Dc = 0.46 cm L = 20 cm ε = 0.5 ED model Da = 1 × 10–2 cm2 s–1

0.03 0.02

GR model Da = 1 × 10–2 cm2 s–1 (upper line)

GR model Da = 1 × 10–3 cm2 s–1 (lower line)

0.01

Eluent concentration (mg mL)

A

0.07

Applicable to GR model only: Dp = 6 × 10–3 cm

0.06

Ds = 5 × 10–9 cm2 s–1

kf = 1 × 10–3 cm s–1

Da = 1 × 10–3 cm2 s–1 0.04 0.03

0

0.05

0.1

0.15

0.2

0.25 Time (h)

0.3

0.35

0.4

0.45

0.5

ED model Da = 1 × 10–2 cm2 s–1

0.02

GR model Da = 1 × 10–3 cm2 s–1 (lower line)

0.01 0

0

ED model

0.05

Common parameters: qm = 5 mg ml–1solid K = 0.5 mg ml–1 Cinj = 10 mg ml–1 Vinj = 0.01 ml Dc = 0.46 cm L = 20 cm ε = 0.5

0

0.05

0.1

0.15

0.2

0.25

GR model Da = 1 × 10–2 cm2 s–1

(upper line)

0.3

0.35

0.4

0.45

0.5

Time (h)

Fig. 6 The GR model point out the effect of particle size on the model-generated chromatograms, whereas ED model is not proficient to show such effects. (A) Small particle diameter where diffusion path is short. (B) Large diameter particle where diffusion path is comparatively long.

Modeling Chromatographic Separation

767

0.12

0.10 Eluent concentration (mg ml–1)

Common parameters: qm = 5 mg ml–1solid Cinj = 10 mg ml–1 Vinj = 0.01 ml–1 Dc = 0.46 cm L = 20 cm ε = 0.5 Da = 1 × 10–3 cm2 s–1

Applicable to GR model only: Dp = 2 × 10–3 cm kf = 1 × 10–3 cm s–1 Ds = 5 × 10–9 cm2 s–1

0.08

0.06 ED model k = 0.5 mg ml–1 0.04 GR model K = 0.5 mg ml–1

0.02

0

0

0.2

ED model GR model –1 K = 0.1 mg ml–1 K = 0.1 mg ml

0.4

0.6

0.8

1

Time (h) Fig. 7

The effect of Langmuir adsorption isotherm parameter, K, on the ED and GR model generated chromatograms.

appropriate. In the former case, the peak tail, appearing to the right on the chromatogram, can be attributed to the slower solute elution rates of the GR model, where mass transfer resistances hinder the rapid desorption of the solute in the stationary phase. With fronting, the reverse is the case. Furthermore, case study 6 reveals that the effect of Da on the GR model-generated peak band broadening is small while it is the main band-broadening parameter for the ED model. Since dispersion is only effective for solute molecules still present in the eluent phase, ie, not for those in the adsorbed phase, with the ED model it is possible to generate chromatograms different in nature than that of the GR model.

2.51.4.4

Case Study 4

The present case study illustrates the effect of mass transfer resistance parameters on model-generated peaks, where only the GR model accounts for external and intraparticle mass transfer resistances. Fig. 8 illustrates the effect of film mass transfer coefficient, kf, where, as the chromatography theory predicts, the band broadening of the simulated chromatogram decreases with increasing kf

0.03 Common parameters: qm = 5 mg ml–1solid K = 0.5 mg ml–1 Cinj = 10 mg ml–1 Vinj = 0.01 ml–1 Dp = 2 × 10–3 cm Dc = 0.46 cm L = 20 cm ε = 0.5 Da = 1 × 10–2 cm2 s–1 Ds = 1 × 10–7 cm2 s–1

Eluent concentration (mg ml–1)

GR model kf = 5 × 10–3 cm s–1 0.02 kf = 1 × 10–3 cm s–1

0.01 kf = 5 × 10–4 cm s–1

0

0

0.1

0.2

0.3 Time (h)

Fig. 8

The effect of film mass transfer coefficient, kf, on the GR model generated chromatograms.

0.4

0.5

768

Modeling Chromatographic Separation

0.03 Common parameters: qm = 5 mg ml–1 solid K = 0.5 mg ml–1 Cinj = 10 mg ml–1 Vinj = 0.01 ml DP = 2⫻10–3 cm Dc = 0.46 cm L = 20 cm ε = 0.5 Da = 1⫻10–3 cm2 s–1 kf = 1⫻10–3 cm s–1

Eluent concentration (mg cm–3)

GR model

0.02 Ds = 1⫻10–7 cm2 s–1

0.01

Ds = 1⫻10–8 cm2 s–1 Ds = 1⫻10–9 cm2 s–1

0

0

0.1

0.2

0.3

0.4

0.5

Time (h) Fig. 9

The effect of intraparticle diffusivity on the GR model-generated chromatograms.

values as long as other parameters are kept constant. Fig. 9 illustrates the effect of internal mass transfer resistance where homogeneous diffusivity, Ds, is changed in order to observe its effect on the GR model-generated chromatograms. It is apparent that the band broadening of the simulated chromatogram decreases with the increase in Ds. The behavior of the GR model-generated chromatograms clearly indicates the importance of mass transfer resistances in the modeling of chromatographic separation.

2.51.4.5

Case Study 5

The fifth case study considers the TD model-generated chromatograms and their agreement with the GR model-generated chromatograms by trying various km values in the TD model. For this purpose, first the effect of sample injection concentration is investigated. Fig. 10 illustrates the GR model-predicted profiles for two different injection concentrations having a five-fold difference. Employing the same parameter values of the GR model that are applicable to the TD model, the agreement between the two models is searched. As long as the particle diameter is kept constant, the km value is very slightly concentration dependent. As indicated by Fig. 10, it is possible to obtain a reasonable agreement between the two model predictions by employment of the same km value in 0.06 Applicable to GR model: DP = 2⫻10–3 cm kf = 1⫻10–3 cm s–1 Ds = 5⫻10–9 cm2 s–1

Eluent concentration (mg ml–1)

0.05

Common parameters: qm = 5 mg ml–1solid K = 0.5 mg ml– Vinj = 0.01 ml Dc = 0.46 cm L = 20 cm ε = 0.5 Da = 1⫻10–2 cm2 s–1

Applicable to TD model: km = 8⫻10–2 s

0.04

0.03

0.02

TD model Cinj = 50 mg ml–1

TD model Cinj = 10 mg ml–1

0.01

0

GR model Cinj = 50 mg ml–1

GR model Cinj = 10 mg ml–1

0

0.1

0.2

0.3

0.4

0.5

Time (h) Fig. 10

The effect of sample injection concentration for the calculation of km by comparison of TD and the GR model-generated chromatograms.

Modeling Chromatographic Separation

Eluent concentration (mg ml–1)

0.025

Common parameters: qm = 5 mg ml–1solid Cinj = 10 mg ml–1 Vinj = 0.01 ml Dc = 0.46 cm L = 20 cm ε = 0.5 Da = 1⫻10–2 cm2 s–1

Applicable to GR model: kf = 1⫻10–3 cm s–1 Ds = 5⫻10–9 cm2 s–1

0.020

0.015

TD model (lower curve) km = 2.5⫻10–1 s

TD model (lower curve) km = 1.0⫻10–2 s

0.005

0

GR model (upper curve) DP = 1⫻10–3 cm

GR model (upper curve) DP = 6⫻10–3 cm

0.010

0

0.1

769

0.2

0.3

0.4

0.5

Time (h) Fig. 11

The effect of particle diameter for the calculation of km by comparison of TD and the GR model-generated chromatograms.

the TD model although the km value does not reflect any physical meaning that is usually attributed to it. This can easily be illustrated if we examine the GR and the TD models generated profiles for different particle diameters, for the same injection concentrations. Fig. 11 illustrates that there is an extreme difference between the km values used in the TD model in order to obtain a satisfactory fit with the GR model-generated profiles. So the TD model, being too simple for the complex phenomena of chromatographic separation, cannot properly take the mass transfer resistances into account.

2.51.4.6

Case Study 6

The sixth case study considers scenarios of multi-component elution bands generated via GR model. For multi-component studies, pivot parameters for the two-component system under investigation is shown in Table 1. The methodology of the GR model-predicted eluent concentrations at the chromatography column outlet is explained in Section 2.51.3.2.2, where a competitive multi-component Langmuir isotherm is employed. The Langmuir parameters in Table 1 indicate that component B is more favorably adsorbed than component A. Fig. 12 shows the significance of particle size on elution bands. It is worth mentioning that the sole role of particle size cannot be demonstrated by other chromatography models except GR model. Fig. 12 clearly illustrates that for large particle size and for unfavorable adsorption, such as component A with a particle diameter of 30 mm, it is likely to encounter overloaded peaks. This situation is represented by the first peak at the dotted line of component A, which corresponds to an overloading peak of A. However, for component B, which is favorably adsorbed, there is no overloading peak, even at same particle size (ie, 30 mm) if all other Table 1

Pivot parameters for multi-component elution bands

Component A (less favorably adsorbed)

Component B (more favorably adsorbed)

qmA ¼ aA =bA ¼ 15 mg=cm3solid

qmB ¼ aB =bB ¼ 15 mg=cm3solid 3

KA ¼ 1=bA ¼ 2 mg=cm Film mass transfer coefficients Homogeneous solid diffusivities Axial dispersion coefficients Sample concentration Injection volume Eluent flow rate Stationary phase particle diameter Eluent flow rate Stationary phase particle diameter Column inside diameter Bed height Bed voidage

KB ¼ 1=bB ¼ 1 mg=cm3 kfA¼kfB¼510–5 cm/s DsA¼DsB¼2.510–9 cm2/s DaA¼DaB¼1.010–3 cm2/s A¼B¼30 mg/cm3 VinjA¼VinjB¼30 mL F¼1.2 mL/min rpA¼rpB¼20 mm F¼1.2 mL/min rpA¼rpB¼20 mm 0.46 cm 15 cm 0.45 (-)

770

Modeling Chromatographic Separation

Fig. 12

Multi-component GR model-generated elution bands on the course of elution: the effect of particle size.

Fig. 13

Multi-component GR model-generated elution bands on the course of elution: the effect of homogeneous solid diffusivity.

parameters are the same. This might be attributed to the adsorption capacity in the case of favorable adsorption. When the particle diameter is reduced to 20 mm, then the overloading peak of A disappears since the diffusion path in the particle is reduced. Fig. 13 shows the significance of solid diffusion rate on elution bands for a multi-component system. This figure represents that a 10-fold change in the particle diffusivity results in different trends on the peak maximum times of the components, ie, how fast the solute leaves the column. For the less favorably adsorbed component A, a decrease in the particle internal mass transfer rate (ie, a decrease in the diffusivity) results in a slightly earlier peak maximum time, since characteristics of both the adsorption isotherm and the particle internal mass transfer do not augment the retention of solute. However, if the particle internal mass transfer increases, this will cause a better chance in regard of the retention of less favorably adsorbed component A, and thus a later peak maximum time develops. Contrary to this explanation, it is possible to argue that an increase in the particle internal mass transfer rate would boost desorption rate and would result earlier peak maximum time. But in the case of less favorably adsorbed components, such as A, the driving force for desorption, ie, solid internal and solid interphase concentration difference, would be small. Consequently, it all depends on the balance between competitive adsorption isotherm effects and desorption mass transfer rate. Fig. 13 indicates that under the specific case presented here, the odds are on the side of retention. For the more favorably adsorbed component B, Fig. 13 clearly indicates that the desorption and retention due to isotherm characteristics yield a completely different portrait. The peak

Modeling Chromatographic Separation

771

maximum time decreases with an increase in particle internal mass transfer rate. This can be attributed to the enhancement of desorption rate both due to higher particle concentration depicted from the adsorption isotherm, ie, increase in the driving force, as well as the increase of desorption rate. As a consequence, GR modeling studies disclose several peculiar features of chromatograms encountered in chromatographic separations both in laboratory and industrial scale.

2.51.5

Summary

Chromatographic separation processes are now better understood, as demonstrated by their successful modeling under a variety of experimental conditions, covering a great segment of the applications. It appears that different models of chromatography might serve for different purposes. The ideal model (IM) is easy to understand and requires the minimum knowledge of algebra and calculus. It is a useful tool for the analysis of the qualitative behavior of the process and the influence of some of the process parameters. But it is no secret that before the introduction of more elaborate models, the design and scaling-up of preparative chromatography columns were largely based on experience. For some applications, mainly in continuous chromatography, this might still be true. But generally speaking, quantitative description of a chromatography columns through modeling studies is now possible. Both the equilibrium dispersive (ED) and the transport dispersive (TD) models are used to describe and understand preparative chromatography, where mass transfer resistances in the chromatographic separation unit is negligible such as separation of small molecules through a column filled with highly porous small-sized particles. In these models, the required numbers of parameters are limited. However, the involved lumped dispersion and mass transfer coefficients do not have an actual physical meaning, and in some cases, concentration-dependent parameters are used in the peak fitting procedure. For the preparative separations of several large molecular compounds such as proteins, the diffusion of solute in the stationary phase is slow. In modeling of such chromatographic separation processes, use of accurate mathematical models is required, where mass transfer resistances are involved. Therefore, instead of using the ED or TD models, sophisticated models such as the general rate model (GR) are used in order to account for the contributions of mass transfer resistances. The solution of the GR model equations is usually carried out by the employment of advanced numerical techniques that requires considerable machine time. This explains why the GR model is not as popular as the ED or the TD models, even though the personal computers of today significantly reduced the computation time. On the other hand, there is renewed interest in large-scale chromatographic separation. This interest is largely because of increased need in the pharmaceutical and biotechnological industries. Scaling up of a preparative separation will inevitably bring wider recognition to the GR model since in these industries, the challenge generally lies in the separation of biomolecules where mass transfer resistances are usually high.

References [1] LeVan, M.D., Carta, C., 2008. Adsorption and ion exchange. In: Green, D.W., Perry, R.H. (Eds.), Perry’s Chemical Engineers’ Handbook, eighth ed. McGrawHill, New York, pp. 16.1–16.69. [2] Sonoda, A., Makita, Y., Hirotsu, T., 2006. Boron isotope fractionation in column chromatography with glucamine type resins. Journal of Nuclear Science and Technology 43, 437–440. [3] Rodrigues, A.E., 1997. Review – Permeable packings and perfusion chromatography in protein separation. Journal of Chromatography B 699, 47–61. [4] Broughton, D., 1966. Continuous Simulated Counter-Current Sorption Process Employing Desorbent Made in Said Process. U.S. Patent 3,291,726. [5] Juza, M., Mazzotti, M., Morbidelli, M., 2000. Simulated moving bed chromatography and its application to chirotechnology. Trends in Biotechnology 18, 108–118. [6] Borren, T., Fricke, J., Schmidt-Traub, H., 2006. Reactive liquid chromatography. In: Schmidt-Traub, H., Górak, A. (Eds.), Integrated Reaction and Separation Operations. Modelling and Experimental Validation. Springer Verlag, Berlin, pp. 191–240. [7] Uretschlager, A., Jungbauer, A., 2002. Preparative continuous annular chromatography (P-CAC), a review. Bioprocess and Biosystems Engineering 25, 129–140. [8] Wiesel, A., Schmidt-Traub, H., Lenz, J., Strube, J., 2003. Modelling gradient elution of bioactive multicomponent systems in non-linear ion-exchange chromatography. Journal of Chromatography A 1006, 101–120. [9] Martin, A.J.P., Synge, R.L.M., 1941. A new form of chromatogram employing two liquid phases. Biochemical Journal 35, 1358–1368. [10] Skoog, D.A., West, D.M., Holler, F.J., Crouch, S.R., 2003. Fundamentals of Analytical Chemistry, eighth ed. Thomson-Brooks Cole, Belmont. [11] Özdural, A.R., Alkan, A., Kerkhof, P.A.J.M., 2004. Modeling chromatographic columns: Non-equilibrium packed-bed adsorption with non-linear adsorption isotherms. Journal of Chromatography A 1041, 77–85. [12] Gu, T., 2015. Mathematical Modeling and Scale-Up of Liquid Chromatography: With Application Examples, second ed. Springer Verlag, New York. [13] Guiochon, G., Lin, B., 2003. Modeling for Preparative Chromatography. Academic Press, San Diego. [14] Guiochon, G., Felinger, A., Shirazi, D.G.G., Katti, A.M., 2006. Fundamentals of Preparative and Nonlinear Chromatography, second ed. Elsevier, Amsterdam. [15] Felinger, A., 2002. Optimization of preparative separations. In: Rathore, A.S., Velayudhan, A. (Eds.), Scale-Up and Optimization in Preparative Chromatography. Marcel Dekker, New York, pp. 77–121. [16] Qamar, S., Sattar, F.A., Abbasi, J.N., Seidel-Morgenstern, A., 2016. Numerical simulation of nonlinear chromatography with core-shell particles applying the general rate model. Chemical Engineering Science 147, 54–64. [17] Horváth, K., Felinger, A., 2015. Influence of particle size and shell thickness of core-shell packing materials on optimum experimental conditions in preparative chromatography. Journal of Chromatography A 1407, 100–105. [18] Close, E.J., Salm, J.R., Bracewell, D.G., Sorensen, E., 2014. A model based approach for identifying robust operating conditions for industrial chromatography with process variability. Chemical Engineering Science 116, 284–295. [19] Coquebert de Neuville, B., Tarafder, A., Morbidelli, M., 2013. Distributed pore model for bio-molecule chromatography. Journal of Chromatography A 1298, 26–34. [20] Glueckauf, E., Coates, J.I., 1947. Theory of chromatography. Part IV. The influence of incomplete equilibrium on the front boundary of chromatograms and on the effectiveness of separation. Journal of the Chemical Society 1315–1321.

772

Modeling Chromatographic Separation

[21] Freitag, R., 2007. Displacement chromatography of biomacromolecules. In: Subramanian, G. (Ed.), Bioseparation and Bioprocessing, second ed., vol. 1. Wiley-VCH, Weinheim, pp. 151–204. [22] Seidel-Morgenstern, A., Guiochon, G., 1993. Modelling of the competitive isotherms and the chromatographic separation of two enantiomers. Chemical Engineering Science 48, 2787–2797. [23] Butler, J.A.V., Ockrent, C., 1930. Studies in electrocapillarity. Part III. The surface tension of solutions containing two surface-active solutes. The Journal of Physical Chemistry 34, 2841–2859.

2.52

Aqueous Two-Phase Systems

J Benavides and M Rito-Palomares, Tecnológico de Monterrey, Monterrey, Mexico JA Asenjo, University of Chile, Santiago, Chile © 2011 Elsevier B.V. All rights reserved. This is a reprint of J. Benavides, M. Rito-Palomares, J.A. Asenjo, 2.49 - Aqueous Two-Phase Systems, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 697-713.

2.52.1 2.52.2 2.52.2.1 2.52.2.2 2.52.2.2.1 2.52.2.2.2 2.52.2.2.3 2.52.2.2.4 2.52.2.3 2.52.2.3.1 2.52.2.3.2 2.52.2.3.3 2.52.2.3.4 2.52.3 2.52.3.1 2.52.3.2 2.52.3.3 2.52.3.4 2.52.3.4.1 2.52.3.4.2 2.52.3.4.3 2.52.3.4.4 2.52.3.4.5 2.52.3.5 2.52.4 References

Introduction Theoretical Background Phase Formation, Binodal Curves, Tie-Line Length, Volume Ratio, and Phase Separation Types of ATPSs and Their Typical Applications Polymer–Polymer and Polymer–Salt ATPSs Alcohol–Salt ATPS Micellar and Reverse Micelle ATPSs Ionic Liquid-Based ATPS Parameters That Influence Partitioning in ATPSs Size-Dependent Partition Electrochemical-Dependent Partition Hydrophobic-Dependent Partition Specific Affinity-Dependent Partition Application of ATPSs for the Recovery of Biological Products Multistage Extraction and Continuous Separation Using ATPSs Process Integration Using ATPSs Bioaffinity-Enhanced Partitioning on ATPSs Products of Interest Partitioned Using ATPS Proteins Nucleic Acids Virus, Virus-like Particles, and Other Bionanoparticles Viable Cells and Organelles Low-Molecular-Weight Compounds Other Applications of ATPSs Conclusions

773 774 774 777 777 777 778 778 778 779 779 779 780 780 780 781 782 783 783 786 786 787 787 787 788 788

Glossary Binodal curve Curve that denotes the condition at which two distinct phases may coexist in equilibrium. Downstream processing Sequence of unit operations addressing the separation, recovery, and purification of molecules. Micelle Loosely bound aggregation of molecules (such as surfactants and other amphipathic compounds) forming a colloidal particle. Tie line Line on a phase diagram that joins the two points representing the composition of phases of liquid–liquid system in equilibrium. Volume ratio Relation between the volume of the top phase and the bottom phase of a liquid–liquid system.

2.52.1

Introduction

Over the past 30 years, the developments in the production of biological products have had a great impact in the biotechnology industry. The genetic modification of expression systems, both prokaryotic and eukaryotic, and the development of novel bioreactors and culture media for the optimization of the large-scale production of biologicals are examples of the extensive work currently being conducted. In this context, biomass and product yields have increased significantly, particularly for microorganisms well characterized for large-scale production of heterologous proteins. Such achievements have resulted in process effluents with high concentration of both solutes and biomass. Consequently, the development of efficient new technologies and strategies for the recovery and purification of biomolecules from these highly concentrated effluents is one of the major concerns in bioprocess engineering.11,16

Comprehensive Biotechnology, 3rd edition, Volume 2

https://doi.org/10.1016/B978-0-444-64046-8.00100-2

773

774

Aqueous Two-Phase Systems

The methods traditionally used in the biotechnology industry for the recovery, separation, and purification of biomolecules include membrane-based technology (microfiltration, ultrafiltration, and nanofiltration), precipitation (salting out, isoelectric, and affinity), and liquid chromatography (ion exchange, reverse phase, hydrophobic interaction, size exclusion, and affinity, among others). The use of organic-aqueous systems formed when an organic solvent and an aqueous solution are mixed is generally not suitable for the separation and recovery of biological macromolecules due to the low solubility of proteins in such systems. In this context, aqueous two-phase system (ATPS) as a downstream process technology has demonstrated to be highly suitable, particularly for process integration and for continuous operation.33,44,45 ATPS is a liquid–liquid separation technology that has been used to develop bioprocesses for the primary recovery and partial purification of a variety of biological products such as proteins, genetic material, bionanoparticles, and phytochemicals, as well as cells and cell organelles.4,11,42,54 ATPSs form when polymers (polyethylene glycol, dextran, and polypropylene glycol), salts (phosphates and sulfates), lowmolecular-weight alcohols (ethanol and propanol), surfactants (n-decyl tetraethylene oxide and octylphenol ethoxylate), and/or ionic liquids (1-butyl-3-methylimidazolium hexafluorophosphate and 1-ethyl-3-methylimidazolium acetate) are combined over critical concentrations, resulting in the formation of two phases. Both phases are in principle hydrophilic, although usually the top phase tends to be somewhat more hydrophobic. The main advantages of this technique include scale-up potential, process integration capability, use of low-toxicity-forming chemicals, and biocompatibility. Process integration results when one single unit operation achieves the process objective of two or more discrete stages, reducing the number of steps needed. Continuous operation is clearly a main potential feature of ATPS.17,44,45 For continuous operation, a critical element is which phase will be the continuous and which the dispersed one will be,13,56 since this can result in very different settling rates and hence processing times. The first studies involving ATPS date from the late 1950s and early 1960s when Albertsson demonstrated the application of this technique for the downstream processing of particles and macromolecules.2 Since then, extensive research regarding the use of ATPS-based strategies for the recovery and purification of biomolecules has been carried out. These studies may be divided into three main areas: (1) characterization of the partition behavior of biologicals, (2) development of new or modified liquid–liquid systems to increase separation selectivity, and (3) development of strategies and approaches for the practical application of ATPS-based technology at large scale. Additionally, ATPSs have demonstrated to have analytical applications for the physicochemical characterization (particularly surface-related properties such as hydrophobicity and electrochemical charge) of compounds, bionanoparticles, and cells. The present article is an overview of the theoretical basis of ATPS-based techniques as well as their application in the separation, recovery, and purification of biological products.

2.52.2

Theoretical Background

2.52.2.1

Phase Formation, Binodal Curves, Tie-Line Length, Volume Ratio, and Phase Separation

The mechanisms related to the separation of two aqueous phases as well as the partition behavior of solutes and particles between such phases are dominated by the thermodynamic equilibrium of the system. As already mentioned, there are several types of ATPSs based on the chemical nature of their two main constituents (phase-forming chemicals).3,31,38 The specific chemical interactions involved in phase formation in ATPS are highly dependent on the type of ATPS.18 Even though the mechanisms involved in the formation and separation of the aqueous phases are extremely complex, they may be simplified based on the hydration enthalpy and the entropy net balance. Although the two main constituents of the ATPS are primordially hydrophilic, the enthalpy of hydration between such components differs. As a result, two thermodynamic scenarios are possible. If the amount of energy in the system is high enough to overcome the net difference between entropy and hydration enthalpy, the two main chemical constituents may coexist at their present concentration in a single homogeneous phase. Otherwise, the separation of the two constituents is energetically favored, promoting the formation of two phases. Since the thermodynamic equilibrium is specific for each particular system at defined conditions (temperature, pH, and pressure), the information of the equilibrium curve, typically known as binodal curve, is critical when working with ATPS. In a phase diagram, the binodal curve represents the concentration boundary separating the monophasic from the biphasic region in an appropriate phase diagram.64 A composition (% w/w) over the binodal curve (Figure 1, line A–B) should be selected in order to form a biphasic aqueous system. Commonly, in the biphasic diagram, the axis of the ordinate is used for the top-phase-rich constituent, while the axis of the abscissa is used for bottom-phase-rich constituent. Additionally, the tie line (TL; which depicts the thermodynamic equilibrium between both phases; Figure 1, line C–D) allows determining the composition of the coexisting phases and the specific tie-line length (TLL) value for a particular system. The TLL is calculated as qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (1) TLL ¼ DC21 þ DC22 where DC1 and DC2 are the absolute difference in concentration of the phase-forming constituents 1 and 2, respectively, in the top and bottom phases. Although all ATPSs within the same TL have identical top- and bottom-phase compositions, their volume ratio (VR) changes according to the global composition of the system (Figure 1). Therefore, although systems 1, 2, and 3 depicted in Figure 1 have the same top- and bottom-phase compositions, determined by points C and D, respectively, their VR changes depending on the global composition of the system. The VR is calculated as VR ¼

VTP VBP

(2)

Aqueous Two-Phase Systems

775

A 1

C Top-phase constituent (% w/w)

VR > 1 2

VR = 1

3

Biphasic region

VR < 1

Monophasic region

D B Bottom-phase constituent (% w/w) Figure 1 Schematic representation of a binodal curve diagram. Concentrations above the binodal curve (line A–B) render aqueous biphasic systems that are characterized by parameters such as tie-line (line C–D) length and the volume ratio (VR).

where VTP and VBP are the volumes of the top and bottom phases, respectively. Many of the mechanisms that influence the partition behavior of solutes and bionanoparticles in ATPS are directly related to the TLL and the VR. Therefore, the proper characterization of these two system parameters is desirable in order to quantify their effect on the fractionation of biological compounds. An important element is which phase will be continuous and which the dispersed one. Settling rates have been studied in polyethylene glycol (PEG)/salt ATPS.13,56 Given the high viscosity, rates are much smaller when the PEG top phase is the continuous phase. Phase-separation times for PEG-4000-phosphate ATPS have been studied35 for small-scale (5-g) and larger-scale (1300-g) systems. Profiles of dispersion height for both larger- and small-scale systems were represented as a fraction of the initial height and were found to be independent of the geometrical dimensions of the separator. Furthermore, by plotting time as a fraction of the initial height, the total time of separation can be calculated for a given height of system in a particular system, as shown in Figure 2(A), for systems of two different sizes. This generalization is important for the design of large-scale aqueous two-phase separators. Phase-separation times were also found to be dependent on which of the phases is continuous. A characteristic change in phase-separation time was also observed at the phase-inversion point (i.e., where the dispersed phase changes to a continuous phase and vice versa), and this point tends toward higher volume ratios as the TLL is increased. Furthermore, the phase-inversion point at each TLL corresponds to a fixed phosphate concentration for this system, as shown in Figure 2(B). For the determination of the TLs, a simple gravimetric method can be used,43 which makes any chemical analysis unnecessary. The method is as follows. A point is selected in the phase plane, through which the TL will pass. The components required are then weighed and mixed thoroughly. The mixture is left at 20  C in a decanter overnight, and then the two fractions are weighed. Calling M, T, and B the points representing the mixture, the top phase, and the bottom phase, respectively, X the weight fraction of phosphate and Y the weight fraction of PEG the following mass balance equations can be written as YT ¼ ðYM =aÞ  ðð1  aÞ=aÞYB

(3)

XT ¼ ðXM =aÞ  ðð1  aÞ=aÞXB

(4)

YT ¼ f ðXT Þ

(5)

YB ¼ f ðXB Þ

(6)

where f(X) is the function representing the binodal giving Y as a function of X, and a is the following measured ratio: a ¼ ðweight of the top phaseÞ=ðweight of the mixtureÞ

(7)

776

Aqueous Two-Phase Systems

A

1.0

Ls-pure Ss-pure

0.8

H*

0.6

0.4

0.2

0.0 0.0

0.2

0.4

0.6

0.8

1.0

t /ts B

50 Continuous top phase

PEG 4000% (w/w)

40

30 Continuous bottom phase 20

10

0 0

5

10 15 Phosphate% (w/w)

20

25

Figure 2 (A) Change in the relative separation time (t/ts) with relative height of dispersion (H* ¼ H/Ho) for 5-g (Ss, small scale; open symbols) and 1300-g (Ls, large scale; closed symbols) systems. Third-order polynomial fit with R ¼ 0.95. (B) Phase diagram PEG-4000-phosphate at pH 7, phaseinversion point (-). To the left of the vertical line, systems have a continuous top phase and to the right a continuous bottom phase.

This is a system of four equations with four unknowns that can readily be solved, and the equation for the TLs can be obtained: Y ¼ Y0 þ SX

(8)

S ¼ ðYT  YB Þ=ðXT  XB Þ

(9)

where the slope S is given by and Yo is the point corresponding to X ¼ 0. Furthermore, having obtained several TLs, an expression of the slope of the family of TLs, g(XB) can be obtained in terms of XB. The point where the derivative of the binodal equals this slope can be obtained now from gðXÞ ¼ df ðXÞ=dðXÞ

(10)

The value of phosphate concentration XC that fulfills this equation corresponds to the critical point, which is obtained therefore without any other experiments.

Aqueous Two-Phase Systems 2.52.2.2

777

Types of ATPSs and Their Typical Applications

ATPSs can be categorized into five main groups: polymer–polymer, polymer–salt, alcohol–salt, micellar systems, and ionic liquidbased systems. Although all of them are used for the fractionation of biomolecules, bionanoparticles, and/or biomass (cells and organelles), there are preferred uses for each type of system. Table 1 shows the types of ATPSs and their typical downstream processing applications as well as alternative applications.

2.52.2.2.1

Polymer–Polymer and Polymer–Salt ATPSs

Polymer–polymer and polymer–salt systems are the most characterized ATPSs, as they have been studied and used for more than 50 years.2,3 These systems are formed when two polymers (PEG, dextran, and polypropylene glycol) or one polymer and one salt (phosphate, sulfate, and citrate) are combined over critical concentrations forming two phases whose main component is water. The first studies involving ATPS date from the late 1950s and early 1960s when Albertsson demonstrated the great potential of polymer–polymer systems for the primary recovery of biological macromolecules. Given that the ionic strength on the polymer–polymer systems is extremely low, they are used preferentially for the separation, recovery, and purification of solutes extremely sensitive to ionic environments as well as viable cells and organelles susceptible to osmotic shock. Besides the traditional polymer–polymer systems, there are ATPSs that exploit the use of copolymer of ethylene oxide and propylene oxide that have proved to be efficient for the development of fractionation processes. Polymer–salt ATPSs are used for a great variety of biomolecules, mainly proteins, and particles, their only inconvenience being the relatively high ionic strength in the salt-rich phase.

2.52.2.2.2

Alcohol–Salt ATPS

Alcohol–salt ATPSs are formed when low-molecular-weight aliphatic alcohols (such as ethanol, propanol, isopropanol, and butanol) and a highly concentrated salt (phosphate, sulfate, and citrate) solution are mixed over certain concentrations yielding Table 1

Types of aqueous two-phase systems and their typical applications in downstream processing and analytical methodologies (common constituent (phase-forming chemicals) mixtures are presented)

Type of ATPS

Constituent 1

Constituent 2

Typical applications

Polymer–polymer

Polyethylene glycol

Dextran

Polymer–salt

Polyethylene glycol Polyvinilpyrrolidone Ficoll Polyethylene glycol

Poly(vinyl methyl ethyl ether) Dextran Dextran (NH4)2SO4

Fractionation, recovery, and purification of proteins, nucleic acids, bionanoparticles, cells, and organelles. Characterization of protein, bionanoparticles, and cell surface physicochemical properties.

Alcohol–salt

Polyethylene glycol Polyvinylpyrrolidone Polyethylene glycol Ethanol

Na2SO4 K2HPO4/KH2PO4 Na2CO3 citrate K2HPO4/KH2PO4

Ethanol 1-Propanol 2-Methyl-2-propanol 1-Butyl-3-methylimidazolium chloride

Na2SO4 K2HPO4/KH2PO4 K2HPO4/KH2PO4 K2HPO4/KH2PO4

1-Butyl-3-methylimidazolium chloride 1-Butyl-3-methylimidazolium BF4 1-Butyl-3-methylimidazolium BF4 Octylphenol ethoxylate

K2CO3

Ionic liquid-based

Micellar

n-Decyl tetra(ethylene oxide) Dioctanoyl phosphatidylcholine Alkyltrimethylammonium bromide

Na3C6H5O7/Na2C4H4O6 Na2CO3 Nonionic/ionic surfactants, salts and/or low-molecular-weight solvents (optional)

Fractionation, recovery, and purification of proteins, nucleic acids, bionanoparticles, and low-molecular-weight compounds. Characterization of protein and bionanoparticles’ physicochemical properties.

Fractionation, recovery, and purification of low-molecular-weight compounds and suitable macromolecules.

Fractionation, recovery, and purification of low-molecular-weight compounds and suitable macromolecules.

Fractionation, recovery, and purification of proteins, nucleic acids, bionanoparticles, and low-molecular-weight compounds.

778

Aqueous Two-Phase Systems

two immiscible phases.28 Although these systems have not been as extensively studied as the polymer–polymer and polymer–salt systems, they have been used for the fractionation of phytochemicals, amino acids, and even proteins such as enzymes.34,50 Specific advantages of these systems include the use of inexpensive constituents (particularly compared with some expensive polymers and co-polymers), easy constituent recovery and reutilization, reduced settling times, and low viscosity. The major disadvantage of alcohol–salt ATPS is that even though they have been used for the separation of some enzymes, many proteins are incompatible with the alcohol-rich phase. Therefore, a conformational shift leading to agglomeration, precipitation, and even denaturation is a risk for macromolecules partitioned on this type of system. As a result, most of the investigation addressing the use of alcohol–salt systems have focused on the fractionation of low-molecular-weight products.34

2.52.2.2.3

Micellar and Reverse Micelle ATPSs

Micellar and reverse micelle ATPSs exploit the capability that concentrated surfactant solutions exhibit to form two immiscible phases when a proper stimulus, such as a rise in temperature or a change in pH, is carried out. This results in a micelle-rich phase and a micelle-poor phase, between which the solutes and nanoparticles will distribute.7,39,40,51,60 The size and shape of the micelles can be controlled by varying parameters such as the surfactant concentration, temperature, pH, and ionic strength. Consequently, salts and some low-molecular-weight solvents can be added in order to further control the partition selectivity. This provides relative control over the partition behavior of the solutes in the system. These systems are usually formed using a surfactant. pH and ionic strength of the system have an important effect on the partition behaviour of proteins clearly related to the protein properties.7 Lysozyme, ribonuclease, and horseradish peroxidase have been extracted in such systems.40,51 Continuous extraction of lysozyme was successfully carried out in a spray column.41 The application of mixed micellar systems (constituted for two or more nonionic and/or ionic surfactants) is becoming popular due to the selectivity features that such mixed ATPSs have shown. The use of nonionic surfactants (such as n-decyl tetraethylene oxide and octylphenol ethoxylate) as constituents for the formation of micellar ATPS is preferred for the fractionation of products extremely sensitive to ionic strength. Although the use of nonionic surfactants is typical for this application, charged constituents may be used in order to exploit the potential effect of the electrochemical interactions on the partition behavior of solutes and particles.

2.52.2.2.4

Ionic Liquid-Based ATPS

Ionic liquids are organic salts that are in liquid state at standard atmospheric conditions. Ionic liquid-based ATPSs are formed when ionic liquids (such as 1-butyl-3-methylimidazolium chloride, 1-butyl-3-methylimidazolium hexafluorophosphate on 1-ethyl-3methylimidazolium acetate) are mixed with concentrated inorganic salt (phosphate, sulfate, or citrate) solutions.25,38 The partition behavior of solutes in the system can be manipulated based on the chemical nature of the ionic liquid (the presence of aliphatic chains, cyclic groups, and electrochemical charge) as well as its concentration. As in the case of alcohol–salt systems, the application of ionic liquid-based ATPS for the fractionation of macromolecules that are extremely sensitive to ionic strength and/or mild hydrophobic environments is limited. However, reports addressing the fractionation of proteins such as enzymes in ionic liquid-based ATPS may be found in the literature.25

2.52.2.3

Parameters That Influence Partitioning in ATPSs

The fractionation behavior of biomolecules and nanoparticles in ATPSs is generally described using the partition coefficient (Kp), a parameter which relates the concentration of a particular solute in the top and bottom phase, and is defined as follows: Kp ¼

XTP XBP

(11)

where XTP and XBP are the concentrations of the species X in the top and bottom phases, respectively. Although the partition behavior of proteins, nanoparticles, cells, and organelles has been extensively studied, the mechanisms governing the fractionation of these species are not totally understood due to the thermodynamic complexity of the physicochemical interactions involved. Several reports addressing the effect of the ATPS parameters as well as the solutes’ physicochemical properties on the partition behavior of biological products may be found in the literature.4,10 The solutes or particles to be partitioned interact with the constituents of the biphasic system via ionic and hydrophobic interactions, hydrogen bonds, van der Waals forces, and other noncovalent interactions. Additionally, the solutes being partitioned may interact with themselves, particularly when the concentration of a particular solute is high8,58 and other solutes present in the systems such as contaminants and additives. Since the net effect of all these interactions is likely to be different between the top and the bottom phases, the solutes partition selectively based on the achievement of the most favorable energy state (thermodynamic equilibrium). Although difficult to model accurately, the partition behavior of a particular solute on ATPS may be empirically estimated based on the solute physicochemical properties as well as the system parameters.4,10 The physicochemical properties of a specific solute or nanoparticle strongly influence its partition behavior in ATPS. The three most influential physicochemical properties are size (molecular weight and hydrodynamic diameter), superficial electrochemical charge,49 and hydrophobic character.10 Regarding cells and organelles, their superficial properties and equivalent diameters influence their partition behavior in ATPS.63 The selection of the system parameters (type of system, system constituents, TLL, and system pH) must be conducted based on the physicochemical properties of the product and contaminants, in order to achieve an effective separation.12 The mechanisms that influence the partition behavior can be exploited separately or in conjunction in

Aqueous Two-Phase Systems

779

order to achieve an effective migration toward one phase while contaminants fractionate to the opposite phase. The principles governing the partition behavior in ATPS can be categorized into four main groups: (1) size dependent, (2) electrochemical dependent, (3) hydrophobicity dependent, and (4) specific bioaffinity dependent.

2.52.2.3.1

Size-Dependent Partition

Since the solutes and nanoparticles to be partitioned in the system have a defined size (molecular weight and hydrodynamic diameter) as well as geometry (tridimensional conformation), they are subjected to the steric effects imposed by the constituents of the system. These steric effects are typically related to the available volume for the solutes to be fractionated toward a particular phase and is generally known as the free volume effect. Polymer–polymer and polymer–salt ATPSs are well known for promoting size-dependent partition on the solutes and nanoparticles. As these systems are constructed using polymers, with most of them of considerable length and molecular weight, the free volume available in polymer-rich phases is limited. This effect is enlarged as the TLL of the system increases, since the concentration of the polymer is also increased. Furthermore, the use of polymers of higher molecular weight creates significant steric effects, since the polymeric arrangement generates cavities reduced in volume. Salt-rich phases and ionic liquid-rich phases also exert steric limitations related to the free volume available. Hence, the volume occupied by the ions reduces the volume available for the fractionation of solutes and particles. The partition of solutes of high molecular weight, bionanoparticles, cells, and organelles are thus influenced by the free volume effect, which results in the partition of whole cells and organelles toward the interface. By contrast, low-molecular-weight compounds such as secondary metabolites and phytochemicals are not strongly influenced by the free volume effect since the steric restrictions on them are limited.

2.52.2.3.2

Electrochemical-Dependent Partition

Some ATPS constituents, such as salts and ionic liquids, are ionizable species. Furthermore, although some other constituents such as polymers (PEG and dextran) do not ionize in solution, they present weak dipole moments due to the presence of functional groups with strong electronegativity. Consequently, the electrochemical interactions have an important role on the partition behavior of proteins, other solutes, and particles.10,49 As opposite charges attract, the presence of charged constituents may generate a selective fractionation of oppositely charged solutes and particles toward a specific phase. The influence of the pH on the electrochemical interactions is fundamental. Therefore, the pH of the system may be manipulated in order to promote selective separation. PEG, the most widely used polymer in ATPS, has a positive dipole moment due to the presence of terminal hydroxyl groups. Therefore, the use of pH values above the isoelectric point (pI) of proteins and other macromolecules and nanoparticles may induce an additional affinity toward the PEG-rich phase.

2.52.2.3.3

Hydrophobic-Dependent Partition

Hydrophobic interactions play a major role in the fractionation of solutes, bionanoparticles, and cells in ATPS. Two well-known effects are involved in such interactions: the phase hydrophobicity effect and the salting-out effect.8 The phase hydrophobicity effect is directly related to the chemical identity of the constituents of the system as well as their concentration.10,59 Although such constituents are in principle hydrophilic, their relative hydrophobicity may vary considerably. Therefore, although both phases of the ATPS are rather hydrophilic, the top phase (PEG or other) is usually more hydrophobic. This favors the partition of amphipathic and less hydrophilic solutes and particles toward that particular phase. In polymer–salt systems, the phase hydrophobicity may be manipulated by varying the TLL, the polymer molecular weight, and by the addition of a salt such as NaCl. When the TLL increases, an intrinsic reduction of the water content is achieved. Therefore, the system becomes more hydrophobic as less water is available. Regarding polymer molecular weight, as this parameter increases an induced hydrophobicity is generated due to the presence of extensive hydrophobic areas. This is particularly evident for polymers such as PEG, which only has hydrophilic functional (hydroxyl) groups at its extremes, whereas the rest of the chain is primordially hydrophobic. Therefore, as the molecular weight of the polymer increases, the ratio of hydrophilic groups to hydrophobic area decreases, reflecting a rise in hydrophobicity. Over the last 20 years, a number of publications have clearly shown that rather hydrophobic proteins can be separated from their contaminants extremely efficiently in PEG/salt ATPSs with the addition of NaCl at a concentration of up to 10%. In most cases, the partitioning of the hydrophobic protein can be increased even several orders of magnitude, whereas the partitioning of contaminants under such conditions is not affected and tends to prefer the lower phase.11 Proteins that have been successfully separated in such systems include thaumatin, a-amylase, tissue plasminogen activator (tpA), and monoclonal antibodies. Regarding the salting-out effect, this is also related to the hydrophobic-dependent partitioning in ATPS. This effect is observed in systems with at least one highly ionic phase (polymer–salt ATPS, ionic liquid-based ATPS, etc.). In these cases, since the amount of water needed to dissolve the salts in the system is high, the solutes to be partitioned are only partially hydrated. Consequently, partitioning toward the less hydrophilic phase is favored under such circumstances. However, as the ionic strength further increases, the capability of the system to partially hydrate the solutes is lost. In such cases, a suspension generated by the aggregation of the solutes is observed.8 This effect is similar to that observed in protein salting-out precipitation procedures. Furthermore, in hydrophobic partitioning described in the previous paragraph in the systems with low and high concentration of NaCl an excellent correlation was found between log K and the inverse of the point at which a protein begins to be salted out (m*).11 The use of alcohol–salt ATPS presents several advantages for the exploitation of hydrophobicity for the selective fractionation of low-molecular-weight compounds such as phytochemicals and secondary metabolites. Several families of such bioactive compounds are soluble in low-molecular-weight alcohols such as ethanol and propanol. Therefore, their fractionation toward the alcohol-rich phase is highly favored. Finally, the use of micellar ATPS allows the fractionation of solutes and particles mainly

780

Aqueous Two-Phase Systems

based on their hydrophobicity. The micellar ATPSs are formed using a surfactant that forms micelles or reverse micelles under particular conditions of pH and temperature. These micelles interact with the biomolecules and particles in the system promoting their partition toward the micelle-rich or micelle-poor phase, depending mainly on their hydrophobicity.

2.52.2.3.4

Specific Affinity-Dependent Partition

Over the last 30 years, strategies focused on increasing the selectivity of ATPS have been developed. Such strategies involve the inclusion of specific ligands with affinity for particular solutes.5 This generates a controlled fractionation/adsorption of the product of interest toward a specific phase. Since biomolecules are well known for their specific biological activity, affinity ligands for most of the products of interest may be identified. The application of such strategies greatly improves the recovery and purification of products of interest in ATPS. The application of such affinity strategies on ATPS is further discussed later in this article.

2.52.3

Application of ATPSs for the Recovery of Biological Products

2.52.3.1

Multistage Extraction and Continuous Separation Using ATPSs

Depending on the specific recovery and purification needs, ATPS may be used as a single separation stage or a multistep extraction strategy.4,17 In some cases, the use of a single ATPS stage meets the recovery and purification needs of a particular process. Otherwise, the use of consecutive batch ATPS stages can be considered if the yield and/or purity requirements are not met using a single fractionation stage. Alternatively, system parameter and process conditions in consecutive stages can be varied to attempt a further increase in process performance. A common strategy comprises the use of two consecutive fractionation stages. The first stage is used as a selective extraction step, while the second is a back extraction. Figure 3 shows simplified flow diagrams representing single-stage and multistage ATPS-related strategies. ATPS strategies may be carried out in batch or continuous mode. Additional approaches such as countercurrent distribution (CCD) and countercurrent chromatography (CCC) for the fractionation, recovery, and purification of biological products using ATPS have been characterized.4,57 In CCD, molecules are separated based on their affinity for two immiscible liquid phases. These immiscible phases move in opposite directions (countercurrent) and are mixed and settled to allow phase separation. The top phase is transferred off in one direction and the bottom phase in the opposite direction in order to generate a new thermodynamic equilibrium. Figure 4 presents a simplified representation of CCD. CCC is a high-resolution chromatography technique based on the CCD principle. The use of CCC has advantages over high-performance liquid chromatography (HPLC) such as low pressure drops, higher processing capability, and lower maintenance cost.57 For CCC, the use of low-molecular-weight and low-viscosity system constituents is preferred for several reasons, including A

Top phase

Sample

Bottom phase

Product in top phase

Contaminants

B

Top phase

Bottom phase

Sample

Product in top phase

Back extraction solvent

Product in back extraction solvent

Contaminants

Figure 3 (A) Single-stage and (B) multistage extraction strategies for the fractionation, recovery, and purification of biological products. Product () and contaminants (,).

Aqueous Two-Phase Systems

781

A

B

Response

C

1

2

3

4

5

6

7 8 Tube

9

10

11

12

13

Figure 4 Countercurrent distribution of biological products using aqueous two-phase systems. (A) The biological sample containing three products with different partition coefficients is partitioned on a first ATPS (Kp ¼ 5 (,), Kp ¼ 1 (B), and Kp ¼ 0.2 (D)); (B) the top phase of the systems are sequentially transferred promoting the fractionation of the products on different tubes based on their partition coefficients; (C) the countercurrent distribution procedure generates fractions enriched in each product.

enhanced mass transfer between phases, reduced constituent deposition on the inner walls of the CCC equipment, and lower pressure drops.

2.52.3.2

Process Integration Using ATPSs

The use of too many unit operations in the downstream processing of biological products usually results in lower recovery yields. In this context, the application of ATPSs for the development of process integration strategies is interesting, although it can be difficult to operate practically. Process integration involves the use of one unit operation able to substitute two or more operations achieving similar process objectives. As a direct result, the total number of recovery and purification steps can be diminished and the process yield may increase without compromising the purity of the product. Four major strategies regarding process integration using ATPS can be identified: (1) extractive bioconversion, (2) extractive fermentation, (3) extractive disruption, and (4) extractive purification.17 A simplified representation of these four strategies is shown in Figure 5. Extractive conversion involves the synthesis of a particular biological product in one phase of the ATPS and the continuous fractionation of the synthesized product toward the opposite phase in order to be recovered (Figure 5(A)). Since the product does not accumulate in the reactive phase, the synthesis is favored due to thermodynamic equilibrium.37 An example is the enzymatic hydrolysis of starch with amylase and amyloglucosidase using thermoseparating ATPS. The extractive fermentation strategy involves the growth of an expression system synthesizing an extracellular product in one of the phases of the system, while the product being expressed partitions toward the opposite phase to be recovered (Figure 5(B)). The biphasic systems are inoculated with the expression system which grows selectively in a particular phase. The continuously secreted product partitions, at least partially, toward the opposite phase. This strategy allows the continuous removal of the product of interest from the fermentation phase helping to overcome low productivity yields due to inhibition. In this approach, restrictions regarding compatibility between the expression system and the ATPS composition are critical. Chen and Lee22 reported the production of Serratia marcescens extracellular chitinase in PEG–dextran ATPS using the extractive fermentation strategy. While the biomass (S. marcescens cells) concentrated in the top phase of the system, the enzyme was recovered from the bottom phase. ChavezSantoscoy et al.21 reported the application of PEG–dextran ATPS for the potential extractive fermentation of cyanobacterial products, specifically lutein and b-carotene. The expression system used by the authors was the cyanobacteria Synechocystis sp. The lutein (one of the monitored products) exhibited affinity toward the bottom phase, while b-carotene remained mainly in the top phase. When the product of interest is not secreted into the culture media, cell disruption is required. In this context, the application of ATPS for the integration of cell disruption and primary recovery represents an innovative approach (Figure 5(C)). In such a strategy, cell disruption or permeation can be done by mechanical or chemical means within the ATPS. After cell disruption or permeation, the homogenate is allowed to settle in order to promote the separation of the two aqueous phases. Therefore, the fractionation and separation of the released product from contaminants such as other biomolecules, biomass, and cell debris are achieved in a single unit operation. A paradigmatic process integration case is the large-scale in situ isolation of periplasmic human insulin-like growth factor I (IGF-I) from Escherichia coli.30 IGF-I accumulated mainly as refractile inclusion bodies in the periplasmic space. After fermentation, a strong chaotrope and reductant was added to the crude broth achieving solubilization and extraction of 90% of all IGF-I. Then PEG-8000 was added to generate an ATPS that allowed 90% recovery and 97% purity in the light phase. This procedure was

782

Aqueous Two-Phase Systems

B

A

Fresh top phase with substrate X

Bottom phase with product Y

X

Top phase depleted in X

Y

Y

Y

Y

Y

Y

Y

Fresh top phase with culture media (CM)

Fresh bottom phase

CM

Bottom phase with product Y

Y

Y

Y

Y

Y

Y

Y

Top phase depleted in CM

Fresh bottom phase

D

C Biomass with intracellular product Top phase

Affinity ligand

Bottom phase

Top phase Top phase with released product

Sample with product Bottom phase Top phase with product–ligand complex

Cell debris

Bottom phase with contaminants

Bottom phase with contaminants

Figure 5 Process integration aqueous two-phase system–based strategies. (A) Reaction þ extraction ¼ extractive bioconversion. (B) Fermentation þ extraction ¼ fermentative extraction. (C) Cell disruption þ extraction ¼ extractive disruption. (D) Affinity þ extraction ¼ extractive purification.

scaled up to 1000 l. Rito-Palomares and Lyddiatt53 studied the extractive disruption strategy using commercial baker’s yeasts disrupted in a bead mill integrated with PEG–phosphate ATPS. The authors concluded that ATPS parameters (such as PEG molecular weight, TLL, and VR) could be manipulated in order to achieve a selective fractionation of the product of interest (G3PDH) to a particular phase while the cell debris partitions to the opposite phase. The use of ATPS with chemically modified polymers as constituents and the addition of affinity ligands usually results in a significant increase in the selectivity of the extractive system (Figure 5(D)). The strategies involving the use of affinity-enhanced ATPS are vast and the application of such systems is discussed in the following section.

2.52.3.3

Bioaffinity-Enhanced Partitioning on ATPSs

Bioaffinity-enhanced ATPSs allow the integration of the primary recovery and purification of biological products. Affinity fractionation in ATPS may be achieved following two different strategies: (1) free ligand addition and (2) ligand-coupled constituents. Figure 6 depicts these two strategies:

• •

Free ligand addition. The ligand (a molecule with specific affinity for the product of interest) is added to the system and the ligand partitions unevenly between the phases. When the product of interest is introduced into the system it presents selectivity for the phase where the ligand is present (Figure 6(A)). Ligand-coupled constituents. The ATPSs are totally or partially constructed using derivate constituents with covalently attached ligands that exert affinity toward the product of interest. Consequently, when the product is introduced on the biphasic system, it partitions preferently toward the ligand-coupled rich phase (Figure 6(B)). The covalent coupling of the ligand to one of the system constituents usually requires complex chemical procedures that involve an activation and a grafting stage. The use of ligand-coupled constituents is preferred since no previous characterization of the free ligand partition behavior is required.

Andrews et al. reported the use of PEG–dextran ATPS for the recovery and purification of thaumatin (a protein flavor enhancer and sweetener) and trypsin (a proteolytic enzyme) with the inclusion of glutathione and trypsin inhibitor as biospecific ligands.5 Thaumatin presented a 17-fold increase in Kp (from 0.27 to 4.6) in modified aqueous biphasic systems with PEG–glutathione. Trypsin presented a dramatic increase in Kp from 0.5 to 16 (a 32-fold increase) in systems containing PEG–trypsin inhibitor conjugates, demonstrating that recovery and purification of these proteins could be dramatically increased by using modified ATPS with biospecific ligands attached to the top phase polymer. Barbosa et al.14 characterized the partition behavior of prepurified plasmid DNA (pDNA) from model buffer solutions using PEGylated zinc finger-glutathione-S-transferase fusion protein (PEG-ZnF-GST) in PEG–dextran ATPS.14 In the presence of pDNA containing a specific oligonucleotide recognition sequence, the zinc finger moiety of

Aqueous Two-Phase Systems

A

Bioaffinity ligand ( ) Top phase polymer (TPP)

Top phase polymer (TPP)

Product of interest ( ) and contaminants ( )

Top phase with ligandproduct complex ( )

Product of interest ( ) and contaminants ( )

Top phase with TPP-ligand-product ) complex (TPP

Bottom phase polymer (BPP)

TPP-bioaffinity ligand complex (TPP )

B

783

Bottom phase polymer (BPP)

TPP

TPP TPP

TPP TPP

TPP TPP

TPP

Figure 6 Strategies for the bioaffinity-enhanced partitioning on aqueous two-phase systems. (A) Free ligand addition and (B) Ligand-coupled constituent.

the PEG-ZnF-GST bound to the plasmid promoting its selective partition toward the top phase (PEG-rich phase). No pDNA was detected in the bottom phase (dextran-rich phase).

2.52.3.4

Products of Interest Partitioned Using ATPS

The use of ATPS for the fractionation, recovery, and purification of biotechnology-related products is well characterized. Biomolecules such as proteins, nucleic acids, viruses, virus-like particles, bionanoparticles, and low-molecular-weight compounds (such as phytochemicals and metabolites) have been fractionated in ATPS. Additionally, the fractionation of biomass such as viable cells, organelles, and cell debris has also been studied and characterized. Table 2 shows examples of studies addressing the separation, recovery and purification of biomolecules, bionanoparticles, and biomass using ATPS. Such examples represent just a small fraction of all the studies available in the literature. Next, a general description of the application of ATPS for the recovery and purification of these biotechnological products is presented.

2.52.3.4.1

Proteins

Proteins are the most versatile biomolecules in terms of functionality and structure. These macromolecules are used in all biotechnology-related applications. The use of ATPS has focused mainly on the separation, recovery, and purification of proteins such as enzymes, therapeutic proteins, and functional proteins as additives.

Representative studies addressing the recovery, separation, and purification of biological products using aqueous two-phase systems

784

Table 2

Product

ATPS (type of system)

Main objective

Main result

Reference

Proteins

Lipase

2-Propanol/K2HPO4 (alcohol–salt) PEG 4000/sulfate with 8.8% NaCl (polymer–salt) Ethylene oxide copolymer (PEO–PPO)/MgSO4 (polymer–salt)

Recovery yield of 99% and purification factor of 13.5 was achieved 53-fold purification with 86% purity

50

a-Amylase

Recovery and purification from Burkholderia pseudomallei Recovery and purification from Bacillus subtilis Characterize partition behavior and carry out the extractive bioconversion of starch to glucose Recovery and purification from Escherichia coli contaminants Production of Serratia marcescens extracellular chitinase using an extractive fermentation strategy Recovery and purification from transgenic sheep milk Characterize the effect of using PEG–glutathione and PEG–trypsin inhibitor on the bioaffinity enhanced partition In situ isolation from E. coli

Amylase and amyloglucosidase Thaumatin Chitinase Human-a-antitrypsin Proteins

Nucleic acids

Thaumatin and trypsin

PEG 6000/phosphate with 2.0 M NaCl (polymer–salt) PEG/dextran (polymer–polymer) PEG 1500/phosphate with up to 15% NaCl (polymer–salt) PEG/PEG-ligand/dextran (polymer–polymer)

Insulin-like growth factor (IGF)-I a-Galactosidase

PEG 8000/salts in fermentation media (polymer–salt) PEG/K2HPO4 (polymer–salt)

Monoclonal Antibody (MAb/IgG)

PEG 1450/phosphate þ12% NaCl (polymer–salt)

Recovery and purification from Hybridoma supernatant

Recombinant human immunoglobulin gamma (IgG) Human recombinant interferon a1 (rhIFN-a1)

PEG/K2HPO4 (polymer–salt)

B-phycoerythrin

PEG/K2HPO4 (polymer–salt)

Recovery and purification from Chinese hamster ovary (CHO) using a multistage extraction strategy Recovery and purification from genetically modified E. coli using a multistage extraction strategy Primary recovery and partial purification from Porphyridium cruentum

Plasmid DNA (pUT649) and RNA

Isooctane/ethylhexanol/methyltrioctyl ammoniumchloride (reverse micellar)

Fractionate RNA and plasmid DNA from preconditioned cleared E. coli lysate

Plasmid DNA vector (6.1 kbp)

PEG/(NH4)2SO4 (polymer–salt)

Primary recovery and partial purification from E. coli lysate

Polyplexes

PEG/(NH4)2SO4PEG/PEG–ligand/dextran (polymer–salt and polymer)

Multistage extraction strategy with PEGylated polyethyleneimine (pPEI) as affinity complex

PEG/K2HPO4 (polymer–salt)

Primary recovery and partial purification from Aspergillus oryzae

Enzymes and starch partitioned toward the bottom phase while glucose migrated to the top phase HPLC analysis. Virtually all thaumatin in top phase and E. coli in bottom Biomass growth selectively at the top phase while the chitinase partitioned toward the bottom phase 91% yield with 73% purity

58 37 19 22 29

The Kp of thaumatin and trypsin presented a 17- and 32-fold increase due to PEG–ligand conjugates

5

70% cumulative recovery and 97% purity (up to 1000-l scale-up) Selective fractionation toward the bottom phase with a recovery yield of 88% and a purification factor of 3.6 Extraction with 12% NaCl and no NaCl for back-extraction. A 80% purity was achieved An 89% recovery yield and a 75% purity was achieved

30

Affinity toward the top phase achieving a recovery yield of >99% and a purification factor of 25 Affinity toward the top phase achieving a recovery yield of 90% and a purification factor of 4 The recovery and purification of plasmid DNA was achieved simultaneously removing RNA Affinity toward the bottom phase with 85–100% recovery yield depending on sample loading (% w/w) A recovery yield of 100% was achieved while contaminants such as RNA and proteins were completely removed

27

48 9 55

16 60 62 24

Aqueous Two-Phase Systems

Type of product

Bionanoparticles

Biomass

Bacteriophages (4X174, P22, and T4)

n-Decyl tetra (ethylene oxide) (micellar)

Characterize partition behavior and correlate with their geometry and size

Virus-like particles (VLPs)

PEG/sulfate different-size PEGs (polymer–salt)

Recovery and purification from yeast proteins and debris

Rotavirus-like particles (dlRLPs)

PEG/K2HPO4 (polymer–salt)

Nanospheres, nanowires, and DNA-derivatized nanowires Human platelets

PEG/dextran (polymer–polymer)

Primary recovery and partial purification from a Trichoplusia ni ovary–baculovirus expression system In situ fractionation, assembly and recovery of functionalized nanoparticles

Several polymer/blood mixtures (polymer-based) PEG/dextran (polymer–polymer)

Obtention of a platelet-rich fraction from whole blood and remotion of contaminants Fractionate cells with different concentrations of anthocyanins.

Plasma membranes and Golgi apparatus

PEG/dextran (polymer–polymer)

Codeine and papaverine (opium alkaloids) 2,3-butanediol

1-Butyl-3-methylimidazolium chloride/salt (ionic liquid-based) Ethanol/K2HPO4 (alcohol–salt)

Fractionation and isolation of plasma membrane and Golgi apparatus from mammalian cells Fractionation and analysis from aqueous samples of Pericarpium papaveris Separation from Klebsiella pneumoniae complex fermentation broth

Lutein and b-carotene

PEG/dextran (polymer–polymer)

Cultured strawberry cells

Low-molecular-weight compounds

Evaluate the application of ATPS for the extractive fermentation of cyanobacterial products

Independently of the shape and size of the bacteriophages partitioned toward the micelle-poor phase Different strategies: VLPs in top phase and proteins in bottom or VLPs in top and debris in bottom An overall recovery yield of 85% and a purification factor >50 were achieved

39 6 15

In situ binding of Au nanospheres with Au nanowires via selective DNA hybridization at the ATPS interface A platelet-rich fraction suitable for clinical applications was obtained The strawberry cells fractionated into two population based on their anthocyanin content Different partition behaviors were observed among different plasma membrane types

32

Recovery yields >90% and >99% for codeine and papaverine, respectively Recovery yield >98% was achieved simultaneously removing most of the contaminants Synechocystis sp. growth selectively at the top phase while lutein partitioned toward the bottom phase

38

61 26 46

34 21

Aqueous Two-Phase Systems 785

786

Aqueous Two-Phase Systems

Several studies have been conducted regarding the recovery of enzymes with application in the food, detergent, and paper industries, as well as in bioremediation and medical treatments. Extensive research concerning the application of ATPS for the primary recovery and partial purification of therapeutic proteins has been conducted. The ever-increasing interest in proteins used for the treatment and prevention of a vast range of diseases becomes obvious considering the number of reports related to the production and purification of these types of macromolecules. Monoclonal and polyclonal antibodies, a-interferon, as well as blood/serum-related proteins are examples of therapeutic products of which fractionation behavior in ATPS has been studied.11 It was found that when adding important concentrations of NaCl (up to 10%) to PEG/salt systems the partition coefficient of the protein thaumatin could be directed to the more hydrophobic PEG phase.19 Later, it was shown that this effect could be much more dramatically observed for the very hydrophobic protein a-amylase. A similar effect was observed for the hydrophobic protein tpA and also, very interestingly, for the partition and purification of monoclonal antibodies that are also very hydrophobic proteins. An industrial serum-free, crude, concentrated culture supernatant of hybridoma produced murine immunoglobulin G (IgG) with a relatively low level of protein contaminants (14% IgG purity) was processed in this system. After the back extraction the contaminants were reduced 18-fold giving IgG with 80% purity. A 5.9-fold purification was obtained out of a 7.3 maximum possible (at 100% pure IgG).9 All of the IgG were recovered. This pioneering finding has been exploited in a European project which has studied in detail the partitioning and important scale-up and processing factors of the very hydrophobic MAbs in ATPS.55 Guan et al.27 reported the use of a two-stage ATPS extraction strategy for the recovery and purification of human recombinant interferon a1 (rhIFN-a1) expressed in genetically modified E. coli. Transgenic sheep milk containing the protein human a-antitrypsin (AAT) was partitioned in PEG–sulfate and PEG–phosphate biphasic systems.29 Individual partition coefficients for AAT and some of the milk proteins were determined in these systems. The effects of PEG molecular weight, pH, and the inclusion of NaCI on the partitioning of the proteins were also studied. It was found that increasing the concentration of NaCI and decreasing the molecular weight of the PEG resulted in an increase of the partition coefficients of the proteins to the upper (PEG) phase. This partitioning effect was greater for the more hydrophobic proteins and particularly in systems having a pH close to the isoelectric point of the protein. Under the most favorable conditions using a 4% w/w loading of transgenic ovine milk, a 91% yield of AAT in the PEG phase with a purity of 73% was obtained. The recovery and purification of functional proteins with use in the food and cosmetic industries as well as in molecular biology applications have also been reported. Phycobiliproteins (colored proteins that form part of the photosynthetic apparatus of cyanobacteria and red algae) such as C-phycocyanin (CPC) and B-phycoerythrin (BPE) have been fractioned and partially purified using processes involving ATPS. These colored proteins have applications as color agents for food and cosmetic products. Additionally, phycobiliproteins are used in molecular biology and microscopy applications as fluorescent colorant used for labeling and immunodetection. Benavides and Rito-Palomares16 characterized the recovery and partial purification of BPE from Porphyridium cruentum using PEG–phosphate ATPS. Under optimum system parameters, a BPE recovery yield and a purification factor of 90% and 4, respectively, were achieved.

2.52.3.4.2

Nucleic Acids

The development and implementation of molecular biology and genetic engineering processes usually require the efficient isolation of nucleic acid-based biomolecules such as DNA and RNA (genomic and plasmidic) as well as genetic vectors. Studies regarding the application of ATPS for the fractionation, recovery, and partial purification of genetic material, particularly plasmid DNA vectors, have been reported. Trindade et al.62 reported the application of PEG–sulfate ATPS for the primary recovery and partial purification of a plasmid DNA vector extracted from E. coli. Streitner et al.60 reported the application of reverse micellar ATPS for the purification of pharmaceutical-grade plasmid DNA. The partition behavior of a 4.6-kb plasmid (pUT649) and E. coli RNA was studied in micellar systems of isooctane, ethylhexanol, and the surfactant methyltrioctyl ammoniumchloride. Selected ATPSs were able to efficiently and selectively fractionate the plasmid DNA from the E. coli RNA. Furthermore, the authors demonstrated that some fractionation systems were even able to selectively extract the supercoiled form of the plasmid. The recovery of polyplexes (genetic vectors conformed by DNA and protective polymer material with potential use in gene therapy) has been characterized in ATPS. Duarte et al.24 reported the recovery and purification of polyplexes using polymer–salt and polymer–polymer ATPS. The process developed by the authors consisted in a two-stage ATPS extraction comprising a PEG– sulfate and an affinity-enhanced PEG–dextran system, followed by ultrafiltration. PEGylated polyethyleneimine (pPEI) was used in the second ATPS stage as an affinity ligand for the product of interest.

2.52.3.4.3

Virus, Virus-like Particles, and Other Bionanoparticles

The production, downstream processing, and application of bionanoparticles are of great scientific interest due to the various applications these complex structures have. Common applications for these particles include the production of vaccines, delivery vectors for gene therapy, and molecular assemblies for drug delivery. Andrews et al.6 carried out two strategies, each with two steps, to purify yeast virus-like particles (VLPs). The first strategy included a PEG 400 or 600 and (NH4)2 SO4 first stage for debris separation and a PEG 4000 or 8000 and (NH4)2 SO4 with added NaCl or phosphate for VLP purification from the proteins. The second strategy included VLP recovery in the interphase and total purification from the debris into the top PEG phase in the second stage. Benavides et al.15 reported the use of PEG–phosphate ATPS for the recovery and partial purification of double-layered rotavirus-like particles (dlRLPs) produced using a Trichoplusia ni ovary–baculovirus expression system using ATPS PEG–potassium phosphate. Liu et al.39

Aqueous Two-Phase Systems

787

studied the partition behavior of several proteins as well as bacteriophages (such as 4X174, P22, and T4) on micellar ATPS using n-decyl tetraethylene oxide (C10E4) as surfactant. The use of ATPS for the in situ fractionation, assembly, and recovery of functionalized nanoparticles was done by Helfrich et al.32 who conducted studies in order to characterize the partition behavior of gold (Au) and silver (Ag) nanospheres, nanowires, and DNA-derivatized nanowires in PEG–dextran ATPS. The authors reported that the partition behavior of Au and Ag nanospheres in ATPS greatly depends on their size in accordance with the free volume theory. While small nanospheres ( 0, kd may have a peak value in the water content range considered if a > 0, b < 0, kd would just reduce as water content decreases if a < 0, b > 0, kd may have a minimum value in the water content range considered

In order to evaluate the effects of these values of the survival of probiotics, Eq. (7) needs to be integrated over the sample size. For a simple slab, it can be expressed as:   RL þbX ∅dx k0 0 expðaXÞexp  EdRT d∅ ¼ (13) dt L d∅ k0 ∅ ¼ dt L

ZL 0

    b Ed Xexp exp a  RT RT

(14)

Eq. (9) can be further simplified to mimic Frank-Kamanetskii into: 1 d∅ k0 Ed ¼  exp ∅ dt L RTref

! ZL

where ∅ ¼ 


0

  b Xexpð∅Þdx exp a  RT

RTref

2

E 
d  T  Tref

(15)

(16)

It can be observed from Eq. (10) that the inactivation rate increases with the increase of temperature and increase of moisture content. During thermal dehydration, the temperature increases but the moisture content reduces. This indicates that the overall inactivation rate during drying is a result of competitive process between increase of temperature and reduction of moisture content. On top of the above mentioned models,72 implemented probabilistic-based modeling, represented as Weibull equation to describe the probiotic survival. Generally, the models describe well the experimental data. Nevertheless, the models can be improved by incorporating protecting medium, oxidative stress and osmotic stress.40

Drying in Biotechnology

2.55.4

Drying of Enzymes

2.55.4.1

Spray Drying Implemented for Drying of Enzymes

829

Spray drying is commonly used for encapsulation of enzymes through which the activity and mobility can be effectively altered.89,90 The spray drying may potentially engineer the enzymes to have better stability and adjustable bioavailability.89,90 On top of its effectiveness, spray drying is preferred as an encapsulation method due to its low cost and ease for scaling up.89,90 Lysosome is a spheroidal hydrolase with antimicrobial properties which can serve as natural food preservatives.82 The functionality and solubility of the lysosomes are dependent on the tertiary structures governed by the conformational changes.4 While spray drying can be used for encapsulating lysosomes, it may lead to product degradation due to high shear, high temperature, air/water interface and water removal. These stresses may unfold the protein which eventually results in aggregation and denaturation.5 Nevertheless, sugars can protect the degradation through the formation of hydrogen bonds.48 Interactions with sugars may also increase the extent of protein secondary structure.32 Similarly, nisin is a small peptide which possesses antibacterial activity against spoilage microorganisms.5 It can be added in food packaging materials as antimicrobial. The application without encapsulation may lose the activity. The encapsulation can help to protect nisin from environmental conditions, deter interactions with food materials and manipulate the release properties.30 Due to its high isoelectric point, nisin can be complexed with polysaccharides through electrostatic interactions.5 Spray drying has been used for encapsulation of lysosome by complexing it with pectin. Due to the high proportion of ionised carboxylic acid groups, the macromolecules become negatively charged resulting in electrostatic interactions with lysosomes at neutral pH.5 The encapsulation through spray drying was effective to deter the aggregation which could be because of the configurational changes of the complex molecules. Complexation with intermediate concentration of pectin (between 0.2 and 0.4 g L1) was able to protect the lysosome activity. However, the low (below 0.2 g L1) and high concentration (above 0.4 g L1) of pectin did not give the similar degree of protection which may be because of insufficiency of protection and limited accessibility of substrates toward the active site. The high concentration of pectin may raise the attractive forces between the pectin chains which may destabilise the lysosome structure.5 When used for encapsulating nisin, pectin and alginate were able to enhance the hydrophobicity.4 Based on the measurement of turbidity, the complexation with pectin and alginate through spray drying was shown to be able to preserve the activity. The nisinalginate complex gave higher turbidity than the nisin-pectine complex which indicated that the size and number of the nisinalginate complex was more important.4 In addition, due to the heat-induced aggregation, the turbidity of the untreated complex was lower than that of reconstituted spray-dried nisin-pectine complex.4 In terms of the antimicrobial activity, nisin microcapsules had the highest antimicrobial activity compared to nisin-pectin and nisin-alginate capsules. This may be because the nisin structure and activity were altered due to the complexation with pectin or alginate.

2.55.4.2

Freeze Drying Implemented for Drying of Enzymes

Freeze drying is suitable for drying of enzymes since enzymes are susceptible to three main types of degradation, i.e., denaturation, hydrolysis and formation of aggregates.56 Heating may decrease the enzyme’s activity since it destroys the chemical forces that support the conformity leading to unfolding, involving formation of secondary and tertiary structure.31 Since water content and storage temperature affect thermal stability and activity of the enzymes, freeze drying parameters need to be carefully adjusted. Martinez et al.57 reported the study of freeze drying of lysosome from chicken egg white. For the samples with water content 4%, no signs of denaturation were observed. However, they absorb the water from air in very short storage period and therefore standardised method for controlling the water content is important. For preserving the activity and adjusting the release profiles, enzymes can also be encapsulated in freeze dried matrix involving biomaterials.39 Through modification of gel formation between cationic and anionic polyelectrolytes, the release parameters can be altered. The enzyme entrapment behavior can be further amended by adding filler such as montmorillonite (MNT) nanoclay.52 The recombinant firefly luciferase was encapsulated into xanthan gum and chitosan. The drying step was carried out on the heat exchanger at 20  C for 30 h at a chamber pressure of 10–20 Pa. The freeze drying successfully stabilised the enzyme in the foam as evidenced by the characteristics of bioluminescence reactions. Due to the increase of viscosity resulted by the addition of MNT, smaller pores were produced. The MNT seems to contribute to stiff polymeric structure providing the stability at acidic conditions. The release profiles were dependent on the pH of external solution, and it was found that the release was higher in pH of 6 than in pH of 8. Through addition of MNT, the release rate can also be adjusted.52

2.55.4.3

Spray-Freeze Drying Implemented for Drying of Enzymes

While freeze drying can be used to produce dried enzymes, the applications are limited by its high operating and capital costs. Spray drying, on the other hand, offer a fast drying but the conformity of biological molecules may be affected due to high temperature.27,83,102 By integrating the advantages of both freeze and spray drying, spray freeze drying can be implemented for production of enzymes. In addition, a unique micro-fluidic jet spray dryer has been constructed and used to produce uniform microparticles where the enzymes are embedded in the structure.93 This dryer implements specially fabricated micro-fluidic nozzles capable to generate monodisperse and uniform droplets. Trypsin was encapsulated on trehalose through spray freeze drying. In this process, the

830

Drying in Biotechnology

precursor solutions were atomised by using microfluidic nozzle above a plate containing liquid nitrogen. The suspended frozen droplets were then transferred to vacuum freeze drying to remove the liquid nitrogen.93 The spray freeze dryer implementing micro-fluidic mentioned above successfully produced trypsin-immobilised microparticles which are spherical and open porous regardless of the composition. The trypsin molecules were partially concentrated on the surfaces indicated by formation of thin surface trypsin enriched layer.103 The protection of trypsin was firstly because of hydrogen bonding between trehalose and trypsin molecules which stabilised the trypsin.51,103 In addition, the aggregation and unfolding of trypsin was retarded by the glass matrix generated by the amorphous trehalose.103 The equal proportion of trehalose and trypsin gave the maximum protection (97.3%) of trypsin activity. The higher ratio of mass of trehalose to trypsin decrease the trypsin’s activity which is probably because of more enhanced adsorption of trypsin at the air/water interface during drying. This is in agreement with80 showing the increase of trypsinogen at the surface along with the rise of trehalose.

2.55.5

Drying of Vaccines

Most vaccines are distributed in liquid form whose storage requires low temperature (2–8  C) to ensure the stability. However, this complicates the distribution of vaccines as cold chain is required. Dried vaccines can prevent this complexity since they have longer shelf-life at higher temperature.78,86 Spray drying has been implemented to produce a number of vaccines including influenza, measles, hepatitis B and tuberculosis vaccines.12,50,54,75

2.55.5.1

Spray Drying Implemented for Production of Vaccines

Production of influenza dried vaccines have been attempted by Kanojia et al.43 by using closed loop spray drying where nitrogen was used as drying medium. The spray drying employed inlet temperature between 110 and 160  C, gas flow rate between 7.3 and 17.5 L min1 and feed flow rate between 1 and 4.5 mL min1. Trehalose was chosen as an excipient and the concentration was varied in the range of 100 and 150 mg mL1. Drying under this condition resulted in powder moisture content between 1.2% and 4.9% and yield ranged from 42% to 80%. The hem agglutination test indicated that there was no significant loss in antigenicity resulted from spray drying. The storage stability test showed that the vaccines remained stable for 3 months during storage at 60  C.43 By using trehalose also as excipients, Sou et al.81 (in v2) produced influenza dried vaccines with stability for 2 months under storage at 40  C. Similarly, trehalose containing influenza vaccines was shown to be stable for 2 months under storage at 50  C.95 Ohtake et al.67 implemented spray drying for manufacturing of dried measles vaccines where the vaccines were formulated in sucrose, trehalose, gelatin, glycerol and Pluronic F68 in potassium phosphate buffer at pH 7. The spray drying conditions are as follows: the feed rate, outlet temperature and atomisation pressure was 0.5 mL min1, 40  C and 15 psia, respectively. Similar to the production of dried influenza vaccines, nitrogen was employed as drying gas. The powders produced by the spray drying were then freeze dried for 12 h at shelf temperature of 15  C. Based on these conditions, residual moisture content ranging from 1.7% to 3.9% were produced with glass transition temperature between 48.8 and 64.3  C.67 In terms of stability, most formulations showed 10) and the six-volume reference work Comprehensive Biotechnology, both published by Elsevier. His honors include the premier awards of the Canadian Society for Chemical Engineering and the American Chemical Society, Biochemical Technology Division. He is an elected fellow of the American Institute for Medical and Biological Engineering (FAIMBE), “one of the highest honors for a bio-engineer”, and of the Royal Society of Canada (FRSC), “the highest accolade for a Canadian scholar”.

v

This page intentionally left blank

VOLUME EDITORS Michael Butler is the Chief Scientific Officer (CSO) of the National Institute for Bioprocessing Research and Training (NIBRT), Ireland; Adjunct Full Professor in University College Dublin; and Distinguished Professor Emeritus of the University of Manitoba, Canada. Born in Wales, he gained his BSc degree in Biochemistry at Birmingham University, MSc at the University of Waterloo, and PhD at King’s College, University of London. Following this, he became a Lecturer and subsequently Principal Lecturer in Biotechnology at Manchester Metropolitan University (1974–90). He was appointed in 1990 to an Industrial Chair in Fermentation Technology at the University of Manitoba, where he was awarded the title Distinguished Professor in 2008. He has published more than 150 articles in peer-reviewed journals as well as written and edited 7 books in the area of animal cell technology. He is an editor for Biotechnology Advances and Comprehensive Biotechnology as well as being on the editorial board of Biotechnology and Bioengineering. He has always collaborated closely with industry and is a past recipient of the prestigious Canadian national Synergy Award for University-Industry innovation (2004). He was the director of MabNet, a Canadian sponsored network for monoclonal antibody production. His previous experience has included a period as Associate Dean Science at Manitoba and several periods as Visiting Scientist or Professor at MIT, Animal Virus Institute (Pirbright), and Universities of Oxford and Rio de Janeiro. In the past he has served on several grant awarding committees, including NSERC (Ottawa), NSF (Washington), MHRC (Winnipeg), and Alberta Ingenuity (Edmonton). He was a founding member of Protein Expression in Animal Cells (PEACe) and elected to the executive committee of ESACT. He was a leader of a biotechnology subgroup for a Canadian government Department of Foreign Affairs mission to Brazil; advisory panel to Swedish program on biotechnology; and chair of SFI advisory panel (Dublin). He founded Biogro Technologies Inc., a research spin-off company on media development. His research work focuses on the development of bioprocesses using mammalian cells for the production of recombinant proteins, monoclonal antibodies, and viral vaccines. He is particularly interested in the bioprocess conditions that can be used to control the biochemical structure of glycoproteins and hence the quality of biopharmaceuticals.

Colin Webb is a Professor of Chemical Engineering at the University of Manchester, UK. He graduated in 1976 as a Chemical Engineer and added a PhD in Biochemical Engineering in 1980 both at the University of Aston. He joined UMIST as a Research Associate in 1979, taking up a lectureship in biotechnology in 1983. He was appointed to a newly created industrially sponsored chair in 1994 and established the Satake Centre for Grain Process Engineering (SCGPE) at Manchester in the same year. As Director of the Centre, he raised awareness of the novel uses that cereals could be put to, particularly as sustainable feedstocks for alternative chemicals and was honored, in 1999, as the UK’s first Distinguished Fellow of the International Academy of Food Science and Technology. His research is at the interface between biotechnology and chemical engineering and is largely directed toward the sustainable bioconversion of agricultural raw materials and the development of integrated biorefinery systems. He has supervised to successful completion a total of 130 students for higher degrees, including 41 PhDs, and has more than 300 publications, including 11 books. Colin’s H-index is 51 based on more than 9000 citations. He is Editor of The Biochemical Engineering Journal and is an editorial board member of several other biotechnology related journals. Within UMIST and the University of Manchester, he has been Head of Department, Head of School, and Associate Dean. Colin has been an external advisor to a large number of universities worldwide, including international scientific advisor to Kobe University, Japan (2007–10), and a visiting professor at universities in Spain and Australia. As a practicing chemical engineer, he has been Vice-President of IChemE (2012–18), Chair of Accreditation (2004–12), and recipient of five UK government SMART awards, the IChemE Hanson Medal (2006), Council Medal (2019), Donald Medal (2019), and the IChemE Bioprocessing Prize (2011).

vii

viii

Volume Editors Dr. Antonio Moreira is Vice Provost for Academic Affairs and Professor of Chemical, Biochemical and Environmental Engineering at UMBC (University of Maryland, Baltimore County). He was chairman of the Chemical and Biochemical Engineering Department, associate dean for the College of Engineering, and associate provost for Academic Affairs. Prior to UMBC, he spent nearly 10 years in industry, with senior management positions with International Flavors and Fragrances, Inc. and Schering-Plough Corp. (now Merck, Inc.). He has significant experience with R&D and commercialization of biopharmaceuticals. He holds a BS degree in Chemical Engineering from the University of Porto, Portugal, and MS and PhD degrees in Chemical and Biochemical Engineering from the University of Pennsylvania. He has an active research program in bioprocess engineering, is author or coauthor to more than 200 publications and presentations, has overseen more than $20 million in contracts and grants, and consults with various biotechnology and pharmaceutical companies. He received a NATO Senior Fellowship and the Parenteral Drug Association’s James Agalloco Award. He was founding president for the Chesapeake Bay Area Chapter of the International Society for Pharmaceutical Engineers, chair of the Council for Biotechnology Centers for BIO, and serves on various scientific advisory boards. He is a graduate of Leadership Maryland. He has been recognized with honorary international membership by The Brazilian Academy of Pharmaceutical Sciences and received the Order of Merit in Public Education by the President of Portugal.

Professor Bernard Grodzinski earned his BSc (Toronto), MSc and PhD (York University, 1974) before becoming a Postdoctoral Fellow of Botany, Oxford, UK. Between 1975–79 he was on faculty of the Botany School of the University of Cambridge. In 1979, he returned to Canada where he serves as Professor in the Department of Plant Agriculture of the University of Guelph. He is CoDirector of Guelph’s Closed Environment Systems Research Facility (CESRF) and also heads the Biotron’s Low-Temperature Research team an initiative of Guelph and Western Universities. The unique infrastructure developed at Guelph has helped foster strong collaborative research for colleagues, students and researchers in Canada and internationally. His primary interests remain understanding photosynthesis, respiration, translocation and crop productivity investigating plant growth and homeostasis in both the natural field and artificially controlled environments (CE) that include commercial greenhouses and specialized chambers being tested for manned space programs. His efforts have led to a better understanding of natural ecotype variation and phenotype responses to environmental stresses. Genetic approaches currently being pursued include selecting plants for better light interception and improved carbon and nitrogen metabolism that control source-sink development and enhance crop yield and quality.

Zhanfeng Cui has been the Donald Pollock Professor of Chemical Engineering, University of Oxford, since the Chair was established in 2000. He is the founding Director of the Oxford Centre for Tissue Engineering and Bioprocessing (OCTEB). He was educated in China and got his BSc degree from Inner Mongolia University of Technology (1982) and MSc (1984) and PhD (1987) from Dalian University of Technology. After a postdoctoral experience in Strathclyde University in Scotland, he joined Edinburgh University as a lecturer in Chemical Engineering (1991). He then held academic appointments at Oxford Engineering Science Department as University Lecturer (1994–98) and Reader (1999–2000). He was a Visiting Professor of Georgia Institute of Technology, USA (1999); the Braun Intertec Visiting Professor to the University of Minnesota, USA (2004); and a Chang-Jiang Visiting Professor to Dalian University of Technology, China (2005). He is a Chartered Engineer, a Chartered Scientist, and a Fellow of the Institution of Chemical Engineers. In 2009, he was awarded a Doctor of Science (DSc) by Oxford University to recognize his research achievement. In 2013, he is elected to Fellow of the Royal Academy of Engineering. His research interests include tissue engineering and stem cell technologies, bioseparation and bioprocessing, and membrane science and technology. He and his coworkers have published more than 120 articles in refereed journal papers and filed 7 patent applications in the last 5 years. He is the academic founder of Zyoxel Limited, an Oxford University spin-off in 2009.

Volume Editors

ix

Dr. Hua Ye is an Associate Professor in Engineering Science and Deputy Director of the CRMI Technology Centre for Regenerative Medicine in the Institute of Biomedical Engineering at the University of Oxford. Dr. Ye holds a degree in Chemical Engineering from Dalian University of Technology (1998), China, and a PhD in Biochemical Engineering from the University of Oxford (2005). She spent 1 year exploring the commercial value of her PhD project on hollow fibre membrane bioreactor as an enterprise fellow at Begbroke Science Park of the University of Oxford before taking up the post of postdoctoral research associate at Imperial College London in 2005. Her research interests lie in stem cell bioprocessing (expansion and differentiation) and biomaterials and bioreactors for tissue engineering. Dr. Ye has published more than 50 journal articles, 6 book chapters, and secured research funding of more than £11M both from the Research Councils UK and industry.

Spiros N. Agathos has been a Professor of Bioengineering (since 1993) at the Catholic University of Louvain, Earth and Life Institute, Belgium, and since 2015, he is the Inaugural Dean of Life Sciences and Biotechnology at Yachay Tech, the first research-intensive university in Ecuador. He obtained his Dipl. Eng. in Chemical Engineering from the National Technical University of Athens, his M. Eng. in the same field from McGill University, and his PhD in Biochemical Engineering from the Massachusetts Institute of Technology (MIT). Professor Agathos was a faculty member at the University of Western Ontario (1982–85) and at Rutgers University (1985–93) and served as a visiting professor in Europe and the Americas. He has published more than 200 articles, 4 books, and 4 patents. He has been Editor or Editorial Board member of many journals and on numerous committees for science and technology policy. He is a consultant to governments and industry, while his former students and postdocs have significant academic and industrial positions across the globe. Among his many awards, he is an elected fellow of the American Academy for Microbiology (AAM), the International Water Association (IWA), the American Institute for Medical and Biological Engineering (AIMBE), and the Society for Industrial Microbiology and Biotechnology (SIMB). His research interests include fungal, insect, and mammalian cell cultures in bioreactors, biocatalyst development, bioprocess optimisation, bioreactor design and scale-up, pollutant biodegradation and site bioremediation, microbial ecogenomics, and biotechnology for sustainability.

Ben A. Stenuit is Full Professor of Industrial and Environmental Biotechnology (since 2017) at Polytech Montpellier, University of Montpellier, Joint Research Unit of Agropolymer Engineering and Emerging Technologies (IATE, UMR 1208), Montpellier, France. He is also Invited Lecturer (from 2015) on biological treatment of wastewater at the Louvain School of Engineering, Catholic University of Louvain (UCL), Louvain-la-Neuve, Belgium. He obtained his PhD in 2009 in Agricultural Sciences and Bioengineering from the Faculty of Bioscience Engineering at the Catholic University of Louvain (with Prof. Spiros N. Agathos). From 2010 to 2013, he was a postdoctoral researcher at the University of California, Berkeley, at the Civil and Environmental Engineering Department (with Prof. Lisa Alvarez-Cohen). From 2013 to 2017, he was working as a postdoctoral researcher at UCL (Earth and Life Institute). To date, he has coauthored more than 25 peer-reviewed publications. His research interests include environmental molecular microbiology, molecular systems ecology, biorefinery, industrial and environmental biotechnology, and bioprocess design and modeling.

This page intentionally left blank

SECTION EDITORS Feng-Wu Bai received his BSc and MSc degrees from Dalian University of Technology, China, and PhD from the University of Waterloo, Canada. He has been a visiting professor at world-famous universities, including MIT, and a consultant for government and industry. Currently, he is a Biochemical Engineering Professor at Dalian University of Technology, China, and his research interest comprises the combination of chemical engineering principles with biotechnology advances for the production of biofuels, bioenergy, and bio-based chemicals at large scale as an alternative to petroleum-based products. To date, his academic achievement and technical innovation have produced 2 books, 3 invited book chapters, more than 120 peer-reviewed articles, and 2 patents that have been commercialized in fuel ethanol production. He is an editor of Biotechnology Advances and editorial board member of Journal of Biotechnology and Chinese Journal of Biotechnology.

Pavneesh Madan is an Assistant Professor in the Department of Biomedical Sciences, Ontario Veterinary College, the University of Guelph. He received his DVM and MVSc degrees from College of Veterinary Sciences, Hisar (India), and PhD in Animal Reproductive Biotechnology from the University of British Columbia (UBC), Vancouver, Canada. His research interest is in understanding cellular, molecular, and genetic mechanisms and regulating preimplantation embryo development and arrest.

Massimo Francesco Marcone earned his B.Sc., B.A. and Ph.D., from the University of Guelph and now serves as a full professor of Food Science in the Ontario Agriculture College at the University of Guelph in Ontario, Canada. His institution, as a whole, ranks as the fourth most comprehensive University in Canada, with his internationally renowned academic department ranking fourth in the world in the CWUR World University Rankings. During his academic career, he has specialized in the area of food analysis and holds an active research program with several M.Sc. and Ph.D. students. He has published over 100 peer-reviewed papers in international science journals leading him to be named a Fellow of the Royal Society for Chemistry in the United Kingdom. His opinions and works have made him highly sought after by the media, especially to determine between fact or fiction when it comes to a plethora of exotic foods and delicacies. His research work has allowed him to travel around the world with several documentary journalists examining many foods of public interest and leading to the publication of three creative non-fiction books on his research work. He teaches a variety of courses taken by approximately 1,600 undergraduate students a year. His teaching accomplishments have been well-recognized through numerous prestigious teaching awards received over the years including the OAC Distinguished Teaching Award, G.P. McRostie Faculty Award, OAC Distinguished Extension Award, and University of Guelph GSA Teaching Fellowship Excellence Award, as Professor of the Year. Quality teaching and engagement of students has always been what he has strived for and giving the “A, B, C” (most accurate, balanced, and current) perspective on his subject area.

xi

This page intentionally left blank

CONTRIBUTORS TO VOLUME 3 B Adamczyk Dublin-Oxford Glycobiology Laboratory, The National Institute for Bioprocessing Research and Training, Dublin, Ireland Jung Ho Ahn Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Republic of Korea AR Alcántara Complutense University of Madrid, Madrid, Spain H Bach University of British Columbia, Vancouver, BC, Canada RA Baffi BioMarin Pharmaceutical Inc., Novato, CA, United States Feng-Wu Bai Shanghai Jiao Tong University, School of Life Sciences and Biotechnology, Shanghai, China SP Banik Maulana Azad College, Kolkata, India Edward A Bayer Department of Biomolecular Sciences, The Weizmann Institute of Science, Rehovot, Israel D Beacom Cargill Incorporated, Minneapolis, MN, United States Yannick J Bomble Biosciences Center, National Renewable Energy Laboratory, Golden, CO, United States Trygve Brautaset SINTEF Materials and Chemistry, Trondheim, Norway K Brorson Center for Drug Evaluation and Research, Food and Drug Administration, Bethesda, MD, United States MP Campbell Dept. of Chemistry and Biomolecular Sciences, Macquarie University, Sydney, NSW, Australia

J-M Cappia Sartorius Stedim France SAS, Aubagne, France JOB Carioca Federal University of Ceará, Fortaleza-Ceará, Brazil; and Ceará State University, Fortaleza-Ceará, Brazil CT Carson BD Biosciences e Pharmingen, La Jolla, CA, United States F Chen The University of Hong Kong, Hong Kong, China; and Peking University, Beijing, China Guo-Qiang Chen Center for Synthetic and Systems Biology, School of Life Sciences, Tsinghua University, Beijing, China H Chen Merck & Co., Inc., Union, NJ, United States S Chowdhury Indian Institute of Chemical Biology (Unit of CSIR, Govt. of India), Kolkata, India A Clements-Egan Centocor Research & Development, Inc., Radnor, PA, United States R Cordoba-Rodriguez Food and Drug Administration, Bethesda, MD, United States Stephanie M Curley Colleges of Nanoscale Science & Engineering, State University of New York Polytechnic Institute, Albany, NY, United States J-Y Dai Dalian University of Technology, Dalian, China Jinwei Dao Center for Synthetic and Systems Biology, School of Life Sciences, Tsinghua University, Beijing, China Ångela de Carvalho River Stone Biotech Aps, Copenhagen, Denmark

xiii

xiv

Contributors to Volume 3

M De Jesus ExcellGene SA, Monthey, Switzerland AL Demain Drew University, Madison, NJ, United States M Doherty Dublin-Oxford Glycobiology Laboratory, The National Institute for Bioprocessing Research and Training, Dublin, Ireland Hongjun Dong Institute of Microbiology, Chinese Academy of Sciences, Beijing, China

GR Gunn, III Centocor Research & Development, Inc., Radnor, PA, United States DL Gutnick Tel Aviv University, Tel Aviv, Israel Patrick C Hallenbeck Life Sciences Research Center, Department of Biology, United States Air Force Academy, USAF Academy, Colorado, CO, United States; and Département de microbiologie, infectiologie et immunologie, Université de Montréal, Montréal, Montréal QC, Canada

Wei Du Department of Chemical Engineering, Tsinghua University, Beijing, China

Chuanshu He CAS Key Laboratory of Urban Pollutant Conversion, University of Science and Technology of China, Hefei, China

Lothar Eggeling IBG-1, Biotechnology, Research Centre Jülich, Jülich, Germany

MJ Hernáiz Complutense University of Madrid, Madrid, Spain

Trond E Ellingsen SINTEF Materials and Chemistry, Trondheim, Norway

Michael E Himmel Biosciences Center, National Renewable Energy Laboratory, Golden, CO, United States

AY Elliott Office of Assistant Secretary for Preparedness and Response (ASPR), Washington, DC, United States N Emre BD Biosciences e Pharmingen, La Jolla, CA, United States LE Erickson Kansas State University, Manhattan, KS, United States

D Hirsch Food and Drug Administration, Bethesda, MD, United States Nancy WY Ho Purdue University, School of Chemical Engineering & Laboratory of Renewable Resources Engineering, West Lafayette, IN, United States

TC Fong BD Biosciences, San Jose, CA, United States

H-P Hohmann Biotechnology R&D, DSM Nutritional Products, Basel, Switzerland

A Fosmer Cargill Incorporated, Minneapolis, MN, United States

S Hou University of Delaware, Newark, DE, United States

C García-Estrada NBIOTEC, Instituto de Biotecnología de León, León, Spain

P Hoyos Complutense University of Madrid, Madrid, Spain

S Ghorai Maulana Azad College, Kolkata, India

Ling Hua Group Biotechnology, Clariant Produkte (Deutschland) GMBH, Planegg, Germany

U Gottschalk Sartorius Stedim Biotech, Goettingen, Germany L Graham Food and Drug Administration, Bethesda, MD, United States Yang Gu Key Laboratory of Synthetic Biology, CAS Center for Excellence in Molecular Plant Sciences, Institute of Plant Physiology and Ecology, Chinese Academy of Sciences, Shanghai, China

H Huang Nanjing Normal University, Nanjing, China WMM Ingledew University of Saskatchewan, Saskatoon, SK, Canada; and Lallemand Ethanol Technology, Parksville, BC, Canada Carlos A Jerez University of Chile, Santiago, Chile

Contributors to Volume 3

xv

L Jiang Zhejiang University, Hangzhou, China; and South China University of Technology, Guangzhou, China

MRLV Leal Brazilian Bioethanol Science and Technology Laboratory (CTBE)/National Center for Research in Energy and Materials (CNPEM), Sao Paulo, Brazil

Weihong Jiang Key Laboratory of Synthetic Biology, CAS Center for Excellence in Molecular Plant Sciences, Institute of Plant Physiology and Ecology, Chinese Academy of Sciences, Shanghai, China

Jong An Lee Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Republic of Korea

MW Jornitz Sartorius Stedim North America Inc., Edgewood, NY, United States T Kaeding Technische Universität Hamburg-Harburg, Hamburg, Germany Amaranta Kahn Department of Biomolecular Sciences, The Weizmann Institute of Science, Rehovot, Israel AD Kaiser US Food and Drug Administration, Silver Spring, MD, United States Rasool Kamal Dalian Institute of Chemical Physics, Chinese Academy of Sciences, Dalian, China JJ Kattla Dublin-Oxford Glycobiology Laboratory, The National Institute for Bioprocessing Research and Training, Dublin, Ireland B Kelley Genentech, San Francisco, CA, United States S Kennett Food and Drug Administration, Bethesda, MD, United States S Khowala Indian Institute of Chemical Biology (Unit of CSIR, Govt. of India), Kolkata, India Kohtaro Kirimura Department of Applied Chemistry, Faculty of Science and Engineering, Waseda University, Tokyo, Japan D Laudert Biotechnology R&D, DSM Nutritional Products, Basel, Switzerland Carolina Zampol Lazaro Département de microbiologie, infectiologie et immunologie, Université de Montréal, Montréal, Montréal QC, Canada

K Lee Food and Drug Administration, Bethesda, MD, United States KH Lee University of Delaware, Newark, DE, United States Sang Yup Lee Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Republic of Korea SB Levy Paratek Pharmaceuticals, Inc., Boston, MA, United States; and Tufts University School of Medicine, Boston, MA, United States Yin Li Institute of Microbiology, Chinese Academy of Sciences, Beijing, China Y-H Lin University of Saskatchewan, Saskatoon, SK, Canada Chen-Guang Liu State Key Laboratory of Microbial Metabolism, and School of Life Sciences & Biotechnology, Shanghai Jiao Tong University, Shanghai, China J Liu The University of Hong Kong, Hong Kong, China Xiaofeng Liu Chengdu Institute of Biology, Chinese Academy of Sciences (CIB, CAS), Chengdu, China FX Malcata LEPABE, Laboratory of Engineering of Processes, Environment, Biotechnology and Energy, University of Porto, Porto, Portugal Sana Malik Department of Bioinformatics & Biotechnology, Government College University Faisalabad, Faisalabad, Pakistan GL Marcone University of Insubria, Varese, Italy F Marinelli University of Insubria, Varese, Italy

xvi

Contributors to Volume 3

J-F Martín Universidad de León, León, Spain C McIntyre BD Biosciences, San Jose, CA, United States NL McKnight Genentech, San Francisco, CA, United States T McMullin Cargill Incorporated, Minneapolis, MN, United States Muhammad Aamer Mehmood Department of Bioinformatics & Biotechnology, Government College University Faisalabad, Faisalabad, Pakistan and State Key Laboratory of Microbial Metabolism, and School of Life Sciences & Biotechnology, Shanghai Jiao Tong University, Shanghai, China

E Pungor BioMarin Pharmaceutical Inc., Novato, CA, United States H Quitmann Hamburg University of Technology, Hamburg, Germany OL Ramos CEB, Centre of Biological Engineering, University of Minho, Braga, Portugal; and LEPABE, Laboratory of Engineering of Processes, Environment, Biotechnology and Energy, University of Porto, Porto, Portugal G Rao University of Maryland Baltimore County, Baltimore, MD, United States AS Rathore Indian Institute of Technology, New Delhi, India

JC Menezes Technical University of Lisbon, Lisbon, Portugal

R Rawat Food and Drug Administration, Bethesda, MD, United States

C Miller Cargill Incorporated, Minneapolis, MN, United States

DH Reifsnyder Genentech, San Francisco, CA, United States

N Moazami Institute of Biotechnology, Iranian Research Organization for Science & Technology, Tehran, Iran

BL Rellahan Food and Drug Administration, Bethesda, MD, United States

A Moreira University of Maryland Baltimore County, Baltimore, MD, United States Yang Mu CAS Key Laboratory of Urban Pollutant Conversion, University of Science and Technology of China, Hefei, China S Mukherjee Indian Institute of Chemical Biology (Unit of CSIR, Govt. of India), Kolkata, India ML Nelson Paratek Pharmaceuticals, Inc., Boston, MA, United States Sarah E Nicoletti Colleges of Nanoscale Science & Engineering, State University of New York Polytechnic Institute, Albany, NY, United States Filipa Pereira European Molecular Biology Laboratory, Structural and Computational Biology Unit, Heidelberg, Germany WS Prince BioMarin Pharmaceutical Inc., Novato, CA, United States

NM Ritter Biologics Consulting Group, Inc., Alexandria, VA, United States PM Rudd Dublin-Oxford Glycobiology Laboratory, The National Institute for Bioprocessing Research and Training, Dublin, Ireland B Rush Cargill Incorporated, Minneapolis, MN, United States W Sabra Hamburg University of Technology, Hamburg, Germany Emrah Sagır Department of Biology, Faculty of Arts and Sciences, Osmaniye Korkut Ata University, Osmaniye, Turkey H Sahm Research Center Jülich, Jülich, Germany R Saldova Dublin-Oxford Glycobiology Laboratory, The National Institute for Bioprocessing Research and Training, Dublin, Ireland Nicholas S Sarai Biosciences Center, National Renewable Energy Laboratory, Golden, CO, United States

Contributors to Volume 3

G Shankar Centocor Research & Development, Inc., Radnor, PA, United States Susan T Sharfstein Colleges of Nanoscale Science & Engineering, State University of New York Polytechnic Institute, Albany, NY, United States Z Shi Zhejiang University, Hangzhou, China DM Shlaes Anti-infectives Consulting, LLC, Stonington, CT, United States V Sluzky BioMarin Pharmaceutical Inc., Novato, CA, United States WB Struwe Dublin-Oxford Glycobiology Laboratory, The National Institute for Bioprocessing Research and Training, Dublin, Ireland P Suominen Cargill Incorporated, Minneapolis, MN, United States J Swisher Center for Drug Evaluation and Research, Food and Drug Administration, Bethesda, MD, United States

Ying Wang Department of Biochemical Engineering, School of Chemistry and Chemical Engineering, Beijing Institute of Technology, Beijing, China OP Ward University of Waterloo, Waterloo, ON, Canada Z Wen Iowa State University, Ames, IA, United States Qiaqing Wu National Engineering Laboratory for Industrial Enzymes, Tianjin Engineering Research Center of Biocatalytic Technology, Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences, Tianjin, China Youduo Wu School of Life Science and Biotechnology, Dalian University of Technology, Dalian, China DM Wuest University of Delaware, Newark, DE, United States FM Wurm Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland; and ExcellGene SA, Monthey, Switzerland Z-L Xiu Dalian University of Technology, Dalian, China

Adam Takos SNIPR Biome Aps, Copenhagen, Denmark

Z Xu Zhejiang University, Hangzhou, China

WE Tente Humacyte, Inc., Durham, NC, United States

Chuang Xue School of Life Science and Biotechnology, Dalian University of Technology, Dalian, China

Sebastian Theobald DTU Biosustain, Novo Nordisk Foundation Center for Biosustainability, Kongens Lyngby, Denmark D Tjahjasari Technische Universität Hamburg-Harburg, Hamburg, Germany P Vaishnav GIDC, Ankleshwar, Gujarat, India JR Vallejos University of Maryland Baltimore County, Baltimore, MD, United States Katherina Garcia Vanegas DTU Bioengineering, Department of Biotechnology and Biomedicine, Technical University of Denmark, Kongens Lyngby, Denmark D Verma Indian Institute of Chemical Biology (Unit of CSIR, Govt. of India), Kolkata, India

xvii

Zhengying Yan Chengdu Institute of Biology, Chinese Academy of Sciences (CIB, CAS), Chengdu, China ST Yang The Ohio State University, Columbus, OH, United States Shihui Yang Hubei University, College of Life Sciences, Wuhan, China Isato Yoshioka Department of Applied Chemistry, Faculty of Science and Engineering, Waseda University, Tokyo, Japan Zhengbo Yue Hefei University of Technology, Hefei, China James F Zawada Sutro Biopharma, Inc., South San Francisco, CA, United States

xviii

Contributors to Volume 3

A-P Zeng Technische Universität Hamburg-Harburg, Hamburg, Germany B Zhang The Ohio State University, Columbus, OH, United States K Zhang The Ohio State University, Columbus, OH, United States Yanping Zhang Institute of Microbiology, Chinese Academy of Sciences, Beijing, China Chunhua Zhao Institute of Microbiology, Chinese Academy of Sciences, Beijing, China

Xin-Qing Zhao State Key Laboratory of Microbial Metabolism, and School of Life Sciences & Biotechnology, Shanghai Jiao Tong University, Shanghai, China Zongbao K Zhao Dalian Institute of Chemical Physics, Chinese Academy of Sciences, Dalian, China J-J Zhong Shanghai Jiao Tong University, Shanghai, China Dunming Zhu National Engineering Laboratory for Industrial Enzymes, Tianjin Engineering Research Center of Biocatalytic Technology, Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences, Tianjin, China L Zhu Zhejiang University, Hangzhou, China

CONTENTS OF VOLUME 3 Editor in Chief

v

Volume Editors

vii

Section Editors

xi

Contributors to Volume 3 Foreword Preface

xiii xxiii xxv

3.01

Industrial Enzymes Dunming Zhu, Qiaqing Wu, and Ling Hua

1

3.02

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems Nicholas S Sarai, Michael E Himmel, Yannick J Bomble, Amaranta Kahn, and Edward A Bayer

14

3.03

Ethanol Production From Sugar-Based Feedstocks JOB Carioca and MRLV Leal

24

3.04

Ethanol From Starch-Based Feedstocks WMM Ingledew and Y-H Lin

35

3.05

Fuel Ethanol Production From Lignocellulosic Biomass Feng-Wu Bai, Shihui Yang, and Nancy WY Ho

49

3.06

Biodiesel Wei Du, Rasool Kamal, and Zongbao K Zhao

66

3.07

Biofuels and Bioenergy: Acetone and Butanol Chuang Xue, Youduo Wu, Yang Gu, Weihong Jiang, Hongjun Dong, Yanping Zhang, Chunhua Zhao, and Yin Li

79

3.08

Long-Chain Liquid Biofuels Sana Malik, Chen-Guang Liu, Xin-Qing Zhao, and Muhammad Aamer Mehmood

101

3.09

Biogas Chuanshu He, Yang Mu, Xiaofeng Liu, Zhengying Yan, and Zhengbo Yue

110

3.10

Biohydrogen Patrick C Hallenbeck, Carolina Zampol Lazaro, and Emrah Sagır

128

3.11

Biofuel From Microalgae Z Wen, J Liu, and F Chen

140

xix

xx

Contents of Volume 3

3.12

Microbial Production of 2,3-Butanediol W Sabra, H Quitmann, A-P Zeng, J-Y Dai, and Z-L Xiu

147

3.13

Citric Acid Kohtaro Kirimura and Isato Yoshioka

158

3.14

Gluconic and Itaconic Acids Kohtaro Kirimura and Isato Yoshioka

166

3.15

Organic Acids: Succinic and Malic Acids Jong An Lee, Jung Ho Ahn, and Sang Yup Lee

172

3.16

Fumaric Acid ST Yang, K Zhang, B Zhang, and H Huang

188

3.17

Industrial Production of Lactic Acid C Miller, A Fosmer, B Rush, T McMullin, D Beacom, and P Suominen

208

3.18

Acetic and Propionic Acids Z Xu, Z Shi, and L Jiang

218

3.19

Acrylic Acid Z Xu, L Zhu, and H Chen

229

3.20

Butyric Acid Z Xu and L Jiang

235

3.21

Polyhydroxyalkanoate/Polyhydroxybutyrate Ying Wang, Jinwei Dao, and Guo-Qiang Chen

244

3.22

1,3-Propanediol and Polytrimethyleneterephthalate D Tjahjasari, T Kaeding, and A-P Zeng

258

3.23

Antibiotics: The Miracle Menaced DM Shlaes

271

3.24

Penicillins and Cephalosporins C García-Estrada and J-F Martín

283

3.25

Tetracyclines and Tetracycline Derivatives ML Nelson and SB Levy

297

3.26

Microbial Secondary Metabolites F Marinelli and GL Marcone

312

3.27

Plant Secondary Metabolites J-J Zhong

324

3.28

Biocatalyzed Production of Fine Chemicals P Hoyos, MJ Hernáiz, and AR Alcántara

334

3.29

Production of Recombinant Proteins by Microbes and Higher Organisms AL Demain and P Vaishnav

374

3.30

Vaccines AY Elliott

387

3.31

Manufacturing Recombinant Proteins in kg-ton Quantities Using Animal Cells in Bioreactors M De Jesus and FM Wurm

396

Contents of Volume 3

xxi

3.32

Recent and Emerging Trends and Concerns Related to the Manufacturing and Testing of Monoclonal Antibodies Intended for Clinical Use BL Rellahan, L Graham, D Hirsch, S Kennett, R Rawat, K Brorson, and J Swisher

402

3.33

Therapeutic Enzymes and Biomimetic Substrates: A Case Study of Recombinant Human Arylsulfatase B (NaglazymeÒ) Substrate Selection and Application WS Prince, E Pungor, V Sluzky, and RA Baffi

414

3.34

Cell-Free Production of Pharmaceutical Proteins James F Zawada

3.35

Combination Products Are Not Solely Biological Products, Drugs, or Devices: A Regulatory Perspective AD Kaiser

3.36

Cellular Therapies CT Carson, N Emre, C McIntyre, and TC Fong

446

3.37

Gene Therapies WE Tente

460

3.38

Regulatory Aspects of Chemistry Manufacturing and Controls for Investigational New Drug Applications and Biologic License Applications to the United States Food and Drug Administration K Lee

427

435

473

3.39

Raw Materials in the Manufacture of Biotechnology Products R Cordoba-Rodriguez

487

3.40

Characterization of Biotechnological/Biological/Biosimilar Products NM Ritter

493

3.41

Protein Glycosylation JJ Kattla, WB Struwe, M Doherty, B Adamczyk, R Saldova, PM Rudd, and MP Campbell

501

3.42

Immunogenicity Assay Development and Validation A Clements-Egan, GR Gunn, III, and G Shankar

521

3.43

Process Analytical Technology in Bioprocess Development and Manufacturing JC Menezes

535

3.44

Process Validation DH Reifsnyder, NL McKnight, and B Kelley

545

3.45

Follow-on Protein Products: Scientific Issues, Developments, and Challenges AS Rathore

554

3.46

Amino Acid Production L Eggeling and H Sahm

563

3.47

Lysine Industrial Uses and Production Trygve Brautaset, Trond E Ellingsen, and Lothar Eggeling

572

3.48

Food-Grade Enzymes OL Ramos and FX Malcata

587

3.49

Proteases OP Ward

604

xxii

Contents of Volume 3

3.50

Application of Enzymes and Microbes for the Industrial Production of Vitamins and Vitamin-Like Compounds D Laudert and H-P Hohmann

3.51

Fungal Biotechnology in Food and Feed Processing S Ghorai, SP Banik, D Verma, S Chowdhury, S Mukherjee, and S Khowala

635

3.52

Metabolic Engineering DM Wuest, S Hou, and KH Lee

647

3.53

Synthetic Biology: An Overview Ångela de Carvalho, Katherina Garcia Vanegas, Filipa Pereira, Sebastian Theobald, and Adam Takos

659

3.54

Industrial Biotechnology and Commodity Products: Single-Use Technologies for Biomanufacturing MW Jornitz, J-M Cappia, and G Rao

3.55

Bioreactors for Commodity Products LE Erickson

3.56

Integrating Process Scouting Devices With Bench-Scale Devices: Challenges and Opportunities for Mammalian Cell Culture JR Vallejos, A Moreira, G Rao, and K Brorson

3.57

Overview of Downstream Processing in the Biomanufacturing Industry U Gottschalk

698

3.58

Applications of Nanotechnology to Bioprocessing Stephanie M Curley, Sarah E Nicoletti, and Susan T Sharfstein

712

3.59

Biosurfactants DL Gutnick and H Bach

731

3.60

Bioleaching and Biomining for the Industrial Recovery of Metals Carlos A Jerez

758

3.61

Biological Control N Moazami

772

616

671 683

690

FOREWORD For a long time, biotechnology has been the young stepson of basic biomedical sciences, identified mostly with the food industry and a bit later with development of vaccines. It was not highly regarded by basic researchers who focused on decipheringdamong other problemsdmetabolic pathways, the genetic code, and the underlying pathogenetic mechanisms of diseases. Drug discovery at the beginning of the century was incidental (e.g., aspirin, penicillin, and even insulin), and the idea of curing a genetic disease was not even a dreamdthey were regarded irreversible. Nobody has dreamt on the possibility of genetic manipulation. This landscape has since changed dramatically, and biotechnology has become the respected inseparable Siamese twin of basic biomedical sciences. Efforts to translate fundamental discoveries into useful products are recognized now as an important and integral hallmark in both academia and industry, and the traditional border between the two has lost its sharp lines of demarcation. In many places around the world, we witness a process where the biotechnological industry, and mostly its research and development branches, is growing and developing around universitiesdwhich serve as nuclei of crystallization for their flourishing, generating a new type of collaborative ecosystem we have not known before. MIT and the pharmaceutical industry in Cambridge, Massachusetts, USA, are probably prime examples of this process. The border between science and its technological applications has almost disappeared. The fences among faculties of biomedical basic sciences, medicine, biotechnology, chemical engineering, chemistry, physics, and civil engineering, among many others, have been lowered, and they hire these daysdmore and more frequentlydscientists with similar profilesdactually competing with one another. Industry is investing more and more efforts in research and development, and has tightened its ties with academia. Consequently, it is employing leading and world-renowned scientists that academia would have been proud to have, and is using state-of-the-art, most advanced, and innovative technologies that are far from being just scaled up pilot processes developed in academia. Working nowadays in the biotechnological industry is no longer regarded as a second choice for academics, and the movement between the two enterprises has become bilateral compared to the unilaterality that domineered the relationship between the two for a long time: experience in industry is regarded nowadays by many in academia as an advantage. Not surprisingly, even the basic nomenclature is changing, and biotechnology has lost its defined traditional boundaries and merged into the actively evolving conglomerate of biomedical sciences, along with cell and molecular biology, immunology, biochemistry, genetics, agriculture, biomedical engineering, and pharmaceutical sciences, among other areas. What are the roots of this revolution? When did it all start? Like many other revolutions, including political ones, we can easily identify points of time when they erupted, but find it more difficult to identify the underlying streams and developments that led to the eruption. It is widely accepteddI thinkdthat the seeds of the revolution were planted with the deciphering of the mechanisms that underlie the central dogma of biologydthe discovery of the double helical structure of DNA which immediately disclosed its mechanism of replication, and then the discovery of mRNA, the specific tRNAs, and the mechanisms of protein translation. These discoveries that were made in the 1950s led to the development of an avalanche of subsequent technologies which enabled us to sequence and then manipulate genes, express or silence them in different organismsdfrom viruses to mammalsdto analyze the other elements in the central dogma that are located hierarchically above DNA (RNA, proteins, small moleculesde.g., the different omics). Last but not least, they enabled us to edit genes (CRISPR-Cas) and to generate proteins that are better in many aspects than the natural ones (e.g., synthetic biology). There is currently no field in biomedical and life sciences that is not deeply involved with biotechnology: from the development of vaccines and small molecules, through drug targeting in tissue-specific liposomes or

xxiii

xxiv

Foreword

nanoparticles; from the development of insects- or fungi-resistant plants to animals that were “cleaned” from viral-hosting genes and that have humanized HLA system (and can serve therefore as an unlimited source of organs for transplantation); from yeast that were engineered to generate antimalarial drugs (e.g., artemisinin), to microorganisms that can make cannabinoids or odd chain fatty acids that are rare in nature; from mammals that secrete protein hormones and/or antibodies in their milk or to their circulation, to microorganisms that dissolve biofilms or clean water from biological or chemical contaminantsdthe list in nonexhaustive. The newly evolving biotechnology will revolutionize our lives in every aspectdfrom preventive and therapeutic medicine, to agriculture, to biomanufacturing and the environment. The third edition of Comprehensive Biotechnology tries to encompass all these developments and in particular the more recent ones. However, due to the exponential growth of the field, the task is daunting, and is becoming more and more difficult. Can we prophesize that future editions will have to devote a separate volume to each of the numerous subfields of what used to be “classical” biotechnologydone devoted to development of vaccines, the other to microorganisms that produce small molecules, and another one to engineered plants that can grow without using insecticides or fungicides. When this will happen, we shall be able to say that the revolution of biotechnology is completed, no need to see it as a separate area.

Aaron Ciechanover The Rappaport Faculty of Medicine and Research Institute Technion Israel Institute of Technology Haifa, Israel

PREFACE Increasingly the life sciences area is impacting virtually all technologies and vice versa; hence, the ongoing importance of biotechnology worldwide. My fellow editors and I believe that it is time to produce a third edition of Comprehensive Biotechnology to bring up to date this unique “one-stop-shopping” authorized reference. It covers the comprehensive knowledge base of this multidisciplinary field and can be readily be accessed online by neophytes as well as veterans in the field, serving a wide range of stake-holders, via the prestigious publisher, Elsevier. This third edition of Comprehensive Biotechnology follows the tradition of the previous two editions: it covers the biotechnology field comprehensively, in six volumes, and is edited by internationally renown biotechnologists for the related subdisciplines. We continue to be responsible for the development of this major reference work (MRW) as the primary source of information with a range of wide value and interest: for teachers, researchers, and administrators in academia, industry, and government. The volume editors, associate editors, and section editors have relevant expertise to handle all essential constituent components of the multiauthored contents indicated below. Since the publication of the second edition, most of the foundation knowledge base has changed to the extent that new material could be incorporated by the expedient of postscripts to several chapters. However, other topics needed new or completely revised chapters. Notably among these are those which dealt with the recent innovations in gene editing, artificial intelligence, digital impaction, renewable resources, and climate change on the field. Comprehensive Biotechnology 3rd edition is published as an electronic online publication, with the following six volumes. Vol 1: Scientific Fundamentals of Biotechnology, edited by Prof. Michael Butler. Vol 2: Engineering Perspectives in Biotechnology, edited by Prof. Colin Webb. Vol 3: Industrial Biotechnology and Commodity Products, coedited by Prof. Antonio Moreira and Prof. Fengwu Bai. Vol 4: Agricultural and Related Biotechnologies, coedited by Prof. Bernard Grodzinski, Prof. Pavneesh Madan and Prof. Massimo Marcone. Vol 5: Medical Biotechnology and Healthcare, coedited by Prof. Zhanfeng Cui and Prof. Hua Ye. Vol 6: Environmental and Related Biotechnologies, coedited by Prof. Spiros Agathos and Prof. Benoit Stenuit Obviously Comprehensvie Biotechnology, third edition, would not have been possible without the combined expertise of the collective editors with their extensive professional connections. As with the previous two editions, this edition provides the following important points.

• All six volumes are published at the same time online, not as a series; this is not a conventional encyclopedia

but a symbiotic integration of brief articles on established topics and longer chapters on new or emerging areas. • Hyperlinks provide sources of extensive additional, related information instantaneously; material authored and edited by world-renown experts in all aspects of the broad multidisciplinary field of biotechnology. • Scope and nature of the work are vetted by a prestigious International Advisory Board including three Nobel laureates. • Most chapters carry a glossary and a professional summary of the authors indicating their appropriate credentials.

xxv

xxvi

Preface

• An extensive index for the entire publication gives a complete list of the many topics treated in the increasingly expanding field.

• To facilitate the one-stop shopping for rapid customer service, the work is designed as an integration of six

mini-encyclopedias. The first two ones provide the bases of the scientific and engineering principles of biotechnology, and the remaining four volumes treat the four major branches that impact on industry, medicine, agriculture, and the environment. Each volume has several sections which addresses the various aspects of the given branch. For convenience each section identifies the relevant topics in alphabetical order. To accomplish the work with appropriate expertise, many authors have contributed to CB3.

As before, this edition is blessed by a contribution of a Nobel laureate; this time, Prof. Aaron Ciechanover (Technion, Haifa); for Comprehensive Biotechnology, 2nd edition, Prof. Werner Arber (U Basel, Switzerland); for Comprehensive Biotechnology, 1st edition, Prof. Don Glaser (UC, Berkeley, US). Murray Moo-Young, PhD, FAIMBE, FRSC Comprehensive Biotechnology, 3rd edition Editor-in-Chief Distinguished Professor Emeritus, Department of Chemical Engineering Advisory board, Centre for Bioengineering and Biotechnology University of Waterloo, Ontario, Canada Postscript As of possible interest are reviews of the first and second editions of Comprehensive Biotechnology It would be very difficult to thoroughly cover the scope of biotechnology in one book. But this new edition of Comprehensive Biotechnology (1st ed., 1989) accomplishes what the title claims; the six-volume set provides detailed synopses of the applications, instruments, methodologies, and principles of modern biotechnology. Volume 1 provides the science background needed to understand biotechnology; it covers the essential biochemistry, biology, biophysics, chemistry, and computer science used in biotechnology applications and research. An explanation of engineering principles relevant to biotechnology follows in volume 2. The authors focus on engineering concepts appropriate to biotechnology product manufacturing. The third volume builds on the first two volumes in its coverage of biotechnology applications in industry and commodity products, including coverage of food ingredients, clinical products, and specialty chemicals. Summing Up: Highly recommended. Lower-division undergraduates through professionals. CHOICE

Review of the First Edition Murray Moo-Young and his colleagues have brought off a notable success in producing this work... Comprehensive Biotechnology will be an essential purchase for all departments and institutions, academic or industrial, that claim an interest in any aspect of the ill-defined field popularly known as biotechnology. Nature, Volume 321 (1986).

PERMISSIONS ACKNOWLEDGEMENT The following material is reproduced with kind permission of Oxford University Press Figure 3. Sulfur Metabolism in Plants and Related Biotechnologies. www.oup.com The following material is reproduced with kind permission of American Association for the Advancement of Science Figure 5a. Microfluidic Technology and Its Biological Application. www.aaas.org The following material is reproduced with kind permission of Nature Publishing Group Figure 10. Microbial Growth Dynamic Table 4. Microbial Growth Dynamic Figure 4. Integrated Production and Separation. http://www.nature.com

i

Industrial Enzymesq

3.01

Dunming Zhu and Qiaqing Wu, National Engineering Laboratory for Industrial Enzymes, Tianjin Engineering Research Center of Biocatalytic Technology, Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences, Tianjin, China Ling Hua, Group Biotechnology, Clariant Produkte (Deutschland) GMBH, Planegg, Germany © 2019 Elsevier B.V. All rights reserved. This is an update of D. Zhu, Q. Wu, N. Wang, 3.02 - Industrial Enzymes, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 3-13.

3.01.1 3.01.2 3.01.3 3.01.4 3.01.5 3.01.6 3.01.7 3.01.8 3.01.9 3.01.10 3.01.11 References

Introduction Protease Lipase Amylase Pullulanase Pectinase Xylanase Laccase Transglutaminase Phytase Perspective

1 2 4 5 7 7 8 9 10 10 11 11

Glossary Bioavailability A measurement of the extent to which a nutrient can be used by a human being or an animal. Carbohydrase An enzyme that acts on a carbohydrate, and either promotes the synthesis of carbohydrates or hydrolysis of carbohydrates into disaccharide molecules. Dynamic kinetic resolution A special case of kinetic resolution where both (reactant) enantiomers engage in a chemical equilibrium and exchange, thus it is possible to convert the achiral reactant with 100% completion. This is called dynamic kinetic resolution. Enantiomeric excess ee For a mixture of (R)- and (S)-enantiomers, with composition given as the mole or weight fractions F(R) and F(S) (where F(R)þF(S)¼1), the enantiomeric excess is defined as jF(R)-F(S) j (and the percent enantiomer excess by 100 times jF(R)-F(S) j). Frequently this term is abbreviated as ee. Hydrolase An enzyme that catalyzes the hydrolysis of a chemical bond, and is classified as EC 3 in the EC number classification of enzymes. Industrial enzyme An enzyme that is used to facilitate industrial processes or the production of industrial products. Kinetic resolution Two enantiomers show different reaction rates in a chemical reaction, thereby creating an excess of the less reactive enantiomer. This phenomenon is called kinetic resolution. Metagenomics The application of modern genomics techniques to the study of communities of microbial organisms directly in their natural environments, bypassing the need for isolation and lab cultivation of individual species. pH optima The pH at which the enzyme exhibits highest activity. Substrate specificity A characteristic feature of enzyme activity in relation to the kind of substrate on which the enzyme or catalytic molecule acts. Temperature optima The temperature at which the enzyme exhibits highest activity.

3.01.1

Introduction

Enzymes are proteins that catalyze chemical reactions. As such, enzymes have been widely used to facilitate industrial processes and the production of products, and these enzymes are referenced as industrial enzymes. Although it has dated back to the ancient times when enzymes were used in baking, brewing, cheese making etc., they were used as either fermenting microorganisms or crude

q

Change History: October 2018. Z. Dhu, made the changes in entire sections of the chapter, Fig. 1 changed with updated information, and some new references added.

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00148-8

1

2

Industrial Enzymes

Figure 1

The projected enzyme markets in 2024 based on application sectors.

preparations such as calves’ rumen or papaya fruit. Only during the last few decades, the development in recombinant DNA technology and advanced bioprocesses made it possible to produce and purify enzymes on a large scale, allowing the wide application of enzymes in various industrial products and processes, such as chemical, detergent, textile, food, animal feed, leather, pulp and paper industries. The latest developments in protein engineering and site-directed evolution have enabled us to create novel enzymes with new activities and/or for new processes.1,2 This makes industrial enzymes accessible to diverse industries that infiltrate all aspects of our daily life.3–5 As predicted in a report by Applied Market Research, the global market for industrial enzymes was $7.1 billion in 2017, and will reach $10.5 billion by 2024 at a compound annual growth rate (CAGR) of 5.7% from 2018 to 2024. The estimated enzyme markets in 2024 based on application sectors is shown in Fig. 1. A continuous stream of innovative enzyme products and their new and emerging applications are driving the enzyme industry to an unprecedented expanding era. Among the currently used industrial enzymes, hydrolases, including proteases and lipases, remain the dominant enzyme types, which are extensively used in the detergent, dairy and chemical industries. Various carbohydrases, primarily amylases and cellulases, represent the second largest group.3,4,6 Table 1 lists the enzymes and the industries where they find valuable applications. An enzyme may be used in various industries, while several enzymes are often needed in the same application for high efficiency. This review will cover the major industrial enzyme types and their various applications, excluding cellulases that have been addressed specifically elsewhere in this volume, and organized by enzyme types. The application of enzymes in various industry has resulted in significant savings in resources such as raw materials and water consumption, and the improvement of energy efficacy. This greatly benefits the industry in question and the environment and will continue to play an important role in improving the sustainability of our society and the quality of our life for the coming generations.

3.01.2

Protease

Proteases, also known as proteinases or proteolytic enzymes, are a large group of enzymes that catalyze the hydrolysis of peptide bonds in proteins and polypeptides. They differ in properties such as substrate specificity, active site and catalytic mechanism, pH and temperature optima and stability profile. There are several schemes for classifying proteases, which provide a wealth of relevant information about each protease. According to the Enzyme Commission (EC) classification, proteases belong to hydrolases (group 3), which hydrolyze peptide bonds (sub-group 4). Proteases can be classified into exopeptidases and endopeptidases, in which the former cleave N- or C-terminal peptide bonds and the latter break internal peptide bonds. Among them, endopeptidases find more commercial applications than exopeptidases. Based on proteolytic mechanism, proteases are divided into six broad groups: serine proteases, threonine proteases, cysteine proteases, aspartic proteases, metalloproteases and glutamic acid proteases. Alternatively, proteases may be classified into acidic, neutral and alkaline (basic) proteases according to the optimal pH. The proteases with pH optima in the range of 2.0–5.0 are called acid proteases, proteases having pH optima around 7.0 are neutral proteases, and alkaline proteases have pH optima in the range of 8.0–11.0. Acidic proteases are mainly fungal in origin, while neutral proteases are mainly of plant origin, and some bacteria and fungi also produce neutral proteases. Some of the industrially important alkaline proteases are those from Bacillus and Streptomyces species. There are thousands of different protease molecules that have been isolated and characterized. Among them, several hundred proteases are commercially relevant, and have been used in laundry and dishwashing detergents, food processing, animal feed additives, leather processing, waste treatment, pharmacology and drug manufacture. For applications in the biopharmaceutical industry, proteases are usually produced in small amounts with high purity, they thus require large-scale purification processes. For the food, detergents and other industries, proteases are used as crude preparations and should be manufactured in large quantities at low cost. Proteases are one of the three largest groups of industrial enzymes and account for about 60% of total global enzyme sales.7,8 The main industrial application of proteases is their use as detergent additives to remove protein deposits and stains, and the major player is subtilisins.9–11 They are used in the detergents for dish washers, laundry detergents as well as in laundry bleach additives. The subtilisin is very effective and used at very low concentration in detergent products. The typical concentration ranges between 0.007% and 0.1%, depending on the type of product. In addition to detergent, subtilisins are also used in the textile and cosmetics industry and in other technical applications such as protein hydrolysate production and leather treatment.

Industrial Enzymes Table 1

3

Industrial enzymes and their valuable applications

Enzyme

Substrate

Industry

Protease

Protein, polypeptide

Lipase

Oil, fat

Amylase

Carbohydrate

Pullulanase

Polysaccharide

Pectinase

Pectin

Xylanase

Xylan

Laccase

Benzenediol

Transglutaminase

Protein/amine

Phytase

Phytate

Laundry and dishwashing Detergents Food processing Animal feed additives Leather processing Waste treatment Pharmacology and drug manufacture Fat and oil Detergents Food and baking Pulp and paper Fine chemicals Materials Starch and fuel Baking Detergents Textile Pulp and paper Beverage Starch Food Beverage Textile Food Animal feed Pulp and paper Baking and food Animal feed Pulp and paper Textile Food Bioremediation Food Textile Animal feed Food

According to the catalytic mechanism, subtilisins can be defined as serine proteases. Aspartic acid, histidine and serine form the catalytic triad of subtilisins, but the amino acid sequences and three dimensional structures of subtilisins are apparently different from those of the other serine proteases, such as chymotrypsin and carboxypeptidase. The molecular weight of subtilisins varies greatly from 18 to 90 kDa, but all the subtilisins used in detergents consist of 269–275 amino acids, with an average molecular weight of 27 kDa. The pH range of subtilisins is from pH 6 to 11, with an optimal pH range between 9 and 11. Currently, the alkaline subtilisins from Bacillus species are the major proteases used in the detergent industry.8,12 The following factors may explain this situation. Firstly, the relatively low substrate specificity and high stability of subtilisins make themselves excellent candidates for detergent additives; Secondly, their production as extracellular enzymes greatly simplifies the purification of the enzyme from the fermentation broth and facilitates other downstream processing steps. In addition, the ability of Bacillus strains to produce enzymes over a very short period warrants a high production efficiency.13 Over the past 20 years, enormous efforts have been made to develop new and improved proteases for use in detergents. The protein engineering of known subtilisins and searching the metagenome for new subtilisins are two widely used approaches.14,15 When bleach-containing products are used in the cleaning process, hydrogen peroxide and peroxo acids are usually generated and oxidize certain methionine residues to sulfoxides, which was known to be responsible for the inactivation by hydrogen peroxide. The first genetically engineered subtilisin was reported in 1985 to address the sensitivity of subtilisin to oxidation by peroxide.16 Since then, most of the amino acid residues of subtilisin have been modified either by site-directed mutagenesis or by random mutagenesis. Gene shuffling has also been performed with subtilisins to improve their properties such as stability in organic solvents, at high temperature and high or low pH. In addition to classical microbiological screening methods, the exploitation of genomic data and metagenomic screening methods have also been used for search of new proteases. This enables us not only to develop subtilisins with better performance, but also to find completely new protease backbones. Proteases are a powerful tool for modifying the properties of food proteins and producing bioactive peptides from proteins. They are widely used in the production of value-added food ingredients and food processing for improving the functional, nutritional

4

Industrial Enzymes

and flavor properties of proteins.17,18 The function of proteases is to catalyze the hydrolysis of proteins, which has been exploited for the production of high-value protein hydrolysates from different sources of proteins such as casein, whey, soy protein and fish meat. These protein hydrolysates exhibit antioxidant, antithrombotic, antihypertensive, anticarcinogenic, satiety regulator or immunomodulatory properties and have a variety of applications in infant food formulations, specific therapeutic food products, fortifying fruit juices and soft drinks, other functional food additives and animal feed. The most significant property of acidic proteases is the ability to coagulate proteins. Microbial acidic proteases have largely replaced the calf enzyme (rennet) in the dairy industry for their ability to coagulate milk protein (casein) to form curds from which cheese is prepared. A protease from Pseudomonas fluorescens R098 has found application as a debittering agent for its ability to hydrolyze the peptides responsible for the bitter taste in cheese. Alkaline proteases have been applied to the fodder production using waste feathers or keratin-containing materials as the resources. For example, B. subtilis and B. licheniformis proteases have shown keratinolytic activity and can be used to hydrolyze feather keratin to give a protein concentrate for fodder production.19,20 Alkaline proteases with elastolytic and keratinolytic activity can be used in leather-processing industries starting from soaking of hides to final products by replacing the hazardous chemicals used in soaking, dehairing and bating processes.21,22 The biotreatment of leather using proteases is preferable as it not only prevents pollution problems, but also is more effective in saving energy. For example, a keratinase from B. subtilis has shown the potential to replace sodium sulfide in the dehairing process.23 As shown above, microbial proteases have already played a critical role in various industries. The pursuit of discovery strategies targeting new dimensions of molecular diversity and novel technologies to improve performance characteristics of existing proteases will certainly be the main focus of development in coming years and will lead to enzymes with much more efficient performance and novel applications.24–26

3.01.3

Lipase

Lipases (triacylglycerol ester hydrolases EC 3.1.1.3) are a class of hydrolases which catalyze the hydrolysis of triglycerides to glycerol and free fatty acids.27 Lipases are widely present in bacteria, fungi, plants and animals. The active sites of lipases are generally characterized by the triad composed of serine, histidine and aspartate. Acyl-enzyme complexes are the crucial intermediates in all lipasecatalyzed reactions. Lipases contain a helical oligopeptide unit that shields the active site of the enzyme. This so-called lid, upon interaction with a hydrophobic interface such as a lipid droplet, undergoes movement in such a way that the active site is exposed to provide free access for the substrate. This phenomenon, called interfacial activation, of lipases occurs at the lipid–water interface. In fact, because of the opposite polarity of lipase enzyme (hydrophilic) and their substrates (hydrophobic), the reaction occurs at the interface of the aqueous and organic phases. Recent studies have demonstrated that substrate inaccessibility rather than enzyme denaturation or inactivation is more responsible for lipase activity, which is a function of interfacial composition.28 While hydrolysis occurs under aqueous conditions, lipases can catalyze reversed ester-forming reactions, such as esterification, interesterification, and transesterification, in nonaqueous media. Therefore, lipases are the enzymes of choice for potential applications in numerous industrial processes including such areas as oils and fats, food and baking, diary, detergents, leather and paper processing, cosmetics and perfume, biodiesel and bioremediation.29,30 Furthermore, lipases are also widely utilized as biocatalysts in synthetic organic chemistry. The chemo-, regio- and/or stereoselective hydrolysis of carboxylic acid esters and the reverse reaction, which occur in aqueous media or organic solvents, respectively, make access to many chiral compounds of pharmaceutical importance.31,32 In the fat and oil industries, a few new enzyme-based processes have been introduced to replace conventional procedures. Cocoa butter fat is a high value product in food, confection and cosmetics industries, while palm oil is a low value product. Conversion of palm oil into cocoa butter fat substitute has been achieved by using lipase -catalyzed interesterification in organic solvent and is now a commercial process.33 The process costs can be dramatically lowered by using immobilized lipases due to the high cost of the enzymes. Lipases are regiospecific and fatty acid specific, and could be exploited for upgrading of vegetable oils to low caloric oils enriched with nutritionally important structured triacylglycerols and oleic acid, which possess tailored physicochemical properties and have a great potential in the future market. In another recently introduced process (‘de-gumming of vegetable oils’), a highly selective microbial phospholipase is used to remove phospholipids in vegetable oils. The esters of short chain fatty acids and alcohols are known flavor and fragrance compounds and used as food additives to improve flavor, lipases can be used to synthesize these compounds under green conditions.33 Nowadays, one or more enzymes including lipase are usually used in detergents for better performance and reduction of the environmental impact. Lipases are used in detergent formulations for efficient removal of lipid stains. Laundering is usually carried out under alkaline conditions, and lipases should be active under such conditions. As early as 1988, a selected strain of the fungus Humicola was used to produce the first lipase, which is capable of dissolving fatty stains, but its production yield was too low for commercial application. In 1994, Novo Nordisk commercialized Lipolase, a fungus lipase of T. lanuginosus origin which was expressed in A. oryzae. One year later, Genencor International introduced two bacterial lipases, Lumafast and Lipomax from P. mendocina and P. alcaligenes, respectively. Lipases used in detergents also include those from the species of the genera of Candida, Chromobacterium, and Acinetobacter. In leather industry, a key process, i.e., degreasing, is to remove fats and grease from fatty raw materials such as animal skins and hides. The conventional methods for degreasing are using organic solvents and surfactants, which resulted in serious environmental

Industrial Enzymes

5

threats such as volatile organic compound (VOC) emissions. Triglyceride is the main form of fat stored in animal skins. Lipase catalyzes the hydrolysis of triglyceride to glycerol and free fatty acids, resulting in removal of fats and grease from skins and hides. Both alkaline active and acid stable lipases have been used in degreasing of animal skins and hides. While alkaline active proteases are used to facilitate the degradation of fat cell membranes and sebaceous gland components, acid stable lipases can be used to treat skins in a pickled state. Lipases are also utilized in deliming and bating processing stages. In pulp and paper industry, pitch results in the major problems in the processes. Various lipases have been used to remove pitches, which are composed of glycerol esters of fatty acids, fatty acids, resin acids, sterols, other fats and waxes. Especially, pitch is accumulated during the production of paper from pulps with high resin content, e.g., sulfite and mechanical pulps from pine, and lipases can be used to reduce the pitch accumulation. This ecofriendly and nontoxic biotechnological method has been used as a routine operation in the large-scale paper-making processes since the early 1990s. Lipases are also used to remove ink from recycled paper for lower residual ink and higher brightness than recycled pulps delinked by chemical methods. Biodiesel consisting of short-chain alkyl (methyl or ethyl) esters is an alternative diesel fuels, which are typically made by transesterification of vegetable oils or animal fats. The conventional method for producing biodiesel, involving acid or base catalysts to form fatty acid alkyl esters, results in high downstream processing costs and serious environmental problems. Thus, lipase-catalyzed transesterification presents an excellent alternative for biodiesel production. As such, enzymatic processes using lipases have recently been developed from waste cooking oils and animal fats, which are available throughout the world. This presents a likely solution to their disposal problem and possible contamination of water and land resources by transforming them into a valuable product of biodiesel.34,35 However, the cost of lipases remains a hurdle for their industrial implementation in the production of biodiesel. Therefore, considerable endeavors have been made to develop cost-effective enzyme systems. Protein engineering is being used to improve the catalytic efficiency of lipases for biodiesel production. Using the tools of recombinant DNA technology for lipase production warrants a sufficient supply of suitable lipases for biodiesel production. Immobilization of lipases on suitable support materials enables their reuse and further reduces their cost.36 It can be expected that environmentally friendly, cost-effective processes for the industrial production of biodiesel is approaching. Lipases catalyze the chemo-, regio- and/or stereoselective hydrolysis and formation of carboxylic acid esters in aqueous media or organic solvents, respectively. This makes lipases the widely used enzymes in synthetic organic chemistry and have been used to prepare the chiral building blocks for pharmaceuticals, agrochemicals and pesticides via (dynamic) kinetic resolution of racemic mixtures.31,32 For example, the kinetic resolution of piperidine atropisomers can afford a chiral intermediate for the synthesis of the farnesyl protein transferase inhibitor SCH66336, an anti-cancer agent showing activity in the nanomolar range. The conventional kinetic resolution is limited by an inherent disadvantage that a maximum of 50% conversion cannot be exceeded. The combination of a lipase catalyzed acylation and ruthenium-catalyzed racemization of the substrate allows for dynamic kinetic resolution of chiral alcohols, amines and a-hydroxy esters in good yields and excellent ee values. Lipase catalyzed chemoselective and regioselective protective group incorporation, and cleavage have found useful application in the chemical manipulation of multifunctional molecules. The combination of lipase-catalyzed kinetic resolution and metal-catalyzed racemization has become a relatively mature and practical synthetic method for the preparation of optically pure alcohols, amines and carboxylic acids in organic and pharmaceutical chemistry. Recently, enzymatic polymerization has rapidly developed and provides an important synthetic technique for the construction of functional polymeric materials. Especially, the lipase-catalyzed ring-opening polymerization and polycondensation reaction have been proven successful for synthesis of tailor-made polymeric materials with unique properties. While lipase-catalyzed polymerization offers a sustainable tool for the production of polymeric materials, the cost and the catalytic activity of enzymes still limit its industrial applications. The development of novel biocatalysts with high activity, selectivity and stability are highly desired for scaling-up of processes to an industrial scale. This will be addressed by searching natural sources or through protein engineering and immobilization strategies, and it is expected that lipase-catalyzed synthesis of polymeric materials will become a promising industrial platform for the green production of biodegradable and biocompatible polymers in the near future.37 Notwithstanding current achievements in the applications of lipases in various industries, there is still a quest for lipases with improved/novel catalytic features and improved stability under the varied operational conditions. In addition to enzyme immobilization and protein engineering,38 marine microorganisms are sought as the source of enzymes with high activity under extreme conditions, because these extremophiles live in harsh environments such as deep sea hydrothermal vents, polar seas or extreme saline conditions.39 For example, cold-active lipases from marine Antarctic origin show high activity in alkaline environments and at low temperatures, allowing their use in detergents for cold-washing and energy saving.

3.01.4

Amylase

Amylases are glycoside hydrolases that act on a-1,4-glycosidic bonds and break starch down into sugars.40 Amylases of microbial origin are classified into exo-acting and endo-acting enzymes. Exo-amylases include glucoamylases and b-amylases. Glucoamylases (EC 3.2.1.3) catalyze hydrolysis of a-1,4 and a-1,6 glucosidic bonds with lower rate for a-1,6 cleavage to give b-D-glucose from the non-reducing ends of starch and related poly- and oligosaccharides. b-Amylases (EC3.2.1.2) cleave the a-1,4-glycosidic linkages in starch from the non-reducing ends to give maltose. Endo-amylases (a-Amylases, E.C. 3.2.1.1.) hydrolyze internal a-1,4 linkages and bypass a-1,6 bonds in amylopectin and glycogen at random in an endo-fashion, producing malto-oligosaccharides of varying chain lengths.41

6

Industrial Enzymes

a-Amylase is one of the most important kind of industrial amylases and has widely applied in numerous industrial processes in food, textile, paper, detergent, and pharmaceutical industries.42 a-Amylases are involved in carbohydrate metabolism, and distributed in plants, animals, and microbes. Thus a-amylases have been isolated from various sources. The amylases of microorganisms have a broad spectrum of industrial applications as they are more stable than those of plant and animal origin. Although a-amylase has been derived from various species, the dominant players in industrial sectors are enzymes originated from fungal and bacterial sources.43 Among bacteria, B. subtilis, B. stearothermophilus, B. licheniformis, and B. amyloliquefaciens are widely used for the production of thermostable a-amylases for various applications. The bacterial a-amylases have a broad profile of physicochemical properties. The pH range of these enzymes are from 1.0 (Bacillus sp.) to about 11.5 (Bacillus No. A-40-2 a-amylase). They are active at a wide temperature range with temperature-activity optima from about 25  C (Alteromonas haloplanctis a-amylase, AHA) to around 90  C (B. licheniformis a-amylase, BLA). The substrate specificity also varies greatly, they can act on the substrates with different chain length and are capable of cleaving close to the a-1,6 branch points in amylopectin and other branched glucose polymers. Starch consists of two glucose polymers: amylose and amylopectin. The former is exclusively a-1,4 linked, while the latter contains many a-1,6 branch points in addition to the a-1,4 linkages found in amylose. D-glucose (dextrose) was previously produced from starch by acid hydrolysis with a low yield of about 85% and concomitant formation of undesirable bitter sugar (gentiobiose) and coloring materials. The inevitable formation of large amount of salt from subsequent neutralization with alkali presented additional disadvantage. Enzymatic processes have now largely replaced the use of strong acid and high temperature processes. Two essential and distinct steps, liquefaction and saccharification, are usually involved in the enzymatic break down of starch to glucose.40,41,43 During liquefaction, a-amylase hydrolyzes a-1,4 linkages of the gelatinized starch at random to a dextrose equivalent (DE) of 10–15. The optimal pH for the reaction is 6.0–6.5 and a structural factor Ca2þ, which maintains the stability of the enzyme protein but does not participate in catalysis, is required. The Ca2þ binding site has been modified using protein engineering to improve its binding affinity and to lower Ca2þ levels needed for the stabilization. Liquefied and partially hydrolyzed starches are known as maltodextrins and are widely used in the food industry as thickeners. In the saccharification step, the reaction is usually carried out at 55–60  C, pH4.0–5.0. The amount of glucoamylase needed in the process and reaction times (24–72 h) are dependent on the percent of glucose desired in the product. Efficiency of saccharification with glucoamylase can be improved by adding pullulanase or isoamylase, and a glucose yield of 95%–97.5% can be achieved. Isoamylases and pullulanases hydrolyze only a-1,6 linkages, and thus are debranching enzymes. Addition of pullulanase or isoamylase can reduce the saccharification time from 72 to 48 h, allow for increased substrate concentrations (to 40%, DS), and lower the use of glucoamylase by up to 50%. The fermentable sugars obtained by these two enzymatic steps can be converted into ethanol using an ethanol fermenting microorganism such as the yeast Saccharomyces cerevisiae. Since solubility of starch is highly temperature-dependent, the removal of starch stains in low and medium-temperature laundering becomes increasingly problematic. a-Amylases catalyze hydrolysis of starch into oligosaccharides which are soluble in water. Thus, the use of a-amylases in detergents formulations enhances their ability to remove tough starch stains.42,43 When enzymes are used in detergents, the requirements for their stability and activity are extremely high, as they are often used under very alkalic and oxidizing conditions. Amylases usually can maintain high activity at low temperature, alkaline pH and oxidizing environment. These unique properties make amylases the second type of enzymes used in the formulation of enzymatic detergents for laundry and dishwashing to degrade the residues of starchy stains, and 90% of all liquid detergents contain these enzymes. a-Amylases used in the detergent industry are mainly derived from Bacillus or Aspergillus. The natural starch gel usually has very high viscosity, that is too high for paper sizing in the pulp and paper industry. a-Amylases has been used to partially degrade the polymer in a batch or continuous processes for the production of starch gel with low-viscosity and high molecular weight. Besides being a good coating for the paper, the resulting starch gel is a good agent for the sizing of paper.42,43 The sizing process enhances the stiffness and strength of paper and improves the writing quality of the paper. For example, some amylases obtained from microorganisms, Amizyme, Termamyl, Fungamyl, BAN and a-amylase G9995, have been commercialized and used in paper industry. a-Amylases are extensively employed in the bakery industry.44 During baking, the a-amylase added in the dough of bread catalyzes the degradation of starch in the flour into smaller dextrins. During the subsequent fermentation by the yeast dextrins are converted to CO2. This process results in enhancements in the volume and texture of the product. In addition, a-amylases in bread baking generate some extra sugar in the dough to improve the taste, crust color and toasting qualities of the bread. Furthermore, a-amylases have effects of reducing the staling of bread and extending their softness retention time, thus increasing the shelf life of these baked goods. An example of commercial a-amylases currently used in the bakery industry is a thermostable maltogenic amylase of B. stearothermophilus. a-Amylases are also used in other food processing industries such as brewing, production of fruit juices, digestive aids and starch syrups. Use of amylases in the pretreatment of animal feed can improve the digestibility of fiber. Amylases are used in textile industry for desizing process. Because starch is cheap and easily available, it is widely employed to yarn before fabric production to avoid fracturing of the warp thread during the weaving process. Starch is later removed from the woven fabric in a wet-process, in which the size is selectively removed by the addition of a-amylases and the fibers remain undamaged. Although a large number of a-amylases from various sources have been reported, only few of them meet the criteria of thermostability, pH tolerance, calcium independency, oxidant stability and high starch hydrolyzing efficiency for the diverse industrial applications.45 As such, thermoacidophilic, organic-solvent tolerant microbial a-amylases have been searched from extremophiles, or through improving the stability by immobilization, chemical modification, protein engineering.46–48 During last decade or so, many strategies have been developed to stabilize a protein for industrial purposes, and several stabilizing mutations of the

Industrial Enzymes

7

a-amylases have been obtained. For example, various mutations within domain B of B. licheniformis a-amylase have been found to enhance its stability. These mutant enzymes have been used in high-temperature applications. As the bacterial a-amylases have been widely utilized in various industrial sectors in recent years, search for a-amylases with specific pH-activity profiles and tailor-made substrate and product specificities are the research focus and will undoubtedly benefit many industrial applications. The initial engineering of both pH-activity profile and substrate specificity has resulted in some better a-amylases. It seems that theories and methods for addressing the engineering of both substrate specificities and pH-activity profiles are needed and such theories and methods will be rapidly developed, leading to novel a-amylases with desired properties.

3.01.5

Pullulanase

Starches are one of the most widely used polysaccharides, predominantly a-1,4 linked D-glucopyranoside polymers with different degree of a-1,6 branches. In addition to the enzymes which hydrolyze the a-1,4 linkages, complete starch hydrolysis also requires a-1,6-glycosidic cleaving enzymes including pullulanases. Pullulanases (EC 3.2.1.41) are found among animals, plants, fungi and bacteria. Type I pullulanases specifically attack a-1,6-glycosidic bonds in branched oligosaccharides such as starch, amylopectin, and glycogen, forming maltodextrins linked by a-1,4-glycosidic linkages. Type I pullulanases and isoamylase (EC 3.2.1.68) exclusively hydrolyze a,1-6 glycosidic linkages and are called as debranching enzymes. The major difference of these two enzymes is that isoamylase is not capable of hydrolyzing pullulan, a polysaccharide with a repeating a,1-6 linkage of maltotriose unit. Pullulanases can hydrolyze the a,1-6 glycosidic linkage in amylopectin and pullulan, while isoamylase exclusively hydrolyze the a,1-6 bond in amylopectin. In addition to the capability of hydrolyzing the a-1,6-glycosidic linkages of pullulan, Type II pullulanases can also hydrolyze a-1,4 linkages, and they are referred to as a-amylase–pullulanase or amylopullulanase.49,50 Pullulanase enables the complete and efficient conversion of the branched polysaccharides into small fermentable sugars, and has been widely applied in the starch-based industries. Products of enzymatic pullulan degradation are used in the pharmaceutical and food industry.50,51 For example, maltotriose syrup is produced by enzymatic hydrolysis of the polysaccharide ‘pullulan’ using the debranching enzyme, pullulanase. This syrup possesses many excellent properties, e.g., low freezing point depression, mild sweetness, retention of moisture, prevention of retrogradation of starch in foodstuffs, less color formation compared with maltose syrups, glucose syrups or sucrose. These properties are useful in food and pharmaceutical industries. High maltotriose syrup have been applied in the food industry for the manufacturing of desserts, baking and brewing, as well as in the pharmaceutical industry for replacing glucose in intravenous feeding. However, the commercial application of pullulanase was restricted due to deactivation at higher temperatures or low activity under the operational conditions. At present, starch debranching was performed by incubating pullulanases with gelatinized starch usually at 60  C. The overall energy consumption could be largely reduced when it was performed at moderate temperature. The “cold starch debranching” requires pullulanase with high activity at ambient temperature. As such, in recent years great interest has been attracted to search for new cold-active pullulanases with high catalytic efficiency at ambient temperatures.52 On the other hand, alkali-active, alkali-stable and detergent-resistant pullulanases were needed for the applications in dishwashing and laundry detergents and textile industries. Because most known pullulanases were active in the acidic or neutral pH ranges, it is highly desired to seek for pullulanases with alkali-activity, detergent-resistance from either natural sources or by protein engineering.53,54

3.01.6

Pectinase

Pectinases or pectinolytic enzymes are a group of related enzymes that hydrolyze the pectic substances.55,56 Pectic substances are high molecular weight, negatively charged, acidic complex glycosidic macromolecules (polysaccharides) that are mostly present in the plant. The primary backbone of pectic substances consists of a-1,4 linked a-D-galacturonate units, with 2%–4% of L-rhamnose units b-1,2 and b-1,4 linked to the galacturonate units. The main chain is connected with the side chains of arabinan, galactan, arabinogalactan, xylose or fucose through their C1 and C2 atoms. There are four main types of pectic substances: (1) Protopectins, the water insoluble pectic substances which are restrictedly hydrolyzed to yield pectins or pectic acids; (2) Pectic acids/pectates, the water soluble polygalacturonans with negligible amount of methoxyl groups; (3) Pectinic acids/pectinates, the polygalacturonans that contain less than 75% methylated galacturonate units. (4) Pectin, the polymethyl galacturonate with at least 75% of the carboxyl groups of the galacturonate units being esterified with methanol. Pectinases are classified into three broader groups: (1) Protopectinases are enzymes which catalyze the degradation of insoluble protopectin to soluble pectin; (2) The enzymes, which catalyze the de-esterification of pectin to remove methoxy esters, are called esterases; (3) Depolymerases are enzymes which are capable of cleaving the a-1,4-glycosidic linkages in the D-galacturonic acid units of the pectic substances. Based on the substrate preference, the cleavage mechanism and the splitting mode (random or endwise) of the glycosidic bonds, depolymerases are further divided into different categories: (1) Polymethylgalacturonases (PMG) which catalyze the hydrolysis of a-1,4-glycosidic bonds of pectin, including Endo-PMG and Exo-PMG; (2) Polygalacturonases (PG) which catalyze cleavage of a-1,4-glycosidic linkages in pectic acid including End-PG and Exo-PG; (3) Polymethylegalacturonate lyases (PMGL) which break pectin by trans-eliminative cleavage, including Endo-PMGL and Exo-PMGL; (4) Polygalacturonate lyases (PGL) which catalyze cleavage of a-1,4-glycosidic linkage in pectic acid by trans-elimination, including Endo-PGL and Exo-PGL.55

8

Industrial Enzymes

The commercial application of pectinases was dated back in 1930 for the clarification of apple juice. Since then, pectinases have been used in various industries such as food and beverage, animal feed, textile, paper and pulp and saccharification of agricultural wastes.57 In the production of fruit juice, pectinases are widely used in fruit juice extraction and clarification.56,57 The fruit juice extraction begins with washing, sorting and crushing of the fruits in a mill. After the fruit pulp is stirred in a holding tank for 15–20 min, which removes the enzyme inhibitors (polyphenols) by oxidizing them with naturally occurring polyphenol oxidases present in the fruit or added oxidizing agent polyvinyl pyrrolidone, it is treated with pectinase. During this incubation process, the pectinase degrades the soluble pectin in fruit pulp, which facilitates pressing, juice extraction and the separation of a flocculent precipitate by sedimentation, filtration or centrifugation. Otherwise, pectin makes juice extraction from the pulp difficult and blocks drainage channels in the pulp through which the juice must pass, thus hampering juice extraction. Treatment time with pectinase depends upon the nature and amount of the enzymes used, the reaction temperature and the type of fruit. In order to increase the pressing efficiency during juice extraction, other enzymes such as cellulases, arabinases and xylanases have been used in combination with pectinases. Pectins also increase the viscosity and turbidity of fruit juice. Pectinases and amylases are used together to remove the suspended matter in fruit juices, resulting in sparkling clear juices (free of haze). Pectinases also find some application in textile industry.56 These enzymes have been used in combination with amylases, lipases, cellulases and other hemicellulolytic enzymes to remove sizing agents before the fabric can be dyed. This enzymatic process has avoided the use of harsh chemicals in the textile processing, resulting in a lower discharge of waste chemicals to the environment, a safer working condition for the workers and better fabric products for the customers. Pectin is like a powerful biological glue and binds the waxes and proteins together. These noncellulosic impurities together with ‘neps’ make the cotton undyeable. In order to achieve uniform dyeing and finishing, caustic alkaline solution (3%–6% aqueous sodium hydroxide) used to be employed in the scouring of cotton at high temperature. The process was very water- and energy-consuming and produced a large amount of waste. Bioscouring offers a cost-effective and eco-friendly alternative strategy for cotton wet processing. This process uses specific enzymes to specifically degrade the noncellulosic impurities. For example, pectinases are used for the decomposition of pectinic substances, proteases for proteins, lipases for fats. Pectinases facilitate efficient interruption of the matrix without cellulose destruction, thus fiber damage is drastically limited. The pH and temperature compatibility, process efficiency and end-product quality determines the selection of pectinases for bioscouring. Bast plant fibers, such as sunn hemp, jute, ramie, flax and hemp, are excellent natural textile materials. They grow groups outside xylem in the cortex, phloem or pericycle. These fibers contain gum, which consists mainly of pectic- and hemi-celluloses and should be removed before their use for textile processing. For example, skinned ramie fibers contain 20%–30% ramie gum and must be degummed before further processing. The traditional chemical degumming treatment is polluting, toxic and non-biodegradable. The enzymatic treatment of bast fibers using xylanases together with pectinases is an economic and ecofriendly degumming process with no damage to the fibers. Although a large number of alkaline pectinases of different origins have been reported in recent years, there is a pressing need to search for new enzymes with high stability and activity under the processing and develop economically efficient and environmentally friendly bioscouring and degumming technologies.58 Pectin-containing wastewaters are generated as by-product in vegetable food processing industries. Alkaline pectinase and alkalophilic pectinolytic microbes have been used in the pretreatment of these wastewaters, facilitating the removal of pectinaceous material.59 For example, the pectic substances from industrial wastewater can be cleaned up using an extracellular endopectate lyase which originates from an alkalophilic soil isolate, Bacillus sp. GIR 621 and has optimal pH 10.0. Pectinases together with glucanases, xylanases, proteinases and amylases are also used in the feed enzyme preparation. These enzymes facilitate the liberation of nutrients either by hydrolysis of non-biodegradable fibers or by releasing nutrients blocked by these fibers.56 This increases absorption of nutrients by animal and poultry, and reduces the amount of feces released by them, thus increasing feed conversion ratio, i.e., animal weight gain with the same amount of barley. Pectinases is used by microbes to metabolize pectic substances. Studies on the biochemical, regulatory, and molecular aspects of these enzyme systems will help us in understanding the catalytic mechanisms of pectinases and guide us in engineering their properties. The future research will be directed toward the economical, ecofriendly and green biotechnology for using pectin polysaccharide in nature.

3.01.7

Xylanase

On average, plant biomass consists of 23% lignin, 40% cellulose, and 33% hemicellulose by dry weight. Hemicelluloses include xylan, mannan, galactan, and arabinan. D-xylose is the monomeric unit of xylans with traces of L-arabinose. Several hydrolytic enzymes with diverse reaction specificity and modes are required for complete conversion of xylans, in which these enzymes act cooperatively to convert xylan into sugars. The multifunctional xylanolytic enzyme system include a repertoire of hydrolytic enzymes: b-1,4-endoxylanase (endo-1,4-b-xylanase, E.C.3.2.1.8), b-xylosidase (xylan 1,4-b-xylosidase, E.C.3.2.1.37), a-L-arabinofuranosidase (a-L-arabinofuranosidase, E.C.3.2.1.55), a-glucuronidase (a-glucosiduronase, E.C.3.2.1.139), acetyl xylan esterase (E.C. 3.1.1.72), and phenolic acid (ferulic and p-coumaric acid) esterase (EC3.1.1.73), which are quite widespread among fungi, actinomycetes, and bacteria. Endo-b-1,4-xylanases catalyze the breakdown of the backbone of xylan to produce xylooligosaccharides, which in turn can be converted to xylose by b-xylosidase. Strain Aspergillus and Trichoderma are employed for the production of xylanases at the industrial scale.60,61

Industrial Enzymes

9

In the last decade, xylanolytic enzymes from microorganism have shown significant application potential in a number of processes in food, feed, and pulp and paper industries.62,63 Currently, the most promising application of xylanases is in pulp and paper industries.64 Kraft process is the pulping process in which wood chips are cooked in a solution of Na2S/NaOH at about 170  C for 2 h. Lignin is degraded and solubilized during this process. The resulting pulp showed a characteristic brown color, and the color intensity is a function of the amount and chemical state of the residual lignin, which has to be removed by bleaching process. Removal of remaining lignin from Kraft pulp is physically and chemically restricted by hemicelluloses, because lignin is covalent binding to the hemicellulose and perhaps to cellulose fibers. The hydrolysis of xylan catalyzed by xylanases facilitates release of lignin from the pulp. This biobleaching reduces the level of usage of chlorine as the bleaching agent and improves the brightness and brightness stability of the pulp. The xylanases used for biobleaching should be active at higher temperature, thermostable, alkalophilic and cellulose-free. Xylanases are increasingly used in producing cellulose from dissolving pulps for the production of rayon, cellophane and several chemicals like cellulose esters (acetates, nitrates, propionates and butyrates) and cellulose ethers (carboxymethyl cellulose, methyl and ethyl cellulose). Xylanases are used to improve dough handling properties, to enhance bread quality, and to extend shelf life by reducing the staling rate.63,65 Xylanase transforms water-insoluble hemicellulose into soluble form, which binds water in the dough. The retention of water decreases dough firmness, increases volume and creates finer and more uniform crumbs, thus significantly improving dough handling properties, and the bread quality. The staling of bread is associated with the gradual increase in crumb firmness, and xylanases have an anti-staling effect during bread storage. The addition of xylanases in the dough can substitute emulsifiers and other chemical additives used in bread production. The use of xylanases in combination with protease, lipase and a-amylase at optimum levels provide better results as compared to its sole use. Because of the synergistic effects, it is a good strategy to use xylanase with other enzymes in bread production. Xylanases have also been used in the production of oligosaccharides from isolated xylans by specific hydrolysis. The oligosaccharides are then used as functional food additives or alternative sweeteners. In cereals like barley and rye, arabinoxylans form the major non-starch polysaccharide and cause high intestinal viscosity of animal feed based on cereals, resulting in less weight gain and feed conversion efficiency. Xylanase-catalyzed hydrolysis of arabinoxylan in cereal-based animal feed results in reduced intestinal viscosity, thus improving poultry growth and feed conversion efficiency. Xylanases have also been used in the pretreatment of forage crops to enhance the digestibility of ruminant feeds. Wastes from agricultural and food industries contain xylan in large amounts. Therefore, xylanases can be used in the treatment of waste water, in which xylan is converted into xylose.63 This opens new avenue for treating hemicellulosic wastes and efficient enzymatic processes are needed to be developed. In terms of improving energy security and reducing greenhouse emissions, the abundant lignocellulosic plant biomass presents one of the sustainable and environment friendly energy sources. The major obstacle for the conversion of the biomass into the biofuel is the difficulty in converting lignocellulosic materials to the fermentable sugars. Xylanase, together with other hydrolytic enzymes, such as cellulases and laccases have been used to realize this critical step in the production of biofuels, such as ethanol, from lignocellulosic biomass. Robust thermostable xylanases are usually required for the harsh processing conditions during deconstruction of lignocellulose to the fermentable products.66

3.01.8

Laccase

Laccases (benzenediol: oxygen oxidoreductases; EC 1.10.3.2) are enzymes that oxidize a large variety of organic substrates using molecular dioxygen as the oxidizing agent. Molecular dioxygen is reduced to water. The substrates can be ortho- and paradiphenols, polyphenols, polyamines, aminophenols, lignins and aryl diamines as well as some inorganic ions. The laccase molecule is a dimeric or tetrameric glycoprotein. There are four copper atoms located in three redox sites in each monomer.67 Laccase alone cannot oxidize non-phenolic compounds. When laccases are used in combination with an electron transfer mediator (Laccasemediated systems, LMS), their substrate range can be extended to non-phenolic compounds. Laccase-mediated systems have shown great potential in array of industrial applications, e.g., oxidation of organic pollutants, pulp delignification, and the development of biosensors or biofuel cells.67,68 For example, in the presence of 1-hydroxybenzotriazole (HBT), laccase has been used in pulp delignification. Phenothiazines have been used as an electron transfer mediator of laccase to bleach and prevent redeposition of azo dyes in textile manufacturing. The laccase/mediator system (LMS) offers one of the promising approaches for an environmentally benign pulp-bleaching process. The lignin in wood pulp must be separated and degraded in production of paper. The conventional chlorine-based delignification/bleaching procedures are highly polluting. It is urgent to replace these procedures with milder and cleaner bio-bleaching technologies. At present, very few enzymatic processes are comparable with current chemical bleaching technologies in terms of the delignification/brightening capabilities. One of the few exceptions is the application of laccase/mediator system in the delignification of kraft pulps. Although many efforts have been made to develop alternative bio-bleaching systems, all these bio-bleaching studies are limited to wood pulps. There are over 7105 t of dyestuff produced annually, and most of them are difficult to decolorize due to their synthetic origin. The textile industry uses up two-thirds of these dyes in the market. This presents an urgency for developing technologies for the removal of dyes from textile industrial effluents. Bioprocesses based on laccases provide a promising solution to this problem because of their capability of degrading dyes of diverse chemical structure.67 Laccases have been used in textile bleaching, for

10

Industrial Enzymes

example, in 1996 Novozyme launched DeniLite, the first industrial laccase which was used together with a mediator in denim finishing. The degradation of polycyclic aromatic hydrocarbons (PAHs) and other xenobiotics is very important, because they constitute a major source of contamination in the environment. Laccases can be used in bioremediation to decompose these PAHs, including those arising from fossil fuel utilization and natural oil deposits.67,69 Laccases have also been employed in various areas of the food industry such as beverage processing, sugar beet pectin gelation and baking etc.67 Phenolics in fruit juice, beer and wine make them browning and cause haze formation and turbidity development, so laccases are applied to eliminate these undesirable compounds in beverage processing. A fungus laccase from Trametes hirsuta is employed in baking, since its ability to cross-link biopolymers makes the dough more resistant and less extensible. Given the eco-friendliness (molecular oxygen from air is used as the final electron acceptor and they only release water as byproduct) and a wide array of applications, there is a high industrial demand for the laccases with broad substrate spectrum, wide pH range, high thermostability, and tolerance to alkaline environments. In this context, rational, semi-rational and directed evolution approaches have been employed to engineer laccases with these desired biotechnological properties.70

3.01.9

Transglutaminase

Transglutaminases (EC 2.3.2.13) are a class of enzymes that catalyze the formation of a covalent bond between the g-carboxamide group of protein- or peptide-bound glutamine (acyl donors) and the free amine group of protein- or peptide-bound lysine (acyl acceptors). Proteins can be modified by transglutaminase-catalyzed intra- or inter-molecular crosslinking, thus improving the properties of the protein.71,72 Without substrate bearing amino groups, water molecules can be used as acyl acceptors in transglutaminase-catalyzed reaction, resulting in the deamination of glutamine residues. Transglutaminases are widely present in animal tissues and body fluids, plants, fishes, and microorganisms. Mammalian transglutaminases are calcium-dependent, while microbial transglutaminases are calcium-independent and have smaller molecular weight. The activity of transglutaminases varies greatly dependent on their origin. Transglutaminases have been utilized to enhance the functional properties of food proteins by catalyzing the cross-linking reactions of proteins. This is applied to the processing of whey proteins, soya proteins, gluten, meat and fish protein,71,72 which showed various advantages. First, lysine in food proteins is protected from numerous chemical reactions, thus increasing the shelf-life of food. Second, different proteins containing complementary limiting essential amino acids can be cross-linked using transglutaminase as catalyst for the production of food proteins with higher nutritive value. Third, various functional groups can be incorporated onto the glutamine residues of protein using transglutaminases, enhancing the functional properties of the protein. Transglutaminases also can encapsulate lipids and/or lipid-soluble materials, form heat- and water-resistant films, improve elasticity and waterholding capacity of modified proteins. Transglutaminases have been used in the modification of wool fabrics. This improves the dyeing properties of wool fabrics, as well as their physical/mechanical properties. As such, this enzyme have found various applications in remediating damage of wool, antibacterial finishing, hydrophilic finishing, and color fixing in dyeing.73 In recent years, microbial transglutaminases (mTG) have been successfully utilized in the site-specific attachment of cytotoxic drugs to antibodies, generating homogeneous and reproducible antibody drug conjugates (ADCs). The role of position, linker and payload were studied in an effort to generate homogeneous molecules of potentially effective therapy with better safety profile than non-conjugated cytotoxics.74

3.01.10 Phytase Phytases (myo-inositol hexakisphosphate phosphohydrolase, EC 3.1.3.8) catalyze the hydrolysis of phytate [myoinositol(1,2,3,4,5,6)hexakisphosphate] to the inorganic phosphate and less-phosphorylated myoinositol derivatives.75 They are present in plants, microorganisms and in some animal tissues, and belong to a subgroup of phosphatases which initiate the stepwise dephosphorylation of phytate. Phytate is the most abundant inositol phosphate in nature. Dephosphorylation can be initiated at different carbon in the myo-inositol ring of phytate. According to the initiation position, phytases are classified into 3-phytases (E.C. 3.1.3.8), 6-phytases (E.C. 3.1.3.26) and 5-phytases (E.C. 3.1.3.72). Phytases have been grouped into acid and alkaline phytases based on their pH optima. Phytases can also be referred to as histidine acid phytases, b-propeller phytases, cysteine phytases or purple acid phytases according to their catalytic mechanisms. The major application of phytases is as animal feed additives, which are largely for swine, poultry, and fish.75 Most commercially available phytases are of fungal or bacterial origin genes, but expressed in yeasts for production purposes. The first industrial phytase product was Natuphos launched in 1991. It has been shown that supplementing dietary phytases can effectively enhance phytatephosphorus bioavailability in animal feed because the enzyme catalyzes the hydrolysis of phytate to inorganic phosphate. For example, utilization of phytate-phosphorus by poultry and pigs can be enhanced 20%–45% by supplementation with microbial phytase to their diets.76 The improvement derived by the supplementation of phytases is also diet-dependent. The maize-based diets showed little intrinsic phytase activity, and addition of phytase generally has significant effects; while less effect was observed for addition of phytase in barley/wheat-based diets which usually possess significant phytase activity. Thus plant phytates are a very valuable resource of phosphorus for animal nutrition. High dietary phosphorus bioavailability avoids the need of supplementing

Industrial Enzymes

11

the animal feed with inorganic phosphorus such as mono- and dicalcium-phosphate (MCP, DCP). Not only the supplementation with microbial phytases in animal feed improves the phosphate utilization in the feed, but also the phosphate content in animal manure is substantially reduced, thus decreasing the discharged phosphate into the water bodies. Since the supplementing the animal feed with inorganic phosphorus and excretion of high phosphorus by monogastric animals creates an environmental problem of phosphorus pollution, application of phytases as animal feed additives offers an effective approach to deal with the phosphorus pollution generated from animal agriculture.77 Phytases have been increasingly used in food processing to improve bioavailabilities of minerals, since degradation of phytates results in the release of minerals from the complexes of phytic acid and the lower myo-inositol phosphates with divalent metals.78 The following strategies are applied to optimize phytate degradation during food processing, e.g., more favorable conditions during food processing for the phytase activity, use of raw material with a high intrinsic phytate-degrading activity, addition of isolated phytases to the production process. Furthermore, phytases may find application in the production of functional foods or food supplements with health benefits. Phytase-catalyzed phytate degradation is also helpful for breadmaking, corn wet milling, the fractionation of cereal bran and production of plant protein isolates. The effectiveness of phytases as food/feed additive is dependent on their pH/temperature optima, stability and resistance to the action of proteases. These properties of microbial phytases are affected by their sources. The initially marketed phytases were fungal in origin, principally from Aspergillus species. This first generation phytases from were usually thermolabile and primarily active at neutral pH. The bacterial phytases have some unique properties such as resistance to proteases, acidic to alkaline pH optima, requirement of metal ions and high substrate specificity that are different from fungal and other phytases.79 However, these enzymes are not thermos-tolerant through pelleting temperatures and do not perform effectively under various conditions of digestive systems. Therefore, in recent years, many efforts are still made to develop tailor-made phytases for the diverse application purposes by employing environmental screening, genome mining and enzyme engineering.80

3.01.11 Perspective As discussed above, as effective catalysts working under mild conditions, enzymes have found numerous applications in different industrial products and processes. Recent advances in protein engineering and design have promoted the efficient development of new enzymes with enhanced properties for established technical applications and tailor-made for entirely new areas of application where enzymes have not previously been used. Because of the constant advances in modern biotechnology, the development pace of novel enzymes will accelerate to meet the increasing industrial needs in the years to come. In addition, the application of enzymes will be rapidly expanded to new areas where it would not be expected before. Successful developments in novel enzymes and new applications will result in significant savings in energy and water for the benefit of both the industry in question and the environment.81 This will greatly help us to address the challenges which our society faces in resources and environment in the coming generations.

References 1. Jackson, C. J.; Gillam, E. M. J.; Ollis, D. L. Directed Evolution of Enzymes. In Comprehensive Natural Products, Vol. ii, Lew, M., Hung-Wen, L., Eds.; Elsevier: Oxford, 2010; pp 723–749. 2. Tang, W. L.; Zhao, H. Industrial Biotechnology: Tools and Applications. Biotechnol. J. 2009, 4, 1725–1739. 3. Kirk, O.; Borchert, T. V.; Fuglsang, C. C. Industrial Enzyme Applications. Curr. Opin. Biotechnol. 2002, 13, 345–351. 4. Sch fer, T.; Borchert, T.; Nielsen, V.; et al. Industrial Enzymes. White Biotechnol. 2007, 59–131. 5. Schäfer, T.; Borchert, T. W.; Nielsen, V. S.; et al. Industrial Enzymes. Adv. Biochem. Eng. Biotechnol. 2007, 105, 59–131. 6. Saha, B. C.; Jordan, D. B.; Bothast, R. J. Enzymes, Industrial (Overview). In Encyclopedia of Microbiology; Moselio, S., Ed., Academic Press: Oxford, 2009; pp 281–294. 7. Li, Q.; Yi, L.; Marek, P.; et al. Commercial Proteases: Present and Future. FEBS (Fed. Eur. Biochem. Soc.) Lett. 2013, 587, 1155–1163. 8. Sharma, K. M.; Kumar, R.; Panwar, S.; Kumar, A. Microbial Alkaline Proteases: Optimization of Production Parameters and Their Properties. J. Gen. Eng. Biotechnol. 2017, 15, 115–126. 9. Maurer, K.-H. Detergent Proteases. Curr. Opin. Biotechnol. 2004, 15, 330–334. 10. Bhaskar, V.; Jones Raj, T. R.; Kandasamy, S. K. J.; Vijaykumar, P.; Achary, A. Optimization of Production of Subtilisin in Solid Substrate Fermentation Using Response Surface Methodology. Afr. J. Biotechnol. 2008, 7, 2286–2291. 11. Gupta, R.; Beg, Q. K.; Lorenz, P. Bacterial Alkaline Proteases: Molecular Approaches and Industrial Applications. Appl. Microbiol. Biotechnol. 2002, 59, 15–32. 12. Kumar, D.; Thakur, N.; Verma, R.; Bhalla, T. C. Microbial Proteases and Application as Laundry Detergent Additive. Res. J. Microbiol. 2008, 3, 661–672. 13. Contesini, F. J.; de Melo, R. R.; Sato, H. H. An Overview of Bacillus Proteases: from Production to Application. Crit. Rev. Biotechnol. 2018, 38, 321–334. 14. Bryan, P. N. Protein Engineering of Subtilisin. Biochim. Biophys. Acta 2000, 1543, 203–222. 15. Vojcic, L.; Pitzler, C.; Körfer, G.; et al. Advances in Protease Engineering for Laundry Detergents. N. Biotech. 2015, 32, 629–634. 16. Estell, D. A.; Graycar, T. P.; Wells, J. A. Engineering an Enzyme by Site-Directed Mutagenesis to Be Resistant to Chemical Oxidation. J. Biol. Chem. 1985, 260, 6518–6521. 17. Sumantha, A.; Larroche, C.; Pandey, A. Microbiology and Industrial Biotechnology of Food-Grade Proteases: A Perspective. Food Technol. Biotechnol. 2006, 44, 211–220. 18. Tavano, O. L.; Berenguer-Murcia, A.; Secundo, F.; et al. Biotechnological Applications of Proteases in Food Technology. Compr. Rev. Food Sci. Food Saf. 2018, 17, 412–436. 19. Dalev, P. G. Utilization of Waste Feathers from Poultry Slaughter for Production of a Protein Concentrate. Bioresour. Technol. 1994, 48, 265–267. 20. Cheng, S. W.; Hu, M. N.; Shen, S. W.; et al. Production and Characterization of Keratinase of a Feather-Degrading Bacillus licheniformis pwd-1. Biosc. Biotech. Biochem. 1995, 59, 2239–2243. 21. Arunachalam, C.; Saritha, K. Protease Enzyme: An Eco-Friendly Alternative for Leather Industry. Indian Journal of Science and Technology 2009, (12), 29–32. 22. Mukhtar, H.; Ikram, U. H. Production of Alkaline Protease by Bacillus subtilis and its Application as a Depilating Agent in Leather Processing. Pakistan J. Bot. 2008, 40, 1673–1679.

12

Industrial Enzymes

23. Arunachalam, C.; Saritha, K. Protease Enzyme: An Eco-Friendly Alternative for Leather Industry. Indian Journal of Science and Technology 2009, 2, 29–32. 24. Rai, S. K.; Roy, J. K.; Mukherjee, A. K. Characterisation of a Detergent-Stable Alkaline Protease from a Novel Thermophilic Strain Paenibacillus tezpurensis Sp. Nov. As-s24-ii. Appl. Microbiol. Biotechnol. 2010, 85, 1437–1450. 25. Doukyu, N.; Ogino, H. Organic Solvent-Tolerant Enzymes. Biochem. Eng. J. 2010, 48, 270–282. 26. Ali, S.; Rana, A.; Rasool, H.; Masood, M. B. E. Protease: An Enzyme with Multiple Industrial Applications. Europ. J. Pharmaceut. Med. Res. 2017, 4, 63–70. 27. Brockman, H. L. Lipases. In Encyclopedia of Biological Chemistry; William, J. L., Lane, M. D., Eds., Elsevier: New York, 2004; pp 571–575. 28. Reis, P.; Holmberg, K.; Watzke, H.; Leser, M.; Miller, R. Lipases at Interfaces: A Review. Adv. Colloid Interface Sci. 2009, 147, 237–250. 29. Hasan, F.; Shah, A. A.; Hameed, A. Industrial Applications of Microbial Lipases. Enzym. Microb. Technol. 2006, 39, 235–251. 30. Sarmah, N.; Revathi, D.; Sheelu, G.; Rani, K. Y.; Sridhar, S.; Mehtab, V.; Sumana, C. Recent Advances on Sources and Industrial Applications of Lipases. Biotechnol. Prog. 2018, 34, 5–28. 31. de Miranda, A. S.; Miranda, L. S. M.; de Souza, R. O. M. A. Lipases: Valuable Catalysts for Dynamic Kinetic Resolutions. Biotechnol. Adv. 2015, 33, 372–393. 32. Verho, O.; Bäckvall, J. E. Chemoenzymatic Dynamic Kinetic Resolution: a Powerful Tool for the Preparation of Enantiomerically Pure Alcohols and Amines. J. Am. Chem. Soc. 2015, 137, 3996–4009. 33. Rajendran, A.; Palanisamy, A.; Thangavelu, V. Lipase Catalyzed Ester Synthesis for Food Processing Industries. Braz. Arch. Biol. Technol. 2009, 52, 207–219. 34. Bajaj, A.; Lohan, P.; Jha, P. N.; Mehrotra, R. Biodiesel Production through Lipase Catalyzed Transesterification: An Overview. J. Mol. Catal. B Enzym. 2010, 62, 9–14. 35. Tacias-Pascacio, V. G.; Virgen-Ortíz, J. J.; Jiménez-Pérez, M.; et al. Evaluation of Different Lipase Biocatalysts in the Production of Biodiesel from Used Cooking Oil: Critical Role of the Immobilization Support. Fuel 2017, 200, 1–10. 36. Tan, T.; Lu, J.; Nie, K.; Deng, L.; Wang, F. Biodiesel Production with Immobilized Lipase: A Review. Biotechnol. Adv. 2010, 28, 628–634. 37. Zhang, J.; Shi, H.; Wu, D.; Xing, Z.; Zhang, A.; Yang, Y.; Li, Q. Recent Developments in Lipase-Catalyzed Synthesis of Polymeric Materials. Process Biochem. 2014, 49, 797–806. 38. Dasetty, S.; Blenner, M. A.; Sarupria, S. Engineering Lipases: Walking the Fine Line between Activity and Stability. Mater. Res. Express 2017, 4, 114008. 39. Navvabi, A.; Razzaghi, M.; Fernandes, P.; Karami, L.; Homaei, A. Novel Lipases Discovery Specifically from Marine Organisms for Industrial Production and Practical Applications. Process Biochem. 2018, 70, 61–70. 40. Gangadharan, D.; Sivaramakrishnan, S. Amylolytic Enzymes. Biotechnol. Agro Ind. Resid. Utilis. 2009, 359–369. 41. Bijttebier, A.; Goesaert, H.; Delcour, J. Amylase Action Pattern on Starch Polymers. Biologia 2008, 63, 989–999. 42. van der Maarel, M. J. E. C.; van der Veen, B.; Uitdehaag, J. C. M.; Leemhuis, H.; Dijkhuizen, L. Properties and Applications of Starch-Converting Enzymes of the Alpha-Amylase Family. J. Biotechnol. 2002, 94, 137–155. 43. Souza, P.; Magalh es, P. Application of Microbial Alpha-Amylase in Industry-A Review. Braz. J. Microbiol. 2010, 41, 850–861. 44. Goesaert, H.; Slade, L.; Levine, H.; Delcour, J. Amylases and Bread Firming-An Integrated View. J. Cereal. Sci. 2009, 50, 345–352. 45. Saini, R.; Saini, H. S.; Dahiya, A. Amylases: Characteristics and Industrial Applications. J. Pharmacogn. Phytochem. 2017, 6, 1865–1871. 46. Asoodeh, A.; Emtenani, S.; Emtenani, S.; Jalal, R.; Housaindokht, M. R. Molecular Cloning and Biochemical Characterization of a Thermoacidophilic, Organic-Solvent Tolerant a-amylase from a Bacillus Strain in Escherichia coli. J. Mol. Catal. B Enzym. 2014, 99, 114–120. 47. Dey, T. B.; Kumar, A.; Banerjee, R.; Chandna, P.; Kuhad, R. C. Improvement of Microbial a-amylase Stability: Strategic Approaches. Process Biochem. 2016, 51, 1380–1390. 48. Sindhu, R.; Binod, P.; Madhavan, A.; Beevi, U. S.; Mathew, A. K.; Abraham, A.; Pandey, A.; Kumar, V. Molecular Improvements in Microbial a-amylases for Enhanced Stability and Catalytic Efficiency. Bioresour. Technol. 2017, 245, 1740–1748. 49. Shim, J. H.; Park, J. T.; Hong, J. S.; et al. Role of Maltogenic Amylase and Pullulanase in Maltodextrin and Glycogen Metabolism of Bacillus subtilis 168. J. Bacteriol. 2009, 191, 4835–4844. 50. Singh, R. S.; Saini, G. K.; Kennedy, J. F. Maltotriose Syrup Preparation from Pullulan Using Pullulanase. Carbohydr. Polym. 2010, 80, 401–407. 51. Hii, S. L.; Tan, J. S.; Ling, T. C.; Ariff, A. B. Pullulanase: Role in Starch Hydrolysis and Potential Industrial Applications. Enzym. Res. 2012, 2012, 921362. 52. Elleuche, S.; Qoura, F. M.; Lorenz, U.; Rehn, T.; Brück, T.; Antranikian, G. Cloning, Expression and Characterization of the Recombinantcold-Active Type-I Pullulanase from Shewanella arctica. J. Mol. Catal. B Enzym. 2015, 116, 70–77. 53. Lu, Z.; Hu, X.; Shen, P.; Wang, Q.; Zhou, Y.; Zhang, G.; Ma, Y. A pH-Stable, Detergent and Chelator Resistant Type I Pullulanase from Bacillus pseudofirmus 703 with High Catalytic Efficiency. Int. J. Biol. Macromol. 2018, 109, 1302–1310. 54. Nisha, M.; Satyanarayana, T. Characteristics, Protein Engineering and Applications of Microbial Thermostable Pullulanases and Pullulan Hydrolases. Appl. Microbiol. Biotechnol. 2016, 100, 5661–5679. 55. Jayani, R. S.; Saxena, S.; Gupta, R. Microbial Pectinolytic Enzymes: A Review. Process Biochem. 2005, 40, 2931–2944. 56. Jacob, N. Pectinolytic Enzymes. In Biotechnology for Agro-Industrial Residues Utilisation; nee’Nigam, P. S., Pandey, A., Eds., Springer ScienceþBusiness Media B.V, 2009; pp 383–396. 57. Garg, G.; Singh, A.; Kaur, A.; Singh, R.; Kaur, J.; Mahajan, R. Microbial Pectinases: an Ecofriendly Tool of Nature for Industries. 3 Biotech 2016, 6, 47. 58. Zhou, C.; Ye, J.; Xue, Y.; Ma, Y. Directed Evolution and Structural Analysis of Alkaline Pectate Lyase from Alkaliphilic Bacillus sp. N16-5 for Improvement of Thermostability for Efficient Ramie Degumming. Appl. Environ. Microbiol. 2015. https://doi.org/10.1128/AEM.01017-15. 59. Kohli, P.; Gupta, R. Alkaline Pectinases: A Review. Biocatal. Agric. Biotechnol. 2015, 4, 279–285. 60. Beg, Q. K.; Kapoor, M.; Mahajan, L.; Hoondal, G. S. Microbial Xylanases and Their Industrial Applications: A Review. Appl. Microbiol. Biotechnol. 2001, 56, 326–338. 61. Biely, P.; Singh, S.; Puchart, V. Towards Enzymatic Breakdown of Complex Plant Xylan Structures: State of the Art. Biotechnol. Adv. 2016, 34, 1260–1274. 62. Corral, O. L.; Villaseñor-Ortega, F. Xylanases. In Advances in Agricultural and Food Biotechnology; Guevara-González, R. G., Torres-Pacheco, I., Eds.; 2006; pp 305–322. 63. Chakdar, H.; Kumar, M.; Pandiyan, K.; Singh, A.; Nanjappan, K.; Kashyap, P. L.; Srivastava, A. K. Bacterial Xylanases: Biology to Biotechnology. 3 Biotech 2016, 6, 150. 64. Walia, A.; Guleria, S.; Mehta, P.; Chauhan, A.; Parkash, J. Microbial Xylanases and Their Industrial Application in Pulp and Paper Biobleaching: a Review. 3 Biotech 2017, 7, 11. 65. Butt, M. S.; Tahir-Nadeem, M.; Ahmad, Z.; Sultan, M. T. Xylanases and Their Applications in Baking Industry. Food Technol. Biotechnol. 2008, 46, 22–31. 66. Uday, U. S. P.; Choudhury, P.; Bandyopadhyay, T. K.; Bhunia, B. Classification, Mode of Action and Production Strategy of Xylanase and its Application for Biofuel Production from Water Hyacinth. Int. J. Biol. Macromol. 2016, 82, 1041–1054. 67. Couto, S. R.; Herrera, J. L. T. Industrial and Biotechnological Applications of Laccases: A Review. Biotechnol. Adv. 2006, 24, 500–513. 68. Mate, D. M.; Alcalde, M. Laccase: a Multi-Purpose Biocatalyst at the Forefront of Biotechnology. Microbial Biotechnol. 2017, 10, 1457–1467. 69. Minussi, R. C.; Miranda, M. A.; Silva, J. A.; et al. Purification, Characterization and Application of Laccase from Trametes Versicolor for Colour and Phenolic Removal of Olive Mill Wastewater in the Presence of 1-Hydroxybenzotriazole. Afr. J. Biotechnol. 2007, 6, 1248–1254. 70. Mate, D. M.; Alcalde, M. Laccase Engineering: From Rational Design to Directed Evolution. Biotechnol. Adv. 2015, 33, 25–40. 71. Martins, I. M.; Matos, M.; Costa, R.; Silva, F.; Ananias Pascoal, A.; Estevinho, L. M.; Choupina, A. B. Transglutaminases: Recent Achievements and New Sources. Appl. Microbiol. Biotechnol. 2014, 98, 6957–6964. 72. Gaspar, A. L. C.; de Góes-Favoni, S. P. Action of Microbial Transglutaminase (MTGase) in the Modification of Food Proteins: A Review. Food Chem. 2015, 171, 315–322. 73. Fu, J.; Jing Su, J.; Wang, P.; Yu, Y.; Wang, Q.; Cavaco-Paulo, A. Enzymatic Processing of Protein-Based Fibers. Appl. Microbiol. Biotechnol. 2015, 99, 10387–10397. 74. Strop, P. Versatility of Microbial Transglutaminase. Bioconjug. Chem. 2014, 25, 855–862. 75. Lei, X. G.; Porres, J. M.; Mullaney, E. J.; Brinch-Pedersen, H. Phytase: Source, Structure and Application. In Industrial Enzymes: Structure, Function and Applications; Polaina, J., MacCabe, A. P., Eds., Springer, 2007; pp 503–529.

Industrial Enzymes

13

76. Maenz, D. D. Enzymatic Characteristics of Phytases as They Relate to Their Use in Animal Feed. In Enzymes in Farm Animal Nutrition; Bedford, M. R., Partridge, G. G., Eds., CABI Publishing: Wallington, UK, 2001; pp 61–84. 77. Kumar, A.; Chanderman, A.; Makolomakwa, M.; Perumal, K.; Singh, S. Microbial Production of Phytases for Combating Environmental Phosphate Pollution and Other Diverse Applications. Crit. Rev. Environ. Sci. Technol. 2016, 46, 556–591. 78. Greiner, R.; Konietzny, U. Phytase for Food Application. Food Technol. Biotechnol. 2006, 44, 125–140. 79. Jain, J.; Singh, B. Characteristics and Biotechnological Applications of Bacterial Phytases. Process Biochem. 2016, 51, 159–169. 80. Vasudevan, U. M.; Krishna, S.; Jalaja, V.; Pandey, A. Microbial Phytase: Impact of Advances in Genetic Engineering in Revolutionizing its Properties and Applications. Bioresour. Technol. 2017, 245, 1790–1799. 81. Sheldon, R. A.; Woodley, J. M. Role of Biocatalysis in Sustainable Chemistry. Chem. Rev. 2018, 118, 801–838.

3.02 Fundamentals and Industrial Applicability of Multifunctional CAZyme Systemsq Nicholas S Sarai*, Michael E Himmel, and Yannick J Bomble, Biosciences Center, National Renewable Energy Laboratory, Golden, CO, United States Amaranta Kahn and Edward A Bayer, Department of Biomolecular Sciences, The Weizmann Institute of Science, Rehovot, Israel © 2019 Elsevier B.V. All rights reserved. This is an update of Q. Xu, Y. Luo, S.-Y. Ding, M.E. Himmel, L. Bu, R. Lamed, E.A. Bayer, 3.03 - Multifunctional Enzyme Systems for Plant Cell Wall Degradation, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 15–25.

3.02.1 Introduction 3.02.2 Natural Diversity of Multifunctional Enzymes 3.02.2.1 Clues Into Function 3.02.3 Industrial Applicability of Multifunctional Enzymes Acknowledgments References

3.02.1

14 17 17 20 22 22

Introduction

Lignocellulosic biomass, the primary organic product of photosynthesis, represents the most abundant renewable carbon resource on Earth. An evolutionary “tug-of-war” between plants and lignocellulose-utilizing organisms has resulted in plants armed with recalcitrant cell-wall structures and organisms displaying sophisticated biomass depolymerization mechanisms. Understanding the mechanisms that contribute to turnover and utilization of lignocellulosic carbon can enhance the science of carbon cycling throughout Earth’s biosphere and lend actionable insights into ways of closing the cycle. Furthermore, taking lessons from the myriad of depolymerization strategies evolved by microorganisms offers a promising route for human-kind’s goals of biomass conversion to biofuels, bio-derived chemicals, and advanced biomaterials (Figs. 1–3). Multi-scale structural, compositional, and molecular features of plant cell walls constitute natural recalcitrance to enzymatic attack. Moreover, fermentable sugars are trapped in plant cell walls composed of a complex array of cellulose, hemicellulose, and lignin linkages. Enzymatic access to cellulosic and hemicellulosic polysaccharides is impeded by prominent structural features, such as coverings of bark and rind, as well as internal vascularity. At the molecular level, the crystallinity of cellulose microfibrils and the resulting hydrogen bonding networks limit enzymatic access and efficiency.1 Cellulose microfibrils are encased in a hemicellulose barrier composed of various polysaccharides, such as xylan, xyloglucan, mannan, arabinose, and pectin.2 This network further limits accessibility to cellulose chains and dramatically increases the combinatorial complexity of chemical linkages, demanding a correspondingly large set of enzymes to attack these bonds.3 Finally, non-carbohydrate components, such as lignin, further rigidify the cell wall and limit access. To enable organisms to access carbon stored in plant biomass, several enzymatic paradigms for lignocellulose depolymerization have been identified throughout the tree of life, including free monofunctional enzymes, cellulosomes, and multifunctional enzymes. In all three of these systems, synergistic enzyme classes with divergent substrate specificity and mechanisms can be employed to depolymerize the network of bonds in lignocellulosic biomass. The bulk of enzymes expressed by cellulolytic fungi and bacteria are glycoside hydrolases (GHs) and carbohydrate esterases.4 More than 140 different structure-based families of glycoside hydrolases are known.4 These hydrolytic enzymes function by hydrolyzing glycosidic linkages in cellulose and other polysaccharides via an acid-base catalytic mechanism that either retains or inverts the stereochemistry of the anomeric carbon.5 There exist three main classes of biomass degrading enzymes, based on their modes of action, endoglucanases, exoglucanases, and cellobiases. Endoglucanases are nonprocessive cellulases, which hydrolyze glycosidic bonds anywhere in the polysaccharide chain, but may prefer the less crystalline regions. Exoglucanases processively cleave cellodextrins from the reducing or non-reducing end of the cellulose chain. Cellobiases are deployed to convert cellobiose into monomeric glucose. The diversity of mechanisms and substrate specificities within GH families creates intermolecular synergy between GHs. Hemicellulose networks in plant biomass are often decorated by acyl groups, which limit accessibility to enzymes. Carbohydrate esterases serve an important role in de-acylating hemicellulose. Beyond GHs yet present in less abundance are accessory enzymes, such as cellodextrin phosphorylases, polysaccharide lyases, and auxiliary activity enzymes (AA) which endow an organism’s cellulolytic arsenal with further diversity in mechanisms and modes of action.

* Present Address: Division of Chemistry and Chemical Engineering, California Institute of Technology, 1200 East California Boulevard, Pasadena, CA 91125, USA. q Change History: November 2018. M. Himmel, Y. Bomble, N.S. Sarai, A. Kahn, and E.A. Bayer added the section, “Clues into function” and updated the section dealing with industrial applicability of multifunctional enzymes.

14

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-12-809633-8.09128-7

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems

15

Figure 1 The hyperactive cellulase CelA from Caldicellulosiruptor bescii actively bound to cellulose. CelA is the most active single-gene product on cellulose thus far isolated in the biosphere. The two glycoside hydrolase domains (GH9 in brown and GH48 in light blue), family 3c carbohydrate binding modules (violet), family 3b carbohydrate binding modules (red), and proline/threonine-rich linker peptides (gray) including putative O-linked glycosylation (yellow) are shown.

Figure 2 Domain organization of multifunctional CAZymes characterized as full-length constructs in the Caldicellulosiruptor pangenome. Hyperthermophilic bacteria in the genus Caldicellulosiruptor dwell in hot springs throughout the terrestrial biosphere. The unique demands on life in this harsh and nutrient-poor ecosystem have guided the evolution of a unique reliance on systems of multifunctional CAZymes for carbon uptake. Species throughout the genus exhibit different combinations of these enzymes and demonstrate correspondingly diverse lignocellulose deconstruction performance. The modular domain architecture of these enzymes is shown, highlighting families of glycoside hydrolases (GH) and carbohydrate binding modules (CBM). Multifunctional enzymes that are secreted are displayed with an N-terminal signal peptide (gray).

It has been proposed that microorganisms in nature rely on secreting enzyme arrays that are not only synergistic, but also tailored to the specific biomass substrate encountered in the extracellular milieu. As such, cellulolytic microorganisms are often well adapted to deconstruct certain types of biomass. In the native environment, microorganisms fill a specialized deconstruction niche and rely on cellulolytic microbes occupying different niches to provide soluble sugars or increase accessibility within recalcitrant biomass. This inter-organismal synergy allows diverse and specialized species to provide “public goods” to other ecosystem members, enabling survival in harsh and substrate-limited environments.6 For example, opportunist microbes that utilize short oligosaccharides as their carbon source, do indeed help the cellulose degraders by attenuating the buildup of inhibitory cellodextrins.7

16

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems

Figure 3 Industrial applications of multifunctional plant biomass degrading enzymes. Multifunctional CAZymes can be harnessed as green biocatalysts for a myriad of industrial reactions using lignocellulosic biomass as an abundant, inexpensive feedstock. Via schemes implementing Consolidated Bio Processing (CBP), engineered microbial strains utilizing multifunctional CAZymes can produce fuel molecules (A) and bio-derived building block chemicals (B) with diverse chemistries from lignocellulose. These molecule classes can replace petroleum derived fuels and serve as feedstocks for bio-polymer synthesis, respectively. In addition to their advantages in CBP processes, these enzymes could be highly efficient when incorporated into in vitro enzyme cassettes for industrial biomass degradation (C), such as in simultaneous saccharification and fermentation (SSF) biofuel production strategies. Recently, multifunctional enzymes have been applied to the nascent nanobiomaterials industry where they have demonstrated performance benefits in creating monodisperse nanocellulose (D).

The free enzyme mode of depolymerization predominates in fungi and bacteria.8 Acting individually, these carbohydrate-active enzymes (CAZymes) are composed of a single catalytic domain (CD) tethered to a carbohydrate binding module (CBM). Free enzymes exist with an array of substrate specificities and modes of action that allow these individually-acting secreted enzymes to work synergistically to depolymerize the complex network of bonds in lignocellulosic biomass. Some anaerobic bacteria produce cellulose degrading macromolecular complexes, known as cellulosomes. This multi-enzyme assembly is composed of non-catalytic structural proteins called “scaffoldins” and a suite of CAZymes active on different biomass linkages. The scaffoldins are populated with cellulosomal CAZymes via high affinity interactions between scaffoldin-borne cohesin and dockerin domains that are tethered to the CAZymes.9 The cellulosome is found in two forms, one tethered to the cell via a surface-layer homology (SLH) domain, and another that is secreted and free from the cell.10 Recent super resolution microscopy studies have provided structural detail into the assembly and structure of the cellulosome.9 A paradigm that is in some sense intermediate in scale and function between the monofunctional enzyme paradigm and the cellulosomal paradigm relies on enzymes with multiple CDs and often multiple catalytic functions. These multifunctional CAZymes are most often secreted; but, similarly to cell-bound cellulosomes, some are also found tethered to microbial cells via SLH domains.11 Multifunctional biomass degrading enzymes have been found in bacteria and eukaryotes but have been characterized predominantly in thermophilicbacteria.5 The unique architecture and mode of action of these enzymes often confers them with high activity on model substrates and plant biomass.5 Indeed, the most cellulolytic single polypeptide yet discovered is CelA, a multifunctional enzyme composed of a GH9-CBM3c processive endoglucanase and a GH48exoglucanase interspersed by two CBM3b modules, from the (hyper)thermophilic bacterium Caldicellulosiruptor bescii.12 Over the past decade, the known diversity of multifunctional enzymes has dramatically expanded as has our understanding of their function. Here, we present a survey of the diversity of multifunctional enzymes in the biosphere, with an emphasis on bacterial enzymes. We review the current state-of-the-art in multifunctional CAZyme biochemistry and biophysics. Finally, we present a forward-looking perspective on the industrial applications possible with these multifunctional CAZymes. Continued investigation into these fascinating biocatalysts holds promise for expanding our understanding of biomass deconstruction in Nature, microbially facilitated biogeochemical cycling, and for inspiring novel strategies for the conversion of biomass to biofuels, bio-derived chemicals, and advanced biomaterials.

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems

3.02.2

17

Natural Diversity of Multifunctional Enzymes

Multifunctional enzymes have evolved to take advantage of synergism and colocalization between two or more catalytic activities and accessory modules. Depending on the composition and chemistry of the biomass substrate encountered in the native environment, microorganisms are equipped with a diversity of multifunctional enzymes with varying domain architectures and combinations of activities. Although the majority of known multifunctional CAZymes originate in bacteria, they are also found in eukaryotes. Berlemont and co-workers, in their most recent analysis of sequenced bacterial genomes, reported 217 proteins with multiple GHs out of 40,729 identified proteins with at least one CAZyme.13 Using a hidden Markov model (HMM) profile based bioinformatic pipeline, they were able to identify proteins across all sequenced bacterial genomes that were involved in carbohydrate targeting and processing. Those proteins that were identified as containing putative GHs by the HMM were confirmed against the PFAM protein family database and their substrate specificity retrieved from the Carbohydrate-Active enZYmes Database (CAZy). The architectures of the bioinformatically classified multifunctional enzymes are not highly conserved across genera.13 Bacterial biomass degraders are the most prevalent reservoir of multifunctional CAZymes and have a correspondingly large diversity of enzyme domain architectures and functionalities. One hundred and five of the 217 identified bacterial multifunctionals contain GH domains from the same family (homo-GH), whereas 112 of these contain GH domains from different families (heteroGH).13 In addition to divergent GH architectures, there exists a wide range of combinations of substrate specificities within a given multifunctional enzyme. Within a given substrate specificity class, different combinations of enzymatic mechanisms also serve to unlock cooperative activity not possible in monofunctional enzymes. For example, some cellulase–cellulase enzymes contain multiple CDs with endoglucanase activity, whereas others contain both exo- and endoglucanase activities. Following cellulase– cellulase enzymes in abundance are those containing two chitinases, and then enzymes containing two xylanases.13 There are also enzymes targeting various combinations of hemicelluloses and various combinations of cellulose and hemicelluloses. Beyond glycoside hydrolases, some enzymes have been found that contain hemicellulose targeting GHs coupled to carbohydrate esterases, which remove polysaccharide side chains, thus increasing the accessibility of glycoside bonds to glycoside hydrolases.14 Although multifunctional genes have been found in a number of bacterial phyla, the genus Caldicellulosiruptor is a reservoir of multifunctional enzymes. Caldicellulosiruptor species are gram-positive lignocellulose degrading bacteria that are found in geothermally active terrestrial environments across the globe. The genomes of 13 Caldicellulosiruptor species have been sequenced to date, which include the most thermophilic bacteria capable of lignocellulose utilization.15 Recent analysis of the Caldicellulosiruptor pangenome has indicated that it is open, supporting the hypothesis that horizontal gene transfer is important for determining the arsenal of environment-specific multifunctional CAZymes in a given species.15 A similar HMM profile analysis of fungal genomes also identified multifunctional enzymes, although the frequency and diversity was greatly reduced in comparison to bacterial genomes.16 The reduced abundance of multifunctionals could reflect the presence of oxidative enzymes in fungal genomes. Homo-GH multifunctionals are highly enriched in fungal genomes compared with heteroGH enzymes. The fungal genus Orpinomyces is a clear exception as it encodes for several multifunctional CAZymes, mostly containing GHs of family 10 and 11.16 The approach detailed by Berlemont and co-workers is relevant for discovering polysaccharide targeting proteins in the ever-increasing number of sequenced genomes for both culturable microorganisms, microbiota, and environmental consortia.17 New approaches have resulted in an increase in identified proteins over the 64 multifunctionals reported in 2011 by Xu et al14; the most notable examples are depicted in Table 1, others can be found in Ref. 16. Despite the diversity of multifunctional CAZymes distributed throughout the tree of life, few have been characterized as fulllength enzymes (Table 2) and fewer still have been investigated for industrial potential. Mining genomes and metagenomes for biomass targeting enzymes may unlock multifunctional CAZymes with novel modes of action and functionalities in understudied and unculturable taxa.

3.02.2.1

Clues Into Function

As multifunctional enzymes have received more attention in the past decade, numerous studies have explored the characteristics of these intriguing biocatalysts and their unique mode of action. Although biomass deconstruction capabilities and specificities vary widely throughout the Caldicellulosiruptor genus, multifunctional CAZymes are responsible for the deconstruction capabilities of each species.18,19 A common core set of multifunctional CAZymes is found throughout the genus, although each species contains a different set of enzymes that are relevant for the biomass substrates encountered in the respective native environment. The presence of certain multifunctional CAZymes within a species is correlated with high cellulolytic ability.20 In two highly cellulolytic species, Caldicellulosiruptor bescii and Caldicellulosiruptor obsidiansis, the most highly secreted multifunctional is CelA, composed of an N-terminal GH9-CBM3c processive endoglucanase, two CBM3b domains, and a C-terminal GH48 exoglucanase, interspersed by threonine/proline rich linker peptides.15,21 CelA was first isolated from a hyperthermophilic bacterium Anaerocellum thermophilum (now C. bescii) in 1998 by Zverlov and co-workers and demonstrated to have cellulase and weak xylanase activity.22 Caldicellulosiruptor bescii was found in terrestrial hot springs containing plant material in the Valley of Geysers, a geothermal hot spot along the Pacific Rim in Kamchatka, Russia.22 CelA was found to be the product of a single gene, representing the first multifunctional CAZyme isolated from its native host.22 CelA is the most important enzyme for lignocellulose depolymerization in Caldicellulosiruptor species. A mutant C. bescii strain with a CelA deletion was constructed to investigate its role in carbohydrate deconstruction.23 After 96 h of growth on Avicel, a model

18

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems Table 1

Examples of fungal and bacterial multifunctional CAZymes with novel domain architectures

Domain architecture

Gene

Source organism

CBM1-GH5-GH6 GH3-Fn3-CBM10-GH6-CBM10-CBM10 GH11-CBM10-GH11-CBM10-GH11 AA9-AA9 GH48-GH9-CBM3-CBM GH5-CBM2-GH6 GH5-GH5-CBM11-doc SLH-SLH-SLH-GH18-GH18-GH18 GH5-GH5 CBM4-CelD-GH6-GH5 GH18-GH8-TSP3-TSP3 doc-GH9-GH6 GH5-fascin-GH18-fascin CBM4-CBM4-GH10-GH10-CBM9 GH48-Big2-GH48-Big2 GH11-GH11-GH11 GH11-GH10-GH10 GH11-CBM5-GH5 GH9-CBM3-PKD-PKD-PKD-PKD-GH48 CBM22-GH10- CBM22-CE1 GH11-CBM6-GH10

Sisni1_485627 Orpsp1_1176522 Orpsp1_1176522 Ompol1_4884 Clocl_3038 NocChr264435_4426 CloThe140364_2441 PaeSp21646_2557 PseOle229656_3973 KriCat249956-1146 ThiNiv169169_1596 ActMis238267_2628 ChiPin99107_2562 BacHem325102_4465 CytFer165886_3007 RumFla296891_0962 RumFla311330_0842 TerTur246850_0350 AquLat268023_3217 HVS_04850 HVS_00470

Sistotremastrum niveocremeum Orpinomyces sp. Orpinomyces sp. Omphalotus olearius Clostridium Clariflavum Nocardiopsis chromatogenes Clostridium thermocellum Paenibacillus sp. HGH0039 Pseudomonas oleovorans Kribbella catacumbae Thiothrix nivea Actinoplanes missouriensis Chitinophaga pinensis Bacillus hemicellulosilyticus Cytophaga fermentans Ruminococcus flavefaciens Ruminococcus flavefaciens Teredinibacter turnerae Aquimarina latercula Hungateiclostridium saccincola Hungateiclostridium saccincola

Adapted from Refs. [13, 16].

Table 2

Multifunctional CAZymes characterized as full-length constructs

Domain architecture

Gene

Source organism

GH9-CBM3-CBM3-CBM3-GH48 GH9-CBM3-CBM3-CBM3-GH5 GH10-CBM3-CBM3-GH48 GH5-CBM3-CBM3-GH44 GH5-CBM3-CBM3-CBM3-GH5 GH74-CBM3-CBM3-GH48 GH12-GH5-CBM3-CBM3-CBM3-GH48 GH9-CBM3-CBM3-CBM3-GH44 GH74-CBM3-CBM3-CBM3-GH44 GH10-CBM3-GH12-GH48

Cbes_1867 Cbes_1865 Cbes_1857 Cbes_1859 Cbes_1866 Cbes_1860 Cmor_0543 Cmor_0545 Cmor_0943 Cdan_2053

Caldicellulosiruptor Caldicellulosiruptor Caldicellulosiruptor Caldicellulosiruptor Caldicellulosiruptor Caldicellulosiruptor Caldicellulosiruptor Caldicellulosiruptor Caldicellulosiruptor Caldicellulosiruptor

References bescii bescii bescii bescii bescii bescii morganii morganii morganii danielii

12, 22, 24, 26, 32, 49 26, 32, 34 26, 32 26, 32 26 26 35 35 35 35

microcrystalline cellulose substrate, the CelA deletion strain had a 77% reduction in overall cell count compared with the parent strain. Interestingly, growth was less abated on lignocellulosic biomass from the important feedstock species Populus trichocarpa, Panicum virgatum, and Arabidopsis thaliana, with 38%, 20%, and 27% reduced cell counts compared to the parental strain, respectively.23 To determine the contribution of the domains in CelA to biomass degradation, the secretome of C. bescii was incubated with Avicel and carboxymethylcellulose (CMC), a substrate used to estimate endoglucanase activity. In contrast to the 15-fold reduction in sugar release on Avicel by the secretome of the CelA deletion strain relative to the parent strain, the CelA deletion mutant displayed similar sugar release on CMC; this suggests that the exoglucanase GH48 domain is essential for the cellulolytic activity of C. bescii.23 Throughout the genus Caldicellulosiruptor, the presence of GH48 domains is the strongest determinant of the ability to degrade crystalline cellulose.20 The genome of C. bescii contains two other GH48 containing multifunctional enzymes, a GH10/GH48, and a GH74/GH48, suggesting that CelA in particular possesses a unique ability to process crystalline substrates, possibly from the purported synergy between GH9 and GH48 domains. Caldicellulosiruptor bescii CelA is the most active single-gene cellulase yet discovered. The activity of CelA on the crystalline cellulosic substrate Avicel was found to outperform that of a combination of the most prevalent industrial exoglucanase, Trichoderma reesei Cel7A, and the Acidothermus cellulolyticus Cel5A endoglucanase.12 These results suggest that the combination of endoglucanase and cellobiohydrolase activities in a single polypeptide most likely results in an intramolecular synergy not observed when the activities are in separate proteins. Transmission electron microscopy imaging of Avicel particle surface morphology after digestion revealed that CelA has a unique mode of action wherein it excavates cavities into cellulose and into biomass.12,24 Despite its prolific ability to deconstruct Avicel and its unique mode of action, CelA had widely varying performance on complex lignocellulosic biomass.12 Lignin content is a significant barrier to efficient hydrolysis of biomass by CelA; pretreatment chemistries that reduce

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems

19

lignin content allowed more effective deconstruction by CelA.24 Enzyme adsorption to lignin is responsible for the loss of CelA activity on highly lignified substrates. On the other hand, CelA is agnostic to cellulose crystallinity; exhibiting similar extent of sugar conversion on low crystallinity and high crystallinity cellulosic substrates, whereas TrCel7A is inhibited by high crystallinity. The insensitivity of CelA to cellulose crystallinity has not been demonstrated in any other enzyme. Given that many multifunctional enzymes, including CelA, operate at elevated temperature, it is worth asking whether their elevated activity is due to the catalytic Arrhenius effects of thermal energy or intrinsic differences in the mechanism and mode of action dictated by their architecture. Comparison of CelA and TrCel7A Avicel conversion at 50  C, the optimal temperature of TrCel7A, demonstrated that CelA still achieves higher levels of conversion.24 Thus, Arrhenius effects are not solely responsible for the exceptional activity of CelA. As early as 1998, it was observed via SDS-PAGE that CelA exhibited a higher molecular mass than predicted by amino acid sequence alone; this discrepancy was attributed to post-translational modification, likely glycosylation.22 Further evidence for glycosylation is provided by the remarkable stability of such a large enzyme containing long, intrinsically disordered linker regions. Staining of C. bescii CelA via a glycoprotein specific Periodic Acid-Schiff stain revealed that the full-length protein homologously expressed in the native host is glycosylated, along with other enzymes in the C. bescii secretome.25 Many multifunctional CAZymes contain such linker regions, implicating an important role for glycosylation in the maintenance and modulation of the biochemical and biophysical properties of these fascinating biocatalysts. Although the function of glycosylation in individual multifunctional CAZymes has not yet been investigated in detail, two recent studies have investigated the role of glycosylation in the secretome of C. bescii.26,27 Bioinformatics analyses conducted by Russell et al.27 highlighted a putative membrane bound Oglycosyltransferase (Cbes_1864) located in close proximity to CelA. These two studies demonstrated that a knockout of this single glycosyltransferase abolishes glycosylation of the C. bescii secretome, including multifunctional CAZymes. Russell et al. found that CelA and the other multifunctional CAZymes secreted by C. bescii are unstable in the glycosyltransferase knockout strain and exhibit dramatically reduced molecular weight, possibly due to proteolytic cleavage.27 This result is in accordance with the paradigmatic role of glycosylation in conferring stability to proteins. Conway et al. found that the Cbes_1864 knockout did not result in diminished crystalline cellulose hydrolysis after 7 days relative to the parent and wild-type strains.26 In contrast, Russell et al. found that their Cbes_1864 deletion exhibited a 77% decrease in growth on crystalline cellulose after 24 h, nearly matching the 78% decrease exhibited by a CelA deletion mutant.27 To confirm that polar effects were not responsible for their observations, Russell et al. utilized a complementation plasmid which restored glycosylation and stability to CelA.27 Although fungal cellulase glycosylation has been well studied,28 investigation into the type and role of glycosylation in bacteria and their multifunctional CAZymes will deepen our understanding of their biochemical and biophysical properties, while enabling glycoengineering approaches.29 The biomass deconstruction potential of Caldicellulosiruptor species varies widely and is controlled by the combination of multifunctional CAZymes in the species’ secretome. Within the genome, multifunctional cellulolytic enzymes are colocalized in one locus, termed the glucan degradation locus (GDL).18 Annotated glycoside hydrolases in Caldicellulosiruptor genomes vary from 37 to 77.15 Only some Caldicellulosiruptor species are capable of microcrystalline cellulose degradation; yet the sheer number of GHs does not appear to control this ability. Rather, the presence of GH48 domains, and CelA in particular, within a given species is requisite for high cellulolytic capability.15 The genomes of the most highly cellulolytic species (C. bescii, C. kronotskyensis, C. danielii, C. morganii, and C. naganoensis) each have three GH48 domains and also a comparatively high number of CBM3 domains. Species with weak or moderate cellulolytic capability lack the same abundance of GH48s and CBM3s.15 The cellulosic substrates present in a given environment further dictate the composition of the C. bescii secretome.30 The repetitive nature of the individual multifunctional CAZymes and the GDL with its many GH48s, GH5s, CBM3s, and highly similar linker sequences, points evolution by gene duplication, recombination, and domain shuffling.15,31 Other multifunctional CAZymes are key in determining the biomass deconstruction capabilities of Caldicellulosiruptor spp. Including CelA, the GDL of C. bescii contains six multifunctional CAZymes.26 The three next abundant after CelA are CbCel9B/ Man5A (Cbes_1865), CbXyn10A/Cel48B (Cbes_1857), and CbMan5B/Cel44A (Cbes_1859).32 CbMan5B/Cel44A and CbCel9B/ Man5A both possess residual cellulase and xylanase activity, but are predominantly mannanases as they show high activity on para-nitrophenol mannose (pNP-mannose), and azurine cross-linked-labeled mannan (AZCL-mannan).32 Previous studies had also shown that these enzymes were primarily mannanases, with some degree of intramolecular synergy in the case of CbCel9B/ Man5A.33,34 CbXyn10A/Cel48B is primarily a xylanase, a substrate specificity granted primarily by its GH10 module, although the enzyme’s GH48 module may also contribute residual xylanolytic activity as previously shown for another GH48 in a multifunctional.32 In addition to these, there also exists CbMan5C/Cel5A (Cbes_1866) and CbCel74A/Cel48C (Cbes_1860). CbCel74A/ Cel48C demonstrates high activity on glucan substrates and low but detectable activity on xylan and mannan substrates.26 Recently, four uncharacterized multifunctional CAZymes from C. morganii and C. danielii were expressed in C. bescii to examine their properties.35 Two of these contained two catalytic domains:CmCel9A/Cel44A(Cmor_0545) and CmCel74A/Cel44B(Cmor_0943). The remaining two harbored three catalytic domains: CmCel12A/Man5A/Cel48A(Cmor_0543) and CdXyn10A/ Cel12A/Cel48A(Cdan_2053), the first triple CD CAZymes yet characterized. Notably, these triple CD enzymes are the first Caldicellulosiruptor GH12s. All of the enzymes showed high activity on glucan substrates, whereas only Cdan_2053, Cmor_0545, and Cmor_0943 had xylanase activity. Of these enzymes, Cmor_0543 was the highest performer. Cmor_0543 displayed 75% activity on crystalline cellulose relative to CelA and retained quite similar performance to CelA on other glucan substrates. It also had some activity on mannan. Further characterization of these enzymes, particularly the benefits of triple CD enzymes over those containing only two CDs, is needed.

20

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems

Multifunctional CAZymes are not only found free in the secretome, but also in a cell-tethered form. The surface layer (S-layer) is a two-dimensional protein array surrounding the cell that is composed of self-assembling surface layer proteins. Many bacterial proteins can associate non-covalently with the surface layer and its component proteins via N- or C-terminal SLH domains. SLH domains are important in bacterial lignocellulose utilization. In bacteria that utilize cellulosomes, SLH domains anchor noncatalytic scaffoldin modules to the cell surface.9 Surface display of cellulosomes and other biomass active enzymes serves to colocalize degradation enzymes with the cell surface; this mechanism may be partially responsible for enzyme-microbe synergy not seen in in vitro CAZyme cocktails.5 In order to gain the advantages given by proximity to growth substrate, Caldicellulosiruptor species associate with the substrate via non-catalytic proteins known as tapirins.36 Another mechanism evolved by Caldicellulosiruptor to localize with lignocellulosic substrate relies on proteins that bind to the S-layer via SLH domains. Two S-layer associated Caldicellulosiruptor saccharolyticus proteins were shown to bind insoluble cellulosic substrates.37 Caldicellulosiruptor kronotskyensis contains 77 annotated glycoside hydrolase domains30 and 19 proteins containing SLH domains,11 both the highest of any sequenced Caldicellulosiruptor species. C. kronotskyensis encodes the largest CAZyme in the genus Caldicellulosiruptor; Calkro_0111, composed of 2435 amino acids is a multifunctional GH16/GH55 composed of 15 domains. To date, this is the only multifunctional SLH-containing CAZyme that has been characterized. Calkro_0111 contains three N-terminal SLH domains and two C-terminal CBM32s which are interspersed by one CBM54, one GH16 endoglucanase, several fibronectin type III domains and actin-crosslinking-like RICIN superfamily domains (ACL), and one GH55 exoglucanase.11 As predicted by the presence of the SLH domains, the enzymes were found localized on the cell surface via transmission electron microscopy and immunofluorescence microscopy. Just as CBMs localize enzymes on cellulosic substrates, ACL domains may localize enzymes onto biomass, in particular the actin filaments produced by algae and higher plants.38 The presence of ACL domains which combined with the GH set in this multifunctional enzyme suggests activity on b-1,3-glucans, which are often found in algae and lichen. The fact that this enzyme is unique to C. kronotskyensis coupled with genetic constraints on maintaining such a large gene suggests that it evolved to uniquely dismantle polysaccharides from organisms endogenous to its native environment of Kamchatka, Siberia. Structural characterization of multifunctional enzymes such as Calkro_0111 may reveal the architectural parameters that endow it with its unique functional ability, opening a route to engineering multifunctional chimeras designed to dismantle particular substrates. A key question regarding the efficiency of multifunctional enzymes is whether monofunctional cellulases can improve their deconstruction potential. Expression of the A. cellulolyticus E1 endoglucanase in C. bescii improved the cellulolytic capability of its secretome.39 Activity against CMC, measured by assaying sugar release, increased by 10.8% and 12.6% at 65  C and 75  C, respectively, while an increase in sugar release on Avicel of 17.5% and 16.4% was observed at those temperatures.39 These results suggest that the ability to generate reducing ends might be a limiting factor in the efficiency of the C. bescii secretome. One study attempted to express heterologous family 10 xylanases from A. cellulolyticus in C. bescii, which resulted in a dramatic increase in growth on xylan substrates, but only a very slight increase in in vitro secretome activity on xylan.40 Another study attempted to express two genes from A. cellulolyticus, a GH6-CBM3-GH12-CBM2 (Acel_0615) multifunctional and the E1 endoglucanase, in C. bescii, and the multifunctional Acel_0615 improved the activity of the secretome on CMC and Avicel when combined with the endoglucanase E1.41 It would be interesting to study whether these results hold true on native plant biomass. When Acel_0615 and E1 were added as a cassette to the secretome, an increase in sugar release of 25% on CMC and 37% on Avicel was observed.41 These studies highlight the fact that despite the catalytic strength of multifunctional enzymes in C. bescii, room for engineering further synergy at the enzymatic and organismal levels exists. It is interesting to think about the evolution of large, complex gene products like these multifunctional enzymes. Most of the multifunctional enzymes known have been discovered in thermophilic and hyperthermophilic species. It is possible that in addition to selection for high, synergistic activity on biomass, constraints imposed by high temperature environments have guided the evolution of these enzymes. Multi-domain proteins in general are abundant across the domains of life. Often, domains from these proteins are not able to fold in isolation.42 Studies of some multi-domain proteins have suggested that the nature of interactions between domain interfaces lends additional stability to these proteins.42–44 Although it is yet unknown whether such effects are at play in the observed stability of multifunctional CAZymes, research to elucidate this possibility would prove insightful. Beyond kinetic and thermodynamic stability, the architectures of multifunctional CAZymes might possess the advantage of colocalization of active sites in high-temperature environments with increased Arrhenius reaction coefficients. Indeed, the presence of multiple CBMs on many multifunctional CAZymes would seem to serve to limit these effects.21 Further study of the fundamentals of multifunctional CAZymes, their role in microbial consortia, and their applicability to industrial development is needed. Gleaning insight from current and future studies will improve our understanding and ability to engineer these diverse and interesting biocatalysts.

3.02.3

Industrial Applicability of Multifunctional Enzymes

Multifunctional enzymes and their catalytic potential and unique mode of action offer the nascent advanced biofuels industry a substantial efficiency benefit. One of the most promising schemes for biofuel production is Consolidated Bioprocessing (CBP), wherein saccharification of biomass and fermentation to a fuel molecule are co-performed by a single microorganism.45 Organisms that couple the native ability to deconstruct plant biomass and the native ability to produce useful fuel or chemical feedstock molecules in high yields and titers are not known in nature. The most successful strategy for engineering a CBP host organism

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems

21

has been to engineer metabolic pathways in cellulolytic organisms to produce aimed products. Introducing the ability to deconstruct complex lignocellulosic substrates to high-titer fuel producers is another potential route to developing host organisms for CBP applications. Furthermore, improving the deconstruction potential of already cellulolytic microorganisms complements metabolic engineering of these organisms for CBP host development. The current bioethanol industry relies on canonical fungal enzyme formulations that have not seen much improvement beyond incremental steps in a decade. Multifunctional CAZymes offers several benefits for improving the deconstruction potential of microorganisms or industrial enzyme formulations. The high activity and synergism inherent to multifunctional CAZymes endows them with the ability to deconstruct multiple biomass linkages more effectively than an equal number of monofunctional CAZymes. Thus, introducing genetic cassettes composed of genetically tractable multifunctional CAZymes in selected microorganisms might allow greater biomass deconstruction. Indeed, a cassette consisting of the four most highly expressed multifunctional enzymes in the secretome of C. bescii exceeds the deconstruction potential of the secretome itself on a variety of plant biomass substrates.32 This cassette represents a proof of principle that genetically tractable numbers of genes may be used to confer or improve cellulolytic ability in microorganisms that are already promising for their metabolic or cellulolytic properties. As biomass substrates differ in composition and diversity of chemical linkages, different strategies are needed in a biorefinery setting to most effectively deconstruct biomass. Cassettes based upon multifunctional enzymes may be tailored for optimal activity on different types of biomass. Further investigation into inter-molecular synergy between multifunctional CAZymes and intramolecular synergy within a given multifunctional CAZyme will inform construction of more efficient cassettes. Despite the catalytic prowess of multifunctional enzymes from Caldicellulosiruptor spp., engineering space exists for improving and tailoring the functionality of these enzymes. Secretion of multifunctional enzymes in C. bescii is thought to occur via the highly conserved Sec pathway, wherein a positively charged N-terminal region containing a signal peptide and a polar C-terminal region targets a protein for export and translocation.46 It has been demonstrated that altering non-signal peptide areas of the N-terminal region affect transport and catalytic efficiency of secreted enzymes.46 To investigate whether CelA could be improved, tags consisting of varying numbers of repeated aspartate residues were inserted on theN-terminal ends of the GH9 and GH48 domains.46 Digestion assays of engineered C. bescii secretome on the model cellulosic substrates CMC and Avicel demonstrated increased sugar release over the WT secretome.46 Engineering the N-terminal end of GH9 with nine aspartate residues resulted in an 82% increase in activity on CMC; tags with fewer aspartate residues and tags on the N-terminal end of GH48 also resulted in increased activity.46 Only one mutant, GH48 with five aspartate residues, resulted in increased activity on Avicel, used to assay exoglucanase activity.46 It is possible that the aspartate tags modified the structure and packing of the ordered domains of CelA, thereby modulating stability or dynamics. Caldicellulosiruptor bescii strains expressing the CelA variants with engineered GH9 domains exhibited dramatically higher growth rates on Avicel than WT C. bescii.46 This study reveals that multifunctional enzymes can be engineered for even greater efficiency, despite their strong native abilities. In engineering a CBP microbe, introduction of heterologous cellulase systems to improve the biomass deconstruction capabilities of microorganisms is requisite. Porting heterologous multifunctional CAZymes to non-native organisms brings a number of engineering challenges. The long linker peptides that are common to multifunctional CAZymes are often glycosylated. Glycosylation modulates the biochemical and biophysical properties of these enzymes and plays an important role in defining their activity and optimal conditions. Therefore, non-native expression organisms must be carefully screened and/or engineered to ensure that they possess the appropriate glycosylation pathways for a particular multifunctional to achieve its optimal activity. This also holds true when using traditional expression platforms to produce enzymes for inclusion in commercial cellulase preparations. These expression hosts may not have glycosylation pathways optimized for conferring optimal glycosylation patterns, motifs, and composition to heterologous multifunctional enzymes. Discovering the glycosyltransferases responsible for decorating multifunctional enzymes in their native host may allow for facile heterologous glycosylation of multifunctional CAZymes.27 Further understanding of the linkage between glycosylation and important industrial properties such as stability, longevity, and activity is needed. In particular, detailed structure–function studies of full-length multifunctional enzymes with native and engineered glycosylation can elucidate this relationship. Gleaning insight from biochemically and biophysically characterized multifunctional CAZymes may also inspire artificial CAZyme chimeras modeled after multifunctionals found in nature. Taking inspiration from natural glycosylation patterns, linker properties, and enzyme architectures may allow for the creation of multifunctional chimeras that are tailored to deconstruct the bonds of particular biomass compositions. Further engineering of the glycosylation patterns and glycosylated motifs of these chimeras may unlock functionality under the acidic and thermal conditions potentially encountered in an industrial biorefinery setting. SLH domains may also be incorporated into these engineered chimeras, or fused to a natural multifunctional enzyme to achieve desired targeting of a multifunctional enzyme to a particular biomass substrate. Such an approach would serve to colocalize microorganisms and their hydrolytic enzymes, hypothetically boosting enzyme-microbe synergy. Beyond applications within bioenergy, multifunctional enzymes have shown promise throughout the advanced bioeconomy, including in nanobiotechnology. Much recent research in materials science has been dedicated to materials with high strength to weight ratios, such as carbon fiber and carbon nanotubes.47 Innovation and efficiency in the aerospace, automotive, and construction sectors relies on such materials, yet cost remains a barrier to widespread implementation. Cellulose is a physically strong material; stronger than steel and approaching the strength of carbon nanotubes on a density basis.47 Nanocellulose morphologies, such as cellulose nanocrystals (CNC) and cellulose nanofibrils (CNF) derived from lignocellulose may be useful for a wide range of structural starting materials. Beyond usage as building blocks, CNCs are relevant for photonic films that may be used as a diverse array of sensors.47 Current CNC and CNF production relies on mechanical, enzymatic, and chemical production

22

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems

methodologies. A recent study has demonstrated that the C. bescii secretome outperforms a fungal cellulase system in the production of nanocellulose. The unique mode of action of the multifunctional enzymes, notably CelA, in the C. bescii system facilitated greater total cellulose conversion and sugar release and also improved the monodispersity of the cellulose nanocrystals and cellulose nanofibrils produced.48 Coproduction of biofuels and nanobiomaterials thus offers a promising pathway for future biorefineries to tailor their product stream to meet economic barriers. The synergistic and catalytic benefits of multifunctional cellulolytic enzymes lend promise to a multitude of applications within the advanced bioeconomy. In order to realize and harness the full scope of these advantages, further research is needed into the fundamental chemical and physical parameters governing multifunctional enzyme activity, substrate recognition, and intramolecular synergy. Multifunctional enzymes display diverse types and patterns of glycosylation due to the presence of disordered linker domains. Understanding the roles and patterns of glycosylation will further build fundamental understanding of multifunctional enzyme folding, substrate binding kinetics, and stability. Such fundamental studies will aid in further elucidating the role of these intriguing biocatalysts in the ecosystem and in global carbon cycling. At the applied level, these studies will inform the engineering of effective multifunctional enzyme chimeras and multifunctional enzyme cassettes which have the potential to catalyze the development of the advanced bioeconomy.

Acknowledgments This work was authored in part by Alliance for Sustainable Energy, LLC, the Manager and Operator of the National Renewable Energy Laboratory for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes. Funding provided by the BioEnergy Science Center (BESC) and the Center for Bioenergy Innovation (CBI), from the U.S. Department of Energy Bioenergy Research Centers supported by the Office of Biological and Environmental Research in the DOE Office of Science. A.K. greatly appreciates scholarships received from the Ministry of Immigrant Absorption, Jerusalem, Israel and from the ministry of Foreign Affairs, Paris, France. A.K. is a Sustainability and Energy Weizmann Fellow and is grateful for receipt of a fellowship from the Sustainability and Energy Research Initiative (SAERI) from the Weizmann Institute of Science. E.A.B. is the incumbent of The Maynard I. and Elaine Wishner Chair of Bio-organic Chemistry. The authors would like to acknowledge Daehwan Chung for helpful discussions and insights regarding multifunctional biomass degrading enzymes, Brandon Knott for providing images of CelA models, and Ruoran Zhang and Peter Ciesielski for the image of a nanocellulose film.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.

Himmel, M. E.; et al. Biomass Recalcitrance: Engineering Plants and Enzymes for Biofuels Production. Science 2007, 315 (5813), 804–807. Cosgrove, D. J. Growth of the Plant Cell Wall. Nat. Rev. Mol. Cell Biol. 2005, 6 (11), 850–861. Brigham, J. S.; Adney, W. S.; Himmel, M. E. Hemicellulases: Diversity and Applications. In Handbook on Bioethanol, Taylor & Francis: Washington, DC, 1996; pp 119–141. Lombard, V.; et al. The Carbohydrate-Active Enzymes Database (CAZy) in 2013. Nucleic Acids Res. 2014, 42 (Database issue), D490–D495. Bomble, Y. J.; et al. Lignocellulose Deconstruction in the Biosphere. Curr. Opin. Chem. Biol. 2017, 41, 61–70. Jiménez, D. J.; et al. Ecological Insights into the Dynamics of Plant Biomass-Degrading Microbial Consortia 2017, 25 (10), 788–796. Nguyen, S. T. C.; et al. Function, Distribution, and Annotation of Characterized Cellulases, Xylanases, and Chitinases from CAZy. Appl. Microbiol. Biotechnol. 2018, 102 (4), 1629–1637. Payne, C. M.; et al. Fungal Cellulases. Chem. Rev. 2015, 115 (3), 1308–1448. Artzi, L.; Bayer, E. A.; Morais, S. Cellulosomes: Bacterial Nanomachines for Dismantling Plant Polysaccharides. Nat. Rev. Microbiol. 2017, 15 (2), 83–95. Xu, Q.; et al. Dramatic Performance of Clostridium Thermocellum Explained by its Wide Range of Cellulase Modalities 2016, 2 (2), e1501254. Conway, J. M.; et al. Multidomain, Surface Layer-Associated Glycoside Hydrolases Contribute to Plant Polysaccharide Degradation by Caldicellulosiruptor Species. J. Biol. Chem. 2016, 291 (13), 6732–6747. Brunecky, R.; et al. Revealing Nature’s Cellulase Diversity: the Digestion Mechanism of Caldicellulosiruptor bescii CelA. Science 2013, 342 (6165), 1513–1516. Talamantes, D.; et al. Natural Diversity of Cellulases, Xylanases, and Chitinases in Bacteria. Biotechnol. Biofuels 2016, 9, 133. Xu, Q.; et al. Multifunctional Enzyme Systems for Plant Cell Wall Degradation. In Comprehensive Biotechnology, Academic Press: Waltham, MA, 2011; pp 15–25. Lee, L. L.; et al. Genus-Wide Assessment of Lignocellulose Utilization in the Extremely Thermophilic Genus Caldicellulosiruptor by Genomic, Pangenomic, and Metagenomic Analyses. Appl. Environ. Microbiol. 2018, 84 (9), e02694–e02717. Berlemont, R. Distribution and Diversity of Enzymes for Polysaccharide Degradation in Fungi. Sci. Rep. 2017, 7 (1), 222. Berlemont, R.; Martiny, A. C. Glycoside Hydrolases across Environmental Microbial Communities. PLoS Comput. Biol. 2016, 12 (12), e1005300. Zurawski, J. V.; et al. Comparative Analysis of Extremely Thermophilic Caldicellulosiruptor Species Reveals Common and Unique Cellular Strategies for Plant Biomass Utilization. Appl. Environ. Microbiol. 2015, 81 (20), 7159–7170. Conway, J. M.; et al. Functional Analysis of the Glucan Degradation Locus in Caldicellulosiruptor bescii Reveals Essential Roles of Component Glycoside Hydrolases in Plant Biomass Deconstruction. Appl. Environ. Microbiol. 2017, 83 (24), e01828–e01917. Blumer-Schuette, S. E.; et al. Caldicellulosiruptor Core and Pangenomes Reveal Determinants for Noncellulosomal Thermophilic Deconstruction of Plant Biomass. J. Bacteriol. 2012, 194 (15), 4015–4028. Lochner, A.; et al. Use of Label-free Quantitative Proteomics to Distinguish the Secreted Cellulolytic Systems of Caldicellulosiruptor bescii and Caldicellulosiruptor obsidiansis. Appl. Environ. Microbiol. 2011, 77 (12), 4042–4054. Zverlov, V.; et al. Properties and Gene Structure of a Bifunctional Cellulolytic Enzyme (CelA) from the Extreme Thermophile ‘Anaerocellum Thermophilum’ with Separate Glycosyl Hydrolase Family 9 and 48 Catalytic Domains. Microbiology 1998, 144 (2), 457–465. Young, J.; et al. Deletion of Caldicellulosiruptor Bescii CelA Reveals its Crucial Role in the Deconstruction of Lignocellulosic Biomass. Biotechnol. Biofuels 2014, 7 (1), 142. Brunecky, R.; et al. The Multi Domain Caldicellulosiruptor Bescii CelA Cellulase Excels at the Hydrolysis of Crystalline Cellulose. Sci. Rep. 2017, 7 (1), 9622.

Fundamentals and Industrial Applicability of Multifunctional CAZyme Systems

23

25. Chung, D.; et al. Homologous Expression of the Caldicellulosiruptor bescii CelA Reveals that the Extracellular Protein Is Glycosylated. PLoS One 2015, 10 (3), e0119508. 26. Conway, J. M.; et al. Parsing In Vivo and In Vitro Contributions to Microcrystalline Cellulose Hydrolysis by Multidomain Glycoside Hydrolases in the Caldicellulosiruptor bescii Secretome. Biotechnol. Bioeng. 2018. 27. Russell, J.; et al. Deletion of a Single Glycosyltransferase in Caldicellulosiruptor bescii Eliminates Protein Glycosylation and Growth on Crystalline Cellulose 2018, 11 (1), 259. 28. Payne, C. M.; et al. Fungal Cellulases. Chem. Rev. 2015, 115 (3), 1308–1448. 29. Beckham, G. T.; et al. Harnessing Glycosylation to Improve Cellulase Activity 2012, 23 (3), 338–345. 30. Poudel, S.; et al. The Diversity and Specificity of the Extracellular Proteome in the Cellulolytic Bacterium Caldicellulosiruptor bescii Is Driven by the Nature of the Cellulosic Growth Substrate. Biotechnol. Biofuels 2018, 11, 80. 31. Gibbs, M. D.; et al. Multidomain and Multifunctional Glycosyl Hydrolases from the Extreme Thermophile Caldicellulosiruptor Isolate Tok7B.1. Curr. Microbiol. 2000, 40 (5), 333–340. 32. Brunecky, R.; et al. High Activity CAZyme Cassette for Improving Biomass Degradation in Thermophiles. Biotechnol. Biofuels 2018, 11, 22. 33. Ye, L.; et al. Molecular and Biochemical Analyses of the GH44 Module of CbMan5B/Cel44A, a Bifunctional Enzyme from the Hyperthermophilic Bacterium Caldicellulosiruptor bescii. Appl. Environ. Microbiol. 2012, 78 (19), 7048–7059. 34. Su, X.; Mackie, R. I.; Cann, I. K. Biochemical and Mutational Analyses of a Multidomain Cellulase/mannanase from Caldicellulosiruptor bescii. Appl. Environ. Microbiol. 2012, 78 (7), 2230–2240. 35. Conway, J. M.; et al. Novel Multidomain, Multifunctional Glycoside Hydrolases from Highly Lignocellulolytic Caldicellulosiruptor Species. AIChE J. 2018. 36. Blumer-Schuette, S. E.; et al. Discrete and Structurally Unique Proteins (Tapirins) Mediate Attachment of Extremely Thermophilic Caldicellulosiruptor Species to Cellulose. J. Biol. Chem. 2015, 290 (17), 10645–10656. 37. Ozdemir, I.; Blumer-Schuette, S. E.; Kelly, R. M. S-layer Homology Domain Proteins Csac_0678 and Csac_2722 Are Implicated in Plant Polysaccharide Deconstruction by the Extremely Thermophilic Bacterium Caldicellulosiruptor saccharolyticus. Appl. Environ. Microbiol. 2012, 78 (3), 768–777. 38. Charrier, B.; Le Bail, A.; de Reviers, B. Plant Proteus: Brown Algal Morphological Plasticity and Underlying Developmental Mechanisms. Trends Plant Sci. 2012, 17 (8), 468–477. 39. Chung, D.; et al. Expression of the Acidothermus Cellulolyticus E1 Endoglucanase in Caldicellulosiruptor bescii Enhances its Ability to Deconstruct Crystalline Cellulose 2015, 8 (1), 113. 40. Kim, S. K.; et al. Heterologous Expression of Family 10 Xylanases from Acidothermus Cellulolyticus Enhances the Exoproteome of Caldicellulosiruptor bescii and Growth on Xylan Substrates. Biotechnol. Biofuels 2016, 9 (1), 176. 41. Kim, S. K.; et al. In Vivo synergistic Activity of a CAZyme Cassette from Acidothermus Cellulolyticus Significantly Improves the Cellulolytic Activity of the C. bescii Exoproteome. Biotechnol. Bioeng. 2017, 114 (11), 2474–2480. 42. Bhaskara, R. M.; Srinivasan, N. J. S.r. Stability of Domain Structures in Multi-Domain Proteins 2011, 1, 40. 43. Han, J.-H.; et al. The Folding and Evolution of Multidomain Proteins 2007, 8 (4), 319. 44. Kataeva, I. A.; et al. The Fibronectin Type 3-like Repeat from the Clostridium Thermocellum Cellobiohydrolase CbhA Promotes Hydrolysis of Cellulose by Modifying its Surface 2002, 68 (9), 4292–4300. 45. Lynd, L. R.; et al. Cellulosic Ethanol: Status and Innovation 2017, 45, 202–211. 46. Kim, S. K.; et al. Engineering the N-Terminal End of CelA Results in Improved Performance and Growth of Caldicellulosiruptor bescii on Crystalline Cellulose. Biotechnol. Bioeng. 2017, 114 (5), 945–950. 47. Zhu, H.; et al. Wood-derived Materials for Green Electronics, Biological Devices, and Energy Applications. Chem. Rev. 2016, 116 (16), 9305–9374. 48. Yarbrough, J. M.; et al. Multifunctional Cellulolytic Enzymes Outperform Processive Fungal Cellulases for Coproduction of Nanocellulose and Biofuels. ACS Nano 2017, 11 (3), 3101–3109. 49. Yi, Z.; et al. Molecular and Biochemical Analyses of CbCel9A/Cel48A, a Highly Secreted Multi-Modular Cellulase by Caldicellulosiruptor bescii during Growth on Crystalline Cellulose. PLoS One 2013, 8 (12), e84172.

Ethanol Production From Sugar-Based Feedstocksq

3.03

JOB Carioca, Federal University of Ceará, Fortaleza-Ceará, Brazil; and Ceará State University, Fortaleza-Ceará, Brazil MRLV Leal, Brazilian Bioethanol Science and Technology Laboratory (CTBE)/National Center for Research in Energy and Materials (CNPEM), Sao Paulo, Brazil © 2017 Elsevier B.V. All rights reserved. This is a reprint of J.O.B. Carioca, M.R.L.V. Leal, Ethanol Production From Sugar-Based Feedstocks, Reference Module in Life Sciences, Elsevier, 2017.

3.03.1 3.03.2 3.03.2.1 3.03.2.2 3.03.2.3 3.03.2.4 3.03.3 3.03.4 3.03.4.1 3.03.4.1.1 3.03.4.1.2 3.03.4.1.3 3.03.5 3.03.5.1 3.03.5.2 3.03.5.3 3.03.6 References

Introduction Feedstocks Sugarcane Sweet Sorghum Sugar Beets Molasses Land Availability Scenarios Feedstock Processing for Ethanol Sugarcane Ethanol Sugar extraction Fermentation Purification Energy Generation and Utilization Sugar Beets Ethanol Sweet Sorghum Ethanol Molasses Ethanol Conclusions

24 25 26 26 27 27 27 27 28 28 30 31 31 32 32 33 33 33

Glossary Agricultural yields Quantity of agricultural product produced per unit area (acre, hectare) per year. Donnelly chute Vertical trough that is installed at the inlet of a mill unit to facilitate the prepared cane feeding and to increase the mill throughput. Greenhouse gas (GHG) Gases that are responsible for retaining in the atmosphere part of the energy, as heat, from the solar radiation reemitted from the Earth’s surface. Milling The most common process to extract the juice (with sugars) from sugarcane that is based on squeezing the cane between successive pairs of rolls under pressure. Molasses Residue from the sugar production process that is still rich in sugars and is often used as substrate for fermentation processes. Parity price Biofuel price that makes it economically indifferent to use the biofuel or the equivalent fossil fuel. Stalks Cylindrical juicy trunk of the tall grasses where the leaves are attached. Strike Discharge from the vacuum pans of the mixture of sugar crystals and mother liquor (part of the sirup that did not crystallize), normally called massecuite, at the end of a crystallization process. Utilities Fluids and energy sources used in an industrial plant as auxiliary inputs to the processes (cooling water, hot water, steam, compressed air, electricity, etc.).

3.03.1

Introduction

Biotechnology offers a clean and powerful tool to produce effective results for the production of biofuels to mitigate carbon dioxide (CO2) emissions from fossil fuels. In this sense, there exist two large avenues concerning the use of genetic engineering to modify crops yields as well as microorganism performance related to bioconversion processes. In the next decades, industrial

q

Change History: July 2016. J.O.B. Carioca and M.R.L.V. Leal updated the text and references mostly related to energy and investment data and land use estimates; it also included the recent experiences in Brazil with sweet sorghum integration with sugarcane to extend the sugar mills and distilleries operating period.

24

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-12-809633-8.09129-9

Ethanol Production From Sugar-Based Feedstocks

25

biotechnology expects to face a big challenge concerning the development of new bioprocesses as well as how to improve old ones. This article focuses on the ethanol production from sugar-based feedstocks such as sugarcane, sweet sorghum, sugar beets, and molasses, its potential and competitiveness. At this moment, it is important to point out that the green revolution has promoted enormous increases in agricultural yields, reducing market prices of the feedstocks. On the other side, the large demand of the energetic market on biofuels has established new paradigms for the agricultural production, once the production of biofuels must be developed harmoniously with the traditional food markets without negative environmental and social impacts. In some countries, policy incentives to use or produce bioenergy further added to the demand for agricultural production and lowered the parity price equivalent to the point where other uncompetitive feedstocks became economically feasible. A technological revolution concerning biofuels is in progress to incorporate new feedstocks and residues into ethanol production that do not interfere with those used for food production, as well as policy prices aiming at incentives. As a good example, soon the enzymatic process to convert cellulose polysaccharides into fermentable sugars will become feasible, and more waste materials will be turned into alcohol fuel, such as forest and crop residues and garbage. Unfortunately, these technologies are still at a noneconomic stage, but it is expected that within a few years, they will become feasible. While cellulose raw materials are the most abundant materials on the Earth, sugar-based feedstocks appear to contribute currently to more effective ethanol production in the world. The total world production of ethanol in 2015 from various feedstocks was about 97.2 billion liters according to the renewable fuels association (RFA) [1]. The United States and Brazil accounted for 85% of this total. So, there is an urgent need to increase this production, incorporating new areas and/or new feedstocks to mitigate CO2 emissions. Around the 1960s, the Club of Roma focused on this problem in a much more political and questionable way. They asked: is this process reversible? How is it possible to invert this problem? Recently, two routes have been tracked to mitigate climate change phenomena according to the Intergovernmental Panel on Climate Change (IPCC). The first is to substitute fossil fuels in the electric energy generation by alternative energy, the most important source of CO2 production in global terms; and the second is the production of biofuels to replace fossil fuels use, mainly in transportation systems. Besides the industrial emissions, it is possible in developing countries to observe substantial CO2 emissions due to deforestation and solid waste being discarded in free areas around the cities, where the organic material could ferment and generate methane, a much more dangerous greenhouse gas (GHG). According to the Clean Energy Trends 2014 [2], global production and wholesale revenues of ethanol and biodiesel reached $97.8 billion in 2013 and are projected to grow to $145.6 billion by 2023. The global biofuels production in 2013 was about 89 billion liters of ethanol and 28 billion liters of biodiesel. In Brazil, ethanol production achieved more than 50% by volume of its national Otto cycle automotive transportation fuels from bioethanol around 2005, surpassing gasoline use for the first time, but has decreased its participation to 43% in 2014. An also impressive investment is being projected to wind power generation in 2023 of approximately, $93.8 billion. A larger value has been projected to be achieved through photovoltaic systems, around $158.4 billion by 2023. Differently from the use of alternative sources of energy, its investments and prices, the production of biofuels has other implications, such as land use and availability, food competitiveness, agro-based policies and practices, and parity prices with petroleum-based fuels. The discussion of impacts of biofuels in food prices has become very hot, but more on the emotional side. There are articles and reports showing contradictory conclusions, but some more equilibrated analyses have shown that if done right, bioenergy can contribute both to energy and food security [3].

3.03.2

Feedstocks

Many types of agricultural feedstocks can be converted into ethanol such as fruits, sugarcane, sugar beets, sweet sorghum, molasses, and polysaccharides as starch, inulin, and cellulose. It is important to point out that polysaccharides require their conversion into fermentable sugars a priori to the fermentation process. The most important sugar-based feedstocks are sugarcane, sugar beets, sweet sorghum, molasses, and ripe fruits. Sugarcane grows easily in humid tropics, presenting high yields of sugar per hectare, while sugar beets grow in temperate climate. The types of sugars from sweet sorghum are similar to those in sugarcane, which are present in its stalks, but its sowing occurs through grains, which are cultivated in a period of about of 120–130 days. Besides that, sweet sorghum produces grains that could be utilized for feed or ethanol production. The choice of the adequate raw material for supplying the demand of biofuels in a specific region depends on several factors, such as land quality and availability, labor, climate condition, yield, level of mechanization, cost of production, and parity price with gasoline, among others. Also, the type of feedstock used for ethanol production has implications related with the feedstock storage and the yearly production period in which it is converted into ethanol in an industrial plant. Sweet sorghum, sugarcane, and sugar beets have a short storage life in their harvest period; molasses can be stored for long periods and is easier to transport for long distances. The storage life of the feedstock is then considerably lengthened. However, ripe fruits have a very short storage life, needing to be processed quickly. Traditionally, they have been used to produce fine beverages of many types. Once the gasoline prices increase, the idea of turning agricultural crops into ethanol will be considered a good deal, if crop prices continue depressed, mainly the sugar-based feedstocks. Recently oil prices have plunged to low levels and are persisting around US$50/barrel, creating uncertainties about the future of biofuels in particular and renewable energy in general. However, the Conferences of the Parties at the end of 2015 (COP 21) has created a commitment of the participating countries to reduce drastically the greenhouse gas (GHG) emissions in order to limit the global warming tendency to 2
C. This will require a significant reduction in fossil fuel use, creating a necessity to accelerate the introduction of renewable energy alternatives that will compete among themselves, not against fossil fuels.

26

Ethanol Production From Sugar-Based Feedstocks

3.03.2.1

Sugarcane

Sugarcane is one of 6–37 species (depending on the taxonomic system) of tall perennial grasses of the genus Saccharum (family Poaceae, Andropogoneae). Native to warm temperate to tropical regions of Asia, they have stout, jointed, fibrous stalks that are rich in sugars and measure 2–6 m (6–19 ft) tall. All sugarcane species interbreed and the major commercial cultivars are complex hybrids according to the Wikipedia encyclopedia site (sugarcane). Brazil produces about 40% of the world’s sugarcane, with more than 500 varieties commercially available, but with the main seven varieties occupying around 60% of the cane area. Brazilian production (768 million tons per year in 2013) is aimed at sugar (sucrose) and ethanol, which has an average country productivity of 6700 L ha1 year1. During the period of 1971–97, the average sugarcane production increased 5.5% per year; the total area increased 3.9% per year, and the field productivity 1.6%. Since 1970, Brazil implemented three main genetic improvement programs that are responsible for such good performance. One of these programs was initially developed by Copersucar Technology Centre in Piracicaba, São Paulo (now Sugarcane Technology Center), the largest one by Interuniversity Network for Development of the Sugar/Ethanol Sector (RIDESA), and the third one by Agronomic Institute of Campinas (IAC). It is possible to find more than 3000 genotypes, including a collection of 423 wild species of Saccharum officinarum, 187 species of Saccharum spontaneum, 65 species of Saccharum robustum, 61 species of Saccharum barberi, and 32 species of Saccharum sinense, among others. It is estimated that yearly about 1,420,000 seedlings are produced [4]. According to the estimate of the Food and Agricultural Organization of the United Nations (FAO): Economic and Social Department: The Statistical Division (2013), the 10 top producers (million tons) of sugarcane in the world are Brazil (768,1), India (341.2), People’s Republic of China (128.2), Thailand (100.0), Pakistan (63.7), Mexico (61.1), Colombia (34.9), Philippines (31.9), Indonesia (28.4) and the United States (27.9), totalizing 1898 million tons.

3.03.2.2

Sweet Sorghum

Archeological evidence suggests that early domestication of sorghum might have occurred 5000 years ago or even earlier. After the initial domestication, sorghum spread to West and East Africa. Further human and natural selection, supplemented by introgression with local wild and weedy taxa, lead to the cultivated varieties of sorghum that are included under Sorghum bicolor subsp. bicolor. Sorghum is stated to have further spread from Ethiopia to the Near East and then to India and China by way of the ancient silk routes probably around the beginning of the Christian era. It was introduced into the United States from Africa in 1857. The plants are cultivated in warmer climates worldwide. Species are native to tropical and subtropical regions. Sorghum belongs to the Poaceae family and subfamily of the Andropogoneae [5]. Sweet sorghum has been widely cultivated in the United States for use in sweeteners, primarily in the form of sorghum sirup. By the early 1900s, the United States produced 20 million gallons of sweet sorghum sirup annually. Sweet sorghum (Sorghum bicolor L. Moench) is similar to grain sorghum with a sugar-rich stalk, almost similar to sugarcane. Besides having wide adaptability, rapid growth, and high sugar accumulation and biomass production potential, sweet sorghum is tolerant to drought, water logging, soil salinity and acidity toxicity. It has great potential for sugar, sirup, and alcohol (most importantly gasohol, which is ethanol blended with gasoline) production, according to Wikipedia encyclopedia site (sweet sorghum). The sugar content in the juice extracted from sweet sorghum varies from 16 to 23% Brix. The most common types of sorghum species are those used for grain production. Concerning productivity, there are many varieties of sweet sorghum that are being experimented in different countries. In the United States, the National Association of Sorghum Producers and Processors [6] provides a large variety of sorghum seeds. Most of these developments were studied in connection with American universities (Iowa, Texas, North Carolina, among others). From these studies, it can be observed that the average ethanol productivity from the stalks is around 3000–4000 L ha1 year1. Many varieties of sweet sorghum have been studied to supply the scientific basis to large-scale cultivation as well as the fermentation parameters for alcohol production. The study carried out in China and published by FAO [7] analyzed the relationship that exists among varieties and the sugar content (as Brix), in different stages of growing, according to different species studied: RIO US, most recommended variety; Shennong No. 2 (Shenyang Agricultural University); 6AX1022 (Liaoning Academy of Agricultural Science); Jitian-2 (Jilin Academy of Agricultural Science); Longshi-1 (Heilongjiang Academy of Agricultural Science); and 6AXN249 (Shenyang Agricultural University). It was observed that there are plenty of sugars in the juice of sweet sorghum stem of these varieties. The “enthrone spectrophotometer – test” showed the presence of the following kinds of sugars in the juice of sweet sorghum stem: xylose, ribose, arabinose, fructose, sorbose, galactose, mannose, sucrose, polyglucose, and glucose. The results indicated that Longshi-1 showed higher Brix than other varieties before waxen maturity and along with maturity, the differences of Brix were becoming smaller, as well as that the total sugar content was affected very significantly by varieties and growth stages. Among the varieties, RIO US and Shennong No. 2 indicated higher Brix, while Longshi-1 and 6AXN249 indicated lower. The total sugar content was very high in sweet sorghum stem juice, but not all of the sugars were fermentable. The fermentable glucose, fructose, and sucrose were dominant. Studies were made on nine varieties of sweet of sorghum to know GHG emissions in comparison with sources of lignites that occur in Turkey [8]. In Brazil during the first decade of the National Alcohol Program, sweet sorghum was one of the feedstocks extensively tested, but could not compete with sugarcane. More recently there have been experiences in Center-South region of Brazil to use sweet sorghum to extend the ethanol production in sugarcane distilleries by harvesting sweet sorghum in the cane off season. Taking advantage of the short cycle of the plant, around 120 days, it is being planted in October/November and harvested in February/ March, in the land where the cane ratoon is eradicated to replant the cane of a new cycle, in a kind of rotation cane/sweet sorghum/cane. The targeted minimum ethanol yield for the process to be economically viable (3000 L ha1 year1) is not being easily met, so the interest of the mills is declining, but the effort to develop new, more productive varieties and improved crop

Ethanol Production From Sugar-Based Feedstocks

27

management practices continues [9]. If successful, this system would improve the factory utilization factor by adding 1 or 2 months to the 7 month average operation per year and increasing the ethanol and surplus electricity annual production, improving the sustainability of the Brazilian ethanol.

3.03.2.3

Sugar Beets

Food crops for fuel are a short-term idea. Although beets have been grown as vegetables for feed since antiquity, their use as a sugar crop is relatively recent. As early as in 1590, the French botanist Olivier de Serres extracted sweet sirup from beetroot, but the practice was not widely used. The Prussian chemist Andreas Sigismund Marggraf used alcohol to extract sugar from beets (and carrots) in 1747, but the methods were not conducive to industrial-scale production. Sugar beet (Beta vulgaris L.), a member of the Chenopodiaceae family, is a plant whose root contains a high concentration of sucrose, and it is grown commercially for sugar production. The sugar beet is directly related to the beetroot, chard, and fodder beet, all descended by cultivation from the sea beet. The European Union (EU), produces 70% of the world sugar beet and France is the main producer. Outside EU, the Russian Federation and the United States are important sugar beet producers. Sugar beet accounts for 30% of the world’s sugar production, but less than 1% of the global ethanol production. Sugar beet seems to be an ideal product to turn into alcohol due to its high alcohol-per-hectare yield, Chile is the sugar beet yield champion with 100 t ha1 that could be converted to 10,000 L ha1, and France with some 9000 L ha1 year1 follows closely, but the average EU yield was only 79 t ha1 in 2014. The wild forms of beet from which cultivated forms are thought to derive are seacoast plants of Europe and Asia and are very variable in habit and duration. Original sugar beet contained only about 4% sugar but careful selection and breeding have raised this to a maximum of 20%. Sugar beet plants have white roots of conical shape, growing deep into the soil with only the crown exposed. Many cultivars of sugar beet exist, almost all are capable of giving root yield of approximately 40 t ha1 year1 at 15.5–18% sugar content, giving 6–7 t sugar per hectare. According to Içöz and collaborators [10], the ethanol productivity from sugar beet species in Turkey was around 6662 L ha1 year1.

3.03.2.4

Molasses

Molasses is probably the cheapest feedstock for ethanol production; it is a by-product of the sugar factories so it is available in any sugar-producing region and there is an international trade of this commodity. There are many applications for molasses ranging from beverages, glycerol, acetic acid, baker’s yeast, and lysine through fermentations, and as an animal feed component or even fertilizer. The composition of molasses varies considerably depending on the non-sucrose in the raw juice and processing technology, but it can be considered approximately, for sugarcane molasses, as 75–85% of total solids, 30–36% sucrose, 10–17% (fructoseþglucose), 10–16% ash, and some smaller quantities of polysaccharides, oligosaccharides, organic acids, proteins, and nitrogen compounds [11]. With some 50% content of fermentable sugars, it is indeed a good feedstock for the fermentation process. The sugar beet molasses has a similar composition, but lower concentration of reducing sugars and higher sucrose.

3.03.3

Land Availability Scenarios

According to data on land availability presented by Leite and Leal [12] and available on the FAO-statistics site, the total world land area is 13,400 million hectare (Mha), from which the main uses are arable land¼1400 Mha; perennial crops¼136 Mha; under grassland¼3400 Mha; and under forests¼3900 Mha. An important portion of the unused land is not suitable for cultivation, such as deserts, iced-land, urban areas, and mountains, leaving some 3300 Mha for rain-fed cultivation. Considering these numbers, an International Energy Agency (IEA) study [13] proposed two distinct scenarios to evaluate future land required for biofuels in 2030. In the first, named Reference Scenario (RS), consider that biofuels will meet 4% of world road-transport fuel demand at 2030, up from 1% in 2005. This would require 35.5 Mha, which represent about 2.5% of arable land. In the second scenario, named by Alternative Policy Scenario (APS), consider that biofuels production rises much faster, and biofuels will meet 7% of world roadtransport fuel demand, requiring 52.8 Mha, which represent around 3.8% of the arable land. A more recent study evaluated the biofuels production in the 34 largest producing countries, responsible for more than 90% of the world production, and estimated the area used for the biofuel feedstock growth in the 2000–10 period as 25 Mha. Deducting the credits for the co-products (mostly distillers dried grains for corn and meals for oil seeds) the net land used was 13.5 Mha, representing about 2.4% of the total arable land in the 34 countries (471 Mha) [14]. Doornbosch and Steenblik [15] estimated that by 2050 around 440 Mha would be available for bioenergy production, without compromising the production of food, feed, and fiber. Therefore, in a global context, land availability is not expected to be a problem to meet the demand for biofuels suggested by the IEA scenarios; nevertheless, at the local levels some problems may be encountered especially if distortion in agricultural prices is introduced by subsidies.

3.03.4

Feedstock Processing for Ethanol

With the recent legislation approved in Europe and the United States, it is becoming apparent that differentiation between biofuels is being made. Initially, the main focus was on GHG-emission reduction potential of the biofuel but other issues such as fossil fuel displacement (energy security), impacts on biodiversity, natural resources demand, production costs, and social and environmental

28

Ethanol Production From Sugar-Based Feedstocks

Fig. 1

Main sugar-based feedstocks processing.

impacts need to be addressed. Therefore, the use of high-yield feedstocks and an efficient and low fossil fuel-demanding conversion process are key issues to have a successful biofuel option. In general, the sugar-based feedstocks have an advantage, compared to starch and oil-based biofuels, in terms of agricultural yields, which gives them an advantage in terms of long-term sustainability. In terms of efficiency in the conversion of the substrates into biofuels, there are some similarities among the different options, but with respect to energy efficiency in the processing of the feedstocks, the differences start to appear. The first and most important issue is the availability of residues that can be used as fuel for the processing plant and the second is the energy demand to process the feedstock, in terms of gigajoules and kilowatt-hours per liter of biofuel and the ratio of the biofuel energy to the fossil fuel used in the whole production chain (normally called net energy ratio or NER). Today, all these aspects have been evaluated with respect to the economics of the options, but in the near future, the other sustainability aspects will be included in the evaluations. In this section, the processing of the four main sugar-based feedstocks for ethanol production (sugarcane, sugar beet, sweet sorghum, and molasses) will be briefly described and the key issues in conversion efficiency and energy economy highlighted. For better understanding, the whole process will be divided into sections; some of them are common to any feedstock case: sugar extraction, juice cleaning, fermentation, distillation/dehydration, and energy and effluent/waste disposal. Special attention will be given to the sugar extraction and fermentation processes since they are responsible for the largest part of the distillery sugar losses. Fig. 1 represents the processing of the sugar-based feedstocks in a general way and the differences, mostly in the sugar extraction process, will be pointed out in each case.

3.03.4.1

Sugarcane Ethanol

Sugarcane is the second most used feedstock for ethanol production, after corn, and has a well-developed processing technology, especially with the sections common with sugar production, since ethanol was used in the first days of the automobile, before the gasoline took over entirely. In Brazil, ethanol fuel has been used, in different proportions, since the mid-1920s, and its blending in gasoline became mandatory in 1931 (5%). However, the real technological development took place after the launching of the National Alcohol Program in November 1975 that resulted in the Brazilian model of producing sugar and ethanol in the same plant in a highly integrated process. In this section, an independent distillery will be the focus with minor discussion of the integrated sugar/ethanol processing.

3.03.4.1.1

Sugar extraction

Sugarcane normally arrives at the mills in trucks, and it is unloaded onto a feeding table that has the function of transferring the cane load from the trucks to the conveyors that will carry the cane to the preparation equipment and cleaning it in the process (in the case of whole-cane harvesting). Downstream of the feeding table, the cane preparation equipment are found: one or two sets of knives (rotating shaft with rows of fixed or hinged blades that chops the cane, in the case of whole-cane harvesting, and maintains the cane layer at an uniform height in the conveyor) and a shredder that pulverizes the cane, opening the cells and exposing the sugars to facilitate extraction; the minimum values for open cell percentage (called preparation index) are 82 and 92% when the extraction processes are milling or diffusion, respectively. Approximately 25% of the factory power requirements come from the preparation equipment [11]; therefore, it is a very important item in the energy balance. The sugar extraction process can be through milling or diffusion. Milling (Fig. 2) is more commonly used worldwide, but diffusion (Fig. 3) is used in nearly all mills in South Africa and is becoming popular in the green field mill projects built since 2004 in Brazil. Due consideration must be given to the selection of the technology and the design of the juice extraction system in sugarcane mills because this is the area with the second highest sugar losses and the highest power consumption in the plant, making this system a key component in the factory recovery efficiency and energy optimization process. Considering that milling is the most

Ethanol Production From Sugar-Based Feedstocks

Fig. 2

Modern milling tandem with Donnelly chutes.

Fig. 3

Diffuser.

29

widespread technology in the world and in Brazil (which produces more than 90% of the sugarcane ethanol in the world), it will be described only briefly here. The milling tandem is formed by four to seven milling units, with the optimum number being a tradeoff between investment/operating costs and sugar extraction efficiency. As the prepared cane exits the shredder, it is transported by conveyor to the Donnelly chute that has the double function of pressurizing the cane being fed to the mill, thus increasing the cane flow, and providing a column of prepared cane to be used by the control system to adjust the feed conveyor speed and the rotation of the mill rolls. The prepared cane as it goes through the first mill has a ratio of juice to fiber reduced from around 7 to 2–2.5, making the task of the second mill to extract more juice very difficult; the solution to this problem is to soak the cane with the juice exiting the second mill, in a process called imbibition, that consists of injecting water in the cane between the last two mills in the tandem and then injecting the juice from one mill to the inlet of the upstream mill. Top extraction efficiencies in Brazilian mills are in the range of 97.0–97.5% but the average is around 96%. The power consumption, including the cane preparation system, is in the range of 16–18 kWh t1 of milled cane, representing around 60% of the total power consumed by the plant. The tendency today is

30

Ethanol Production From Sugar-Based Feedstocks

to replace the steam turbine drives with high-efficiency electric motors/inverters or electric motors with hydraulic torque converters; as a result, the capacity to generate surplus power to sell to the national grid is increased considerably. From the milling tandem, two streams emerge: the extracted juice with the sugars and the bagasse. The juice has different types of dissolved and undissolved solids (sand, clay, and bagacillo), colloidal impurities, and microorganisms that could be detrimental to the fermentation yield and robustness, thus making their removal necessary. The juice treatment normally used [16] consists in passing the juice through a series of screens of reducing opening sizes, to remove the coarse impurities, followed by a chemical treatment and a clarifier (settling tank). The first screening step is already integrated with the milling tandem, commonly using a fixed screen (cush–cush) or a rotating screen, and at the end of the screening sequence a hydrocyclone is sometimes used to improve the removal of sand and clay. The chemical treatment is a sequence of operations starting with the addition of milk of lime (Ca(OH)2) adjusting the pH to the 6.8–7.2 range to reduce the formation of organic acids and the decomposition of sucrose, and to form calcium salts that improve the separation of impurities from the juice in the following steps. The juice is heated up to 105–110
C, degassed in a flash tank, and directed to a clarifier with previous addition of polymers to facilitate the coagulation and flocculation of the colloids and nonsugar protein and sedimentation of the flocks. The clarified juice is removed from the top of the clarifier while the mud that contains the impurities and some sugars is pumped from the bottom and directed to the vacuum rotary filter for sugar recovery; bagacillo (fine bagasse particles) is added to the mud as filter aid and most of the mud liquid is removed by vacuum becoming the so-called filtered juice. The filtered juice can be mixed either with the clarified juice or, preferably, with the mixed juice from the mills. The clarified juice is subjected to a process called high-temperature short-time (HTST) sterilization to kill the microorganisms and then subjected to an immediate regenerative cooling counterflow with the mixed juice to around 60
C and a final cooling to 32
C [17]. This juice sterilization system and the equipment and piping between it and the distillery should be designed and built according to the standards of sanitary systems, avoiding flow stagnation areas and bypasses that facilitate the growth of microorganism populations and recontamination. After that, the juice is ready for the concentration step in evaporators to adjust the Brix (concentration of soluble solids) to the 17–22% range, depending on the desired final ethanol concentration in the wine.

3.03.4.1.2

Fermentation

The fermentation process is the critical phase of ethanol production in the factory, where a close control of the process microbiology is necessary and the highest efficiency losses take place. The most widely used process in Brazil is the Melle-Boinot with cell recycle either of the batch type or the multistage continuous fermentation type. The latter is used only in a few distilleries despite its lower investment costs, easier automation, low energy use, and smaller footprint, due to its increased susceptibility to contaminations that are hard to control. In the fermentation, the sucrose is first hydrolyzed to glucose and fructose by the yeast before being fermented according to the simplified Gay Lussac equations [16] indicated in sequence C12 H22 O11 þH2 O/C6 H12 O6 þC6 H12 O6

(1)

C6 H12 O6 /2C2 H5 OH þ 2CO2 þ 23:5 kcal

(2)

Therefore, the stoichiometric conversion of inverted sugars to ethanol and CO2 is 0.511 and 0.489 kg, respectively, per kilogram of sugars. However, according to Pasteur [11] maximum possible yield is 94.5% of the theoretical value, which means 0.483 kg of ethanol kg1 of inverted sugars, due to the formation of other products, besides ethanol and CO2, such as glycerol, organic acids, and yeast mainly. In real fermentations, unfavorable conditions, such as high osmotic pressure and contamination, can increase the amounts of these compounds, and the need to keep the fermentation time at reasonable value will result is some unfermented sugars in the final wine, thus reducing the fermentation yields to values below 91% in most of the cases. The heat generated in the fermentation process must be removed from the fermentors and the temperature kept in the range of 30–34
C. Prior to fermentation, the mash preparation takes place with the adjustment of the Brix to a value compatible with the desired final ethanol concentration in the wine, which should not be below 8% to keep the energy consumption and vinasse production within reasonable values. In the case of most Brazilian mills, the combined production of sugar and ethanol makes it easy to perform this adjustment by means of a combination of mixed juice from the mills with final molasses from the sugar factory. For an independent distillery, this adjustment will require the juice concentration by means of evaporators requiring the use of process steam. The fermentation with cell recycle in Brazil permits the use of higher yeast concentrations, around 12% in volume, that results in shorter fermentation time, and reduces the consumption of sugars to propagate the yeast before each batch, increasing the fermentation yields. At the end of each batch, the yeast is separated from the wine by high-speed centrifuges and is submitted to a treatment with sulfuric acid at a pH around 2.5 for a couple hours before returning to the vat for a new cycle. Fermentation time in Brazil is in the 6–12 h range. In other countries, the cell recycle is not normally used, which requires a prefermenter to multiply the yeast in an aerobic, air blown, fermentation, which consumes sugars and increases the fermentation cycle length; the yeast concentration is also lower than in the case of Brazil and all this add up to a much longer fermentation cycle, normally exceeding 20 h. Contamination is the most serious problem in fermentations; therefore, robustness is something to be pursued; sterile mash, temperature control, low osmotic pressure, healthy yeast, adequate alcohol concentrations, and continuous monitoring of the microbiology are among the important precautions. Concerning the microorganisms that can be used in the fermentation process (yeasts, molds, and bacteria), there exists nowadays throughout the world a large number of standardized culture collections having its biochemistry and taxonomy well known.

Ethanol Production From Sugar-Based Feedstocks

31

Basic information on this matter can be obtained in the World’s Federation for Culture Collections (WFCC), where a number of collections there exist primarily to provide cultures and services for the academic and industrial applications. Other regional culture collections are available at country level, such as MIRCENs, ECCO, and ATCC, among others. All over the world, and especially in Brazil, Saccharomyces cerevisiae is the species most used to convert the main sugars present in the sugar-based feedstocks (sucrose, fructose, and glucose) into ethanol.

3.03.4.1.3

Purification

The separation of the ethanol from the water is done by distillation and if anhydrous ethanol is the end product, it requires two steps. The first one, the stripping and rectification, delivers the hydrous ethanol with a concentration of 96.5% by volume, which is an azeotropic mixture ethanol/water that does not permit further separation by simple distillation. The distillation process in Brazil uses five columns grouped in two sets: one called stripping column concentrates the ethanol from 8–10 to 40–50% and removes most of the volatile contaminants, and the second, the rectification column takes the ethanol concentration to 96.5% by volume; this is called hydrous fuel ethanol and is used in neat ethanol or flexible fuel car engines. To be blended with the gasoline, anhydrous ethanol is needed. For the removal of additional water to produce the anhydrous ethanol, with a maximum water content of 0.3% by volume, the hydrous ethanol undergoes what is called dehydration process that is normally one of the three types: 3.03.4.1.3.1 Azeotropic distillation A solvent is added to hydrous ethanol, normally cyclohexane, to form a triple azeotropic mixture ethanol/water/cyclohexane with a boiling point lower than anhydrous ethanol. This mixture goes to a distillation column where ethanol is removed from the bottom, at 99.7% concentration, and the remaining azeotropic mixture is removed from the top of the column, condensed, and sent to a settling tank where two phases are formed; the top phase, rich in cyclohexane, returns to the dehydration column and the bottom one, with ethanol and water, is directed to the solvent recovery column. In the solvent recovery column, the mixture is distilled to separate the cyclohexane at the top, which is returned to the dehydration column, and a water/ethanol mixture that is directed to the rectification column. 3.03.4.1.3.2 Molecular sieves A bed of zeolite beads is the dehydrating agent. Ethanol in the vapor form leaving the rectification column is heated more and made to flow through the zeolite bed where the water is retained in the bead porous. After the bed is saturated with water, the regeneration is performed using ethanol vapor to strip the water from the bed and the ethanol/water vapor mixture is sent to distillation. 3.03.4.1.3.3 Solvent extraction Monoethylene glycol has a strong affinity for water and it is added at the top of the dehydration column and it strips the water vapor leaving anhydrous ethanol to be condensed at the top of the column and solvent/water mixture is removed from the bottom [8] and sent to a separation column and the solvent is recycled. Distillation and dehydration are the largest process steam consumers in an ethanol plant; in the case of sugarcane where the bagasse provides for all the energy needs, this is of little concern and steam consumption is on the order of 4.5–5.0 kg of steam (at 2.5 bar, saturated) per liter of anhydrous ethanol, while grain and beet ethanol plants that use external fuels are much more efficient and steam consumption is normally below 2 kg of steam per liter anhydrous ethanol. The technology for more efficient distillation and dehydration is available and mature, but the higher investment needed must be justified by economic reasons, as is the case where external fuel is used, but they have not attracted much interest in Brazil where the free bagasse is the only fuel. This situation may change in the future where better economic uses for bagasse, such as surplus power generation and cellulosic ethanol, become attractive; then multipressure distillation and molecular sieve dehydration will become the technologies of choice for sugarcane ethanol.

3.03.5

Energy Generation and Utilization

The energy area in sugarcane ethanol distilleries is a very important sector due to the high investments required and the growing importance of surplus power generation in several countries, including Brazil. Under average conditions, the primary energy of sugarcane is divided as one-third for the sugars and two-third for the fibers; the latter equally divided between the fibers in the stalk and in the leaves. Today, the sugars are used to produce ethanol, the stalk fibers are inefficiently used to produce all the energy needed to process the sugarcane, and the leaves are totally wasted, either burned in the preharvest or left in the field to decompose. In some countries, there is a slow but steady trend to improve the primary energy conversion to useful products, especially in the energy area (second-generation fuels and electricity), but within the biorefinery concept, the range of options is quite broad. The sugarcane mills, whether for ethanol, sugar, or ethanol/sugar production, have evolved in the past three decades mainly to become self-sufficient in energy, meaning that no external fuel, nor external electricity, would be needed to run the industrial plant. Operating in the cogeneration mode, where high-pressure steam coming from the bagasse fired boilers is used in back-pressure steam turbines to drive the electric generators and main equipment in the plant (knifes, shredder, mills, boiler fans, and large pumps) and the turbine exhaust steam at around 2.5 bar pressure provided for the thermal energy needed for juice concentration and

32

Ethanol Production From Sugar-Based Feedstocks

Table 1

Average main sugar-based feedstocks parameters

Parameters

Units

Sugarcane

Sorghum

Sugar beets

Molasses

Feedstock yield Ethanol yield Production cost References

t ha1 L ha1 $ L1 NA

65.0 4550 0.17–0.33 [15,18]

50 3000 NA [6,19]

46.0 5060 0.42–0.55 [18,15]

NA 260 NA [8]

Abbreviations: NA, not available or not applicable.

heating, distillation, dehydration, and other uses. The equilibrium point where nearly all begasses are consumed and all thermal and electromechanical energy demands are met was reached with a high-pressure steam condition of 22 bar/300
C and process steam consumption around 500 kg t1 of sugarcane processed; this is today the most popular situation around the world. Still operating in the cogeneration mode, surplus power can be produced for sale by increasing the boiler pressure/temperature conditions to values such as 42 bar/400
C, 65 bar/480
C, and 100 bar/520
C with increasing amounts of surplus power and increasing investment costs; the local electricity sale price and equipment cost will dictate the best economic or strategic choice. At this point, it is also important to consider the plant-related energy consumption and technologies to reduce the internal energy consumption, such as electric mill drives, multipressure distillation, molecular sieve dehydration, membrane separation, condensation/extraction turbine generators, and, last but not least, recovery and use of the sugarcane tops and leaves left in the fields after unburned cane harvesting. Several countries are moving in this direction, notably Brazil, India, Mauritius, Reunion Island, and Guatemala due to the specific local conditions. In Brazil, the steady increase of green cane harvesting (without burning) is producing significant amounts of sugarcane straw (normally called trash) that is being collected and used to increase power generation, still in an incipient way. In the future, cellulosic ethanol production may use this residue as a preferential feedstock. The main effluents in a sugarcane ethanol distillery are vinasse and filter cake. The former is the liquid residue from the distillation and the latter is the waste from the rotary vacuum filters of the clarifier mud. Vinasse is produced in large amounts, from 10 to 15 L per liter of ethanol, and is mostly disposed at the cane fields with or without treatment, since it is rich in organic matter and potassium, partially displacing the use of some chemical fertilizers. The filter cake is also rich in organic matter, phosphorus, and other nutrients and used, composted with bagasse and boiler ashes or not, in cane planting, also replacing some amounts of chemical fertilizers. The residual waters are normally mixed with the vinasse for field disposal. The main parameters for the four cases of sugar-based feedstocks for ethanol production are shown in Table 1 and will be discussed in this section.

3.03.5.1

Sugar Beets Ethanol

Except for the extraction and juice treatment processes, the ethanol production from sugar beets is essentially the same as described above for sugarcane. Since no residue is available to be used as fuel, the processing of sugar beets to ethanol tends to use energyefficient processes such as multipressure distillation and molecular sieve dehydration. The juice extraction is performed by diffusion. The beets are sliced into pieces called corsettes to increase the surface area and facilitate the diffusion of the sugars; the diffuser extracts the juice that is very clean and free from suspended solids, making the treatment much easier, or even unnecessary, compared with sugarcane juice. The corsettes are dried in two stages, the first one in a press and the second one in rotary drum-type dryers that consumes a large portion of the plant thermal energy. The dried pulp is normally sold as animal feed. Two main points are important also to differentiate sugar beets and sugarcane as ethanol feedstocks: the much short harvesting period (around 3 months compared with 6–8 months for sugarcane) and the disposal of the vinasse since sugar beets are normally planted in a crop rotation system and the land used for beet crops is not normally owned by the distillery. These topics will not be treated here.

3.03.5.2

Sweet Sorghum Ethanol

Sweet sorghum has similarities with sugarcane and sugar beets: it has a very short harvesting period (around 60 days) like sugar beets and the sugars are in the stalks as in sugarcane. The juice extraction process is quite similar to the sugarcane case, using the same type of equipment and the bagasse can be used likewise as fuel for the whole process. The inverted sugar content is higher than in sugarcane but this is only a problem for sugar production not for ethanol; the sorghum stalks starts to deteriorate much faster than sugarcane, requiring the processing to be done as soon as possible after the harvest. The short agricultural cycle of 110–150 days allows for two crops per year if the weather pattern permits, thus improving the economics of the processing plant. Also the harvesting period can be increased by the use of different varieties with different cycle and sowing time, and/or the use of irrigation. As explained above, sweet sorghum is being tried in Brazil as a supplemental feedstock to sugarcane to be harvested and processed in the sugarcane off-season, to increase the distillery operation period. However, sweet sorghum yields are still below the minimum level to make this option economically attractive. The case of vinasse disposal presents difficulties in a degree between sugarcane and sugar beets.

Ethanol Production From Sugar-Based Feedstocks 3.03.5.3

33

Molasses Ethanol

The ethanol production in a distillery annexed to a sugar factory seems to be the simplest and direct route to start a large-scale ethanol production in any sugar-producing country. This set up takes advantage of a common utility facility, with low-cost electricity, steam, cooling water, and effluent handling and avoids the difficulties in handling and storing a high-viscosity product. Surprisingly, it is not becoming as popular as one would expect, maybe due to established markets for the molasses competing with other uses. Brazil is using this alternative in large scale since the mid-1920s and especially after 1931 when the government mandated a 5% blend of ethanol in the gasoline; after the launching of the National Alcohol Program in 1975, the sugarcane industry expanded quickly and the ethanol/sugar joint production became the Brazilian model; today, approximately 67% of all ethanol and more than 95% of all sugar are produced in sugar factories with fully integrated annexed distillery. With the joint production of sugar, ethanol, and electricity, the commercial risks from price fluctuations are highly reduced. The sugar factory operates with more comfort, requiring less energy and producing a better quality sugar since only two strikes are done, instead of the traditional three strikes and the final molasses is sent to the distillery with a higher purity. The mash preparation is very simple because it is essentially a dilution of the molasses to the proper sugar content and the fermentation and distillation are carried out as mentioned in Fig. 1. On an average, the ethanol yield is about 260 L t1 of molasses.

3.03.6

Conclusions

Sugar-based feedstocks represent the best choice to be considered for ethanol production when considering the existing technologies all over the world, and the net energy balance in the production system places sugarcane and sweet sorghum among the best feedstocks used to ethanol production. This could significantly reduce the use of fossil fuel even in the fields, contributing effectively to positive GHG abatement potential. However, substituting biofuels for fossil fuels in the agricultural equipment would slightly increase the cane production costs. Brazil achieved the most competitive ethanol cost, around $0.25 L1 that prevailed until the initial phases of the fast expansion initiated in 2003; today, this figure is higher due to several reasons, but can be brought down to competitive values compared to gasoline. Sweet sorghum has been so far primarily used for the production of fodder and molasses, mainly in the United States where there is a traditional industrial production well established. Currently, several countries all over the world have chosen it for ethanol production, such as Turkey, China, Thailand, and some Latin American countries, but is has not become an important ethanol feedstock. Sugar beets present the best option in terms of sugar yield per hectare, but unfortunately its energy balance is only slightly positive. Molasses is a by-product of sugar industry that acts as a commodity, but it has been used as a feedstock to the production of ethanol in annexed distilleries in Brazil for a long time. The existence of a large experience already established in many countries concerning the availability of species and yeast strains to process the sugar-based feedstocks does not constitute any impediment to launch the industrial production all over the world. Certainly, qualified human resources to manage the industrial production present some restrictions for some less developed countries. Concerning land availability, the IEA scenarios present some restrictions concerning the substitution to achieve the total world demand of biofuels, but they are expected to reach 16% of global energy use in transport by 2040, with ethanol representing near three-fourths of this amount [20]. The use of by-products of the sugar-based feedstocks, such as bagasse, sugar beets residues, CO2, and the vinasse, constitutes a great concern aiming at improving ethanol yields, feed production, and the reduction of environmental impacts. The establishment of bilateral and multilateral cooperative program between nations could improve the capability for ethanol production in the world, as well as contribute to the reduction of CO2 emissions. It does not make sense to establish a tariff to restrict the world market of ethanol between countries, considering the actual CO2 concentration in the atmosphere.

References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

Renewable Fuels Association, 2016. World fuel ethanol production. Report of the Renewable Fuels Association (RFA), Washington, DC. Clean Edge Inc., 2014. Clean energy trends 2014. Report of the Clean Energy Inc. Portland, OR: Clean Energy Inc. Kline, K.L., Msangi, S., Dale, V., et al., 2016. Reconciling food security and bioenergy: Priorities for action. GCB Bioenergy. https://doi.org/10.1111/gcbb.12366. Macedo, I.C., 2000. A Energia da Cana de Açúcar. UNICA, São Paulo, Brazil. Raemakers, R.H., 2001. Crop Production in Tropical Africa. Directorate General for International Co-operation, Brussels, Belgium. National Sweet Sorghum Producers and Processors Association, 2009. Sources of sorghum seed. Report of NSSPPA. Lexington, KY: NSSPPA. FAO – NRMED, 2009. Integrated energy systems in China, FAO. Report on the Cold Northeastern Region Experience, Rome. Türe, S., Uzum, D., Türe, E., 1997. The potential use of sweet sorghum as a non-polluting source of energy. Energy 22, 17–19. Brazilian Enterprise of Agricultural Research (EMBRAPA). Agroindustrial system of sweet sorghum in Brazil and the Public Private Partnership (in Portuguese), EMBRAPA Document No. 138. Içöz, E., Tugrul, M., Saral, A., Içöz, E., 2009. Research on ethanol and use from sugar beet in Turkey. Biomass and Bioenergy 33, 1–7. Rein, P., 2007. Cane Sugar Engineering. Verlag Dr. Albert Bartens KG, Berlin. Leite, A.L.S., Leal, M.R.L.V., 2008. Biomass future based supply. In: Carioca, J.O.B. (Ed.), Brazilian Network on Green Chemistry: Awareness, Responsibility and Action. Edições UFC, Fortaleza, Brazil, pp. 217–233. International Energy Agency, 2006. World energy outlook 2006, IEA. Report of the International Energy Agency, Paris. Woods, J., Lynd, L.R., Laser, M., et al., 2015. Land and bioenergy. Bioenergy and Sustainability: Bridging the gaps, São Paulo, Brazil, pp. 258–300. Doornbosch R., Steenblik R., 2007. Biofuels: Is the cure worse than the disease? Paper prepared for the Round Table on Sustainable Development at the OECD – Organization for Economic Co-operation and Development, SG/SD/RT (2007), vol. 3.

34

Ethanol Production From Sugar-Based Feedstocks

[16] Copersucar Technology Center, 1999. Fundamentos dos Processos de Fabricação de Açúcar e Álcool; Caderno Copersucar Série Industrial no 020, Piracicaba, Brazil, CTC. [17] Center for Management and Strategic Studies-CGEE, 2009. Bioetanol combustível: Uma oportunidade para o Brasil, Prepared by Interdisciplinary Center of Energy Planning of the University of Campinas (NIPE/UNICAMP), CGEE, Brasilia/Brazil. [18] FAO, 2008. Biofuels: Prospects, risks and opportunities. Report of FAO. Rome: FAO. [19] Horta Nogueira, L.A.H., 2008. Sugarcane-Based Bioethanol. Banco Nacional de Desenvolvimento Econômico – BNDES, Rio de Janeiro. [20] International Energy Agency, 2015. World energy outlook 2015, IEA. Report of the International Energy Agency, Paris.

3.04

Ethanol From Starch-Based Feedstocks

WMM Ingledew, University of Saskatchewan, Saskatoon, SK, Canada; and Lallemand Ethanol Technology, Parksville, BC, Canada Y-H Lin, University of Saskatchewan, Saskatoon, SK, Canada © 2011 Elsevier B.V. All rights reserved. This is a reprint of W.M.M. Ingledew, Y.-H. Lin, 3.05 - Ethanol from Starch-Based Feedstocks, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 37-49.

3.04.1 3.04.2 3.04.3 3.04.3.1 3.04.3.2 3.04.3.3 3.04.4 3.04.4.1 3.04.4.2 3.04.4.3 3.04.4.3.1 3.04.4.3.2 3.04.4.3.3 3.04.4.4 3.04.4.5 3.04.4.6 3.04.5 3.04.5.1 3.04.5.2 3.04.5.3 3.04.5.4 3.04.5.5 3.04.5.6 3.04.6 3.04.6.1 3.04.6.2 3.04.6.3 3.04.6.4 3.04.7 3.04.8 References

Introduction Biochemistry of the Ethanol Process Yeasts Used in the Process Yeast Yeast Formats Yeast Production in an Alcohol Plant Unit Operations Relevant to Ethanol Production Grain Processing Enzymatic Processing and Cooking Fermentation Batch Fermentation VHG Fermentation Continuous Fermentation Alcohol Processing Stillage Processing Carbon Dioxide Production Environmental Requirements in Fermentation pH Temperature Aeration Osmotic Stress Chemical Stress Freedom from Microbial Infection Yield Coefficient and Net Rate Expression Yield and Fermentation Efficiency Yield and Feedstock Yield and Overall Plant Performance Net Rate Expression in Continuous Fermentations Metabolic Flux Analysis Summary

36 36 37 38 38 38 38 38 39 39 39 39 40 41 41 41 42 42 42 43 43 44 44 44 44 45 45 45 46 47 47

Glossary Active dry yeast (ADY) A dehydrated form of yeast – called instant dry yeast (IDY) by the baking industry as it can be used without rehydration - is commonly used by fuel alcohol producers to directly inoculate an ethanol fermentation or to inoculate a propagator/conditioning tank prior to inoculation of the main fermentor. Distillation and dehydration Unit operations designed to concentrate ethanol to 194
US Proof using a still, followed by an adsorption/exclusion process using a molecular sieve to obtain 200
Proof ethanol. Distillers grains Insoluble mash components obtained from the bottom of a beer still and usually separated by centrifugation (called distillers wet grains, DWG). When dried, the feed produced is called distillers dried grain (DDG). When DDG is partially mixed with concentrated thin stillage or backset from the centrifuge (the syrup), it is called DDG with solubles (DDGSs). PDH bypass A metabolic pathway shunting to the reduced metabolic route governed by pyruvate decarboxylase from the oxidized route governed by pyruvate dehydrogenase in order to replenish nicotinamide adenine dinucleotide (NADþ) during ethanol fermentation by yeast.

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00149-X

35

36

Ethanol From Starch-Based Feedstocks

Simultaneous saccharification and fermentation (SSF) A fermentation strategy to minimize osmotic stress resulting from the presence of a high concentration amount of glucose. Yeasts are ‘spoon fed’ by the production of glucose from dextrin at a rate similar to the rate of glucose utilization during anaerobic growth and fermentative production of ethanol. Very-high-gravity (VHG) fermentation A fermentation strategy to increase ethanol concentration in batch fermentation by providing nutrients, optimal conditions, and a high wort concentration of dextrins at 28–38
Brix.

3.04.1

Introduction

The growth of the fuel ethanol industry has been remarkable, and ethanol made by fermentation for fuel and industrial purposes now approaches the total production volume of ethanol for beverage use – based on 200 Proof alcohol. Despite the retrenchment in the fuel alcohol industry in 2008/2009 and the slowdown of new capacity planned, currently, some 7.04  1010 L of capacity exists worldwide for fuel (at virtually 200 Proof), while the distilled beverage industry produces about 1.65  1010 L of 40% alcohol spirits; the beer industry produces 1.04–1.63  1011 L at 5% alcohol; and the wine industry produces about 7.17  1011 L at 14% alcohol. If such quantities are accurate and the average alcohol content in them is reasonably averaged, the potable ethanol industry would therefore produce (at 200 Proof) 6.6  109 L of spirits alcohol, 5.2–8.2  109 L of beer alcohol, and 100  109 L of wine alcohol for a total beverage alcohol production of 1.12–1.15  1011 L. The production of alcohol from substrates containing sugars or starches is well known, and the technological advances in science, equipment, and unit processing are improving with time such that, despite the fact that up to 80% of the cost of making alcohol is the cost of substrate, properly engineered plants operate profitably, provided that the debt load of the plants is not heavy. Without any major process modifications of existing plants and/or with minor alterations of process operations, profit margins of plant could be increased. New advances, alternate equipment, expansion of ethanol production to cellulose utilization, and yield improvement will ensure that ethanol has become the world’s premier biotechnological commodity.

3.04.2

Biochemistry of the Ethanol Process

In fuel alcohol production, mashes made from grains or sugars are inoculated either directly with a specially selected and prepared strain of active dry (really instant dry) Saccharomyces yeast, stabilized liquid yeast, a pressed or cake yeast, or by dry yeasts conditioned (acclimatized) by a preincubation for 20–30 min at 30–38
C in water or diluted mash (or for several hours in a prefermentor at a temperature of 30
C in diluted mash) and inoculated to achieve initial viable cell numbers in fermentors between 5  106 and 1  107 mL1 of final mash. It is the yeast that makes ethanol in the plant; all other front-end or back-end unit operations can only lose ethanol. The fermentor is the only place where ethanol is synthesized. Growth of the inoculated yeast in batch fermentors takes place under semianaerobic conditions with recommended levels of oxygen approaching 20 ppm per yeast generation time. Under such conditions, ethanol, carbon dioxide, and some glycerol are produced along with new yeast biomass, the latter reaching levels of about 10% of that produced under the aerobic conditions seen in a true batch propagator in a yeast plant under strict conditions of very low substrate and very high aeration. In batch fermentation, a level of approximately 0.05 g of yeast per gram of sugar used is expected and ethanol is made. In a yeast production plant scenario, approximately 0.5 g of yeast is made from every gram of sugar present – no ethanol is produced. In the alcohol plant, even when aeration is supplied, anaerobic conditions prevail as the substrate level is high and the aeration level is inadequate for more extensive yeast growth.4 However, even so, it is interesting to see that in a 0.95  106 L fermentor inoculated at 107 yeasts mL1, the number of cells grown in the tank could reach 2.5  108 mL1 for a total of 2.38  1017 yeasts. If 1 g of yeast dry weight contains 4.87  1010 yeasts,4 then the weight of cells grown in the fermentor above would approximate 4877 kg. If these cells are 5–10 mm in diameter, 1 g of yeast would have a surface area between 1.7 and 6.9 m2, and the 4877 kg of yeast in the fermentor would have a surface area between 8.3  106 and 33.6  106 m2. This is an area equivalent ranging from 1869 to 7584 US football fields. It is no wonder then that these yeast cells can together convert 232 750 kg of sugar (33% solids at 67% starch) to 16% v/v ethanol – excreted with CO2 through the yeast cell membrane. This is the power of the microbe. The biochemistry of the alcohol process is known2 and relatively straightforward.1,11 As illustrated in Figure 1, 12 enzymes in the Saccharomyces convert approximately 90% of the sugars supplied or made from starch into carbon dioxide and ethanol. Other end products are the cells made during fermentation, glycerol, and much smaller quantities of aldehydes, esters, organic acids, and other chemicals – all made by the many enzymes in the cell responsible for making intermediates needed for metabolism and cell growth. Pyruvate located at the pivotal location of the glycolytic pathway is synthesized by pyruvate kinase. There are three key enzymes that modulate flux distribution at the pyruvate node: pyruvate decarboxylase (PDC, converting pyruvate to acetaldehyde and CO2), pyruvate dehydrogenase (PDH, converting pyruvate to acetyl-CoA and CO2), and pyruvate carboxylase (PYC, converting pyruvate to oxaloacetate in the presence of CO2). It has been shown that over 80% of carbon flux is channeled through the PDC route. In the presence of the enzyme alcohol dehydrogenase, the yeast produces ethanol from acetaldehyde and thereby regenerates nicotinamide adenine dinucleotide (NADþ) from accumulated NADH þ Hþ to maintain an internal redox balance within the yeast cytoplasm. In the presence of an electron acceptor such as oxygen, acetyl CoA can be made that is one of the

Ethanol From Starch-Based Feedstocks

37

Outside Thousands of other enzymes are also located in the cell

Inside (cytoplasmic) Substrates Hexose transporter

Glucose Fructose

Glucose HXK ATP GLK Glucose-6-phosphate

HXT Fructose

P l a s m a m e m b ra n e

ATP

HXK

Fructose 1,6-diphosphate FBA Dihydroxyacetone Glyceraldehyde 3-phosphate TPI phosphate TDH NAD+ 1,3-diphosphoglycerate NADH, H+

Ethanol

Ethanol

PGI

Fructose-6-phosphate ATP PFK

ADH Acetaldehyde

End products

PGK

ATP

3-phosphoglycerate PGM 2-phosphoglycerate ENO Phosphoenol pyruvate

PDC CO2

CO2

PYK

ATP

Pyruvate

Figure 1 Normal glycolysis performed by yeasts in sugar-containing mashes. Adapted from Boulton RB, Singleton VL, Bisson LF, and Kunkee RE (1996) Yeast and biochemistry of ethanol fermentation. In: Boulton RB, Singleton VL, Bisson LF, and Kunkee RE (eds.) Principles and Practices of Wine Making. New York: Chapman and Hall and Pretorius IS (2000) Tailoring wine yeast for the new millennium. Novel approaches to the ancient art of winemaking. Yeast 16: 675–729.

key constituents needed for lipid synthesis to keep the cellular membrane intact. When very-high-gravity (VHG) fermentation is complicated by the lack of oxygen, metabolic flux going through the anaplerotic pathway governed by PYC is about sevenfold faster than the flux going through the pathway governed by PDH. As a result, the tricarboxylic acid (TCA) cycle is operated in a reductive mode, and detection of succinate can be observed.6 Glycerol is made by yeasts during alcoholic fermentation in amounts ranging from about 0.7% to 1.6%. Hydrogen ion and electrons that accumulate in the cell when biochemical intermediates in the lower right side of Figure 1 are removed from the glycolytic pathway due to the need for metabolic intermediates for anaerobic cell growth. This leads to excess reduced NADH and Hþ that slows or impedes fermentation unless sequestered by substrate levels of an alternate ‘metabolic sink’ of dihydroxyacetone phosphate (in equilibrium with glyceraldehyde-3-phosphate). Then, redox status (and the rate of glycolysis in Saccharomyces yeast) is returned and further metabolism of glyceraldehyde down the glycolytic pathway through acetaldehyde to ethanol and carbon dioxide resumes.

3.04.3

Yeasts Used in the Process

The production of yeast in a yeast factory and the equipment used for this highly aerobic process have recently been detailed and should be examined by those in the alcohol production industry to ensure that it is understood that aerobic yeast growth cannot be

38

Ethanol From Starch-Based Feedstocks

achieved in alcohol plants due to the fact that for efficient growth to occur, substrate concentrations must be kept very low (less than 0.2% w/v) and aeration must be supplied at more than one volume of air per volume of mash during the process. The incremental feeding regime of medium for this growth and the specialized equipment needed to produce yeast are not found in alcohol production plants.4

3.04.3.1

Yeast

Saccharomyces cerevisiae is virtually the exclusive yeast used in fuel alcohol and beverage alcohol production. Although there are many identified strains stored in culture collections all over the world and in use commercially, only minor metabolic differences appear to be found between them – other than in specific congeners important for beverage flavor and aroma. These yeasts have been carefully preserved under laboratory supervision to ensure purity and genetic identity. Genetically modified yeasts are not used in this industry – probably because the yeasts end up in the distillers’ dried grains or wet grains marketed and the industry would be under great regulatory pressure if such yeasts were present. However, the cellulose-to-ethanol industry will, almost without doubt, have to use genetically engineered yeasts or bacteria, and this will lead to a good deal of governmental control and data requirements for this new segment of the ethanol production industry.

3.04.3.2

Yeast Formats

Today, yeasts are available in a number of different formats – all with good viability and utility.4 One of the largest breakthroughs in the supply of yeast to industry has been the development of instant dry yeast (called active dry yeast (ADY) in the fuel industry). This product has extremely high viable yeast cell counts of 2.2–5  1010 g1, and a product shelf life of up to 3 years. These yeasts have a number of advantages over other formats of yeast (yeast cream, stabilized liquid yeast, and compressed or cake yeast) but normally require conditioning prior to their ability to grow. What they lack in rapid initiation of growth is probably made up for in most cases by advantages in shipping, storage, and handling.

3.04.3.3

Yeast Production in an Alcohol Plant

An alternative to eliminate, in part, the purchase of yeast from a yeast supplier is a process called ‘propagation’ but more accurately called conditioning or metabolic acclimatization. In this process, seen more often in continuous plants than in batch plants, a single vessel prior to a sequence (train) of fermentation tanks is used. It is inoculated to start the process of anaerobic yeast growth, but sized and run at a rate such that inocula can continuously be added to the first fermentor in the train – leaving sufficient volume behind with augmentation of new mash so that the viable cell numbers in the propagator do not decrease – leading to constant cell numbers entering the first fermentor and thus (with growth taken into account) a constant but higher number (viable and nonviable) in each fermentor down the train. This type of in-house yeast production is not recommended by microbiologists due to the fact that any contamination from bacteria or faster-growing wild yeasts will competitively increase faster than the culture yeast itself as time proceeds in the conditioning tank. Loss of ethanol yield and significant problems in regaining control of fermentation ensue, and these issues are difficult to resolve. In batch plants, efforts are often taken to condition ADY so that they are close to or entering the logarithmic phase of growth at the time they are inoculated into the production mash to levels approaching 1  107 viable cells per milliliter. All equipment, aeration and transfer lines should be kept virtually sterile so that bacteria and wild yeast contamination are reduced. Batch processing here allows control of cleanliness and sanitation.

3.04.4

Unit Operations Relevant to Ethanol Production

The sequence of operations normally used in a batch ethanol plant (where all processes except fermentation are run in a continuous fashion) is briefly described. It should be noted that this process is now not the only configuration of equipment used in alcohol plants; many alternative unit operations have now made the planning and operation of new facilities more difficult than in the past.

3.04.4.1

Grain Processing

In North America, the most common substrate is #2 yellow dent corn, with some grain sorghum and wheat used in local areas. Worldwide utilization of both starchy and sugar substrates and areas of their availability and processing variability have been published recently.9 Corn is very widely used in North America and is described here. Milling is the first unit process required, and the purpose is to provide access to kernel components for water and appropriate contact for enzymatic attack after either cooking or cold-cook processing. The hammer mill is most often used to achieve appropriate particle size – not too large for effective starch breakdown and not too small to lead to lost fines in the process. The hammers in the mill rotate at high speed and shatter kernels of corn on screens of specified aperture. Sieve analyses should be made on the grain passing from the screen to ensure reproducible operation from day to day.

Ethanol From Starch-Based Feedstocks 3.04.4.2

39

Enzymatic Processing and Cooking

Starch hydrolysis and cooking begin in the mingler or slurry tank, where the corn meal is mixed with heated backset (recycled water from the beer still) and other recycled water streams. The industrial enzyme a-amylase (normally, today a high temperature-tolerant a-amylase) is required in the cooking process to convert starch particles to dextrins during or after their liquefaction. The water is added normally poised at a pH of about 5.5–6.0 at an optimal temperature of 82–85
C. Under such conditions, significant random hydrolysis of starch to high-molecular-weight dextrins occurs, the latter being more soluble in water and much less viscous. Often, a jet cooker is used that provides steam and turbulence to aid in the dispersion of starch granules and their breakdown to dextrins. Other ethanol plants may not use such vigorous conditions but provide temperatures below gelatinization temperatures in a no-cook or cold-cook process. In both cases, the a-amylase enzyme used hydrolyzes a-1,4 linkages randomly in the starch polymer. The dextrins so produced are of random molecular weight but are more soluble and await the introduction of glucoamylase in simultaneous saccharification and fermentation (SSF) where the high-molecular-weight dextrins are converted through low-molecular-weight dextrins in a measured fashion from the nonreducing end of the glucose polymers to glucose that yeasts will handle as production proceeds. It should be noted that continuous front-end processing (prior to fermentation) and continuous back-end processing (postbeer well) are done in all alcohol plants across the industry. Only the fermentation area (and beer well) is either batch or continuous depending on engineering designs used.

3.04.4.3 3.04.4.3.1

Fermentation Batch Fermentation

Batch fermentation is a process where all the substrate and nutrients are added at zero time or soon after inoculation takes place, and the vessel is allowed under a controlled environment to proceed until maximum end product concentration is achieved. More than 84% of all alcohol in North America is made by batch fermentation – mainly due to the flexibility in a batch plant, the higher product concentrations possible, the ability of industry to minimize infection/loss of yield, the ease of the practice, and low maintenance costs. Despite the above-noted advantages, batch fermentation requires frequent cleaning, sanitizing, and filling of fermentors, resulting in loss of productivity. To maintain the same productivity, the initial (i.e., capital) investment for a batch fermentation is higher than its continuous counterpart.

3.04.4.3.2

VHG Fermentation

Until about 1987, ethanol was made in concentrations from 8% to 10% v/v in commercial batch fuel alcohol plants. VHG fermentation was exploited in a fuel alcohol context as a method to reduce costs of water, labor, and utilities per unit of alcohol.14 It took many years before full commercial realization of VHG technology was married with equipment modifications such that 20% v/v alcohol became the norm in approximately 20 plants across the US Midwest. VHG fermentation to this alcohol concentration is only possible in batch fermentors currently. Depending on the initial glucose level (Figure 2 at 300 gL1;), VHG fermentation in lab experiments exhibited four observable growth phases when one simultaneously examines profiles of biomass, viability, viable counts, and redox potential. As illustrated in Figure 2, a rapid decline of redox potential correlated with rapid yeast propagation, resulting in net NADH production. During this phase of growth (called Region I), the oxidizing power contributed from the fermentation medium and by constant agitation could not compensate for the reducing power resulting from rapid yeast growth. Therefore, a reduction of redox potential was obtained. Region I ends in the logarithmic phase (around the 18th hour of the fermentation), followed by the onset of stationary phase. When measured, redox potential was lower than the redox potential set point (150 mV in this case), so sterilized air was sparged into the fermentor in order to maintain redox potential at the set point (Region II). This region, although the ethanol concentration continually built up in the fermentor, resulted in maintenance of the number of viable cells at a nearly constant level equivalent to the stationary phase of yeast growth. Compared with Region I, the propagation of yeast population slowed, implying that the net production of NADH was less than that produced in Region I. As a result, the total number of electron donors (mainly contributed by NADH) could be balanced by the total number of electron acceptors (mainly attributed to dissolved oxygen via air sparging and agitation and to acetaldehyde). Therefore, a controlled redox potential was maintained; the duration of Region II lasted about 12 h (i.e., from 18 to 30 h). As fermentation proceeded, Region III (between 30 and 36 h) was characterized by a rise of redox potential, where the oxidizing power (i.e., the number of electron acceptors) was greater than the reducing power (i.e., the number of electron donors). In this region, the air sparging was stopped and agitation at 150 rpm was maintained. Although the change of number of viable cells was not significant, a reduction of both viability and biomass was noticeable. Region III might signify that the effect of ethanol was gradually switching to toxification from inhibition on yeast growth, which might be regarded as a transition period from the late stationary phase to the commencement of the death phase. As the severity of ethanol toxicity worsened, a moderate decline of yeast viability (decreased to 87% at 48 h from 97% at 36 h) became obvious, showing a large amount of yeast death. Region IV (classified as the death phase) in Figure 2 showed that the response of redox potential profile was relatively sluggish compared with that in Region III. The sluggishness of redox potential profile in this region was probably attributed to the inability of yeast to propagate due to ethanol toxicity, even with a gradually saturated dissolved oxygen level provided through constant agitation. As a result, a positive redox potential profile was recorded and eventually leveled off.

Ethanol From Starch-Based Feedstocks

8

Biomass (g l–1)

Redox potential (mV)

III

II

I

100

0

IV

40 90

6 Biomass Viability Viable counts Redox potential

4

80

−100

30

20

70 10

2

−200

50

100

0

Viable counts (× 107 cells ml−1)

10

200

Viability (%)

40

0

10

20

30

40

60 50

0

Time (h) Figure 2 Characteristic growth pattern of VHG fermentation using sugar (cultivation conditions: glucose, 300 gL1; aeration, 1.3 vvm; redox potential setting, 150 mV).

It can be noticed that between 6 and 8 h (circled part of Figure 2), a ‘shoulder-like’ pattern was found on the redox potential profile. This pattern is correlated to the glucose level; that is, the higher the glucose feed, the clearer the pattern (data not shown). By cross-examining biomass profile shown in the same figure, this pattern corresponds to the transition period from lag to logarithmic phase during yeast growth. It is postulated that the presence of this pattern is linked to the initial glucose concentration. The secret of VHG fermentation is to reduce osmotic pressure, keeping substrate in dextrin form, but in equilibrium with smaller amounts of glucose liberated by added glucoamylase enzyme. In this way, fermentation is prolonged and alcohol is produced at the fastest rate possible. In addition, the highest viable cell number of yeast (the catalyst) is allowed to be made, and bacterial and wild yeast are discouraged. The key to fast growth and high cell numbers is nutrition. Careful research has shown that usable nitrogenous compounds are the most important nutrients now at suboptimal levels in almost all mashes and, therefore, required as fermentation additives (yeast foods). Mashes made from all grains and especially from sugars and semipurified starches are devoid of usable nitrogen as well as other nutrients such as usable phosphorous and sulfur, and minerals and vitamins that yeast requires for enzyme stability and activity. For this reason, yeast growth is poor and the amount of cellular biomass is insufficient to end ferment even moderate amounts of carbohydrate in such prepared ‘mashes’. This is documented in a number of papers and summarized in more general terms.3 The guidelines for implementing VHG fermentation technology and the advantages of operating VHG process are also listed in Table 1.

3.04.4.3.3

Continuous Fermentation

Batch fermentation is easy to operate and gives closer control of bacterial contamination. However, some prefermentation processes are required. Due to the transient nature of batch fermentation, ethanol production can also vary, making it difficult for process analyses. As an alternative, continuous fermentation maintains the process at steady state with constant production rates, making it easier for process optimization. In reality, almost no continuous plants operate in steady state, and continuous fermentation leads to less than 16% of all alcohol made in North America today. This is primarily due to large and continuing losses in alcohol due to infection – normally after a relatively long period of low infection and probably correlating with establishment of a ‘house flora’ of microbes adapted to fermentation conditions. These microbes rob yield throughout the process – normally in dual trains of 4–5 fermentors sometimes inoculated by a continuous propagation system that grows contaminants faster than it grows yeasts, and continually adds contaminant inocula into the first vessels in the train(s). The recovery from infection is problematic in continuous plants due to the use of heat exchangers and countercurrent carbon dioxide collection headers as well as blanked ‘Ts’ in piping and heat exchangers – all of which must also be cleaned and sanitized in a prolonged downtime that eliminates the productivity advantages associated and described for continuous systems.12

Ethanol From Starch-Based Feedstocks Table 1

41

Guidelines for implementing VHG fermentation technology and advantages of operating VHG process

Guidelines for implementing VHG fermentation technology • Prepare mashes with increasingly high solids (less water) • Remove solids (with rinse) prior to fermentation – if it can be done • Supply sterile oxygen to the fermentation – cold side of heat exchanger – 20 ppm per generation time or 5 ppm per hour – as needed for strong yeast cell membranes • Supply enough yeast-usable nitrogen so that mashes are not ‘usable N’ deficient. More stimulates fermentation rate and increases yeast cell numbers • Supply other nutrients as needed • Gelatinize and liquefy but do not saccharify starch in mash before fermentation • Ensure that pumps can handle mashes of more than 32
Brix • Use a yeast that tolerates alcohol well and thrives in high sugar media • Carry out SSF in fermentors (simultaneous saccharification and fermentation) • Work up slowly in specific gravity to ensure that problems (bottlenecks) do not occur • ‘Condition’ or prepare yeast in lower gravity mashes to inoculate VHG mash. Do not reuse yeast • Keep fermentor temperature down (without much loss in rate of ethanol made due to adequate nitrogen supplied) • Keep mash free from contaminating bacteria and bacterial end products Advantages of operating a VHG process • Increased plant capacity or reduction in capital costs • Increased alcohol to 20% v/v • Increased fermentor space (removal of insolubles and water) • Increased plant efficiencies • Reduction in labor costs per liter of ethanol produced • Reduction in energy per liter (much lower insolubles in fermentor and still, less water through the still, optimal distillation, reduced inputs of water, reduction in fermentor downtime as well as cleaners/ sanitizers) • Reduced survival and proliferation of bacteria • Opportunities for food quality by-products and to harvest high-protein yeast

3.04.4.4

Alcohol Processing

When fermentation is over, the fermented mash from batch or continuous fermentors is emptied and taken to the beer well. This large vessel is really nothing more than a surge tank that accumulates ethanol at production concentration (along with all other solubles and insolubles), which will be taken forward for distillation. The beer well is the least studied vessel in the process, and little attention has been given to losses in ethanol in this liquid through the action of bacteria and wild yeasts. The fermented mash from the beer well moves to the beer still where the concentration of the alcohol rises in the stripper and then in the rectifier column – from fermentation concentrations to near the azeotrope of 194
Proof. This well-known but somewhat complicated system is well described.8 Later, dehydration of this alcohol is normally achieved by molecular sieves in a process where the previous technology called azeotropic distillation, using entrainers such as benzene or cyclohexane, has now been replaced by pressure swing adsorption using special synthetic zeolites with a specific pore size of 3 Å that provide a significant increased volume of the columns for the 2.8 Å water molecules than for the molecules of ethanol that are closer to 4.4 Å in size. Zeolites work by separation due to molecular size as well as water adsorption. These columns are reusable after regeneration – done by recycling of a pure ethanol vapor by reverse flow in a cyclical process.13 Produced dry alcohol is normally denatured with gasoline for shipping and storage. In this way, the alcohol is suitable for incorporation at virtually any concentration for sale.

3.04.4.5

Stillage Processing

The effluent from the bottom of the beer still is centrifuged with the liquid phase recycled as backset or evaporated to syrup. Solids from the centrifuge are processed to distillers wet grains (DWG) and/or distillers dried grains (DDG), and the DDG is mixed with solubles (syrup) to make DDG with solubles (DDGS). By-products are sold to animal feed producers. This part of the plant, the back end or dryhouse, requires expensive equipment and energy use for operation. Consistent quality and composition are needed for incorporation into animal and poultry feeds.

3.04.4.6

Carbon Dioxide Production

Carbon dioxide is liberated in the fermentors and passes through scrubbers in an attempt to remove any ethanol, volatile organic compounds, and fines that might be entrained in the gas as it leaves the vessel. Under some conditions, carbon dioxide is collected,

42

Ethanol From Starch-Based Feedstocks

purified, and used for soft drinks, dry ice, or other industrial processes. Reduction or sequestration of CO2 is a goal of the industry so that the greenhouse gas benefits of making ethanol (in comparison with petroleum) will be even more beneficial. Even so, ethanol production is now considered to be carbon neutral as the CO2 generated in fermentation and on using the ethanol as a fuel is virtually the same quantity that was taken up by the plant and converted to starch during the crop year.

3.04.5

Environmental Requirements in Fermentation

3.04.5.1

pH

Maintaining fermentation pH near the intracellular pH of a yeast cell is essential. A deviation between extracellular and intracellular pH results in a change of yeast growth rate and cell yield, particularly under VHG conditions. There are many sources that contribute to pH change during ethanol fermentation. Addition of acid to reduce pH in mashing, microbial infection leading to elevated acetic and lactic acids, CO2 liberation during glucose conversion by a yeast cell, production of acidic end products, as well as usage of nitrogenous cations from salts are the key sources. All cause the reduction of mash pH. The application of ADY during industrial-scale ethanol fermentation is a common practice. However, ADY contaminated by lactobacilli and other acid-producing bacteria are not unusual although numbers are comparatively low.4 As illustrated in Figure 3, when the fermentation pH (i.e., extracellular pH) is lower than the intracellular pH inside a yeast cell, acetic and lactic acids (in their undissociated forms) originally produced by microbial contaminants can freely be transported across cell membranes. When inside the cell, the intracellular pH value is greater than the pKa values of both acids, resulting in acid dissociation and a lowering of intracellular pH due to the formation of dissociated protons. To maintain intracellular pH homeostasis,  1 mol of adenosine triphosphate (ATP) is consumed to pump each proton to the outside of the cell. There are only 2 mol of ATP being generated per mole of glucose by S. cerevisiae during ethanol synthesis. As a result, one could then extrapolate that the overall fermentation efficiency would be reduced due to reduction in cell growth resulting from the shortage of ATP availability. To carry out a contaminant-free ethanol fermentation, buffering fermentation pH close to the intracellular pH inside a yeast cell might significantly improve ethanol fermentation although cost is likely prohibitive.

3.04.5.2

Temperature

Temperature exerts an ultimate impact on yeast thermodynamics during ethanol production under VHG conditions. As shown in the left panel of Figure 4, the final ethanol concentration is affected by the choice of fermentation temperature irrespective of initial dissolved solid concentrations. The difference in ethanol yield subjected to temperature variation is shown as the dissolved solid concentrations increase, indicating that proper temperature management is important particularly under VHG conditions. An improper choice of temperature during ethanol fermentation can lower fermentation rate. Figure 4 (right panel) also shows that

Outside

HAc, Hlac

P l a s m a m e m b ra n e

pH (intracellular) = 5.4 HAc ≥ H+ + Ac– HIac ≥ H+ + lac– pH – pKa = log [X–]/[HX] pKa (HAc) = 4.74 pKa (Hlac) = 3.86 ATP

Figure 3

H+

Effect of external pH on internal pH (HAc ¼ acetic acid; Hlac ¼ lactic acid). See more detailed discussion in Ref. 3.

Ethanol From Starch-Based Feedstocks

Commercial yeast #1 18

Glucose, 36 °C

12

Glucose, 40 °C

250

Ethanol, 36 °C Ethanol, 40 °C

Concentration (g l–1)

Ethanol (% v/v)

Commercial yeast #2 300

14% 19% 24.7% 30.3% 36.5%

15

43

200

150

100

9 50

6 15

20

25

30

Temperature (°C)

35

0

0

10

20

30

40

50

60

70

Time (h)

Figure 4 Effect of high temperature in high solid mashes with two commercial yeasts. Data are not comparable due to differences in mashes used but illustrate the stress on yeasts at high ethanol levels. Left panel, reproduced from Jones AM and Ingledew WM (1994) Fuel alcohol production: Optimization of temperature for efficient very-high-gravity fermentation. Applied and Environmental Microbiology 60: 1048–1051 without extra usable nitrogen added; right panel, reproduced from data provided by the supplier of Commercial Yeast #2.

the performance of a yeast cell cultivated in 36
C exceeds that at 40
C. As a result, faster glucose utilization rates and ethanol production rates are observed. In addition, to operate the ethanol fermentation process at the optimal growth temperature of yeast used, the selection of a yeast strain that has an optimal growth temperature close to the temperature of geological location of a fuel alcohol plant is necessary. A large portion of energy consumption can be saved when the temperature difference between the optimal growth temperature of a yeast cell and that of plant site is minimized. It should also be noted that temperature becomes more stressful to yeasts under VHG conditions or whenever other stresses are present. Industry has adopted ‘temperature staging’ to lower temperature stress as ethanol increases.5

3.04.5.3

Aeration

Air provision to VHG fermentation is required because the mass accumulation of CO2 resulting from glucose conversion can hamper yeast growth, and thus lower ethanol productivity. Sparging small volumes of air directly into fermentation broth not only helps in liberation of CO2 but also to create an even distribution of temperature in the fermentor. Sparging small volumes of air facilitates syntheses of unsaturated fatty acids and sterols within yeast, thereby indirectly promoting biomass formation, resulting in faster fermentation rates and shorter fermentation times. These lipids are essential constituents when making a new yeast cell to maintain the membrane integrity of an actively growing yeast population.10 Figure 5 illustrates the effect of aeration on yeast propagation under VHG conditions. VHG fermentation with aeration can accelerate yeast duplication rate and extend its logarithmic phase, thereby increasing glucose utilization and ethanol production when compared with those grown with inadequate aeration. Additionally, aeration during VHG fermentation can reduce fermentation time, resulting in complete glucose utilization (42 h with aeration versus incomplete glucose uptake after 48 h, Figure 5). From Figure 5, it is estimated that yeast cells subjected to such aeration can make 1.32-fold more ethanol than those without aeration. Sparging air to VHG fermentation is essential, but sparging with the right timing and time interval is necessary. Based on the growth regions shown in Figure 2, four aeration schemes (0–8, 8–18, 18–28, and 28–38 h) were tested to explore the possibly of further increasing fermentation rate under VHG conditions. Results shows that yeast aerated between 8 and 18 h (i.e., in the logarithmic phase) could completely utilize glucose in 36 h (as compared with 42 h). At the optimized aeration rate, VHG fermentation could be completed in 30 h. No air stripping of ethanol was noticed. It is not necessary to aerate VHG fermentation throughout the course of fermentation. The ethanol yield obtained from a full-course aeration scheme did not provide any advantage over ‘right timing’ aeration scheme. Only operating costs were increased.

3.04.5.4

Osmotic Stress

Stresses are referred to as any environmental factor that hinders yeast growth. During VHG fermentation, high glucose levels would be the cause of osmotic stress, which is counteracted by accumulation of intracellular glycerol as a compatible solute. Glycerol acts

44

Ethanol From Starch-Based Feedstocks

175

14

350

300

12

125

250

10

200

8

100

75

50

Ethanol

Biomass

6

150

Biomass (g l–1)

150

Glucose (g l–1)

Ethanol (g l–1)

w/o Aeration w/ Aeration

4

100 Glucose

25

50

0

0

2

0

10

20

30

40

0 50

Time (h) Figure 5

Effect of aeration during VHG glucose fermentation. Air supplied at 0.82 vvm.

as an intracellular osmoregulant to protect yeasts when the extracellular osmolarity is high. Stress of this kind is reduced by not hydrolyzing starch to glucose in the mashing stage, but instead creating dextrins that are then processed to glucose in the fermentor by SSF. In this process, sugars liberated from medium to large limit dextrins by glucoamylase enzyme are made at a rate almost equal to the rate that yeast are able to take up and biochemically convert them to end products.

3.04.5.5

Chemical Stress

Chemical stress not only can occur from the organic acids such as acetic and lactic acid produced by microbes that contaminate the fermentation but also can be attributed to sulfide compounds added during the wet-milling process, to recycled backset, and to sanitizer residues. Salts and other chemicals build in fermentation primarily due to the recycling of backset (the water from the centrifuging of whole stillage). Backset will build in concentration of ions and solubles as well as insolubles as it is used over and over again in the process to reduce the amount of water needed for the process. Other materials such as sodium ion (sodium hydroxide cleaning agent/sanitizer) can also accumulate in backset to above 500 ppm. This ion is more inhibitory if other stressful agents including temperature, pH, and organic acids are present simultaneously.

3.04.5.6

Freedom from Microbial Infection

Microbial infection is primarily caused by the presence in the process of Lactobacillus and related species, or wild yeasts such as Schwanniomyces (Dekkera). Both of these microbes are able to live in the environment of the mash but at times outnumber the culture yeasts due to either faster growth rates or to their relative intolerance to the organic acids they make (lactic and acetic acids). These acids affect yeast growth and metabolism. Normal ethanol factories are estimated to lose 1%–4% of alcohol due to microbial contamination when operating well. Badly infected plants may lose 17% or more as the occasional rampant infection leads to dumping of fermentor contents, shutdown, and complete cleaning of the facility. Bacterial control is presently done using antibiotics, but their use is becoming a matter of controversy, and new antimicrobial compounds are needed to reduce spoilage and ethanol loss in this industry.

3.04.6

Yield Coefficient and Net Rate Expression

3.04.6.1

Yield and Fermentation Efficiency

Yield coefficients, usually defined as the ratio of change of the desired metabolites to a limiting substrate, are commonly used to evaluate the performance of microorganisms grown in a single fermentor under either batch or continuous conditions. Yp/s (the ratio of ethanol produced to that of glucose utilized) is the most used yield coefficient in ethanol fermentation. The theoretical maximum Yp/s is 0.511, which is calculated according to the stoichiometric relation that 1 mol of glucose is converted into 2 mol of ethanol and 2 mol of CO2. By comparing it with the experimentally obtained Yp/s, the fermentation efficiency under

Ethanol From Starch-Based Feedstocks Table 2

45

Calculation of ethanol yield based on starch (S) and moisture (M) content

M ¼ 12, S ¼ 71

M ¼ 14, S ¼ 67

Corn (a) ¼ 880 kg ¼ 1000*(1–0.12) Starch (b) ¼ 625 kg ¼ (a)*0.71 Glucose (c) ¼ 694 kg ¼ (b)*180/162 Ethanol ¼ 355 kg ¼ (c)*0.511, or 450 L ethanol ton1, or 3.03 US gallons ethanol bu1 Assume 90–93% of the maximum attainable yield 405–419 L ethanol ton1, or 2.72–2.82 US gallons ethanol bu1

Corn (a) ¼ 860 kg ¼ 1000*(1–0.14) Starch (b) ¼ 576 kg ¼ (a)*0.67 Glucose (c) ¼ 640 kg ¼ (b)*180/162 Ethanol ¼ 327 kg ¼ (c)*0.511, or 415 L ethanol ton1, or 2.79 US gallons ethanol bu1 374–386 L ethanol ton1, or 2.52–2.60 US gallons ethanol bu1

Use 1000 kg of corn grain as the basis. Note that both starch and moisture concentrations in crops vary greatly. 1 bu (or bushel) ¼ 56 lb (as defined for alcohol plants). 1 L ton1 ¼ 6.725 10 3 US gallon bu1.

one particular process operating condition is estimated, such that various fermentation conditions or process settings can be quantitatively evaluated so that the most effective parameters can then be selected. A fermentation efficiency ranging from 90% to 93% of the theoretical maximum yield in an ethanol plant operated in batch mode is now a norm, and a yield in a range of 2.6–2.9 US gal per bushel of corn is common.

3.04.6.2

Yield and Feedstock

Depending on how corn grain is processed, and stored, it can vary in moisture content (11%–16%), starch content (67%–72%), and other components, ultimately affecting ethanol yield. The expected ethanol yield estimated from two corn grains having different moisture and starch contents is illustrated in Table 2. The proper selection of feedstock with high starch and low moisture contents for ethanol production is necessary because it can significantly change the projected ethanol productivity and profit margin in a factory. Grain purchase specifications are key to high yields.

3.04.6.3

Yield and Overall Plant Performance

Losses of yield result from retrogradation of starch, Maillard reactions, yeast cell growth, glycerol production, minor amounts of organic acids, higher alcohols, esters and aldehydes, as well as losses in factory operation due to spillage, bacterial and wild yeast infections, stuck fermentations, inadequate nutrition, out-of-specification environmental conditions, and effects of stress on yeast. A number of steps (Table 3) have been taken in an attempt to improve yields and the economics of fermentative ethanol.

3.04.6.4

Net Rate Expression in Continuous Fermentations

When a series of fermentors are cascaded or interconnected, the use of a single ‘yield coefficient’ as the measure to evaluate such a process configuration may not be appropriate. Instead, using the ‘net rate’ approach is a feasible evaluation criterion.7 This approach takes the difference of concentration of a metabolite present in the outlet and inlet streams and multiplies it by the

Table 3

Steps to enhance overall yield of an ethanol plant

• Pay a premium for high starch, low moisture corn • Increase dissolved solids in mash • Increase fermentation rates and ethanol contents while eliminating grain solids in the fermentor and still • Build plants with excess cooling and evaporative capacity • Reduce backset – reduce organic acids and ions • Reduce harmful recycled solutions • Use heat exchangers to cool mash – recycle energy • Know your yeast, its properties, and contaminants • Select your yeast and understand effect of yeast foods • Do not use continuous or continuous-intermittent propagation • Buy the best enzymes you can and know how to use them • Understand cleaning and sanitation • Take advantage of continuous education for plant employees

Ethanol From Starch-Based Feedstocks

Net ethanol production rate (g l–1 h–1)

Net glucose consumption rate (g l–1 h–1)

46

15 A

15.2% w/v 19.1% w/v 22.5% w/v 25.4% w/v 31.2% w/v

10

5

0 6

F1

F2

F3

F4

F5

F1

F2

F3 Fermentor

F4

F5

B

5 4 3 2 1 0 –1

Figure 6 Illustration of the use of net rate expression applied to a multistage chemostat fermentation system under different levels of glucose. F1–F5 refers to five chemostat fermentors connected in series. Reproduced from Lin Y-H, Bayrock DP, and Ingledew WM (2002) Evaluation of Saccharomyces cerevisiae grown in a multistage chemostat environment under increasing levels of glucose. Biotechnology Letters 24: 449–453.

corresponding dilution rate. By incorporating the dilution rate into the net rate expression, the performance of each fermentor in a multistage chemostat fermentation process can be compared, and the rate-limiting stage in the fermentation train can be pinpointed. Proper actions can then be taken to enhance fermentation efficiency. Figure 6 illustrates the use of net rate expression in a 5-chemostat multistage fermentation system. It shows that lower levels of glucose in the feed stream results in higher net glucose consumption rates (NGCRs). The NGCR progresses from F1 to F2 for glucose feed between 15.2% and 25.4% w/v. With further increases in glucose feed up to 31.2% w/v, the NGCR decreases from F1 to F5. Such a linear decline in NGCR might be attributed to osmotic shock caused by glucose. To overcome the problem, SSF could be one of the possible options. A high NGCR is also accompanied by an increase in ethanol production from F1 to F2, resulting in a high net ethanol production rate (NEPR). As fermentation proceeds, abrupt decreases of NGCR are paralleled by similar declines in NEPR, indicating increased ethanol toxicity in the later stages of the fermentation system. One possible choice to reduce such a drastic effect is to incorporate extractive fermentation into the process. The above example portrays the applicability of net rate expression to locate process bottlenecks in multistage continuous VHG fermentation systems.

3.04.7

Metabolic Flux Analysis

Instead of using a ‘black box’ approach such as yield coefficient, to evaluate the performance of ethanol fermentation by yeast cells, metabolic flux analysis (MFA) provides a different mechanism to probe intracellular enzymatic activities as a whole within a yeast cell. MFA requires the construction of a metabolic map that is phenotypic to specific fermentation conditions. The map consists of several pathways that are interrelated by associated biochemical reactions. Each bioreaction is linked through stoichiometric coefficients that represent the involvement of metabolites and equilibria between enzymes, substrates, and products. Providing a phenotypic metabolic map with measured metabolite concentrations, a metabolic flux for each bioreaction in the map is estimated and unmeasured metabolite concentrations are evaluated. The magnitude of an estimated flux correlates to the strength of each enzyme that catalyzes reactions in the corresponding bioreactor. MFA has been applied to analyze a 5-chemostat multistage fermentation process operated under high gravity conditions (15.2% w/v glucose), and illustrated that at the pyruvate node, the activities of PDC and PDH were maintained at a nearly constant level even though ethanol accumulated from the upstream

Ethanol From Starch-Based Feedstocks

Storage

CO2 Close proximity Cropland

Ethanol plant

Ethanol

Stillage Energy

Cattle farm

47

Refinery Excess heat Gasohol

Manure Consumer Figure 7

Illustration of one possible cooperative business scheme or biorefinery using various substrates.

fermentor to the downstream one in the fermentation train and reached inhibitory concentrations. Comparatively, a relatively large deviation of the fluxes on the pathway regulated by PYC possibly related to the change of redox status in the TCA cycle enzymes still operating in yeast cells under semianaerobic conditions.6

3.04.8

Summary

The fuel ethanol production sector is a large-volume-low-value industry. Any process modification and/or operation that can increase ethanol productivity by a small fraction will dramatically enhance profits for the business. Various fermentation process configurations and optimizations have been suggested to improve product yields and productivity in an attempt to reduce costs. Additionally, profits can be further increased when the relevant industries are working cooperatively. These industries may include crop growers, fuel ethanol producers, cattle farms, refinery, and power companies. As illustrated in Figure 7, a fuel alcohol production plant can be built in close proximity to areas of crop production, such that the cost of transporting feedstock to the plant site is reduced. Anhydrous ethanol is blended with gasoline provided by a refinery to make a blended product and sold to consumers for profit. The excess heat from the refinery is reutilized by the ethanol plant to lower energy cost. The stillage collected from the beer still contains various nutritional substances, making it an ideal supplement to nourish ruminants, and the manure from these animals can be further processed by means of solid-phase fermentation to form methane, which is then utilized to heat the ethanol plant to lower the utility bill. The fermented manure can be applied to cropland as fertilizer. Carbon dioxide can be collected and stored underground to minimize CO2 emissions to the atmosphere. Turning by-products or waste products from one industry into feedstock for others has the potential to sustain the limited supply of natural resources, and creates new and greener businesses. This may also be the mechanism for cellulose ‘front-end’ incorporation in the future so that substrates are optimized for cost and availability.

References 1. Boulton, R. B.; Singleton, V. L.; Bisson, L. F.; Kunkee, R. E. Yeast and Biochemistry of Ethanol Fermentation. In Principles and Practices of Wine Making; Boulton, R. B., Singleton, V. L., Bisson, L. F., Kunkee, R. E., Eds., Chapman and Hall: New York, 1996. 2. Ingledew, W. M. Alcohol Production by Saccharomyces cerevisiae: A Yeast Primer. In The Alcohol Textbook; Jacques, K., Lyons, T. P., Kelsall, D. R., Eds., 4th ed.; Nottingham University Press: Nottingham, 2005; pp 49–87. 3. Ingledew, W. M. Yeast Stress in the Fermentation Process. In The Alcohol Textbook; Ingledew, W. M., Kelsall, G. D., Austin, G. D., Kluhspies, C., Eds., 5th ed.; Nottingham University Press: Nottingham, 2009; pp 115–126. 4. Ingledew, W. M.; Austin, G. D.; Kraus, J. K. Commercial Yeast Production for the Fuel Ethanol and Distilled Beverage Industries. In The Alcohol Textbook; Ingledew, W. M., Kelsall, G. D., Austin, G. D., Kluhspies, C., Eds., 5th ed.; Nottingham University Press: Nottingham, 2009; pp 127–144. 5. Jones, A. M.; Ingledew, W. M. Fuel Alcohol Production: Optimization of Temperature for Efficient Very-High-Gravity Fermentation. Appl. Environ. Microbiol. 1994, 60, 1048–1051. 6. Lin, Y.-H.; Bayrock, D. P.; Ingledew, W. M. Metabolic Flux Variation of Saccharomyces cerevisiae Cultivated in a Multistage Continuous Stirred Tank Reactor Fermentation Environment. Biotechnol. Prog. 2001, 17, 1055–1060. 7. Lin, Y.-H.; Bayrock, D. P.; Ingledew, W. M. Evaluation of Saccharomyces cerevisiae Grown in a Multistage Chemostat Environment under Increasing Levels of Glucose. Biotechnol. Lett. 2002, 24, 449–453. 8. Madson, P. W. Ethanol Distillation: The Fundamentals. In The Alcohol Textbook; Ingledew, W. M., Kelsall, G. D., Austin, G. D., Kluhspies, C., Eds., 5th ed.; Nottingham University Press: Nottingham, 2009; pp 289–302. 9. Monceaux, D. A. Alternative Substrates for Fuel Alcohol Production. In The Alcohol Textbook; Ingledew, W. M., Kelsall, G. D., Austin, G. D., Kluhspies, C., Eds., 5th ed.; Nottingham University Press: Nottingham, 2009; pp 47–71. 10. O’Connor-Cox, E. S. C.; Ingledew, W. M. Effect of the Timing of Oxygenation on Very High Gravity Brewing Fermentations. J. Am. Soc. Brew. Chem. 1990, 48, 26–32. 11. Pretorius, I. S. Tailoring Wine Yeast for the New Millennium. Novel Approaches to the Ancient Art of Winemaking. Yeast 2000, 16, 675–729.

48

Ethanol From Starch-Based Feedstocks

12. Sheppard, J. D. Continuous Ethanol Fermentation. In The Alcohol Textbook; Ingledew, W. M., Kelsall, G. D., Austin, G. D., Kluhspies, C., Eds., 5th ed.; Nottingham University Press: Nottingham, 2009; pp 259–273. 13. Swain, R. L. B. Molecular Sieve Dehydrators: Why They Became the Industry Standard and How They Work. In The Alcohol Textbook; Ingledew, W. M., Kelsall, G. D., Austin, G. D., Kluhspies, C., Eds., 5th ed.; Nottingham University Press: Nottingham, 2009; pp 379–384. 14. Thomas, K. C.; Hynes, S. H.; Jones, A. M.; Ingledew, W. M. Production of Fuel Alcohol from Wheat by VHG Technology. Effect of Sugar Concentration and Fermentation Temperature. Appl. Biochem. Biotechnol. 1993, 43, 211–226.

3.05

Fuel Ethanol Production From Lignocellulosic Biomass

Feng-Wu Bai, Shanghai Jiao Tong University, School of Life Sciences and Biotechnology, Shanghai, China Shihui Yang, Hubei University, College of Life Sciences, Wuhan, China Nancy WY Ho, Purdue University, School of Chemical Engineering & Laboratory of Renewable Resources Engineering, West Lafayette, IN, United States © 2019 Elsevier B.V. All rights reserved. This is an update of N.W.Y. Ho, M.R. Ladisch, M. Sedlak, N. Mosier, E. Casey, 3.06 - Biofuels from Cellulosic Feedstocks, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 51-62.

3.05.1 3.05.2 3.05.2.1 3.05.2.2 3.05.2.3 3.05.2.4 3.05.3 3.05.3.1 3.05.3.2 3.05.3.3 3.05.3.4 3.05.4 3.05.4.1 3.05.4.2 3.05.4.3 3.05.5 3.05.5.1 3.05.5.2 3.05.6 3.05.7 References

3.05.1

Introduction Lignocellulosic Biomass Cellulose Hemicelluloses Lignin Other Components Pretreatment Physicochemical Pretreatment Chemical Pretreatment Solvent Extraction Biological Pretreatment Enzymatic Hydrolysis of Cellulose and Co-Fermentation of C5 and C6 Sugars Separate Hydrolysis and Co-Fermentation Simultaneous Saccharification and Co-Fermentation Consolidated Bioprocessing Strain Development Engineering S. cerevisiae Engineering Z. mobilis Unit Integration and System Optimization Conclusions

49 51 51 52 52 53 54 54 55 57 58 58 58 59 60 60 61 62 63 64 64

Introduction

As the largest liquid biofuel, fuel ethanol is now being produced predominantly from sugar- and starch-based feedstocks. While the US is producing fuel ethanol mainly from corn with a total production capacity of 15.8 billion gallons in 2017, Brazil is producing fuel ethanol primarily from molasses, a by-product of sugar production, with a total production capacity of 7.1 billion gallons in 2017 (World Fuel Ethanol Production: https://ethanolrfa.org/resources/industry/statistics/). Apparently, fuel ethanol produced from sugar- and starch-based feedstocks is not sustainable, particularly when its production capacity is expanded drastically to address challenges raised with sustainable transportation and environment through reducing dependence on fuels derived from crude oil, taking into account of the increase of global population and consequent demand for food supply as well as the shrinkage of arable land for grain production due to the speedup of industrialization and urbanization, especially in developing countries with large population. Lignocellulosic biomass, particularly agricultural residues, is abundantly available on the earth, which is a sustainable feedstock for producing fuel ethanol at large scales.1 Fuel ethanol produced from lignocellulosic biomass is not competing for food commodities directly or land use indirectly, which has been termed as the second generation (2G) fuel ethanol or cellulosic ethanol compared to the first generation (1G) fuel ethanol produced from sugar- and starch-based feedstocks.2 However, as shown in Fig. 1, major components in lignocellulosic biomass are cellulose, hemicelluloses and lignin, which are entangled to form lignin-carbohydrate complexes (LCCs) with vascular structures for transporting nutritional components absorbed by root from soil to other parts, and in the meantime providing support and protection for plants.3 Natural evolution has made lignocellulosic biomass recalcitrant to degradation, and thus pretreatment is necessary for destructing the LCCs to separate the cellulose component from lignin for enzymatic hydrolysis to release glucose as feedstock for microbial fermentation.4,5 On the other hand, although hemicelluloses are easily hydrolyzed, they are heterologous carbohydrates, and significant amounts of pentose sugars including xylose and arabinose are produced during pretreatment, which are not fermentable for the brewing yeast Saccharomyces cerevisiae used for fuel ethanol production from sugar- and starch-based feedstocks, and recombinant strains engineered with pentose metabolic pathways for ethanol production are needed.6 The characteristics of lignocellulosic biomass make the production of 2G fuel ethanol significantly different from those established processes for fuel ethanol production from sugar- and starch-based feedstocks. In this chapter, fundamentals and cutting-edge

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00150-6

49

50

Fuel Ethanol Production From Lignocellulosic Biomass

Figure 1

Schematic diagram of plant cell walls and the impact of pretreatment on their structures. Reprinted from [3] with permission.

Fuel Ethanol Production From Lignocellulosic Biomass

51

technologies for cellulosic ethanol production are presented, with a focus on the understanding for the structure of lignocellulosic biomass, leading pretreatment technologies, enzymatic hydrolysis of the cellulose component, co-fermentation of hexose and pentose sugars released from the hydrolysis of cellulose and hemicelluloses, and process development through unit integration and system optimization. Challenges and perspectives for the commercial production of 2G fuel ethanol are also highlighted.

3.05.2

Lignocellulosic Biomass

Understanding lignocellulosic biomass, particularly its molecular structure and chemical composition, is a prerequisite for developing effective pretreatment technologies to destruct the LCCs, designing enzymes to liberate sugars, particularly cellulases to release glucose from the cellulose component, as well as engineering microorganisms for robust production of ethanol by converting sugars more efficiently. Lignocellulosic biomass is mainly with plant cell wall, which composes of structural carbohydrates of homogeneous cellulose and heterogeneous hemicelluloses as well as phenolic polymer lignin as primary components. Although contents of these components vary substantially, depending on plant species and variety, climate, soil conditions and fertilization practice, for agricultural residues such as corn stover, wheat and rice straw that have been acknowledged as preferred feedstock for the production of 2G fuel ethanol, they contain
40% cellulose,
25% hemicelluloses and
15% lignin.1 The distinctive feature of plant cell wall is the two-part structure, as illustrated in Fig. 2.7 Primary cell wall is developed with cell division, and then enlarged during cell growth to a fiberglass-like structure, with crystalline cellulose microfibrils synthesized by the complex of cellulose synthase from glucose and embedded into the matrix composed mainly of hemicelluloses and lignin. The primary wall of adjacent cells is held together by a sticky layer called the middle lamella composed of pectin to form the conducting tissue system arranged in numerous vascular bundles. On the other hand, when cells cease to grow, secondary cell wall is gradually deposited between the plasma membrane and the primary cell wall for better mechanical strength and structural reinforcement through the incorporation of lignin into xylem fibers, which accounts for the bulk of lignocellulosic biomass that can be converted to fuels and chemicals.8 The development of the conducting tissue system with the rigid secondary cell wall is a critical adaptive event associated with the evolution of land plants, which not only facilitates the transport of water and nutrients as well as extensive upright growth, but also raises the recalcitrance to degradation due to the interaction and cross-linking of cellulose, hemicelluloses and lignin.9

Figure 2

Diagrammatical structure of plant cell wall. Reprinted from [7] with permission.

3.05.2.1

Cellulose

Cellulose is a polysaccharide composed of linear glucan chains that are linked together by b-1,4-glycosidic bonds with cellobiose residues as the repeating unit at different degree of polymerization (DP), and packed into microfibrils which are held together by intramolecular hydrogen bonds as well as intermolecular van der Waals forces.10 Although polymorph has been documented for cellulose, X-ray diffraction analysis indicates that native cellulose occurs as cellulose I, which is a mixture of polymorph Ia and Ib.11 Cellulose Ia is synthesized simultaneously with the extension of the microfibril network, which is dominant in some bacteria and lower plants to form the primary wall, but cellulose Ib is deposited mainly within the secondary wall of higher plants for strength. The decipherment of crystalline structure for cellulose indicates that cellulose Ia is characterized by the triclinic unit containing one chain, while there are two chains in the monoclinic unit of cellulose Ib for more intramolecular hydrogen bonds, making it more stable. Therefore, harsh conditions are needed to transform cellulose Ib in plant biomass into amorphous polymorph characterized by crystallinity index (CI) for more efficient hydrolysis by cellulases.12

52

Fuel Ethanol Production From Lignocellulosic Biomass

3.05.2.2

Hemicelluloses

Hemicelluloses are heterogeneous polysaccharides with the b(1/4)-linked backbone structure of pentose (C5) sugars such as xylose and arabinose and hexose (C6) sugars including mannose, galactose and glucose as the repeating units, which have the same equatorial configuration at C1 and C4, as illustrated in Fig. 3. The structural similarity of hemicelluloses with the b-1,4glycosidic bonds of cellulose molecule benefits for a conformational homology, which can lead to a strong non-covalent association with cellulose microfibrils. However, unlike cellulose that is crystalline and resistant to degradation, hemicelluloses are amorphous and thus easily hydrolyzed into monomer sugars. However, hemicelluloses are embedded and interact with cellulose and lignin, which significantly increase the strength and toughness of plant cell wall. Xyloglucan and xylan are major hemicelluloses in plant biomass. Xyloglucan is abundant in the primary wall, with the oligosaccharide composed of xylose (X) and glucose (G) with various side chains, XXXG or XXGG for vascular plants including grain crops, as the repeating unit. Xylan is polysaccharide with b-(1/4)-linked xylose residues as a backbone, which is often acetylated at the O-3 position of xylose residues and/or modified by a-(1/2)-linked glucuronosyl and 4-O-methyl glucuronosyl residues. Xylan, also known as glucuronoxylan, is the dominating noncellulosic polysaccharide in the secondary wall of dicots. A schematic illustration of xyloglucan and xylan is illustrated in Fig. 4. Therefore, major sugars in the hydrolysate of hemicelluloses are xylose, arabinose, glucose and galactose.

Figure 3

Repeating units of hemicelluloses. Reprinted from [13] with permission.

Xyloglucan (The arrow indicates the typical cleavage)

Glucuronoxylan, typical dicot structure Figure 4

Diagram of xyloglucan and xylans. Reprinted from [13] with permission.

3.05.2.3

Lignin

Although lignin is a non-sugar based polymer and cannot be used as feedstock for ethanol production through microbial fermentation, it exerts significant impact on the techno-economic performance of the production of 2G fuel ethanol. On the one hand, phenolic inhibitors of microbial growth and fermentation are produced from the degradation of lignin during pretreatment that

Fuel Ethanol Production From Lignocellulosic Biomass

53

is needed to render cellulose amenable to cellulase attack. On the other hand, lignin could deposit onto the surfaces of cellulose molecules, even absorb cellulases, compromising the effectiveness of enzymatic hydrolysis of cellulose.14,15 Moreover, when lignin is not removed before fermentation, it occupies reaction space within fermenters, and limits solid uploading of the substrate for more sugars to produce ethanol at high titers. As a result, it is almost impossible for ethanol fermentation from lignocellulosic biomass to achieve high ethanol titers as that achieved for ethanol fermentation from sugar- and starch-based feedstocks, and much more stillage with large amounts of solid residues are discharged after ethanol distillation. However, lignin yields more energy when it is burnt, and thus it is a good selection for combined heat and power (CHP) production as an eco- and enviro-friendly energy supply for fuel ethanol production.16 Moreover, lignin is an excellent starting material for various products including transportation fuels and chemicals as well as value-added chemicals, which would add credits on the bioconversion process and make cellulosic ethanol more economically competitive in case it was valorized properly.17 Understanding fundamentals with lignin biosynthesis is a prerequisite for developing more efficient pretreatment and conditioning processes and enzymatic hydrolysis of cellulose as well as engineering microorganisms with improved tolerance to phenolic inhibitors. As illustrated in Fig. 5, lignin biosynthesis starts with the deamination of phenylalanine for cinnamic acid, followed by the modification of the aromatic ring by hydroxylation and O-methylation and the reduction of the side chain to an alcohol moiety, resulting in three major monolignols: p-coumaryl, coniferyl and sinapyl alcohols, which are exported across the plasma membrane into the apoplast. The fraction of these monolignols varies substantially among plant species and tissues in same plant as well as subcellular locations, which is also affected by stages with plant development and environmental stimuli. In addition to the three canonical monolignols, many other compounds are also involved in the biosynthesis of lignin, particularly ferulates, coniferaldehyde and acylated monolignols,18 which might be liberated during the pretreatment of lignocellulosic biomass.

Once incorporated into lignin, the three major monolignols are referred to as p-hydroxyphenyl (H), guaiacyl (G) and syringyl (S) phenylpropanoid units, respectively. The most frequent inter-unit linkage is the β-O-4 linkage, which is also the one most easily broken chemically. The other linkages such as the β-5 linkage are more resistant to chemical degradation.

Figure 5

Schematic diagram of lignin biosynthesis. Reprinted from [18] with permission.

3.05.2.4

Other Components

In addition to major components of cellulose and hemicelluloses to be hydrolyzed into sugars for ethanol fermentation and lignin left before or after ethanol fermentation, other components like proteins and ashes also affect the process economics, which have not been addressed adequately up till now, even completely neglected, since no commercial production of fuel ethanol from lignocellulosic biomass has been practiced. For example, extra nutrients are needed to nourish ethanologenic microorganisms, either

54

Fuel Ethanol Production From Lignocellulosic Biomass

S. cerevisiae or Zymomonas mobilis that are engineered for ethanol production from lignocellulosic biomass, due to insufficient nutrients in the feedstock, which inevitably increases cost for the production of 2G fuel ethanol. Like cellulose, hemicelluloses and lignin, nutritional components in lignocellulosic biomass also vary with species and variety, climate, soil conditions and fertilization practice. For major agricultural residues such as corn stover, wheat and rice straw, the content of proteins is 3%–5%, much lower than that in starch-based feedstock such as corn. Moreover, ethanologenic microorganisms cannot break down these proteins into assimilable amino acids, and the supplementation of proteases that has been practiced in ethanol production from starch-based feedstock is also needed to hydrolyze proteins with lignocellulosic biomass, providing nitrogen sources to support microbial growth and ethanol fermentation. Meanwhile, ammonia or urea can be used directly as assimilable nitrogen sources, and corn steep liquor (CSL), another cost-effective nutrient, can also be supplemented, which not only provides assimilable nitrogen sources but also trace minerals and vitamins, particular for cellulosic ethanol production from corn stover, since CSL is a by-product of corn wet-milling processes with a reliable supply nearby. As for macronutrients such as phosphor, potassium, calcium and other minerals, they are usually sufficient due to high ash contents in lignocellulosic biomass.

3.05.3

Pretreatment

The self-assembly architecture of plant cell wall with crystalline cellulose microfibrils interacted and entangled with hemicelluloses and lignin creates LCCs,3 making cellulose inaccessible for cellulases to bind onto for hydrolysis. Therefore, after a preliminary size reduction through mechanical methods such as chopping, pretreatment is needed to destruct LCCs for efficient enzymatic hydrolysis of the cellulose component. An ideal pretreatment process should maximize sugar yield from cellulose and hemicelluloses, and in the meantime minimize energy consumption and environmental impact with less consumption of chemicals. Comprehensive review on pretreatment is available elsewhere, and some leading technologies are commented.

3.05.3.1

Physicochemical Pretreatment

Pure physical pretreatment does not use any chemicals. The size reduction by mechanical methods such as chopping is one of them, through which the surface area of lignocellulosic biomass is increased, and the DP and crystallinity of cellulose may also decrease to some extent under ultra-fine milling conditions, but energy consumption for reducing the feedstock from the size of millimeters to fine particles of micrometers is extremely high, which is unacceptable from the viewpoint of engineering design and process operation. Radiation like microwave that can penetrate and heat lignocellulosic biomass instantly has also been studied, but it is problematic to process the feedstock at mass quantity for fuel ethanol production, needless to say the big power requirement for generating the radiation. Therefore, more attention has been focused on hydrothermal pretreatment by liquid hot water (LHW), during which physicochemical reactions occur to hydrolyze some components in lignocellulosic biomass through autocatalysis, particularly hemicelluloses that are easily hydrolyzed under the pretreatment conditions to expose cellulose for enzymatic hydrolysis. As illustrated in Fig. 6, slurry is pre-heated via a heat exchanger, which not only saves steam consumption for heating the slurry, but also cooling water to cool down the pretreated material. The pre-heated slurry is further heated by steam via another heat exchanger, and then passes through the reactor with a designated holding time for pretreatment. Theoretically, the reactor should be operated at plug flow. Therefore, tubular reactors are preferred, and residence time and temperature can be optimized for different feedstocks. This process is also termed as physicochemical pretreatment, and underlying mechanism is assumed to be the partial degradation of LCCs catalyzed by organic acids, mainly acetic acid released from the hydrolysis of acetylated hemicelluloses, making the process autohydrolytic in nature.19 For LHW pretreatment, the pH of the feedstock might drop below 4,

Figure 6

Schematic diagram of the liquid hot water (LHW) process. Reprinted from [8] with permission.

Fuel Ethanol Production From Lignocellulosic Biomass

55

resulting in the formation of inhibitors due to the degradation of sugars under acidic conditions, and a pH control strategy can be applied to the process to maintain the pH above 4, preferably between 5 and 7 by adding base.20 Since the alkali is not a catalyst as in alkaline pretreatment to be addressed thereafter, this process is termed as the pH-controlled LHW pretreatment. Another physicochemical pretreatment of lignocellulosic biomass is steam explosion (SE), during which the feedstock is heated to elevated temperature by statured steam, maintaining for a short period of time at designated temperature and pressure, followed by depressurizing instantly to disrupt LCCs by explosion. In addition to the autohydrolysis of hemicelluloses by weak acids as that occurs during LHW, lignocellulosic biomass is torn up by force created during the depressurizing process, and in the meantime, hydrolyzed products are released more efficiently for cellulose to be exposed more completely.21 Moreover, compared to LHW operated at slurry conditions, much higher solid uploading can be applied to SE. Due to the advantages of lower capital investment, less impact on environment and simple process design and operation, SE has been intensively studied and tested at pilot scales around the world, which is highlighted in the report Process Design and Economics for Biochemical Conversion of Lignocellulosic Biomass to Ethanol released by the National Renewable Energy Laboratory (www.nrel.org), Golden, CO, USA. Acid (H2SO4) can be supplemented into the SE process at low concentrations so that it can be operated at less severe conditions to improve sugar yield, making the process a hybrid of physicochemical and chemical processes, which will be addressed in more detail in Section 3.05.3.2: Chemical pretreatment. Both batch and continuous processes have been developed for SE, as illustrated in Fig. 7. The batch process is very simple. Humidified feedstock is fed through a screw feeder into the reactor, which is then pressurized by saturated steam and maintained for a period of time. After the reaction, the material is discharged into the explosion tank operated at atmospheric pressure, in which volatile components are separated, and pretreated biomass is left for washing to collect sugars released during the hydrolysis of hemicelluloses. To overcome the disadvantage of the discontinuity, multiple reactors can be designed and operated alternatively. In contrast, the continuous system is more productive and effective, but the design of the reactor and discharger is more complicated due to high solid uploading with the feedstock as well as high pressure required by the pretreatment.

Figure 7

Diagram for batch (A) and continuous (B) steam explosion. Reprinted from [8] with permission.

3.05.3.2

Chemical Pretreatment

High temperature applied during physicochemical pretreatment under LHW and SE conditions dehydrates sugars and produces inhibitors such as furfural from xylose and hydroxymethylfurfural (HMF) from glucose. To address this problem, acids can be supplemented to facilitate the destruction of LCCs under less severe conditions, either lower temperature or shorter reaction time. Among various acids, sulfuric acid is most commonly used. Although temperature for concentrated acid pretreatment is much lower, acid recovery presents a big challenge. Therefore, dilute acid with concentration less than 2% is preferred, which can be neutralized conveniently by lime or ammonium during conditioning.22

56

Fuel Ethanol Production From Lignocellulosic Biomass

Dilute acid pretreatment has been intensively studied for various feedstocks, and Fig. 8 highlights the process previously developed at NREL. Milled corn stover is conveyed into a screw feeder, and dilute acid is injected at the discharge. The feedstock is then fed into a mixing and heating screw, and further conveyed into the vertical presteamer. Hot water is added to make the effluent at 30% total solids. The presteamer is operated at 100  C, with a retention time of 10 min. The feedstock is discharged through the screw feeder, and acid is added again at the discharge to a concentration of 18 mg g1 dry biomass before feeding into the horizontal reactor, which is operated at 158  C (0.55 MPa), with a residence time of 5 min. The feedstock from the horizontal reactor is discharged into a blowdown tank operated at 130  C (0.28 MPa). The slurry from the blowdown tank goes into the oligomer conversion tank, where an additional 4.1 mg acid/g feedstock is added, making the total acid loading to 22.1 mg g1 dry biomass. The oligomer conversion tank is also maintained at 130  C, with a residence time of 20–30 min. After that, the feedstock is discharged into a flash tank operated at atmospheric pressure. At this stage, the hydrolysate containing 30% total solids and 16.6% insoluble solids is pumped into the conditioning tank, in which the slurry is diluted to 20% for enzymatic hydrolysis and cooled down to 75  C. Ammonia is sparged into the dilution water to adjust the hydrolysate pH to 5 as well as provide nitrogen source for subsequent microbial growth and ethanol fermentation. All volatile components from the blowdown tank, oligomer conversion tank and flash tank are condensed and collected.

Figure 8 Process diagram of the steam explosion and dilute-acid pretreatment of corn stover developed by NREL. Reprinted from [8] with permission.

Although dilute acid pretreatment seems more economically competitive, some disadvantages like corrosion that requires expensive acid-resistant stainless steel or coating and inhibitors produced during the pretreatment under high temperature endeavor to explore alternatives, and alkaline pretreatment is one of them. Various alkalis including sodium hydroxide, sodium carbonate, lime and aqueous ammonia have been studied.23 Basically, alkaline pretreatment is a delignification process, and underlying mechanism is the cleavage of hydrolysable linkages between lignin and glycosidic bonds of polysaccharides to destruct LCCs, causing a reduction of the gross crystallinity of lignocellulosic biomass as well as the disruption of lignin structures for dissolution. In addition, alkaline de-esterification or saponification of intermolecular ester bonds crosslinking xylan hemicelluloses and lignin improves the accessibility of cellulases to cellulose.24 The effectiveness of alkaline pretreatment depends on the characteristics of lignocellulosic biomass and reaction conditions. In general, alkaline pretreatment is more efficient with herbaceous crops and agricultural residues with relatively low lignin contents. Sodium hydroxide is one of the strongest bases with relatively low cost, which has been intensively studied for the pretreatment of lignocellulosic biomass, and its effectiveness is evident. However, significant amounts of salt produced during the process and remained with the pretreated biomass are a big problem, which inhibit microbial growth and ethanol fermentation if they are not being washed away. Moreover, the black liquor with lignin dissolved raises similar environmental concerns as that previously observed with the pulping industry. An alternative solution for these problems is to replace sodium hydroxide with ammonia, but ammonia hydroxide is a weak base, which is less efficient for the delignification of lignocellulosic biomass and many other chemical reactions to destruct LCCs, and special designs are needed for such a process. In general, pressurized reactors for ammonia to penetrate into lignocellulosic biomass more effectively are preferred. Among various technologies that have been developed for the pretreatment of lignocellulosic biomass by ammonia, ammonia fiber explosion/expansion (AFEX), a combination of ammonia pretreatment and explosion, seems more promising due to its relatively high productivity.25 As illustrated in Fig. 9, lignocellulosic biomass is pretreated with ammonia at mild temperature and high pressure within the AFEX reactor. When it is discharged with pressure released, the rapid expansion of ammonia gas causes swelling of the feedstock, which correspondingly disrupts LCCs, creating more accessible surfaces with the cellulose component for

Fuel Ethanol Production From Lignocellulosic Biomass

57

enzymatic hydrolysis, and in the meantime makes ammonia separated from the pretreated biomass for recovery. Since temperature for AFEX is much lower than that applied to SE, not only can energy consumption be saved but also the formation of inhibitory by-products prevented. In addition, washing is not necessary for biomass pretreated by AFEX, which benefits for high solid loading thereafter in enzymatic hydrolysis of the cellulose component. Meanwhile, ammonia remained with the pretreated biomass can be used as nitrogen source for microbial growth and ethanol fermentation. Recently, an improved AFEX process termed as extractive ammonia (EA) was developed, through which lignin was extracted more efficiently, which not only made the cellulose component more accessible for enzymatic hydrolysis to save the enzyme loading but also separated lignin for being processed as value-added products.26

Figure 9

Process diagram for ammonia fiber explosion/expansion pretreatment. Reprinted from [25] with permission.

3.05.3.3

Solvent Extraction

Solvent extraction is a fractioning process, in which an organic solvent is usually used to destruct LCCs.27 Technically, either cellulose or lignin can be targeted for solvent extraction, but lignin is more preferred due to its relatively low contents in lignocellulosic biomass to reduce the amount of solvent to be recovered for reuse. Compared with physicochemical and chemical pretreatments, relatively mild temperature and pressure as well as a neutral pH environment applied to the solvent pretreatment of lignocellulosic biomass can reduce carbohydrate degradation into undesired by-products such as furfural and HMF. Another advantage of solvent extraction is that pure lignin fragments can be recovered as a by-product, which might be processed as value-added products. Alcohols with low boiling points such as ethanol and methanol are favored because of their low cost and easy recovery. However, high boiling point alcohols including ethylene glycol and glycerol offer low demands on temperature and pressure but increase energy consumption for solvent recovery. Recently, ionic liquids (ILs) and IL-based solvent systems have been developed for biomass pretreatment through selective dissolution of cellulose or lignin.28 ILs are molten salts composed completely of paired ions, which are in liquid state at low temperature below 100  C in general, particularly at room temperature for the purpose of biomass pretreatment. Based on the understanding of the chemistry of the anion and cation and their interactions, various ILs can be designed to dissolve either cellulose or lignin from lignocellulosic biomass, and consequently destruct LCCs and crystalline structure of cellulose molecules for subsequent enzymatic hydrolysis. Compared to conventional solvents, IL-based solvents are more environmentally friendly (green). On the one hand, the recovery of ILs is less energy-intensive compared to distillation or evaporation for recovering conventional solvents, since dissolved cellulose or lignin can be precipitated by adding specially designed anti-solvents. On the other hand, vapor pressure is very low with ILs, and thus much less hazards are released into workshops and the environments. However, there are still many challenges for ILs to be economically competitive for the pretreatment of lignocellulosic biomass to produce bulk commodities like ethanol. Recently, a series of tertiary amine-based ILs were synthesized from aromatic aldehydes derived from lignin and hemicelluloses,29 which might open a window for addressing those challenges.

58

Fuel Ethanol Production From Lignocellulosic Biomass

3.05.3.4

Biological Pretreatment

Compared to major pretreatment technologies reviewed previously, biological pretreatment through solid fermentation employs microorganisms that selectively degrade lignin in lignocellulosic biomass to expose the cellulose component for enzymatic hydrolysis at mild conditions without special requirements for instruments.30 Both bacteria and fungi have been explored for such a purpose, but rot fungi associated with wood decay are predominant species for lignin degradation, particularly white-rot fungi due to their abundant ligninolytic enzymes including lignin peroxidase, manganese peroxidase, laccases and other enzymes for better selectivity in lignin degradation.31 Although biological pretreatment is energy-saving and environmentally friendly, its disadvantages are apparent. Firstly, extremely low degradation rate requires time as long as weeks for a significant change in the structure of LCCs, making the process unmatchable with the subsequent hydrolysis of the cellulose component and ethanol fermentation. Secondly, significant biomass is lost during the process, not only the lignin that is mineralized into low-molecular weight fragments which would be further catabolized into final product CO2, but also sugars released from hemicelluloses, even cellulose by hydrolytic enzymes (simultaneous decay with lignin degradation) as carbon source to support the growth of microorganisms. Finally, it is unreliable to control microbial growth and metabolism under open and solid fermentation conditions with mixed species, which inevitably increases contamination risk for subsequent cellulose hydrolysis and ethanol fermentation. Unless these problems are well addressed, biological pretreatment would not be practical for the production of 2G fuel ethanol at commercial scales from the viewpoint of engineering design and process operation.

3.05.4

Enzymatic Hydrolysis of Cellulose and Co-Fermentation of C5 and C6 Sugars

After pretreatment, enzymatic hydrolysis is needed to further depolymerize the cellulose component into glucose, which, together with sugars released from the hydrolysis of hemicelluloses during the pretreatment, can be used for ethanol fermentation. Although intensive R & D effort has been performed around the world for decades, two barriers still remain to be overcome for developing a viable process to make bioethanol economically competitive. Unlike amylases and glucoamylases that are available at low prices for commercial production of various bulk products including ethanol from starch-based feedstocks, cellulases to liberate glucose from cellulose for bioethanol production are more expensive due to low efficiency with their fermentation production as well as the heterogeneous cellulose hydrolysis that significantly compromises the reaction rate and increases the enzyme dosage.32,33 As a result, glucose released from enzymatic hydrolysis of the cellulose component is by no means cheaper as expected, although lignocellulosic biomass is extremely cheap. On the other hand, all ethanologenic species, whether S. cerevisiae for ethanol production from sugar- and starch-based feedstock or Z. mobilis that has been studied for cellulosic ethanol production, cannot ferment pentose sugars released from the hydrolysis of hemicelluloses into ethanol at rate and yield that are acceptable from the viewpoint of industrial production. Although pentose sugars can be converted into other products like furfural through intramolecular dehydration of xylose by chemical catalysis, and xylitol, lactic acid and 2,3-butanediol by fermentations, those processes are not economically competitive, and the co-fermentation of pentose and hexose sugars by engineered microbial strains still stands very firmly for bioethanol production.6 Based on the arrangement for cellulase production, cellulose hydrolysis and ethanol fermentation, different strategies have been developed for enzymatic hydrolysis of the cellulose component and co-fermentation of C5 and C6 sugars, which are schematically shown in Fig. 10.

3.05.4.1

Separate Hydrolysis and Co-Fermentation

For SHCF, cellulose is completely hydrolyzed into glucose by cellulases under optimum conditions, particularly temperature around 50  C that is optimal for the enzymatic hydrolysis, and consequently saves cost for enzyme consumption, but such a high temperature cannot be tolerated by S. cerevisiae or Z. mobilis, which in general performs ethanol fermentation at temperature below 35  C. After a complete cellulose hydrolysis, lignin is left, which can be recovered by a filter to be processed as value-added by-products. As a result, the viscosity of the hydrolysate is low, making it suitable for high gravity (HG) fermentation to increase ethanol titers, saving energy consumption for ethanol distillation as well as reducing stillage discharge. However, the accumulation of glucose during the hydrolysis significantly inhibits the activity of b-glucosidase, which consequently results in the accumulation of cellobiose to further inhibit the activity of cellobiohydrolase. The supplementation of b-glucosidase could be a solution to address this problem if its cost is not too high, for example, b-glucosidase commercially produced by Aspergillus niger. On the other hand, microbial contamination during cellulose hydrolysis, hydrolysate transport through pumps and pipelines and even during ethanol fermentation is another concern with SHCF, which might substantially compromise ethanol yield. At present, SHCF is being explored by researchers and industry. For example, the Switzerland-based company Clariant has developed SHCF for its sunliquid demonstration plant in Straubing, Germany,34 and Project Liberty with the joint venture of POET and DSM in Emmetsburg, USA, also employed SHCF for cellulosic ethanol production, although details are not publically available for comments.

Fuel Ethanol Production From Lignocellulosic Biomass

Figure 10

3.05.4.2

59

Strategies for enzymatic hydrolysis of cellulose and co-fermentation of C5 and C6 sugars. Reprinted from [8] with permission.

Simultaneous Saccharification and Co-Fermentation

For ethanol fermentation from starch-based feedstock, mash is heated to 110–120  C and maintained for 10–20 min, which is then flashed to about 90–95  C and maintained for 60–90 min for amylase to hydrolyze starch completely into dextrin. This two-step process is termed as cooking or liquefaction, which also sterilizes the mash to control contamination with saccharification and ethanol fermentation thereafter. After the liquefaction, mash is cooled down to 60–65  C for glucoamylase to be supplemented to hydrolyze the dextrin into glucose and other fermentable sugars, but the saccharification process is maintained for 20– 30 min only to achieve a dextrose equivalent of 15–20. The mash is further cooled down to 30–34  C for yeast to be inoculated to start ethanol fermentation. Since most dextrin is hydrolyzed into sugars during the fermentation, the process is termed as simultaneous saccharification and fermentation (SSF), which has been well practiced in the industry. When a similar strategy is applied for ethanol production from lignocellulosic biomass, an acronym SSCF is fabricated, taking into account the unique characteristic of the hydrolysate that includes both C5 and C6 sugars. For SSCF, the pretreatment, particularly physicochemical pretreatment under much higher temperature, not only destructs the LCCs to expose cellulose for enzymatic

60

Fuel Ethanol Production From Lignocellulosic Biomass

hydrolysis but also sterilizes lignocellulosic biomass to reduce contamination risk with cellulose hydrolysis and ethanol fermentation. However, the saccharification of dextrin/pretreated cellulose and the fermentation/co-fermentation of C5 and C6 sugars are by no means simultaneous, but sequential in nature. SSCF is simple in design and easy for operation. Most importantly, higher ethanol yield could be achieved due to the alleviation of product inhibition in the activities of cellulases for more complete cellulose hydrolysis. However, temperature for cellulose hydrolysis by cellulases and ethanol fermentation by microbes, either S. cerevisiae or Z. mobilis, is significantly different, making a simultaneous optimization for the two processes impossible. Therefore, SSCF must be operated at lower temperature to accommodate microbial growth and ethanol fermentation, normally at 30–34  C. As a result, rate for enzymatic hydrolysis of cellulose is compromised, and much longer time is needed to complete the process. Moreover, lignin cannot be separated from cellulose prior to ethanol fermentation under SSCF conditions, making the slurry extremely viscous. It is difficult for SSCF to be operated under HG conditions for high ethanol titers, and energy consumption is high for ethanol recovery by distillation due to low ethanol titers achieved during the fermentation and large amounts of stillage discharged from the distillation unit. Fed-batch, even multi-feed batch, can address this problem to some extent, but ethanol titers achieved are still low. For example, fed-batch and multi-feed batch processes were reported for SSCF with the solid uploading of 11.7% and 20% (w/w), respectively, producing 37.5 g L1 and 57.0 g L1 ethanol from steam pretreated wheat straw at 48 and 72 h.35,36 A hybrid process much like SSF practiced in ethanol fermentation from starch-based feedstock can be developed, in which a prehydrolysis under the optimal temperature condition is applied to cellulose hydrolysis by cellulases, followed by the SSCF process to shorten the time required by the hydrolysis and co-fermentation to improve the productivity of the whole system,37 but impact of lignin on the rheology of the slurry still cannot be overcome.

3.05.4.3

Consolidated Bioprocessing

Cellulases are produced separately and supplemented to hydrolyze the cellulose component of the pretreated biomass for ethanol production by the SHCF and SSCF processes, which is one of bottlenecks for cost reduction due to the high cost of cellulases produced predominantly by strains from T. reesei. In nature, many organisms, particularly microbes, can utilize native cellulose as carbon source and energy to support their growth and metabolism through synthesis and secretion of unique cellulases to hydrolyze cellulose by the synergic functions of different enzymes. Such a natural phenomenon inspires scientists to develop mimic systems, either an individual microorganism or a microbial community, to produce ethanol and other chemicals directly from lignocellulosic biomass, even without pretreatment, which was proposed almost one decade ago and termed as CBP.38 However, no natural microorganisms are available for commercial production of bioethanol through such a CBP strategy. Thus, the development of CBP strains is a prerequisite. Currently, both bacterial and yeast species have been explored for this purpose with the following strategies: (1) engineering cellulase producers or microorganisms with ethanol production, and (2) engineering an ethanologen with cellulase production.39 For the first strategy, anaerobic cellulolytic bacteria from the genera of Clostridium are good candidates, and the targets for metabolic engineering of this species include increasing ethanol titer by improving ethanol tolerance through rational designs based on the understanding of mechanism underlying its response to ethanol inhibition and random approaches such as the selection of mutants through evolutionary adaptation, and on the other hand improving ethanol yield by blocking the synthesis of major by-products.40 As for the second strategy, primary concerns are the expression and secretion of functional cellulases in ethanologenic species, particularly S. cerevisiae, which has been engineered with genes encoding glycoside hydrolases including cellulases and hemicellulases through cell surface display.41 Unfortunately, the expression of cellobiohydrolases (CBH I and CBH II) from T. reesei in S. cerevisiae is generally poor, needless to say challenges to engineer the species with more other enzymes or pathways required by efficient production of cellulosic ethanol. Theoretically, the CBP strategy can completely eliminate cellulase production and integrate cellulase production, cellulose hydrolysis and ethanol fermentation into a single unit. However, there are many unknowns to be elucidated in order to make it significant in the production of bioethanol. For example, the production of cellulolytic enzymes, hydrolysis of cellulose and hemicelluloses and co-fermentation of released sugars need to be well coordinated within the single cell and between cells and their surroundings at different magnitudes, from molecular levels involving gene expression and regulation to intracellular metabolic network to kinetics of the heterogeneous hydrolysis with mass transfer limitation.

3.05.5

Strain Development

Unlike conventional sugar- and starch-based feedstocks, hydrolysate of lignocellulosic biomass contains large amounts of pentose sugars such as xylose and arabinose in addition to hexose sugars of glucose, mannose and galactose. Unfortunately, native ethanologenic species, either S. cerevisiae or Z. mobilis, cannot metabolize pentose sugars. If hexose sugars are fermented only with pentose sugars unconverted, feedstock consumption for bioethanol production will be substantially high, and in the meantime unfermented pentose sugars will be discharged with stillage, and consequently increase workload for stillage treatment. Therefore, engineering ethanol producers with the co-fermentation of pentose and hexose sugars for more efficient production of cellulosic ethanol has been pursued endlessly within the past four decades.

Fuel Ethanol Production From Lignocellulosic Biomass

61

Naturally, some yeast species metabolize xylose to xylulose using xylose reductase (XR) and xylitol dehydrogenase (XDH), but bacteria convert xylose directly to xylulose by xylose isomerase (XI). Xylulose can be further metabolized through the non-oxidative pentose phosphate pathway to enter either the Embden-Meyerhof-Parnas (EMP) pathway with S. cerevisiae or the Entner-Doudoroff (ED) pathway with Z. mobilis. In order to speed up xylulose metabolism, xylulokinase (XK) needs to be overexpressed. Thus, an overall strategy for engineering S. cerevisiae or Z. mobilis with xylose metabolism for ethanol production can be illustrated in Fig. 11.42

Figure 11

3.05.5.1

Strategies for engineering xylose-metabolizing pathways in bacteria and yeast. Reprinted from [8] with permission.

Engineering S. cerevisiae

Currently, fuel ethanol is solely produced from sugar- and starch-based feedstocks by S. cerevisiae, which exhibits significant advantages over other species. For example, S. cerevisiae is tolerant to ethanol, and high ethanol titer can be achieved to save energy consumption not only for ethanol distillation but also for stillage treatment due to the significant reduction of the discharge. At present, 15% (v/v) ethanol can be produced from corn powder in the industry. Moreover, S. cerevisiae prefers acidic environments with pH values below 4.5, which can effectively prevent microbial contamination, since as a bulk commodity fuel ethanol is marketed with low price, and fermenters used in the industry are too large to be sterilized by vapor. Since the 1980s, great effort has been devoted to engineering S. cerevisiae with xylose metabolism for cellulosic ethanol production.43 Some xylose-metabolizing yeast such as Pichia stipitis can metabolize xylose to xylulose through the two-step reduction/ oxidation pathway, and S. cerevisiae can further metabolize xylulose through the PPP pathway. If such a heterologous xylosemetabolizing pathway were engineered into S. cerevisiae, together with the overexpression of XK, the recombinant would be able to ferment xylose into ethanol. In the middle of the 1990s, Ho’s group at Purdue University, USA, reported the successful development of S. cerevisiae 1400 (pLNH32) that could ferment xylose nearly completely to ethanol.44 In addition, the recombinant strain could co-utilize glucose and xylose without significant lag period between the fermentation of the two sugars. The Purdue strain was developed by transforming an industrial strain S. cerevisiae 1400 with a high copy number of the 2m plasmid pLNH 32, which contains genes XR, XD and XK. The 2m plasmid was designed for broad hosts. Furthermore, Ho’s group developed a unique gene integration technique, facilitating effective integration of multiple genes into the chromosome of S. cerevisiae in multiple copies, which is easy to perform and guarantees the genes cloned on the integration plasmid to be transferred into the host strains and integrated into their genome in as many copies as desired to provide desired activities for the cloned enzymes. This technique allows integrating XRXDXK together as a cassette into the chromosome of S. cerevisiae with sufficient copies until the resulting recombinant ferments xylose efficiently. Based on the above progress, Ho’s group further developed another strain S. cerevisiae 424A (LNH-ST) for cellulosic ethanol production,45 and Fig. 12 shows the time-course for the co-fermentation of glucose/xylose. This strain has been validated by ethanol producers to be able to co-ferment glucose and xylose in hydrolysates from different lignocellulosic feedstocks, which is currently available for industry to produce cellulosic ethanol. Dr. Ho and her coworkers have continued to improve the strain by making it coferment other sugars like arabinose, together with glucose, xylose, mannose and galactose, and more resistant to inhibitors. A new and improved derivative strain has been developed to ferment all sugars present in hydrolysates from any cellulosic biomass, producing more than 10% (v/v) ethanol without special detoxification to remove inhibitors in the hydrolysates. Although progress has been made by Dr. Ho and her co-workers in engineering S. cerevisiae with the XRXDXK cassette for ethanol production through the co-fermentation of xylose and glucose, other scientists have experienced challenges with recombinant S. cerevisiae strains they developed: fermenting xylose slowly with much less, even no ethanol produced, since the two key enzymes for xylose metabolism in yeast require different co-factors, NADPH/NADPþ for XR and NADþ/NADH for XDH, which cannot be regenerated in engineered S. cerevisiae for a balance with the co-factors, and consequently results in xylitol accumulation.

62

Fuel Ethanol Production From Lignocellulosic Biomass

Figure 12 Co-fermentations of glucose and xylose in simulated medium (A) and wheat straw hydrolysate (B) by the recombinant Saccharomyces cerevisiae 424A(LNH-ST). Reprinted from [8] with permission.

Engineering co-factor preference has thus been explored for S. cerevisiae engineered with the expression of XR and XDH for xylose metabolism to address this challenge so that xylitol could be directed to xylulose. For example, XR with NADH-preference was expressed in S. cerevisiae for xylose metabolism to produce ethanol more efficiently.46 Recently, significant progress has been made in engineering S. cerevisiae with XI overexpression for ethanol production.47,48 In fact, such a strategy was explored at the beginning of engineering S. cerevisiae for xylose metabolism four decades ago, but not successful due to the lacking of scientific fundamentals with the enzyme and advanced tools for genetic modifications on S. cerevisiae. Theoretically, engineering S. cerevisiae with XI overexpression would be more effective for xylose to be metabolized into xylulose than engineering the species with the overexpression of both XR and XDH with an imbalance of the co-factors. In addition, transporter engineering has been also developed and applied to engineered S. cerevisiae strains to further improve their robustness for the co-fermentation of pentose and hexose sugars.49,50

3.05.5.2

Engineering Z. mobilis

Z. mobilis is a facultative anaerobic gram-negative bacterium, which is generally regarded as safe (GRAS). Compared to S. cerevisiae that metabolizes glucose through the EMP pathway with 2 mol of ATP produced from 1 mol of glucose metabolized, Z. mobilis metabolizes glucose into ethanol and CO2 through the ED pathway with only 1 mol of ATP produced from the same amount

Fuel Ethanol Production From Lignocellulosic Biomass

63

of glucose consumed.51 As a result, less biomass is accumulated during ethanol fermentation by this bacterial species, and ethanol yield, the most important factor for fuel ethanol production, would be improved. Moreover, cells of Z. mobilis are much smaller in size to provide more surfaces for glucose uptake to produce ethanol rapidly, making it with a nickname of catabolic highway.52 Albeit with the aforementioned advantages for ethanol fermentation, Z. mobilis has never been commercially used for ethanol production from sugar- and starch-based feedstocks. The reasons for this phenomenon are: (1) Z. mobilis can metabolize only glucose, fructose and sucrose,51 and the narrow substrate spectrum makes it not suitable for ethanol fermentation from starchbased feedstock, since many other sugars are released during the hydrolysis of dextrin by glucoamylase, and (2) when fructose or sucrose is used as substrate, ethanol yield is substantially compromised due to the formation of a large amount of by-product levan, making Z. mobilis not suitable for ethanol production from sugar-based feedstock such as molasses. However, these disadvantages for ethanol production by Z. mobilis from sugar- and starch-based feedstocks are not problematic for ethanol production from lignocellulosic biomass, since the only sugar released from enzymatic hydrolysis of cellulose is glucose. On the other hand, both Z. mobilis and S. cerevisiae cannot metabolize pentose sugars for ethanol production from lignocellulosic biomass, but engineering the bacterial species with pentose metabolism can overcome the intrinsic challenge of co-factor imbalance associated with engineering S. cerevisiae for the same purpose. The pioneer work for engineering Z. mobilis with pentose metabolism for ethanol production was performed at NREL more than three decades ago, and genes isolated from E. coli for xylose assimilation and the pentose phosphate pathway: xylose isomerase (xylA), xylulose kinase (xylB), transketolase (tktA) and transaldolase (tal) were engineered into Z. mobilis, enabling the recombinants to use xylose for growth and ethanol fermentation.53 Shortly, arabinose utilization was also engineered into Z. mobilis by expressing five genes from E. coli encoding L-arabinose isomerase (araA), L-ribulokinase (araB), L-ribulose-5-phosphate-4-epimerase (araD), transaldolase (talB) and transketolase (tktA).54 Furthermore, those genes were engineered into Z. mobilis through genomic DNAintegration for stable expression to ferment xylose and arabinose as well as glucose.55 In order to facilitate pentose metabolism, transporter engineering has been also performed for Z. mobilis.56 Although significant progress has been made in engineering Z. mobilis for ethanol production from lignocellulosic biomass, no commercial applications have been reported so far, due to incomplete understanding on the species as well as the complexity of industrial substrates, particularly the inhibition of various toxic by-products released during the pretreatment of lignocellulosic biomass and molecular mechanism underlying the response of Z. mobilis to environmental stresses, which was explored recently through quantitative proteomics and transcriptomics analysis and experimental studies as well.57–59 With the sequencing of the Z. mobilis genome and elucidation of more functional genes, together with the applications of synthetic and systems biology methodologies,60–62 more efficient bacterial strains are expected to be developed. Under the support of the DOE project: the Integrated Corn-Based Bio-Refinery (ICBR), DuPont and Broin Companies have established a partnership to produce cellulosic ethanol from corn stover by genetically modified Z. mobilis, which might be a milestone for commercial application of this species.

3.05.6

Unit Integration and System Optimization

Unit operations on pretreatment, enzymatic hydrolysis and co-fermentation of C5 and C6 sugars have been developed within the past decades for fuel ethanol production from lignocellulosic biomass. Unit integration is to optimize these units on system levels to improve the techno-economic performance of the whole process, making cellulosic ethanol economically competitive compared to fuel ethanol produced from sugar- and starch-based feedstocks as well as petroleum-based fuels. Unlike ethanol fermentation from sugar- and starch-based feedstocks that can be carried out at HG conditions with about 15% (v/v) ethanol produced, ethanol titer achieved in the fermentation of lignocellulosic biomass is much lower, only 7%–8% (v/v), due to inhibition of toxic by-products released during the pretreatment in microbial growth and ethanol fermentation as well as extremely high viscosity of the slurry when lignocellulosic biomass is uploaded at high solid contents. As a result, more water needs to be supplemented into the production system, which not only compromises productivity of the whole process but also consumes more energy in ethanol distillation and stillage treatment. The most important consideration for unit integration and process optimization is to minimize water usage for the whole process without significant compromise of enzymatic hydrolysis of the cellulose component and ethanol fermentation. Taking the COFCOSINOPECNovozyme 2G fuel ethanol project with an annual production capacity of 62 million liters as an example, we highlight the overall process, which includes feedstock handling, size reduction, pretreatment, conditioning, enzymatic hydrolysis, co-fermentation, ethanol distillation, residues dewatering and biogas production. Feedstock for this project is corn stover, which contains 10%–15% moisture under field-harvesting conditions. After primary size reduction by shredder, the feedstock is screened to remove non-metal impurities and passed through magnetic separators to remove metal impurities. Then, it is further reduced to 20–50 mm by shredder. SE pretreatment is applied with solid contents of 30%–40%. Prior to the pretreatment, the feedstock is pre-heated by flash vapor, which not only saves fresh vapor consumption but also reduces the dilution of the pretreated feedstock by condensed water. The temperature and residence time are controlled at 130–220  C and 5–120 min for the SE pretreatment, depending on feedstock properties such as composition and size, but the higher the temperature is, the shorter the residence time should be. Small amount of sulfuric acid is supplemented to accelerate the hydrolysis of hemicelluloses to destruct the LCCs for cellulases to access the surface of the cellulose component more efficiently. Moreover, the acidic condition can alleviate the severity of conditions for the SE pretreatment to reduce sugar degradation.

64

Fuel Ethanol Production From Lignocellulosic Biomass

SHCF is applied for this project. The pretreated feedstock is transferred into the hydrolysis reactor with initial dry matter contents of 20%–25% after neutralized by alkali such as lime, sodium hydroxide or ammonia. The mixing of the substrate with cellulases presents challenges due to high viscosity of the slurry at the early stage of the hydrolysis. Lab trials and pilot plant operation indicated that feeding substrate and enzymes in a fed-batch mode could improve the mixing performance and facilitate the enzymatic hydrolysis. The temperature and pH are set at 50  C and 5.0, respectively, the optimal conditions for Cellic CTec2 developed by Novozymes for the hydrolysis of cellulosic feedstock, followed by the co-fermentation of hexose and pentose sugars by the genetically engineered S. cerevisiae strain developed by Dr. Ho at Purdue University and licensed to COFCO. The yeast seed is cultured with the hydrolysate supplemented with corn steep liquor. Due to high concentrations of inhibitors and low contents of nutrients in the hydrolysate, an extended culture time is required for the seed culture, and much higher inoculation is needed to initiate ethanol fermentation, which is completed within 96–120 h. The slurry containing ethanol of 5%–7% (v/v) is distilled for ethanol recovery. The stillage discharged from the distillation system is filtered to separate lignin residues remained after the fermentation. The filtrate is digested anaerobically for biogas production, and the lignin cake is dewatered. Both biogas and lignin residues could be co-fired for vapor generation and used for CHP in the future for commercial production at larger scales.

3.05.7

Conclusions

Cellulosic ethanol is a sustainable solution for current issues with reliable supply of transportation fuels and environmental challenges. Although significant progress has been made on pretreatment of lignocellulosic biomass, enzymatic hydrolysis of the cellulose component and co-fermentation of pentose and hexose sugars over the past decades, cellulosic ethanol is still not economically competitive compared to fuel ethanol produced from sugar- and starch-based feedstocks, needless to say petroleum-based transportation fuels, making cost reduction one of the biggest challenges for cellulosic ethanol production. Taking into account of the multidisciplinary nature of the whole process, the portfolio that incorporates deep understanding of the characteristics of lignocellulosic biomass, innovations on developing more efficient microbial strains for enhanced fermentation rate and ethanol yield as well as process integration and system optimization for saving energy consumption will be relentless effort for scientists, researchers and engineers. Moreover, the development of biorefinery to utilize the feedstock more comprehensively with value-added coproducts such as bio-based materials from the lignin component would credit bioethanol production, making it more economically competitive.

References 1. Zabed, H.; Sahu, J. N.; Boyce, A. N.; Faruq, G. Fuel Ethanol Production from Lignocellulosic Biomass: An Overview on Feedstocks and Technological Approaches. Renew. Sustain. Energy Rev. 2016, 66, 751–774. 2. Naik, S. N.; Goud, V. V.; Rout, P. K.; Dalai, A. K. Production of First and Second Generation Biofuels: A Comprehensive Review. Renew. Sustain. Energy Rev. 2010, 14, 578–597. 3. Chundawat, S. P. S.; Beckham, G. T.; Himmel, M. E.; Dale, B. E. Deconstruction of Lignocellulosic Biomass to Fuels and Chemicals. Ann. Rev. Chem. Biomol. Eng. 2011, 2, 121–145. 4. Zhao, X. B.; Zhang, L. H.; Liu, D. H. Biomass Recalcitrance. Part I: the Chemical Compositions and Physical Structures Affecting the Enzymatic Hydrolysis of Lignocellulose. Biofuel Bioprod. Biorefin. 2012, 6, 465–482. 5. Zhao, X. B.; Zhang, L. H.; Liu, D. H. Biomass Recalcitrance. Part II: Fundamentals of Different Pretreatments to Increase the Enzymatic Digestibility of Lignocellulose. Biofuel Bioprod. Biorefin. 2012, 6, 561–579. 6. Kim, S. R.; Ha, S. J.; Wei, N.; Oh, E. J.; Jin, Y. S. Simultaneous Co-fermentation of Mixed Sugars: a Promising Strategy for Producing Cellulosic Ethanol. Trends Biotechnol. 2012, 30, 274–282. 7. Sticklen, M. B. Plant Genetic Engineering for Biofuel Production: towards Affordable Cellulosic Ethanol. Nat. Rev. Genet. 2010, 9, 433–443. 8. Zhao, X. Q.; Zi, L. H.; Bai, F. W.; et al. Bioethanol from Lignocellulosic Biomass. Adv. Biochem. Eng. Biotechnol. 2012, 128, 25–51. 9. Himmel, M. E.; Ding, S. Y.; Johnson, D. K.; et al. Biomass Recalcitrance: Engineering Plants and Enzymes for Biofuels Production. Science 2007, 315, 804–807. 10. Klemm, D.; Heublein, B.; Fink, H. P.; et al. Cellulose: Fascinating Biopolymer and Sustainable Raw Material. Angew. Chem. Int. Ed. 2005, 44, 3358–3393. 11. Poletto, M.; Júnior, H. L. O.; Zattera, A. J. Native Cellulose: Structure, Characterization and Thermal Properties. Materials 2014, 7, 6105–6119. 12. Park, S.; Baker, J. O.; Himmel, M. E.; Parilla, P. A.; Johnson, D. K. Cellulose Crystallinity Index: Measurement Techniques and Their Impact on Interpreting Cellulase Performance. Biotechnol. Biofuels 2010, 3, 10. 13. Scheller, H. V.; Ulvskov, P. Hemicelluloses. Annu. Rev. Plant Biol. 2010, 61, 263–289. 14. Li, H. J.; Pu, Y. Q.; Kumar, R.; et al. Investigation of Lignin Deposition on Cellulose during Hydrothermal Pretreatment, its Effect on Cellulose Hydrolysis, and Underlying Mechanisms. Biotechnol. Bioeng. 2013, 111, 485–492. 15. Martín-Sampedro, R.; Rahikainen, J. L.; Johansson, L. S.; et al. Preferential Adsorption and Activity of Monocomponent Cellulases on Lignocellulose Thin Films with Varying Lignin Content. Biomacromolecules 2013, 14, 1231–1239. 16. Lythcke-Jøgensen, C.; Haglind, F.; Clausen, L. R. Exergy Analysis of a Combined Heat and Power Plant with Integrated Lignocellulosic Ethanol Production. Energy Convers. Manag. 2014, 85, 817–827. 17. Ragauskas, A. J.; Beckham, G. T.; Biddy, M. J.; et al. Lignin Valorization: Improving Lignin Processing in the Biorefinery. Science 2014, 344, 1246843. 18. Bonawitz, N. D.; Chapple, C. The Genetics of Lignin Biosynthesis: Connecting Genotype to Phenotype. Annu. Rev. Genet. 2010, 44, 337–363. 19. Jiang, W.; Chang, S. L.; Qu, Y. S.; et al. Changes on Structural Properties of Biomass Pretreated by Combined Deacetylation with Liquid Hot Water and its Effect on Enzymatic Hydrolysis. Bioresour. Technol. 2016, 220, 448–456. 20. Li, H. Q.; Jiang, W.; Jia, J. X.; Xu, J. pH Pre-corrected Liquid Hot Water Pretreatment on Corn Stover with High Hemicellulose Recovery and Low Inhibitors Formation. Bioresour. Technol. 2014, 153, 292–299.

Fuel Ethanol Production From Lignocellulosic Biomass

65

21. Jacquet, N.; Maniet, G.; Vanderghem, C.; et al. Application of Steam Explosion as Pretreatment on Lignocellulosic Material: A Review. Ind. Eng. Chem. Res. 2015, 54, 2593–2598. 22. Behera, S.; Arora, R.; Nandhagopal, N.; et al. Importance of Chemical Pretreatment for Bioconversion of Lignocellulosic Biomass. Renew. Sustain. Energy Rev. 2014, 36, 91–106. 23. Kim, J. S.; Lee, Y. Y.; Kim, T. H. A Review on Alkaline Pretreatment Technology for Bioconversion of Lignocellulosic Biomass. Bioresour. Technol. 2016, 199, 42–48. 24. Chen, Y.; Stevens, M. A.; Zhu, Y. M.; et al. Understanding of Alkaline Pretreatment Parameters for Corn Stover Enzymatic Saccharification. Biotechnol. Biofuels 2013, 6, 8. 25. Bals, B.; Wedding, C.; Balan, V.; et al. Evaluating the Impact of Ammonia Fiber Expansion (AFEX) Pretreatment Conditions on the Cost of Ethanol Production. Bioresour. Technol. 2011, 102, 1277–1283. 26. Sousa, L. C.; Jin, M. J.; Chundawat, S. P. S.; et al. Next-generation Ammonia Pretreatment Enhances Cellulosic Biofuel Production. Energy Environ. Sci. 2016, 9, 1215–1223. 27. Zhang, K.; Pei, Z. J.; Wang, D. H. Organic Solvent Pretreatment of Lignocellulosic Biomass for Biofuels and Biochemicals: A Review. Bioresour. Technol. 2016, 199, 21–33. 28. Hou, Q. D.; Ju, M. T.; Li, W. Z.; et al. Pretreatment of Lignocellulosic Biomass with Ionic Liquids and Ionic Liquid-based Solvent Systems. Molecules 2017, 22, 490. 29. Socha, A. M.; Parthasarathi, R.; Shi, J.; et al. Efficient Biomass Pretreatment Using Ionic Liquids Derived from Lignin and Hemicellulose. Proc. Natl. Acad. Sci. USA 2014, 111, E3587–E3595. 30. Sindhu, R.; Binod, P.; Pandey, A. Biological Pretreatment of Lignocellulosic Biomass –An Overview. Bioresour. Technol. 2016, 199, 76–82. 31. García-Torreiro, M.; López-Abelairas, M.; Lu-Chau, T. A.; et al. Fungal Pretreatment of Agricultural Residues for Bioethanol Production. Ind. Crop. Prod. 2016, 89, 486–492. 32. Passos, D. F.; Pereira, N.; Castro, A. M. A Comparative Review of Recent Advances in Cellulases Production by Aspergillus, Penicillium and Trichoderma Strains and Their Use for Lignocellulose Deconstruction. Curr. Opin. Green Sustain. Chem. 2018, 14, 60–66. 33. Jeoh, T.; Cardona, M. J.; Karuna, N.; et al. Mechanistic Kinetic Models of Enzymatic Cellulose Hydrolysis  a Review. Biotechnol. Bioeng. 2017, 114, 1369–1385. 34. Rarbach, M.; Söltl, Y. sunliquid®: Sustainable and Competitive Cellulosic Ethanol from Agricultural Residues. Chimia 2013, 67, 732–734. 35. Bondesson, P. M.; Galbe, M. Process Design of SSCF for Ethanol Production from Steam-pretreated, Acetic-acid-impregnated Wheat Straw. Biotechnol. Biofuels 2016, 9, 222. 36. Wang, R. F.; Unrean, P.; Franzén, C. J. Model-based Optimization and Scale-up of Multi-feed Simultaneous Saccharification and Co-fermentation of Steam Pre-treated Lignocellulose Enables High Gravity Ethanol Production. Biotechnol. Biofuels 2016, 9, 88. 37. Cassells, B.; Karhumaa, K.; Nogué, V. S. Hybrid SSF/SHF Processing of SO2 Pretreated Wheat Straw-Tuning Co-fermentation by Yeast Inoculum Size and Hydrolysis Time. Appl. Biochem. Biotechnol. 2017, 181, 536–547. 38. Xu, Q.; Singh, A.; Himmel, M. E. Perspectives and New Directions for the Production of Bioethanol Using Consolidated Bioprocessing of Lignocellulose. Curr. Opin. Biotechnol. 2009, 20, 364–371. 39. Liu, H.; Sun, J. L.; Chang, J. S.; et al. Engineering Microbes for Direct Fermentation of Cellulose to Bioethanol. Crit. Rev. Biotechnol. 2018, 38, 1089–1105. 40. Tian, L.; Papanek, B.; Olson, D. G. Simultaneous Achievement of High Ethanol Yield and Titer in Clostridium thermocellum. Biotechnol. Biofuels 2016, 9, 116. 41. den Haan, R.; van Rensburg, E.; Rose, S. H. Progress and Challenges in the Engineering of Non-cellulolytic Microorganisms for Consolidated Bioprocessing. Curr. Opin. Biotechnol. 2015, 33, 32–38. 42. Aristidou, A.; Penttilä, M. Metabolic Engineering Applications to Renewable Resource Utilization. Curr. Opin. Biotechnol. 2000, 11, 187–198. 43. Kim, S. R.; Park, Y. C.; Jin, Y. S.; et al. Strain Engineering of Saccharomyces cerevisiae for Enhanced Xylose Metabolism. Biotechnol. Adv. 2013, 31, 851–861. 44. Ho, N. W. Y.; Chen, Z.; Brainard, A. Genetically Engineered Saccharomyces Yeast Capable of Effective Cofermentation of Glucose and Xylose. Appl. Environ. Microbiol. 1998, 64, 1852–1859. 45. Bera, A. K.; Ho, N. Y. W.; Khan, A.; et al. A Genetic Overhaul of Saccharomyces cerevisiae 424A(LNH-ST) to Improve Xylose Fermentation. J. Ind. Microbiol. Biotechnol. 2011, 38, 617–626. 46. Lee, S. H.; Kodaki, T.; Park, Y. C.; et al. Effect of NADH-preferring Xylose Reductase Expression on Ethanol Production from Xylose in Xylose-metabolizing Recombinant Saccharomyces cerevisiae. J. Biotechnol. 2012, 158, 184–191. 47. Zhou, H.; Cheng, J. S.; Wang, B. L.; et al. Xylose Isomerase Overexpression along with Engineering of the Pentose Phosphate Pathway and Evolutionary Engineering Enable Rapid Xylose Utilization and Ethanol Production by Saccharomyces cerevisiae. Metab. Eng. 2012, 14, 611–622. 48. Ko, J. K.; Um, Y. S.; Woo, H. M.; et al. Ethanol Production from Lignocellulosic Hydrolysates Using Engineered Saccharomyces cerevisiae Harboring Xylose Isomerase-based Pathway. Bioresour. Technol. 2016, 209, 290–296. 49. Farwick, A.; Bruder, S.; Schadeweg, V.; et al. Engineering of Yeast Hexose Transporters to Transport D-xylose without Inhibition by D-glucose. Proc. Natl. Acad. Sci. USA 2014, 111, 5159–5164. 50. Nijland, J. G.; Shin, H. Y.; de Waal, P. P.; et al. Increased Xylose Affinity of Hxt2 through Gene Shuffling of Hexose Transporters in Saccharomyces cerevisiae. J. Appl. Microbiol. 2017, 124, 503–510. 51. Kalnenieks, U. Physiology of Zymomonas mobilis: Some Unanswered Questions. Adv. Microb. Physiol. 2007, 51, 73–117. 52. Sprenger, G. A. Carbohydrate Metabolism in Zymomonas mobilis: a Catabolic Highway with Some Scenic Routes. FEMS (Fed. Eur. Microbiol. Soc.) Microbiol. Lett. 1996, 145, 301–307. 53. Zhang, M.; Eddy, C.; Deanda, K.; et al. Metabolic Engineering of a Pentose Metabolism Pathway in Ethanologenic Zymomonas mobilis. Science 1995, 267, 240–243. 54. Deanda, K.; Zhang, M.; Eddy, C.; et al. Development of an Arabinose-Fermenting Zymomonas mobilis Strain by Metabolic Pathway Engineering. Appl. Environ. Microbiol. 1996, 62, 4465–4470. 55. Mohagheghi, A.; Evans, K.; Chou, Y. C.; et al. Cofermentation of Glucose, Xylose, and Arabinose by Genomic DNA-integrated Xylose/arabinose Fermenting Strain of Zymomonas mobilis AX101. Appl. Biochem. Biotechnol. 2002, 98100, 885–898. 56. Dunn, K. L.; Rao, C. V. Expression of a Xylose-specific Transporter Improves Ethanol Production by Metabolically Engineered Zymomonas mobilis. Appl. Microbiol. Biotechnol. 2014, 98, 6897–6905. 57. Yang, S.; Pan, C.; Hurst, G. B.; et al. Elucidation of Zymomonas mobilis Physiology and Stress Responses by Quantitative Proteomics and Transcriptomics. Front. Microbiol. 2014, 5. https://doi.org/10.3389/fmicb.2014.00246. 58. Shui, Z. X.; Qin, H.; Wu, B.; et al. Adaptive Laboratory Evolution of Ethanologenic Zymomonas mobilis Strain Tolerant to Furfural and Acetic Acid Inhibitors. Appl. Microbiol. Biotechnol. 2015, 99, 5739–5748. 59. Gu, H. Q.; Zhang, J.; Bao, J. High Tolerance and Physiological Mechanism of Zymomonas mobilis to Phenolic Inhibitors in Ethanol Fermentation of Corncob Residue. Biotechnol. Bioeng. 2015, 112, 1770–1782. 60. Seo, J. S.; Chong, H.; Park, H. S.; et al. The Genome Sequence of the Ethanologenic Bacterium Zymomonas mobilis ZM4. Natural Biotechnology 2005, 23, 63–68. 61. Yang, S.; Pappas, K. M.; Hauser, L. J.; et al. Improved Genome Annotation for Zymomonas mobilis. Natural Biotechnology 2009, 27, 893–894. 62. Wang, X.; He, Q. N.; Yang, Y. F.; et al. Advances and Prospects in Metabolic Engineering of Zymomonas mobilis. Metab. Eng. 2018. https://doi.org/10.1016/ j.ymben.2018.04.001.

Biodieselq

3.06

Wei Du, Department of Chemical Engineering, Tsinghua University, Beijing, China Rasool Kamal and Zongbao K Zhao, Dalian Institute of Chemical Physics, Chinese Academy of Sciences, Dalian, China © 2019 Elsevier B.V. All rights reserved. This chapter replaces S. Tamalampudi, H. Fukuda, 3.07 - Biodiesel, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 63-70.

3.06.1 3.06.2 3.06.2.1 3.06.2.1.1 3.06.2.1.2 3.06.2.1.3 3.06.2.2 3.06.2.2.1 3.06.2.2.2 3.06.2.3 3.06.2.3.1 3.06.2.3.2 3.06.2.3.3 3.06.2.3.4 3.06.2.4 3.06.2.5 3.06.3 3.06.3.1 3.06.3.2 3.06.3.3 3.06.3.3.1 3.06.3.3.2 3.06.3.3.3 3.06.3.3.4 3.06.3.3.5 3.06.3.4 References

3.06.1

Introduction Current Technologies for Biodiesel Production Homogeneous Catalyst-Mediated Process Homogeneous Alkali-Catalyzed Reaction Homogeneous Acid-Catalyzed Reaction Two-Step Catalytic Process With Acid and Alkali Heterogeneous Catalyst-Mediated Process Heterogeneous Solid Alkali-Catalyzed Reaction Heterogeneous Solid Acid-Catalyzed Reaction Lipase-Mediated Process Immobilized Lipase-Mediated Transformation Free Lipase-Mediated Alcoholysis Whole Cell-Mediated Alcoholysis Mechanism and Kinetics of Lipase-Catalyzed Alcoholysis Biodiesel Production in a Super Critical Fluid System Processing of By-product Glycerol Feedstocks for Biodiesel Production Edible Oils (First Generation) Non-edible Oils (Second Generation) Microbial Feedstock for Biodiesel Production (Third Generation) Biodiesel From Yeast Lipids Biodiesel From Algal Lipids Metabolic Engineering of Microbial Cell Factories Novel Feedstocks for Microbial Lipid Production Future Outlook of Microbial Feedstock Constraints and Perspectives of Biodiesel Development

66 66 67 67 68 68 68 68 69 69 69 69 70 70 71 71 71 71 71 72 72 73 73 74 74 75 75

Introduction

The depletion of fossil resources and increasing environmental concerns have stimulated exploitation of alternate sources for production of fuels and chemicals. Biodiesel has emerged as a renewable, biodegradable, and clean substitute for petroleumbased diesel fuels. Biodiesel use substantially reduces emission of SOx, CO, hydrocarbons, soot, and particulate matters1; however, it results in a slight increase in NOx emission, which can be positively influenced by delaying the injection time in engines. Life-cycle emission of CO2 from biodiesel is much lower than that from conventional diesel. Another advantage of biodiesel is that its production is free of seasonal and climatic changes, which benefits societies as it reduces dependence on oil import. Moreover, biodiesel use does not need modifications to currently used engines, as it is environment friendly and similar to existing fuels.1,2 The global biodiesel production has increased dramatically in the past decade. Many countries have now adopted B5, B10, or B20 standards, which use 5%, 10% or 20% biodiesel, respectively, supplemented into petroleum-based diesel. This chapter covers basic technologies regarding biodiesel production by transesterification using various catalysts and different feedstocks, with special emphasis on enzyme-catalyzed processes and microbial oils as an emerging feedstock.

3.06.2

Current Technologies for Biodiesel Production

Biodiesel can be produced by transesterification or esterification reactions, as described in Fig. 1. Transesterification reaction for biodiesel production refers to the formation of fatty acid alkyl-ester (mainly methyl ester or ethyl ester) between triacylglycerides (TAG) and methanol or ethanol. q

Change History: February 2019. Z. K. Zhao, R. Kamal and W. Du rewrote the chapter for this new edition.

66

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00151-8

Biodiesel

O

67

O

O

O R3

O

O

+ R'OH Catalyst

R1

R3

O

O

O R2

TAG

OH O

Alcohol

DAG

+ O

R2

R'OH

Catalyst

O

O

+ R

R' R1

O

R'OH

OH

R' R

HO

O

O

FA

O

OH O

+

R' R2

O

R3 R' : Me, Et

MAG

R'OH

Catalyst

Catalyst: Alkali, Acid, Enzyme O HO

O R' R

Figure 1

: Biodiesel (FAME, FAEE)

O

OH OH

+

R' R3

O

Glycerol

Biodiesel production from oil feedstocks.

Virtually TAG and long chain FA from any sources can be used as raw materials for biodiesel production. Although short chain alcohols, such as methanol or ethanol can be used, methanol is the most widely used because of its low price, strong polarity, and higher reactivity. The transesterification reaction between TAG and methanol can be catalyzed by acid, alkali, or enzymes, or can occur in a supercritical fluid system without catalyst. The transesterification reaction consists of a number of consecutive, reversible reactions (Fig. 1). The first step is the conversion of TAG to diacylglyceride (DAG), which is further converted into monoacylglyceride (MAG) and then glycerol, with concurrent formation of one equivalent ester per each step. The kinetics of transesterification reaction depends on the type of alcohol, molar ratio, type and amount of catalyst, and reaction temperature.1–6 The choice of catalyst is critical in designing an efficient process capable of maximizing the value of the materials while minimizing waste generation and energy use. A wide range of catalysts are used for the afore-mentioned transesterification, including homogeneous and heterogeneous chemical catalysts and biocatalysts. Lipases have been intensively studied, especially at laboratory scale, with the main focus on reducing overall cost of the system. In next section, different types of chemical homogeneous, solid heterogeneous, and biocatalysts are discussed.

3.06.2.1 3.06.2.1.1

Homogeneous Catalyst-Mediated Process Homogeneous Alkali-Catalyzed Reaction

Homogeneous alkali-mediated biodiesel production is a mature technology used widely in many countries (Fig. 2). Commonly used alkali catalysts are sodium hydroxide (NaOH), potassium hydroxide (KOH), and sodium methoxide. Among these, sodium methoxide is more effective than sodium hydroxide, presumably because it can quench a small amount of water to form NaOH and MeOH.2–6 However, NaOH and KOH are also widely used in industrial biodiesel production because of their low costs. The most promising reaction conditions were established as an alcohol-to-oil molar ratio of 6:1, 60
C, 1 h with either 0.5% sodium methoxide or 1% NaOH as preferred catalyst when using methanol as the alcohol.3–5 Generally, the reaction temperature should be slightly below the boiling point of the alcohol, and the alcohol should be used at about 2 equivalent of the theoretical amount to drive the reaction equilibrium. Furthermore, the free fatty acid (FAA) content should be less than 0.5% (acid value 3) such as n-hexane and petroleum ether. Hydrophobic organic solvents have been explored extensively as a reaction medium for immobilized lipase-catalyzed biodiesel production. In reality, methanol and glycerol have poor solubility in such relatively hydrophobic organic solvents. Therefore, the negative effects caused by methanol and glycerol cannot be completely eliminated, and lipase still exhibits poor operational stability. Some relatively hydrophilic organic solvents such as tert-butanol (C4H10O) were proposed as the reaction medium for lipase-mediated alcoholysis for biodiesel production. In such systems, both methanol and glycerol are soluble and the whole system is homogeneous; thus, the adverse effects of methanol and glycerol on lipase performance can be completely overcome. The combined use of Lipozyme TL IM and Novozym 435 in the tert-butanol system was reported with the highest biodiesel yield of 95% with no obvious loss in lipase activity even after being reused for up to 200 cycles. This technology was validated to be suitable for the conversion of various feedstocks into biodiesel, including low grade waste oils. However, for large-scale application, the recovery of the solvent and other related issue should be taken into considerations. Recently, metal–organic frameworks (MOFs) have gathered much interest owing to their excellent characteristics, including great structural tailorability, diversity, extremely high surface area, and porosity.37 These endow MOFs the potential as a great platform for enzyme immobilization. Lipase immobilization onto MOFs was resulted in a better enzyme stability and improved catalytic performance in biodiesel production. More efforts are needed to further broaden the application of MOFs for immobilization of lipases, especially for the production of biodiesel.

3.06.2.3.2

Free Lipase-Mediated Alcoholysis

Free lipase-mediated alcoholysis offers an alternative approach to enzyme-catalyzed biodiesel production; however, it is least reported. Compared with immobilized lipase, free lipase has the advantages of faster reaction rate and lower cost; therefore, free lipase-mediated methanolysis has been studied extensively for biodiesel production in recent years. Reports have revealed that a soluble lipase NS81006 catalyzes methanolysis of triglycerides to produce biodiesel in an oil/water biphasic system and over 90% of biodiesel yield can be achieved after 8 h of reaction.38–43 Although the free lipase-mediated reaction for biodiesel production offers an alternative approach for enzyme-mediated alcoholysis, the recovery of free lipase must be considered. Researchers from Tsinghua University, China, have developed a novel process wherein both free and immobilized lipases are combined for biodiesel production from various oil feedstocks. The operational life of lipases can be extended over 100-fold

70

Biodiesel

compared to that with traditional enzymatic approaches, leading to a significant reduction in the cost associated with lipases. This technology can be applied for poor-quality oils feedstocks, such as waste cooking oil, crude palm oil, algal oil, and presents a great potential for the production of second-/third-generation biodiesel. Additionally, the production process with this approach is environment friendly with zero emission and no acid/alkali usage in the process. This technology has been commercialized successfully with the production of 50,000 ton biodiesel/year in Hunan province, China (Fig. 3). It is the first commercialized facility in the world using enzymatic approaches for biodiesel production. Based on environmental and economic benefits of this facility, a bigger production plant (200,000 ton/year) is under construction in Guangdong, China. This process has also been demonstrated successfully in Brazil. The successful scalability of this technology is drawing attention worldwide, as many companies from the United States, Europe, Brazil, and Asian countries such as Malaysia and Indonesia are exploring the potential of licensing this technology for large-scale biodiesel production from low-quality oil feedstocks.

3.06.2.3.3

Whole Cell-Mediated Alcoholysis

Exploiting whole cell-mediated alcoholysis instead of the conventional immobilized lipases as catalysts for biodiesel production is a potential way to reduce the cost of biocatalyst, since it avoids the complex procedures of lipase fermentation, purification, and immobilization.44–48 Rhizopus oryzae, an intercellular lipase producing fungus species, has been extensively explored for biodiesel production. Immobilization of whole cells can be achieved spontaneously during culturing. To further stabilize the whole-cell biocatalysts, crosslinking treatment with glutaraldehyde (C5H8O2) was suggested, which allows reuse of immobilized whole cell catalysts for many batches. Further study revealed that the activity of such cross-linked immobilized cells was comparable to that of the commercial lipase Novozym 435. Acyl migration was also observed during the whole cell-mediated methanolysis for biodiesel production, which was later promoted with the increase in water content in the reaction mixture. Although whole cell biocatalysts overproducing intracellular lipase are expected to considerably reduce the lipase production cost and to offer a promising prospect for industrial biodiesel production, several challenges such as scaling up and process optimization need further assessment.

3.06.2.3.4

Mechanism and Kinetics of Lipase-Catalyzed Alcoholysis

During lipase-mediated alcoholysis for biodiesel production, three kinetic pathways are usually proposed to be operating (1) direct alcoholysis of glycerides (triglycerides, diglycerides, and monodiglycerides) into fatty acid alkyl esters, (2) two consecutive steps consisting of hydrolysis (conversion of glycerides into FFA) followed by esterification (conversion of FAA into esters), and (3) simultaneous alcoholysis and hydrolysis followed by esterification. Theoretically, Ping-Pong Bi-Bi mechanism is widely accepted for alcoholysis of triglycerides where each product is released between the additions of the substrates. In fact, active site of lipases includes acidic or basic functional groups such as hydroxyl groups and nitrogen atoms of amine groups that receive or donate protons during the reaction where conjugate acids of amines act as proton donors and carboxylate ions as proton acceptors. Eventually, acidic or basic catalytic reactions occur in the active site of the enzyme. Albeit steady-state kinetics analysis have been explored to describe the enzymatic conversion for a series of models with different complexities, the accuracy of those models remains questionable.49–53

Figure 3

Commercial lipase-mediated biodiesel production plant.

Biodiesel 3.06.2.4

71

Biodiesel Production in a Super Critical Fluid System

Super-critical fluid possesses different properties with its density similar to liquid, viscosity similar to gas, and thermal conductivity and diffusion coefficient between gas and fluid.54–58 Methanol is hydrophobic in supercritical conditions, and triglycerides dissolve well in supercritical methanol. As a result, biodiesel production in the supercritical fluid system has advantages of quick reaction rate and high converting yield. As the super-critical fluid method is sensitive to operating temperature and pressure variations, it is feasible to change the operating conditions of the reaction to adjust the physical properties of super-critical fluid, and thus affect the mass transfer, solubility, and reaction dynamics. However, limitations associated with this method are high operating temperature and pressure of approximately 350–400
C and 45–65 MPa, respectively, and its economic viability for industrial application needs further systematic evaluation.

3.06.2.5

Processing of By-product Glycerol

During transesterification of oil feedstocks for biodiesel production, approximately 10% w/w of glycerol is formed as a by-product. Utilization of this glycerol, especially in large-scale biodiesel production units is challenging. Reutilization of the crude glycerol as a carbon source for microbial oil production or integrated production of some other chemical compounds such as 1,3-propanediol (PDO) could make the overall biodiesel production process more promising and profitable in a circular economy. 1,3-PDO is a valuable organic compound and yields polytrimethylene terephthalate (PTT) by copolymerization with terephthalic acid (C8H6O4) or (methyl ester). PTT exhibits excellent properties in comparison with other polymers such as polyethylene terephthalate (PET). 1,3-PDO can be manufactured via chemical methods from petrochemicals or through fermentation from other renewable substrates. However, the chemical method for 1,3-PDO production has many limitations such as low selectivity, requirement of high temperature and pressure, and occurrence of highly explosive and toxic intermediates, such as ethylene oxide and acrolein, during the reaction. Therefore, the production of 1,3-PDO by fermentation has gathered much interest because of its advantages, such as mild reaction conditions, relatively lower investment, and the use of renewable substrates as starting materials, over the chemical method. Professor Liu and co-workers of Tsinghua University have already developed a novel biological process for conversion of crude glycerol into 1,3-PDO. Further, a 20 kton/year 1,3-PDO production facility is operating successfully in Jiangsu, China. The purity of 1,3-PDO reaches up to 99.92% and has been exploited for polymerization.59–63

3.06.3

Feedstocks for Biodiesel Production

Various feedstocks have been used for biodiesel production. The feedstock selection depends upon its availability, abundance, and economic aspects in the concerned region or country. In countries such as USA and Brazil, soybean oil is commonly utilized for biodiesel production, whereas in Finland, UK, Germany, and Italy, the rapeseed oil is more commonly used. Similarly, Asian countries like Malaysia and Indonesia have surplus palm and coconut oils in their coastal belts, which they utilize for biodiesel production. Also, Karanja and Jatropha in the Indian peninsula have been identified as potential feedstocks for biodiesel production. Generally, two types of vegetable oils are used for biodiesel production, namely the edible oils and non-edible oils. Utilization of the edible oils for biodiesel production raises great concerns over its effects on food supply. On the other hand, the non-edible oils available in huge quantities have several advantages such as their portability in liquid form, readiness, low aroma and sulfur (S) content, biodegradability, and more importantly, they have no negative influence on food supply. Furthermore, some emerging feedstocks such as microbial oils from oleaginous microorganisms, especially from yeast and microalgae, are regarded as promising feedstocks for biodiesel production, specifically because of the sustainability of the process. Biodiesel produced from edible oils is known as first-generation biofuel, that produced from non-edible oils is known as second-generation biofuel, biodiesel from microbial feedstocks is known as third-generation biofuel.

3.06.3.1

Edible Oils (First Generation)

In early years, the edible oil feedstock was the main source widely utilized for biodiesel production. The edible oils derived from feedstocks such as soybean, rapeseed, palm, mustard, olive, rice, wheat, coconut, and corn are categorized as first-generation feedstocks for biodiesel manufacturing. Though the first-generation feedstocks possess several advantages like high abundance and simple conversion procedure; however, the major drawback of using this feedstock is the threat of limiting the food supply, which can affect food market supply and prices. On the other hand, the high cost, land usage for cultivating crops, and the climate related problems also limit the utilization of first-generation feedstocks. These limitations induced the researchers to search for other feedstocks for biodiesel production.

3.06.3.2

Non-edible Oils (Second Generation)

Owing to concerns over utilization of edible oil feedstock for biodiesel production, researchers started using a variety of oils derived from non-edible crops. The fuels derived from non-edible feedstocks are known as second-generation biofuels. These oils are

72

Biodiesel

derived from plants such as Calophyllum inophyllum, Jatropha curcas, Mahua indica, Karanja, Neem, Rubber seed, Thevetia peruviana, and Nag champa. The major benefit of using non-edible oils is that it does not put burden on food supply and prices, which first-generation oils put. Additionally, the second-generation feedstocks can be grown on non-agricultural or marginal land; however, for plants like Jatropha, Cammelina, and oil palm, biodiesel yield drops when they are grown on non-agriculture or marginal land. Thus, farmers are forced to cultivate the second-generation feedstock crops on agricultural lands, which in turn affects the food production, supply, and market prices. To overcome the socioeconomic problems associated with the second-generation biofuels, researchers have focused on novel feedstocks such as microbial oil feedstocks for biodiesel production.

3.06.3.3

Microbial Feedstock for Biodiesel Production (Third Generation)

As discussed previously; biodiesel production from different edible feedstocks (first-generation biofuel) and from non-edible residues of food crops or non-edible plants feedstocks (second generation biofuel) are both in conflict with our food supply and put a huge stress on world food market, arable land use, water use, and may lead to a burden on the economy.64–67 Therefore, alternative feedstocks, i.e., microbial oils for biodiesel production have gained considerable attention recently. The fuels produced from microbial feedstocks are called third-generation biofuels, which are considered to be more practical and feasible alternative resource of energy that overcome the key problems associated with production of first- and second-generation biofuels. The advantages of the third-generation biofuels include higher energy density, good miscibility with other fossil fuels, and high compatibility with common combustion engines.65 Moreover, they are not dependent on weather conditions, do not use land, have greater annual productivity, short production cycle, and easy scalability.68 The oil produced by oleaginous microbes from different carbon sources is called microbial oil or single cell oil.69 Oleaginous microbes are those that can store more than 20% lipid of their dry cell weight.70 Mostly, lipid accumulation in these microbes occurs during “imbalanced” growth conditions where the excessive extra-cellular carbon is utilized for the synthesis and accumulation of lipids in nitrogen limiting conditions.71 The triglycerides composed of polyunsaturated fatty acids (PUFAs) is the key constituent of the microbial lipids,71,72 wherein the fatty acids have a high degree of unsaturation (e.g., C16 and C18) and chemical similarity to vegetable and plant oils (e.g., palm oils, rapeseed oils, and soybean oils, etc.).73 This shows the quality and importance of microbial oils as a feedstock for biodiesel production. The lipid producing oleaginous microbes are mainly divided into major three different groups, fungi (yeasts and molds), bacteria, and microalgae,72 among which, yeast and microalgae have been considered highly potential feedstocks for the production of biodiesel (Fig. 4).

3.06.3.3.1

Biodiesel From Yeast Lipids

Recently, oleaginous yeasts have been referred to as microbial cell factories and alternative oil producer to vegetable oil for a more sustainable biofuel industry.74 Among the oleaginous microbes, yeasts have several advantages over algae, molds, and bacteria, because of their higher growth rate, higher cell mass, adaptation to diverse substrates, and higher lipid production yields.75 The lipid storage in oleaginous yeasts satisfies energy demands as it comprises triacylglycerols (TAGs), which are mainly produced in the endoplasmic reticulum (ER) and lipid bodies in the cytosol. Oleaginous yeasts have potential to utilize different varieties of carbon substrates for de novo biosynthesis of lipids. Under nutrient limiting conditions (especially nitrogen limitation), citric acid accumulates in the mitochondrion, which is then exported to the cytoplasm for conversion to acetyl-CoA (a precursor of fatty acid synthesis) by ATP citrate lyase.70 These oil-accumulating yeasts have gathered much attraction because they produce lipid associated molecules and chemicals, including high-value nutritious oils, biodiesel, and adhesives.76–81 Production of biodiesel from yeasts oil is more efficient and sustainable, yet with similar costs compared to that from oilseeds.82 Many oleaginous yeasts have been known to accumulate lipids, e.g., Pichia guilliermondii,83 Trichosporon cutaneum,84 Aureobasidium pullulans85 Rhodotorula mucilaginosa,86

Figure 4

Biodiesel production by oleaginous algae and yeasts.

Biodiesel

73

Rhodosporidium toruloides,87 Yarrowia lipolytica,88 Trichosporon oleaginosus,89,90 and Lipomyces starkeyi.91 Amongst others, R. toruloides has received much attention recently.92 The strain R. toruloides Y4 can produce approximately 100 g L1 cell mass with more than 70% lipid content/dry cell mass under optimal culture conditions.93 Furthermore, this strain is highly resistant to inhibitory compounds, such as the deterrent compounds present in different biomass hydrolysates.94 Furthermore, oleaginous yeasts have the natural ability to utilize a variety of carbon sources for lipid accumulation. For example, the yeast Y. lipolytica utilizes glucose, crude glycerol, and waste oils as carbon substrate for lipid accumulation.94 The fatty acids produced by this yeast have extremely complex composition, which includes palmitoleic acid, palmitic acid, oleic acid, stearic acid, linoleic acid, and linolenic acid. Among which, C16 and C18 carbon containing fractions are dominant, with 50% of oleic acid content. These fatty acids produced by Y. lipolytica are considered to be highly suitable for biodiesel production.95 The lipid storage potential of some of the yeast strains utilizing different carbon substrates is summarized in Table 1.

3.06.3.3.2

Biodiesel From Algal Lipids

Microalgae, also called the sunlight-driven cell factories or cell bioreactors, have the ability to convert carbon dioxide to potential feeds, foods, fuels, and biologically active compounds of a high-value.96,97 Lipid synthesis mechanism of the holophytic microalgae is same as that of the green plants, i.e., they convert carbon dioxide (CO2) into autologous cell mass by fixation of inorganic carbon through photosynthesis, which is followed by its transformation into storage lipids through a series of cellular metabolic reactions. This approach has the leading advantage of high lipid storage.98 A rough estimation shows a 100-fold higher lipid yield per hectare with photosynthetic microorganism (microalgae or cyanobacteria) feedstock than with plant feedstock. In addition to the higher lipid yield, other features of photosynthetic microorganisms include a very short replication time and continuous harvesting. Moreover, the photosynthetic microorganisms do not require arable land and do not compete with food production. These factors indicate the potential of algal feedstock-derived biodiesel to replace large volumes of fossil fuel.97 Table 1 reviews the lipid content produced by some potential algae.

3.06.3.3.3

Metabolic Engineering of Microbial Cell Factories

The major challenge of these biofuel production processes are to advance the carbon utilization for desirable products. Moreover, the practical or the theoretical yields (gram of achieved product per gram of raw material used) based on the current metabolic pathways and the productivity (production rate/volume) of bioprocesses are mostly too low for industrial applications. The need for extremely high production volumes and low cost of biofuels suggests that a high yield and productivity are required, which can be achieved by controlling each step of biofuel generation. Most of these problems can be resolved by improving the existing metabolic pathways of microbes by redesigning the metabolic systems such that the yield and productivity increase but substrate and operational cost requirement are minimized. Particularly, engineering microbial cell factories can directly affect the biodiversity and use of land. For example, an increase of 10% in the yield of biofuel from lignocellulose biomass hydrolysate results in saving 10% land usage by reducing the requirement of raw material. Likewise, a 10% increase in microalgal productivity results in a 10% decrease in cultivation footprint.112 Thus, the efficient conversion of a feedstock into the desirable fuel products can have significant and direct impact on the economic feasibility of fuel production. Fortunately, the recent progress in technologies and understanding of genomics, metabolomics, and proteomics has enabled us to mine genomes of different strains for potential metabolic pathways, allowing design, construction, and characterization of pathways to perform a very specific function. The steady progress made by Table 1

Lipid contents of various oleaginous yeast and microalgae from different substrates

Microbe Yeast Lipomyces starkeyi Cryptococcus curvatus Yarrowia lipolytica Rhodotorula glutinis Rhodosporidium toruloides R. toruloides Cryptococcus albidus Cryptococcus curvatus Lipomyces starkeyi Algae Chlorella vulgaris Chlorella emersonii Chlorella minutissima Chlorella sorokiniana Nannochloropsis sp Isochrysis zhangjiangensis

Substrate

Lipid content %

Reference

Glucose þ Xylose Glucose DSCBH Glycerol Glycerol Glucose Glucose WOPH Glucose/Xylose

52 34 58 53 40 46 56 37 56

99 100 101 102 103 104 105 106 107

Sea Water Watanabe medium Watanabe medium Watanabe medium MWW Seawater

56 63 57 22 60 53

108 109 109 109 110 111

DSCBH, detoxified sugarcane bagasse hydrolysate; WOPH, waste office paper hydrolysate; MWW, municipal wastewater.

74

Biodiesel

researchers thus far toward engineering microbial cell factories to enhance fuels storage will reduce the reliance on petroleum and CO2 emissions. The accumulation of lipids within microbial cells could be efficiently enhanced by more advanced and recent techniques in synthetic biology and cellular metabolic engineering, which offers a platform of cell factories comprising widely engineered homologous and heterologous traits that can be exploited for the production of advanced biofuels.113 The strategies for the production of more advanced biofuels converge upon the four key metabolic routes: the 2-ketoacid pathway, fatty acid synthesis pathway, isoprenoid pathway, and reverse beta-oxidation pathway. In addition, more recently, the polyketide synthesis pathway has been attracting attention as a promising alternative production route for biofuel.114 The fuels produced by the engineered microbes have a great potential to improve processing steps; for example, by fermentation, energy coupling, and utilization of substrate. Metabolic engineering of microbes for biofuel production is one of the novel technologies, which could be a promising alternative for producing truly sustainable, cost-competitive, and technologically feasible biofuels. Therefore, the role of cellular metabolic engineering in production of fuels is critical for the expansion and sustainability of biofuel supply in the near future. There are several confines limitations associated with using microorganisms with natural biofuel production capabilities, including the slow growth, high nutrition requirements, pathogenicity, sensitivity to stress conditions, and low product yields. Hence, metabolic engineering has not only been exploited to advance the conventional microbial fermentation processes, but also to overcome all these problems. Moreover, it is also used to produce others chemicals that are presently used as fuels.115

3.06.3.3.4

Novel Feedstocks for Microbial Lipid Production

Large number of carbon substrates have been utilized for the production of lipids using oleaginous microbes, including low-cost industrially-derived sugars or sugar containing residual wastes, polysaccharides, N-acetylglucosamine, vegetable oils, various product or by-product hydrolysates, FAAs in pure form, fatty acids methyl esters (FAME), fatty by-products or wastes, crude glycerol, ethanol, n-alkanes, and organic acids.116–123 Among these, the most common carbon sources are the glucose and glucose-rich substrates. Considering the yield of microbial oils, some stoichiometric calculations of the overall carbon flux to triacylglycerol from glucose show that 1 mol glucose metabolized via glycolysis produces 2 mol of pyruvic acid. Therefore, approximately 15 mol of glucose are needed to synthesize 1 mol of triacylglycerol or 100 g glucose can provide maximum of 32 g lipid, assuming that glucose is not used for the synthesis of any other products, which of course is, as energy is consumed for synthesis and maintenance of cells and cellular machineries. Therefore, the practical conversion is approximately 20–22 g lipid/100 g glucose. Thus, only 1 ton of lipid can be obtained from 5 tons of glucose. Using only sugar as a carbon source for biofuel production has a little economic viability, because the cost of carbon source alone would be more than the income gained via the lipid product.71,72 Additionally, from the energy transfer point of view, the stoichiometry of oleyl ethyl ester (C20H38O2) in yeast Saccharomyces cerevisiae indicates that five glucose or six of xylose molecules with one oxygen molecule are required to synthesize this ester. During its biosynthesis, carried out by fatty acid synthase via eight cycles of condensation-reduction-dehydration-reduction, 10 molecules of each CO2 and water are produced as by-products. Generally, the input of cellular metabolic process is approximately 14.08 MJ of glucose or xylose, and the output is 12.56 MJ of oleyl ethyl ester (0.89 MJ per MJ input) plus 0.05 MJ captured in ATP.124 Thus, to make the microbial oil production more cost-effective, substantial efforts have been devoted for searching new carbon feedstocks instead of glucose, such as waste frying oils,125 lignocellulosic biomass hydrolysates,126 biodiesel industry-derived crude glycerol,127 and industrially produced organic wastes.128 From industrial point of view, the abundant and low-cost carbon substrates are the sugars derived from lignocellulosic materials. Microbial lipids produced from lignocellulose-derived sugars are cost-effective and have the potential for scaling up.129 Lignocellulosic biomass is a promising and abundant resource that can supplant the expensive carbon sources such as starch and refined sugars. Lignocellulosic biomass generally consists of about 75% carbohydrates (cellulose and hemicellulose) and approximately 25% lignin on a dry-mass basis.130 Thus, the sugars derived from lignocellulosic biomass are of great interest for sustainable microbial-based fuel production. Further, transesterification of triglycerides for biodiesel production results in a production of significant amount of glycerol as a byproduct,131 which accounts for more than 10% of the product.132 Thus, for every 100 kg of biodiesel produced, approximately 10 kg crude glycerol is obtained. Biological conversion of crude glycerol into microbial lipids is another good opportunity to valorize the biodiesel refinery. The microbial lipids produced from crude glycerol can be fed back into the biodiesel production plant, which will reduce the transportation costs. In addition to biodiesel production, use of such lipids to produce various other compounds will make the whole conversion process (i.e., crude glycerol to microbial lipid) more sustainable and economical.132–134 Overall, the process of biodiesel production from microbial feedstock is summarized in Fig. 4.

3.06.3.3.5

Future Outlook of Microbial Feedstock

Microorganisms such as yeast for lignocellulose-derived sugars and microalgae for direct CO2 conversion have the potential to produce the next-generation biofuels. Thus far, the commercial production of microbial lipids is in pipeline. However, neither a single microbial host nor any other lipid candidates has fully supplanted the need for development of biofuel production. Lignocellulosic feedstock is an abundant, renewable resource and can serve as a leading carbon substrate for microbial lipid production. At present, it is easier to transform microbes that naturally possess metabolic pathways for utilizing this resource, such as cellulose degradation, CH4 oxidation, or CO2 fixation, than engineering strains that carry pathways for more efficient utilization of these resources. However, as a long-term goal, achieving the theoretical limit of the fuel production via the designed pathways is of key importance, which can be achieved with the advances in genetic tools and mathematical modeling. Although the natural metabolic networks are significant, substantial room exists for improving the yield and productivity of microbial oil production. Design

Biodiesel

75

and discovery of novel enzymes and microbial metabolic pathways coupled with innovative, cost-effective substrates and highly efficient and economical processing technologies can significantly contribute to ultimate aim of making sustainable biofuel production a reality.

3.06.3.4

Constraints and Perspectives of Biodiesel Development

Owing to global warming and shortage of fossil fuels, biodiesel production is of immense significance. Although sustainable biodiesel production is hindered by many problems, like expensive raw material, low production yield, food supply and other environmental related issues (land and water usage issues), the recent advancements in research and technologies overcome these barriers in biodiesel production process. However, further efforts are necessary for identification of cheap and sustainable feedstocks, increasing the yield, achieving efficient transesterification (cheap and effective catalyst), and process optimization of biodiesel production. Therefore, focusing on microbial feedstock for biodiesel production has the potential to supplant the conventional fuels in a more sustainable and economical way. Microbial oil production is more sustainable in terms of environmental and economic aspects as it is free of all the limitations that the first- and second-generation biofuels face. Therefore, generating microbial cells factories using advanced genetic tools and process optimization for more robust and improved oil production from cost-effective and diverse carbon substrates is a need of the day. With the recent biotechnological advancement, biodiesel production will be reality in near future, which will help to build a healthier tomorrow.

References 1. Sakthivel, R.; Ramesh, K.; Purnachandran, R.; Shameer, P. M. A Review on the Properties, Performance and Emission Aspects of the Third Generation Biodiesels. Renew. Sustain. Energy Rev. 2018, 82, 2970–2992. 2. Baskar, G.; Aiswarya, R. Trends in Catalytic Production of Biodiesel from Various Feedstocks. Renew. Sustain. Energy Rev. 2016, 57, 496–504. 3. Taylor, P.; Talebian-kiakalaieh, A.; Aishah, N.; Amin, S. Single and Two-step Homogeneous Catalyzed Transesterification of Waste Cooking Oil : Optimization by Response Surface Methodology. Int. J. Green Energy 2015, 12 (9), 888–899. 4. Shankar, A. A.; Pentapati, P. R.; Prasad, R. K. Biodiesel Synthesis from Cottonseed Oil Using Homogeneous Alkali Catalyst and Using Heterogeneous Multi Walled Carbon Nanotubes: Characterization and Blending Studies. Egypt. J. Pet. 2017, 26 (1), 125–133. 5. Dwivedi, G.; Sharma, M. P. Application of Box–Behnken Design in Optimization of Biodiesel Yield from Pongamia Oil and its Stability Analysis. Fuel 2015, 145, 256–262. 6. Sánchez, M.; Bergamin, F.; Peña, E.; Martínez, M.; Aracil, J. A Comparative Study of the Production of Esters from Jatropha Oil Using Different Short-Chain Alcohols : Optimization and Characterization. Fuel 2015, 143, 183–188. 7. Avhad, M. R.; Marchetti, J. M. A Review on Recent Advancement in Catalytic Materials for Biodiesel Production. Renew. Sustain. Energy Rev. 2015, 50, 696–718. 8. Mathimani, T.; Uma, L.; Prabaharan, D. Homogeneous Acid Catalysed Transesterification of Marine Microalga Chlorella sp. BDUG 91771 Lipid - an Efficient Biodiesel Yield and Its Characterization. Renew. Energy 2015, 81, 523–533. 9. Verma, P.; Sharma, M. P.; Dwivedi, G. Impact of Alcohol on Biodiesel Production and Properties. Renew. Sustain. Energy Rev. 2016, 56, 319–333. 10. Bharathiraja, B.; Chakravarthy, M.; Kumar, R. R.; Yuvaraj, D.; Jayamuthunagai, J.; Kumar, R. P.; Palani, S. Biodiesel Production Using Chemical and Biological Methods – A Review of Process, Catalyst, Acyl Acceptor, Source and Process Variables. Renew. Sustain. Energy Rev. 2014, 38, 368–382. 11. Oglio, E. L. D.; de Sousa, P. T., Jr.; De Vasconcelos, L. G.; Adriano, C.; Barros, E. F.; Alberto, C. Monitoring of Microwave-Assisted Acid-Catalyzed Transesterification for Biodiesel Production Using FT-NIR Spectroscopy with Continuous-Flow Cell. Curr. Microw. Chem. 2016, 3, 68–77. 12. Vafaeezadeh, M.; Alinezhad, H. Brønsted Acidic Ionic Liquids : Green Catalysts for Essential Organic Reactions. J. Mol. Liq. 2016, 218, 95–105. 13. Sadaf, S.; Iqbal, J.; Ullah, I.; Bhatti, H. N.; Nouren, S.; Nisar, J.; Iqbal, M. Biodiesel Production from Waste Cooking Oil: An Efficient Technique to Convert Waste into Biodiesel. Sustain. Cities Soc. 2018, 41. 14. Kumar, R.; Dixit, A.; Kumar, S.; Gursahib, S. Production and Characterization of Ethyl Ester from Crude Jatropha curcas Oil Having High Free Satty Acid Content. J. Inst. Eng. Ser. A 2015, 96 (3), 229–235. 15. V Lee, H.; Juan, J. C.; Taufiq-yap, Y. H.; Kong, P. S. Advancement in Heterogeneous Base Catalyzed Technology : An Efficient Production of Biodiesel Fuels. J. Renew. Sustain. Energy 2015, 7 (3), 032701. 16. Devi, B. L. A. P.; Lakshmi, K. V.; Gangadhar, K. N.; Prasad, R. B. N. Novel Heterogeneous SO3 Na-Carbon Transesterification Catalyst for the Production of Biodiesel. Chemistry 2017, 2 (5), 1925–1931. 17. Jiang, B.; Wang, Y.; Zhang, L.; Sun, Y.; Yang, H.; Wang, B.; Yang, N. Biodiesel Production via Transesterification of Soybean Oil Catalyzed by Superhydrophobic Porous Poly (Ionic Liquid) Solid Base. Energy Fuels 2017, 31 (5), 5203–5214. 18. Cao, M. O.; Liu, Y.; Zhang, P.; Fan, M.; Jiang, P. Biodiesel Production from Soybean Oil Catalyzed by Magnetic Nanoparticle MgFe2O4@CaO. Fuel 2016, 164, 314–321. 19. Marinkovi, D. M.; Stankovi, M. V.; Veli, A. V.; Avramovi, J. M.; Miladinovi, M. R.; Stamenkovi, O. O.; Veljkovi, V. B.; Jovanovi, M. Calcium Oxide as a Promising Heterogeneous Catalyst for Biodiesel Production : Current State and Perspectives. Renew. Sustain. Energy Rev. 2016, 56, 1387–1408. 20. Paulo, J.; Duarte, A.; Di, L.; Souza, A. Alumina-supported Potassium Compounds as Heterogeneous Catalysts for Biodiesel Production : A Review. Renew. Sustain. Energy Rev. 2016, 59, 887–894. 21. Mardhiah, H. H.; Chyuan, H.; Masjuki, H. H.; Lim, S.; Lee, H. V. A Review on Latest Developments and Future Prospects of Heterogeneous Catalyst in Biodiesel Production from Non-edible Oils. Renew. Sustain. Energy Rev. 2017, 67, 1225–1236. 22. Sahu, G.; Gupta, N. K.; Kotha, A.; Saha, S.; Datta, S.; Chavan, P. A Review on Biodiesel Production through Heterogeneous Catalysis Route. ChemBioEng Rev. 2018, 5 (4), 1–23. 23. Zhang, D.; Zhang, X.; Li, Y.; Wang, S.; Wang, X.; Jiang, Z. Incorporation of Ce3þ Ions into Dodecatungstophosphoric Acid for the Production of Biodiesel from Waste Cooking Oil. Mater. Sci. Eng. C 2018, 92, 922–931. 24. Vieira, S. S.; Graça, I.; Fernandes, A.; Manuel, J.; Lopes, F. M.; Filipa, M.; Magriotis, Z. M. Influence of Calcination Temperature on Catalytic, Acid and Textural Properties of SO4 2/La2O3/HZSM-5 Type Catalysts for Biodiesel Production by Esterification. Microporous Mesoporous Mater. 2018, 270, 189–199. 25. Hassanpour, A.; Lai, X.; Lee, A. F.; Wilson, K. A Magnetically Separable SO4/Fe-Al-TiO2 Solid Acid Catalyst for Biodiesel Production from Waste Cooking Oil. Appl. Catal. B Environ. 2018, 234, 268–278. 26. Souza, R. D.; Vats, T.; Chattree, A.; Siril, P. F. Graphene Supported Magnetically Separable Solid Acid Catalyst for the Single Step Conversion of Waste Cooking Oil to Biodiesel. Renew. Energy 2018, 126, 1064–1073. 27. Mansir, N.; Taufiq-yap, Y. H.; Rashid, U.; Lokman, I. M. Investigation of Heterogeneous Solid Acid Catalyst Performance on Low Grade Feedstocks for Biodiesel Production : A Review. Energy Convers. Manag. 2017, 141, 171–182.

76

Biodiesel

28. Amini, Z.; Ilham, Z.; Chyuan, H.; Mazaheri, H.; Chen, W. State of the Art and Prospective of Lipase-Catalyzed Transesterification Reaction for Biodiesel Production. Energy Convers. Manag. 2016, 141, 339–353. 29. Pereira, E.; Andrade, E.; Fernandez-lafuente, R.; Maria, D.; Freire, G. Support Engineering : Relation between Development of New Supports for Immobilization of Lipases and Their Applications. Biotechnol. Res. Innov. 2017, 1–9. 30. Sankaran, R.; Show, P. L. Biodiesel Production Using Immobilized Lipase: Feasibility and Challenges. Biofuels Bioprod. Biorefin. 2016, 10 (6), 896–916. 31. Lee, O. K.; Lee, E. Y.; Kim, K. H. Nano-immobilized Biocatalysts for Biodiesel Production from Renewable and Sustainable Resources. Catalysts 2018, 8 (2), 68. 32. Shuai, W.; Das, R. K.; Naghdi, M.; Brar, S. K.; Verma, M. A Review on the Important Aspects of Lipase Immobilization on Nanomaterials. Biotechnol. Appl. Biochem. 2017, 64 (4), 496–508. 33. Arumugam, A.; Ponnusami, V. Biodiesel Production from Calophyllum inophyllum Oil Using Lipase Producing Rhizopus oryzae Cells Immobilized within Reticulated Foams. Renew. Energy 2014, 64, 276–282. 34. Luque, S.; Cervero, J. M. Novozym 435-catalyzed Synthesis of Fatty Acid Ethyl Esters from Soybean Oil for Biodiesel Production. Biomass Bioenergy 2014, 61, 131–137. 35. Online, V. A.; Masjuki, H. H.; Ching, J. J.; Chong, W. T. Enzymatic Transesterification for Biodiesel Production: a Comprehensive Review. RSC Adv. 2016, 6 (65), 60034– 60055. 36. Zhao, X.; Qi, F.; Yuan, C.; Du, W.; Liu, D. Lipase-catalyzed Process for Biodiesel Production : Enzyme Immobilization , Process Simulation and Optimization. Renew. Sustain. Energy Rev. 2015, 44, 182–197. 37. Hu, Y.; Dai, L.; Liu, D.; Du, W.; Wang, Y. Progress & Prospect of Metal-Organic Frameworks (MOFs) for Enzyme Immobilization (Enzyme/MOFs). Renew. Sustain. Energy Rev. 2018, 91, 793–801. 38. Wancura, H. C.; Rosset, D. V.; Tres, M. V.; Oliveira, J. V.; Mazutti, M. A.; Jahn, S. L. Production of Biodiesel Catalyzed by Lipase from Thermomyces lanuginosus in its Soluble Form. Can. J. Chem. Eng. 2018, 9999, 1–8. 39. Designs, E.; Fugmann, H. Development of Heat Transfer Surface Area Enhancements : A Test Facility for New Heat Exchanger Designs. Energies 2018, 11 (5), 1322. 40. Lv, L.; Dai, L.; Du, W.; Liu, D. Effect of Water on Lipase NS81006-Catalyzed Alcoholysis for Biodiesel Production. Process Biochem. 2017, 58, 239–244. 41. Li, Y.; Du, W.; Liu, D. Free Lipase-Catalyzed Biodiesel Production from Phospholipids-Containing Oils. Biomass Bioenergy 2014, 71, 162–169. 42. Ren, H.; Du, W.; Lv, L. Study on Free Lipase-Catalyzed Ethanolysis for Biodiesel Preparation in an Oil/Water Biphasic System. J. Am. Oil Chem. Soc. 2011, 88 (10), 1551–1555. 43. Zeng, L.; He, Y.; Jiao, L.; Li, K. Preparation of Biodiesel with Liquid Synergetic Lipases from Rapeseed Oil Deodorizer Distillate. Appl. Biochem. Biotechnol. 2017, 183 (3), 778–791. 44. He, Q.; Shi, H.; Gu, H.; Naka, G.; Ding, H.; Li, X.; Zhang, Y.; Hu, B.; Wang, F. Immobilization of Rhizopus oryzae LY6 onto Loofah Sponge as a Whole-Cell Biocatalyst for Biodiesel Production. BioResources 2016, 11 (1), 850–860. 45. Almyasheva, N. R.; Shuktueva, M. I.; Petrova, D. A.; Kopitsyn, D. S.; Kotelev, M. S.; Vinokurov, V. A.; Novikov, A. A. Mycoscience Biodiesel Fuel Production by Aspergillus niger Whole-Cell Biocatalyst in Optimized Medium. Mycoscience 2017, 3–8. 46. Liu, Y.; Huang, L.; Zheng, D.; Fu, Y.; Shan, M.; Xu, Z. Development of a Pichia pastoris Whole-Cell Biocatalyst with Overexpression of Mutant Lipase I PCLG47I from Penicillium cyclopium for Biodiesel Production. RSC Adv. 2018, 8, 26161–26168. 47. Amoah, J.; Ho, S.; Hama, S.; Yoshida, A.; Nakanishi, A. Conversion of Chlamydomonas sp . JSC4 Lipids to Biodiesel Using Fusarium Heterosporum Lipase-Expressing Aspergillus oryzae Whole-Cell as Biocatalyst. Algal. Res. 2017, 28, 16–23. 48. Chen, G.; Liu, J.; Qi, Y.; Yao, J.; Yan, B. Biodiesel Production Using Magnetic Whole-Cell Biocatalysts by Immobilization of Pseudomonas mendocina on Fe3O4-Chitosan Microspheres. Biochem. Eng. J. 2016, 113, 86–92. 49. Li, Y.; Du, W.; Dai, L.; Liu, D. Kinetic Study on Free Lipase NS81006-Catalyzed Biodiesel Production from Soybean Oil. J. Mol. Catal. B Enzym. 2015, 121, 22–27. 50. Firdaus, M.; Yusoff, M.; Brask, J.; Nielsen, P. M.; Guo, Z.; Fedosov, S. Kinetic Model of Biodiesel Production Catalyzed by Free Liquid Lipase from Thermomyces lanuginosus. J. Mol. Catal. B Enzym. 2016, 133, 55–64. 51. Canet, A.; Bonet-ragel, K.; Benaiges, M. D.; Valero, F. Lipase-catalysed Transesterification : Viewpoint of the Mechanism and Influence of Free Fatty Acids. Biomass Bioenergy 2016, 85, 94–99. 52. Bucio, S. L.; Solaesa, Á. G.; Sanz, M. T.; Melgosa, R.; Beltrán, S.; Sovová, H. Kinetic Study for the Ethanolysis of Fish Oil Catalyzed by Lipozyme ® 435 in Different Reaction Media. J. Oleo Sci. 2015, 441 (4), 431–441. 53. Liangliang, L.Ü.; Wei, D. U.; Dehua, L. I. U. Kinetics of Liquid Lipase NS81006-Catalyzed Alcoholysis of Oil for Biodiesel Production. Chin. J. Catal. 2012, 33 (11), 1857–1861. 54. Obie Farobie, Y. M. State of the Art of Biodiesel Production under Supercritical Conditions. Prog. Energy Combust. Sci. 2017, 63, 173–203. 55. Nan, Y.; Liu, J.; Lin, R.; Tavlarides, L. L. Production of Biodiesel from Microalgae Oil (Chlorella protothecoides) by Non-catalytic Transesterification in Supercritical Methanol and Ethanol: Process Optimization. J. Supercrit. Fluids 2015, 97, 174–182. 56. Francisco, A.; Pilar, J. P.; de los Ríos, A. P. High-yield Non-catalytic Supercritical Transesterification of Soybean Oil to Biodiesel Induced by Gradual Heating in a Batch Reactor. J. Supercrit. Fluids 2016, 111, 135–142. 57. García-martínez, N.; Andreo-martínez, P.; Quesada-medina, J.; Pérez, A.; Ríos, D. L.; Chica, A.; Beneito-ruiz, R.; Carratalá-abril, J. Optimization of Non-catalytic Transesterification of Tobacco (Nicotiana tabacum) Seed Oil Using Supercritical Methanol to Biodiesel Production. Energy Convers. Manag. 2017, 131, 99–108. 58. Olivares-carrillo, P.; Paneque, M.; Palacios-nereo, F. J.; Quesada-medina, J. High-yield Production of Biodiesel by Non-catalytic Supercritical Methanol Transesterification of Crude castor Oil (Ricinus communis). Energy 2016, 107, 165–171. 59. Liu, D.; Xu, J.; Zhao, X.; Du, W. Bioconversion of Glycerol into Lipids by Rhodosporidium toruloides in a Two-Stage Process and Characterization of Lipid Properties. Eng. Life Sci. 2016, 3, 303–313. 60. Yang, B.; Liang, S.; Liu, H.; Liu, J.; Cui, Z.; Wen, J. Metabolic Engineering of Escherichia coli for 1,3-propanediol Biosynthesis from Glycerol. Bioresour. Technol. 2018, 267, 599–607. 61. Zheng, Z.; Guo, N.; Hao, J.; Cheng, K.; Sun, Y.; Liu, D. Scale-up of Micro-aerobic 1,3-propanediol Production with Klebsiella pneumonia CGMCC 1.6366. Process Biochem. 2009, 44, 944–948. 62. Supaporn, P.; Yeom, S. H. Statistical Optimization of 1,3-propanediol (1,3-PD) Production from Crude Glycerol by Considering Four Objectives: 1,3-PD Concentration, Yield, Selectivity, and Productivity. Appl. Biochem. Biotechnol. 2018, 1–18. 63. Tabah, B.; Varvak, A.; Pulidindi, N.; Foran, E. Production of 1,3-propanediol from Glycerol via Fermentation by Saccharomyces cerevisiae. Green Chem. 2016, 18 (17), 4657–4666. 64. Brennan, L.; Owende, P. “Biofuels from Microalgae d A Review of Technologies for Production , Processing , and Extractions of Biofuels and Co-products. Renew. Sustain. Energy Rev. 2010, 14, 557–577. 65. Nigam, P. S.; Singh, A. Production of Liquid Biofuels from Renewable Resources. Prog. Energy Combust. Sci. 2011, 37 (1), 52–68. 66. Kim, J.; Nielsen, J. Bioproduction of Fuels: An Introduction. In Consequences of Microbial Interactions with Hydrocarbons, Oils, and Lipids: Production of Fuels and Chemicals. Handbook of Hydrocarbon and Lipid Microbiology; Lee, S., Ed., Springer: Cham, 2017; pp 3–25. 67. Demirbas, A. Comparison of Transesterification Methods for Production of Biodiesel from Vegetable Oils and Fats. Energy Convers. Manag. 2008, 49, 125–130. 68. Chaturvedi, S.; Bhattacharya, A.; Khare, S. K. Trends in Oil Production from Oleaginous Yeast Using Biomass : Biotechnological Potential and Constraints. Appl. Biochem. Microbiol. 2018, 54 (4), 361–369. 69. Subramaniam, R.; Dufreche, S. Microbial Lipids from Renewable Resources: Production and Characterization. J. Ind. Microbiol. Biotechnol. 2010, 37 (12), 1271–1287.

Biodiesel 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81.

82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113.

77

Papanikolaou, S.; Aggelis, G. Lipids of Oleaginous Yeasts . Part I : Biochemistry of Single Cell Oil Production. Eur. J. Lipid Sci. Technol. 2011, 113, 1031–1051. Ratledge, C.; Cohen, Z. Microbial and Algal Oils : Do They Have a Future for Biodiesel or as Commodity Oils ? Lipid Technol. 2008, 20 (7), 155–160. Ratledge, C.; Wynn, J. P. The Biochemistry and Molecular Biology of Lipid Accumulation in Oleaginous Microorganisms. Adv. Appl. Microbiol. 2002, 51, 1–51. Hassan, M.; Blanc, P. J.; Granger, L. M.; Pareilleux, A.; Goma, G. Lipid Production by an Unsaturated Fatty Acid Auxotroph of the Oleaginous Yeast Apiotrichum curvatum Grown in Single-Stage Continuous Culture. Appl. Microbiol. Biotechnol. 1993, 40 (4), 483–488. Patel, A.; Arora, N.; Sartaj, K.; Pruthi, V.; Pruthi, P. A. Sustainable Biodiesel Production from Oleaginous Yeasts Utilizing Hydrolysates of Various Non-edible Lignocellulosic Biomasses. Renew. Sustain. Energy Rev. 2016, 62, 836–855. Li, Q.; Du, W.; Liu, D. Perspectives of Microbial Oils for Biodiesel Production. Appl. Microbiol. Biotechnol. 2016, (80), 749–756. Beopoulos, A.; Nicaud, J. M.; Gaillardin, C. An Overview of Lipid Metabolism in Yeasts and its Impact on Biotechnological Processes. Appl. Microbiol. Biotechnol. 2011, 90 (4), 1193–1206. Blazeck, J.; Hill, A.; Liu, L.; Knight, R.; Miller, J.; Pan, A.; Otoupal, P.; Alper, H. S. Harnessing Yarrowia lipolytica Lipogenesis to Create a Platform for Lipid and Biofuel Production. Nat. Commun. 2014, 5, 1–10. Matthäus, F.; Ketelhot, M.; Gatter, M. Production of Lycopene in the Non-carotenoid-producing Yeast Yarrowia lipolytica. Appl. Environ. Microbiol. 2014, 80, 1660–1669. Ratledge, C. Fatty Acid Biosynthesis in Microorganisms Being Used for Single Cell Oil Production. Biochimie 2004, 86, 807–815. Rutter, C. D.; Zhang, S.; V Rao, C. Engineering Yarrowia lipolytica for Production of Medium-Chain Fatty Acids. Appl. Microbiol. Biotechnol. 2015, 99, 7359–7368. Xue, Z.; Sharpe, P. L.; Hong, S.; Yadav, N. S.; Xie, D.; Short, D. R.; Damude, H. G.; Rupert, R. A.; Seip, J. E.; Wang, J.; Pollak, D. W.; Bostick, M. W.; Bosak, M. D.; Macool, D. J.; Hollerbach, D. H.; Zhang, H.; Arcilla, D. M.; Bledsoe, S. A.; Croker, K.; Mccord, E. F.; Tyreus, B. D.; Jackson, E. N.; Zhu, Q. Production of Omega-3 Eicosapentaenoic Acid by Metabolic Engineering of Yarrowia lipolytica. Nat. Biotechnol. 2013, 31 (8), 734. Sitepu, I. R.; Garay, L. A.; Sestric, R.; Levin, D.; Block, D. E.; German, J. B.; Boundy-mills, K. L. Oleaginous Yeasts for Biodiesel: Current and Future Trends in Biology and Production. Biotechnol. Adv. 2014, 32 (7), 1336–1360. Wang, G.; Chi, Z.; Song, B.; Wang, Z.; Chi, Z. Bioresource Technology High Level Lipid Production by a Novel Inulinase-Producing Yeast Pichia guilliermondii. Bioresour. Technol. 2012, 124, 77–82. Shen, Q.; Lin, H.; Zhan, J.; Wang, Q.; Zhao, Y. Sweetpotato Vines Hydrolysate Induces Glycerol to Be an Effective Substrate for Lipid Production of Trichosporon fermentans. Bioresour. Technol. 2013, 136, 725–729. Wang, C.; Li, Y.; Xin, F.; Liu, Y.; Chi, Z. Evaluation of Single Cell Oil from Aureobasidium pullulans var. melanogenum P10 Isolated from Mangrove Ecosystems for Biodiesel Production. Process Biochem. 2014, 49 (5), 725–731. Yen, H.; Liao, Y.; Liu, Y. X. Cultivation of Oleaginous Rhodotorula mucilaginosa in Airlift Bioreactor by Using Seawater. J. Biosci. Bioeng. 2016, 121 (2), 209–212. Gao, Z.; Ma, Y.; Wang, Q.; Zhang, M.; Wang, J.; Liu, Y. Effect of Crude Glycerol Impurities on Lipid Preparation by Rhodosporidium toruloides Yeast 32489. Bioresour. Technol. 2016, 218, 373–379. Kuttiraja, M.; Dhouha, A.; Tyagi, R. D. Kinetics of Lipid Production at Lab Scale Fermenters by a New Isolate of Yarrowia lipolytica SKY7. Bioresour. Technol. 2016, 221, 234–240. Wang, J.; Gao, Q.; Zhang, H.; Bao, J. Inhibitor Degradation and Lipid Accumulation Potentials of Oleaginous Yeast Trichosporon cutaneum Using Lignocellulose Feedstock. Bioresour. Technol. 2016, 218, 892–901. Meo, A.; Priebe, X. L.; Weuster-botz, D. Lipid Production with Trichosporon oleaginosus in a Membrane Bioreactor Using Microalgae Hydrolysate. J. Biotechnol. 2016, 241, 1–10. Calvey, C. H.; Su, Y.; Willis, L. B.; Mcgee, M.; Jeffries, T. W. Nitrogen Limitation, Oxygen Limitation , and Lipid Accumulation in Lipomyces starkeyi. Bioresour. Technol. 2016, 200, 780–788. Huang, X.; Liu, J.; Lu, L.; Peng, K.; Yang, G.; Liu, J. Culture Strategies for Lipid Production Using Acetic Acid as Sole Carbon Source by Rhodosporidium toruloides. Bioresour. Technol. 2016, 206, 141–149. Li, Y.; Zhao, Z. K.; Bai, F. High-density Cultivation of Oleaginous Yeast Rhodosporidium toruloides Y4 in Fed-Batch Culture. Enzym. Microb. Technol. 2007, 41 (3), 312–317. Beopoulos, A.; Cescut, J.; Haddouche, R.; Uribelarrea, J. L.; Molina-Jouve, C.; Nicaud, J. M. Yarrowia lipolytica as a Model for Bio-Oil Production. Prog. Lipid Res. 2009, 48 (6), 375–387. Ma, Y.; Gao, Z.; Wang, Q.; Liu, Y. “Biodiesels from Microbial Oils: Opportunity and Challenges. Bioresour. Technol. 2018, 263, 631–641. Walker, T. L.; Purton, S.; Becker, D. K. Microalgae as Bioreactors. Plant Cell Rep. 2005, 24 (11), 629–641. Rittmann, B. E. Opportunities for Renewable Bioenergy Using Microorganisms. Biotechnol. Bioeng. 2008, 100 (2), 203–212. Chisti, Y. Biodiesel from Microalgae. Biotechnol. Adv. 2007, 25 (3), 294–306. Kong, X. L.; Liu, B.; Zhao, Z. B.; Feng, B. Microbial Production of Lipids by Co-fermentation of Glucose and Xylose with Lipomyces starkeyi 2#. Chinese J. Bioprocess Eng. 2007, 5 (2), 36–41. Zhang, J.; Fang, X.; Zhu, X.; Li, Y.; Xu, H.; Zhao, B. Microbial Lipid Production by the Oleaginous Yeast Cryptococcus curvatus O3 Grown in Fed-Batch Culture. Biomass Bioenergy 2011, 35 (5), 1906–1911. Asteraye, Y. T.; Wang, C.; Truong, C.; Ju, Y. Lipid Production from Yarrowia lipolytica Po1g Grown in Sugarcane Bagasse Hydrolysate. Bioresour. Technol. 2011, 102 (19), 9216–9222. Karamerou, C. W.; Eleni, E.; Theodoropoulos, C. Evaluating Feeding Strategies for Microbial Oil Production from Glycerol by Rhodotorula glutinis. Eng. Life Sci. 2016, 17 (3), 314–324. Tchakouteu, S. S.; Kalantzi, O.; Gardeli, C.; Koutinas, A. A.; Aggelis, G.; Papanikolaou, S. Lipid Production by Yeasts Growing on Biodiesel-Derived Crude Glycerol: Strain Selection and Impact of Substrate Concentration on the Fermentation Efficiency. J. Appl. Microbiol. 2015, 118 (4), 911–927. Zhang, C.; Shen, H.; Zhang, X.; Yu, X.; Wang, H.; Xiao, S.; Wang, J.; Zhao, Z. K. Combined Mutagenesis of Rhodosporidium toruloides for Improved Production of Carotenoids and Lipids. Biotechnol. Lett. 2016, 38 (10), 1733–1738. Fu, R.; Fei, Q.; Shang, L.; Brigham, C. J.; Nam, H. “Enhanced Microbial Lipid Production by Cryptococcus albidus in the High - Cell - Density Continuous Cultivation with Membrane Cell Recycling and Two - Stage Nutrient Limitation. J. Ind. Microbiol. Biotechnol. 2018, 1–7, 0123456789. Annamalai, N.; Sivakumar, N.; Oleskowicz-popiel, P. Enhanced Production of Microbial Lipids from Waste Office Paper by the Oleaginous Yeast Cryptococcus curvatus. Fuel 2018, 217, 420–426. Sharif Rahman, R. S.; Arbter, P.; Popovic, M.; Bajpaia, R. Microbial Lipid Production from Lignocellulosic Hydrolyzates : Effect of Carbohydrate Mixtures and Acid-Hydrolysis Byproducts on Cell Growth and Lipid Production by Lipomyces starkeyi. J. Chem. Technol. Biotechnol. 2016, 92 (8), 1980–1989. Liu, Z.; Wang, G. Effect of Iron on Growth and Lipid Accumulation in Chlorella vulgaris. Bioresour. Technol. 2008, 99, 4717–4722. Illman, A. M.; Scragg, A. H.; Shales, S. W. Increase in Chlorella Strains Calorific Values when Grown in Low Nitrogen Medium. Enzym. Microb. Technol. 2000, 27, 631–635. Jiang, L.; Luo, S.; Fan, X.; Yang, Z.; Guo, R. Biomass and Lipid Production of Marine Microalgae Using Municipal Wastewater and High Concentration of CO2. Appl. Energy 2011, 88 (10), 3336–3341. Feng, D.; Chen, Z.; Xue, S.; Zhang, W. Increased Lipid Production of the Marine Oleaginous Microalgae Isochrysis zhangjiangensis (Chrysophyta) by Nitrogen Supplement. Bioresour. Technol. 2011, 102 (12), 6710–6716. Liao, J. C.; Mi, L.; Pontrelli, S.; Luo, S. “Fuelling the Future : Microbial Engineering for the Production of Sustainable Biofuels. Nat. Rev. Microbiol. 2016, 14 (5), 288. Wackett, L. P. Engineering Microbes to Produce Biofuels. Curr. Opin. Biotechnol. 2011, 22 (3), 388–393.

78

Biodiesel

114. Cheon, S.; Kim, H. M.; Gustavsson, M.; Lee, S. Y. Recent Trends in Metabolic Engineering of Microorganisms for the Production of Advanced Biofuels. Curr. Opin. Chem. Biol. 2016, 35, 10–21. 115. Bailey, J. E. Toward a Science of Metabolic Engineering. Science (80–) 1991, 252, 1668–1675. 116. Papanikolaou, S.; Aggelis, G. Biotechnological Valorization of Biodiesel Derived Glycerol Waste through Production of Single Cell Oil and Citric Acid by Yarrowia lipolytica. Lipid Technol. 2009, 21 (4), 83–87. 117. Papanikolaou, S.; Aggelis, G. Yarrowia Lipolytica : A Model Microorganism Used for the Production of Tailor-Made Lipids. Eur. J. Lipid Sci. Technol. 2010, 112, 639–654. 118. Angerbauer, C.; Siebenhofer, M.; Mittelbach, M.; Guebitz, G. M. Conversion of Sewage Sludge into Lipids by Lipomyces starkeyi for Biodiesel Production. Bioresour. Technol. 2008, 99, 3051–3056. 119. Wu, S.; Hu, C.; Jin, G.; Zhao, X.; Zhao, Z. K. Phosphate-limitation Mediated Lipid Production by Rhodosporidium toruloides. Bioresour. Technol. 2010, 101 (15), 6124–6129. 120. Wu, S.; Hu, C.; Zhao, X.; Zhao, Z. K. Production of Lipid from N-Acetylglucosamine by Cryptococcus curvatus. Eur. J. Lipid Sci. Technol. 2010, 112 (7), 727–733. 121. Wu, H.; Li, Y.; Chen, L.; Zong, M. Production of Microbial Oil with High Oleic Acid Content by Trichosporon capitatum. Appl. Energy 2011, 88 (1), 138–142. 122. Papanikolaou, S.; Rontou, M.; Belka, A.; Athenaki, M.; Gardeli, C.; Mallouchos, A.; Kalantzi, O.; Koutinas, A. A.; Kookos, I. K.; Zeng, A. P.; Aggelis, G. Conversion of BiodieselDerived Glycerol into Biotechnological Products of Industrial Significance by Yeast and Fungal Strains. Eng. Life Sci. 2017, 17 (3), 262–281. 123. Liu, J.; Yuan, M.; Liu, J. N.; Huang, X. F. Bioconversion of Mixed Volatile Fatty Acids into Microbial Lipids by Cryptococcus curvatus ATCC 20509. Bioresour. Technol. 2017, 241, 645–651. 124. Caspeta, L.; Nielsen, J. Economic and Environmental Impacts of Microbial Biodiesel. Nat. Biotechnol. 2013, 31 (9), 789–793. 125. Eguchi, S.; Kagawa, S.; Okamoto, S. Environmental and Economic Performance of a Biodiesel Plant Using Waste Cooking Oil. J. Clean. Prod. 2015, 101, 245–250. 126. Kumar, D.; Singh, B.; Korstad, J. Utilization of Lignocellulosic Biomass by Oleaginous Yeast and Bacteria for Production of Biodiesel and Renewable Diesel. Renew. Sustain. Energy Rev. 2017, 73, 654–671. 127. Yang, X.; Jin, G.; Gong, Z.; Shen, H.; Bai, F.; Zhao, Z. K. Recycling Biodiesel-Derived Glycerol by the Oleaginous Yeast Rhodosporidium toruloides Y4 through the Two-Stage Lipid Production Process. Biochem. Eng. J. 2014, 91, 86–91. 128. Dourou, M.; Kancelista, A.; Juszczyk, P.; Sarris, D.; Bellou, S.; Triantaphyllidou, I.; Rywinska, A.; Papanikolaou, S.; Aggelis, G. Bioconversion of Olive Mill Wastewater into High-Added Value Products. J. Clean. Prod. 2016, 139, 957–969. 129. Huang, C.; Chen, X.; Xiong, L.; Chen, X.; Ma, L.; Chen, Y. Single Cell Oil Production from Low-Cost Substrates : The Possibility and Potential of its Industrialization. Biotechnol. Adv. 2013, 31 (2), 129–139. 130. Schreiner, M.; Lopes, G. Engineering Biological Systems toward a Sustainable Bioeconomy. J. Ind. Microbiol. Biotechnol. 2015, 42 (6), 813–838. 131. Uprety, B. K.; Chaiwong, W.; Ewelike, C.; Rakshit, S. K. Biodiesel Production Using Heterogeneous Catalysts Including Wood Ash and the Importance of Enhancing Byproduct Glycerol Purity. Energy Convers. Manag. 2016, 115, 191–199. 132. Uprety, B. K.; Venkatesagowda, B.; Rakshit, S. K. Current Prospects on Production of Microbial Lipid and Other Value-Added Products Using Crude Glycerol Obtained from Biodiesel Industries. Bioenergy Res. 2017, 10 (4), 1117–1137. 133. Uprety, B. K.; Reddy, J. V.; Dalli, S. S.; Rakshit, S. K. Utilization of Microbial Oil Obtained from Crude Glycerol for the Production of Polyol and its Subsequent Conversion to Polyurethane Foams. Bioresour. Technol. 2017, 235, 309–315. 134. Uprety, B. K.; Dalli, S. S.; Rakshit, S. K. Bioconversion of Crude Glycerol to Microbial Lipid Using a Robust Oleaginous Yeast Rhodosporidium toruloides ATCC 10788 Capable of Growing in the Presence of Impurities. Energy Convers. Manag. 2017, 135, 117–128.

3.07

Biofuels and Bioenergy: Acetone and Butanolq

Chuang Xue* and Youduo Wu*, School of Life Science and Biotechnology, Dalian University of Technology, Dalian, China Yang Gu* and Weihong Jiang, Key Laboratory of Synthetic Biology, CAS Center for Excellence in Molecular Plant Sciences, Institute of Plant Physiology and Ecology, Chinese Academy of Sciences, Shanghai, China Hongjun Dong*, Yanping Zhang, Chunhua Zhao, and Yin Li, Institute of Microbiology, Chinese Academy of Sciences, Beijing, China © 2019 Elsevier B.V. All rights reserved. This is an update of H. Dong, Y. Zhang, Y. Zhu, G. Luan, R. Wang, W. Tao, Y. Li, 3.08 - Biofuels and Bioenergy: Acetone and Butanol, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 71-85.

3.07.1 3.07.2 3.07.2.1 3.07.2.2 3.07.2.3 3.07.2.4 3.07.2.5 3.07.2.6 3.07.2.7 3.07.3 3.07.3.1 3.07.3.2 3.07.3.3 3.07.4 3.07.4.1 3.07.4.2 3.07.4.2.1 3.07.4.2.2 3.07.5 3.07.5.1 3.07.5.1.1 3.07.5.1.2 3.07.5.1.3 3.07.5.2 3.07.5.2.1 3.07.5.2.2 3.07.5.2.3 3.07.5.3 3.07.5.4 3.07.5.4.1 3.07.5.4.2 3.07.5.4.3 3.07.5.4.4 3.07.6 3.07.6.1 3.07.6.2 3.07.6.3 3.07.6.4 3.07.7 3.07.7.1 3.07.7.2 3.07.7.3 3.07.7.4 References

*

Introduction History of ABE Fermentation Discovery of Butanol-Producing Bacteria ABE Fermentation for Producing Butanol to Synthesize Rubber ABE Fermentation for Producing Acetone to Manufacture Cordite in World War I ABE Fermentation for Producing Butanol to Manufacture Quick-Drying Lacquer ABE Fermentation for Producing Acetone to Manufacture Cordite in World War II ABE Fermentation Decline Since the 1950s ABE Fermentation for Producing Butanol as a Promising Biofuel Nowadays Microorganisms for Acetone and Butanol Production Saccharolytic Clostridia Gas-Fermenting Clostridia Cellulolytic Clostridia Metabolic Pathways of Acetone and Butanol Formation Metabolic Pathways Genetic Regulation Transcriptional Regulation Post-transcriptional regulation Genetic Engineering Gene Transfer Protoplast Production and Transformation Electrotransformation Conjugation Gene Knockout Homologous Recombination Group II Intron CRISPR/Cas9-Based Editing System Downregulating the Expression of Target Protein and CRISPR Interference Cases of Genetic Engineering Engineering Acid Pathways Engineering Solvent Pathways Engineering Global Transcriptional Factors Reconstruction of Butanol Production Pathways in Other Hosts Systems Biology Genomics Transcriptomics Proteomics Metabolic Modeling Fermentation Substrates Butanol Toxicity and Tolerance Strain Improvement and In Situ Recovery Technology Development of Fermentation Technology

80 81 81 81 81 82 82 82 82 82 83 83 84 84 84 85 85 86 87 87 87 87 87 87 87 88 88 88 89 89 89 90 90 90 90 91 92 93 94 94 94 95 97 98

Authors with equal contributions. Change History: October 2018. C. Xue, Y. Gu, H. Dong, Y. Wu, Y. Li, W. Jiang, Y. Zhang and C. Zhao made updates to the Keywords and Sections to this article.

q

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00152-X

79

80

Biofuels and Bioenergy: Acetone and Butanol

Glossary ABE fermentation Production of acetone, butanol and ethanol via typically clostridial fermentation. Biofuel The fuel produced from biomass, including ethanol, butanol, biodiesel and so on. Butanol Four carbon alcohol (C4H9OH) with a distinct odor, colorless and flammable. Solventogenic bacteria The bacteria that can produce acetone and butanol under anaerobic culture conditions. Butanol toxicity The damage of butanol to cell associated with cellular structure, physiology and metabolism. CRISPR/Cas-based editing system The prokaryotic immune system that provides acquired resistance to exogenous genetic elements and is categorized as type I, type II and type III. asRNA technology The technology for downregulating the expression of target gene by expressing designed antisense RNA.

3.07.1

Introduction

Butanol can be biologically produced via the ABE fermentation, one of the oldest large-scale industrial fermentative processes in the world. In fact, microbial butanol production was first discovered by Louis Pasteur in 1861, while industrial ABE fermentation was initiated at the beginning of the 20th century for synthetic rubber production, and then particularly developed for explosive cordite production during World War I and II. After war, this fermentative process was also performed for nitrocellulose lacquer production in automobile industry. Until the 1950s, industrial ABE fermentation contributed to
66% and
10% of worldwide butanol and acetone supply, respectively. However, considering the increased substrates price and rapid development of petrochemical industry, the prominence of ABE fermentation dramatically declined with most global industrial facilities closed by the 1960s. Afterward, ABE fermentation has received renewed scientific and commercial interests as it could help address global unsustainability of fossil economy and environmental pollution. In the past century, substantial research studies have been conducted to establish feasible strain and process engineering strategies for developing ABE fermentation by Clostridium species,1 which mainly include 1) expanding low-cost feedstocks availability for economically sustainable butanol production; 2) uncovering mechanisms of transcription regulation or signal transduction in sugar uptake, biphasic metabolism, stress response and histidine kinase interactions; 3) reinforcing Clostridium strains with hyper sugar-utilizing, butanol-producing and -tolerating capacities by genetic manipulation; 4) establishing efficient genetic manipulation tools for clostridial genome editing such as CRISPR-Cas toolkit; 5) combining with omics analysis for new insight into clostridial physiology; 6) developing advanced single or integrated technologies for butanol recovery. As the genome sequences of two typical strains Clostridium acetobutylicum ATCC 824 and Clostridium beijerinckii NCIMB 8052 were released in 2001 and 2007, respectively, systematic analyses, including genetics, transcriptomics, proteomics and metabolic modeling have provided substantial knowledge on the clostridial physiology and regulatory mechanisms. The global demand of fossil fuels, such as petroleum, coal, and natural gases, is steadily growing with rapid economic development and population explosion. According to the BP Statistical Review of World Energy published in 2010, these fossil-based energy resources would be depleted within 45
60 years in terms of the current consumption policy, which led to globally increasing prices of petroleum-derived diesel and gasoline. Furthermore, the utilization of fossil fuels is not economically sustainable and environmentally friendly due to its limited availability, greenhouse gas (GHG) emission and environmental pollution of carbon monoxide (CO), sulfur dioxide (SO2), nitrogen oxides (NOx), particulate matter and volatile organic compounds. Therefore, considering the economic and environmental sustainability, renewable biofuels from low-cost feedstocks will be favorable and dominant in the near future. So nowadays, microbial ABE fermentation from various feedstocks is dedicated to produce biobutanol for complementing or replacing diesel and gasoline, beyond the petroleum-based butanol as before. Butanol (butyl alcohol or 1-butanol, C4H9OH, MW 74.12), is a colorless and flammable liquid with a distinct odor, which is widely used as not only an artificial flavorant additive in food and beverage industries but an important solvent, chemical intermediate and extractant in cosmetics and pharmaceutical industries. For instance, butanol is often used to synthesize butyl acrylate and methacrylate esters for latex surface coating, enamels and lacquers, and butyl glycol ether, butyl acetate and plasticizers. Butanol can be used as the diluent for brake fluid formulations, as well as the solvent for hormones, vitamins, and antibiotics production. More importantly, butanol is considered to be an environmentally friendly substitute for fossil-based energy superior to ethanol, owing to its higher energy content, lower water absorption, better blending ability with gasoline, and direct use in conventional combustion engines without modification. Butanol is predicted to replace about 20% global supply of gasoline and diesel in the near future. Actually, around 10–12 billion pounds of butanol is synthetized annually from petrochemical industries, which account for approximately $ 7–8.4 billion in current world market. The market for butanol is expanding at the rate of 3% per year, which is valued at around $10 billion by 2020. A number of companies have been extensively developing projects for butanol production from renewable feedstocks to meet the global demands of biobutanol in the future. Typically, GranBio (Brazilian Biotech Company, Alagoas) and Rhodia (Solvay International Chemical Group, Belgium) collaborated for butanol production from sugarcane bagasse in Brazil. The impending

Biofuels and Bioenergy: Acetone and Butanol

81

biorefinery is estimated to yield 100 kt solvents annually. ButamaxÔ Advanced Biofuels, a joint venture of British Petroleum (UK) and DuPont (USA), is dedicated to develop sustainable technologies to achieve economic production of butanol as advanced biofuel for the future transportation market. ButamaxÔ technology is mainly focused on efficient bioconversion of starch- and sugarbased feedstocks such as corn and sugarcane into butanol based on currently existing biorefining facilities. Additionally, some other global enterprises have made great endeavors for impelling commercialization of butanol as a promising biofuel, including GreenBiologics (UK), Cobalt Biofuels (USA), Tetravitae Bioscience Inc (USA), Gevo (USA), METabolic EXplorer (France), Butalco (Switzerland) and Cathay Industrial Biotech (China). However, there are still two major obstacles for developing ABE fermentation competitive with petroleum-derived butanol. One obstacle is the severe butanol toxicity to bacterial cells that limits the butanol concentration in the fermentation broth, which results in a high cost for butanol recovery from the dilute fermentation broth. Another obstacle is the low efficiency of utilizing lignocellulose-derived substrates for ABE fermentation. Because of the microbial inhibitors present in the hydrolysate of lignocellulosic feedstocks, the solventogenic bacteria exhibited a weak growth and poor metabolic shift toward solventogenesis. Additionally, the low capacity of co-utilization of hexose and pentose remains another significant issue for producing butanol from complexed substrates. To address these problems, developing systematically engineered Clostridium strains with hyper sugarutilizing, butanol-producing and -tolerating capacities would be necessary for economic ABE fermentation. This article will mainly introduce the microbial ABE fermentation, including related history, butanol-producing strains and biochemical engineering approaches. Substantial research studies on the metabolic networks and the regulation mechanism have provided feasible engineering approaches for strain reinforcement via genetic manipulation, wherein butanol production,inhibitors tolerance and sugar utilization were significantly improved. This article will also summarize the current advances in systems analyses of clostridial physiology, strain improvements, fermentation process developments and in situ butanol recovery techniques.

3.07.2

History of ABE Fermentation

3.07.2.1

Discovery of Butanol-Producing Bacteria

The term “ABE fermentation” represents Acetone, Butanol and Ethanol fermentation, which is also called “ABE fermentation” because of Ethanol produced in this fermentative process. The first record about butanol-producing bacteria should date back to 1861, when the famous French scientist Louis Pasteur reported a microbial culture of “Vibrion butyrique”, which might be a mixed bacterial culture comprising Clostridium butyricum or C. acetobutylicum.2 Albert Fitz probably first isolated the pure bacterium named Bacillus butylicus and did some seminal studies about this strain. Over the past century, many scientists discovered some other new butanol-producing bacteria with different designations due to a lack of consistent taxonomic standards.

3.07.2.2

ABE Fermentation for Producing Butanol to Synthesize Rubber

At the beginning of the 20th century, as the shortage of natural rubber resulted in high price, Strange and Graham Ltd. in England then planned to synthesize rubber artificially using isoprene, a product of butanol derivatives isoamyl alcohol or butadiene. The firm employed Perkins and Weizmann of Manchester University and subsequently employed Fernbach and Schoen of the Institute Pasteur to work on this project since 1910. In 1911, Fernbach isolated a culture which was able to ferment potatoes rather than maize starch to produce butanol. After resigning from Strange and Graham Ltd. in 1912, Chaim Weizmann attempted to isolate a butanol-producing bacterium strain BY. This strain was later named C. acetobutylicum with better fermentation performance than Fernbach’s original culture. In the middle of 1913, Strange and Graham Ltd. initiated butanol fermentation using Fernbach’s strain at Rainham. The plant produced butanol and by-product acetone from potatoes.2

3.07.2.3

ABE Fermentation for Producing Acetone to Manufacture Cordite in World War I

With the advent of World War I in 1914, the British needed cordite (smokeless powder) in large quantities to produce munitions. For preparation of cordite, acetone was used as the colloidal solvent for nitrocellulose. Due to the tremendous shortage of chemically produced acetone in England at the outbreak of the war, the unnoticed acetone that was produced during microbial butanol production caught the attention of the British War Office. Strange and Graham Ltd. was then asked to provide acetone to the government. Therefore, the first industrial-scale ABE fermentation was rapidly developed as a response to a high demand for acetone production. In this period, Weizmann developed another ABE fermentation process (Weizmann process) and applied for a patent in March 1915. Because of the apparent advantages of the Weizmann process, acetone production from maize by this process was performed in several plants in England, including Strange and Graham Ltd. This Weizmann process utilized C. acetobutylicum to produce 3000 tons of acetone and 6000 tons of butanol within the next 2 years. Due to the grain shortage caused by the German blockade, the fermentation process was transferred to Canada and the USA, in 1916 and 1917, respectively.2

82

Biofuels and Bioenergy: Acetone and Butanol

3.07.2.4

ABE Fermentation for Producing Butanol to Manufacture Quick-Drying Lacquer

With armistice of World War I in 1918, there was no longer need to produce acetone. However, how to treat the large amount of butanol accumulated as a by-product in acetone production remained an economic problem. Butanol again became a commercially important solvent for the production of quick-drying lacquers, for the rapidly developed automobile industry by Henry Ford. Therefore, many plants performing ABE fermentation were erected to produce butanol for the automobile industry, and intensive research studies were also carried out rapidly. Additionally, molasses became cheaply available in large quantities and became the main carbon substrate for solvent production. At its peak capacity in 1927, the Commercial Solvent Corporation Plant in Peoria, IL (USA) possessed 96 fermentors in operation with a total volume of 18,168 m3, representing the largest fermentation facility for ABE fermentation at that time.2

3.07.2.5

ABE Fermentation for Producing Acetone to Manufacture Cordite in World War II

With the outbreak of World War II, the ABE fermentation was once again performed to produce acetone for the explosive cordite manufacture. Because of the high demand for acetone during the war, continuous distillation was applied in ABE fermentation. The industrial production capacity for acetone and butanol continued to increase until its peak in 1945, when two-thirds of butanol and one-tenth of acetone in the United States were derived from ABE fermentation.2

3.07.2.6

ABE Fermentation Decline Since the 1950s

From the 1950s, the ABE fermentation declined rapidly after the war, and almost ceased in the 1960s due to multiple reasons. First, the formerly cheap substrate molasses began to be used as feeding additives for pigs and cows breeding, and showed decreasing quality by improved sugar processing technology. Second, ABE fermentation was apt to be infected by phage at that time. Third, there was no significant progress on industrial strain improvement and the fermentation process to reduce the production costs, although some efforts had been done, such as screening for new strains and developing a continuous fermentation process. Fourth, ABE fermentation was strongly limited by the butanol toxicity (10
13 g L1), which is still currently a bottleneck for improving the ABE fermentation. Most importantly, when acetone and butanol were also produced cheaply from petroleum oil by the developing petrochemical industry, the ABE fermentation lost its economic competition completely and was discontinued in the industrialized Western countries during the 1960s.2

3.07.2.7

ABE Fermentation for Producing Butanol as a Promising Biofuel Nowadays

Since the 1970s, the global oil crisis renewed academic and industrial interests developing butanol as a sustainable and promising alternative to petroleum-based energy resources. Considering the excellent physicochemical property of butanol as an advanced energy fuel, such as higher energy content, lower water absorption and better blending ability, the butanol biorefining has become a hot point for research area since 2005. Recently, many companies like DuPont, BP and GEVO have initiated projects to produce butanol from renewable biomass. In 2006, BP and DuPont announced their partnership to develop and commercialize biobutanol, and presented plans to produce 30,000 tons of butanol annually in a modified ethanol facility of British Sugar in the UK. In 2008, DuPont and BP proposed that the use of butanol could increase the butanol blending in gasoline fuel. In 2009, this partnership was cleared to take over the US company–butanol LLC. BP and Mazda announced the use of ethanol and butanol blending in the Petit Le Mans Race, USA. Currently, the industrial ABE fermentation has been intensively re-established globally. In general, the economic feasibility of current ABE fermentation is governed by three major factors: feedstocks cost, butanol yield and product recovery, which resulted in numerous important developments of strain reinforcements, fermentation processing improvements, genetic manipulation toolkits, omics analysis and butanol recovery techniques. Thus, biobutanol is predicted to be a feasible substitute biofuel for gasoline in future.

3.07.3

Microorganisms for Acetone and Butanol Production

Starting from Pasteur, various methods for isolating solventogenic clostridia have been well documented. These bacteria were isolated from potatoes, the roots of nitrogen-fixing legumes, other root crops, cereal crops, fruits such as gooseberries, and agricultural soil. Basically, the Clostridium has four common characteristics: gram-positive, spore-forming, obligately anaerobic and no capability of dissimilatory sulfate reduction. According to different substrates used, the industrial Clostridium species can be roughly divided into saccharolytic, gas-fermenting and cellulolytic clostridia. During the studies of phylogenetic relationships between solventogenic clostridial strains, various methods have been used. For example, based on the inter-group similarities, genomic DNA-DNA reassociation and DNA fingerprints, as well as phenotype comparison, currently reported solventogenic clostridia can be roughly classified into four species, C. acetobutylicum, C. beijerinckii, Clostridium saccharobutylicum, and Clostridium saccharoperbutylacetonicum.3 A recent comparative genomic analysis of a total 44 saccharolytic Clostridium strains based on their genomes, including C. acetobutylicum, Clostridium aurantibutyricum, C. beijerinckii, Clostridium diolis, Clostridium felsineum, Clostridium pasteurianum, Clostridium puniceum, Clostridium roseum,

Biofuels and Bioenergy: Acetone and Butanol

83

C. saccharobutylicum, and C. saccharoperbutylacetonicum, have been performed,4 in which these anaerobes were only classified into two large groups, including many subgroups. Therefore, there has not yet been a uniform classification in terms of solventogenic clostridia.

3.07.3.1

Saccharolytic Clostridia

At the initial stage, corn mash (starch) was used as the only substrate for the clostridia-based ABE production (Weizmann process). During this period, the dominant Clostridium species employed in industrial ABE fermentation is C. acetobutylicum, and therefore, this organism is sometimes called the “Weizmann Organism”. C. acetobutylicum shows the high starch-decomposing ability, enabling a rapid conversion of starchy materials into acetone and butanol, as well as a small amount of ethanol. Because of the shortage of corn in the 1930s, molasses were gradually adopted as a substitutive substrate. Thus, many efforts were subsequently made to isolate new strains capable of fermenting molasses to ABE solvents. After the expiration of the Weizmann’s patent in 1936, a lot of patents about molasses-fermenting clostridia were documented. Many different names for these clostridia were produced in a completely haphazard manner and in generally lacked any systematic basis, which led to the proliferation of names for solventproducing clostridia. It is noteworthy that the widely studied typical strain, C. acetobutylicum ATCC 824 is not the first isolate of C. acetobutylicum. This strain, which was isolated from Connecticut garden soil in 1924, was never used in industry. Weyer and Rettger chose the strain, rather than Weizmann’s strain, as the typical strain of the species, for they characterized it. The DSM 792 strain that was used by German research groups and housed with the DSM Culture Collection, is the equivalent type culture strain. Another well-studied strain NCIMB 8052 that is housed with the National Collection of Industrial and Marine Bacteria (NCIMB) in Scotland was formerly recognized as C. acetobutylicum, but now it has been affirmed to be the strain of C. beijerinckii. To date, the conventional and most extensively studied route for biological utilization of lignocellulosic biomass follows the pretreatment-hydrolysis-fermentation (PHF) process. The hydrolysates from different lignocellulosic biomass include multiple monosaccharides, in which glucose and xylose are normally two major components. Many species of saccharolytic clostridia have exceptional substrate diversity, capable of utilizing a variety of carbohydrates, including hexose and pentose sugars, enabling them to be developed for producing chemicals and biofuels by fermenting lignocellulosic hydrolysates. However, similar to many other industrial microorganisms, some typical Clostridium species (e.g., C. acetobutylicum) possess the carbon catabolite repression (CCR) effect, which is mediated by catabolite control protein A (CcpA) and (or) other regulators.5 The CCR effect led to the undesired behavior of Clostridium species in fermenting lignocellulosic hydrolysates, namely the inhibited utilization of non-preferable sugars in the presence of preferred substrates (e.g., glucose). Of course, there are also exceptions, such as C. beijerinckii. Although C. beijerinckii also contains the ccpA gene according its genome data, this Clostridium species did not show very severe CCR effect as that in C. acetobutylicum.6 To eliminate the adverse effect from CCR in solventogenic clostridia, many efforts have been made, including inactivation of key factors associated with CCR,7,8 mutation of catabolite responsive element (CRE),9 intensification of metabolic pathways of nonpreferable sugars.6,10,11 All of these strategies have taken effect, enabling some major Clostridium species, e.g., C. acetobutylicum and C. beijerinckii, to efficiently use different monosaccharides simultaneously. In addition, the tolerance of clostridia to the inhibitors in lignocellulosic hydrolysates is also a key issue in fermenting these materials. It is known that many toxic inhibitors, such as phenolic compounds and furan derivatives will generate during the pretreatment process of lignocellulosic biomass.12 Detoxification via physical and chemical approaches can address this problem, however, will inevitably increase the preparation cost of lignocellulosic hydrolysates, which is unlikely to be the optimal choice. Similar problems have also existed within biotechnological routes, i.e., microbial and enzymatic detoxification. Therefore, a promising and industrially feasible strategy is to develop Clostridium species with high tolerance to inhibitors. However, this appears to be a challenging task for solventogenic clostridia, and there has been almost no progress in this aspect. Actually, many members of the above saccharolytic clostridia can also use other cheap industrial organic matters rather than sugar-contained feedstocks. For example, C. pasteurianum is capable of digesting glycerol, the main by-product from biodiesel industry, as the carbon source to generate butanol.13 C. saccharoperbutylacetonicum can use acetic, butyric and lactic acids to produce butanol.14 These findings demonstrate the substrate diversity of saccharolytic clostridia.

3.07.3.2

Gas-Fermenting Clostridia

The widespread use of fossil fuels is no longer sustainable, and moreover, their use has caused serious environmental pollution and global warming. Thus, environmentally friendly routes to produce chemicals and fuels are urgently need. One promising strategy is to directly capture free carbon sources (e.g., CO2 and CO) before they are incorporated into lignocellulosic biomass, and then converted these C1 gases into various target products, in which microorganisms can play important roles. The microorganisms that are capable of using CO2 or CO as the only carbon sources belong to the autotroph, in which acetogens including gas-fermenting clostridia are a major member. These anaerobic bacteria are able to capture and metabolize CO or CO2, allowing them to grow on different industrial off-gases (e.g., steel manufacture and oil refining, coal and natural gas). Additionally, gas-fermenting clostridia can also consume “synthesis gas” (syngas), which is an important industrial feed gas consisting of CO, CO2 and H2, and can also be produced from the gasification of lignocellulosic biomass, such as forestry and agricultural wastes.

84

Biofuels and Bioenergy: Acetone and Butanol

To date, the identified and well-studied gas-fermenting Clostridium species mainly include Clostridium ljungdahlii, Clostridium autoethanogenums and Clostridium carboxidivorans. All these acetogens employ the Wood-Ljungdahl pathway to fix and metabolize CO and CO2 molecule into the important intermediate,15 acetyl-CoA, which is further catalyzed to synthesize several major chemicals, such as acetic acid, ethanol, butyric acid, 1-butanol and so on. Of note, among the known six natural biological carbon fixation pathways, the Wood-Ljungdahl pathway is the one with the least reaction steps and energy consumption.

3.07.3.3

Cellulolytic Clostridia

An alternative route for biological conversion of lignocellulosic materials to chemicals is the consolidated bioprocessing (CBP) based on cellulolytic microorganisms, which are able to hydrolyze cellulose/hemicellulose and then metabolize the derived sugars. CBP combines the three steps of the conventional bioconversion of lignocellulosic materials (production of saccharolytic enzymes, hydrolysis of lignocellulose/hemicellulose and fermentation) within one reactor, and thus, theoretically has advantage over the conversional route in processing costs. Many Clostridium species, such as Clostridium thermocellum, Clostridium cellulolyticum and C. cellulovorans, have been reported to be able to secrete cellulase and hemicellulase, and thus, can directly grow on lignocellulosic biomass. Using these Clostridium species as the chassis, some have been genetically modified to better meet the requirements of CBP-based biorefinery or produce ethanol and other higher alcohols that are not native products in these anaerobes. However, large-scale industrial application of CBP still remains unexplored due to the low saccharification rate of lignocellulosic biomass based on cellulolytic clostridia, compared to the traditional enzymatic hydrolysis. This rate-limiting step severely drags down the conversion efficiency of the whole CBP system.

3.07.4

Metabolic Pathways of Acetone and Butanol Formation

3.07.4.1

Metabolic Pathways

A typical feature of clostridial solvent production is the biphasic fermentation with yields of butanol, acetone and ethanol in the ratio of 6:3:1. As shown in Fig. 1, the formation of acetone, butanol and ethanol shares the same metabolic pathway from glucose to acetyl-CoA, which is then separated into different pathways. Generally, the enzymes in the pathway of glucose to acetyl-CoA then to butyryl-CoA keep active during the whole fermentation process. Five enzymes, pyruvate-ferredoxin oxidoreductase, thiolase, 3hydroxybutyryl-CoA dehydrogenase, crotonase and butyryl-CoA dehydrogenase (encoded by pflB, thl, hbd, crt and bcd, respectively) are involved in the reactions of converting pyruvate to butyryl-CoA. 2ATP, 2NADH glucose

pyruvate Ferredoxin

pflB (pyruvate-ferredoxin oxidoreductase) CAC0980 ak ATP (acetate kinase) acetate CAC1743

NADH, NADPH NAD, NADP Ferredoxin-H2

pta (phosphotrans acetylase) acetyl-P

CAC1742

NADH acetyl-CoA

NADH acetaldehyde

adhE (acetaldehyde dehydrogenase) CAP0162, CAP0035

thl (thiolase) CAC2873 acetone

acetoacetate adc ctfAB (acetoacetate (CoAdecarboxylase) transferase) CAP0165 CAP0163, CAP0164

ethanol bdhAB, edh (ethanol dehydrogenase) CAC3298,CAC3299, CAP0059 CAP0162, CAP0035

acetoacetyl-CoA NADH hbd (3-hydroxybutyryl-CoA dehydrogenase) CAC2708 3-hydroxybutyrylCoA crt (crotonase) CAC2712 crotonyl-CoA NADH bcd (butyryl-CoA dehydrogenase) CAC2711

ATP butyrate buk (butyrate kinase) CAC3075

Figure 1

NADH butyryl-P

butyryl-CoA ptb (phosphotrans butyrylase) CAC3076

Metabolic pathways in Clostridium acetobutylicum.16

NADH butyraldehyde

adhE (butyraldehyde dehydrogenase) CAP0162, CAP0035

butanol bdhAB (butanol dehydrogenase) CAC3298,CAC3299 CAP0162, CAP0035

Biofuels and Bioenergy: Acetone and Butanol

85

Different pathways function in different phases (as shown in Fig. 1). In the initial phase (acidogenesis) of clostridia-based ABE fermentation, besides the abovementioned five enzymes, the other four enzymes, namely acetate kinase, phosphotransacetylase, phosphotransbutyrylase and butyrate kinase (encoded by ack, pta, ptb and buk, respectively) are also active and contribute to acidogenesis. In this phase, the acetyl-CoA derived from glycolysis is mainly metabolized to produce acetate and butyrate, accompanied by the formation of a large amount of ATPs, which can support cell growth. In addition, to keep the redox balance, NADH that forms during glycolysis is oxidated to support the synthesis of butyrate, resulting in that more butanol is produced than acetate during the solventogenic phase when most of the acids formed during the acidogenic phase begin to be re-assimilated to produce solvents (butanol derives from butyrate, and ethanol derives from acetate). Although the conversion of acetyl-CoA to butyryl-CoA is quite thermodynamically favorable (DrGmq ¼ 14.2 kcal mol1), the condensation reaction of acetyl-CoA to acetoacetyl-CoA is thermodynamically unfavorable (DrGmq ¼ 5.3 kcal mol1), therefore indicating that the condensation reaction is a rate-limiting step in the whole pathway from acetyl-CoA to butyryl-CoA. It has been found that a high concentration of acetyl-CoA facilitated the condensation reaction as well as solvent production.2,16 As the fermentation proceeds, the solventogenesis phase of clostridia begins once the pH value of the culture decreases to a crucial point for acid reassimilation. At these stages, the expression of genes responsible for the formation of ABE solvents are induced, while those for acidogenesis are repressed. The known major genes located in solventogenic pathways include adhE, bdhAB, ctfAB and adc, coding acetaldehyde/butyraldehyde dehydrogenase, butanol dehydrogenase, CoA-transferase and acetoacetate decarboxylase, respectively. During the solventogenic phase, the previously produced acetate and butyrate are reassimilated and converted into acetyl-CoA and butyryl-CoA, respectively, by CoA-transferase. Additionally, acetone production is closely coupled with ethanol and butanol synthesipathways by employing CoA-transferase. In fact, CoA-transferase, rather than acetoacetate decarboxylase, has been considered to be the rate-limiting enzyme for acetone formation. Meanwhile, the conversion of acids to acetyl-CoA and butyryl-CoA contributes to relieve the damage of low pH to cells. During the solventogenic phase, a large amount of NADH is consumed. Actually, the high solvents titer is, to a certain extent, up to the availability of NADH. According to the basic research on the metabolic pathways of solvents production, we can draw an image about the metabolic flux distribution alteration in different phases. Generally, during the acidogenic phase, the expression of ptb and buk genes is upregulated while ack and pta genes are just expressed inductively during the transition phase from acidogenesis to solventogenesis. The important intermediate, acetyl-CoA, is metabolized through some branched pathways, forming acetate and butyrate. To keep the redox balance during the acidogenic phase, most of the acetyl-CoA has to be metabolized to butyrate, which requires a high accumulation of acetyl-CoA to make the thermodynamically unfavorable condensation reaction run smoothly. In contrast, the metabolic fluxes are opposite during the solventogenic phase, and acetate and butyrate are reversed to form acetyl-CoA and butyryl-CoA, and then further produce solvents. It is believed that the elevated expression of ctfAB, adc, adhE and bdhAB contributes to the above phase-dependent pathway shift. Based on this, the fluxes in the acidogenic phase are dramatically distinct from those in the solventogenic phase. Besides, while the solvents pathways are activated, sporulation occurs. However, a detailed understanding of the relationship between sporulation and solventogenesis is still lack, requiring a further investigation.

3.07.4.2

Genetic Regulation

During the whole fermentation process of clostridia, there are two important regulatory points. One is the transition from acidogenesis to solventogenesis and the other is the beginning of sporulation. The low pH is believed to be the important signal for solventogenesis. Additionally, the solventogenesis and sporulation are closely coupled. At the end of the exponential cell growth phase, due to the accumulation of acids, cells encounter the unfavorable low pH conditions. So cells begin to uptake acetate and butyrate that were produced early, and convert them to acetyl-CoA and butyryl-CoA. Consequently, the acetyl-CoA and butyryl-CoA are converted to solvents. However, the solvents, especially butanol, are toxic to cells. Butanol can impact the membrane fluidity, and moreover, lead to the malfunction of cell-binding proteins, Therefore, the sporulation process initiates, which helps clostridial cells survive in the adverse circumstance. However, along with spore formation, clostridial cells gradually lose the solvent-producing ability.

3.07.4.2.1

Transcriptional Regulation

So far, various transcriptional factors are known to be involved in the regulation of sporulation and solvents formation of solventogenic clostridia. 3.07.4.2.1.1 Spo0A The firstly identified global regulator in solventogenic clostridia is Spo0A, which controls sporulation and many other important physiological and metabolic processes.17 The inactivation or overexpression of spo0A could cause significant transcriptional alternation of nearly all essential genes located in solvent synthetic pathways, including the sol operon (adhE1-ctfA-ctfB), adc, bdhA, bdhB, ptb and buk. The enhanced expression of the sol operon after spo0A overexpression can accelerate the solvent formation in Clostridium acetobutyicum. Among the abovementioned genes located in solvent-forming pathways, the sol operon, adc, bdhA and bdhB contain a Spo0A-binding site (TGNCGAA, N¼A, C, G or T) in their upstream noncoding regions, indicating that they are under the direct regulation of Spo0A. In the contrary, no such a binding site was found around ptb-buk and pta-ack. Additionally, Spo0A is known to be phosphorylated by some histidine kinases, and then the phosphorylated Spo0A can exert regulatory function.18

86

Biofuels and Bioenergy: Acetone and Butanol

3.07.4.2.1.2 CcpA CcpA (Catabolite Control Protein A) is an essential global regulator in Gram-positive bacteria and responsible for two fundamental regulatory mechanisms, i.e., CCR (Carbon Catabolite Repression) and CCA (Carbon Catabolite Activation). These two mechanisms are crucial for bacteria, enabling them to adapt quickly in response to different environmental changes. Generally, CcpA can recognize and bind to catabolite responsive elements (cres) in the promoter and ORF (open reading frame) region of the target genes, achieving the transcriptional activation and repression, respectively. To date, different CcpA-binding cre architectures have been identified in solventogenic clostridia, including the typical 14- to 16-nucleotide palindromic sequence and some non-canonical cres with flexible architectures. CcpA has been found to be a crucial regulator for controlling solvent formation in clostridia. Inactivation of the ccpA gene in C. acetobutylicum significantly impaired cell growth and solvent synthesis.7 Among the essential genes located in solvent forming pathways, the sol operon and adhE2 gene were found to be significantly downregulated in transcription after ccpA was inactivated.19 Interestingly, three CcpA-binding sites were discovered in the promoter region of the sol operon,20 indicating that this important gene cluster in C. acetobutylicum is under a complex regulation of CcpA. 3.07.4.2.1.3 AbrB AbrB is capable of controlling quite a few essential cellular processes in Gram-positive bacteria through its pleiotropic function, including sporulation, antibiotic biosynthesis, biofilm formation, cyanotoxin production and so on. In C. acetobutylicum, three AbrB homologues have been identified and named AbrB0310, AbrB1941, and AbrB3647,21 in which AbrB0310 and AbrB3647 were found to be able to affect solvent production by regulating some essential genes located in solvent synthetic pathways. 3.07.4.2.1.4 Rex The Redox-sensing regulator Rex responds to intracellular NADH/NADþ ratio and then modulates alcohol production and oxidative stress tolerance in C. acetobutylicum.22 Many genes/operons involved in some essential metabolic pathways (including solvent synthesis) were found to be regulated by Rex in C. acetobutylicum.22,23 When intracellular NADH/NADþ ratio increases, Rex dissociates from its binding sites in the promoter region of the target genes, and relieve the repression on these genes. Therefore, in C. acetobutylicum Rex plays an important role in regulating the production and ratio of ABE (acetone-butanol-ethanol) solvents. Besides, in gas-fermenting C. ljungdahlii, the Rex-binding site have been identified in the promoter regions of the Wood-Ljungdahl pathway genes.22 This indicates that Rex may regulate metabolism of CO or CO2 in C. ljungdahlii. 3.07.4.2.1.5 CsrA CsrA (Carbon Storage Regulator A) is a post-transcriptional regulator that exists widely in eubacteria. CsrA controls the expression of target genes by binding to 50 -untranslated region of mRNA.24,25 For solventogenic clostridia, CsrA was firstly identified in C. acetobutylicum. Inactivation of csrA caused significantly impaired cell growth and solvent production, indicating its importance in this anaerobe. Moreover, it has been known that CsrA normally executes a pleiotropic regulation, capable of affecting a large number of physiological and metabolic processes. In C. acetobutylicum, the RNA-seq analysis revealed that CsrA may affect the expression of the genes associated with growth, mobility, ion transportation, signal transduction, protein translation and folding, solvent synthesis and central metabolism26; however, the detailed regulatory mechanisms remain to be explored. 3.07.4.2.1.6 Sigma factors Sigma factors play a crucial role in controlling mRNA transcription. For sporulation, sigma factors sE, sF, sG, and sK are activated in turn after separation with the help of sH and spo0A, and each of these sigma factors regulates the transcription of a specific set of genes and operons. At the same time, Spo0A～P can control the sporulation by interacting with sH. It seems that Spo0A participates both in sporulation and solventogenesis process. Spo0A controls the sporulation by binding the regulatory sites of related genes involved in the sporulation. Spo0A is activated when it is phosphorylated by a two-component signal transduction system. It seems that Spo0A, to some extent, plays some kind of balancing role on solvents formation by regulating related genes expression between sporulation and solventogenesis. In addition, similar to the functionality of abrB for sporulation initiation in Bacillus subtilis, a homologous gene abrB310 may function as a regulator at the transition between acidogenesis and solventogenesis in clostridial bacteria. It is believed that the NADH availability in the cells is quite important for the high production of solvents. Meanwhile, some studies suggested that the level of butyryl-P is also very vital for high butanol production. Besides, there is evidence showing that butanol starts to form when butyryl-P concentration reaches its peak. Some studies also suggested that, for butanolproducing strains, the maximum butyryl-P concentration should be higher than 60–70 pmol g1 (DCW), while in the nonbutanol-producing strains lacking the megaplasmid pSOL1, the concentration would never surpass 50 pmol g1 (DCW).

3.07.4.2.2

Post-transcriptional regulation

Post-transcriptional regulation is an essential biological process and well known not only for controlling diverse essential traits but also for the value in genetic modification in organisms. However, in this aspect, it remains poorly unexplored in anaerobic clostridia.

Biofuels and Bioenergy: Acetone and Butanol

87

Small noncoding RNAs (sncRNAs) are major regulatory molecules responsible for the post-transcriptional regulation in organisms. Moreover, sncRNAs are regarded as promising targets for metabolic engineering. However, the number of the identified sncRNAs in bacteria is much smaller than plant and mammalian cells. In addition, these reported bacterial sncRNAs have been mainly limited to only a few typical species, such as Escherichia coli and Bacillus subtilis. For solventogenic clostridia, some sncRNAs are found to be associated with stress responses,27 but lack detailed functional analysis. SolB is the newly reported sncRNA in C. acetobutylicum with a revealed regulatory mechanism,28 showing a specific regulation on expression of the sol locus. Of note, a comprehensive analysis of sRNAs in 21 clostridial genomes based on computational prediction gave a large number of potential sncRNAs in the genus Clostridium.29 This study provided a valuable resource for a continued functional analysis of regulatory sncRNAs in industrial clostridia.

3.07.5

Genetic Engineering

3.07.5.1

Gene Transfer

3.07.5.1.1

Protoplast Production and Transformation

An efficient protoplast transformation protocol for C. acetobutylicum NI-4081 has been developed with a high transformation frequency (up to 106 transformants/g-DNA). The protocol presented in Handbook on clostridia5 is described as follows: Bacteria are grown to mid-exponential phase (approximately 108 cells/mL) in T69 medium containing 0.6 M sucrose. To remove the cell wall, lysozyme and penicillin G are added. The protoplasts are centrifuged at 3000 g and washed twice in protoplast washbuffer (T69 buffer supplemented with 0.6 M sucrose, 0.5% (w/v) bovine serum albumin (BSA) and 1 mM CaCl2). Protoplasts are then resuspended in protoplast buffer (T69 supplemented with 0.5 M xylose, 0.5% (w/v) BSA, 25 mM MgCl2 and 25 mM CaCl2). Plasmid DNA (50–800 ng), polyethylene glycol (PEG) 4000 (35% (w/v)), and 109 protoplasts are mixed and incubated at room temperature for 2 min. The mixture is then diluted 10-fold in T69 medium supplemented with 0.5 M xylose, 0.5% (w/v) BSA, 1 mM CaCl2 and 4 mg mL1 choline. The protoplasts are then centrifuged, washed, and resuspended in the same medium. Dilutions are then added to T69 top agar (T69 supplemented with 0.25 M xylose, 0.5% (w/v) BSA, 1 mM CaCl2, 4 mg mL1 choline and 0.8% (w/v) agar) and then poured onto T69 agar (0.25 mM xylose and 2.5% (w/v) agar) and incubated at 34  C for 20 h. To select for transformants with erythromycin resistance markers, a further 3 mL of top agar (T69 supplemented with 0.25 M xylose, 1 mg mL1 erythromycin and 0.8% (w/v) agar) is overlayed onto the plates and they are incubated for 4–6 days at 37  C.

3.07.5.1.2

Electrotransformation

The protocols for electrotransformation of C. acetobutylicum ATCC 824 and C. beijerinckii NCIMB 8052 have been documented in the literature.5 Because C. acetobutylicum ATCC 824 contains a restriction system, Cac824I, which greatly reduces the electroporation efficiency of foreign DNA lacking the appropriate methylation modification. To circumvent this, plasmid DNA should be methylated in vivo in E. coli containing the Bacillus subtilis phage phi 3T I methyltransferase. Generally speaking, cells are cultured to a proper growth phase, and then are harvest by centrifugation at 4  C. The precooled (4  C) ETB buffer (electroporation buffer, containing 270 mM sucrose, 1 mM MgCl2 and 7 mM NaH2PO4 (pH7.4)) is used to wash and resuspend the cells. Plasmid DNA is then added to the suspension and held on ice for several minutes before electroporation with a 2.0 kV pulse and 25 mF capacitance. The electroporation solution is then mixed with the proper volume of medium and incubated at 37  C for 3–5 h before plating on the agar plate containing the appropriate selective agents. Usually, the PEG Buffer was successfully used to help the electrotransformation of C. beijerinckii NCIMB 8052. In addition, different electrotransformation buffers and electroporation parameters are applied in different protocols for electrotransformation of C. acetobutylicum ATCC 824, C. beijerinckii NCIMB 8052 and other strains such as DSM 1731, DSM 792 and NRRL B592.

3.07.5.1.3

Conjugation

Conjugation is an important process for genetic exchange between bacteria. The process needs mating of donor cell and recipient cell, and involves a cis-acting nick site (oriT) and the trans-acting functions given by a transfer protein. Now the conjugation between two cells has been developed as a useful method for artificial gene transfer. Using Escherichia faecalis or B. subtilis as donor strain, conjugative plasmids pAMb1 and pIP501 have been conjugated into C. acetobutylicum and C. beijerinckii.5 For convenient usage of the conjugation, the transfer of DNA from E. coli into the species of clostridia is very attractive. But in real operation, the successful conjugation was only reported in C. beijerinckii. The commonly used E. coli donor strains are HB101 (carried by an IncP-type helper plasmid, R702) and SM10 (Tra functions are integrated into the chromosome). The mobilizable vectors (suicide or replicative) carry the transfer origin (oriT) of an IncP plasmid, such as RP4 or RK2. The fact that no report about conjugation of C. acetobutylicum and E. coli exists may be due to the strong barrier of the Restriction and Modification system in C. acetobutylicum.

3.07.5.2 3.07.5.2.1

Gene Knockout Homologous Recombination

Although the technology of gene knockout based on the homologous recombination has been a classical way to study the function of a target gene or to genetically engineer a strain, it showed very low efficiency in butanol-producing clostridial strains. Most of successful cases of gene knockout are based on suicide vectors. The first case of gene knockout was reported in C. beijerinckii NCIMB

88

Biofuels and Bioenergy: Acetone and Butanol

8052, in which gutD and spo0A were disrupted via conjugation with the suicide vectors containing corresponding homologous fragments. Green and Bennett also made several gene knockout in C. acetobutylicum via electrotransformation with the suicide vectors containing corresponding homologous fragments, i.e., buk, pta, aad, solR. Considering the low efficiency of the gene knockout strategy based on suicide vector, researchers developed a new strategy of gene knockout based on a replicable vector. This method can make the homologous recombination happen easily due to the efficient replication of vectors containing homologous fragments in cells. pLHKO is such a kind of vector applied in C. acetobutylicum, which contains the gene of chloramphenicol resistance, and the replication origin of pIM13. First, the replication vector is ligated with an erythromycin resistance gene flanked by two homologous fragments of the target gene. Second, after introduction of the replication vector into C. acetobutylicum, the positive transformants containing the target vector are cultured on agar plates without antibiotics for several times and lose the vector via photocopy method. Third, the colonies are transferred to the plates containing erythromycin. Finally, the positive colonies are confirmed by PCR or Southern blotting, and spo0A in C. acetobutylicum was thus disrupted. Maybe due to the laborious work of this method, only one successful application case has been reported by far.

3.07.5.2.2

Group II Intron

In 2007, the gene knockout system for Clostridium based on the retrohoming of group II intron was developed. The system was derived from Lactococcus lactis Ll. LtrB group II intron. The principle of retrohoming is that the group II intron self-excised from the RNA can form a ribonucleoprotein (RNP) complex (intron-encoded protein and excised intron RNA complex) with a conserved intron-encoded protein (IEP), which can recognize the DNA target site and make an invasion event. Further research revealed that the target-site recognition rules were determined by two sites located on the intron RNA, i.e., EBS1d and EBS2. Therefore, the modification of the two sites according to a statistically mathematical model will lead to artificial insertion of intron into target sites. A commercial gene knock-out system, Targetron (Sigma–Aldrich, St. Louis, MO), has been developed based on these rules. Minton’s group (University of Nottingham, UK)6 and Jiang’s Group (Shanghai Institute of Plant Physiology and Ecology, CAS, China) have also reported this kind of gene knockout system in Clostridium. This system is now the most efficient gene knock-out system for Clostridium. However, it should be noticed that, only special target sites can be inserted using this system, and the deletion type of gene knockout is not easily conducted.

3.07.5.2.3

CRISPR/Cas9-Based Editing System

CRISPRs (clustered regularly interspaced short palindromic repeats) is a prokaryotic immune system providing acquired resistance to exogenous genetic elements, e.g., phages and plasmids.30 CRISPR–Cas systems are categorized as type I, type II and type III based on different signature genes, namely Cas3, Cas9 and Cas10, respectively. In type II systems, the signature DAN endonuclease Cas9 works together with the crRNA (CRISPR RNA) and tracrRNA (Trans-activating CRISPR RNA) duplex to mediate interference. A breakthrough for this system was the creation of a single guide RNA (sgRNA) chimera that synthesize the functions of crRNA and tracrRNA. The CRISPR-Cas9-mediated genome editing is programmable via the design of sgRNAs that guide Cas9 proteins to recognize and cleave the target DNA sequences through WatsonCrick base pairing. The sequence of sgRNA can be easily changed to match a sequence of interest and then retarget the Cas9 nuclease to a target (genes or other functional elements) of choice. For clostridia, the CRISPR-Cas9 editing system was first established in two industrially relevant Clostridium species, namely C. cellulolyticum and C. beijerinckii.31,32 This genetic tool overcomes the deficiencies of reported tools in clostridia, including relying on labor intensive screening to identify the desired mutants; polar effects to downstream genes using the group II intron-mediated insertional inactivation; instability of the mutants generated by single crossover integration of plasmids into chromosome. The Streptococcus pyogenes CRISPR-Cas9 system was introduced into these two Clostridum species, enabling efficient gene deletion and insertion of foreign genes into the genome. In the manipulation of C. cellulolyticum, to circumvent the lethality of wild-type Cas9-induced double-strand breaks, Cas9 nickase was adopted, enabling a single-nick-triggered homologous recombination strategy. Subsequently, the CRISPR-Cas9 editing systems were developed in more industrial Clostridium species, including C. acetobutylicum, C. saccharoperbutylacetonicum, and gas-fermenting C. ljungdahlii and Clostridium autoethanogenum, showing a good universality in the genus Clostridium.

3.07.5.3

Downregulating the Expression of Target Protein and CRISPR Interference

Antisense RNA (asRNA) is an efficient means for regulating gene expression. Generally, there are two kinds of mechanisms for inhibiting target RNA translation by binding of asRNA, (1) hindering ribosome binding site (RBS) interactions with ribosomes, and/or (2) stimulation of the degradation of the target RNA by ribonucleases via altering its structure. Desai and Papoutsakis first studied the efficiency of antisense RNA strategies in C. acetobutylicum by applying artificially designed asRNA. However, design of an effective asRNA is not a simple work. In addition, it is not enough to only consider the association rate between asRNA and target mRNA. It was shown that applying Mfold for structural feature-based asRNA design was an effective way. Mfold is a computational algorithm for predicting secondary structure, based on thermodynamics and structural information derived from the studies of known RNA molecules. In this method, free nucleotides (nucleotides in asRNA molecules that are not involved in intramolecular binding) and components (structural features that contain regions of intramolecular binding called duplex RNA) are recognized as the factors for an asRNA. Constructs of antisense RNA can be generated and visualized using DNA representation software such as Gene

Biofuels and Bioenergy: Acetone and Butanol

89

Construction Kit2 (Textco, Inc., West Lebanon, NH). In C. acetobutylicum ATCC 824, the downregulated expression of adc, buk, ctfAB and ptb have been reported as the successful cases of asRNA method.33 In addition, CRISPR-Cas9 system can also be modified to achieve the downregulation of gene expression. Briefly, a catalytically dead Cas9 (usually denoted as dCas9) protein that lacks endonuclease activity is guided by sgRNA to the target genomic loci, achieving the transcriptional regulation of target genes. This technology is termed CRISPR interference (CRISPRi).34 Based on the data from the numerous reports, gene transcription repression mediated by CRISPR-dCas9 is efficient, tunable and reversible. Importantly, simultaneous repression of multiple genes can be easily realized via the CRISPRi approach. To date, the CRISPR-dCas9 system derived from Streptococcus pyogenes has been successfully used in clostridia for transcriptional repression, including C. acetobutylicum, C. beijerinckii and gas-fermenting C. ljungdahlii. This powerful genetic tool will find utility in synthetic biology research in clostridia.

3.07.5.4 3.07.5.4.1

Cases of Genetic Engineering Engineering Acid Pathways

To improve the yields of solvents, it is expected to reduce the carbon fluxes to acids. So researchers managed to inactivate the genes involved in acids formation. The pta-inactivated mutant PJC4PTA made by homologous recombination showed no significant difference with wild type strain in terms of solvents production. In contrast, buk-inactivated mutant PJC4BK performed as a solvent superproducer, producing 16.7 g L1 butanol, 4.4 g L1 acetone, and 2.6 g L1 ethanol at pH 5.0. Because butyrate-formation is accompanied with ATP production, reduction in the butyrate-formation flux might lead to a metabolic burden. It was observed that acetate production (also responsible for ATP production) by PJC4BK increased from 3.1 to 6.7 g L1, which was probably caused by the cellular demand for ATP which was supplied from butyrate-formation in the wild type strain5,16. In this process of PJC4BK fermentation, the switch from acidogenesis to solventogenesis occurred during the exponential growth phase (A600 < 0.5, in contrast to the end of exponential phase in wild type strain) when the undissociated butyric acid level was less than 1 mM (in contrast to the high undissociated butyric acid concentration (17 mM) in wild type strain). Thus, it was indicated that the accumulation of butyryl-P (not butyric acid) might play a regulatory role and trigger solvent production in C. acetobutylicum. The research on the effects of downregulating the phosphotransbutyrylase (PTB) and butyrate kinase (BK) of the butyrate formation pathway by the asRNA method also confirmed that butyryl-P was a signal molecule for triggering the switch from acidogenesis to solventogenesis. Downregulation of BK resulted in increased solvents production which was similar with BK inactivation. In contrast, downregulation of PTB led to much lower solvent production. Since PTB is involved in butyryl-P formation from butyryl-CoA, downregulation of PTB will result in lower butyryl-P level. According to the correlation between butyryl-P concentration and solvent production in engineered strains with changed activities of PTB or BK, it is believed that butyryl-P plays a key regulatory role in the triggering of solventogenesis.5,16 Butyryl-P is assumed to act as a phosphate donor on the onset of solventogenesis, but the detailed mechanism is still unclear at this time.

3.07.5.4.2

Engineering Solvent Pathways

Mermelstein et al. strengthened the acetone pathway in C. acetobutylicum by overexpressing the genes (adc and ctfA/B) involved in acetone formation. It was observed that acetone-formation related genes were expressed earlier in the engineered strain and led to earlier induction of acetone formation compared to the control strain in pH-controlled fermentation process. The engineered strain produced, relative to the control strain, 95%, 37%, and 90% higher final concentrations of acetone (8.7 g L1), butanol (14 g L1), and ethanol (1.4 g L1), respectively, and almost no residual carboxylic acids were detected. The results indicated that the alcohols production could be promoted by amplification of acetone pathway for enhancing the cellular absorption ability of carboxylic acids. For two enzymes contributing to the acetone pathway, acetoacetate decarboxylase (ADC, encoded by adc) and CoAtransferase (CoAT, encoded by ctfA/B), some studies have been performed. Downregulation of ADC by asRNA showed no concomitant effect on acetone formation, while downregulation of CoAT resulted in a drasticdecrease in acetone production. It was believed that CoAT was the key enzyme for acetone synthesis. Downregulation of CoAT by asRNA also resulted in lower production of butanol. The reason is that aad/adhE1 (alcoholaldehyde dehydrogenase) locates on the same polycistronic message as ctfA/B, so asRNA-mediated downregulation of CoAT leads to the degradation of the whole aad-ctfA-ctfB transcript.5,16 Overexpression of the alcohol-aldehyde dehydrogenase (aad) gene in the engineered strain with CoAT downregulated using antisense RNA (ctfb1-asRNA) restored back to a normal butanol production, while producing little acetone. It indicated that acetone pathway was not necessary for butanol production, although amplification of the acetone pathway does enhance the butanol production. The conclusion was confirmed by two cases: 1) in the C. acetobutylicum M5 lacking the megaplasmid pSOL1 containing the genes involved in ABE formation, overexpression of aad could restore the butanol production, but no acetone was produced; 2) the adc-disrupted mutant achieved by employing group II intron method could produce as much butanol as the control strain, but produced much less acetone. AdhE (aldehyde dehydrogenase encoded by adhE gene, also named aad) is responsible for alcohol (butanol and ethanol) formation. Although overexpression of adhE in the pSOL1 lacking mutant M5 could restore the butanol production, overexpression of adhE in wild-type strain (producing around 10 g L1 of butanol) and solR or buk mutants (producing more than 10 g L1 of butanol) showed no effect on the butanol production. It is considered that adhE just contributes to maintain the basic ability of butanol production (less than 10 g L1 of butanol).5,16

90

Biofuels and Bioenergy: Acetone and Butanol

3.07.5.4.3

Engineering Global Transcriptional Factors

SolR was regarded as a putative DNA-binding transcriptional repressor that negatively regulated solventogenic metabolism. In the solR disrupted mutant, the expressions of the solventogenic genes were induced earlier and maintained throughout the solventogenic phase, resulting in 17.8 g L1 butanol. However, it was pointed that solR should not be the repressor of sol operon (contributing to solvents biosyntheses), indicating that the aforementioned result was incorrect, and meanwhile they suggested that Spo0A, rather than SolR, was a putative regulator of the sol operon. Anyway, solR is a good engineering target for improving butanol production.5,16 The solvent tolerance and stress response of C. acetobutylicum are related to the upregulated expression of genes encoding chaperones (e.g., groES, dnaKJ, hsp18, and hsp90). Overexpression of the heat shock proteins GroES and GroEL (encoded by groESL) led to an increase of the final butanol titer (17 g L1, 33% higher than the control strain) and an enhancement of the tolerance to butanol, presumably by stabilizing global proteins.

3.07.5.4.4

Reconstruction of Butanol Production Pathways in Other Hosts

James C. Liao’s group reported the expression of the genes from C. acetobutylicum involved in solvents pathways in E. coli for 1butanol production. They cloned the crt, bcd, etfAB, hbd and adhE2 genes from C. acetobutylicum, and atoB gene from E. coli to construct two plasmids pJCL17 and pJCL60. The engineered E. coli strain JCL187 (DadhE, DldhA, DfrdBC, Dfnr, Dpta, containing pJCL17 and pJCL60) could produce 552 mg L1 butanol using glycerol as substrate. In the same year, Masayuki Inui et al. reported the construction of butanol pathway in E. coli JM109 via overexpressing crt, bcd, etfAB, hbd, adhE1 and thiL genes of C. acetobutylicum. This strain could produce 16 mM butanol using glucose as substrate. In DuPont’s patent, the thl, hbd, crt, bcd, ald and bdhAB genes of C. acetobutylicum were expressed in E. coli, B. subtilis and Streptococcus cerevisiae, resulting in titers of butanol, 0.8, 0.19 and 0.01 mM, respectively. The engineered S. cerevisiae in another independent work reported by Steen et al. could produce 2.5 mg L1 (0.03 mM) butanol. Nielsen et al. reported their work about re-construction of butanol synthetic pathway in E. coli, Pseudomonas putida and B. subtilis in 2009, demonstrating the potential of engineering butanol biosynthesis in a variety of heterologous microorganisms.16

3.07.6

Systems Biology

3.07.6.1

Genomics

Genomics is the study of the genomes of organisms, aiming to identify the functionality of each gene. In order to obtain adequate molecular knowledge to develop genetically or metabolically engineered butanol-producing strains, the genome of the typical model strain C. acetobutylicum ATCC 824, which is well known for its use in a variety of molecular biology and metabolic engineering studies, was sequenced by the shotgun approach in 2001.35 The C. acetobutylicum ATCC 824 genome consists of 3,940,880 bp, encoding 3740 open reading frames (ORF) while the megaplasmid pSOL1 consists of 192,000 bp, encoding 178 ORFs (according to NCBI genome website https://www.ncbi.nlm.nih.gov/genome/?term¼Clostridiumþacetobutylicum). The megaplasmid pSOL1 contains several genes responsible for solvent production, encoding aldehyde alcohol dehydrogenase (adhE1, CAP0162; adhE2, CAP0035), CoA-transferase (ctfA, CAP0163; CftB, CAP0164) and acetoacetate decarboxylase (adc, CAP0165). Additionally, a phosphotransferase system (PTS, CAP0066
CAP0068) belonging to the fructose-mannose-sorbose (Man) family was identified in pSOL1 (according to NCBI genome website https://www.ncbi.nlm.nih.gov/genome/? term¼Clostridiumþacetobutylicum). More importantly, there are 13 complete phosphotransferase systems encoded by the genome of Clostridium acetobutylcum ATCC 824, which comprise a minimum of IIB and IIC domains or lack an IIA domain. These PTS-encoding systems are well consistent with the fact that the C. acetobutylicum ATCC 824 is capable of utilizing a wide range of substrates, including polysaccharides, disaccharides and monosaccharides. It should be noted that although at least 11 proteins were identified as cellulosome components comprising of cohesins and dockerins, cellulose cannot be metabolized effectively by C. acetobutylicum, which is ascribed to the low sequence homology and specific interactions compared to those of other cellulolytic bacteria. On the other hand, a number of reports suggested that another solventogenic bacterium C. beijerinckii NCIMB 8052 might have greater potential for the industrial production of biofuels, owing to its broader substrate range (pentoses, hexoses, starch, and others) and pH optimum for growth and solvent production. The sequence of C. beijerinckii NCIMB 8052 genome was completed by the University of Illinois and DOE Joint Genome Institute in 2007. However, no plasmid was found in the genome while all the solventogenic genes are located in the chromosome (according to NCBI genome website https://www.ncbi.nlm.nih.gov/genome/? term¼Clostridiumþbeijerinckii). Comparative genetic studies between these two Clostridium strains are expected to uncover genetic and regulatory networks in solventogenesis, which thus provide critical biological information for strains improvements via genetic manipulation. Although C. acetobutylicum ATCC 824 as a typical model strain is systematically investigated for research, many other strains to date have been isolated and adopted for various feedstocks under specific fermentation condition. Therefore, comparative genomics studies can provide an essential insight into the physiological and genetic characterization of C. acetobutylicum strains. For instance, Mo et al. reported the draft genome sequence of C. acetobutylicum strain GXAS18-1, a special butanol-producing strain that can directly ferment cassava flour by ammonium acetate addition. A significant difference between C. acetobutylicum ATCC 824 and GXAS18-1 exists that a clustered regularly interspaced short palindromic repeats (CRISPR) system has been found only in the GXAS18-1 genome.36 Additionally, C. acetobutylicum EA 2018 is characterized as a hyper butanol-producing strain and obtained

Biofuels and Bioenergy: Acetone and Butanol

91

from random mutagenesis of N-methyl-N-nitro-N-nitrosoguanidine (NTG) treatment. The complete genome of C. acetobutylicum EA 2018 was sequenced using Roche 454 pyrosequencing. Genomic comparison with C. acetobutylicum ATCC 824 showed many variations in the EA 2018 strain, including a total of 46 deletion sites and 26 insertion sites, which might contribute to the butanol-overproducing physiological characteristics. The biological information is thus valuable to develop genetically engineered C. acetobutylicum for high-efficient butanol production.37 More recently, a hyper butanol-producing and -tolerant strain C. acetobutylicum JB200, obtained from asporogenic C. acetobutylicum ATCC 55025 via rational mutagenesis and adaptation in a fibrous bed bioreactor, can produce up to
21 g L1 butanol during batch ABE fermentation, increased by
67% compared to that of its parental strain (
12.6 g L1). Comparative genomic analysis showed a single-base deletion in cac3319 encoding histidine kinase (HK) found in the genome of C. acetobutylicum JB200, which thus resulted in C-terminal truncation of the histidine kinase. In fact, the biological function of five orphan HKs, encoded by cac0323, cac0903, cac3319, cac2730 and cac0437, respectively, have been proven to be significantly involved in Spo0A phosphorylation associated with endospore formation of C. acetobutylicum. In order to investigate the regulatory effect of HK on fermentation kinetics and solventogenic metabolism, the cac3319 gene in C. acetobutylicum ATCC 55025 was therefore disrupted using the ClosTron technology. Compared to its parental strain, 44.4% and 90% increases on butanol production and productivity were achieved by the cac3319 inactivation mutant, indicating that CAC3319 could play a critical role regulating metabolic activities of C. acetobutylicum. Those results not only shed light on the exploration of multiple HKs’ functions in the complex regulatory network of butanol biosynthesis but also targeted regulation genes editing for high butanol-producing strain development.38

3.07.6.2

Transcriptomics

DNA microarray is a high-throughput, genomic-scale technology for comparison of relative transcriptional levels between two samples. A microarray consists of targets complimentary to RNA transcripts, such as oligonucleotides or PCR products, which are printed or synthesized on substrates like glass slides. The published genome sequences of solventogenic clostridia make the application of DNA microarrays possible for gene expression profiling in genome scale. Until now, substantial transcriptomics studies have been performed to tailor the target metabolic pathways, for the sake of better understanding of complexed genetic networks and identification of gene functions under certain culture condition. Taken together, these transcriptomics studies on C. acetobutylicum strains have been targeted toward 1) genetic diversity of control and mutant strains; 2) gene expression patterns during different physiological phases; 3) response to stress metabolites such as acetate, butyrate, butanol and other microbial inhibitors; 4) response to available nutrients and culture conditions. Based on the above circumstances, a variety of genetically and metabolically engineered C. acetobutylicum have been developed to improve the ABE fermentation performance, thus making biobutanol more economically competitive on a commercial scale. Generally, C. acetobutylicum produces acids and forms granulose during the exponential growth, subsequently re-assimilate acids to generate solvents and sporulates during the stationary phase. In order to investigate the transcriptional programming, timecourse profiles of different physiological phases were established. Zhao et al. investigated the genes expression pattern in the early sporulation. Alsaker et al. compared and examined genes that expressed in the exponential, transitional and early stationary phases, respectively, and found that most genes on pSOL1 were upregulated at the onset of solventogenesis, which indicated that the megaplasmid could play important roles regulating solventogenesis initiation. Compared with Bacillus subtilis, C. acetobutylicum shows a protracted sporulation. Jones’ further transcriptomic studies on the whole process including late stationary phase revealed for the first time the detailed roadmap of clostridial sporulation. Genes were clustered into six groups according to physiological phase division. On the other hand, stationary phase events have been proven to be regulated by Spo0A, a master regulator in B. subtilis, solventogenic clostridia C. acetobutylicum and C. beijerinckii. Regulatory effects of Spo0A have been extensively studied in B. subtilis. As for C. acetobutylicum ATCC 824, several solventogenic genes are of great interests to be subjected to the protein Spo0A. DNA microarray analysis has been further used to investigate gene expression responses of both inactivation and overexpression of spo0A mutant strain SKO1 and strain 824 (pMSPOA).17 The spo0A mutant strain SKO1 is asporogenous, filamentous and severely deficient in solvent biosynthesis (only 5.7% solvent of the control). Solventogenic gene sigF, carbohydrate metabolic genes and part of electron transport genes were downregulated in strain SKO1 while gene abrB and majority of chemotaxis and motility operons were upregulated. Additionally, expression pattern of several genes responsible for sporulation and solventgenesis were both inhibited. Strain 824 (pMSPOA) exhibited enhanced butanol tolerance as demonstrated by prolonged glucose utilization as well as extended solvent production. Differential expression of genes related to fatty acid metabolism, chemotaxis, heat shock, and cell division, exhibited their close relationship with sporulation and solvent responses. Similar to strain SKO1, strain M5 is characterized as an asporogenous and non-solventogenic phenotype due to loss of the megaplasmid pSOL1, which contains all genes necessary for solvent formation. Besides downregulation of all pSOL1 genes and common genes such as spo0A, sigF, expressions of specific ones such as cheA, cheC, fts, gyrases and DNA helicases encoding genes were remarkably regulated. In order to understand cellular stress responses in C. acetobutylicum, transcriptomic studies associated with stress responses to largely toxic metabolites such as butyrate, acetate and butanol have been intensively carried out. Comparative results showed expression adjustment of saturated- and unsaturated-fatty acids biosynthetic genes, which is responsible for membrane composition adaptation, and obvious upregulation of stress proteins. Despite spo0A is counterintuitive in view of the expectation that stresses could trigger sporulation as a threat signal to cell growth, slight induction of sporulation genes such as spo0A was observed, Nevertheless, relationships between stress responses, solventogenesis and sporulation still remain elusive. Furthermore, in order to

92

Biofuels and Bioenergy: Acetone and Butanol

improve robustness for stresses, heat shock gene groESL was overexpressed in strain 824(pGROE1), leading to the 40% and 33% higher solvent production compared to the wildtype and control strains, respectively. DNA microarray analysis showed that motility and chemotaxis genes were upregulated while other stress responses genes were downregulated, indicating that the overexpression of GroESL could contribute to tolerance and sustain homeostasis. Given that C. acetobutylicum strains could naturally utilize a wide range of substrates, it is of great essence to identify sugar-specific mechanisms for transcriptional regulation of both transport and metabolism genes in C. acetobutylicum ATCC 824. Based on a transcriptomic study, DNA microarrays were used to determine transcript levels of total RNA extracted from cells cultured in media using eleven substrates as sole carbon source, including polysaccharide (starch), disaccharides (sucrose, lactose, maltose and cellobiose), hexoses (glucose, mannose, galactose and fructose) and pentoses (xylose and arabinose). The sugar-specific induction of transport and metabolism genes indicates that these biological mechanisms are significantly regulated at the transcriptional level and subject to carbon catabolite repression (CCR). These comparative results also show that C. acetobutylicum can uptake pentose via symporters and ATP-binding cassette (ABC) transporters, while disaccharides and hexoses are primarily taken up via PTS and a gluconate: Hþ (GntP) transporter. More importantly, the expression of some transporter genes was proven to be induced by specific sugar or a subset of the sugars tested.39 Until recently, micronutrient zinc has been demonstrated to play pleiotropic regulatory roles in ABE fermentation by C. acetobutylicum. In order to elucidate the zinc-associated response on sugar utilization and solventogenesis initiation, transcriptional analysis was performed at the exponential growth phase in glucose medium. Correspondingly, the gene glcG (CAC0570) encoding a glucose-specific PTS was upregulated, together with the other two PTS-encoding genes CAC1353 and CAC1354 in the presence of zinc. Several genes involved in the transport and metabolisms of five other substrates (maltose, cellobiose, fructose, mannose, xylose and arabinose) were differentially overexpressed, demonstrating that the regulatory effect of micronutrient zinc is sugar-specific. Additionally, multiple genes responsible for glycolysis (glcK and pykA), acidogenesis (thlA, crt, etfA, etfB and bcd) and solventogenesis (ctfB and bdhA) of C. acetobutylicum prominently responded to micronutrient zinc at differential expression levels.40

3.07.6.3

Proteomics

The shift paradigm of analyzing proteins pools of a cell at a global scale is directly triggered by the availability of genome sequencing. As proteins participate in virtually every process within cells, their variation could be directly correlated with phenotypic changing. In this section, proteomics studies on solventogenic clostridia will be summarized for analyzing the clostridial physiological shift from acidogenesis to solventogenesis as well as cellular response to lignin inhibition during lignocellulose-based ABE fermentation by C. acetobutylicum. Schatter et al. conducted a comparative proteomics study associated with acidogenesis and solventogenesis in C. acetobutylicum. Of 130 actively synthesized proteins, 52 proteins were found with relatively higher biosynthesis rate while 34 with lower rate during solventogenesis. Eleven of these proteins were identified by N-terminal sequencing, of which Adc, Hsp18, DnaK and GroEL expression abundances were dramatically increased in the solventogenic phase. Interestingly, a tetracistronic operon serCAXS was found derepressed during solventogenesis, but little is known about this physiological regulation. The proteins acetoacetate decarboxylase and Hsp18 both occurred in two variants, indicating possible covalent modification of these proteins. Spo0A is known as a master regulator of the metabolic shift from acidogenesis to solventogenesis. In order to examine Spo0Aassociated expression at the proteomic level, Sullivan and Bennett compared the proteomic profiles of the control strain 824(pIMP1), spo0A inactivated strain SKO1 and spo0A overexpressed strain 824(pMSPOA).41 Twenty-three proteins with differential expression abundance were identified, of which twenty had a unique isoform while five had two or more isoforms, indicating the post-translational phosphorylation and glycosylation modifications in C. acetobutylicum. These identified proteins are significantly involved in heat shock response, acid- and solvent-producing pathways, as well as transcription and translation. In order to investigate protein changes at the early stage of stationary phase when solvents are typically produced, proteomic profiles were compared with the wild-type strain C. acetobutylicum ATCC 824, non-solventogenic strain M5 (losing megaplasmid pSOL1), and strain M5(pIMP1E1AB) (expressing plasmid-based CoA-transferase and aldehyde/alcohol dehydrogenase). Consequently, a total of sixty-eight protein spots, corresponding to fifty-six unique proteins, were clearly identified as being differentially regulated at the early stage of stationary phase in those three strains. In addition to proteins identified to be involved in solventogenesis (AdhE1 and CtfB), significant changes were found in enzymes responsible for sugar transport and metabolism, acid- and solvent-producing pathways, heat shock response, translation, as well as amino acid biosynthesis. On the other hand, significant changes of proteins associated with posttranslational modifications were observed in the solventogenic phase of C. acetobutylicum.42 Raut et al. investigated the effect of lignin on cellobiose consumption, growth rate, morphology, ABE production with a quantitative proteomic analysis of altered proteins associated with the cellular response to lignin inhibition. The study employed 8-plex isobaric tags for relative and absolute quantitation (iTRAQ) to profile biological replicates of four sample types. Of 583 identified proteins, 328 proteins were quantified with at least two unique peptides. Up or downregulated proteins were determined by comparing exponential and stationary phases of C. acetobutylicum ATCC 824 grown on cellobiose in the presence and absence of lignin. Of relevance, most genes involved in glycolytic, acidogenic and solventogenic pathways were downregulated during the exponential and stationary phases in presence of lignin. Moreover, expressions of proteins associated with DNA repair, transcription, translation and GTP/ATP-dependent activities were also significantly regulated, resulting in altered cell morphology.43

Biofuels and Bioenergy: Acetone and Butanol 3.07.6.4

93

Metabolic Modeling

The goal of systems biology is to understand, simulate and redesign the biological systems. Mathematic models have been developed to address the problem of investigation, simulation and prediction. Few metabolic models for clostridial fermentative pathways were published even before the genomics age. Papoutsakis developed the first stoichiometric model of the pathways. The model could be used to calculate or estimate the rates of reactions occurring within the pathways in several solventogenic clostridia. In order to resolve the problem of singularity in the stoichiometric matrix, a nonlinear constraint of in vitro kinetics and selectivity of the CoA transferase from C. acetobutylicum, was thus adopted. The original stoichiometric model was reformulated as a nonlinear constrained minimization problem, and it was then validated using indirect external pH value relationship. Using this model, acidogenic pathway was found to play crucial roles during the solventogenic phase. In the acidogenic phase, acid-producing pathway provides essential energy for cell growth. When switched to solventogenesis, however, the butyrate formation flux is inversed for serving in butyrate re-assimilation, which indicates that enhancing activities of phosphotransbutyrylase (PTB) and butyrate kinase (BK) may improve butanol production without extra acetone generation. While acetate biosynthesis by phosphotransacetylase (PTA) and acetate kinase (AK), still carries on and provides energy even during solventogenesis. Due to the complexity of the clostridial metabolic pathways, all derived metabolic models are static (time-independent), except for the kinetic model for C. saccharoperbutylacetonicum N1-4 ATCC13564, a closely related strain of C. acetobutylicum ATCC 824. Prediction analysis showed high consistency with experimental time-course variation of metabolites. Partially consistent with Papoutsakis’ constrained stoichiometric model, butyrate-butyryl-CoA-cycle was suggested to be the key for improving butanol production. It was predicted by the sensitivity analysis that 5% increase in the reverse flux of butyrate production and 5% decrease in the flux of butyrate metabolism by CoA transferase would highly contribute to the high butanol production.44,45 Genome-scale models offer a possibility to investigate quantitatively the cellular behavior of an organism in silico. For instance, these models can be used as tools to predict growth rates, uptake and secretion rates, and internal metabolic fluxes. They provide a framework to predict cellular behavior before laborious experimental work and thus contribute to design experiments. So far, there are two genome-scale metabolic model for C. acetobutylicum ATCC 824. The first one was established by Papoutsakis’ group in 2008. Rapid network building was completed by combining semi-automated reverse engineering algorithm, physiological knowledge and also thermodynamics analysis of specific pathway. The reconstructed model could qualitatively predict correct fluxes distribution in wildtype or mutant strains, and show the dependency on the urea cycle of a-glutamate biosynthesis in incomplete citric acid cycle, which was different from former hypothesis of pyruvate-based biosynthesis. In their following study, the genomescale metabolic model was further improved by applying more constraints, including external proton concentrations to modeling flux dynamic changes during whole batch fermentation. The second model is established by Lee’s group, which is similar to the former one in terms of the composing reactions, metabolites and enzyme-encoding genes, as well as the simulation of wildtype and mutant strains. The main discrepancy existed in the hypothesis of a-glutarate biosynthesis pathway, as Lee followed Noilling’s suggestion. In addition, Lee’s group applied a non-linear objective function to improve modeling accuracy during solventogenesis, based on the hypothesis stemming from minimization of metabolic adjustment. Moreover, Lee’s group investigated the effect of the hydrogen flux on solventogenic metabolism. It was found that the decreased hydrogen production would reserve more reducing equivalents, thus leading to enhanced NADH-dependent butanol biosynthetic pathway of C. acetobutylicum.44,45 Dash et al. developed a second-generation genome-scale metabolic model for C. acetobutylicum ATCC 824, iCac802 and described the use of transcriptomic data to set additional constraints on reaction flux bounds by the CoreReg method. These constraints were calculated for varying levels of butyrate and butanol stress and were used to identify core sets of altered reactions in flux values, which could broadly account for all the changes observed in cellular metabolism. The CoreReg method was used to differentiate between the two stressors, with a larger restriction on cell growth for butanol stress, and predict not only the metabolic response but the candidate focal points of regulation based on the identified core sets. The resulting CoreReg data could be further used for plausible regulatory loops design within these affected metabolic reactions.46 Constraint-based metabolic model has been widely applied for analyzing multispecies interactions, identifying essential drug targets, predicting optimal culture conditions as well as supporting genome annotation and metabolic engineering. As a matter of fact, the genome-scale constraint-based model has a vast solution space. Therefore, the more detailed biological information on the physiological characteristics of solventogenic clostridia is desired, the more constraints are needed. Wallenius et al. described a step-wise optimization procedure to predict solvent production during continuous ABE fermentation with immobilized C. acetobutylicum cells. The modeling approach applies constraint-based metabolic models as previously documented for C. acetobutylicum behavior without direct flux constraints. As a result, the experimental data set consisting of 25 experiments could be successfully simulated with this modeling approach. Without direct constraint requirements and biological information about the dried cell weight (DCW), a stepwise optimization procedure was thus developed successfully.47 Until recently, Wallenius et al. developed genome-scale metabolic model for investigating the metabolism of C. acetobutylicum in chemostat, which stimulated the bacteria with constant butanol stress under glucose-limited condition. The Constraint-Based Reconstruction and Analysis (COBRA) was combined with additional constraints from 13C Metabolic Flux Analysis (13C-MFA) and experimental results. A model consisting of 451 metabolites and 604 reactions was further applied in flux balance analysis (FBA). Considering various optimization objectives such as intracellular NADH/NADPH generation and ATP maintenance, the flux space was investigated under different modelled conditions (especially with butanol stress). The metabolic flux boundaries were obtained from 13C-MFA while the measured flux and labelled amino acids were used to respond to the 13C-MFA models.

94

Biofuels and Bioenergy: Acetone and Butanol

Additionally, an uncharacterized exopolysaccharide (EPS) as previously documented with C. acetobutylicum was thus characterized on monosaccharide composition level, provoking a further study on the possibilities of natural EPS application for butanol recovery.48 Metabolic pathways are complex in the butanol-producing strain C. acetobutylicum ATCC 824. In contrast with the long-lasting steady states in other microorganisms where flux balancing analysis could be simply applied without altered objective functions, at least two metabolic phases exist in the batch ABE fermentation of C. acetobutylicum, which results in changed objective functions in different phases. Besides, cycling pathways prevalent in the metabolic network of C. acetobutylicum ATCC 824 also lead to unnecessary singularity to the linear system. Therefore, extra constraints including regulation information should be well taken into consideration.

3.07.7

Fermentation

3.07.7.1

Substrates

For the economically feasible ABE fermentation, the substrate cost and availability have long been regarded as significant issues from academic and industrial aspects. In the early days of microbial ABE fermentation, potatoes were widely used as the primary substrate but resulted in inefficient fermentative process. Weizmann then used the cooked maize mash for improved ABE production. Whereafter, an economic fermentative process using sugar-based feedstocks like molasses as carbon sources was further developed. During the World War I, horse chestnuts were used as a candidate substrate for ABE fermentation in England, while the cooked maize mash was utilized in America. After World War I, with the increasing demand for butanol and severe competition with petroleum-based products, the industrial ABE fermentation changed its substrate into molasses, which was available at a price strongly competitive with starchy feedstocks. Correspondingly, new Clostridium strains with capacities of fermenting sugars were isolated. Until the 1950s, with the increased price of molasses and rapid development of petrochemical industry routes, microbial ABE fermentation declined and was forced to be closed. Since the oil crisis began in the 1970s, growing concerns on sustainable and environmental issues renewed interest in butanol as a promising alternative to fossil fuel. The high cost of conventional starch- or sugar-based feedstocks such as maize, wheat, millet and molasses is considered as the most important economic factor affecting the commercialization of industrial ABE fermentation. It is suggested that the future research should focus on renewable feedstocks to revive sustainable ABE fermentation, thus relieving the oil crisis and environmental pollution. In recent years, various substrates derived from lignocellulosic and non-lignocellulosic feedstocks have been successfully utilized for clostridial ABE production, including starch, sucrose, lactose, glucose, fructose, mannose, galactose, xylose, arabinose and glycerol. Especially, the utilization of low-cost and widespread lignocellulosic feedstocks is considered as the hot point for economic ABE fermentation. Lignocellulosic feedstocks, such as corn stover, corn fiber and rice straw, have been intensively investigated. In general, lignocellulosic feedstocks composed of cellulose (40%–45%), hemicellulose (20%–30%), lignin (10%–25%), ash and extractives, have long been used as a source of energy for years. Both hemicellulose and cellulose hydrolysates can be utilized by C. acetobutylicum to produce ABE. Xylose is the main component of pentose sugars, which is easy to be released from hemicellulose. It was confirmed that if xylose could be co-utilized with glucose, relatively higher yield of fermentable sugars would be achieved. Among abundant lignocellulosic feedstocks, corn stover (CS) is well accepted as the most favorable feedstock for ABE fermentation due to its widespread availability, high residue yield and sugar content. Followed by indispensable pretreatment and enzymatic hydrolysis, 100 g dry CS can be hydrolyzed into 42–66 g of fermentable sugars (mainly glucose and xylose), which is utilized as carbon sources toward butanol by C. acetobutylicum. Although H2SO4-pretreatment is cost-effective and widely used for preparing corn stover hydrolysates (CSH) at industrial scale, the major challenges for ABE fermentation from lignocellulosic hydrolysate lie in poor sugar utilization, cell growth and metabolic shift from acidogenesis toward solventogenesis, which inevitably suffer from weak acids, furan derivatives, phenolic compounds and soluble salts generated during H2SO4-pretreatment and subsequent neutralization process, making CS-based butanol production less economically competitive with petroleum-driven butanol due to its relatively low production and productivity. Despite physical, chemical and biological detoxification methods have been applied for eliminating the effects of inhibitors on ABE fermentation, the resulting wastewater generation, energy cost and sugar loss will be unfavorable at industrial scale. Hence, renewable substrates, efficient pretreatment and hydrolysis technologies will make the lignocellulosic ABE fermentation more economic and sustainable.

3.07.7.2

Butanol Toxicity and Tolerance

The ABE fermentation has long been limited to low butanol production, yield and productivity due to the severe butanol toxicity to Clostridium strains. The major challenge lies in the clostridial survival from relatively high butanol stress. Among the end products, butanol is the most toxic with a lethal concentration to inhibit the clostridial metabolism, while acetone and ethanol will not cause severe inhibitory effects owing to their low concentrations throughout the whole fermentation. A typical batch fermentation will completely cease with about 20 g L1 total solvents produced, of which butanol concentration rarely exceeds 13 g L1. In nature, once butanol concentration reaches the threshold of about 4.0
4.8 g L1 during the ABE fermentation, the cell growth and overall metabolism will be severely inhibited.

Biofuels and Bioenergy: Acetone and Butanol

95

The butanol toxicity is directly related to its great polarity and hydrophobicity as a long-chain alcohol. Physiological investigation revealed a close correlation between the solvent toxicity and its logPow, a value defined as the logarithm of the partition coefficient of the solvent tested in a defined octanol–water mixture. Generally speaking, as the logPow value decreases, more solvent will be accumulated in the membrane with increased toxicity. Solvents with LogPow below 4.0 are usually considered to be extremely toxic to the cells. LogPow for butanol is 0.8, which means a strong trend to be accumulated in the phospholipid bilayer of cytoplasmic membrane, thus resulting in severe damage of the structural and physiological integrity of the cells. High concentration of butanol accumulated in the cytoplasmic membrane will lead to structural damage, consequential inhibition, as well as disruption of physiological function. Cell membrane is the primary target site of butanol toxicity. The semipermeable membrane surrounding a cell is not only a physical barrier separating the intracellular components from the extracellular environment, but also a functional complex anchoring various types of important enzymes and transport proteins. It plays a vital role in nutrients transportation, signal and energy transduction, regulation of the intracellular environment as well as maintenance of the cellular energy status and osmotic pressure. Hence, it is of great importance to maintain a balance between the rigidity and fluidity of the cytoplasmic membrane. Contrary to short-chain aliphatic alcohols such as ethanol, butanol that partitions into the lipid bilayer will increase the fluidity of the cell membrane and decrease the stability, thus resulting in the loss of related physiological function. The increased fluidity of the cytoplasmic membrane exhibits a higher intensity of lipid dispersal and will lead to more easily breaking of the phospholipid molecules from the membrane and a weaker structure of the membrane skeleton. Therefore, the membrane becomes more sensitive to environmental perturbation with sequentially high possibility of cell lysis. Besides the direct structural damage, butanol toxicity to structural stability can induce an autonomic degradation of the cells. Butanol concentration above the inhibitory threshold will lead to the release of autolysin, which can hydrolyze bacterial components by breaking down the beta 1–4 bond between N-acetylmuramic acid and N-acetylglucosamine molecules. The excessive amount of autolysin will degrade the peptidoglycan matrix in the cell wall and bring about the burst of cells due to cellular osmotic pressure. Instability of the cell membrane caused by high butanol stress will further inhibit or disrupt some membrane-dependent physiological functions and processes. Contrary to ethanol, high concentration of butanol has a significant inhibitory effect on the membrane-bound ATPase activity. Clostridium cells can maintain a constant intracellular pH value through an active manner requiring an effective Hþ-ATPase system. The system couples the ATP hydrolysis and the proton transportation across the membrane, avoiding a drastic decrease of internal pH and keeping a constant pH gradient across the membrane. It was reported that butanol concentration of 10 g L1 will completely abolish the pH gradient. Although the intracellular ATP level shows a significant decrease at the same time with the pH gradient crash, the abolishment of pH gradient and the decrease of intracellular ATP level are considered as two independent events. The nutrients transport, another important membrane-associated function, is also inhibited by butanol toxicity. Uptakes of glucose and amino acids are both greatly demolished when the butanol concentration reaches an inhibitory level. However, relationship between this demolishment and the decreased intercellular ATP level has not been fully understood. Furthermore, besides the membrane-linked inhibitory effects, butanol toxicity will also impact the viable cells through some other ways. For example, together with the energy shortage, the starvation of nutrients caused by the blocked transport system will restrict cell growth. Additionally, protein denaturation is supposed to influence the cellular physiology in a wider range and deeper level. Butanol toxicity is determined by not only the solvent toxicity but also the intrinsic tolerance of the cells. Through the long journey of evolution, butanol-producing strains such as C. acetobutylicum have developed systematic mechanisms to adapt cells to butanol stress. Changed ester-linked fatty acids composition can effectively readjust the membrane fluidity. The ratio of the saturated-to-unsaturated fatty acids, as well as the content of long-chain fatty acids will be increased to buffer the excessive membrane fluidity of C. acetobutylicum cells. These responses are also observed against heat shock and stress from other long-chain aliphatic alcohols. In the meantime, members of the HSP family such as GroES/L and DnaK/J will also be upregulated to maintain the protein biosynthetic quality and to remodel misfolding proteins. Until now, the maximum butanol production of
21 g L1 is achieved by C. acetobutylicum JB200, a mutant with hyper butanolproducing and butanol-tolerating capacities. While C. beijerinckii BA101 could produce 19.6 g L1 butanol and tolerate up to 23 g L1 butanol. Overall, butanol toxicity has complex mechanisms with multiple effects on the cellular structure, physiology and metabolism, and there are still many details and mechanisms on butanol tolerance to be explored and disclosed.

3.07.7.3

Strain Improvement and In Situ Recovery Technology

The major obstacle hindering the development of industrial ABE fermentation is relatively low production and productivity caused by low butanol tolerance of Clostridium strains. To address these problems, tremendous efforts have been made on improvements of butanol-producing and butanol-tolerating capacities via genetic manipulation and rational mutagenesis, as well as establishments of advanced continuous or in situ removal of butanol from fermentation broth. For strain improvements, traditional mutagenesis is one of the most reliable and widely used approaches. During organism breeding, mutations are induced using physical (heavy-ion irradiation, ultraviolet, X-rays, etc.) or chemical (methyl methane sulfonate, hydroxyl amine, N-methyl-N0 -nitro-N-nitrosoguanidine, etc.) mutagens. By harnessing spontaneous mutation, the most well-known strain C. acetobutylicum JB200, a hyper butanol-producing and butanol-tolerant mutant, was obtained from asporogenic C. acetobutylicum ATCC 55025 via rational mutagenesis and adaptation in a fibrous bed bioreactor.49 Besides, the C. beijerinckii BA101 with a butanol tolerance up to 23 g L1 and C. acetobutylicum BKM19 were both successfully obtained from chemical mutagenesis, which exhibited stable butanol production and tolerance.50 Although these mutagenesis strategies managed to generate

96

Biofuels and Bioenergy: Acetone and Butanol

abundant mutants, it takes lots of time for screening of target mutants favorable for ABE fermentation under specific culture conditions. While the rational strain improvement using recombinant DNA technology is significantly restricted to the complexed mechanisms of stress response and the difficulty of clostridial genetic manipulation. Nevertheless, several cases have been successfully documented. For example, the overexpression of HSP GroESL definitely improved butanol tolerance of C. acetobutylicum. More importantly, genetic and metabolic engineering strategies could be also used for improving solventogenic metabolism and co-utilization of complexed substrates, as well as developing non-native butanol-producing strains as host for butanol production. In the past decades, with the development of synthetic biology technology and molecular manipulation methodology, homologous recombination, antisense RNA, mobile group II intron retargeting, and CRISPR-Cas9 editing technologies have been widely applied for genetically engineering Clostridium strains. A variety of genes involved in sugar transport and metabolism, glycolytic, acidogenic and solventogenic pathways, sporulation and signal transduction have been extensively investigated.1 Typically, the butanol biosynthetic genes were successfully introduced into non-native ethanol-producing microorganisms such as S. cerevisiae and E. coli, which indicate the feasibility of applying these non-native hosts for butanol production but the poor butanol productions and productivities. Additionally, the gene cac3319 encoding an orphan histidine kinase in C. acetobutylicum ATCC 55025 was inactivated via the ClosTron technology, leading to 44.4% more butanol production and 90% higher butanol productivity, thus indicating that CAC3319 could play a vital role regulating butanol biosynthesis.51 Undoubtedly, strain improvement for ABE fermentation is expected to achieve a remarkable breakthrough in the near future. To alleviate the butanol toxicity to microbial cells, numerous efforts have been also made to develop in situ product recovery techniques, including gas stripping, liquid-liquid extraction, adsorption, pervaporation, and vapor stripping-vapor permeation. As compared with conventionally well-known distillation separation technology performed off-line at high energy consumption, these advanced alternatives can continuously remove butanol during the fermentation, thus reducing butanol toxicity and improving the ABE fermentation efficiency. Gas stripping is a simple separation technique which has the ability of selective removal of volatile components from the fermentation broth. In this process, the oxygen-free nitrogen or fermentation gases (CO2 and H2) are sparged into the broth to strip solvent away. Then the stripping mixture passes through a condenser, in which the vaporized solvent is condensed and the resting gas is recycled back to the fermentor to strip more butanol. The performance of gas stripping is limited by the solvent concentration in the fermentation broth. Therefore, the intermittent strategy for gas stripping was proposed to control the butanol production in fermentation broth, thereby achieving high solvent production (175.6 g L1).1 Although the butanol production obtained from one-stage gas stripping is much higher than that in the fermentation broth, the condensate still contains a large amount of water. Furthermore, the condensate recovered in the first-stage gas stripping was concentrated by second-stage gas stripping (to 420.3 g L1) or pervaporation process using a carbon nanotube composite membrane (to 521.3 g L1).52,53 It should be noted that large amount of circulating gas will lead to no cell damage but a mass of foam formation, which may cause the addition of antifoam. Since it has several advantages of simple scale up, easy operation, only removal of volatile compounds and so on, gas stripping still is a promising and competitive process for in situ butanol recovery for industrial application after optimizing the parameters such as gas recycle rate, bubble size and condensation temperature. Liquid-liquid extraction is another efficient recovery technology that can remove butanol from the fermentation broth in situ. This method is established on the principle that solubilities of chemicals vary in different solutions and distribution coefficients of the chemicals vary between two immiscible phases. In this process, when a water-insoluble organic extractant is mixed with the liquid fermentation broth, butanol will be selectively concentrated in the organic phase, because it is more soluble than in the aqueous phase. Taking advantage of the immiscibility between the two phases, extractants containing butanol can be simply separated from the fermentation broth without removing substrates, water, nutrients or cells. Various extractants including oleyl alcohol, methylated crude palm oil, decanol and biodiesel have been applied for removal of ABE products from fermentation broth. Especially for oleyl alcohol as an extractant, 35.9 g L1 butanol could be produced via oleyl alcohol extraction when fermenting cassava as the substrate by C. acetobutylicum ATCC 824. Sometimes mixture of several extractants is employed to balance the conflict between low toxicity and high distribution coefficient. One of the biggest challenges for in situ liquid-liquid extraction is to screen a proper extractant which can satisfactorily meet the key characteristics, including low toxicity to butanol producing cells, high distribution coefficient for butanol, immiscibility with the fermentation broth, low cost, and high availability. Notably, to address potential problems associated with liquid-liquid extraction, such as emulsion formation and toxicity, the proposed perstraction process can be considered as an advanced liquid-liquid extraction with a semi-permeable membrane placed between the extractant and the fermentation broth. When liquid-liquid extraction or perstraction is employed for in situ butanol recovery, it is crucial to regenerate and recycle the extractant for reducing the cost of downstream separation process.54 Adsorption is regarded as the oldest energy-efficient technique applied for in situ product recovery from ABE fermentation. In adsorption process, butanol can be adsorbed by adsorbents and then recovered at special conditions such as high temperature, resulting in concentrated butanol solution. However, the relatively low adsorption capacity and selectivity for butanol together with the high price of absorbents have been considered as great obstacles to promote this technology. The adsorption capacity of butanol is closely related to adsorbents surface area. In selecting a suitable adsorbent, the main concerns are the loading capacity and biocompatibility. A wide range of materials investigated as adsorbents include active carbon, zeolites, polyvinylpyridine, polymeric resins.55 Among all these adsorbents, active carbon showed the best performance with a high adsorption capacity for butanol recovery. Xue et al. developed in situ butanol adsorption using active carbon (Norit ROW 0.8) coupled with immobilized cells for

Biofuels and Bioenergy: Acetone and Butanol

97

ABE fermentation, obtaining a condensate containing
167 g L1 butanol by thermal desorption.56 This is the highest butanol concentration ever achieved via adsorption up to date, but still lower than that obtained by gas stripping. Among the selective membrane-based in situ butanol recovery technologies applied for ABE fermentation, pervaporation process is regarded as a promising alternative and characterized as a liquid or solid hydrophobic membrane contacting with the fermentation broth, wherein the solvents selectively diffuse across the membrane as vapors leaving behind nutrients, substrate, and the microbial cells. The compounds selectively removed will be recovered by condensation to keep the concentrate and vapor pressure gradients allowing the diffusion to continue. During reverse osmosis process, suspended microbial cells should be restored with hollow-fiber ultrafilter as a pretreatment. Sequentially, pretreated fermentation broth will be concentrated through a semipermeable membrane allowing only water molecules to pass. Finally, the dewatered fermentation will be further distilled to get butanol with higher purity. Generally, pervaporation performance is primarily governed by the structure and property of membrane. Different types of pervaporation membranes have been evaluated for in situ butanol recovery, of which polydimethylsiloxane (PDMS)-based membranes are well accepted for pervaporation process due to its excellent mechanical and thermal stability. The PDMS-based membrane possessing several advantages such as low energy consumption, no negative effect on the microbial cells, no losses of nutrients and substrates, higher selectivity and so on, is considered as a promising method for in situ butanol recovery. Most recently, the aligned and open-ended CNT/(polydimethylsiloxane) PDMS membranes are controllably fabricated to form a hamburger-like structure that possesses nanochannels (
10 nm) in the intermediate layer as well as angstrom cavities in the embedded PDMS. These aligned CNT membranes surpass the filling content limitation of the nonaligned CNT/PDMS membrane, leading to excellent mechanical properties and a multiplying performance increase of mass flux and selectivity for the separation of alcohols.57 However, the membrane contamination caused by the adsorbed ions, sugars, cells and biomacromolecules, is the main challenge for long-term operation. Apart from the abovementioned recovery techniques, the single-stage close-circulating vapor stripping-vapor permeation (VSVP) is attracting worldwide academic and industrial attention. In the VSVP process, the off-gas generated during ABE fermentation is circulated between bioreactor and membrane module to produce feed vapor mixture, thus volatile solvents are stripped with bubbles and separated into solvent- and water-rich vapor stream by a hydrophobic membrane. Compared with the gas stripping process, the vaporized solvents in the VSVP process permeate through a membrane before condensation, wherein butanol separation factor could be promoted remarkably. Furthermore, high solvent concentration could be obtained from the vaporized solvent contacting with both sides of membrane. A VSVP with temperature-difference control has been recently described and integrated with ABE fermentation. As a result, 212.7 g L1 butanol (339.3 g L1 ABE) was generated in condensate, with remarkably reduced energy consumption of 19.6 kJ g1 butanol, superior to other one-stage recovery techniques.58 As briefly described above, many advances in developments of butanol-tolerating strains and in situ butanol recovery have been achieved. However, neither of them can be considered as a perfect solution owing to different obstacles restricting their wide application, which provoked further investigations on efficient genome editing techniques, and integration of recovery systems followed by process optimization and overall analyses of fermentation performance, ecological and environmental factors and so on.

3.07.7.4

Development of Fermentation Technology

Industrial ABE fermentation is traditionally operated through a batch fermentation process followed by distillation process for final product recovery and purification, due to the simple operation and reduced risk of contamination. Fermentors for biobutanol production have a capacity of 200,000 to 800,000 L without mechanical agitation system. In some cases, a set of 20 or more fermentors are connected in series. Initial medium sterilized with high temperature usually fulfills 90%–95% of the fermentor capacity, and sterilized CO2 will be bubbled across the broth for thoroughly. The cultures preserved in form of spores are heated at 65–100  C for 1–3 min for activation, followed by 2 to 4 buildup stages and subsequent inoculation into the fermentors with an inoculum amount of 2%–4%. ABE fermentation usually runs at a temperature ranging from 29 to 39  C and applied in a fermentation period of 40 to 60 h. The final butanol production ranges from 12 to 22 g L1. After fermentation, cell mass and other suspended solids will be removed by centrifugation and added into cattle feed, the liquid broth will be finally distilled. Sometimes, gases (carbon dioxide and hydrogen) generated during the fermentation are also recovered and separated for a variety of purposes. Depending on the strains used for batch ABE fermentation, final butanol production is usually less than 20 g L1 with a solvent yield varying from 0.30 to 0.40 g g1. Butanol productivity is achieved at relatively low levels of 0.34 to 0.46 g L1 h due to substrate inhibition, lag phase, butanol toxicity and downtime for cleaning, sterilization and charging. The problem of product inhibition can be solved by applying in situ product recovery technology to decrease the butanol concentration in fermentation broth. Due to ease of operation, the batch technology was widely applied for ABE production in the last century, along with low solvent productivities and the requirement of large fermentors. Therefore, substantial efforts have been made to develop advanced fermentation technologies with enhanced productivity for ABE fermentation. Continuous fermentation with free cells can be applied to eliminate the lag phase, and simplify upstream and downstream processes. Fed-batch fermentation can be used to relieve the substrate inhibition. As well, for increasing butanol productivity with high cell density, some researches have been performed on the use of immobilized cell reactors and membrane cell recycle bioreactors. For industrial ABE fermentation, fed-batch technology is employed when high concentration of substrate exhibits inhibitory effects on the cells. Compared to the batch mode, the working volume of fermentor is not high (usually less than 50%) during fed-batch culture, and the initial concentration of substrate is kept below the inhibitory threshold. During the fermentation, concentrated substrate is replenished so that the substrate concentration in the medium remains relatively stable at a non-inhibition level.

98

Biofuels and Bioenergy: Acetone and Butanol

As the volume of fermentation broth increased to approximate 75% of the fermentor capacity, the process will be stopped with broth harvested. However, the application of fed batch in biobutanol production must be combined with the use of in situ product recovery technology due to the high toxicity of butanol. When relieving the substrate and product inhibition simultaneously, butanol yield and productivity will be significantly improved in fed-batch fermentation coupled with online butanol recovery system. Continuous fermentation technology has been well established to improve bioreactor productivity by keeping the culture at a steady state during fermentation. In continuous mode, cells in the fermentor are cultivated to the exponential phase, when fresh medium will be supplemented into the fermentor continuously and product stream will be exported out with the same flow rate to keep a constant working volume of fermentation. Continuous fermentation reduces the lag phase and downtime, and prolongs the solventogenic phase during ABE fermentation, which contributes to improved butanol productivity. However, the complexity of ABE fermentation brings great challenge to the simple single-stage continuous fermentation at the industrial scale, such as the instability of butanol production and the increase of acids generation. To solve these problems, two- or more-stage continuous fermentation technology have been investigated to reduce oscillation and improve final butanol production. For example, in a typical two-stage fermentors, the acidogenic and solventogenic phase of ABE fermentation are carried out in separated fermentors under optimal conditions to avoid the interactive influence. Additionally, the continuous free cell system has another notable problem which high cell concentration cannot be retained due to cell washout at high dilution rates. In the traditional ABE fermentation, the cell density during batch culture is usually limited to no more than 4 g L1 (DCW), which is an important reason for the low solvent productivity. Considering that high cell density greatly contributes to fermentation efficiency, fermentation with immobilized cells and membrane-based cell recycle technology have been investigated to retain high cell density through different ways. Cell immobilization makes the ABE fermentation possible to keep higher cell densities by reducing cell loss and compatible with various reactor configurations (stirred tank bioreactor, packed-bed bioreactor, fluidizedbed bioreactor, etc.). Several cost-effective matrixes were employed for cell immobilization, such as calcium alginate, chitosan, coke, and so on. The matrix can be placed in a tubular bioreactor where continuous fermentation can be operated, with fresh medium constantly imported into the bioreactor and products exported from the bottom. The matrix can usually immobilize a high amount of cells, thus achieve significantly improved productivity. In continuous fermentation with immobilized cells, cell density can reach up to 50–70 g L1(DCW) at high flow rates, which result in a 20- to 40-fold increase in solvent productivity compared to the batch mode. Besides, cell immobilization mode is more stable with longer life spans in comparison with that of free cell operation. However, the immobilized cells culture will be limited to substrate, nutrient, and product diffusion, resulting in cell inactivation or death, especially in the innermost cell layers due to starvation of nutrients and toxicity of butanol. Not all cells immobilized function for solvent synthesis, amounts of them present in the form of inactive spores. Thus, the sporulation of the butanol-producing cells is suggested to be blocked. For the membrane-based cell recycle continuous fermentation, the bioreactor is initially in a batch mode and cells are cultivated to the exponential phase. Prior to reaching the stationary phase, the fermentation broth is circulated, and then microbial cells are captured by a filtration membrane which allows the aqueous production solution to permeate through. When both medium inflow and product removal flow are equal, a constant volume is kept in the fermentor. As a result, cells density can be accumulated up to 100 g L1 (DCW) by introducing cell recycling, which will lead to a significant improvement of butanol productivity as high as 10 g L1 h. However, like all other membrane-based fermentation technology, this type of bioreactors is commonly encountered the obstacle of fouling and clogging.59

References 1. Xue, C.; Zhao, J. B.; Chen, L. J.; et al. Recent Advances and State-of-the-art Strategies in Strain and Process Engineering for Biobutanol Production by Clostridium acetobutylicum. Biotechnol. Adv. 2017, 35, 310–322. 2. Jones, D.; Woods, D. Acetone-butanol Fermentation Revisited. Microbiol. Rev. 1986, 50, 484. 3. Gu, Y.; Jiang, Y.; Wu, H.; et al. Economical Challenges to Microbial Producers of Butanol–Feedstock, Butanol Ratio and Titer. Biotechnol. J. 2011, 6, 1348–1357. 4. Poehlein, A.; Solano, J. D. M.; Flitsch, S. K.; et al. Microbial Solvent Formation Revisited by Comparative Genome Analysis. Biotechnol. Biofuels 2017, 10, 58. 5. Deutscher, J. The Mechanisms of Carbon Catabolite Repression in Bacteria. Curr. Opin. Microbiol. 2008, 11, 87. 6. Xiao, H.; Li, Z. L.; Jiang, Y.; Yang, Y.; et al. Metabolic Engineering of D-xylose Pathway in Clostridium beijerinckii to Optimize Solvent Production from Xylose Mother Liquid. Metab. Eng. 2012, 14, 569–578. 7. Ren, C.; Gu, Y.; Hu, S. Y.; et al. Identification and Inactivation of Pleiotropic Regulator CcpA to Eliminate Glucose Repression of Xylose Utilization in Clostridium acetobutylicum. Metab. Eng. 2010, 12, 446–454. 8. Xiao, H.; Gu, Y.; Ning, Y. Y.; et al. Confirmation and Elimination of Xylose Metabolism Bottlenecks in Glucose Phosphoenolpyruvate-dependent Phosphotransferase Systemdeficient Clostridium acetobutylicum for Simultaneous Utilization of Glucose, Xylose, and Arabinose. Appl. Environ. Microbiol. 2011, 77, 7886–7895. 9. Bruder, M.; Moo-Young, M.; Chung, D. A.; Chou, C. P. Elimination of Carbon Catabolite Repression in Clostridium acetobutylicum–a Journey Toward Simultaneous Use of Xylose and Glucose. Appl. Microbiol. Biotechnol. 2015, 99, 7579–7588. 10. Jin, L.; Zhang, H.; Chen, L. W.; et al. Combined Overexpression of Genes Involved in Pentose Phosphate Pathway Enables Enhanced D-xylose Utilization by Clostridium acetobutylicum. J. Biotechnol. 2014, 173, 7–9. 11. Gu, Y.; Li, J.; Zhang, L.; et al. Improvement of Xylose Utilization in Clostridium acetobutylicum via Expression of the TalA Gene Encoding Transaldolase from Escherichia coli. J. Biotechnol. 2009, 143, 284–287. 12. Ling, H.; Teo, W. S.; Chen, B. B.; et al. Microbial Tolerance Engineering toward Biochemical Production: from Lignocellulose to Products. Curr. Opin. Biotechnol. 2014, 29, 99–106. 13. Lin, D. S.; Yen, H. W.; Kao, W. C.; et al. Bio-butanol Production from Glycerol with Clostridium pasteurianum CH4: the Effects of Butyrate Addition and In Situ Butanol Removal via Membrane Distillation. Biotechnol. Biofuels 2015, 8, 168.

Biofuels and Bioenergy: Acetone and Butanol

99

14. Tracy, B. P.; Jones, S. W.; Fast, A. G.; et al. Clostridia: The Importance of Their Exceptional Substrate and Metabolite Diversity for Biofuel and Biorefinery Applications. Curr. Opin. Biotechnol. 2012, 23, 364–381. 15. Fast, A. G.; Schmidt, E. D.; Jones, S. W.; Tracy, B. P. Acetogenic Mixotrophy: Novel Options for Yield Improvement in Biofuels and Biochemicals Production. Curr. Opin. Biotechnol. 2015, 33, 60–72. 16. Lee, S.; Park, J.; Jang, S.; et al. Fermentative Butanol Production by Clostridia. Biotechnol. Bioeng. 2008, 101, 209–225. 17. Harris, L. M.; Welker, N. E.; Papoutsakis, E. T. Northern, Morphological, and Fermentation Analysis of spo0A Inactivation and Overexpression in Clostridium acetobutylicum ATCC 824. J. Bacteriol. 2002, 184, 3586–3597. 18. Harris, L. M.; Welker, N. E.; Papoutsakis, E. T. Multiple Orphan Histidine Kinases Interact Directly with Spo0A to Control the Initiation of Endospore Formation in Clostridium acetobutylicum. J. Bacteriol. 2002, 184, 3586–3597. 19. Ren, C.; Gu, Y.; Wu, Y.; et al. Pleiotropic Functions of Catabolite Control Protein CcpA in Butanol-producing Clostridium acetobutylicum. BMC Genomics 2012, 13, 349. 20. Yang, Y.; Zhang, L.; Huang, H.; et al. A Flexible Binding Site Architecture Provides New Insights into CcpA Global Regulation in Gram-Positive Bacteria. mBio 2017, 8, e02004– e02016. 21. Scotcher, M. C.; Rudolph, F. B.; Bennett, G. N. Expression of abrB310 and SinR, and Effects of Decreased AbrB310 Expression on the Transition from Acidogenesis to Solventogenesis in Clostridium acetobutylicum ATCC 824. Appl. Environ. Microbiol. 2005, 71, 1987–1995. 22. Zhang, L.; Nie, X. Q.; Ravcheev, D. A.; et al. Redox-responsive Repressor Rex Modulates Alcohol Production and Oxidative Stress Tolerance in Clostridium acetobutylicum. J. Bacteriol. 2014, 196, 3949–3963. 23. Wietzke, M.; Bahl, H. The Redox-sensing Protein Rex, a Transcriptional Regulator of Solventogenesis in Clostridium acetobutylicum. Appl. Microbiol. Biotechnol. 2012, 96, 749–761. 24. Mercante, J.; Suzuki, K.; Cheng, X. D.; et al. Comprehensive Alanine-scanning Mutagenesis of Escherichia coli CsrA Defines Two Subdomains of Critical Functional Importance. J. Biol. Chem. 2006, 281, 31832–31842. 25. Schubert, M.; Lapouge, K.; Duss, O.; et al. Molecular Basis of Messenger RNA Recognition by the Specific Bacterial Repressing Clamp RsmA/CsrA. Nat. Struct. Mol. Biol. 2007, 14, 807–813. 26. Tan, Y.; Liu, Z. Y.; Liu, Z.; et al. Comparative Transcriptome Analysis between csrA-disruption Clostridium acetobutylicum and its Parent Strain. Mol. Biosyst. 2015, 11, 1434–1442. 27. Gaida, S. M.; Al-Hinai, M. A.; Indurthi, D. C.; et al. Synthetic Tolerance: Three Small Noncoding RNAs, DsrA, ArcZ and RprA, Acting Supra-additively against Acid Stress. Nucleic Acids Res. 2015, 41, 8726–8737. 28. Jones, A. J.; Fast, A. G.; Clupper, M.; Papoutsakis, E. T. Small, Low, but Potent: the Complex Regulatory Role of the Small RNA SolB on Solventogenesis in Clostridium acetobutylicum. Appl. Environ. Microbiol. 2018. https://doi.org/10.1128/AEM.00597-18. 29. Chen, Y. L.; Indurthi, D. C.; Jones, S. W.; Papoutsakis, E. T. Small RNAs in the Genus Clostridium. mBio 2011, 2. e00340-10. 30. Hsu, P. D.; Lander, E. S.; Zhang, F. Development and Applications of CRISPR-Cas9 for Genome Engineering. Cell 2014, 157, 1262–1278. 31. Wang, Y.; Zhang, Z. T.; Seung-Oh, S.; et al. Markerless Chromosomal Gene Deletion in Clostridium beijerinckii Using CRISPR/Cas9 System. J. Biotechnol. 2015, 200, 1–5. 32. Xu, T.; Li, Y. C.; Shi, Z.; et al. Efficient Genome Editing in Clostridium cellulolyticum via CRISPR-Cas9 Nickase. Appl. Environ. Microbiol. 2015, 81, 4423–4431. 33. Tummala, S.; Welker, N.; Papoutsakis, E. T. Design of Antisense RNA Constructs for Downregulation of the Acetone Formation Pathway of Clostridium acetobutylicum. J. Bacteriol. 2003, 185, 1923. 34. Qi, L. S.; Larson, M. H.; Gilbert, L. A.; et al. Repurposing CRISPR as an RNA-guided Platform for Sequence-specific Control of Gene Expression. Cell 2013, 152, 1173–1183. 35. Nolling, J.; Breton, G.; Omelchenko, M.; et al. Genome Sequence and Comparative Analysis of the Solvent-producing Bacterium Clostridium acetobutylicum. J. Bacteriol. 2001, 183, 4823. 36. Mo, X. C.; Pei, J. X.; Guo, Y.; et al. Genome Sequence of Clostridium acetobutylicum GXAS18-1, a Novel Biobutanol Production Strain. Genome Announc. 2015, 3. 37. Hu, S. Y.; Zheng, H. J.; Gu, Y.; et al. Comparative Genomic and Transcriptomic Analysis Revealed Genetic Characteristics Related to Solvent Formation and Xylose Utilization in Clostridium acetobutylicum EA 2018. BMC Genomics 2011, 12, 93. 38. Xu, M. M.; Zhao, J. B.; Yu, L.; Yang, S. T. Comparative Genomic Analysis of Clostridium acetobutylicum for Understanding the Mutations Contributing to Enhanced Butanol Tolerance and Production. J. Biotechnol. 2017, 263, 36–44. 39. Servinsky, M. D.; Kiel, J. T.; Dupuy, N. F.; Sund, C. J. Transcriptional Analysis of Differential Carbohydrate Utilization by Clostridium acetobutylicum. Microbiology 2010, 156, 3478–3491. 40. Wu, Y. D.; Xue, C.; Chen, L. J.; et al. Transcriptional Analysis of Micronutrient Zinc-associated Response for Enhanced Carbohydrate Utilization and Earlier Solventogenesis in Clostridium acetobutylicum. Sci. Rep. 2015, 5, 16598. 41. Sullivan, L.; Bennett, G. Proteome Analysis and Comparison of Clostridium acetobutylicum ATCC 824 and Spo0A Strain Variants. J. Ind. Microbiol. Biotechnol. 2006, 33, 298–308. 42. Jang, Y. S.; Han, M. J.; Lee, J.; et al. Proteomic Analyses of the Phase Transition from Acidogenesis to Solventogenesis Using Solventogenic and Non-solventogenic Clostridium acetobutylicum Strains. Appl. Microbiol. Biotechnol. 2014, 98, 5105–5115. 43. Raut, M. P.; Couto, N.; Pham, T. K.; et al. Quantitative Proteomic Analysis of the Influence of Lignin on Biofuel Production by Clostridium acetobutylicum ATCC 824. Biotechnol. Biofuels 2016, 9, 113. 44. Senger, R.; Papoutsakis, E. T. Genome-scale Model for Clostridium acetobutylicum: Part I. Metabolic Network Resolution and Analysis. Biotechnol. Bioeng. 2008, 101, 1036–1052. 45. Senger, R.; Papoutsakis, E. T. Genome-scale Model for Clostridium acetobutylicum: Part II. Development of Specific Proton Flux States and Numerically Determined Subsystems. Biotechnol. Bioeng. 2008, 101, 1053. 46. Dash, S.; Mueller, T. J.; Venkataramanan, K. P.; et al. Capturing the Response of Clostridium acetobutylicum to Chemical Stressors Using a Regulated Genome-scale Metabolic Model. Biotechnol. Biofuels 2014, 7, 144. 47. Wallenius, J.; ViikiläM; Survase, S.; et al. Constraint-based Genome-scale Metabolic Modeling of Clostridium acetobutylicum Behavior in an Immobilized Column. Bioresour. Technol. 2013, 142, 603–610. 48. Wallenius, J.; Maaheimo, H.; Eerikäinen, T. Carbon 13-Metabolic Flux Analysis Derived Constraint-based Metabolic Modelling of Clostridium acetobutylicum in Stressed Chemostat Conditions. Bioresour. Technol. 2016, 219, 378–386. 49. Xue, C.; Zhao, J. B.; Lu, C. C.; et al. High-titer n-butanol Production by Clostridium acetobutylicum JB200 in Fed-batch Fermentation with Intermittent Gas Stripping. Biotechnol. Bioeng. 2012, 109, 2746–2756. 50. Jang, Y. S.; Malaviya, A.; Lee, S. Y. Acetone-butanol-ethanol Production with High Productivity Using Clostridum acetobutylicum BKM19. Biotechnol. Bioeng. 2013, 110, 1646–1653. 51. Xu, M. M.; Zhao, J. B.; Yu, L.; et al. Engineering Clostridium acetobutylicum with a Histidine Kinase Knockout for Enhanced n-butanol Tolerance and Production. Appl. Microbiol. Biotechnol. 2015, 99, 1011–1022. 52. Xue, C.; Zhao, J. B.; Liu, F. F.; et al. Two-stage in situ Gas Stripping for Enhanced Butanol Fermentation and Energy-saving Product Recovery. Bioresour. Technol. 2013, 135, 396–402. 53. Xue, C.; Liu, F. F.; Xu, M. M.; et al. A Novel in Situ Gas Stripping-pervaporation Process Integrated with Acetone-butanol-ethanol Fermentation for Hyper n-butanol Production. Biotechnol. Bioeng. 2016, 113, 120–129. 54. Outram, V.; Lalander, C. A.; Lee, J. G. M.; et al. Applied in situ Product Recovery in ABE Fermentation. Biotechnol. Prog. 2017, 33, 563–579.

100

Biofuels and Bioenergy: Acetone and Butanol

55. Abdehagh, N.; Tezel, F. H.; Thibault, J. Separation Techniques in Butanol Production: Challenges and Developments. Biomass Bioenergy 2014, 60, 222–246. 56. Xue, C.; Liu, F. F.; Xu, M. M.; et al. Butanol Production in Acetone-butanol-ethanol Fermentation with in situ Product Recovery by Adsorption. Bioresour. Technol. 2016, 219, 158–168. 57. Yang, D. C.; Tian, D. X.; Xue, C.; et al. Tuned Fabrication of the Aligned and Opened CNT Membrane with Exceptionally High Permeability and Selectivity for Bioalcohol Recovery. Nano Lett. 2018. https://doi.org/10.1021/ACS.NANOLETT.8B0. 58. Zhu, C.; Chen, L. J.; Xue, C.; Bai, F. W. A Novel Close-circulating Vapor Stripping-vapor Permeation Technique for Boosting Biobutanol Production and Recovery. Biotechnol. Biofuels 2018, 11, 128. 59. Ezeji, T.; Qureshi, N.; Blaschek, H. Butanol Fermentation Research: Upstream and Downstream Manipulations. Chem. Rec. 2004, 4, 305–314.

3.08

Long-Chain Liquid Biofuels

Sana Malik, Department of Bioinformatics & Biotechnology, Government College University Faisalabad, Faisalabad, Pakistan Chen-Guang Liu and Xin-Qing Zhao, State Key Laboratory of Microbial Metabolism, and School of Life Sciences & Biotechnology, Shanghai Jiao Tong University, Shanghai, China Muhammad Aamer Mehmood, Department of Bioinformatics & Biotechnology, Government College University Faisalabad, Faisalabad, Pakistan and State Key Laboratory of Microbial Metabolism, and School of Life Sciences & Biotechnology, Shanghai Jiao Tong University, Shanghai, China © 2019 Elsevier B.V. All rights reserved.

3.08.1 3.08.2 3.08.2.1 3.08.2.2 3.08.2.3 3.08.2.4 3.08.3 3.08.3.1 3.08.3.2 3.08.4 References

3.08.1

Introduction Synthesis of Long-Chain Liquid Biofuels Using Microbial Cell Factories High-Carbon Alcohols Synthesis of Hydrocarbons Fatty Acid–Derived Liquid Biofuels Isoprenoid-Based Liquid Biofuels Challenges and Opportunities in the Commercial Production of Long-Chain Liquid Biofuels Cell Membrane Engineering to Improve Tolerance Exportin-Mediated Secretion to Counter Cellular Toxicity Conclusions and Prospects

101 101 101 102 103 103 104 104 105 106 107

Introduction

The overconsumption of fossil fuels is depleting their reserves quickly and in the meantime increases the global carbon emissions. In order to sustain transportation and reduce carbon emissions, it is essential to produce environmentally friendly transportation fuels on a sustainable basis. Traditional biofuels are produced from food crops such as sugarcane and grains, and vegetable oils extracted from palm fruits, rapeseed, soybean, and jatropha. These biofuels have low energy densities and cannot meet the energy requirements particularly of the aviation industry, which require long-chain fuels with high energy densities. These fuels can be produced by tuning the metabolic pathways of microbes through a combination of synthetic and system biology tools (Table 1). Some microorganisms naturally produce certain long chain liquid fuels, for example, Clostridium species produce butanol and isopropanol.1 However, slow cell growth rates, lower product yields, difficulty in genetic manipulations, less tolerance to fuel toxicity and poor ability to consume carbon sources are some of the major challenges which hinder their commercial production. Therefore, model organisms because of their amenability by genetic engineering approaches have become the focus of research to produce long-chain fuels. However, the production and extraction of these biofuels on commercially feasible levels are still not possible. In this article, we discuss major challenges, which hinder their large-scale production and review the research efforts deployed to address them.

3.08.2

Synthesis of Long-Chain Liquid Biofuels Using Microbial Cell Factories

Advances in recombinant DNA tools have revolutionized the production methodology of many recombinant proteins of industrial importance, particularly in the pharmaceutical and food industry. Revolutions in biotechnology are also expected to transform the biofuel production landscape entirely by increasing the synthesis of industrially viable compounds using microbial platforms. The world is in the dire need of developing robust, advanced and cost-effective biorefinery-based systems for biofuel production through the innovations in metabolic engineering, gene manipulation, protein engineering, and pathway portability using the system and synthetic biology and omics tools.

3.08.2.1

High-Carbon Alcohols

In comparison to ethanol, butanol is generally regarded as the potential replacement of gasoline because of its extraordinary properties like less hygroscopicity and volatility as well as higher energy density. In nature, it is produced through Acetone-Butanol-Ethanol (ABE) pathway by Clostridia spp., but this pathway is not economically feasible on an industrial scale. Exceeding limits of butanol with the cell causes cellular toxicity and consequently yield loss. The pathway was therefore replicated in Saccharomyces cerevisiae and Escherichia coli using synthetic biology tools to make butanol a competitive liquid transportation biofuel.2 Different approaches have been utilized to enhance the yield of butanol in yeast, by overexpressing the mitochondrial and cytosolic pathways and associated enzymes. Initial attempts for the heterologous production of butanol in S. cerevisiae and E. coli met with a low yield of 2.5 mg L
1 and 500 mg L
1, respectively.3 The use of a mutant strain of S. cerevisiae defective in citrate synthase and malate synthase increased the yield of butanol

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00485-7

101

102

Long-Chain Liquid Biofuels

Table 1

Metabolic pathway engineering to produce long-chain liquid fuels in various hosts

Potential biofuel

Native organism

Targeted pathway/genes

Source organism

Heterologous host

Titer

References

Butanol

Clostridia spp.

Butanol biosynthetic pathway and Isozymes

Ralstonia eutropha, Clostridium beijerinckii and E. coli Salmonella enterica

S. cerevisiae

2.5 mg L
1

3

S. cerevisiae

16.5 mg L
1

4

E. coli

S. cerevisiae

120 mg L
1

6

Treponema denticola

E. coli

30g L
1

5

B. brevis, L. lactis and Enterobacter sp. S. elongatus

Synechocystis sp.

0.43 g L
1

9

E. coli

25 mg L
1

11

P. marinus

E. coli

220  3 mg L
1

13

M. hydrocarbonoclasticus Indigenous E. coli and A. baylyi

S. cerevisiae S. cerevisiae Z. mobilis

6.3 mg L
1 5 mg mL
1 11 g L
1

18 19 17

P. frutescens and C. limon

S. cerevisiae

31

E. coli S. enterica

S. cerevisiae S. cerevisiae

0.12 mg L
1 0.49 mg L
1 900 mg L
1 120 mg L
1

Limonene

C. limon

ACS, ALD6, ERG10, ADH2 and cytosolic acetyl CoA Acetyl-CoA pathway and PDH-bypass Modified 1-butanol pathway Catabolic pathway (aldc als, AR) Fatty acid biosynthesis pathway CoA-dependent butyraldehyde pathway Wax ester synthase genes FAS1, FAS2 and ACC1 FAEE biosynthesis pathway and Pdc, Adh, atfA genes limonene synthase genes

Bisabolane Amorphadiene

Origanum vulgare Artemisia annua

bisabolane synthase genes Mevalonate pathway and PDH bypass

2,3-BDO

Klebsiella pneumoniae Cyanobacteria

C13–C17 alkanes Propane

Cryptococcus albidus Oleaginous organisms

FAEE Microdiesel

33 34

precursor molecules to 16.5 mg L
1.4 Introduction of a modified 1-butanol pathway in E. coli along with genes encoding Ter-enzyme and associated enzymes NADH and acetyl-CoA resulted in 30 g L
1 titer.5 Replication of the acetyl-CoA pathway along with the E. coli pyruvate dehydrogenase bypass in S. cerevisiae enhanced 1-butanol yield to 120 mg L
1. 2,3-butanediol (2,3-BDO) is a precursor molecule of many industrial compounds including ketones, 1,3-butadiene, and gbutyrolactone (GBL), which is, therefore, a focus of intensive research. Heterologous expression of acetolactate decarboxylase (aldC), butanediol dehydrogenase (bdh) and acetolactate synthase (als) from Enterobacter cloacae and Bacillus licheniformis into various hosts resulted in the production of 10 g L
1 of 2,3-butanediol from pyruvate. A further increase in the titer was achieved by the overexpression of the glycolysis catalyzing specific enzymes.7 Commercial-scale production of 2,3-BDO was also attained with the productivity of 2.2 g L
1 h.8 Many cyanobacterial species like Synechocystis have also been used for heterologous production of 2,3-BDO. With the introduction of three heterologous genes namely acetolactate decarboxylase (aldc) from Brevibacillus brevis, acetolactate synthase (als) from Leuconostoc lactis and acetoin reductase (AR) from Enterobacter sp., the 2,3-BDO was synthesized by Synechocystis sp. PCC6803 at 0.43 g L
1.9 Later, by utilizing the sunlight and atmospheric CO2 with a different combination of genes, the production level was improved to 2.38 g L
1.10 These studies have highlighted the role of key genes, which can be used for the heterologous production of high-carbon alcohols using selected microbial platforms. However, novel biosynthesis pathways for 2,3-BDO are yet to be designed to achieve required yields and titers with robust strains to meet commercial criteria.

3.08.2.2

Synthesis of Hydrocarbons

Long chain alkanes such as tridecane, pentadecane, and heptadecane have shown to be the most competent substitutes for diesel, kerosene oil and gasoline as transportation fuels. Most commonly used pathways for the synthesis of hydrocarbons include the valine pathway, the fatty acid synthesis pathway, and the butanol pathway. The initial study for the production of the alkanes C13–C17 in E. coli by expressing the cyanobacterial genes encoding a fatty acyl ACP/CoA reductase (FAR) and a fatty aldehyde deformylating oxygenase (FADO) reported the expression of 25 mg L
1.11 Expression of fatty acid pathway genes from various other sources improved their production to 580 mg L
1.12 Propane is also an important candidate for biofuel research due to lower CO2 emission, high energy density and compatibility with the existing petroleum infrastructure. To convert butanol to propane, a combination of four genetic pathways along with aldehyde deformylating oxygenase gene (ADO) from Prochlorococcus marinus were co-expressed in E. coli, and out of four combinations of genes, the atoB-TPC7-ADO was shown to be the most efficient route producing 220  3 mg L
1 of propane.13 In the conversion of isobutyraldehyde into isobutanol, aldehyde reductase (AR) blocks the synthesis of propane, so by deleting 13 AR-coding genes in E. coli, the titer was increased to 267 mg L
1.14 S. cerevisiae has also been employed to synthesize long-chain alkanes by knocking down the Hfd1 gene encoding hexadecenal dehydrogenase as it halts the

Long-Chain Liquid Biofuels

103

alkane synthesis. Overexpression and regulation of the redox system along with the deletion of Hfd1 gene has proven to be a considerable modification which increases the overall titer of alkanes in yeast.15 The photosynthetic organisms, cyanobacteria have also been studied to produce hydrocarbons. Synechococcus elongatus is a model organism which was used for the first time to synthesize isobutyraldehyde, from which a variety of hydrocarbons can be derived via biological/chemical conversion. Introduction of a KDC-dependent pathway in S. elongatus and installing four pathway genes from different organisms resulted in a titer of 723 mg L
1 in 12 days, which was further improved by overexpressing the rbcL and rbcLS gene encoding ribulose 1,5-bisphosphate carboxylase/oxygenase (Rubisco), enhancing the titer up to 1.1 g L
1 in 8 days.16 The use of E. coli, S. cerevisiae, and S. elongatus to produce long-chain hydrocarbon fuels highlights their potentials to deliver cost-effective future biofuels. It is worth noticing that use of non-photosynthetic system needs careful evaluation before industrial production due to environmental concerns from the carbon emission from burning of the alkanes needed for their production. Therefore, using photosynthetic systems such as cyanobacteria and microalgae seem more promising due to their carbon-neutral nature.

3.08.2.3

Fatty Acid–Derived Liquid Biofuels

Fatty acids are precursor molecules of a variety of industrially important compounds including hydrocarbons, fatty alcohols, fatty acid methyl esters (FAMEs), fatty acid ethyl esters (FAEEs). These molecules can be used as transportation fuels due to their hydrophobicity, biodegradability and non-toxic nature. Free fatty acids themselves cannot be used as liquid fuels for internal combustion engines due to the presence of carboxyl moiety. Attempts have been made to synthesize the fatty acid derived molecules which can be used to produce biodiesel. Bacterial acetyltransferase genes have been expressed in S. cerevisiae to synthesize the steryl ester, and later four different types of wax ester synthases and diacylglycerol acetyltransferase (WS/DGAT) were expressed to synthesize FAEEs.17 Nevertheless the low expression of wax ester synthase genes, the study highlighted the possibility of engineering WS/ DGAT to synthesize various lipids in eukaryotes. As a result, extensive efforts have been made to boost the endogenous activity of wax ester synthases in yeast cell factories. Five genes from different microbial origins encoding wax ester synthases were introduced into S. cerevisiae to compare their abilities for FAEEs synthesis, and all of them performed well, but the wax ester synthase from Marinobacter hydrocarbonoclasticus was the best candidate with an overall concentration of 6.3 mg L
1,18 which indicates the importance of genetic bioprospecting to select the best candidate genes to design a robust pathway. Similarly, overexpression of fatty acid metabolism genes namely fatty acid synthase 1–2 (fas1, fas2) and acetyl-CoA carboxylase (acc1) in S. cerevisiae improved the lipid accumulation by four-fold compared to control, resulting in a 400 mg L
1 of free fatty acids (FFAs) and 5 mg mL
1 of FAEEs.19 To further augment the synthesis levels of triacylglycerol (TAG), another oleaginous yeast strain Yarrowia lipolytica was engineered with a translation elongation factor 1-a promoter to overexpress the diacylglycerol acyltransferase (dga) in cells, resultantly the lipid content was increased to 33.8 % on dry cell weight basis, which was further enhanced to 61.75 % by expressing another gene encoding acetyl-CoA carboxylase.20 To boost the lipid synthesis in Y. lipolytica, overexpression of three different genes diacylglyceride acyl-transferase (dga1), acetyl-CoA carboxylase (acc1) and mammalian delta-9 stearoyl-CoA desaturase (scd) resulted in a lipid production of approximately 55 g L
1, four-fold higher compared to that produced by the native strain.21 Fatty acid derived biofuels have also been produced using non-yeast microbial hosts including E. coli and cyanobacteria. In E. coli, the FAEE biosynthesis pathway was introduced and pyruvate decarboxylase (pdc), alcohol dehydrogenase (adh) encoding genes from Z. mobilis, and atfA gene from Acinetobacter baylyi were overexpressed, which resulted the synthesis of 1.2 g L
1 of FAAE (biodiesel) via fed-batch fermentation process, while optimization of the process variables and using glycerol as a substrate further improved this titer to 11 g L
1.22 FAEEs have also been produced in E. coli by utilizing non-carbon sources and linking the atfA gene expression and ethanol production with a thioesterase enzyme which cleaves acyl-ACP and a fatty acid CoA ligase. Optimization of the fed-batch process and deletion of fadE gene increased the production titer, but it was not enough to commercialize the FAEE because of the ethanol toxicity. To counter this problem, a transcription factor fadR was expressed which suppressed the expression of genes involved in ethanol and acetate synthesis, this modification increased the titer by 7.5-fold, and further process optimizations and multigene expression of fatty acid synthesis pathway in E. coli improved the titer up to 8.6 g L
1.23,24 Production of fatty acid-derived biofuels has been reported in cyanobacteria too albeit at much low levels. Overexpression of genes of the fatty acid biosynthesis pathway in Synechocystis sp. and deletion of some of the proteins present in S layer increased productivity to 13312 mg L
1 day.25 Heterologous expression of Rubisco encoding genes increased the FFA production up to 3-fold, with an overall concentration of 130 mg L
1.26 Down-regulating the expression of a transcriptional regulator encoding gene cyAbrB2 and aas gene in Synechococcus sp. coupled with the expression of UcTE gene in a mutant strain capable of producing and secreting the FFA resulted in doubling its production.27 These studies show the potential of non-yeast hosts to produce and secrete fatty acid-derived biofuels. Further optimization of the metabolic pathways leading to the FFA biosynthesis can be employed to increase the production levels.

3.08.2.4

Isoprenoid-Based Liquid Biofuels

Isoprenoids are branched-chain unsaturated hydrocarbons found in plants, bacteria, and microalgae, and are considered as a good quality transportation fuel. These compounds have a higher octane number and do not suffer premature ignition in combustion engines. Their naturally low abundance is the major challenge for their commercialization. So a variety of microbial hosts have been investigated for their potential to overexpress isoprenoids in higher quantities. Terpenoids based transportation fuels can be synthesized with the different arrangements of mainly two building blocks isopentyl diphosphate (IPP) and dimethylallyl

104

Long-Chain Liquid Biofuels

diphosphate (DMAPP) isomer of IPP, and both of these play a key role in the synthesis of isoprenoid compounds. There are two metabolic pathways involved in the synthesis of isoprenoids, deoxy-D-xylulose 5-phosphate pathway (DXP) or mevalonate pathway (MEV). Microalgae accumulate triterpenes in significant quantities like plants and bacteria. In comparison to other species, Botryococcus braunii, the chief source of triterpenes, is capable of accumulating hydrocarbon compounds as high as 86% of its dry cell weight.28 Triterpenoids are constantly synthesized and secreted in the form of botryococcene and squalene in B. braunii, by diverting up to 35 % of carbon assimilates to their synthesis.28 The FPP synthesis in yeast is carried out via the mevalonate (MVA) pathway, hence metabolic engineering can be exploited in yeast cells to synthesize terpenes-based biofuels with higher productivities. Overexpression of MVA-homologues EfmvaE and EfmvaS from Enterococcus faecalis into yeast has been shown to convert Acetyl-CoA into mevalonate at much greater efficiency, and this mevalonate was converted to IPP, which was isomerized to DMAPP.29 Farnesyl diphosphate synthase enzyme performs the condensation of IPP with DMAPP to geranyl diphosphate, and later one more IPP undergoes condensation to synthesize FPP, which is a key precursor molecule in the synthesis of squalene via ssl1 and ssl2 genes, or the synthesis of botryococcene via ssl1 and ssl3 genes.30 Commercial synthesis of these molecules is not straightforward in heterologous microbial platforms due to their cytotoxic nature, low titers and difficult recovery. These challenges are expected to be overcome by a combination of synthetic biology and computational biology tools followed by optimization studies. Among the isoprenoids, limonene is considered the best choice of fuel and a potential replacement of jet fuel due to its unique properties including a branched-chain structure, low freezing point and water immiscibility. Limonene can also be served as a flavoring agent and found in many citrus fruits naturally albeit in low quantities. To test the heterologous production of limonene in S. cerevisiae, a mutant strain defective in ERG20 was co-transformed with a gene encoding limonene synthase from Perilla frutescens and (þ) limonene from Citrus limon, and although limonene could be produced in significant quantities, yeast growth was compromised.31 Similarly, sesquiterpenoids with isoprene units in their structures have also been used as suitable transportation fuels. Among the sesquiterpenoids, bisabolene and farnesene have received significant attention. Farnesene can also be used in the synthesis of flavoring agents and fragrances at a commercial scale. It is also considered as a potential replacement of the jet fuel and diesel because of its remarkable properties like less volatility and high energy content. FPP biosynthesis pathway is used in the synthesis of farnesene in microbial hosts. To achieve the enhanced production of farnesol/farnesene through FPP synthesis in E. coli and S. cerevisiae, the mevalonate pathway was introduced in these microbial hosts, and the synthesis rate of prenyl alcohols including geranylgeraniol, farnesol and nerolidol were increased by overexpressing the gene Hmg encoding hydroxymethylglutarylCoA reductase involved in both mevalonate pathway and prenyl diphosphate pathway in S. cerevisiae.32 Introduction of six bisabolane synthases in S. cerevisiae DHmg1 mutant strain resulted in the higher production of bisabolane, which was then hydrogenized to synthesize bisabolene - a powerful replacement of D2 diesel.33 Amorphadiene is another type of sesquiterpenoids with a higher functional density which makes it a good biofuel candidate. The overexpression of indigenous Ald6 gene and the heterologous expression of Asc1 gene of Salmonella enterica, led to higher production of amorphadiene in the S. cerevisiae, and the level of mevalonate pathway precursor acetyl-CoA, which carries out the synthesis of sesquiterpenes, elevated with the expression of these genes in yeast cells.34 Another strategy has been adopted to increase the flux of this precursor molecule acetyl-CoA by heterologously expressing the phosphoketolase pathway of Aspergillus nidulans in S. cerevisiae to synthesize the various industrially important compounds.

3.08.3

Challenges and Opportunities in the Commercial Production of Long-Chain Liquid Biofuels

Use of microbial platforms for the production of biofuels by re-routing their metabolic pathways to produce long-chain hydrocarbons is not straightforward as it seems. Despite several advancements in the genetic engineering approaches, not a single advanced biofuel molecule has been produced on a commercial scale because of the several challenges like increased cellular toxicity, microbial contamination, costly extraction of fuel molecules from media and un-optimized process conditions. To successfully commercialize the biofuels on an industrial scale, these issues demand a reliable solution. One way to achieve this is by enhancing the carbon assimilation pathways of host organisms and then re-directing the metabolic flux toward the synthesis of desired molecules via natural or synthetic pathways. To further increase the metabolic flux, a balance is required among the expression of these pathways and the enzymes involved. The metabolic pathways present naturally in the organisms can be remodeled or carefully optimized to achieve higher production of desired products. Excessive use of antibiotics to control microbial contamination have made the competing microorganisms antibiotic-resistant, which serves as another blockade in the commercialization of biofuels. So, the metabolic pathways of microorganisms can be designed to enable them to absorb required nutrients present in media in the form of xenobiotics, which will outcompete the contaminating microbes. A schematic approach is shown in Fig. 1.

3.08.3.1

Cell Membrane Engineering to Improve Tolerance

An efficient and operative strategy which can be used to increase cellular tolerance toward biofuel molecules is the cell membrane engineering. Membrane composition can be remodeled by altering the fatty acid chain length and saturation, which will help the cell to increase tolerance against external stresses. Membrane integrity is severely damaged by the overproduction of free fatty acids, which are the reason for cellular membrane stress. The lipid composition of the cell membrane can be modulated with the expression of a gene encoding acyl-ACP thioesterase of Geobacillus, to perform the hydrolysis of unsaturated medium chain length FFAs present in high amounts in the cell. This modification has greatly reduced the free fatty acid content in cell and enhanced the

Long-Chain Liquid Biofuels

105

Figure 1 Schematic flow-chart of steps involved in the synthesis and production of long-chain biofuels using synthetic biology, genetic engineering and process optimization.

tolerance toward FFAs in E. coli.35 By deleting the aas gene responsible for the incorporation of medium chain fatty acids into membrane phospholipids, thereby increasing the chain length, improved the tolerance and fatty acid production by 20% in E. coli.36 By using an evolutionary approach, an E. coli strain MG1655 was evolved with dynamic membrane composition for enhanced tolerance against fatty acids, carboxylic acids and butanol.37 Membrane engineering has gained popularity as a potential target to enhance the robustness of microbial hosts. However, there are some limitations while using this approach as we have very little information about the impact of changes in membrane composition except for membrane integrity. For this reason, we must find other ways to improve the tolerance against biofuel molecules by making their recovery easy from media by exporting them extracellularly.

3.08.3.2

Exportin-Mediated Secretion to Counter Cellular Toxicity

Another reliable and practical approach for reducing cytotoxicity and increasing the biofuel production is the heterologous expression of a special type of proteins known as exportins which help in the export of fuel molecules from the host cell constantly, thereby reducing the chances of cytotoxicity caused by the accumulation of fuel molecules.38 Usually, the molecules are exported through simple diffusion due to the concentration gradient, but this process is not as much direct, because specially encoded transporters usually control the movement of almost all types of molecules across the cell membrane. So, these transporters could be the potential targets to design transporter-based cell secretion systems, which would serve a dual purpose; reduced toxicity due to continuous secretion of the biofuel molecules and their easier recovery. To date, many transporter proteins have been identified involved in the transport (import/export) of various molecules across the cell membrane, so choosing the best candidate for the secretion of a specific biofuel is a task for which a deep understanding is required to explore the cell export mechanisms and the evaluation of different transporters for their potential to export biofuel molecules. Among transporters, the multidrug transporters (MDR) belong to a major class of small efflux pumps which are responsible for drug resistance in bacteria, fungi, yeast, mammals, and parasites. As such, they provide a good model for studying how these organisms export drug molecules. The detailed knowledge may help designing strategies leading toward reduced toxicity and improved recovery of the fuels. There are basically two classes of MDR transporters based on their source of energy: 1) secondary transporters which perform their co-transporter activity relying on proton gradients 2) adenosine triphosphate (ATP) binding cassette (ABC) transporters which depend upon ATP hydrolysis to carry out transporter activity to translocate molecules across the cell membrane. These transporters have been assigned to four superfamilies: a) drug/metabolite transporter superfamily, b) multidrug/oligosaccharide-lipid/polysaccharide lippase family, c) resistance-nodulation division family (RND), and d) greatly diverse major facilitator superfamily.39 ABC transporters are ubiquitous proteins in nature, present in all five kingdoms of life starting from simple bacteria to complex animals. They comprise one of the largest and oldest families of transporter proteins which are involved in the usual export/import of various drugs, xenobiotics, metabolites, lipids, polysaccharides, amino acids, vitamins and hormones across the cell membrane. The ABC transporters owing to their poly-specific nature have also been found to export the wax to the cuticle layer of plant cells, which reflects their potential to export greasy molecules like biofuels. Both types of ABC transporters, exporters and importers, are encoded by prokaryotic genomes while eukaryotic genome code only for exporters.40 Resistance-nodulation-division transporter family also known as RND transporters have been used first time to secrete the biofuel molecules from the host cells. These efflux pumps belong to the Gram-negative type of bacteria like E. coli. RND pumps helped to increase the synthesis of limonene by 1.5-fold in bacteria.41 Owing to its large structure and tri-partite structural composition,

106

Long-Chain Liquid Biofuels

RND transporters are not favorable in the secretion of other biofuel molecules synthesized using a wide variety of hosts including yeast, cyanobacteria, and microalgae.41 Another type of transporter AcrB belongs to AcrAB-TolC transporter system, which has been engineered to enhance the export efficiency of biofuels for their easier recovery from media via directed evolution method, and an increase of 47% and 400% was observed in the export of linear and cyclic molecules n-octane and a-pinene, respectively.42 Due to their ubiquitous nature, the ABC transporters can be heterologously expressed in various species. For instance, heterologous expression of a bacterial ABC transporter (LmrA) in human cells could successfully export drug.43 The export of arenes in Pseudomonas putida was found to increase with the expression of a SrpABC efflux pump belonging to RND transporter family.44 In S. cerevisiae the alkanes export was facilitated with the expression of four types of ABC efflux pumps namely, Snq2p, Pdr5p, Pdr15p, and Yor1p, which also helped to improve the cell tolerance against biofuel molecules, and out of four efflux pumps, two pumps Snq2p and Pdr5p performed more efficiently than others in the export of alkanes from S. cerevisiae.45 Heterologous expression of two transporters Abc2 and Abc3 of Y. lipolytica helped in keeping the levels of alkanes low in the S. cerevisiae which surprisingly increased the cell tolerance against decane and undecane.45 A different approach was used to counter the cellular toxicity issue by genetically boosting the two membrane pumps, proton pump (Pma1) and potassium pump (Trk1), and up-regulation of these pumps in S. cerevisiae elevated the ethanol tolerance and production levels.46 Another study was based on achieving the increased ethanol production and enhanced cell tolerance against many stresses in S. cerevisiae with the deletion of Ady2 acetate transporter.47 By overexpressing a natural ABC transporter Pdr12 in S. cerevisiae the export of short branched-chain fatty acids (BCFA) was increased.48 It was observed that the introduction of an ABC efflux pump MdlB in E. coli increased the isopentenol production levels up to 12%.49 Directed evolution was performed to generate genetic variants of AcrB efflux pumps to evaluate their potential for improving biofuel tolerance in E. coli, and among these variants, AcrAB-TolC transporter has shown a promising potential in increasing tolerance in host cells against n-octane, isobutanol and olefins.42,50 To increase the tolerance against a-pinene, four different types of transporters were overexpressed, and their combined effects on a -pinene tolerance were studied.51 It was observed that from a panel of 19 transporters, only two transporters namely S. enterica ser. Typhimurium MsbA (St-MsbA) and E. coli MsbA (Eco-MsbA) effectively secreted zeaxanthin and canthaxanthin in the growth media.52 Similarly, a mammalian transporter Fatp1 (a free fatty acid importer) started exporting alcohols after exchanging the functional domain involved in fatty acid export between Fatp1 and another transporter Fatp4, and expression of this Fatp1 in S. cerevisiae improved the cell growth viability via acyl-CoA synthase activity and by continuously secreting fatty alcohols from host cells.53 Careful optimization of a specific range for transporter expression in host cells is an essential requirement, as overexpression of transporters can overburden the translational machinery of host cell system, which may lead to halted growth and ultimately lower productivity of biofuels. To address this issue, an expression control system was designed by using a native stress-regulatory promoter of E. coli, PgntK which could prompt negative feedback to control the expression levels of an engineered transporter, and expression of an efflux pump AcrBv2 (a variant of AcrB) enhanced the butanol tolerance and production titers up to 35% in E. coli, driven by the PgntK promoter.54 Fatty acid synthesis pathway was introduced in S. cerevisiae to synthesize 1-alkenes along with a Fatp1 transporter for their secretion, and the dynamic regulation approach was also used to control the expression of this transporter gene in S. cerevisiae, which resulted in an improved alkenes production by countering the cytotoxicity issue.55 Transporters have become the powerful tools in metabolic pathway engineering which confer tolerance to the host cells, consequently improving the biofuel production but their expression sometimes interferes with the membrane functionality of the host cells. The selection or screening of a suitable transporter to export a specific biofuel is the foremost and a crucial factor in transporter engineering, followed by the evaluation of the capability of important transporters to secrete biofuels. Based on this evaluation some transporters are being selected and later their properties and optimum expression range can be modified using directed evolution approach. Owing to their hydrophobic nature, isoprenoids are highly toxic as they interact with mitochondrial membrane, and can damage the membrane integrity. Accumulation of limonene up to a concentration of 0.5–0.8 g L
1 in engineered yeast strains negatively affects the growth of host cells. To counter this issue a library of 43 bacterial efflux pumps was created via the screening of bacterial genome sequences to enhance the tolerance against terpenoids.56 A variant of acetyl-CoA carboxylase, Acc1S1157A was engineered in S. cerevisiae to increase oleic acid levels, which ultimately ameliorated the toxicity caused by octanoic acid and improved its production titer.57 As a replacement for heterologous expression, the overexpression of indigenous transporters in E. coli enhanced the yield of short and medium chain alcohols and their tolerance, and the potential of 44 indigenous transporters of E. coli was evaluated, in which only 4 transporters namely MdtJ, Bcr, and MdtH and YdeA showed the ability to continuously export the monohydroxy, dihydroxy, and trihydroxy alcohols, thereby improving the tolerance by 35%–50% against them.58 The most direct and efficient way to pump out the biofuel molecules to prevent their intracellular accumulation which leads to cellular toxicity is the use of efflux pumps, however, this strategy is not that much easy and poses its own challenges as it requires precise regulation of efflux pumps expression in host cells. Fig. 2 shows a cartoon of an engineered cell capable of synthesis and secretion of long-chain biofuels via heterologously installed synthetic pathways and exportin.

3.08.4

Conclusions and Prospects

Long chain liquid biofuels have potential to replace conventional fossil fuels, but the cost to be incurred on their overall production makes their use impractical due to extremely low titer and difficult recovery from their natural hosts. Therefore, intensive research is required to remodel the metabolic networks of microorganisms with careful and controlled optimization of different parameters to

Long-Chain Liquid Biofuels

Figure 2

107

A schematic diagram of an engineered cell capable of heterologous synthesis and secretion of long-chain liquid biofuels.

develop robust platforms for the production of long-chain liquid fuels. Here robustness refers to the capability of microbial cell factories to maintain a stable phenotype and to withstand the harsh industrial conditions including high pH and temperature, osmotic pressure, substrate/metabolite toxicity, oxidative stress and inhibitor molecules. The efficacy of natural microbes can be enhanced through adaptive evolution, metabolic flux regulation and rewiring of natural pathways to increase the production of desired biofuels on an industrial scale. Moreover, it would be required to engineer and establish photosynthetic biofuel secreting systems to ensure their carbon-neutral nature, low-cost and easier recovery.

References 1. Singh, V.; Mani, I.; Chaudhary, D. K.; Dhar, P. K. Metabolic Engineering of Biosynthetic Pathway for Production of Renewable Biofuels. Appl. Biochem. Biotechnol. 2014, 172 (3), 1158–1171. 2. Jang, Y. S.; Park, J. M.; Choi, S.; Choi, Y. J.; Cho, J. H.; Lee, S. Y. Engineering of Microorganisms for the Production of Biofuels and Perspectives Based on Systems Metabolic Engineering Approaches. Biotechnol. Adv. 2012, 30 (5), 989–1000. 3. Steen, E. J.; Chan, R.; Prasad, N.; Myers, S.; Petzold, C. J.; Redding, A.; Ouellet, M.; Keasling, J. D. Metabolic Engineering of Saccharomyces cerevisiae for the Production of nButanol. Microb. Cell Factories 2008, 7 (36). 4. Krivoruchko, A.; Serrano-Amatriain, C.; Chen, Y.; Siewers, V.; Nielsen, J. Improving Biobutanol Production in Engineered Saccharomyces cerevisiae by Manipulation of AcetylCoA Metabolism. J. Ind. Microbiol. Biotechnol. 2013, 40 (9), 1051–1056. 5. Shen, C. R.; Lan, E. I.; Dekishima, Y.; Baez, A.; Cho, K. M.; Liao, J. C. Driving Forces Enable High-Titer Anaerobic 1-Butanol Synthesis in Escherichia coli. Appl. Environ. Microbiol. 2011, 77 (9), 2905–2915. 6. Lian, J.; Si, T.; Nair, N. U.; Zhao, H. Design and Construction of Acetyl-CoA Overproducing Saccharomyces cerevisiae Strains. Metab. Eng. 2014, 24, 139–149. 7. Xie, N. Z.; Chen, X. R.; Wang, Q. Y.; Chen, D.; Du, Q. S.; Huang, R. B. Microbial Routes to (2R, 3R)-2, 3-butanediol: Recent Advances and Future Prospects. Curr. Top. Med. Chem. 2017, 17 (21), 2433–2439. 8. Park, J. M.; Rathnasingh, C.; Song, H. Metabolic Engineering of Klebsiella pneumoniae Based on in Silico Analysis and its Pilot-Scale Application for 1, 3-propanediol and 2, 3butanediol Co-production. J. Ind. Microbiol. Biotechnol. 2017, 44 (3), 431–441. 9. Savakis, P. E.; Angermayr, S. A.; Hellingwerf, K. J. Synthesis of 2, 3-butanediol by Synechocystis sp. PCC6803 via Heterologous Expression of a Catabolic Pathway from Lactic Acid-And Enterobacteria. Metab. Eng. 2013, 20, 121–130. 10. Oliver, N. J.; Rabinovitch-Deere, C. A.; Carroll, A. L.; Nozzi, N. E.; Case, A. E.; Atsumi, S. Cyanobacterial Metabolic Engineering for Biofuel and Chemical Production. Curr. Opin. Chem. Biol. 2016, 35, 43–50. 11. Schirmer, A.; Rude, M. A.; Li, X.; Popova, E.; Del Cardayre, S. B. Microbial Biosynthesis of Alkanes. Science 2010, 329 (5991), 559–562. 12. Kallio, P.; Pásztor, A.; Thiel, K.; Akhtar, M. K.; Jones, P. R. An Engineered Pathway for the Biosynthesis of Renewable Propane. Nat. Commun. 2014, 5, 4731.

108

Long-Chain Liquid Biofuels

13. Menon, N.; Pásztor, A.; Menon, B. R.; Kallio, P.; Fisher, K.; Akhtar, M. K.; Leys, D.; Jones, P. R.; Scrutton, N. S. A Microbial Platform for Renewable Propane Synthesis Based on a Fermentative Butanol Pathway. Biotechnol. Biofuels 2015, 8 (1), 61–72. 14. Bentley, G. J.; Jiang, W.; Guamán, L. P.; Xiao, Y.; Zhang, F. Engineering Escherichia coli to Produce Branched-Chain Fatty Acids in High Percentages. Metab. Eng. 2016, 38, 148–158. 15. Buijs, N. A.; Zhou, Y. J.; Siewers, V.; Nielsen, J. Long-chain Alkane Production by the Yeast Saccharomyces cerevisiae. Biotechnol. Bioeng. 2015, 112 (6), 1275–1279. 16. Oliver, J. W.; Atsumi, S. Metabolic Design for Cyanobacterial Chemical Synthesis. Photosynth. Res. 2014, 120 (3), 249–261. 17. Kalscheuer, R.; Luftmann, H.; Steinbüchel, A. Synthesis of novel Lipids in Saccharomyces cerevisiae by Heterologous Expression of an Unspecific Bacterial Acyltransferase. Appl. Environ. Microbiol. 2004, 70 (12), 7119–7125. 18. Shi, S.; Valle-Rodríguez, J. O.; Khoomrung, S.; Siewers, V.; Nielsen, J. Functional Expression and Characterization of Five Wax Ester Synthases in Saccharomyces cerevisiae and Their Utility for Biodiesel Production. Biotechnol. Biofuels 2012, 5 (1), 7–17. 19. Runguphan, W.; Keasling, J. D. Metabolic Engineering of Saccharomyces cerevisiae for Production of Fatty Acid-Derived Biofuels and Chemicals. Metab. Eng. 2014, 21, 103–113. 20. Tai, M.; Stephanopoulos, G. Engineering the Push and Pull of Lipid Biosynthesis in Oleaginous Yeast Yarrowia lipolytica for Biofuel Production. Metab. Eng. 2013, 15, 1–9. 21. Qiao, K.; Abidi, S. H. I.; Liu, H.; Zhang, H.; Chakraborty, S.; Watson, N.; Ajikumar, P. K.; Stephanopoulos, G. Engineering Lipid Overproduction in the Oleaginous Yeast Yarrowia lipolytica. Metab. Eng. 2015, 29, 56–65. 22. Elbahloul, Y.; Steinbüchel, A. Pilot-scale Production of Fatty Acid Ethyl Esters by an Engineered Escherichia coli Strain Harboring the p(Microdiesel) Plasmid. Appl. Environ. Microbiol. 2010, 76 (13), 4560–4565. 23. Zhang, F.; Ouellet, M.; Batth, T. S.; Adams, P. D.; Petzold, C. J.; Mukhopadhyay, A.; Keasling, J. D. Enhancing Fatty Acid Production by the Expression of the Regulatory Transcription Factor FadR. Metab. Eng. 2012, 14 (6), 653–660. 24. Xu, P.; Gu, Q.; Wang, W.; Wong, L.; Bower, A. G.; Collins, C. H.; Koffas, M. A. Modular Optimization of Multi-Gene Pathways for Fatty Acids Production in E. coli. Nat. Commun. 2013, 4, 1409. 25. Lu, X. A Perspective: Photosynthetic Production of Fatty Acid-Based Biofuels in Genetically Engineered Cyanobacteria. Biotechnol. Adv. 2010, 28 (6), 742–746. 26. Ruffing, A. M. Improved Free Fatty Acid Production in Cyanobacteria with Synechococcus sp. PCC 7002 as Host. Front. Bioeng. Biotechnol. 2014, 2, 17. 27. Kawahara, A.; Sato, Y.; Saito, Y.; Kaneko, Y.; Takimura, Y.; Hagihara, H.; Hihara, Y. Free Fatty Acid Production in the Cyanobacterium Synechocystis sp. PCC 6803 is Enhanced by Deletion of the cyAbrB2 Transcriptional Regulator. J. Biotechnol. 2016, 220, 1–11. 28. Niehaus, T.; Kinison, S.; Okada, S.; Yeo, Y.-S.; Bell, S. A.; Cui, P.; Devarenne, T. P.; Chappell, J. Functional identification of Triterpene Methyltransferases from Botryococcus braunii Race B. J. Biol. Chem. 2012. https://doi.org/10.1074/jbc.M111.316059. 29. Peng, B.; Plan, M. R.; Chrysanthopoulos, P.; Hodson, M. P.; Nielsen, L. K.; Vickers, C. E. A Squalene Synthase Protein Degradation Method for improved Sesquiterpene Production in Saccharomyces cerevisiae. Metab. Eng. 2017, 39, 209–219. 30. Niehaus, T. D.; Okada, S.; Devarenne, T. P.; Watt, D. S.; Sviripa, V.; Chappell, J. Identification of Unique Mechanisms for Triterpene Biosynthesis in Botryococcus braunii. Proc. Natl. Acad. Sci. Unit. States Am. 2011, 108 (30), 12260–12265. 31. Jongedijk, E.; Cankar, K.; Ranzijn, J.; Van der Krol, S.; Bouwmeester, H.; Beekwilder, J. Capturing of the Monoterpene Olefin Limonene Produced in Saccharomyces cerevisiae. Yeast 2015, 32 (1), 159–171. 32. George, K. W.; Alonso-Gutierrez, J.; Keasling, J. D.; Lee, T. S. Isoprenoid Drugs, Biofuels, and ChemicalsdArtemisinin, Farnesene, and Beyond; In: Biotechnology of Isoprenoids, vol. 148; Springer, 2015. 33. Peralta-Yahya, P. P.; Ouellet, M.; Chan, R.; Mukhopadhyay, A.; Keasling, J. D.; Lee, T. S. Identification and Microbial Production of a Terpene-Based Advanced Biofuel. Nat. Commun. 2011, 2, 483. 34. Shiba, Y.; Paradise, E. M.; Kirby, J.; Ro, D.-K.; Keasling, J. D. Engineering of the Pyruvate Dehydrogenase Bypass in Saccharomyces cerevisiae for High-Level Production of isoprenoids. Metab. Eng. 2007, 9 (2), 160–168. 35. Lennen, R. M.; Pfleger, B. F. Modulating Membrane Composition Alters Free Fatty Acid Tolerance in Escherichia coli. PLoS One 2013, 8 (1), e54031. 36. Sherkhanov, S.; Korman, T. P.; Bowie, J. U. Improving the Tolerance of Escherichia coli to Medium-Chain Fatty Acid Production. Metab. Eng. 2014, 25, 1–7. 37. Royce, L. A.; Yoon, J. M.; Chen, Y.; Rickenbach, E.; Shanks, J. V.; Jarboe, L. R. Evolution for Exogenous Octanoic Acid Tolerance improves Carboxylic Acid Production and Membrane integrity. Metab. Eng. 2015, 29, 180–188. 38. Teixeira, M. C.; Godinho, C. P.; Cabrito, T. R.; Mira, N. P.; Sá-Correia, I. Increased Expression of the Yeast Multidrug Resistance ABC Transporter Pdr18 Leads to increased Ethanol Tolerance and Ethanol Production in High Gravity Alcoholic Fermentation. Microb. Cell Factories 2012, 11 (1), 98–106. 39. Putman, M.; van Veen, H. W.; Konings, W. N. Molecular Properties of Bacterial Multidrug Transporters. Microbiol. Mol. Biol. Rev. 2000, 64 (4), 672–693. 40. Karpowich, N.; Martsinkevich, O.; Millen, L.; Yuan, Y.-R.; Dai, P. L.; MacVey, K.; Thomas, P. J.; Hunt, J. F. Crystal Structures of the MJ1267 ATP Binding Cassette Reveal an induced-Fit Effect at the ATPase Active Site of an ABC Transporter. Structure 2001, 9 (7), 571–586. 41. Dunlop, M. J.; Dossani, Z. Y.; Szmidt, H. L.; Chu, H. C.; Lee, T. S.; Keasling, J. D.; Hadi, M. Z.; Mukhopadhyay, A. Engineering Microbial Biofuel Tolerance and Export Using Efflux Pumps. Mol. Syst. Biol. 2011, 7 (1), 487–493. 42. Foo, J. L.; Leong, S. S. J. Directed Evolution of an E. coli inner Membrane Transporter for improved Efflux of Biofuel Molecules. Biotechnol. Biofuels 2013, 6 (1), 81–92. 43. Gutmann, D. A.; Ward, A.; Urbatsch, I. L.; Chang, G.; van Veen, H. W. Understanding Polyspecificity of Multidrug ABC Transporters: Closing in on the Gaps in ABCB1. Trends Biochem. Sci. 2010, 35 (1), 36–42. 44. Sun, X.; Zahir, Z.; Lynch, K. H.; Dennis, J. J. An Anti-repressor, SrpR, is involved in Transcriptional Regulation of the SrpABC Solvent Tolerance Efflux Pump of Pseudomonas Putida S12. J. Bacteriol. 2011, 193, 2717–2725. 45. Chen, B.; Ling, H.; Chang, M. W. Transporter Engineering for improved Tolerance against Alkane Biofuels in Saccharomyces cerevisiae. Biotechnol. Biofuels 2013, 6, 21. https://doi.org/10.1186/1754-6834-6-21. 46. Lam, F. H.; Ghaderi, A.; Fink, G. R.; Stephanopoulos, G. Engineering Alcohol Tolerance in Yeast. Science 2014, 346 (6205), 71–75. 47. Zhang, M.; Zhang, K.; Mehmood, M. A.; Zhao, Z. K.; Bai, F.; Zhao, X. Deletion of Acetate Transporter Gene ADY2 improved Tolerance of Saccharomyces cerevisiae against Multiple Stresses and Enhanced Ethanol Production in the Presence of Acetic Acid. Bioresour. Technol. 2017, 245, 1461–1468. 48. Yu, A.-Q.; Juwono, N. K. P.; Foo, J. L.; Leong, S. S. J.; Chang, M. W. Metabolic Engineering of Saccharomyces cerevisiae for the Overproduction of Short Branched-Chain Fatty Acids. Metab. Eng. 2016, 34, 36–43. 49. Foo, J. L.; Jensen, H. M.; Dahl, R. H.; George, K.; Keasling, J. D.; Lee, T. S.; Leong, S.; Mukhopadhyay, A. Improving Microbial Biogasoline Production in Escherichia coli Using Tolerance Engineering. mBio 2014, 5 (6). e01932-14. 50. Mingardon, F.; Clement, C.; Hirano, K.; Nhan, M.; Luning, E. G.; Chanal, A.; Mukhopadhyay, A. Improving Olefin Tolerance and Production in E. coli Using native and Evolved AcrB. Biotechnol. Bioeng. 2015, 112 (5), 879–888. 51. Turner, W. J.; Dunlop, M. J. Trade-offs in improving Biofuel Tolerance Using Combinations of Efflux Pumps. ACS Synth. Biol. 2014, 4 (10), 1056–1063. 52. Doshi, R.; Nguyen, T.; Chang, G. Transporter-mediated Biofuel Secretion. Proc. Natl. Acad. Sci. USA 2013, 110, 7642–7647. 53. Hu, Y.; Zhu, Z.; Nielsen, J.; Siewers, V. Heterologous Transporter Expression for improved Fatty Alcohol Secretion in Yeast. Metab. Eng. 2018, 45, 51–58.

Long-Chain Liquid Biofuels

109

54. Boyarskiy, S.; López, S. D.; Kong, N.; Tullman-Ercek, D. Transcriptional Feedback Regulation of Efflux Protein Expression for increased Tolerance to and Production of nButanol. Metab. Eng. 2016, 33, 130–137. 55. Zhou, Y. J.; Hu, Y.; Zhu, Z.; Siewers, V.; Nielsen, J. Engineering 1-Alkene Biosynthesis and Secretion by Dynamic Regulation in Yeast. ACS Synth. Biol. 2018, 7 (2), 584–590. 56. Parveen, M.; Hasan, M. K.; Takahashi, J.; Murata, Y.; Kitagawa, E.; Kodama, O.; Iwahashi, H. Response of Saccharomyces cerevisiae to a Monoterpene: Evaluation of Antifungal Potential by DNA Microarray Analysis. J. Antimicrob. Chemother. 2004, 54 (1), 46–55. 57. Besada-Lombana, P. B.; Fernandez-Moya, R.; Fenster, J.; Da Silva, N. A. Engineering Saccharomyces cerevisiae Fatty Acid Composition for increased Tolerance to Octanoic Acid. Biotechnol. Bioeng. 2017, 114 (7), 1531–1538. 58. Zhang, Y.; Dong, R.; Zhang, M.; Gao, H. Native Efflux Pumps of Escherichia coli Responsible for Short and Medium Chain Alcohol. Biochem. Eng. J. 2018, 133, 149–156.

3.09

Biogasq

Chuanshu He and Yang Mu, CAS Key Laboratory of Urban Pollutant Conversion, University of Science and Technology of China, Hefei, China Xiaofeng Liu and Zhengying Yan, Chengdu Institute of Biology, Chinese Academy of Sciences (CIB, CAS), Chengdu, China Zhengbo Yue, Hefei University of Technology, Hefei, China © 2019 Elsevier B.V. All rights reserved. This is an update of X. Liu, Z. Yan, Z.-B. Yue, 3.10 - Biogas, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 99-114.

3.09.1 Introduction 3.09.2 Microbiological Aspects 3.09.3 Interspecies Electron Transfer in AD 3.09.3.1 Traditional Interspecies Electron Transfer 3.09.3.2 Direct Interspecies Electron Transfer 3.09.4 Key Factors in the AD Process 3.09.4.1 Inoculums 3.09.4.2 Substrate 3.09.4.3 Temperature 3.09.4.4 pH 3.09.4.5 Oxidation–Reduction Potential 3.09.4.6 Retention Time 3.09.4.7 Mixing 3.09.4.8 Nutrients 3.09.4.9 Toxicity and Inhibition 3.09.4.10 Pretreatment 3.09.5 AD Modeling 3.09.5.1 The IWA AD Model No. 1 3.09.5.2 Computational Fluid Dynamics 3.09.6 Monitoring of AD Process 3.09.6.1 Gas Chromatography 3.09.6.2 Spectroscopic Devices 3.09.7 AD Processes 3.09.7.1 Classification 3.09.7.2 Household AD Process 3.09.7.2.1 Hydraulic Pressure Digester 3.09.7.2.2 Floating Gas Holder Digester 3.09.7.3 Digestion Process of Large- and Medium-Scale Biogas Plants 3.09.7.3.1 Process Flowchart 3.09.7.3.2 Types of Digesters 3.09.8 Application of Biogas Technology 3.09.8.1 Modes of Biogas Plant 3.09.8.2 Farm Biogas Plant 3.09.8.3 Organic Municipal Solid Waste Treatment Biogas Plant 3.09.9 Utilization of Biogas 3.09.9.1 Biogas Upgrading Technologies 3.09.9.2 Biogas Compression and Storage 3.09.10 Utilization of Biogas Residue and Slurry 3.09.10.1 Utilization of Biogas Residue 3.09.10.2 Utilization of Biogas Slurry 3.09.11 Perspectives References Relevant Websites

q

111 111 113 113 113 113 113 114 114 114 114 114 115 115 115 116 116 117 117 117 117 118 118 118 118 118 119 119 119 119 120 120 120 121 122 123 124 125 125 125 125 126 127

Change History: August 2018. Chuanshu He, Yang Mu, Xiaofeng Liu, Zhengying Yan, and Zhengbo Yue updated the sections.

110

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00154-3

Biogas

111

Glossary Anaerobic digestion (AD) AD is a biological process in which microorganisms break down biodegradable solids and soluble substances in the absence of oxygen. It is widely used for the treatment of industrial and domestic wastes with the production of biogas. AD microbiology AD microbiology is a systemic science that studies anaerobic microorganisms, including their isolation, identification, and culture techniques, and analyses their physiological and biochemical characteristics, the functions, interrelations, and evolution of microbial communities, and populations in AD processes, which is the basis of biogas production. Eco-friendly energy production An eco-friendly energy production mode means that the biogas plant works as a part of the integrated ecological system that incorporates waste treatment, livestock breeding, and other farmland production activities. In such a system, animal feces and other wastes from livestock breeding are used as feedstock for the biogas plant. Biogas can be used to produce electricity or heat, while the digested sludge serves as fertilizer for farmlands. According to theoretical calculations, the biogas plant and farmlands are built up to meet the requirement of the ecological and sustainable development of agricultural production and to minimize the investment and operation costs while increasing energy output. Environment-friendly energy production An environment-friendly energy production mode is one in which a simultaneous treatment of wastes and production of clean energy are achieved. It is a systematic project that aims to effectively treat wastes and recover energy at the same time. Interspecies electron transfer Electrons flowing from one species of organism to other by shuttle components, cell-cell contact, or conductive mineral. It plays an important role in anaerobic microbial communities that degrade biodegradable organic matter. In methanogenic environments, interspecies H2/formate transfer is the primary mechanism for interspecies electron exchange. Direct interspecies electron transfer (DIET) Species exchange electrons through conductive mineral or biological electrical connections. It is a potential alternative to interspecies H2/formate transfer.

3.09.1

Introduction

With growing population and increasing energy demand, energy shortage has stimulated the worldwide explorations into produce biofuel products, such as ethanol, butanol, and biogas, for replacing the traditional fossil fuels. AD not only works as a promising biological technique for organic wastes recycling and environmental pollution control but also provides a renewable energy product – biogas, a mixture formed mainly of methane and carbon dioxide, with low investment, abundant low-cost feedstock and convenient management. Biogas technology involves several aspects, including waste collection, equipment manufacture, biogas production and distribution, organic fertilizer, and farming. The advancement of modern biotechnology, chemical engineering, and the manufacturing industry is stimulating the rapid progress of biogas technologies. It is estimated that about 5% of the hundred billion tons of biomass derived from plant photosynthesis annually is converted into wastes and decomposed thereafter by microorganisms in an anaerobic environment. By collecting the methane in produced biogas for electricity generation and other energy consumptions, a significant contribution is possibly made to supplement global energy shortage. This chapter presents the basic principles and updated technologies with respect to the production of biogas as a renewable, ecoand environment-friendly energy resource, with a focus on its bulk production. The technical considerations and recent advances of anaerobic processes, as well as the applications of biogas, are also reviewed.

3.09.2

Microbiological Aspects

The AD process includes a series of bioprocesses in which different dominant microorganisms break down biodegradable solids and soluble substances under anaerobic or oxygen-free conditions, producing biogas. Biogas is mostly composed of methane (40%– 70%) and carbon dioxide (30%–60%). According to the category of the predominant microbes, the AD process is divided into four stages: hydrolysis, acidogenesis, acetogenesis, and methanogenesis (Fig. 1). Hydrolytic bacteria. Hydrolysis reaction is known as the most activated reaction of the consortia of bacteria involved in AD.1 Within the first stage, hydrolytic bacteria, from the group of relative anaerobes of species Clostridium, Proteus, Peptococcus, Bacteroides, Bacillus, Vibrio, Acetivibrio cellulolyiticus, Staphylococcus, Micrococcus,1 can excrete extracellular enzymes, such as cellulases, amylases, proteases, and lipases, to hydrolyze macromolecular organic substances (proteins, carbohydrates, lipids, cellulose, etc.) into soluble monomeric or dimeric substrates, which could be used as the nutrients for self-growth, and utilized by other microorganisms as well within the system. Fermentative bacteria. In the acidogenesis stage, the fermentative bacteria convert simple monomers produced by the hydrolysis into short-chain volatile fatty acids (SCVFAs), trace alcohols, ketones, CO2, NH3, H2S, and H2. Among the products of fermentation,

112

Figure 1

Biogas

Stages of the AD process.2

acetate and carbon dioxide can be directly used by methanogenes as substrate and energy source. In the acetogenesis process, other compounds such as lactate, ethanol, propionate, butyrate can only be used by methanogenic bacteria after they are converted by obligatory bacteria of the following species: Lactobacillus, Eschericia coli, Staphylococcus, Bacillus, Pseudomonas, Micrococcus, Eubacterium limosum, Clostridium, Eubacterium, Bifidobacterium, Acetobacterium, Syntrophomonas wolfei.1 Hydrogen-producing acetogenic bacteria. This group of bacteria metabolizes C3 or higher organic acids (propionate, butyrate, etc.), ethanol, and certain aromatic compounds such as benzoate into acetate, formate, H2, and CO2, which is not favorable thermodynamically and thus rate-limiting. Up till now, only few hydrogen-producing acetogenic bacteria have been identified, including Syntrophomonas, Syntrophospora, Syntrophobacter, Fusobacterium, and Pelobacter. Bacteria related to Syntrophomonas genus oxidise C4–C7 fatty acids and caproic acid to acetic acid and H2. However, acetogens as obligate hydrogen producers are unable to survive in high partial hydrogen pressures. Coincidentally, the hydrogen-consuming methanogenic microorganisms can rapidly scavenge hydrogen and maintain the partial pressure of hydrogen at an extremely low level. Homoacetogens. As a mixotrophic bacterium capable of autotrophy and heterotrophy, the homoacetogens can use either H2/ CO2 via the acetyl-CoA pathway or a wide range of sugars, alcohols, methoxylated aromatic compounds and single carbon compounds (methanol and formate) to produce acetic acid. The existence of the homoacetogens not only increases the concentration of acetic acid for methane production but also keeps the hydrogen partial pressure low in the anaerobic system. However, the functions of these bacteria in the anaerobic process are still under debate. It is estimated that acetic acid produced by these bacteria accounts for 1%–4% in mesophilic digesters and 3%–4% in thermophilic digesters. Currently, approximately 100 homoacetogenic species phylogenetically classified in 21 genera have been identified, for instance, Acetobacterium woodii, Acetobacterium wieringae, and Clostridium thermoautotrophicum.3 Methanogens. Methanogens from the Euryarchaeota kingdom of Archaebacteria are characterized by high physiological specialization and extremely strict anaerobiosis. They can convert products of previous phases, that is, acetic acid, H2/CO2 and formate and methanol, methylamine or dimethyl sulfide into methane and carbon dioxide. Methanogens can be divided into three groups: hydrogenotrophic, aceticlastic, and methylotrophic methanogens. Currently, over 200 species in 25 genera and 3 classes of methanogens have been isolated and identified with three representative orders: Methanobacteriales, Methanomicrobiales and Methanosarcinales. Despite the fact that only few bacteria are capable to produce methane from acetic acid, and hydrogen-consuming methane production is the more effective process of energy capture by methanogens, a vast majority of CH4 (about 70%) is found to result from acetic acid conversions by heterotrophic methane bacteria.1 Methanogens with various morphologies like rod, cocci, and spiral share the following characteristics: 1. 2. 3. 4. 5.

Extremely low growth rate – for example, the doubling time of Methanosaeta is 4–9 days Strict anaerobiosis – they are sensitive to oxygen and oxidants and thus cannot survive with exposure to oxygen or air Limited simple compounds as their nutrition sources Living in neutral or weak alkaline environment with suitable temperature Biogas as their major end product, which mainly consists of methane and carbon dioxide

Interaction. Microbes in the AD process constitute a complex and balanced ecosystem by mutualism and symbiosis. Nonmethanogens offer nutrition for the growth and reproduction of methanogens, remove toxic substances to the methanogens, and create suitable conditions for methane production. Meanwhile, methanogens eliminate product inhibition for nonmethanogens. The synergistic action of these two groups ensures an appropriate pH condition for the AD system. The complex anaerobic microbe often aggregates in the form of flocs or granule by creating a network of cells and extracellular polymeric

Biogas

113

substances (EPS) in different wastewater treatment systems. The protein contents in the EPS were discovered to significantly contribute to the granule formation and stability.4 Anaerobic flocs shows a loose and homogenous structure, while granular sludge has a denser and multilayer structure with microbes closely attached to each other and embedded in an extracellular matrix, ensuring a high level of metabolic activity. The formed suspended flocs can cause bulking and foaming problems with relatively slow settling velocity, whereas the granular sludge has an excellent settling property and dense microbial structure for withstanding high-strength organic wastewater and its shock loading.4

3.09.3

Interspecies Electron Transfer in AD

Electrons flowing from one species of organism to other by shuttle components, cell–cell contact, or conductive mineral, known as interspecies electron transfer, play an important role in anaerobic microbial communities that degrade biodegradable organic matter. In methanogenic environments, interspecies H2/formate transfer is the primary mechanism for interspecies electron exchange.5,6 Recently discovered direct interspecies electron transfer (DIET), in which species exchange electrons through conductive mineral or biological electrical connections, is a potential alternative to interspecies H2/formate transfer.7

3.09.3.1

Traditional Interspecies Electron Transfer

Traditional interspecies electron transfer relies on the diffusion of electron carriers between species. The reduced electron carriers (like NADH, FADH, ferredoxins (Fd)) released by acetogenic bacteria are regenerated to the oxidized state via the reduction of protons to H2.6 Since first proposed over 50 years ago, interspecies hydrogen transfer has been considered to be essential to the generation of methane in anaerobic digesters.7 However, H2 is poorly soluble in water, therefore formate can serve as a substitute for H2.8 Mainly based on Fick’s law, the results of diffusion models showed that interspecies formate transfer could sustain a 100fold higher conversion rate than interspecies hydrogen transfer.5 Formate has been successfully proven to be used as an electron transfer molecule in co-cultures thriving on fatty acids (propionate and butyrate) or proteins.8 Some syntrophic interactions could also use both formate and H2 to transfer electrons between species.8

3.09.3.2

Direct Interspecies Electron Transfer

DIET is a recently discovered syntrophic metabolism in which microorganisms exchange electrons between cells to cooperatively degrade organic compounds via electrical contact.9 From 2006, the research groups of Stams and others suggested that DIET could happen between obligate H2-producing acetogens bacteria and methanogenic archaea in some environments.9 DIET is considered to be potentially more effective for interspecies electron transfer than traditional strategies under certain conditions. Biological DIET. The possible existence of biological DIET was first discovered during the study of interspecies electron exchange mechanism in the natural conductive methanogenic aggregates in a simulated anaerobic wastewater digester.8 Interestingly, Geobacter species are abundant in most of the methanogenic environments reporting biological DIET, which is probably because Geobacter species form networks using metallic-like conductive pili.8 The stacking of p–p orbitals of five aromatic amino- acids in the pilin monomer has been proposed to contribute to the metallic-like conductivity of Geobacter species.8 Cytochromes that are abundantly present outside the cell may facilitate electron transfer to or from the pili.7 Cell aggregation in methaneproducing cultures can be considered as a strategy to facilitate the direct interspecies electron transfer.7,9 Facilitating DIET. Since a relative long time (about 30 days) is required for the initial adaption of Geobacter species to transfer electron between species via pili-mediated DIET, conductive materials could be a solution to facilitate DIET. Conductive materials can mediate electron transfer between cells during DIET, demanding less energetic investment because it would be unnecessary to produce extracellular components for biological electrical connections.8 So far, conductive additives such as nano-magnetite, akaganeite, goethite, granular activated carbon, biochar carbon cloth, and anthraquinone disulfonate have been shown presumably able to accelerate DIET and thus enhance methane production from organic wastes under anaerobic conditions.

3.09.4

Key Factors in the AD Process

3.09.4.1

Inoculums

Inoculums with active methane-producing microorganisms are critical for the rapid and successful start-up of an anaerobic digester. The influence of the inoculum on the AD process mainly depends on the four factors: origin/source, concentration, activity, and pretreatment. The impact of pretreatment will be discussed in Section 3.09.4.10, while other three factors are elaborated below. Origin/source. Digested sludge from agricultural wastes, discarded food, restaurant wastes, cattle and swine excrement, digested municipal sludge, and mixtures of swine and digested municipal sludge have all been used as inoculum source. The amount of inoculum required and the performance of the AD process will be varied with the type of the inoculum used.1 Digested sludge that can be found worldwide from a municipal wastewater treatment plants is generally used.10

114

Biogas

Concentration. Normally, the volume of the inoculum should be ranged from 10% to 80% of the total working volume of the digester, whose content of biomass proportional to the volatile solids (VS) content of the inoculum can range in manures and granular sludge from 2%–3% to 10%.10 20%–30% supplementation of inoculums would favor the smooth start-up of AD digesters. Activity. In general, inoculum activity is expressed by assessing specific methanogenic activity (SMA) using different positive control substrates. As reported previously, the experimental SMA values should be close to the theoretical ones.10

3.09.4.2

Substrate

The substrates commonly used for biogas generation in AD process are complex solid organic materials, including agricultural wastes, discarded food, restaurant wastes, cattle and swine excrement, and digested municipal sludge. Anaerobic biodegradability of organic matter expressed by SMA on a VS basis is related to its characterization, particle size, and concentration. Since proteins, lipids and extracted fractions of carbohydrates are usually the soluble parts of organic substances, biodegradability is limited by the crystallinity of the cellulose and the lignin content.10 A particle size of
10 mm for the lignocellulosic wastes is suggested, otherwise it would be difficult for microorganisms to carry out hydrolysis following digestion process.10 The amount of volatile solids added or fed to the digester per day per unit volume, represented by organic loading rates (OLR), should be balanced between the concentrations for the successful operation of digester and the effective generation of biogas. In food waste anaerobic digestion systems under mesophilic conditions, the optimal OLR were found to be 22.65 kg VS/m3 day for hydrogen fermentation reactor and 4.61 kg VS/m3 day for methane fermentation reactor in an integrated two-stage digester, while 9.2 kg VS/m3 day in a single-stage reactor.11 Thermophilic system has greater potential to relieve the overloading inhibition compared with mesophilic one.12

3.09.4.3

Temperature

The temperature of different digesters varies between 8 and 65  C. Digestion temperature can be divided into three subranges: low temperature below 25  C (psychrotrophic digestion), moderate temperature from 25 to 45  C (mesophilic digestion), and high temperature from 45 to 65  C (thermophilic digestion). Most anaerobic digesters run under mesophilic or thermophilic conditions. Although the reaction rates, high-load bearing capacity and gas production rate increase with the increase of temperature, acidification is easy to occur because of the inadaptability of methanogenesis during thermophilic AD. Therefore, the optimal conditions for organic matter biodegradation and biogas production would be a two-phase anaerobic digestion process, namely thermophilic hydrolysis/acidogenesis and mesophilic methanogenesis.12

3.09.4.4

pH

Each group of microorganisms in AD digester has a different optimum pH range, for example, 6.5–7.2 for methanogen, 4.0–8.5 for acid-forming bacteria, and 5.5–6.5 for hydrolysis and acidogenesis bacteria.13 Methanogenic microorganisms are most sensitive to pH, hence the pH value of AD digester designed to obtain biogas is generally controlled at 6.5–7.5. Biogas production is generally inhibited or even fails at pH values below 6 or above 8. However, faster growth of fermentative bacteria than the methanogens will accumulate acids, resulting in the fast drop of pH. A buffering capacity of 1000–3000 mg L1 CaCO3 or a molar ratio of at least 1.4:1 of bicarbonate/VFA should be required to maintain a stable digestion process.13

3.09.4.5

Oxidation–Reduction Potential

Methanogens are strict anaerobic microorganisms that can be inhibited or even killed by trace oxygen. Therefore, a low oxidation– reduction potential (ORP), which varies linearly with the logarithm of oxygen concentration, is necessary to maintain their activities. Normally, the ORP of non-methanogenesis is in a range from 100 to 100 mV, while that of methanogenesis is in a range from 400 to 150 mV.

3.09.4.6

Retention Time

Retention time is the duration required to complete the degradation of organic matter. It depends on the generation time of methanogens, the process temperature, OLR and substrate compositions.12 Hydraulic retention time (HRT) and solids retention time (SRT) are the terms commonly used to denote the average time that substrate and bacteria (solids) spend in the AD, respectively. In conventional low-rate digesters or reactors without recycle or supernatant withdrawal, the SRT equals to the HRT.13 However, in high-rate digesters, SRT should be much higher than HRT by way of attached growth or suspended growth, for instance, SRT is about three times higher than the HRT in a typical high-rate anaerobic digester treating dairy manure.14 The critical SRT values for designing a digestion with temperature ranging from 18 to 40  C are shown in Table 1. Since these SRT values were established in ideal conditions, a multiplication factor of about minimum 2.5 is recommended in practice for safety.13

Biogas Table 1

3.09.4.7

115

Suggested SRT for the design of completely mixed high-rate digesters14

Operating temperature ( oC)

Minimum SRT in theory (d)

Minimum SRT in practice (d)

18 24 30 35 10

11 8 6 4 4

28 20 14 10 10

Mixing

Mixing in anaerobic digesters is essential to transfer substrates to microorganisms, dilute inhibitory substances, ensure uniform pH and temperature, and prevent stratification and by-passing flow. Mixing can be accomplished with gas and draft tubes with mechanical mixers or by recycling pumps, while mixing by circulating the biogas produced with blowers is more economic, particularly for large-scale biogas plants.

3.09.4.8

Nutrients

The microbial population responsible for anaerobic digestion requires nitrogen enzymes/protein biosynthesis, and phosphate for formation of energy carriers ATP and NADP. The appropriate C/N ratio is 20–30 in AD process with a ratio of 25 being the most commonly used.12 Substrates with low C/N ratio would increase the risk of ammonia toxicity to methanogens and cause insufficient utilization of carbon sources.12 The phosphorus demand is about 15% that of the nitrogen demand. In order to achieve high biogas production, it is desirable to use different raw materials simultaneously to maintain the proper C/N ratio in the digester. Inorganic elements such as iron, nickel, cobalt, and zinc are the micronutrients required by some microorganisms in relatively small quantities to initiate the digestion process. Ni is a stimulatory in both biogas production and the methane content of biogas.12 Generally, dosages of some micronutrients in chloride form are recommended as follows: FeCl2 1.0 mg L1, CoCl2 0.1 mg L1, NiCl2 0.1 mg L1, and ZnCl2 0.1 mg L1.1

3.09.4.9

Toxicity and Inhibition

A wide variety of inorganic and organic substances either already present in the digester substrate or generated during digestion have been reported to inhibit the AD processes. Inhibition is usually indicated by a reduction of the microbial population and steadystate rate of methane production, the accumulation of VFAs, and the reduction of pH.1 Ammonia. Except for extra addition, degradation of nitrogenous matter, mainly proteins and urea, would produce ammonia.13 Ammonia ( HRT; the microbes are fixed onto surface of the inert filler; effectively reduces the loss of microbes; applicable for the treatment of organic wastewater with a low chemical oxygen demand (COD) and/or with low solid sludge

A biogas plant is a systematic project that integrates biogas production, resource utilization, and pollution control. The adopted process techniques and system design of biogas plants are significantly different in various countries, due to different local climate, energy policies, fossil energy reservation conditions and economic status. According to their functions, production objectives and impact on environment, biogas plants can be classified into two categories: eco-friendly energy production and environmentfriendly energy production modes. The substance and energy flows of these two modes are shown in Fig. 5. Normally eco-friendly energy production mode is suitable for areas possessing farmland, fishing ponds and vegetation plantation, which can receive and consume the residue digestate of the biogas plant. It is an ideal mode since the low investment and operation costs are required for a good comprehensive performance. Such an eco-friendly energy production mode has been employed in many farm biogas plants in Europe and in some household and livestock manure biogas plants in China. In the environment-friendly energy production mode, both the treatment of wastes and production of clean energy are achieved simultaneously. The biogas plants that use municipal solid wastes, organic wastewater and excess sludge as the raw materials are typically running in an environment-friendly energy production mode.

3.09.8.2

Farm Biogas Plant

A farm biogas plant is built to digest farm wastes including manure and other organic substances and produce biogas, which can enhance the economic benefit of the farm. Farm biogas plants can function both decentralized and centralized modes.26 In decentralized farm-based plants, the self-produced digestion materials of the farms are utilized for ecological energy production. Such plants are relatively small in size and have a low biogas production efficiency. Centralized farm biogas plants are the combination of multiple farm units, with the livestock manure and other organic substances collected and transported from adjacent farms, which are usually established in the central area of the multiple farms to reduce cost and time for feedstock transportation and the convenient utilization of biogas produced. A typical farm biogas plant is shown in Fig. 6, and its technical features mainly include fermentation materials and process designs.

Figure 4

Schematic diagram of batch dry digesters: (A) one stage, (B) sequential and (C) hybrid.

Biogas

121

Figure 5 Schematic diagram of substance and energy flow in two modes of a biogas plant: (A) eco-friendly energy production mode, and (B) environment-friendly energy production mode.

3.09.8.3

Organic Municipal Solid Waste Treatment Biogas Plant

Municipal solid waste (MSW) contains a variety of biodegradable organic substances. Biogas plants using MSW as raw materials are becoming attractive, and more and more commercial plants will be established across the world. On the other hand, these biogas plants can effectively reduce the facilities required for MSW landfill or incineration.27 The anaerobic digestion process of MSW

Figure 6

Process flowchart of a typical farm biogas plant.

122

Biogas

MSW

Biogas

Fuel Power generation

Purification

Fraction

Soil conditioner Organic MSW

Feedstock pretreatment

Anaerobic Digestate Residue Dehydration digestion

Thermochemical treatment Landfill

Digestate backset

Treatment and controlled discharge

Slurry

Liquid fertilizer Figure 7

Typical process flowchart of organic MSW treatment biogas plant.

normally includes four steps as can be seen from Fig. 7. Several typical processes have been developed for MSW anaerobic digestion as follows. Eco Technology JVV Oy process. This system illustrated in Fig. 8 is an one-stage continuous wet process established by Bottrop, Germany, at a plant with an annual treatment capacity of 6,500 ton organic MSW. Currently, several plants with the Eco Technology JVV Oy process are under construction. This process requires a temperature of 35–55  C for thermophilic digestion, an HRT of 15–20 days and a reactor volume of 5000 m3. The feedstock in the tank is mixed by the produced biogas, and digestate is pasteurized for 30 min at 70  C to ensure its safety for farmland use. Dranco process. This process is a dry continuous one-stage process under thermophilic conditions with the temperature of 48–58  C (Fig. 9). In Europe, currently four large organic MSW treatment plants are using the Dranco process, with the capacity of 11,000–35,000 tons per year. AD process features quick start-up, mature technologies for construction and management and convenient material feeding and discharge. These advantages make it the prevailing selection for biogas production. However, it suffers several disadvantages such as large tank volume, difficulty to separate slurry and residue, as well as more wastewater. The TS of the organic fraction in MSW is normally over 20% and is an appropriate feedstock for dry digestion. Compared with wet digestion, dry digestion has higher performance in treating organic fraction of MSW and thus exhibits a bright prospect. In recent years, the effective treatment of excess sludge has received increasing attention. Co-digestion of organic MSW and excess sludge presents a promising technological process for treatment and recycling of MSW and sludge.

3.09.9

Utilization of Biogas

Prior to the utilization, biogas needs to be purified or upgraded, which involves a series of processes including dehydration, desulfurization and removal of organic halides, carbon dioxide and heavy metals (Fig. 10). The degree of purification varies with the purpose of its utilization. Sorted organic wastes Anaerobic digester

Primary crushing

Biogas storage tank

Biogas stirring

Metal separation Digestate Drum Screening

RFD Incineration

Feed preparation

Process water

pasteurization

Buffer tank

Liquid fertilizer

Dehydration

Solid humus

Wastewater Figure 8

Flowchart of the Eco Technology JVV Oy process.

Biogas

123

Organic wastes Heat and electricity

Lost heat of the gas turbine

CHP production

Screening

Metal removal

Biogas Anaerobic digester

Feedstock feeder

Digestate

Partial stream

Crushing

Sludge

Aerobic composting

Filtration

Mixing Dehydration

Solid humus Figure 9

Flowchart of Dranco process.

3.09.9.1

Biogas Upgrading Technologies

Carbon dioxide removal. CO2 accounts for about 40% of biogas, greatly affecting the heating value of biogas. Therefore its removal is the core process for biogas purification. Table 7 summarizes various physical and chemical methods for the removal of CO2 in biogas. Water removal. Untreated biogas is usually saturated with water vapor when it leaves the digester. Compared with other impurities in biogas, water removal is relatively simple. Besides for the condensation method, flowing through the solution

Desulphurization dehydration

Biomethane CH4 97-99%

Natural gas pipeline system

Biogas CH4 60-70%

Compression /liquefaction

Figure 10

Gas station

Gas cylinders

Vehicles

Consumers

Overview of biogas utilization.

Boilers

Gas turbines

CHP

Fuel cells

On-site utilization

Biogas production

Off-site utilization

Upgrading

124

Biogas

Table 7

Summary of techniques for CO2 removal and their main principles

Techniques

Principle

Pressure swing adsorption (PSA)

Separating CH4, CO2 and N2 from the biogas mixture based on their affinity to different adsorbent materials (active carbon, silica gel, molecular sieve, Al2O3, zeolite) Taking advantage of the solubility of CO2 and insolubility of CH4 in some special solution to remove CO2 without any chemical reactions A simple process to remove CO2 since it has a higher solubility in water than methane CO2 reacting with special chemical substance including alkali solutions like sodium/potassium/ calcium hydroxides, K2CO3, and amine compounds such as mono ethanol amine (MEA) and di-methyl ethanol amine Based on the different permeability of the component in the membrane material (hollow fibers, spiral wound modules, etc.), CO2 can pass through the membrane driven by the high pressure while CH4 are retained

Physical adsorption Water Scrubbing Chemical absorption Membrane technology

(e.g., hygroscopic salts like CaCl2 and LiCl2, glycols) or solid adsorbents (e.g., SiO2, activated charcoal or molecular sieves, and compound desiccant) with high hydroscopic property is another commonly used approach to separate water from biogas. H2S removal. Various methods listed in Table 8 have been proposed to remove H2S in biogas. Dry and wet desulfurization methods are regarded as the main practical approaches nowadays. However, the disadvantages of these two kinds of traditional methods cannot be ignored, such as highly contaminated, high-cost and low-efficiency. Therefore, environment-friendly and cost-effective biological treatment, taking advantage of the metabolism of microbes to transform the H2S into element S or sulfate, is a newly developing technology for desulfurization. Trace gases removal. Oxygen and nitrogen can mix into the biogas if the gas is collected under pressure or the purification process is not controlled cautiously. Presence of nitrogen and oxygen might result in the failure of certain subsequent desulfurization and decarbonization processes. Currently, the removal of both nitrogen and oxygen can be achieved principally by physisorption with activated carbon, molecular sieves or membranes. Oxygen can also be removed via catalytic reduction with hydrazine, sodium sulfite, or pyrogallol and catalytic oxidation of hydrogen or hydrocarbons.

3.09.9.2

Biogas Compression and Storage

The critical temperature and pressure of biogas for liquefaction are 82.5  C and 47.5 bar, respectively, so it does not liquefy under pressure at ambient temperature, causing difficulties for biogas storage. The compression of the biogas is able to reduce the storage requirements, concentrate the energy content and provide pressure to overcome the resistance to gas flow. The storage facilities vary depending on the storage pressure, as illustrated in Table 9.13 Table 8

Summary of techniques for H2S removal and their main principles

Techniques

Methods

Principle

Dry desulfurization

Chemical adsorption

Processes involving chemical adsorption of H2S on solid adsorbents such as active carbon Catalytic conversion of H2S into elemental S in the presence of catalysts like vanadium oxide, Fe contained materials, and activated carbon under high reaction temperature Taking advantage of the solubility of H2S and insolubility of CH4 in some special solutions (e.g., methanol, sulfoxide) to remove H2S One of the oldest methods for H2S removal involves alkaline solutions scrubbing to form compounds Converting H2S into elemental S via catalysts containing Fe or Cu or V elements in liquid phase at room temperature A purification apparatus with microorganisms immobilized in the form of a biofilm fixed on a packed bed comprised of material such as peat, soil, and activated carbon. When the biogas enters into the biofilter and reaches the 0.5–1 m biofilm, microorganisms will convert H2S into S0 and SO42 in the presence of oxygen The pollutant is first transferred from the gas phase to the liquid phase in the contact tower containing adsorbents, following with being oxidized in the eration tank composed of activated sludge Similar to a biofilter, with the difference being that the packed bed is trickled over with a nutrient solution

Catalytic hydrogenation Wet desulfurization

Physical adsorption Chemical adsorption Wet oxidation

Biological treatment

Biofilter

Bioscrubber Biotrickling filter

Biogas

Table 9

125

Commonly used biogas storage options

Pressure

Storage device

Material

Low (0.14–0.41 bar) Low Medium (1.05–1.97 bar) High (200 bar)

Water sealed gas holder Gas bag Propane or butane tank Commercial gas cylinders

Steel Rubber, plastic, vinyl Steel Alloy

3.09.10 Utilization of Biogas Residue and Slurry 3.09.10.1 Utilization of Biogas Residue Crop fertilizer. Biogas residue is the bottom slag produced by AD process existing in organic and inorganic solid form, and rich in various kinds of vitamins, large amounts of humic acid, indole-3-acetic acid, N, P, K and other trace elements, a range of active substances secreted by microorganisms. Biogas residue with a certain percentage of the chemical fertilizer can increase the yield of vegetables and fruits, significantly reduce the nitrate content in leafy vegetables, and improve the content of Vitamin C, organic acids and sugar in fruit. Soil conditioner. Biogas residue is a good soil conditioner, which can supplement phosphorus and potassium to soil, benefit the nutrient balance, and improve soil permeability. It is found that biogas residues have higher amounts of organic matter compared with chemical fertilizer. The increase of organic matter in soil is conducive to microbial activity and the formation of soil aggregate structure, hence strengthening the stability of soil. Feedstuff. Biogas residue is rich in nitrogen, phosphorus, potassium and other trace elements and contains a certain amount of hormones and vitamins, which can help the growth and propagation of fungi, earthworms, fish. The nutrients in biogas residues are rich and easy to absorb, so feeding pigs and chicken with residue can make them appetite vigorous and do not get diseases or seldom ill.

3.09.10.2 Utilization of Biogas Slurry Organic foliar fertilizer. Biogas slurry also contains large amounts of nutrients similar to biogas residues. Biogas slurry is also often used as liquid fertilizer directly to improve crops yield and quality as its rich content of nutritive elements. The most frequently applied way of slurry is used as foliar fertilizer, which can improve crops yield and quality, swell of fruits, inhibit diseases and insects. Seed soaking agent. The soluble nutrients in biogas slurry will be easily absorbed by seed with the function of osmosis. It can effectively activate the enzyme in both seed embryo and endosperm, stimulate germination and growths of seeds, accelerate the transformation of nutrient from dormancy to seedlings, promote metabolism, and meanwhile kill the pathogenic bacteria.28 Prevention of crop diseases and insect pests. Practical and scientific experiments have confirmed that biogas slurry can partially substitute pesticides to control 23 kinds of diseases, 14 types of pests from 13 kinds of agronomic crops including grain, vegetables, and fruit trees.28 Biogas slurry was called biological pesticides as its non-residue, non-pollution and nonresistance. Butyric acid, a kind of phytohormone – gibberellin, indole acetic acid and vitamin B12 contained in biogas slurry have obvious inhibitory effect on pathogenic bacteria. Ammonia, ammonium salts and some antibiotics in biogas slurry are able to directly kill pests. Fish culture. Being rich in various kinds of amino acids and trace elements makes biogas slurry a good compound feed for animal, especially fish. Feeding with biogas slurry can increase the amount and quality of planktons as fish crops, take control of diseases and pests, and save feedstuff. However, due to the risk of inadequately killing pathogens by anaerobic digestion, the relative management departments in China generally prohibit the use of biogas slurry as feed additives.

3.09.11 Perspectives Biogas production in AD has two elegant advantages. It doesn’t need high quality feedstock and provides an efficient and environment-friendly treatment method for various wastes including municipal wastes, farm wastes, food process and other industrial wastewaters. On the other hand, biogas recovery is much simpler and easier, compared with the production of other biofuels. Thus, many commercial plants have been constructed across the world. For the further development of biogas AD, the authors recommend that attention should be given to the following aspects: (1) AD is a biochemical process driven by microorganisms and microbial management can improve early diagnosis and process optimization. Establishment of effective online detection systems based on reliable microbial indicators is necessary to complement routine process monitoring and control. (2) Efforts are required to reduce the inhibitory effects of ammonia, VFAs, H2, and sulfides, and some other new inhibitory substances,

126

Biogas

including nanoparticles. (3) A better mechanistic understanding of the role and potential of DIET in complex AD environments will allow for biogas production improvement at more fundamental levels. (4) Integration of AD and other technologies, such as pyrolysis, may overcome defects in each individual process and improve both resource use and operation efficiency. Recently, lignocellulosic biomass is being investigated extensively as a renewable feedstock to produce biofuels. By 2050, lignocellulosic biomass is predicted to provide approximately 17% of the world’s electricity and 38% of the world’s direct fuel.29 In fact, the AD process for biogas production can be well integrated into the biorefinery processes with the by-products of the biofuels production as substrates, which can enhance the net energy balance, and in the meantime, partly recycle valuable nutrients. However, two aspects need to pay more attention: effective and economical pretreatment of the substrates due to their recalcitrance to biodegradation, and novel technologies for more efficient biogas production and utilization. Without doubt, advances in engineering science and biotechnology fundamentals will facilitate the progress of the R&D of the biogas production and its commercial applications in the future. Some environmental and ecological concerns still exist with the AD process. One of the most important concerns is the toxic components, such as antibiotics and inorganic additives that are used to overcome the disadvantages of AD process, which would contaminate the environment and enter the food chain through the contaminated soil. Posttreatment of such digestion effluent is critical to remove and degrade these toxic components. Some technologies including wetland systems, aerobic treatment and advanced oxidation processes have been developed. Another is the remained nutrients in the effluent such as nitrogen and phosphorous, which could result in the eutrophication of the water body. Processes aiming at the recovery of these nutrients have been developed, but more detailed studies need to be performed to make them more economically viable.

References 1. Amani, T.; Nosrati, M.; Sreekrishnan, T. R. Anaerobic Digestion from the Viewpoint of Microbiological, Chemical, and Operational Aspects - a Review. Environ. Rev. 2010, 18, 255–278. 2. Madsen, M.; Holm-Nielsen, J. B.; Esbensen, K. H. Monitoring of Anaerobic Digestion Processes: A Review Perspective. Renew. Sustain. Energy Rev. 2011, 15, 3141–3155. 3. Drake, H. L.; Küsel, K.; Matthies, C. Acetogenic Prokaryotes, Springer: New York, 2006. 4. McSwain, B. S.; Irvine, R. L.; Hausner, M.; Wilderer, P. A. Composition and Distribution of Extracellular Polymeric Substances in Aerobic Flocs and Granular Sludge. Appl. Environ. Microbiol. 2005, 71, 1051–1057. 5. Stams, A. J.; Plugge, C. M. Electron Transfer in Syntrophic Communities of Anaerobic Bacteria and Archaea. Nat. Rev. Microbiol. 2009, 7, 568–577. 6. Baek, G.; Kim, J.; Kim, J. S.; Lee, C. S. Role and Potential of Direct Interspecies Electron Transfer in Anaerobic Digestion. Energies 2018, 11, 107. 7. Lovley, D. R. Reach Out and Touch Someone: Potential Impact of DIET (Direct Interspecies Energy Transfer) on Anaerobic Biogeochemistry, Bioremediation, and Bioenergy. Rev. Environ. Sci. Biotechnol. 2011, 10, 101–105. 8. Shrestha, P. M.; Rotaru, A. E. Plugging in or Going Wireless: Strategies for Interspecies Electron Transfer. Front. Microbiol. 2014, 5, 237. 9. Dubé, C. D.; Guiot, S. R. Direct Interspecies Electron Transfer in Anaerobic Digestion: A Review. Adv. Biochem. Eng. Biotechnol. 2015, 151, 101–115. 10. Raposo, F.; Rubia, M. D. L.; Fernández-Cegrí, V.; Borja, R. Anaerobic Digestion of Solid Organic Substrates in Batch Mode: An Overview Relating to Methane Yields and Experimental Procedures. Renew. Sustain. Energy Rev. 2012, 16, 861–877. 11. Nagao, N.; Tajima, N.; Kawai, M.; Niwa, C.; Kurosawa, N.; Matsuyama, T.; Yusoff, F. M.; Toda, T. Maximum Organic Loading Rate for the Single-Stage Wet Anaerobic Digestion of Food Waste. Bioresour. Technol. 2012, 118, 210–218. 12. Mao, C. L.; Feng, Y. Z.; Wang, X. J.; Ren, G. X. Review on Research Achievements of Biogas from Anaerobic Digestion. Renew. Sustain. Energy Rev. 2015, 45, 540–555. 13. Appels, L.; Baeyens, J.; Degrève, J.; Dewil, R. Principles and Potential of the Anaerobic Digestion of Waste-Activated Sludge. Prog. Energy Combust. Sci. 2008, 34, 755–781. 14. Abbasi, T.; Tauseef, S. M.; Abbasi, S. A. Anaerobic Digestion for Global Warming Control and Energy GenerationdAn Overview. Renew. Sustain. Energy Rev. 2012, 16, 3228–3242. 15. Chen, J. L.; Ortiz, R.; Steele, T. W.; Stuckey, D. C. Toxicants Inhibiting Anaerobic Digestion: a Review. Biotechnol. Adv. 2014, 32, 1523–1534. 16. He, C. S.; He, P. P.; Yang, H. Y.; Li, L. L.; Lin, Y.; Mu, Y.; Yu, H. Q. Impact of Zero-Valent Iron Nanoparticles on the Activity of Anaerobic Granular Sludge: From Macroscopic to Microcosmic Investigation. Water Res. 2017, 127, 32–40. 17. Donoso-Bravo, A.; Mailier, J.; Martin, C.; Rodrã-Guez, J.; Aceves-Lara, C. A.; Vande, W. A. Model Selection, Identification and Validation in Anaerobic Digestion: a Review. Water Res. 2011, 45, 5347–5364. 18. Harris, P. W.; McCabe, B. K. Review of Pre-treatments Used in Anaerobic Digestion and Their Potential Application in High-Fat Cattle Slaughterhouse Wastewater. Appl. Energy 2015, 155, 560–575. 19. Yu, L.; Wensel, P. C.; Ma, J. W.; Chen, S. L. Mathematical Modeling in Anaerobic Digestion (AD). J. Biorem. Biodegrad. 2013, 5. S4-003. 20. Takahashi, H.; Li, Y. Y.; Noike, T. CFD Simulation of Mixing in Anaerobic Digesters. Bioresour. Technol. 2009, 100, 2228–2233. 21. Nguyen, D.; Gadhamshetty, V.; Nitayavardhana, S.; Khanal, S. K. Automatic Process Control in Anaerobic Digestion Technology: A Critical Review. Bioresour. Technol. 2015, 193, 513–522. 22. Fatih Demirbas, M.; Mehmet, B. Progress and Recent Trends in Biogas Processing. Int. J. Green Energy 2009, 6, 117–142. 23. Bolzonella, D.; Pavan, P.; Mace, S.; Cecchi, F. Dry Anaerobic Digestion of Differently Sorted Organic Municipal Solid Waste: a Full-Scale Experience. Water Sci. Technol. 2006, 53, 23–32. 24. Cho, S. K.; Kim, D. H.; Jeong, I. S.; Shin, H. S.; Oh, S. E. Application of Low-Strength Ultrasonication to the Continuous Anaerobic Digestion Processes: UASBr and Dry Digester. Bioresour. Technol. 2013, 141, 167–173. 25. Weiland, P. Biomass Digestion in Agriculture: a Successful Pathway for the Energy Production and Waste Treatment in Germany. Eng. Life Sci. 2006, 6, 302–309. 26. Chen, Q.; Liu, T. B. Biogas System in Rural China: Upgrading from Decentralized to Centralized? Renew. Sustain. Energy Rev. 2017, 78, 933–944. 27. Hartmann, H.; Ahring, B. K. Strategies for the Anaerobic Digestion of the Organic Fraction of Municipal Solid Waste: an Overview. Water Sci. Technol. 2006, 53, 7–22. 28. Feng, H.; Qu, G. F.; Ning, P.; Xiong, X. F.; Jia, L. J.; Shi, Y. K.; Zhang, J. The Resource Utilization of Anaerobic Fermentation Residue. Procedia. Environ. Sci. 2011, 11, 1092–1099. 29. Feng, Q. J.; Lin, Y. Q. Integrated Processes of Anaerobic Digestion and Pyrolysis for Higher Bioenergy Recovery from Lignocellulosic Biomass: A Brief Review. Renew. Sustain. Energy Rev. 2017, 77, 1272–1287.

Biogas

Relevant Websites Agrobiogas. http://www.agrobiogas.eu. Bioferm. http://www.bio-gas.de. Bundesministerium fur Wirtschaft und Technologie. http://www.german-renewable-energy.com. United States Environmental Protection Agency. http://www.epa.gov. Krieg & Fischer Ingenieure GmbH. http://www.kriegfischer.de. RosRoca. http://www.rosroca.com. Topagrar. http://www.mt-energie.com. lws01. http://www.biogasenergysolutions.com. ZORG Biogas. http://www.zorg-biogas.com.

127

3.10

Biohydrogenq

Patrick C Hallenbeck, Life Sciences Research Center, Department of Biology, United States Air Force Academy, USAF Academy, Colorado, CO, United States; and Département de microbiologie, infectiologie et immunologie, Université de Montréal, Montréal, Montréal QC, Canada Carolina Zampol Lazaro, Département de microbiologie, infectiologie et immunologie, Université de Montréal, Montréal, Montréal QC, Canada Emrah Sagır, Department of Biology, Faculty of Arts and Sciences, Osmaniye Korkut Ata University, Osmaniye, Turkey © 2019 Elsevier B.V. All rights reserved. This is an update of M.-S. Kim, D.-H. Kim, J.K. Lee, 3.11 - Biohydrogen, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 115-125.

3.10.1 3.10.2 3.10.2.1 3.10.2.2 3.10.2.2.1 3.10.2.2.2 3.10.3 3.10.4 3.10.4.1 3.10.4.2 3.10.4.3 3.10.5 3.10.5.1 3.10.5.2 3.10.5.3 3.10.5.4 3.10.5.5 3.10.6 3.10.6.1 3.10.6.2 3.10.7 References

Introduction Biophotolysis General Microorganisms Involved in Biophotolysis H2 Production by Cyanobacteria H2 Production by Green Algae Limitations for Practical Application Dark Fermentation General Factors Affecting Dark Fermentation Limitations of Dark Fermentation for Practical Application Photofermentation by Photosynthetic Bacteria General Factors Affecting Hydrogen Production Strain Improvement Co-cultures Limitations MECs General Mechanism Limitations Closing Remarks

129 130 130 130 130 130 130 131 131 132 133 133 133 134 135 135 136 136 136 137 138 138

Glossary Biohydrogen Hydrogen produced by microbes from renewable substrates Biophotolysis The microbial production of hydrogen using light energy to split water Direct biophotolysis The captured light energy is used to directly reduce ferredoxin and then hydrogenase Indirect biophotolysis The captured light energy is stored as carbohydrate and then later used to reduce ferredoxin and hydrogenase Dark fermentation The microbial anaerobic conversion of organic compounds to various products; organic acids, alcohols, hydrogen MEC (Microbial Electrolysis Cell) Electrons extracted from organic substrates by electrogenic bacteria are boosted with supplemental voltage, producing hydrogen at the cathode. Photofermentation The light-driven conversion of organic compounds to hydrogen by photosynthetic bacteria CSTR Continuous stirred-tank reactor AnSBR or ASBR Anaerobic sequencing batch reactor EGSB Expanded granular sludge bed UASB Up-flow anaerobic sludge blanket ADSBR Anaerobic down-flow structured-bed reactor AnMBR Anaerobic membrane bioreactors

q

Change History: October 2018. P.C. Hallenbeck, C. Zampol Lazaro and E. Sagir updated the keywords and sections to this article.

128

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00155-5

Biohydrogen

129

AFBR Anaerobic fluidized-bed reactors APBR Anaerobic packed-bed reactors CMTR Continuous multiple tube reactor

3.10.1

Introduction

Hydrogen has attracted great interest over the past few decades as a potential future energy carrier. It has a number of unique attributes that make it a desirable fuel; it is easily combusted or readily converted to electricity using fuel cells at relatively high efficiencies and in principal its sole oxidation product is water. However, hydrogen is presently almost exclusively produced using steam reforming of hydrocarbons, a process which is hardly sustainable due to its intense fossil fuel consumption. Therefore, the use of microorganisms to carry out biological hydrogen production (biohydrogen) has attracted much attention in the last few decades due to the possibility of producing clean energy by consuming renewable resources such as agro-industrial wastes or even water. In fact, the microbial world has a great diversity of organisms capable of carrying out some type of hydrogen metabolism and thus different biohydrogen processes are being developed, each with its own potential advantages as well as technical barriers practical application.1–3 In the absence of oxygen, various types of bacteria carry out fermentation of organic substrates and this dark fermentation produces in many cases hydrogen in addition to other products (organic acids, alcohols, etc.).4 Although it has been possible to achieve high volumetric rates of hydrogen production from waste streams with dark fermentation, the low yields obtained and consequently inefficient waste treatment are problematic. Yields are limited by the nature of the microbial metabolic pathways, a restriction that might be mitigated by metabolic engineering.5,6 As well, the large quantities of side products that are produced by dark fermentation might be converted to additional hydrogen, or another biofuel, methane, through the introduction of a second stage. MECs, microbial electrolysis cells have been relatively recently developed where electrogenic bacteria extract electrons and protons from organic substrates at an anode and, with the addition of supplemental voltage, hydrogen is evolved at the cathode. These are suitable for the direct conversion of substrates, such as monomeric sugars, to hydrogen, but obviously could also find use as a second stage to a dark fermentative process since MECs are capable of the thermodynamically unfavorable production of hydrogen from the organic acids produced by dark fermentation since the additional energy required can be supplied as an external electrical current.7 In addition, different types of microorganisms are capable of capturing light energy and using it to drive hydrogen evolution from water (biophotolysis) or from the conversion of organic substances into hydrogen (photofermentation).8 There are a number of challenges for the implementation of light processes for hydrogen production including low efficiencies of light conversion (biophotolysis and photofermentation) and oxygen inhibition of hydrogen production (biophotolysis). Possible workarounds for these problems include reducing the photosynthetic antenna size, the creation of an oxygen tolerant hydrogenase, decoupling oxygen and hydrogen evolution through two-stage or two-phase processes. The high cost of photobioreactors might be overcome by the development of high-tech hydrogen impermeable plastics. Many previous studies have mainly used pure cultures of bacteria or algae, simple substrates, such as easily fermentable sugars and organic acids, and synthetic media. Moving forward towards practical applications, research is examining the use of cheap and abundant substrates, mainly carbohydrate-rich agricultural and industrial wastes as well as investigating various inoculum sources with highly active organisms. Optimization of process economics probably requires choosing the appropriate inoculum and substrate source based on the availability and abundance of agro-industrial wastes of each region/country. In addition, there are technical barriers that must be overcome to make this environmentally friendly process economically feasible. In particular, low hydrogen yields and rates are significant barriers for the practical application of biological hydrogen production. During dark fermentation, microbial metabolic constraints dictate that a significant fraction of the organic substrate that is consumed is used to make side products other than hydrogen, restricting yields to a maximum of 4H2/glucose, or 33% of the theoretical maximum. In reality, yields are even lower due to a variety of other factors, including in some cases hydrogen consumption by microorganisms present in the reactors. The rates and yields of hydrogen production by the light-driven processes of biophotolysis and photo-fermentation are significantly limited by the inefficient utilization of the solar light input due to several factors including the inability to adequately adapt to variations in light intensity and limited effective light penetration due to self-shading. Likewise, MECs, although attractive in principal, require further development to increase volumetric hydrogen production, reduce the overvoltages required, and increase the columbic efficiency by decreasing electron consuming side metabolic reactions, such as methane production. Approaches to R&D for the improvement of biohydrogen processes include engineering, to improve reactor performance and increase process performance, materials sciences to develop better electrode materials for MECs and biophotoreactors, and synthetic biology to develop new metabolic pathways and to increase the efficiencies of existing ones, including the possible development of oxygen resistant hydrogenases. In addition, molecular techniques are being used to understand how the microbial community structure is related to the process efficiency and how it changes relative to the changes in environmental conditions.

130

Biohydrogen

3.10.2

Biophotolysis

The microbial production of hydrogen using light energy to split water.

3.10.2.1

General

Hydrogen production using photosynthetic microorganisms are being investigated as a means of harvesting some of the immense free energy available in the global solar flux. A very attractive option is to use green algae or cyanobacteria which carry out watersplitting photosynthesis and can couple this to proton reduction to hydrogen in a process that called biophotolysis.9–11 Biophotolysis can be either direct or indirect. In direct biophotolysis, electrons from the water splitting reaction are transported through photosystem II (PSII) and photosystem I (PSI) to ferredoxin which then directly reduces a hydrogen producing enzyme. In indirect biophotolysis, reduced ferredoxin is used to fix carbon and produce form some form of carbohydrate, which can then later be used to produce hydrogen. Hydrogen production by heterocystous cyanobacteria, and green algae, principally sulphur-deprived cultures of the green alga, Chlamydomonas, has been relatively extensively studied. The mechanism differs somewhat in the two systems, with each having its own specific limitations while both systems must cope with the essentially incompatible reactions of simultaneous oxygen evolution and proton reduction with an oxygen sensitive enzyme. Different solutions have been developed for the two types of systems, which, however, also impose additional restrictions on yields.

3.10.2.2 3.10.2.2.1

Microorganisms Involved in Biophotolysis H2 Production by Cyanobacteria

Nitrogenase is responsible for the majority of hydrogen production by heterocystous cyanobacteria, prokaryotes that grow in filaments and produce specialized cells, heterocysts, when limited for nitrogen. Heterocysts provide a micro-anaerobic environment through a number of mechanisms, allowing oxygen sensitive processes, such as nitrogen fixation or hydrogen production to occur. Some hydrogen is formed in a side reaction to nitrogen fixation, but nitrogenase turnover continues in the absence of exogenous substrates, reducing protons to hydrogen gas, and hence hydrogen evolution is much greater in the absence of N2. This system, studied for over three and a half decades, is inherently robust and very reasonable conversion efficiencies (conversion of total incident light energy to free energy of hydrogen), capable of being sustained for days to weeks, were achieved using nitrogen-limited cultures early on; 0.4% laboratory studies, with cultures incubated under natural sunlight, 0.1%. Although these efficiencies have not been much improved upon over the subsequent years, there is some room for improvement since theoretical efficiencies with this nitrogenase-based system are around 4.6%. Improvements might be made by reducing the photosynthetic antennae size, substituting hydrogenase for nitrogenase, or increasing heterocyst frequencies. Some unicellular cyanobacteria are able to fix nitrogen or produce hydrogen when cells are grown with a light-dark cycle. Hydrogen can also be produced with these organisms using indirect biophotolysis in which first-stage photosynthesis is used to fix carbon from which reductant that can be used in a second stage to produce hydrogen.

3.10.2.2.2

H2 Production by Green Algae

Some species of green algae contain a [FeFe] hydrogenase and re-illumination of darkened, anaerobic, cultures leads to a short burst of hydrogen production. Relatively recently sustained hydrogen production by illuminated cultures was demonstrated using two stages: 1) initial normal photosynthesis and growth, followed by; 2) sulphur deprivation treatment which greatly decreases photosystem II activity due to photo-damaged D1 protein. This allows the rate of respiratory oxygen consumption to surpass the low rate of remaining oxygen evolution, making the culture anaerobic, thus promoting hydrogenase activity, with hydrogen production continuing for days.

3.10.3

Limitations for Practical Application

Light conversion efficiencies are a critical limiting parameter for any light-based system and are a significant barrier to the practical use of hydrogen production by either cyanobacteria or green algae. Low light conversion efficiencies are thought to be due to a number of factors, with perhaps the most important being inefficient light absorption at high light intensities. Under these conditions, light energy above that which can be usefully used in photosynthesis is absorbed by antenna complexes and wasted as heat or fluorescence. The efficiency of light utilisation might therefore be increased by decreasing the photosystem antenna size, increasing useful light penetration into the culture. Indeed, in several cases, mutants with reduced pigment complexes have been shown to have higher specific rates of hydrogen production under some circumstances. Various aspects of hydrogen production by sulphur-deprived systems have been examined, including immobilization, different sulphur deprivation regimes, and the effects of mutations in starch metabolism. For systems based on sulfur deprivation it would seem that lowered efficiencies are unavoidable since this process is based upon reduction in PSII activity to allow oxygen removal through respiration, permitting sustained functioning of hydrogenase. However, to achieve this, photosynthetic rates must be drastically reduced, and respiration must be fed with fixed carbon compounds, leading to decreases in light conversion efficiencies. Therefore, the sulphur-deprived system is incapable of achieving efficiencies that would support the development of a practical system.

Biohydrogen

131

One solution which has been proposed and studied to some extent is to use a hydrogen evolving enzyme that is relatively stable in the presence of oxygen. However, so far no naturally occurring hydrogenase has been found which is sufficiently oxygen resistant and so efforts have turned to trying to create one. A number of strategies might be employed. For one, understanding the molecular details of oxygen inactivation, and, in particular, the access and movement of gases, oxygen and hydrogen, through the hydrogenase enzyme could permit the design and engineering of a more oxygen-tolerant hydrogenase. Another approach would be to use molecular evolution and gene shuffling techniques with the selection of more oxygen tolerant enzymes. For both cyanobacteria and green algae, yields and longevity of hydrogen production need to be increased without a significant reduction in the fitness of the culture. A number of metabolic engineering changes could be made in attempts to do this. One possibility is to create PSII mutants with lowered activity. In principle this might allow for hydrogen production with green algae without the need for debilitating sulfur deprivation, which also often requires significant physical manipulations. Indeed, use of a Chlamydomonas reinhardtii cy6Nac2.49 mutant that activates photosynthesis in a cyclical manner, demonstrated that controllable expression of PSII can be used to increase hydrogen production under nutrient replete conditions.12 Another approach is to downregulate or eliminate metabolic pathways that compete for photosynthetic reducing power with hydrogenase. In these organisms, carbon fixation by Rubisco is the major sink for highly reducing electrons. Several different strategies for reducing electron utilization by Rubisco are being investigated. In one study, hydrogen production was increased 10- to 15-fold when a mutated small subunit of Rubisco was expressed during sulfur deprivation.

3.10.4

Dark Fermentation

Biohydrogen can also be obtained through a process called dark fermentation (DF) in which very diverse groups of microorganisms convert organic substrates into fermentation products, including hydrogen and carbon dioxide, in an oxygen and light free environment. DF can take place in very simple operational setups. However, maximization of hydrogen yields and production rates requires careful optimization of reactor configuration and process parameters. Dark fermentation is part of a complex process, anaerobic digestion (AD), in which organic matter can be totally oxidized for biofuel production and wastewater treatment. The first step (hydrolysis) of AD includes the breakdown of the complex substances, such as proteins, lipids and carbohydrates, into soluble and easily biodegradable monomers. Fermentative bacteria consume the monomers, e.g., glucose, sucrose, producing organic acids, alcohols, hydrogen and carbon dioxide. The general idea of DF fermentation, the factors influencing the process, its limitations and perspectives will be reviewed in this section.

3.10.4.1

General

In recent years many researchers worldwide have studied different biological ways to generate hydrogen. Diverse groups of microorganisms ranging from mesophilic to hyperthermophilic, including facultative (Enterobacter), strict anaerobes (Clostridium) and even aerobes (Bacillus), are capable of producing hydrogen from a variety of substrates, including simple sugars and various agricultural, industrial, or domestic waste streams.13,14 The production of hydrogen in DF happens because in the absence of external electron acceptors (anaerobiosis), some microorganisms reduce protons to hydrogen in order to eliminate excess electrons generated by metabolism. The hydrogen yield varies accordingly to the fermentation type. The maximum theoretical yield would be 4 mols H2 per mol of glucose if there was no microbial growth and acetate was the sole end-product. The reported yields are much lower than the theoretical values, with few exceptions. Even if it were be possible to achieve the maximum theoretical yield, it represents just 33% of the maximum hydrogen that could be extracted from one mol of glucose (12 mols H2/mol glucose). This occurs because the organic matter cannot be totally oxidized and remain in the medium in the form of organic acids and alcohols, a major disadvantage of dark fermentation (low COD removal).15 The DF process can be carried out by pure cultures, co-cultures, or mixed microbial consortia. Pure cultures have been widely used in a number of fundamental studies and, in general allow for a better understanding of a process on a metabolic and mechanistic basis. However, from a practical point of view, when wastes are used as carbon sources, pure cultures would not be costeffective at large scale due to the low product value, which precludes running a sterile operation. The main disadvantage to using mixed cultures is that the overall hydrogen yield can be decreased by the consumption of hydrogen by various microorganisms, including in particular methanogens, homoacetogens, sulfate reducers, etc.16 Therefore, it is important to pretreat inocula using a method that preferentially selects for spore-forming, hydrogen-producing microorganisms, and inhibits hydrogen-consumers. However, such a simple operation will not eliminate some spore-forming hydrogen-consumers, such as Desulfotomaculum, but on the other hand, it eliminates useful non-spore-forming hydrogen-producing bacteria such as Enterobacter and Citrobacter that might be present in the sludge. Diverse pre-treatment protocols have been tested: heat-shock, acidic/alkaline shocks, microwave irradiation, aeration, treatment with BESA (2-bromoethanesulfonic acid) or chloroform; however, it is not yet possible to say which one is more effective. Moreover, there is also some interest in attempting to isolate new organisms that might possess special properties on their own, or that might be useful in constructing co-cultures or artificial microbial consortia. Besides inoculum source, substrate type is another key factor of the process. As noted previously, hydrogen production from simple sugars has been widely investigated but is economically unfeasible. Therefore, recent studies are testing the possibility to extracting hydrogen from carbohydrate-rich materials such as agricultural, agro-industrial and food wastes. Lignocellulosic wastes

132

Biohydrogen

(e.g., corn stover, barley and wheat straw), which are very abundant worldwide, have the potential to be used for hydrogen production after certain pre-treatments, which are, however, still costly and need further development. A variety of pre-treatments are available: mechanical and thermal pre-treatments (physical)17, acidic and alkaline shocks (chemical), steam explosion (physicochemical), treatment with fungi enzymes (biological). Physico-chemical methods carried out at very high temperatures generate compounds (phenolic compounds and furfurals) that can strongly inhibit dark fermentation. Biological technologies of lignocellulose deconstruction do not pose this problem, but take much longer and are more costly. DF is a process that can take place in a very simple setup, anaerobiosis being the only requirement if using a strict anaerobe as inoculum; however, this requirement can be neglected if using a co-culture (facultative and anaerobe) or a mixed culture, because under these conditions oxygen can be rapidly consumed. Nevertheless, in order to improve process efficiency, many different types of bioreactor designs have been used to investigate biological hydrogen production, including well-known configurations, such as CSTR (continuous stirred-tank reactor), AnSBR or ASBR (anaerobic sequencing batch reactor), EGSB (expanded granular sludge bed) and UASB (upflow anaerobic sludge blanket). Other novel reactor’s configurations have also been tested: ADSBRs (anaerobic downflow structured-bed reactors), AnMBRs (anaerobic membrane bioreactors), AFBR (anaerobic fluidized-bed reactors), APBR (anaerobic packed-bed reactors), CMTR (continuous multiple tube reactor). Each reactor configuration has its own specific advantages and disadvantages, suggesting particular uses. Many studies are carried out in a batch mode; however, fed-batch and continuous mode are also applied. The latter operational strategies overcome potentially toxic inhibition by the substrate or by-products, and thus present an interesting alternative when highly concentrated or complex feedstocks are used as substrates.18 Inoculation strategy is also a parameter that is under investigation. In many studies the DF process is carried out using suspended cells, which permits better mixing, however, it can also lead to biomass wash out when short hydraulic retention times are used. Therefore, immobilizing the biomass on some sort of support material is an alternative which permits cellular retention even under high flow rates. Cell immobilization can be achieved through a number of techniques; adsorption, encapsulation, and entrapment, with adsorption being the one most commonly used. A variety of support materials have been used; granular activated carbon, polyester fibers, low-density polyethylene granules, expanded clay, porous ceramic, ground tire, etc, for cell immobilization and, there seems to be a correlation between the type of support material used and the characteristics of the biofilm formed. Bioaugmentation, a technique that consists of adding desired, specialized and actively growing microbial strains to native microflora, has been suggested as a strategy to overcome issues related to instability, particularly during long-term continuous hydrogen production.19 This approach can be used to enrich the reactor with a known microorganism giving high yields and hydrogen production rates, thus attempting to allow hydrogen-producing bacteria to dominate the bacterial community over hydrogenconsumers. Furthermore, the addition of microorganisms capable of degrading complex biomaterials can potentially increase hydrogen production by facilitating the breaking down of such otherwise unassimilable substrates into readily assimilated ones. Bioaugmentation also seems advantageous because agricultural and industrial wastes that are used as substrates contain indigenous microorganisms that could easily compete with hydrogen producing microbes. Furthermore, bioaugmentation appears to enhance hydrogen production during the start-up phase. Another strategy to improve substrate conversion yields and production rates is the construction of metabolically engineered microorganisms, an approach that has become feasible for most organisms due to the dramatic improvements in genetic engineering techniques over the past two decades. Notably, metabolic engineering also requires a qualitative and quantitative analysis of the metabolic pathways of the microorganism to be modified. The desired changes can be introduced either by modifying an existing pathway or through the introduction of a new one. A typical approach is the genetic modification of an existing pathway with the aim of increasing the conversion of a wider range of monomers to key metabolic intermediates (e.g., pyruvate). Another possibility would be knocking out a gene that encodes uptake hydrogenases, an enzyme responsible for hydrogen oxidation.

3.10.4.2

Factors Affecting Dark Fermentation

As with any industrial biotechnological process, process efficiencies, substrate conversion yields, and volumetric production rates can usually be improved through the judicious determination of the optimal operational conditions. A variety of known factors that impact overall hydrogen yields and rates can be targeted: pH, temperature, availability of macronutrients and micronutrients, HRT (hydraulic retention time), OLR (organic loading rate), as well as the liquid (organic acids) and gaseous (H2) by-products of DF20,21. One parameter that has been examined is the influence of temperature, and its role in controlling fermentations occurring at various possible temperature ranges; mesophilic (37
C), thermophilic (55
C) and hyperthermophilic conditions (70
C). In reality, although one particular range might appear to offer advantages in some cases over others, the reactors, and thus the type of fermentation involved, would most likely be a function of the waste substrate to be used. For example, high-temperature waste streams could be subjected to a thermophilic fermentation, with its several advantages, possibly higher yields, reduced contamination risk, etc., without incurring a heating penalty which would otherwise have to be applied. pH is another parameter with significant influence on fermentative hydrogen production. Fermentation of carbohydrates typically leads to the production of organic acids in addition to hydrogen, potentially acidifying the culture sufficiently to inhibit partially or completely further hydrogen production by inducing metabolic pathway change or by impacting on cell viability. There is no consensus which pH value is the ideal for improving the process efficiency; however, it seems that the pH 5.5–6.5 range may be useful in many cases in maintaining high yields and rates of hydrogen production while avoiding low pHs which trigger solvetogenesis, reducing hydrogen evolution.

Biohydrogen

133

As the generation of liquid products (organic acids) influences the efficacy of the production of hydrogen, the accumulation of that gas in the headspace of the bioreactors also plays an important role on the microbial metabolism. At high hydrogen partial pressure, the concentration of the gas increases in the liquid phase and then takes place the reduction of ferredoxin. In this case, the bacterial metabolism shifts to the production of reduced compounds, lactate, ethanol, acetone and butanol, reducing overall hydrogen production. A possible alternative for solving this problem would be applying vigorous mixing or using a gas sparging technique (sparging with CO2 or N2), which, however, dilutes the produced hydrogen.1 Within a certain limit/threshold, it has been reported that the higher the substrate concentration the higher the volumetric hydrogen production rate; however, in many studies the yield follows the opposite trend, it decreases with increasing substrate concentration. The negative effect of high organic loading in the systems can occur due to the accumulation of organic acids, low pHs and high hydrogen partial pressure. It’s also possible that toxic compounds present in the wastes are also responsible for process inhibition rather than the carbohydrate content itself. This is the case for residues that contain high amounts of macro and micronutrients (e.g., sugarcane stillage), or the presence of toxic compounds produced during biomass pre-treatments. There is no consensus on the ideal substrate concentration needed to improve the hydrogen yield and it seems to be very dependent on the type of substrate and the cell concentration (F/M ratio). An alternative to suppress the negative effect of high OLR would be to maintain a high biomass concentration using immobilization techniques. All the mentioned parameters can contribute to a selection/enrichment of a specific microbial community. Therefore, various genetic molecular analyses have been performed in order to characterize the microbial population present in different inoculum sources, to monitor the bacterial diversity according to the changes in the operational parameters (e.g., temperature, pH, HRT, or OLR), to visualize its special distribution in granular sludge, and to quantify specific microbial populations. By using these techniques, it is possible to establish correlations between the composition and structure of a specific microbial community and its influence on the overall process. It is also possible to identify new indigenous strains that could have great potential as hydrogen-producers. Recently, the routine use of molecular biology analyses has been more generally used given the significant cost reductions over the past decade. Various culture-independent molecular methods have recently been tested, such as Cloning-sequencing Libraries, Denaturing Gradient Gel Electrophoresis (DGGE), Terminal Fragment Length Polymorphism (T-RFLP), High-Throughput Pyrosequencing, quantitative real-time PCR (qrt-PCR), etc.

3.10.4.3

Limitations of Dark Fermentation for Practical Application

Still the limitation on the hydrogen production through dark fermentation is the efficiency of the process in terms of yield and hydrogen production rate. As mentioned previously, in most cases hydrogen yields are much less than the theoretical values. Even if these were improved somewhat, it would still necessary to polish the dark fermentation effluent since in the organic matter in the waste stream is not completed oxidized, leaving reduced compounds (COD). Thus, it’s mandatory to have a second system in which the organic acids and alcohols produced could be consumed. A second step could be either a methanogenic reactor, for methane production, or a hydrogen production photo-fermentative system. In both cases, the addition of a second step makes the process laborious and costly in comparison to a single stage. Furthermore, methanogens preferences for substrates may mean that not all the organic acids generated are consumed. On the other hand, photo-fermentative bacteria are more flexible in terms of nutritional preferences; however, the light input requirement for a photofermentative process is a definite disadvantage. Genetically engineered microorganisms that could be more efficient than wild type strains are under investigation. As well, efforts are underway to isolate indigenous, robust/resilient, high hydrogen yield and rate hydrogen microorganisms. Their use in a bioaugmentation strategy could be an interesting alternative. How to produce hydrogen production from cheap and abundant substrates in ways that don’t compete with food production is another challenge to be addressed. Production from lignocellulosic substrates fulfills this goal, but requires the development of low cost pre-treatment methods which eliminate the inhibitory effects of the toxic compounds produced during the treatment. Furthermore, substrate manipulation prior to dark fermentation should be kept to a minimum. Obviously, sterilization methods, filtration or autoclave, used to eliminate harmful contaminant microorganisms (e.g., hydrogenotrophic methanogens, homoacetogens, sulphate-reducing bacteria, nitrate-reducing bacteria and lactic acid bacteria) cannot be used due to the high energy costs potentially involved.

3.10.5

Photofermentation by Photosynthetic Bacteria

3.10.5.1

General

Photofermentation (PF) is a biological hydrogen production process that can be realized mostly by purple non-sulfur bacteria (PNSB) (Fig. 1). Purple non-sulfur bacteria are a diverse metabolic group of microorganisms capable of producing hydrogen under anaerobic light conditions.22,23 The use of PNSB offers several specific advantages over other biohydrogen production processes; lack of oxygen generation, the ability to use variety of substrates, a wide range of the light spectrum (400–950 nm) for can be used for photosynthesis, and metabolic versatility under a variety of conditions. Biomass-derived materials present a plentiful, renewable, and accessible resource for hydrogen production by PNSB, carried out either by pure cultures or various co-cultures.24 In this hydrogen production process, electrons produced metabolically from organic substrates from ferredoxin (Fd) to nitrogenase. Nitrogenase is the enzyme primarily involved in photofermentation, and it requires ATP (2ATP/e) for the reduction of

134

Biohydrogen

Photosystem Central Metabolism

eATP

Organic acids & Sugars

Reverse Electron Transport

eNitrogenase ADP

H2 Figure 1 Photofermentation is carried out by the photosynthetic bacteria under anaerobic and light illumination. The electrons produced by utilization of sugars and organic acids are sent to nitrogenase through reverse electron transport. The electrons coming from the results of central metabolism and light excitation of photosystem are transferred by the specific coenzymes to the nitrogenase. Nitrogenase catalyzes the conversion of protons into molecular hydrogen by using ATP.

protons to molecular hydrogen. Nitrogenase synthesis and activity is sensitive to the availability and concentration of oxygen and ammonia in the environment. Both depress hydrogen production by inhibiting the catalytic activity of nitrogenase. Photofermentative hydrogen production is carried out under anaerobic nitrogen-deficient conditions, presented in Eq. (1). 16ATP þ 8H þ 8e /4H2 þ 16ADP þ 16Pi

(1)

Biological hydrogen production through photofermentation is a clean and attractive technology since it uses organic wastes and solar energy. For hydrogen production, volatile fatty acids, such as acetic acid, lactic acid, and butyric acid, are effectively utilized by the photofermentative bacteria under nitrogen limited, anaerobic and illuminated conditions. Biohydrogen production from various organic acids through photofermentation is shown in Eqs. (2)–(4). Acetic acid : C2 H4 O2 þ 2H2 O þ light energy/4H2 þ 2CO2

(2)

Lactic acid : C3 H6 O3 þ 3H2 O þ light energy/6H2 þ 3CO2

(3)

Malic acid : C4 H6 O5 þ 3H2 O þ light energy/6H2 þ 4CO2

(4)

A great variety of substrates ranging from organic acids to proteins and carbohydrates have been used for hydrogen production via photofermentation. In addition, many different kinds of industrial, agricultural and municipal wastes are potential substrate sources. Depending upon the availability of feedstocks, biohydrogen production by photofermentation could be part of a centralized or distributed power supply for urban or rural areas. It is worth emphasizing that using organic wastes not only generates hydrogen but also contributes to waste removal. A maximum of 12 mol of hydrogen can be theoretically obtained in the utilization of 1 mol of glucose through DF or PF stages. Biohydrogen has been demonstrated successfully by photofermentation directly in a single step instead of dark fermentation (DF) or integrated systems.23 Recently, some novel technologies, developing biofilm of photosynthetic bacteria biofilms, advanced bioreactor designs, are being applied to enhance photofermentation.

3.10.5.2

Factors Affecting Hydrogen Production

The ultimate aim of research in biohydrogen is to establish and sustain a large-scale industrial hydrogen production system. To achieve this, each critical parameter influencing the overall process must be examined in detail by starting from lab-scale to pilot-scale studies. Numerous studies have attempted to understand the underlying reasons for low hydrogen yields and rates in hydrogen production operations. Since the overall reaction mechanism of hydrogen production by photosynthetic bacteria depends on the light received by the photosystem unit in the cell membrane, the source, distribution and intensity of light are of great importance.25 Thus, the rate of hydrogen can be highly influenced by light-dependent factors. The amount of illumination should be kept in a uniform manner throughout the surface of the bioreactor. Also, organic wastes may prevent light penetration due to their dark color, thereby leading lower hydrogen production.

Biohydrogen

135

Although most of the previous research focused on photobiological hydrogen production through microalgae-oriented systems, many photosynthetic bacteria were investigated for their hydrogen potential in the last years. Recently, studies in the area of biohydrogen have focused on large-scale photobioreactors for purple non-sulfur bacteria (Adessi et al., 2014). A large number of reactor designs and configurations are available. Among them, flat panel and tubular photobioreactors, have more suitable geometries and characteristics enabling a feasible photobiological production of hydrogen (Adessi et al., 2012). Besides, one of the most critical issues in PBR design is to ensure the prevention of hydrogen leakage from the bioreactor. Temperature is a very critical parameter in terms of metabolic activities of photofermentative bacteria throughout the hydrogen production process. pH control is also a key factor influencing the culture and the sustainability of the overall process. Therefore, the pH of the system must be regulated with a convenient buffer system as the changes in pH could adversely affect the microbial metabolism. The type and concentration of carbon source have considerable impacts on hydrogen production performance. Organic acids such as lactate, malate, and acetate have been mostly used as feedstock for the hydrogen generation through photofermentation. In addition, carbohydrates, especially glucose and sucrose, are considered promising sources. Each carbon source has its own metabolic pathway, and feeding strategies and processes should be optimized with regards to critical factors to increase the efficiency of hydrogen production. In addition, the nitrogen source used is vital for the growth of the culture and to maintain the microbial culture for long-term hydrogen production. Thus, nitrogen availability in the bioreactor medium determines the fate of the whole system. However, the quantity of nitrogen must not be considered as a sole factor in the process. Carbon to nitrogen (C/N) ratio also affects biohydrogen performance in a bioreactor. This concept must be considered in feeding strategies during operation of the system. It is worth noting that photofermentative bacteria require some microelements during growth and hydrogen generation. Fe and Mo are major required elements for photoheterotrophic purple non-sulfur bacteria for maintenance and hydrogen production.

3.10.5.3

Strain Improvement

Although they share common features, each strain of purple non-sulfur bacteria has its unique structural and metabolic diversity. It is worth noting that an efficient microorganism is quite vital to make an economically viable hydrogen production system. Notably, genetic modifications of metabolic pathways could enhance the capabilities for hydrogen generation. Also, metabolic engineering can provide significant improvement by introducing new pathways and eliminating pathways that compete with hydrogen generation. A number of strategies might be employed to enhance hydrogen production.25 Blocking the reductant pathways in the cell may enhance the electron flow in favor of proton reduction. Inactivation or elimination of the Calvin Benson Basham cycle (CBB) by mutating specific genes could also be an approach to improve substrate conversion and hydrogen efficiency. One attempt could be blocking the polyhdroxybutyrate (PHB) synthesis pathway to divert the electron flow of the cell in favor of hydrogen production. The interruption or inactivation of uptake hydrogenase by mutation could enhance hydrogen yield as molecular hydrogen is consumed by hydrogenase itself. Besides, hydrogenases from various species can be genetically transferred to the desired organism as the turnover rates of hydrogenases are higher than that of nitrogenases (Hallenbeck et al., 2012). Another approach could be reduction of photosynthetic pigment content of the cells via mutation in order to increase light conversion efficiency by allowing for deeper light penetration into the culture.

3.10.5.4

Co-cultures

Research in biohydrogen has demonstrated that both dark fermentation or photo-fermentation have drawbacks, such as insufficient substrate conversion and low hydrogen rates, when they are used alone. To overcome this problem, dark fermentative bacteria (DFB) can be grown simultaneously with photofermentative bacteria (PFB) in a single photobioreactor (Fig. 2). The hydrogen yield could be increased up to 12 mol H2/mol glucose by the co-culture systems. The main aim of using a co-culture is that organic acids produced by DFB can be used as a substrate by PFB for hydrogen production resulting in complete substrate utilization (Eqs. 5–6). Therefore, a combined dark and photofermentation process can possibly be used to achieve maximum hydrogen yields and rates. Furthermore, the use of a co-culture system enables a reduction in the total reactor volume, reduces the overall fermentation time, and lowers the overall cost with a simplified reactor operation.26 Dark fermentation by selective anaerobes: C6 H12 O6 þ 2H2 O/2CH3 COOH þ 4H2 þ 2CO2

(5)

Photo-fermentation by photosynthetic bacteria: 2CH3 COOH þ 4H2 O þ light energy/8H2 þ 4CO2

(6)

Overall H2 yield by co-culture: 4H2 þ 8H2 ¼ 12H2 A variety of microorganisms could be used together, and their interactions affect and determine the performance of the whole process during their growth and hydrogen production. The most widely used microorganisms belong to the genera Clostridium and Rhodopseudomonas for DF and PF, respectively.

136

Biohydrogen

Organic wastes

Pre-treatment Cellulose Sucrose Starch

Dark Fermentation

Organic acids

Photofermentation

H2

H2 Co-culture

Figure 2 Co-culture of dark and photofermentation. In a co-culture system, both dark and photofermentative bacteria are cultivated in a single bioreactor for hydrogen production.

Co-culture systems permit the use of high concentrations of substrates at start-up of the process as the short chain fatty acids, produced in the DF stage as the results of the substrate degradation, can be consumed by the photosynthetic bacteria in situ, thereby eliminating the need for dilution prior to utilization. As for the DF and PF systems, diverse organic agro-industrial wastes can be used as substrates (Dasgupta et al., 2010). Nevertheless, it’s necessary to point out that fixed nitrogen, if present at high levels, can inhibit hydrogen production (Hallenbeck & Liu, 2016). Another aspect that should be taken into account is the faster growth rate of DFB in comparison to PFB. The imbalance between these growth rates can cause a rapid culture medium acidification making the environment unsuitable for photo-fermentation. Possible solutions could be the use of an appropriate buffer, or the use of a high concentration of PF cells; however, excess biomass could lead to self-shading issues. Since those microbial groups have different nutritional requirements the addition of micronutrients (Fe, Mo, Ni) could be necessary for the overall process (Keskin et al., 2011). Also, it has been shown that hydrogen production yields can be improved by the co-culture technique, with achievement of 70%–80% of the theoretical value. The selection and combination of certain microorganisms and the determination of the optimal culture conditions for both of them can optimize process efficiency by enabling them to develop their maximum rates.

3.10.5.5

Limitations

Although recent large-scale studies in photofermentative hydrogen production have been encouraging, there are still limitations for practical application. The limitations are derived from either individual parameters or combination of multiple critical factors. The parameters affecting hydrogen performance must be considered at each level from lab scale to large industrial scale. The main challenges to the commercialization of biohydrogen are the low hydrogen production rates and yields. Obviously, commercialization of photofermentative hydrogen production is still in the future and practical and sustainable use of biohydrogen will require novel approaches, ultimately perhaps permitting decentralized hydrogen production in certain locations where organic wastes are abundant and readily available (Adessi and De Philippis, 2014). One key barrier to industrial scale photofermentative hydrogen production is a low photosynthetic efficiency leading to low hydrogen rates. In addition, as photofermentation depends upon solar energy, the natural day/night cycle limits the overall efficiency of hydrogen production. Current strategies to improve biohydrogen production include appropriate culture selection, process optimization, genetic engineering of productive strains, and development of low-cost effective photobioreactors.23 Oxygen and ammonia sensitivity of the nitrogenase is also a limiting factor, particularly in large-scale photobioreactors. Overall, the technical and economic aspects must be considered together when evaluating the future prospects of photofermentative hydrogen.

3.10.6

MECs

3.10.6.1

General Mechanism

There has been great interest in microbial electrolysis cells (MECs), especially in the last decade. In a MEC system, carbon dioxide, protons and electrons are produced as the result of microbial activities at the anode (Fig. 3). These electrons are then transferred to

Biohydrogen

CO2 e-

Influent

Microorganism

Effluent

A n o d e

Power Supply

137

H2 e-

H+

H+

C a t h o d e

Proton Exchange Membrane

Figure 3 A general scheme of a typical microbial electrolysis cell (MEC). Microorganisms transfer electrons as the result of substrate oxidation in the anode chamber. The protons diffuse from a proton exchange membrane (PEM) to cathode chamber in which protons are reduced to form molecular hydrogen. Carbon dioxide is evolved as a side product of cell metabolism. A power supply (electrical energy) is employed for the required endothermic energy in the system.

the cathode through a wire. When an external voltage (0.2–0.8 V) is applied, hydrogen is evolved at the cathode. Using such as system, hydrogen can be generated from acetate by a MEC system (Eqs. 7–8). CH3 COOH þ 2H2 O/8Hþ þ 8e þ 2CO2 ðanodeÞEanode : 0:28V

(7)

8Hþ þ 8e /4H2 ðcathodeÞEcathode : 0:42V

(8)

As the result of metabolic activities, electrons produced by microorganisms are transferred to the anode. The transfer of electrons from cells to the anode is highly critical and several factors determine the efficiency of the transfer. The type of material, surface area, charge, conductivity, compatibility, stability, other physical and chemical characteristics of the electrodes are major parameters in an efficient MEC system. The anode electrodes are made mostly of carbon- or graphite-based materials and fabricated in response to the desired properties. Platinum and carbon are commonly used materials for cathode electrodes. In general, most MECs are twochamber devices, with a proton exchange membrane between the MEC anode and cathode chambers. These membranes, which facilitate the transfer of protons from the anode to the cathode, can be made from different materials and are either cationic or anionic. Although most of the MEC reactors used so far have been two-chambered in order to prevent hydrogen escape to the anode where it could possibly be consumed. However, single chamber MEC systems without a membrane have been recently designed and configured, resulting in a significant improvement in hydrogen rates.27,28 A variety of MEC reactors of different size and shape have been designed and operated for hydrogen production so far. Despite the various designs and configurations, most of the MEC reactors are still not suitable for a large-scale sustainable hydrogen production system. Among the several technical issues is proper electrode selection, since it is desirable to have a high surface/volume ratio, while at the same time maintaining low cost. The operation of an MEC starts with the selection of the microorganisms used in the anode chamber. The microorganisms for inoculation could be either a pure culture or mixed wastewater seed. After inoculation of the microorganism at the anode, a power is provided in the MEC for production of molecular hydrogen at the cathode. However, the applied voltage should be carefully adjusted and must not exceed the threshold in order not to affect the metabolism of the bacteria adversely. There are many factors affecting the efficiency of hydrogen production in MEC systems. The type of substrate and its concentration are highly critical for the operation of MEC reactors. Short-chain volatile fatty acids (e.g., acetate) or dark fermentation effluents have been demonstrated to serve as substrate and potentially promote anodic biofilm formation. As well, hydrogen production from various organic wastes, including glucose and cellulose, has been examined in a combined dark fermentation and MEC system for hydrogen production. The combination of dark or photofermentation with microbial electrolysis cells can increase the overall hydrogen yield as the fermentation products could be used simultaneously. This ensures the possibility of using organic wastes including industrial and agricultural for waste treatment or biotechnological production purposes.

3.10.6.2

Limitations

Most of the attention and efforts in microbial electrolysis cell (MEC) studies have been given to the optimization of the process and configuration of the system. Although some of the advantages of microbial electrolysis cell technology for enhanced hydrogen

138

Biohydrogen

production have been demonstrated, the current status of this methodology is not yet sufficient for its practical application. Further studies development, including advances in design and construction of suitable cells, electrode material, buffers, and optimization of external voltage applied must be considered in order to enhance hydrogen yields and rates (Hallenbeck, 2014). Wastewater treatment can also be accomplished using microbial electrolysis cells after full optimization of this system. Practical application of this technology will require the design and configuration of a highly efficient and low-cost MEC system. In fact, several critical parameters influence and restrict the application of MEC system. They can be either structural (e.g., electrode material, design) or functional (e.g., surface charge, biocompatibility) depending on the configuration of the platform. Organic wastes which will be used in MEC operations must be well characterized and the overall optimization of process is required for a large-scale system. The hurdles in start-up of MEC systems limit its implementation and up scaling. One recent advance is the development of biocathodes made from microorganisms which provide an interesting alternative to metallic cathode catalysts. The electro–catalytic activities of the microorganisms play important roles in the efficiency of the MEC system. Also, the interaction of biofilm with electrodes, substrate composition and concentration determine hydrogen production performance of the MEC.

3.10.7

Closing Remarks

As the effects of climate change due to fossil fuel combustion become more and more apparent daily, it has become obvious that drastic changes in energy sources and utilization are required to avoid even more disastrous effects. One approach that could potentially go a long way towards the decarbonization of the energy economy is the use of sustainably generated hydrogen as energy carrier. Biological hydrogen production offers a variety of means for accomplishing this, from photosynthetic microorganisms that are capable of using water and sunlight as substrates to produce hydrogen, to photosynthetic bacteria, capable of degrading a variety of organic wastes to hydrogen with the input of solar energy, to hydrogen production in the dark from various waste streams, either through some type of fermentation or through the use of MECs. As discussed here, each system has its advantages but also particular challenges that must be overcome to reach market deployment. A variety of strategies are being used in attempts to enhance the yields and rates of hydrogen production; metabolic engineering, process parameter optimization, strain isolation and bioaugmentation, and the development of two-stage systems. In the near to medium term some improvements in efficiencies should be seen, and, given the relatively active effort in this area, major advances in one or more areas may also occur.

References 1. Hallenbeck, P. C. Bioenergy from Microorganisms: an Overview. In Microbial BioEnergy: Hydrogen Production; Zannoni, D., De Philippis, R., Eds.; Advances in Photosynthesis and Respiration (Including Bioenergy and Related Processes), vol 38; Springer: Dordrecht, 2014; pp 3–21. 2. Dincer, I.; Acar, C. Review and Evaluation of Hydrogen Production Methods for Better Sustainability. Int. J. Hydrogen Energy 2015, 40 (34), 11094–11111. 3. Van Niel, E. W. J. Biological Processes for Hydrogen Production. Adv. Biochem. Eng. Biotechnol. 2016, 156, 155–194. 4. Hallenbeck, P. C. Fermentative Hydrogen Production: Principles, Progress, and Prognosis. Int. J. Hydrogen Energy 2009, 34 (17), 7379–7389. 5. Hallenbeck, P. C.; Abo-Hashesh, M.; Ghosh, D. Strategies for Improving Biological Hydrogen Production. Bioresour. Technol. 2012, 110, 1–9. 6. Abo-Hashesh, M.; Wang, R.; Hallenbeck, P. C. Metabolic Engineering in Dark Fermentative Hydrogen Production; Theory and Practice. Bioresour. Technol. 2011, 102 (18), 8414–8422. 7. Liu, H.; Hu, H. Microbial Electrolysis: Novel Biotechnology for Hydrogen Production from Biomass. In Microbial Technologies in Advanced Biofuels Production; Hallenbeck, P. C., Ed., Springer, 2012; pp 93–106. 8. Hallenbeck, P. C.; Lazaro, C. Z.; Sagir, E. Chapter 1 Photosynthesis and Hydrogen from Photosynthetic Microorganisms. In Microalgal Hydrogen Production: Achievements and Perspectives, The Royal Society of Chemistry, 2018; pp 1–30. 9. Batyrova, K.; Hallenbeck, P. C. Sustainability of Biohydrogen Production Using Engineered Algae as a Source. In Biohydrogen Production: Sustainability of Current Technology and Future Perspective; Singh, A., Rathore, D., Eds., Springer India: New Delhi, 2017; pp 163–180. 10. Nagarajan, D.; Lee, D. J.; Kondo, A.; Chang, J. S. Recent Insights into Biohydrogen Production by Microalgae - from Biophotolysis to Dark Fermentation. Bioresour. Technol. 2017, 227, 373–387. 11. Torzillo, G.; Scoma, A.; Faraloni, C.; Giannelli, L. Advances in the Biotechnology of Hydrogen Production with the Microalga Chlamydomonas reinhardtii. Crit. Rev. Biotechnol. 2015, 35, 485–496. 12. Batyrova, K.; Hallenbeck, P. C. Hydrogen Production by a Chlamydomonas reinhardtii Strain with Inducible Expression of Photosystem II. Int. J. Mol. Sci. 2017, 18, 647. 13. Kapdan, I. K.; Kargi, F. Bio-hydrogen Production from Waste Materials. Enzym. Microb. Technol. 2006, 38 (5), 569–582. 14. Ghimire, A.; Frunzo, L.; Pirozzi, F.; Trably, E.; Escudie, R.; Lens, P. N. L.; Esposito, G. A Review on Dark Fermentative Biohydrogen Production from Organic Biomass: Process Parameters and Use of By-products. Appl. Energy 2015, 144, 73–95. 15. Lin, C. Y.; Nguyen, T. M. L.; Chu, C. Y.; Leu, H. J.; Lay, C. H. Fermentative Biohydrogen Production and its Byproducts: A Mini Review of Current Technology Developments. Renew. Sustain. Energy Rev. 2018, 82, 4215–4220. 16. Hung, C.-H.; Chang, Y.-T.; Chang, Y.-J. Roles of Microorganisms Other than Clostridium and Enterobacter in Anaerobic Fermentative Biohydrogen Production Systems a Review. Bioresour. Technol. 2011, 102 (18), 8437–8444. 17. Rafieenia, R.; Lavagnolo, M. C.; Pivato, A. Pre-treatment Technologies for Dark Fermentative Hydrogen Production: Current Advances and Future Directions. Waste Manag. 2018, 71, 734–748. 18. Sivagurunathan, P.; Kumar, G.; Bakonyi, P.; Kim, S.-H.; Kobayashi, T.; Xu, K. Q.; Lakner, G.; Tóth, G.; Nemestóthy, N.; Bélafi-Bakó, K. A Critical Review on Issues and Overcoming Strategies for the Enhancement of Dark Fermentative Hydrogen Production in Continuous Systems. Int. J. Hydrogen Energy 2016, 41 (6), 3820–3836. 19. Goud, R. K.; Sarkar, O.; Chiranjeevi, P.; Mohan, S. V. Bioaugmentation of Potent Acidogenic Isolates: A Strategy for Enhancing Biohydrogen Production at Elevated Organic Load. Bioresour. Technol. 2014, 165, 223–232. 20. Bundhoo, M. A. Z.; Mohee, R. Inhibition of Dark Fermentative Bio-hydrogen Production: A Review. Int. J. Hydrogen Energy 2016, 41 (16), 6713–6733. 21. Elbeshbishy, E.; Dhar, B. R.; Nakhla, G.; Lee, H.-S. A Critical Review on Inhibition of Dark Biohydrogen Fermentation. Renew. Sustain. Energy Rev. 2017, 79, 656–668.

Biohydrogen

139

22. Adessi, A.; De Philippis, R. Hydrogen Production: Photofermentation. In Microbial Technologies in Advanced Biofuels Production; Hallenbeck, P. C., Ed., Springer ScienceþBusiness Media: New York, USA, 2012; pp 53–75. 23. Hallenbeck, P. C.; Liu, Y. Recent Advances in Hydrogen Production by Photosynthetic Bacteria. Int. J. Hydrogen Energy 2015, 41, 4446–4454. 24. Keskin, T.; Abo-Hashesh, M.; Hallenbeck, P. C. Photofermentative Hydrogen Production from Wastes. Bioresour. Technol. 2011, 102 (18), 8557–8568. 25. Adessi, A.; De Philippis, R. Photobioreactor Design and Illumination Systems for H2 Production with Anoxygenic Photosynthetic Bacteria: A Review. Int. J. Hydrogen Energy 2014, 39 (7), 3127–3141. 26. Adessi, A.; De Philippis, R.; Hallenbeck, P. C. Combined Systems for Maximum Substrate Conversion. In Microbial Technologies in Advanced Biofuels Production; Hallenbeck, P. C., Ed., Springer, 2012; pp 107–126. 27. Kitching, M.; Butler, R.; Marsili, E. Microbial Bioelectrosynthesis of Hydrogen: Current Challenges and Scale-up. Enzym. Microb. Technol. 2017, 96, 1–13. 28. Zhen, G.; Lu, X.; Kumar, G.; Bakonyi, P.; Xu, K.; Zhao, Y. Microbial Electrolysis Cell Platform for Simultaneous Waste Biorefinery and Clean Electrofuels Generation: Current Situation, Challenges and Future Perspectives. Prog. Energy Combust. Sci. 2017, 63, 119–145.

3.11

Biofuel From Microalgae

Z Wen, Iowa State University, Ames, IA, United States J Liu, The University of Hong Kong, Hong Kong, China F Chen, The University of Hong Kong, Hong Kong, China; and Peking University, Beijing, China © 2011 Elsevier B.V. All rights reserved. This is a reprint of Z. Wen, J. Liu, F. Chen, 3.12 - Biofuel from Microalgae, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 127-133.

3.11.1 3.11.2 3.11.3 3.11.3.1 3.11.3.2 3.11.3.3 3.11.3.4 3.11.4 3.11.4.1 3.11.4.2 3.11.4.3 3.11.4.4 3.11.4.5 3.11.5 References

Introduction and Scope Major Algal Composition Different Types of Biofuels From Microalgae Biogas Ethanol Biodiesel Bio-Oil and Syngas Algal Biodiesel Production Pipeline Algal Physiology and Genetic Engineering Mass Algal Culture Algae Harvesting and Dewatering Biomass Processing for Oil Extraction Conversion of Algal Oil Into Biodiesel Conclusion and Perspectives

140 141 141 141 142 142 142 142 142 143 144 145 145 146 146

Glossary Biogas A gas mixture containing carbon dioxide and methane as major components that are generated through breakdown of organic matters by bacteria and/or archaea without oxygen. Bio-oil A synthetic liquid fuel that is extracted by treating biomass in a reactor at temperature of about 500  C without oxygen. Microalgae Commonly photosynthetic organisms that primarily use water, carbon dioxide, and sunlight to produce biomass and oxygen. Syngas A gas mixture that contains varying amounts of carbon monoxide and hydrogen that are generated by gasification of coal or biomass. Thermochemical conversion A process by which biomass is treated at high temperature with various catalysts to produce various liquid and/or gaseous fuels.

3.11.1

Introduction and Scope

Microalgae are mostly photosynthetic organisms that primarily use water, CO2, and sunlight to produce biomass and O2. The nutrients required for growing algae are nitrogen, phosphorus, mineral salts, trace elements, and silicon (for diatom). Most of those nutrients are available from municipal, industrial, and agricultural wastewater. Compared with terrestrial plants, microalgae have a high oil content and growth rate. Algal cells generally contain 4%–12% oil (dry basis) but can be as high as 77% depending on species and growing conditions.1,2 Mass cultivation of microalgae can be performed on unexploited lands using saline water in arid regions, thus avoiding competition for limited arable lands. Due to these merits, microalgae have long been considered a promising alternative and renewable feedstock for biofuel production. Depending on the biomass composition, microalgae can be processed into various types of biofuels including biogas, alcohol, biodiesel, and jet fuels. However, current algal biofuel production is still far from economical due to several major challenges such as low oil yield, high harvest cost, and the contamination of the native species.3 Developing an economic algal biofuel production requires a collaborative effort between algal biologists and bioprocess engineers. This article provides an overview of the current status of algal biofuel production. The algal biomass composition and the various types of biofuels that can be produced from algae are discussed. At last, we use algal biodiesel production as an example to illustrate the production chain elements of algal biodiesel production.

140

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00156-7

Biofuel From Microalgae

3.11.2

141

Major Algal Composition

Microalgal biomass contains carbohydrate, proteins, and lipids as major compositions. In general, proteins account for 40%–60% of dry biomass, followed by carbohydrate (20%–30%), and lipids (10%–20%).1 Syntheses of these components are highly regulated by the culture conditions. For example, under nitrogen-limitation and high light conditions, microalgal cells tend to accumulate lipid instead of the starch.4 In addition to these three major components, algal cells also contain small amounts (1%–5%) of nucleic acids,1 and various pigments such as carotenoids. From the biofuel production point of view, lipids (oils) are the most interesting group of components. In general, algal lipids are divided into two classes: neutral lipids and polar lipids. Triacylglycerol (TAG) is the major neutral lipid found in algae. In addition to TAG, algae also contain small amounts of other neutral lipids such as monoacylglycerol, diacylglycerol, and sterols. Polar lipids are more complex than neutral lipids, of which glycolipids and phospholipids are the two most important and popular groups. Lipid composition and content are important factors to assess the potential of algae as biodiesel feedstock. Over the past few decades, numerous algal species have been screened and characterized for their lipid production potentials. The lipid composition and content of these oleaginous algae are species- and/or strain-dependent and may vary greatly. Under optimal growth conditions, algae generally synthesize a small amount of lipids with polar lipids being the main components; whereas under unfavorable environmental or stress conditions, algae may accumulate large quantities of lipids with neutral lipids, particularly TAG as the major components. This might be due to the shift of lipid metabolism from membrane polar lipids to storage neutral lipids. Algae can produce lipids up to 77% of dry weight, with TAG accounting for as much as 80% of total lipids. The synthesized TAGs are deposited in lipid bodies located in cytoplasm of algal cells. Unlike higher plants in which individual classes of lipids may be synthesized and localized in a specific cell, tissue, or organ, algae produce these different lipids in a single cell. From a biodiesel production point of view, TAGs are preferred to phospholipids or glycolipids because of their high proportions of fatty acids and lack of phosphate. Algal fatty acids are in either saturated or unsaturated form, and the unsaturated fatty acids may vary in the number and position of double bonds on the acyl chain. Based on the number of double bonds, unsaturated fatty acids are classified into monounsaturated fatty acids and polyunsaturated fatty acids. Many algae have been investigated for their fatty acid profiles. The fatty acids of algae are commonly in medium length, ranging from 16 to 18 carbons, although composition of those fatty acids varies greatly. In general, the major fatty acids are C16:0, C18:1, and C18:2 or C18:3 in green algae, C16:0 and C16:1 in diatoms, and C16:0, C16:1, C18:1, and C18:2 in cyanobacteria. However, it should be noted that these data are obtained from algal species under specific conditions that may vary greatly when the algal cells are exposed to different environmental or nutritional conditions such as light intensity, temperature, and nitrogen concentration.

3.11.3

Different Types of Biofuels From Microalgae

3.11.3.1

Biogas

Anaerobic digestion is widely used for treating various waste streams such as municipal sludge or animal waste. For treating microalgae using anaerobic digestion, the digesting materials can be either raw algal biomass or the residue after oil extracted from the biomass. Methane produced from anaerobic digestion can be used as a heat source or for electricity generation. Anaerobic digestion process can also mineralize the organic nitrogen and phosphorus contained in the algal biomass, resulting in a flux of ammonium and phosphate that can be used as a substrate for microalgae, thus reducing the use of fertilizer in the microalgal culture. The use of raw algal biomass for methane production can avoid the biomass-harvest and oil-extraction processes used in algal biodiesel production, and significantly reduce the production cost and energy debt. Anaerobic digestion of the cell residues after lipid extraction is strongly recommended for balancing both the energy and economy of the algal biodiesel production. There are disadvantages of using anaerobic digestion for treating algal biomass. In general, algal cell contains a ‘tough’ cell wall that is difficult to be digested. The proteins contained in the biomass will release ammonia when degraded; a high level of ammonia can inhibit the microorganism in the digesters. The inhibition will become more severe when digesting the lipid extracted algal residues because the protein content is even higher. In addition, some marine algal species require high levels of sodium ions for growth. Nevertheless, sodium ions at high concentrations are strongly inhibitory to the anaerobic microflora.5 All these factors will reduce the methane yield when raw microalgal biomass is being digested. Indeed, it has been reported that the degradation rates of Chlorella and Scenedesmus species are only 60%–70% of that in active sludge digestions.6 In another study on anaerobic digestion of Chlorella vulgaris, it is found that 50% of the biomass does not undergo anaerobic digestion even at a long retention time of 28 days.6 To increase the anaerobic digestion efficiency, pretreatment of algal biomass is needed so the organic substrates in the algal cells are more accessible to anaerobic microflora and readily biodegraded. Various pretreatment methods that are developed for treating other waste materials such as animal waste and municipal sewage sludge can be applied to treating algal biomass. These include the physical treatment (mechanical maceration, ultrasonic lysis, and heat treatment), chemical treatment (acid, base, neutral detergent, and ozonation), and biological and enzymatic treatment. The high protein content of the algal biomass usually leads to a low C/N ratio, which is imbalanced for anaerobic digestion. For example, freshwater algae have an average C/N ratio of 10.2, while terrestrial plants have an average C/N ratio of 36. Co-digestion of algal biomass with other organic matters such as wastepaper to ensure a balanced C/N ratio of the influent composition can increase the digestion performance.7

142 3.11.3.2

Biofuel From Microalgae Ethanol

Certain microalgae are capable of producing high levels of carbohydrates such as starch or cellulose as reserve materials, which are ideal feedstocks for ethanol production. Compared to terrestrial plants, algae have a high photosynthetic efficiency and can synthesize and accumulate large quantities of carbohydrate biomass. When making ethanol from terrestrial biomass such as corn stover and switchgrass, a harsh pretreatment step is usually needed to break down the complex structure of those lignocellulosic materials, so that the cellulose can be converted via hydrolysis into fermentable sugars. Aquatic algal cells, however, are buoyant and do not contain those structural biopolymers such as hemicellulose and lignin. This greatly simplifies the algal bioethanol production process by eliminating the complex and expensive pretreatment steps.8 Ethanol from microalgae can be produced through the conventional method, that is, extract the starch or cellulose from the algal biomass, hydrolyze the starch/cellulose to be sugars, and ferment the sugars to produce ethanol by appropriate ethanol producers. First, the harvested algal cells are treated through mechanical means such as ultrasonic, explosive disintegration, mechanical shear, or enzymatic dissolution of cell walls. The starch is then extracted with water or an organic solvent. Once the starch is extracted, it can be further fermented to ethanol using the technology similar to other starch-based feedstocks, that is, saccharification and fermentation. This can be done through either a sequential step or a single step (simultaneous saccharification and fermentation). The ethanol is then purified by distillation to remove water and other impurities in the diluted alcohol product (10%–15% ethanol), and then condensed into concentrated form (95% ethanol). In addition to the above mentioned conventional methods, some algal species are capable of producing ethanol through a dark, anaerobic-based self-fermentation process. When microalgae grow in dark and in the presence of oxygen, the algal cells usually consume storage starch for their maintenance, with water and CO2 as the starch-decomposition products. Under anaerobic conditions, however, the decomposition is incomplete, and a variety of products such as hydrogen, CO2, ethanol, lactic acid, formic acid, and acetic acid can be produced. Based on this mechanism, dark anaerobic algal fermentation process is developed for ethanol production. For example, the green microalga Chlamydomonas reinhardtii produced around 1% (w/w) of ethanol with 30%–40% of the theoretical yield of 0.56 g ethanol/g of starch conversion rate through dark anaerobic fermentation.9 The alga Chlorococcum littorale is also reported to produce ethanol through dark anaerobic fermentation, and 27% of the cellular starch is consumed within 24 h at 25  C.10

3.11.3.3

Biodiesel

Algal oil is ideal for biodiesel production. Compared with plant-based oil, algal oil has relatively high carbon and hydrogen contents, and low oxygen content. These characteristics make algal oil attractive for biodiesel production because it may lead to high-energy content, low viscosity, and low density. The basic chemical reaction required to produce biodiesel is the esterification of lipids with alcohol. Glycerol is produced as byproduct. High lipid-containing algae are most desirable for biodiesel production, and the neutral lipids (TAGs) contained in the algal cells are an ideal feedstock for producing biodiesel. It is noticeable that some microalgae are capable of producing high levels of TAGs but their growth rates11 are relatively low. Many marine microalgal species may produce higher levels of phospholipids than TAGs. Phospholipids, however, are not desirable in the transesterification process. All these factors need to be carefully considered before a process for algal biodiesel production can be developed.

3.11.3.4

Bio-Oil and Syngas

The organic components present in algal biomass can also be converted into crude bio-oils or syngas fuels through thermochemical conversion processes. Depending on the temperature and the availability of oxygen, the thermochemical conversion process can be categorized as gasification, pyrolysis, and thermochemical liquefaction. The end products vary from gas, liquid, to solid fuel, depending on different processes used. In the gasification process, the carbonaceous materials in the algal biomass are converted into synthetic gas (syngas) by means of partial oxidation at a temperature ranging from 800 to 900  C. The major compositions in the syngas are CO2, CO, CH4, and H2. Ammonium is another major component of syngas for biomass with high nitrogen content.12 Pyrolysis (particularly the fast pyrolysis) converts biomass into bio-oil, charcoal, and gaseous fraction by heating the algal biomass at around 500  C in the absence of the oxygen. A drying process is usually needed prior to the pyrolysis in order to save the energy used. Liquefaction is usually performed in an aqueous solution of alkali or alkaline salt at around 300  C and 10 MPa. The advantage of liquefaction is that wet biomass can be directly treated without involving a drying process. The major products of the liquefaction are bio-oils. The gaseous phase of the liquefaction contains CO2, but not H2 and CO.13

3.11.4

Algal Biodiesel Production Pipeline

3.11.4.1

Algal Physiology and Genetic Engineering

Phototrophic microalgae require several things to grow, including a light source, carbon dioxide, water, and inorganic salts. Algal lipid production depends largely on the growth conditions, including the nutrients, temperature, light intensity, growth phase, and

Biofuel From Microalgae

143

physiological status. The growth medium must contain inorganic elements that help make up the algal cell, such as nitrogen, phosphorus, iron, and sometimes silicon. In general, oleaginous algae produce only small quantities of neutral lipids (TAG) under optimal growth conditions. Syntheses and accumulation of TAG are facilitated by placing the algae under stress conditions that are imposed by chemical (nutrient limitation, salinity, and pH) or physical (temperature and light intensity) stimuli. Of those nutrients, nitrogen is the most crucial factor influencing lipid metabolisms. An increase in lipid/TAG accumulation under nitrogen limitation conditions has been observed in numerous algal species. In diatoms, silicon is another important nutrient affecting lipid metabolism. It has been reported that silicon-deficient Cyclotella cryptic cells had a higher neutral lipid than siliconreplete cells.14 Phosphate and sulfate limitation also promote lipid accumulation for certain algal species. Temperature influences algal lipid production through altering the fatty acid composition. A general trend is that low temperature tends to increase the unsaturation of the fatty acids and vice versa. By contrast, however, there is no general trend in the effects of temperature on the total lipid production by microalgae. Light intensity also significantly influences algal lipid production. Typically, low light intensity induces the formation of polar lipids, particularly the membrane polar lipids associated with chloroplast, whereas high light intensity decreases total polar lipids and increases the neutral storage lipids. In addition to the use of traditional approaches such as algal physiology for lipid production, genetic engineering is another important means that may improve algal productivity and, thus, the economics of algal biodiesel production. Understanding the lipid biosynthesis is of great help to genetically engineer algal lipid production. Theoretically, overexpressing the genes involved in fatty acid synthesis would be able to increase lipid accumulation as fatty acids are the precursors for lipid biosynthesis. Because neutral lipids (TAG) are the most preferred type of lipids for biodiesel production, increasing TAG/total lipids ratio or cellular TAG content through genetic engineering is the most focusing area. Overexpressing genes involved in TAG assembly has been found to significantly increase TAG production in higher plants. Such strategies may also be applied to microalgae. Commonly, microalgae produce large amounts of lipids under unfavorable conditions, which go beyond the log growth phase. The enhanced lipid biosynthesis through genetic engineering, therefore, is likely to reduce algal proliferation and biomass production. In such a case, the genes involved in lipid biosynthesis need to be overexpressed after the target algae have achieved a high cell density and entered the stationary phase. Another feasible approach to increasing the cellular lipid content is inhibiting metabolic pathways that lead to other carbon storage compounds, such as starch. Starch synthesis shares common carbon precursors with lipid synthesis in algae. Blocking starch synthesis is able to redirect carbon flux to the lipid biosynthetic pathway, resulting in the overproduction of fatty acids and thus total lipids.15 The important properties of biodiesel, such as cetane number, viscosity, cold flow, and oxidative stability, are largely determined by the characteristics of biodiesel feedstocks such as the carbon chain length and unsaturation degree of the fatty acids of the oil.16 Thus, the genetic modification of algal fatty acid composition is of interest. Generally, saturated fatty esters possess high cetane number and superior oxidative stability; whereas unsaturated, especially polyunsaturated, fatty esters have improved lowtemperature properties.17 It is suggested that the modification of fatty esters, for example, the enhanced proportion of oleic acid (C18:1) ester, can provide a compromise solution between oxidative stability and low-temperature properties.17 Oleic acid is converted to linoleic acid (C18:2) catalyzed by a D12 desaturase enzyme encoded by the FAD2 gene. Inactivation of this desaturation step can greatly increase the proportion of oleic acid in soybean and may represent a possible strategy for elevated accumulation of oleic acid in microalgae. Genetic engineering can also be used potentially to improve tolerance of algae to stress factors such as temperature, salinity, and pH. These improved attributes will allow for the cost reduction in algal biomass production and be beneficial for growing selected algae under extreme conditions that limit the proliferation of invasive species. Photoinhibition could also be addressed by genetic engineering. The adoption of engineered algae having a higher inhibition light threshold will significantly improve biodiesel production economics. Although the great potential of using engineered microalgae for biodiesel production has been proposed for some time, the research in genetic engineering of microalgae is still in its infancy. The lack of full or near-full genome sequences and robust transformation systems makes genetic engineering of algae lag much behind that of bacteria, fungi, and higher eukaryotes. Although certain algal species have been reported for efficient transformation, it proves to be difficult to produce stable transformants. Currently, sophisticated genetic engineering whereby several genes are concurrently overexpressed or downregulated is only realistically applicable to the green alga C. reinhardtii.

3.11.4.2

Mass Algal Culture

There are a variety of photoautotrophic-based microalgal culture systems. For example, in those culture systems, the algae may be grown in suspension culture or attached on solid surface. Each system has its own advantages and disadvantages. Currently, the suspension-based open ponds and enclosed photobioreactors are commonly used for algal biofuel production. An open pond culture system usually consists of a series of raceway-type of ponds placed outdoors, while a photobioreactor is a sophisticated reactor design that can be placed outdoors (in most cases) or indoors (e.g., in greenhouse). Open ponds are the oldest and simplest systems for mass cultivation of microalgae. In this system, the shallow pond is usually about 1 ft deep; algae are cultured under conditions identical to their natural environment. The pond is designed in a raceway configuration, in which a paddle wheel provides circulation and mixing of the algal cells and nutrients.1 The raceways are typically made from poured concrete, or they are simply dug into the earth and lined with a plastic liner to prevent the ground from soaking

144

Biofuel From Microalgae

up the liquid. Baffles in the channel guide the flow around bends in order to minimize space. The system is often operated in a continuous mode, that is, the fresh feed (containing nutrients including nitrogen, phosphorus, and inorganic salts) is added in front of the paddle wheel, and algal broth is harvested behind the paddle wheel after it has been circulated through the loop.1 Depending on the nutrient requirements by algal species, a variety of wastewaters can be used for algal culture such as dairy/swine lagoon effluent and municipal wastewater. For some marine type microalgae, seawater or water with high salinity can be used. Although open ponds cost less to build and operate than enclosed photobioreactors, the open pond system has its intrinsic disadvantages. Because of the open-air nature, the open pond often experiences a lot of water loss due to evaporation. Thus, in open ponds the microalgae fail to use carbon dioxide efficiently, and thus biomass production is limited.2 Biomass productivity is also limited by contamination with unwanted algal species as well as organisms that feed on algae. In addition, optimal culture conditions are difficult to maintain in open ponds, and recovering the biomass from such a dilute culture is expensive.18 Enclosed photobioreactors have been employed to overcome the contamination and evaporation problems encountered in open ponds.18 These systems are made of transparent materials and generally placed outdoors for illumination by natural light. The cultivation vessels have a large surface-area-to-volume ratio. The most widely used photobioreactor is a tubular design, which has a number of clear transparent tubes, usually aligned with the Sun‘s rays.2 The tubes are generally less than 10 cm in diameter to maximize sunlight penetration. The medium broth is circulated through a pump to the tubes, where it is exposed to light for photosynthesis, and then back to a reservoir. A portion of algal cells is usually harvested after the solar collection tubes. In some photobioreactors, the tubes are coiled spirals to form what is known as a helical tubular photobioreactor, but these sometimes require artificial illumination, which adds to the production cost, so this technology is only used for high-value products, not for biodiesel feedstock. The algal biomass is prevented from settling by maintaining a highly turbulent flow within the reactor using either a mechanical pump or an airlift pump.2 The result of photosynthesis will generate oxygen. In an open raceway system, this is not a problem as the oxygen is simply returned to the atmosphere. However, in the closed photobioreactor, the oxygen levels will build up until they inhibit and poison the algae. The culture must periodically be returned to a degassing zone, an area where the algal broth is bubbled with air to remove the excess oxygen. In addition, the algae use carbon dioxide, which can cause carbon starvation and an increase in pH. Therefore, carbon dioxide must be fed into the system in order to successfully cultivate the microalgae on a large scale. Photobioreactors require cooling during daylight hours, and the temperature must be regulated in night hours as well. This may be done through heat exchangers located either in the tubes themselves or in the degassing column. The advantages of the enclosed photobioreactors are obvious. It can overcome the problems of contamination and evaporation encountered in open ponds.18 The biomass productivity of photobioreactors can be 13 times more than that of a traditional raceway pond on average.2 Harvest of biomass from photobioreactors is less expensive than that from a raceway pond, since the typical algal biomass is about 30 times as concentrated as the biomass found in raceways.2 However, enclosed photobioreactors also have some disadvantages. For example, the reactors are difficult to scale-up. Moreover, light limitation cannot be entirely overcome because light penetration is inversely proportional to the cell concentration. Attachment of cells to the tube walls may also prevent light penetration. Although the enclosed photobioreactor systems can enhance the biomass concentration, the growth of microalgae is still suboptimal due to variations in temperature and light intensity.

3.11.4.3

Algae Harvesting and Dewatering

Algal harvesting is the concentration of diluted algal suspension into a thick algal paste, with the aim of obtaining slurry with at least 2%–7% algal suspension on dry matter basis. In general, algal biomass harvest is a very challenging step in the algal biofuel production chain. Because the size of the algal cells is very small (3–30 mm diameter) and cell concentration is very dilute (1 g L 1 for open pond system and 90% yield. Industrially, maleic acid is produced from maleic anhydride [C2H2(CO)2O], which is produced from either benzene or n-butane via catalytic oxidation. These reactions are summarized below. n-Butane oxidation to maleic anhydride: C4 H10 þ 3:5O2 /C4 H2 O3 þ 4H2 O Hydrolysis of maleic anhydride C4 H2 O3 þ H2 O/C2 H4 O4 It is noted that fumaric acid can be recovered as a by-product during the production of maleic anhydride. This not only improves the production of maleic anhydride but also turns the waste fumaric acid into valuable product and simplifies the downstream process. The oxidation of n-butane uses catalysts based on vanadium and phosphorus oxides. For the isomerization of maleic

192

Fumaric Acid

Raw material: Maleic acid

Isomerization

Output: Fumaric acid crystal Figure 4

Centrifugation

Drying

Decolorization (activated charcoal)

Crystallization

Filtration

A flowchart for fumaric acid production from maleic acid via isomerization.

acid to fumaric acid, various types of catalysts have been developed. They can be categorized into three types: (1) mineral acids, (2) peroxy compounds with bromides and bromates, and (3) sulfur-containing compounds such as thiourea and its derivatives. The isomerization process has been well established since the 1970s. Fig. 4 shows a simplified process flowsheet. Decolorization and filtration are applied to remove the impurities present in the product solution. Fumaric acid with a high purity can be obtained through crystallization, washing, and drying. However, high toxicity of catalysts, harsh production conditions, and harmful exhausts generated in this process cause serious pollution and health concerns. Alternatively, enzymatic conversion of maleic acid to fumaric acid can be carried out with maleate cis–trans isomerase or maleate isomerase under mild conditions. Microorganisms producing this enzyme include Pseudomonas species and Alcaligenes faecalis. Instead of using purified enzymes, whole-cell catalysis is preferred for industrial production due to simplified procedures and lower costs for enzymes used in this process. However, maleate isomerases from these organisms were unstable even at a moderate temperature. Using thermostable maleate isomerases derived from Bacillus spp. alleviated this problem and enhanced fumaric acid production. A high fumaric acid productivity of 6.98 g L1 h and yield of 95% from maleic acid was achieved using Pseudomonas alcaligenes strain XD-1 cells treated at 70  C for 1 h to inactivate fumarase, which otherwise would convert fumaric acid to malic acid, without affecting the activity of maleate isomerase.12 However, there has been no follow-up study, probably because the process requires high maleic acid concentration as the feedstock, which would increase the production cost, especially at the industrial scale, considerably.

3.16.3.2

Fermentation of Sugar to Fumaric Acid

During the 1940s and before the rising of the petrochemical industry, fumaric acid was produced from sugars by fermentation with R. arrhizus.3–5 Fig. 5 shows the flowsheet for fermentation and subsequent downstream processes once used in industry for fumaric acid production from sugar.8 In general, the fermentation process involved two steps: (1) seed culture preparation and (2) acid production. After
24 h seed culture, cells were harvested and transferred into the production fermentor for fumaric acid production. The downstream process depended on the neutralizing agent used during the fermentation. In the CaCO3 process, the fermentation broth containing calcium fumarate, cells, and excessive CaCO3 was acidified with H2SO4 to pH 1.0 and heated to 160  C. After filtration to remove insoluble particles (cells and CaSO4), the filtrate was cooled to below the room temperature to recover fumaric acid as precipitate via crystallization. When Na2CO3 was used as the neutralizing agent, the fermentation broth was first

Process with Na2CO3 Seed culture Filtration

Acidification (Sulfuric acid to pH 1)

Fermentation Acidification at 160°C (Sulfuric acid to pH 1)

Filtration

Process with CaCO3

Filtration Figure 5

Drying

Flowsheet for fumaric acid production via fermentation.13

Output: Fumaric acid crystal

Cooling to 20°C

Fumaric Acid Table 1

193

Comparison of fumaric acid production from various substrates and fermentor systems by Rhizopus spp.

Strain

Fermenter

Substrate

Product titer (g L1)

Yield (g g 1)

Productivity (g L1 h)

References

Rhizopus arrhizus

Shake flask and stirred tank Fluidized bed Shake flask Shake flask Shake flask Shake flask Shake flask Shake flask Shake flask Stirred tank Shake flask and stirred tank RBC with adsorption RBC 10 L air lift Bubble column Shake flask Shake flask Shake flask SSF Shake flask Shake flask SSF Shake flask 7-L bioreactor Shake flask Shake flask Stirred tank

Glucose

38–130

0.33–1.0

0.46–2.0

13–17

Molasses Xylose Potato flour Corn starch Grape must Xylose Crude glycerol and glucose Glucose and xylose Glucose Glucose

17.5 15.3 43.5 71.9 24.1 45.3 22.8 46.8 30.3 35.8–66.5

0.36 0.23 0.58 0.60 0.23

0.36 0.07 0.42 0.50 0.17 0.47 0.16 0.6 0.4 0.48–0.90

18 19 20 21 22 23 24 25 26 27, 28

Glucose Glucose Glucose Glucose Dairy manure Glucose Apple industry waste biomass Cornstarch Pulp and paper solid waste Lignocellulosic syrup Corncob hydrolysate Glucose Apple juice Cassava bagasse

92.0 0.85 75.5 0.75 37.8 0.75 37.2 0.53 25 50.2 0.72 25.2 – 52.0 g kg1 dry weight 44.1 0.44 23.5 41.5 g kg1 dry weight 34.20 0.43 49.1 – 14.7–20.0 0.50–0.66 33.1 – 21.3 –

4.25 3.78 0.81 1.03 0.17 0.35 0.35

15 29 30 31 32 33 34 34 35 36 36 37 38 39, 40 41 42

Rhizopus oryzae

Rhizopus nigricans Rhizopus formosa

– 0.58 – 0.33–0.67

0.53 0.49 0.24 – 0.25 0.23 –

RBC, Rotary biofilm contactor.

filtered to remove cells and then acidified to recover fumaric acid crystals. This process was simpler as no heating and cooling steps were involved and no gypsum (CaSO4), a solid waste, was produced. The conventional industrial fermentation process for fumaric acid production became unfavorable due to the high production cost that can be attributed to the high sugar price (vs. petroleum feedstock) and relatively low product yield ( M 2. Judgment could lead to conclusion of effectiveness. 4. C –T point estimate favors C and upper bound of 95% CI > M 1, indicating 330 there is no evidence of effectiveness for test drug. Figure 6

Active control–test drug differences (point estimate, 95% CI).10

278

Antibiotics: The Miracle Menaced

This is then the fourth level of discounting. So we get to a point, rather unscientifically, where the noninferiority margin is some unrealistically small number that results in infeasible trial designs. The FDA has now issued guidelines on noninferiority margins.9 They require a justification for the proposed margin. This must include a rationale for the assumption that the comparator is superior to no therapy (or placebo) (M1) that is the underlying assumption upon which noninferiority trials are based. This is extremely challenging for antibiotics where no placebocontrolled trials have been conducted since the 1940s. Why? Because it has been thought that since antibiotics had such dramatic treatment effects when they were compared to placebo it would be unethical to withhold them from patients with serious infections (some might argue that this would be the case even for not so serious bacterial infections). During the heyday of antibiotics, the Infectious Diseases Society of America and the FDA developed a series of guidelines for conducting trials for approval in specific indications such as skin infections, intra-abdominal infections, community and hospital-acquired pneumonia, etc. These guidelines were based on the idea that clinical trials had to be feasible. So, the statistical requirements of the trials were less stringent in those indications where recruitment is difficult. In reviewing these guidelines in the late 1990s, the FDA decided that there was no scientific basis for the statistical stringency required in different indications. They began a campaign to increase this stringency, and therefore trial size. Several companies, including Wyeth, balked at this increased stringency believing that the increased costs associated with the increased trial size requirements would reduce their return on investment to nil. The data upon which Wyeth based their decision is detailed in Fig. 7. Wyeth proposed running trials for tigecycline at a 15% noninferiority margin, while the FDA proposed a 10% margin. The projected difference in patient numbers is shown. The FDA‘s proposal increased patient numbers by two- to threefold. The situation led to a letter written by myself and Bob Moellering entitled ‘The FDA and the end of antibiotics’.22 Eventually, the FDA backed down on some of their demands – but that was transitory. As the leadership at the agency changed, trial stringency increased. Wyeth was one on the last companies to be able to run trials at the lower 15% noninferiority margin. Then, the FDA began to examine the issue of milder, community-acquired infections such as otitis media (middle ear infection), acute bacterial sinusitis, and acute bacterial exacerbations of chronic bronchitis. They felt that the evidence that antibiotic therapy was not better than placebo was not convincing. They, therefore, now require superiority trials with a comparison to placebo for approval. Surprisingly, The European regulatory authority has now agreed with the FDA in their stance, although they seem to remain more flexible. Scientifically, in the case of otitis where it is not clear that antibiotics improve disease in the majority of patients, this might be reasonable. It is at best controversial for acute bacterial sinusitis and it is probably scientifically completely incorrect for patients with severe exacerbations of chronic bronchitis. For exacerbations of chronic bronchitis with increased cough and sputum purulence, the prestigious Cochrane Report has concluded that antibiotics, regardless of choice, reduce the risk of short-term mortality by 77%, decrease the risk of treatment failure by 53%.19 But, the science is almost irrelevant, since it has been impossible to convince enough patients to enroll in such trials in a reasonable period of time to allow for the trials to be practical. Who with severe otitis, sinusitis, or life-threatening bronchitis wants to take a placebo after all? Therefore, these indications are now out of the reach of pharmaceutical companies. Several billion dollars in marketplace potential has been wiped off the map. The future reality is that if resistant respiratory pathogens arise in a period where no antibiotic active against them is approved for these indications, physicians will use unapproved antibiotics that might work off label. In its latest re-examination of indications for antibiotics, the FDA has focused on pneumonia and skin infections. The FDA met with its Anti-Infectives Drug Advisory Committee in December 2009 to discuss how industry should design clinical trials for new

Indication

Cure rate

90% power 10% delta

90% power 15% delta

CAP (total number for two studies) 70% evaluability

85%

1532

688

Skin (total number for two studies) 60% evaluability

80%

2248

1000

IAI (total number for two studies) 60% evaluability

70%

2948

1316

65%

1598

710

8326

3714

80% power 10% delta

80% power 15% delta

6226

2770

HAP (total number for one study) 60% evaluability Total

Total

Figure 7 Clinical trial size projections for tigecycline.22 Numbers indicate number of patients required to enroll to achieve statistical targets as indicated.

Antibiotics: The Miracle Menaced

279

antibiotics to treat pneumonia. It was the clinicians versus the statisticians and, maybe, just maybe, we all will win. The FDA stated in their summary that infeasible trial designs were unacceptable. They also showed that, based on clinical grounds alone, antibiotics have an enormous effect on pneumonia by day three of illness. Dr. Mary Singer of the FDA showed data2,8,9,16 from the 1930s through the 1950s demonstrating that, compared to no effective therapy, antibiotics made 30%–70% more people significantly better by day 3 of therapy. Of course, every clinician who has ever seen a patient knows this already. One problem will be to better define what ‘better’ means today. The advisory committee essentially rejected a call by Public Citizen and others to make mortality the only possible endpoint for clinical trials in pneumonia. Most felt that such trials would not be feasible since they would have to enroll up to 50,000 patients per trial for two trials. In spite of this, the statisticians insisted that mortality was the only scientifically acceptable endpoint. The committee also voted that only those patients where you could show a bacterial pathogen should be evaluated for efficacy in the trial and that patients who had received any prior antibiotic should be excluded from trials. This is still difficult since, under the best of circumstances, only about 30% of patients enrolled in modern trials have a demonstrable bacterial pathogen and about 40% have already received at least one dose of another antibiotic. In the United States, quality criteria require that patients receive antibiotics for pneumonia within 6 h of presentation to the hospital. To get a patient enrolled within this time frame is going to be challenging as well. However, there is light at the end of the tunnel. If the FDA would agree to decrease the statistical stringency they require, everything becomes easier. They can justify this scientifically because the effect of antibiotic therapy at day three is so great. Then, either we need to be able to pool patients who have a definite bacterial pathogen demonstrated as the cause of their infection across two trials, or we need to be able to use investigational diagnostic tools to increase our diagnosis rates in the trial. If all of this could be done, we can again start to develop new antibiotics for pneumonia. One brave company, Advanced Life Sciences, recently announced that they would carry out a superiority trial in pneumonia. They will attempt to demonstrate that their ketolide antibiotic (a member of the general macrolide class) is superior to another macrolide, azithromycin, in the treatment of serious community acquired pneumonia caused by macrolide-resistant pathogens where their ketolide is expected to be active. The problem with this approach is that even against macrolide-resistant organisms, therapy will still succeed in the majority of patients for a variety of reasons. Therefore, they might only see an absolute difference of say 10%–15% improvement with their ketolide compared to azithromycin. To prove this will require a very large number of patients infected with resistant strains. This design will be at best difficult to execute. Sometime in early March, the FDA let it be known through the Infectious Diseases Society and others that they no longer knew how to set a noninferiority margin for designing trials in serious infections of skin and skin structures (ABSSSI). They reviewed pneumonia and decided that the greatest clinical effect of antibiotics in the treatment of pneumonia occurred early in the treatment course – in the first 72 h or so. The FDA then went back to re-review the old data on skin infections. They became concerned that, like pneumonia, the greatest clinical effect occurred early. If so, they reasoned, then the current endpoints looking at cure much later, usually after 2 weeks or so, were inappropriate. They further worried that the old data, comparing to placebo where the time to response was studied were so limited as to preclude their ability to actually define the extent of this response in a reliable way. The Infectious Diseases Society of America has published their approach to clinical trial design in skin infections.28 Finally, one company, Trius, announced that they had come to an agreement with the FDA on endpoints for their proposed Phase III trial in skin infections.30 They said, ‘The double-blind pivotal study will compare the efficacy and safety of once-daily oral administration of 200 mg of torezolid phosphate over six days of treatment to twice-daily oral administration of 600 mg of linezolid (Zyvox) for 10 days of treatment. The primary efficacy endpoint will be the cessation of spread of infected lesions and absence of fever at 48–72 h following initiation of treatment. Secondary endpoints will include, among other things, sustained clinical response at the end of therapy visit, and the investigator’s assessment of clinical response at all visits and clinical success at the post treatment evaluation visit. Provided noninferiority is met, an assessment of superiority of torezolid phosphate to linezolid with respect to the primary efficacy endpoint will also be made. This is the first indication that the FDA has indeed applied, which could possibly be the criteria for their new guidelines for the study of skin infections. At least these trials appear to be feasible. The FDA has other indications under review such as nosocomial pneumonia where they seem to be leaning to mandating 30-day all-cause mortality as the endpoint for the trials. It is not clear what the NI margin will be, but if it is under 15%, the trials will again be infeasible.

3.23.7

Large Pharmaceutical Companies Exit and Biotechnology Enters

According to IMS Health, the global pharmaceutical market, as of 2008, was around $770 billion. Of that, antibiotics comprised about $38 billion of which $21 billion represented sales of oral drugs in the community and $16 billion came from sales of parenteral antibiotics (mostly) in hospital. The hospital segment appears to be growing where the community segment seems flat or may even be declining in dollar volume. With a market size of $38 billion, one would think this would be attractive to industry. However, the antibiotics market is highly saturated, competition is fierce, marketing costs can be high, and a number of large selling products are becoming generic. As the costs of clinical trials and the risk of failure have increased for all therapeutic areas within the pharmaceutical industry, large companies have carried out an ever more stringent prioritization process. In this process, they attempt to focus resources on the most commercially promising projects where they are most likely to make a substantial return on their investment. The Tufts Center has examined retrospectively the net present value (corrected for inflation) of various drug classes (see Table 2).

280

Antibiotics: The Miracle Menaced Table 2

Table 3

Net present value (lifetime earnings minus lifetime costs) of drugs4

Net present value (NPV) of drugs 1990–94

Mean NPV

All drugs Antibiotics Statins SSRI antidepressants

$0.8 billion $1.1 billion $15 billion $11 billion

Large pharmaceutical companies active in antibacterial research24,25

Large pharmaceutical companies active in antibacterial research in 1990 Abbott Bayer Bristol Meyers Ciba Glaxo Hoechst Johnson & Johnson Lederle Marion Merrell Dow Merck Parke-Davis Pfizer Roche Rhone Poulenc SmithKline Beecham Squibb Upjohn Zeneca

Companies active today

Companies not pursuing antibacterial research today

Pfizer-Wyeth Astra-Zeneca Glaxo SmithKline Novartis Merck-Schering Plow

Abbott Bayer Bristol Meyers Squibb Lilly Roche Johnson & Johnson Sanofi-Aventis

In this retrospective analysis, it is clear that antibiotics are, at best, average in return and do not compare to the statins for cholesterol control or to antidepressants. This result is partly related to the fact that antibiotics are only taken for a few days to 2 weeks where other drugs, since they cannot cure disease, are taken for a lifetime or for more prolonged periods. It is easy to understand how, in the pharmaceutical company prioritization process, antibiotics would be lower on the priority list. Physicians and hospitals, in their attempt to slow the emergence of resistance, frequently use new antibiotics active against resistant strains very sparingly. Thus, antibiotics are one of the few areas where a good new product will simply be put on the shelf and only used when absolutely necessary. The other dynamic weighing on antibiotics research is consolidation within the industry. A number of years ago, Dr. Karen Bush examined the effect of consolidation on six companies (see Table 3) from 1983 to 2003.24,25 These six 2003 companies derived from the prior consolidation of 70 companies resulting in over a 90% consolidation. With the recent purchase of Wyeth by Pfizer and Schering-Plough by Merck, the consolidation within the industry is now over 95% during the last four decades. Putting consolidation together with companies that have announced (or sometimes not) their departure from antibiotics research (substantially), Dr. Steven Projan and the author constructed (Table 4) that the author has continued to modify with the latest events.24,25 It is estimated that almost all mid- to large-sized pharmaceutical companies were still actively engaged in antibiotics research in 1990. Today, 20 years later, only four large pharmaceutical companies (over $10 billion in revenue yearly) remain in the area. Of interest is the fact that among those who have left the area, only one, Johnson and Johnson, has licensed in new antibiotics from biotechnology companies and they only recently left the area. The explanation is partly that when the companies close down their antibiotics research effort, they lose most of their expertise in the area and can no longer carefully evaluate opportunities that might be presented. With the departure of large pharmaceutical companies from antibiotics research, a number of small biotechnology companies have entered the fray. While most have procured cast-offs from large companies, either when the large companies left the field and spun-off their antibiotics assets or by in-licensing products, a few have actually started to discover novel compounds. Of 17 biotech antibiotics placed into late stage development (phase II or beyond) over the last decade or so, only six were actually discovered at the biotech company (see Table 5). All but one are descendants of previously discovered classes of antibiotics. The numbers for biotech appear to be increasing, though, and we can look forward to a pipeline bubble from biotech soon. The question for biotech is the following. Will they be able to finance the phase III trials at about $30 million per trial? This is

Antibiotics: The Miracle Menaced Table 4

281

Consolidation within the pharmaceutical industry 1980–200324,25

2003 pharmaceutical company

Number of original companies since 1980

Aventisa Bristol-Meyers-Squibb Glaxo Smith Kline Novartis Pfizerb Wyeth

17 8 12 7 12 14

Not shown: Merck has now purchased Schering Plow. a Now Sanofi-Aventis. b Pfizer has now purchased Wyeth.

Antimicrobial compounds in development by biotech24,25

Table 5 Compound

Manufacturer

Company of origin

Class origin

Status

Dalbavancin Iclaprim Oritavancin Telavancin Ceftibiprole Cethromycin Doripenem EDP-420 Faropenem Ceftaroline NXL-104 NXL-103 PTK-0796 Torezolid Radezolid ACHN-490 BC-3781

Durata from Pfizer from Vicuron Arpida Targanta Theravance Basilea-Johnson&Johnson (J&J) Advanced Life Sciences Peninsula J&J Enanta Replidyne Cerexa/Forest Novexel Novexel Paratek – Novartis Trius Rib-X Achaogen Nabriva

Merrell Marion Dow Roche Lilly Theravance Roche Abbott Shionogi Enanta Daiichi Suntory Takeda Aventis Aventis Paratek Dong-A Rib-X Achaogen Nabriva

Glycopeptide Trimethoprim Glycopeptide Glycopeptide Cephalosporin Macrolide Carbapenem Macrolide Penem Cephalosporin Novel Streptogrammin Tetracycline Oxazolidinone Oxazolidinone Aminoglycoside Pleuromutilin

Unknown Failed in the 1 US Failed in the US Approved Failed in the US & Europe Failed in the US Approved Ph. II Failed in the US Submitted for approval Phase II Phase II Phase III Ph. II Ph. II Ph. II Ph. II

Class origin – drug class root for new compound. Compounds in bold were discovered in Biotech. One compound of novel class, NXL-104 is indicated as novel in bold under class origin.

beyond the range of venture capital funding and usually requires a large pharmaceutical company partner (a shrinking list of possibilities there) or the company has to go to the public markets as Trius did recently. However, those markets are not very welcoming right now.

3.23.8

Conclusions

As long as we use antibiotics, whether it is on the farm, for our domestic animal friends or four ourselves, resistance to those antibiotics will follow eventually. The struggle between bacteria and the natural toxins produced by other microorganisms in their ecological niche has been going on since the beginning of time. We are just accelerating the process. Infection control, antimicrobial stewardship and other similar interventions can only delay the onset of resistance. Therefore, we must have a continuing pipeline of new antibiotics to sustain the miracle we have enjoyed for the last 75 years. However, the extraordinary difficulty of new antibiotic discovery, an uncertain and sometimes hostile regulatory environment and market pressures are driving the pharmaceutical industry, upon whom we still depend for new antibiotics, out of the field. Moreover, once companies leave the domain, they so far seem unable or unwilling to reenter. The success of products such as Zyvox (linezolid) and Cubicin (daptomycin) and of biotechnology companies such as Cubist are beacons of light in this otherwise dismal environment. One can only hope that regulatory agencies, especially the FDA, will understand their own contribution to our current dearth of new antibiotics and that they will quickly rectify the situation by once again providing feasible paths forward for antibiotic development. This is the one major obstacle that we can quickly remove. The biotechnology industry is beginning to innovate antibiotics, but they must be able to bring them all the way to market. That will require a major investment that seems less and less likely. Beyond that, we still await the true birth of antibiotics from genomics. Finally, the industry must find a way to bring antibiotics forward in a way that provides a reasonable return on their investment.

282

Antibiotics: The Miracle Menaced

References 1. Armstrong, G. L.; Conn, L. A.; Pinner, R. W. States during the 20th Century: Trends in Infectious Disease Martality in the United States. J. Am. Med. Assoc. 1999, 281 (1), 61–66. https://doi.org/10.1001/jama.281.1.61. 2. Bullowa, J. G. M. The Course, Symptoms and Physical Findings. In The Management of Pneumonias, Oxford University Press: New York, NY, 1937. Chapter II. 3. Cubist Pharmaceuticals (n.d.) Recent results and 2010 plans. http://www.cubist.com/investor_relations. 4. DiMasi, J. A.; Grabowski, H. G.; Vernon, J. Research and Development Costs and Returns by Therapeutic Category. Drug Inf. J. 2004, 38, 211–223. 5. Drusano, G. L. Antimicrobial Pharmacodynamics: Critical Interactions of ‘bug and Drug’. Nat. Rev. Microbiol. 2004, 2, 289–300. 6. Duncan, G.; Warner, W. P.; Dauphinee, J. A.; Dickson, R. C. The Treatment of Pneumococcal Pneumonia with Dagenan. Can. Med. Assoc. J. 1939, 40, 325–332. 7. Fine, M. J.; Auble, T. E.; Yealy, D. M.; et al. A Prediction Rule to Identify Low-risk Patients with Community-acquired Pneumonia. N. Engl. J. Med. 1997, 336 (4), 243–250. 8. Flippin, H. F.; Lockwood, J. S.; Pepper, D. S.; Schwartz, L. The Treatment of Pneumococcic Pneumonia with Sulfapyridine. J. Am. Med. Assoc. 1939, 112, 529–534. 9. Food and Drug Administration. Issues in Clinical Trial Design for Community Acquired Bacterial Pneumonia, 2009. http://www.fda.gov/downloads/AdvisoryCommittees/ CommitteesMeetingMaterials/Drugs/Anti-InfectiveDrugsAdvisoryCommittee/UCM195620.pdf. 10. Food and Drug Administration. Guidance for Industry Non-inferiority Clinical Trials, 2010. http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/ Guidances/UCM202140.pdf. 11. Hirai, K.; Ito, A.; Abe, Y.; et al. Comparative Activities of AM-715 and Pipemidic and Nalidixic Acids against Experimentally Induced Systemic and Urinary Tract Infections. Antimicrob. Agents Chemother. 1981, 19 (1), 188–189. 12. Kuck, N. A.; Redin, G. S.; Forbes, M. Activity of Minocycline and Other Tetracyclines against Tetracycline-sensitive and -resistant Staphylococci. PSEBM Proc. Soc. Exp. Biol. Med. 1971, 136 (2), 479–481. 13. Leigh, D. A. Antibacterial Activity and Pharmacokinetics of Clindamycin. J. Antimicrob. Chemother. 1981, 7 (Suppl. A), 3–9. 14. Maggi, N.; Pasqualucci, C. R.; Ballotta, R.; Sensi, P. Rifampicin: A New Orally Active Rifamycin. Chemotherapy 1966, 11 (5), 285–292. 15. McCormick, M. H.; McGuire, J. M.; Pittenger, G. E.; et al. Vancomycin, a New Antibiotic. I. Chemical and Biologic Properties. Antibiot. Annu. 1955–1956, 3, 606–611. 16. Meakins, J. C.; Hanson, F. R. The Treatment of Pneumococcic Pneumonia with Sulfapyridine. Can. Med. Assoc. J. 1939, 40 (4), 333–336. 17. O‘Shea, R.; Moser, H. E. Physicochemical Properties of Antibacterial Compounds: Implications for Drug Discovery. J. Med. Chem. 2008, 51 (10), 2871–2878. 18. Payne, D. J.; Gwynn, M. N.; Holmes, D. J.; Pompliano, D. L. Drugs for Bad Bugs: Confronting the Challenges of Antibacterial Discovery. Nat. Rev. Drug Discov. 2007, 6, 29–40. 19. Ram, F. S.; Rodriguez-Roisin, R.; Granados-Navarrete, A.; et al. Antibiotics for Exacerbations of Chronic Obstructive Pulmonary Disease. Cochrane Database Syst. Rev. 2006, (2), CD004403. 20. Retsema, J.; Girard, A.; Schelkly, W.; et al. Spectrum and Mode of Action of Azithromycin (Cp-62,993), a New 15-membered-ring Macrolide with Improved Potency against Gram-negative Organisms. Antimicrob. Agents Chemother. 1987, 31 (12), 1939–1947. 21. Rosenblatt, J. E.; Barrett, J. E.; Brodie, J. L.; Kirby, W. M. Comparison of in Vitro Activity and Clinical Pharmacology of Doxycycline with Other Tetracyclines. Antimicrob. Agents Chemother. 1966, 6, 134–141. 22. Shlaes, D. M.; Moellering, R. C., Jr. The FDA and the End of Antibiotics. Clin. Infect. Dis. 2002, 34, 420–422. 23. Shlaes, D. M.; Projan, S. J.; Edwards, J. E. Antibiotic Discovery: State of the State. Am. Soc. Microbiol. News 2004, 70, 275–281. 24. Shlaes, D. M.; Projan, S. J. Antimicrobial Resistance versus the Discovery and Development of New Antimicrobials. In Antimicrobial Resistance; Mayers, D. L., Ed.; 2009; pp 43–50. Heidelberg, London, New York: Humana Press; Dordecht: Springer. 25. Shlaes, D. M. Antibiotics: The Perfect Storm, Springer: Dordrecht, Heidelberg, London, New York, 2010. 26. Shlaes, D. M. Antibiotics, the Perfect Storm Weblog, 2010. http://antibiotics-theperfectstorm.blogspot.com. 27. Slama, T. G. Gram-negative Antibiotic Resistance: There Is a Price to Pay. Indiana Univ. School Med. Critic. Care 2008, 12 (Suppl. 4), S4 (doi:10.1186). 28. Spellberg, B.; Talbot, G. H.; Boucher, H. W.; et al. Antimicrobial Availability Task Force of the Infectious Diseases Society of America. Antimicrobial Agents for Complicated Skin and Skin-structure Infections: Justification of Noninferiority Margins in the Absence of Placebo-controlled Trials. Clin. Infect. Dis. 2009, 49 (3), 383–339. 29. Spellberg, B.; Powers, J. H.; Brass, E. P.; et al. Trends in Antimicrobial Drug Development: Implications for the Future. Clin. Infect. Dis. 2004, 38 (9), 1279–1286. 30. Trius Therapeutics. Trius Therapeutics Obtains Special Protocol Assessment with FDA for Phase 3 Study of Torezolid Phosphate, 2006. http://www.triusrx.com/triustherapeutics-news-100616.php. 31. Waitz, J. A.; Moss, E. L., Jr.; Drube, C. G.; Weinstein, M. J. Comparative Activity of Sisomicin, Gentamicin, Kanamycin, and Tobramycin. Antimicrob. Agents Chemother. 1972, 2 (6), 431–437.

Penicillins and Cephalosporinsq

3.24

C Garcı´a-Estrada, NBIOTEC, Instituto de Biotecnología de León, León, Spain J-F Martı´n, Universidad de León, León, Spain © 2019 Elsevier B.V. All rights reserved. This is an update of C. García-Estrada, J.-F. Martín, 3.24 - Penicillins and Cephalosporins, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 255-268.

3.24.1 Introduction to Penicillins and Cephalosporins 3.24.2 Structure and Mechanism of Action of Penicillins and Cephalosporins 3.24.3 Penicillin and Cephalosporin Biosynthesis 3.24.3.1 Penicillin and Cephalosporin Biosynthetic Pathways 3.24.3.2 Organization and Expression of Penicillin and Cephalosporin Biosynthetic Genes 3.24.3.3 Compartmentalization of the Penicillin Biosynthetic Pathway 3.24.3.4 Compartmentalization of the Cephalosporin Biosynthetic Pathway 3.24.4 Biotechnological Implications in the Biosynthesis of Penicillins and Cephalosporins 3.24.4.1 Strategies Applied to the Production of Penicillins 3.24.4.1.1 Industrial Strain Improvement and Genetic Engineering 3.24.4.1.2 Production of Semisynthetic Penicillins 3.24.4.2 Production of Cephalosporins and Genetic Engineering of P. chrysogenum and A. chrysogenum 3.24.4.2.1 Penicillin-Derived Cephalosporins 3.24.4.2.2 Cephalosporin C-Derived Cephalosporins 3.24.4.2.3 Semisynthetic Cephalosporins 3.24.5 Future Outlook References Relevant Websites

283 284 285 285 287 288 290 291 291 291 292 292 293 293 294 295 295 296

Glossary 7-ACA 7-Aminocephalosporanic acid. ACVS L-d(a-aminoadipyl)-L-cysteinyl-D-valine synthetase. 7-ADCA 7-Aminodeacetoxycephalosporanic acid. ad7-ACCCA Adipyl-7-amino-3-carbamoyloxymethyl-3-cephem-4-carboxylic acid. 6-APA 6-Aminopenicillanic acid. 7-DAC 7-Aminodecetylcephalosporanic acid. DAC Deacetylcephalosporin C. DAOC Deacetoxycephalosporin C. IAT Acyl CoA:isopenicillin N-acyltransferase IPN Isopenicillin N.

3.24.1

Introduction to Penicillins and Cephalosporins

Antibiotics are low-molecular-weight organic compounds produced by microorganisms that at low concentrations are able to selectively inhibit the growth of other microorganisms. These compounds are considered secondary metabolites, since they are not required for normal growth, development, or reproduction of the microorganisms that produce them. The biological role of antibiotics is controversial and it has been suggested that antibiotic production may confer an ecological advantage for survival in natural habitats where nutrients are limiting for microbial growth1. The discovery of b-lactam antibiotics is one of the most significant milestones in human history and it entailed a revolution in modern chemotherapy. The members of the family of b-lactam antibiotics stand out from the other family members because of their high activity and low toxicity, for which they are among the most commonly prescribed drugs. The use of these compounds has helped medicine to face up to infectious diseases, which have been the cause of death of millions of human beings throughout history, and dramatically reduce the mortality rate. q

Change History: October 2017. C. García-Estrada and JF. Martín updated all sections, mainly the penicillin and cephalosporin biosynthesis, including updated information on transport systems for beta-lactam intermediates. Both authors updated figures and the reference list with some more up to date research.

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00166-X

283

284

Penicillins and Cephalosporins

Sir Alexander Fleming’s accidental discovery of antimicrobial activity generated by a fungus contaminating a Petri dish containing Staphylococcus sp. in 1928 represents the starting point for modern antibiotic therapy. Fleming was not the first person to observe the antagonism between fungi and bacteria. Several scientists had observed this phenomenon before, although the effect was not initially attributed to the production of a specific compound. At the end of the 19th century, several experiments established that microbial antagonism was due to the action of a diffusible substance produced by one organism on another and this phenomenon was given the name antibiosis. Later, other researchers such as Ernest Duchesne and Andre Gratia (who identified the lytic agent ‘mycolysate’) also showed the antibiosis effect between fungi and bacteria.2 Fleming initially identified the mold responsible for the antibacterial effect as Penicillium rubrum3 and, together with his assistants Dr. Stuart Craddock and Mr. Frederick Ridley, set out to purify the lytic agent, dubbed penicillin.2 Due to the low amounts of penicillin purified from culture broths, Fleming did not extend his work to clinical study and the use of penicillin as a therapeutic agent did not happen until the 1940s. Only a few scientific articles related to penicillin were published in the next 10 years following the discovery of penicillin. The most important one described the correct identification of the Fleming‘s isolate as Penicillium notatum and reported that the b-lactam antibiotic penicillin was the active compound inhibiting the bacterial growth.4 In 1939, efforts were made by Howard W. Florey, Ernst B. Chain, and Norman Heatley, among other scientists, to produce a stable penicillin and develop a massive penicillin production method, initially in Oxford and after the beginning of World War II, at the Northern Regional Research Laboratory of the Department of Agriculture, Peoria, IL, USA. Penicillin was initially isolated from P. notatum cultures, but the low titers produced by this microorganism and the antibiotic demand derived from World War II made the selection of new strains of paramount importance. An improvement was observed after the isolation of Penicillium chrysogenum NRRL 1951 from an infected cantaloupe in a local market at Peoria, IL, USA. This strain was more suitable than P. notatum for penicillin production in submerged cultures, a fact that attracted the interest of several pharmaceutical companies, such as Pfizer & Co., E.R. Squibb & Sons, and Merck & Co., as well as the government.2 Industrial strain improvement programs were developed, and these resulted in penicillin high-producing strains and mass production of penicillin in 1944. In 1945, Ernst B. Chain, Howard W. Florey, and Sir Alexander Fleming were awarded the Nobel Prize for Physiology and Medicine, and a year after this award, penicillin was finally available in the open market. The penicillin precursor 6-aminopenicillanic acid (6-APA) was detected in fermented broths in 1957 and was isolated in 1959. This finding was the starting point for the synthesis of semisynthetic penicillins, which is achieved through the addition of different side chains to 6-APA by a chemical process, and contributed to the beginning of a new era in chemotherapy.5 The history of cephalosporins began in 1945, when the fungus Cephalosporium acremonium was isolated by Giuseppe Brotzu from the bay water at Cagliari, Italy.6 Sir Howard W. Florey, Professor of Pathology at Oxford University, agreed to undertake the work on the products of Brotzu‘s fungus and Edward Abraham, later helped by Guy Newton, started investigations at The Sir William Dunn School of Pathology in September 1948. Brotzu‘s strain was deposited at the Commonwealth Mycological Institute, Kew, as strain CMI 49137 and was renamed Acremonium chrysogenum, which in Greek means ‘gold-producing branches’, whereas C. acremonium means ‘branches with head-like seeds‘, which appeared to be less apt. The group at Oxford found that this fungus produced at least three types of antimicrobial compounds, which were isolated and identified.7 The first compounds isolated in 1949 were members of the cephalosporin P complex and were later found to be tetracyclic triterpenes chemically related to helvolic acid (fumigacin). Cephalosporin P received this name because it was exclusively active against Gram-positive bacteria. Later, the same year, a second compound, initially named cephalosporin N, was found in culture filtrates from which cephalosporin P had been removed. This compound was active against Gram-negative and Gram-positive bacteria and was found to be a penicillin with a D-a-aminoadipic side chain. Therefore, it was renamed penicillin N. Finally, in an experiment carried out in 1953 to determine the molecular weight of penicillin N, cephalosporin C was isolated. This antibiotic showed two interesting aspects–it was active against Gram-negative and Gram-positive bacteria and it was not hydrolyzed by penicillinase. The latter aspect was especially interesting due to the appearance of penicillin-resistant bacteria. The main drawback of cephalosporin C was the weak antibacterial activity, but the isolation of overproducing mutants and the biosynthesis of semisynthetic cephalosporins have solved this problem.

3.24.2

Structure and Mechanism of Action of Penicillins and Cephalosporins

The b-lactam antibiotics, like many other secondary metabolites, have unusual chemical structures. All b-lactams contain a fourmembered b-lactam ring closed by an amide bond. Penicillins contain a bicyclic ‘penam’ nucleus (Fig. 1) formed by fused b-lactam and sulfur-containing thiazolidine rings and an acyl side chain, which depends on the precursors present in the culture medium, bound to the amino group at C-6. Hydrophobic penicillins are exclusively synthesized by filamentous fungi from the genera Penicillium (e.g., P. chrysogenum) and Aspergillus (Aspergillus nidulans), whereas hydrophilic penicillins are synthesized by filamentous fungi (A. chrysogenum), actinomycetes (Streptomyces sp.), and some Gram-negative bacteria. Cephalosporins contain the ‘cephem’ nucleus (Fig. 1), a six-membered dihydrothiazine ring fused to the b-lactam ring. Cephalosporin C has a D-a-aminoadipyl side chain attached to the C-7 amino group, which is identical to that of hydrophilic penicillin N but differs from that of hydrophobic penicillins. Cephalosporins are produced by the fungi A. chrysogenum, Paecilomyces persicinus, Kallichroma tethys, and some other deuteromycetes, although Gram-positive actinomycetes such as Streptomyces clavuligerus or Nocardia lactamdurans and Gram-negative bacteria such as Lysobacter lactamgenus also synthesize cephalosporins as intermediates of cephamycin and cephabacin biosynthetic pathways.8–10 The mechanism of action of penicillins and cephalosporins consists of the inhibition of peptidoglycan biosynthesis, which weakens the bacterial cell wall during cellular division, leading to cytolysis and death. These antibiotics covalently bind to the active

Penicillins and Cephalosporins

H N

R

S

O O

N

CH3

H2N

CH3

H

COOH

Penam

H N

D

COOH

R1

O O

285

S

N

CH 2R2 COOH

Cephem

Figure 1 Chemical structure of the penicillins (penam nucleus) and cephalosporins (cephem nucleus). The D-configuration of the a-aminoadipic side chain is indicated on carbon 1 of this amino acid by D.

site of penicillin-binding proteins (PBPs), which catalyze the linking of peptidoglycan molecules in bacteria in the last step of the bacterial cell wall biosynthesis, because of the structural similarity between b-lactam antibiotics and the last two amino acids (acyl-D-alanine-D-alanine) of the pentapeptide that links the peptidoglycan molecule. PBP enzymes, including transglycosylases (PBP1 complex), transpeptidases (PBP3), and carboxypeptidases (PBP4, PBP5, and PBP6), are irreversibly inhibited by b-lactam antibiotics, no longer catalyzing the linking reaction. In addition, these antibiotics trigger the activation of bacterial cell wall hydrolases and autolysins, which lead to cell lysis.2 The activity of these antibiotics is initially higher against Gram-positive bacteria. In these microorganisms, PBPs are located on the cytoplasmic membrane exposed to the environment, unlike Gram-negative bacteria, where PBPs are present in the periplasmic space protected by the external outer membrane, which acts as a barrier for different molecules. However, the incorporation of new molecules to the penam and cephem nuclei has given rise to the synthesis of semisynthetic antibiotics with a higher activity against Gram-negative microorganisms.

3.24.3

Penicillin and Cephalosporin Biosynthesis

Classical b-lactam antibiotic biosynthesis pathways have been characterized in detail.8–10 The different steps of the biosynthetic pathways take place in different subcellular compartments, involving intracellular traffic of enzymes, precursors, and intermediates.

3.24.3.1

Penicillin and Cephalosporin Biosynthetic Pathways

The biosynthesis of b-lactam compounds (Fig. 2) involves sequential reactions including the formation of the penam nucleus from a linear tripeptide that is cyclized (early biosynthetic steps), ring expansion of the penam to the cephem nucleus (intermediate steps), and modification of the b-lactam nucleus (late decorating steps). The basic structure of classical b-lactam antibiotics originates from three amino acids: L-a-aminoadipic acid, L-cysteine, and Lvaline. L-valine and L-cysteine are common amino acids, unlike L-a-aminoadipic acid, which is a nonproteinogenic amino acid formed by a specific pathway related to the fungal lysine biosynthetic pathway. The two early enzymatic steps are common to all classical b-lactam producers. These steps lead to the biosynthesis of isopenicillin N (IPN), which is the first compound in the pathway with antibiotic activity. The first enzyme of the pathway is the nonribosomal peptide synthetase L-d(a-aminoadipyl)-L-cysteinyl-D-valine (ACV) synthetase (ACVS), which is a very large multifunctional protein (MW around 420 kDa). This protein is encoded by a single structural 11 kb intronless gene called pcbAB (acvA), which is present in both fungal and bacterial penicillin and cephalosporin (and cephamycin) gene clusters. The ACVS consists of three well-conserved domains that activate each of the three amino acids. This protein sequentially activates the three substrate amino acids with ATP as aminoacyl adenylates, binds them to the enzyme as thioesters, epimerizes the L-valine to D-valine, catalyzes the condensation of the three amino acids to form the tripeptide ACV, and, finally, releases this tripeptide from the enzyme by means of the internal thioesterase activity. In the second step of the early biosynthetic stage, the IPN synthase (cyclase) encoded by the intronless pcbC (ipnA) gene, which is an intermolecular dioxygenase that requires Fe2þ, molecular oxygen, and ascorbate, catalyzes the removal of four hydrogen atoms from the ACV tripeptide. This reaction leads to the oxidative ring closure of the tripeptide and the formation of the bicyclic structure (penam nucleus) of IPN, which constitutes the branch point of penicillin and cephalosporin biosynthesis. Hydrophobic penicillin-producing fungi (e.g., P. chrysogenum and A. nidulans) contain, in addition to the pcbAB (acvA) and pcbC (ipnA) genes common to filamentous fungi and bacteria, a third gene in the penicillin cluster. This gene, named penDE (aatA), encodes the peroxisomal acyl CoA: IPN acyltransferase (IAT), and it is not present in cephalosporin C- or cephamycinproducing microorganisms. One distinctive feature of the penDE (aatA) gene is that unlike the other two genes in the cluster, it contains three introns, which suggests a eukaryotic origin.11 The IAT is synthesized as a 40-kDa precursor protein (proIAT) that undergoes an autocatalytic processing between residues Gly102 and Cys103. The active protein is a heterodimer of two subunits:

286

Penicillins and Cephalosporins

L-α-aminoadipic H2N H

L-cysteine

L

H2N

COOH

H2N

SH

COOH

COOH

COOH

pcbAB (acvA) δ(-L-α-aminoadipyl)-L-cysteinyl-D-valine (LLD-ACV)

L-valine

ACV synthetase H N

L

H2 N H

O

COOH

SH NH

O

COOH

pcbC (ipnA) Isopenicillin N (L-isomer)

H N

L

H2N H

Isopenicillin N synthase

O

common steps to all classical β -lactam producers

PAA-CoA L-α-aminoadipic

L-α-aminoadipic

penDE (aatA) Isopenicillin N acyltransferase

H N

S

O O

N

PAA-CoA

COOH

S

H2N

HS-CoA

Benzylpenicillin

O

S

O

COOH

N COOH

cefD1 cefD2

IPN-CoA synthetase IPN-CoA epimerase H N

D

H2N H

COOH

N

S

O

N

O

COOH

COOH

Penicillin N (D-isomer)

6-APA

Deacetoxycephalosporin C (DAOC) synthase

cefEF Penicillin producers (P. chrysogenum)

H N

D

H2N

DAOC

O

H COOH

S N

O

COOH

Deacetylcephalosporin C (DAC) synthase

cefEF H2N

H N

D

H COOH

O

S O

cmcH

COOH

O-Carbamoyl-DAC (OCDAC) cmcI, cmcJ

H2N

H N

D

H COOH

O

H2N

CH3 S

Carbamoyl transferase

cefG

Cephamycin C Cephamycin C producers (S. clavuligerus)

S O

OH

N

DAC-acetyltransferase H N

D

COOH

O

COOH

H2N OCONH2

N

H N

D

H COOH

Methoxyl-transferase O

O

DAC

OCONH2

N

H COOH

O

S O

OCOCH3

N COOH

Cephalosporin C Cephalosporin C producers (A. chrysogenum)

Figure 2 Biosynthetic pathways of benzylpenicillin, cephalosporin C, and cephamycin C in Penicillium chrysogenum, Acremonium chrysogenum, and Streptomyces clavuligerus, respectively. The first two steps (upper part of the figure) are common to all classical b-lactam producers. The L- or D-configuration of the a-aminoadipic side chain is indicated on carbon 1 of this amino acid by L or D.

a (11 kDa, corresponding to the N-terminal fragment) and b (29 kDa, corresponding to the C-terminal region).10,12 This heterodimeric enzyme removes the a-aminoadipic side chain of IPN and exchanges it for hydrophobic acyl molecules, which have to be previously activated by aryl-CoA ligases before they become substrates for the IAT.13 Due to the broad specificity of IAT and to the presence of several isoenzymes, a wide range of side chains may serve as substrates for this enzyme. Thus, natural penicillins, such as penicillin F (D3-hexenoic as side chain) and K (octanoic acid as side chain), are synthesized under natural conditions. However, feeding the cultivation media with phenylacetic or phenoxyacetic acids directs the biosynthesis mainly toward benzylpenicillin (penicillin G) or phenoxymethylpenicillin (penicillin V), respectively.8 In addition to the aatA gene, A. nidulans possesses

287

Penicillins and Cephalosporins

another gene, which has been recently characterized. This gene, named aatB, is not clustered with the rest of the penicillin genes and encodes a cytosolic protein, likely an acyl-CoA transferase, which plays a role in penicillin biosynthesis. The P. chrysogenum aatB homologue (named ial), however, appears to differ in function from that of A. nidulans, since the activity of the protein encoded by this gene was not related to penicillin biosynthesis.8,14 In addition to these key enzymes, other enzymes are also required for penicillin biosynthesis, such as the aryl-CoA ligases, which activate the side-chain aromatic acid, and the phosphopantetheinyl transferase (PPTase), which activates the nonribosomal ACVS.10 In the cephalosporin biosynthesis intermediate steps, IPN is converted to its D-isomer (penicillin N), which is the precursor of antibiotics containing the cephem nucleus (i.e., cephalosporins and cephamycins). This conversion is carried out in a single step in bacterial strains by a classical pyridoxal phosphate-dependent epimerase encoded by the cefD gene, which was found to be located in the cephamycin gene cluster.9 However, epimerization of IPN in A. chrysogenum is carried out by the concerted action of two proteins encoded by two linked genes.15 These genes, cefD1 and cefD2, are located in the ‘early’ cephalosporin gene cluster (Fig. 3). The first gene, cefD1, has four introns and encodes a 71 kDa protein with similarity to fatty acid acyl-CoA synthetases. The second gene, cefD2, contains one intron and encodes a protein homologous to a-methylacyl-CoA racemases of eukaryotic origin. The proposed model for epimerization includes three biochemical steps: CefD1 converts IPN into isopenicillinyl N-CoA; then CefD2 isomerizes the compound into penicillinyl N-CoA, which seems to be released from the enzyme by the third enzyme, a thioesterase. The following step in the cephalosporin/cephamycin pathway is the oxidative opening of the five-membered thiazolidine ring of penicillin N, forming a six-membered dihydrothiazine ring upon reclosure. Deacetoxycephalosporin C (DAOC) synthase (expandase) is the enzyme catalyzing the ring expansion in both A. chrysogenum and bacteria.9 The next step of the pathway corresponds to the hydroxylation of the methyl group at C-3 of DAOC, giving rise to deacetylcephalosporin C (DAC). Both reactions are catalyzed in A. chrysogenum by the same cefEF-encoded enzyme DAOC synthase (expandase)/DAC synthase (hydroxylase), whereas in S. clavuligerus, one enzyme for each reaction has been found: the DAOC synthase (encoded by the cefE gene) and the C-3 hydroxylase (encoded by the cefF gene). The genes cefE and cefF encode proteins with about 70% identity in amino acids, which are 60% identical to the cefEF-encoded protein. In fact, these two enzymes have related molecular mechanisms and each has retained approximately 10% of the residual activity of the other one. These features point to a likely gene duplication event as the origin of these two genes, which gave rise to two proteins with the ability to perform different, although mechanistically related, functions (expandase and hydroxylase).10 The late (and final) step in cephalosporin C biosynthesis is the conversion of DAC into cephalosporin C, a reaction catalyzed by the DAC acetyltransferase, which uses acetyl-CoA as the donor of the acetyl group.10 This enzyme (49 kDa) is evolutionarily similar to O-acetylhomoserine acetyltransferases and is encoded by the cefG gene, which contains two introns and is linked to the cefEF gene, but in the opposite orientation (Fig. 3). In cephamycin-producing actinomycetes, DAC undergoes carbamoylation, followed by hydroxylation and transfer of a methyl group to the hydroxyl present at C-7.9

3.24.3.2

Organization and Expression of Penicillin and Cephalosporin Biosynthetic Genes

Those bacterial and fungal microorganisms that produce b-lactam antibiotics show the typical distribution of the biosynthetic genes in clusters. In the filamentous fungi that produce penicillin, the pcbAB–pcbC (acvA–ipnA) genes are always grouped and are located next to the penDE (aatA) gene. As indicated before, in addition to the aatA gene, the A. nidulans genome contains the aatB gene (Fig. 3). This gene is not clustered with the rest of the penicillin genes and encodes a probable acyl-CoA transferase involved in the last step of the A. nidulans penicillin biosynthesis pathway.8 A remarkable phenomenon undergone by penicillin-overproducing

pcbAB

pcbC

penDE

ipnA

aatA

P. chrysogenum (Chr. I)

acvA

aatB aatA

A. nidulans (Chr. ?)

A. nidulans (Chr. VI)

cefR cefT

cefP cefT

cefT

ORF3

pcbAB

A. chrysogenum (Chr. VII)

pcbC

cefD2

cefD1

cefM

cefEF

cefG

A. chrysogenum (Chr. I)

Figure 3 Penicillin gene clusters in Penicillium chrysogenum and Aspergillus nidulans and cephalosporin gene clusters in Acremonium chrysogenum. Genes of prokaryotic origin are depicted in white, whereas genes of eukaryotic origin are shaded.

288

Penicillins and Cephalosporins

strains of P. chrysogenum is the amplification of the genomic region that includes the penicillin gene cluster.16,17 For more details on this topic, see Section 3.24.4.1.1. In the cephalosporin producer A. chrysogenum, the biosynthetic genes are organized in at least two clusters located on different chromosomes (Fig. 3). In this fungus, the ‘early‘ gene cluster, located on chromosome VII (4.6 Mb), contains the genes pcbAB and pcbC, encoding the enzymes for the first two steps of the pathway; cefD1 and cefD2, responsible for the epimerization of isopenicillin N; and the genes cefT and cefM, which encode transporter proteins required for the transport of hydrophilic b-lactams and penicillin N, respectively. However, the cefEF and cefG genes, whose protein products are involved in the final steps of cephalosporin biosynthesis, are arranged in the ‘late‘ gene cluster located on the 2.2 Mb chromosome I. Biosynthesis of hydrophobic penicillins in P. chrysogenum and A. nidulans is affected by several factors through complex regulatory processes.10,11,18 Easily utilizable carbon, nitrogen, and phosphorus sources dramatically affect the production of this antibiotic. The transcriptional regulation of the genes responsible for penicillin biosynthesis has been studied in detail and regulatory elements have been identified, such as an enhancer element placed at the divergent promoter region of the pcbAB–pcbC genes of P. chrysogenum, which binds a transcriptional activator named PTA1. Since no penicillin pathway-specific regulatory genes have been found in the amplified region containing the three biosynthetic genes,16,17 regulation of the biosynthesis of penicillin seems to be controlled directly by global regulators (e.g., CreA, PacC, and Nre) rather than by pathway-specific regulators. One of these regulators is the LaeA protein, which is a nuclear methyltransferase controlling expression of the penicillin genes in P. chrysogenum and the synthesis of sterigmatocystin, lovastatin, and penicillin, and pigmentation in several aspergilli. LaeA also regulates the synthesis of gliotoxin and the virulence of Aspergillus fumigatus. The LaeA protein contains an S-adenosylmethionine-binding site characteristic of methyltransferases and is predicted to function at the level of chromatin modification. It has been proposed that LaeA regulates the gene clusters through heterochromatin reorganization, perhaps by interacting with methylases or deacetylases that are associated with heterochromatin.19,20 It is also interesting that in A. nidulans, the aatB gene is regulated by the same regulators AnCF and AnBH1 as the aatA gene, which suggests that these two genes are paralogues derived by duplication of a common ancestor gene.8 In A. chrysogenum, cephalosporin biosynthetic gene expression is also controlled by several global regulators, such as the carbon catabolite repressor CreA, the pH regulator PacC, the winged helix transcriptional factor CPCR1, or the veaA gene-encoded Velvet protein. In addition to the effect of these global regulators, DL-methionine is a well-known inducer of cephalosporin biosynthesis.10

3.24.3.3

Compartmentalization of the Penicillin Biosynthetic Pathway

The different biochemical steps involved in the biosynthesis of penicillin take place in different cellular compartments.10 This implies that precursors, enzymes, and the regulation and optimization of the processes involved are spatially separated, assuring optimal environmental conditions for each step of the biosynthetic pathway. Therefore, enzymes, precursors, intermediates, and products must be efficiently transported inside organelles (Fig. 4). The first enzyme of the pathway, ACVS, was associated in early studies to membrane structures identified as Golgi-like organelles. Further cell fractionation experiments located this enzyme attached to or inside vacuoles. However, the fact that the optimal pH for in vitro ACVS activity was higher than that of the vacuolar pH, together with the cofactor requirement and protease sensitivity, indicated that this enzyme was a cytosolic enzyme. The use of immunocytochemistry and electron microscopy techniques confirmed that ACVS was a cytosolic enzyme. As indicated before, the enzyme PPTase activates the ACVS through the covalent attachment of a 40 -phosphopantetheine moiety derived from coenzyme A. In addition, PPTase also activates the a-aminoadipate reductase, which catalyzes the conversion of a-aminoadipate into a-aminoadipate semialdehyde in the fungal lysine biosynthetic pathway occurring in the cytosol. Due to the location of target enzymes, the PPTase is likely to be located in the cytosol. Mitochondria are essential organelles for the synthesis of L-a-aminoadipic acid, since enzymes responsible for the biosynthesis of this amino acid, such as homocitrate synthetase and homoisocitrate dehydrogenase, are located within the mitochondrial matrix. Once the precursor amino acids are synthesized, they are stored in vacuoles. These organelles serve to regulate the levels of L-a-aminoadipate and cysteine, which are toxic at moderate concentrations. Once ACV is synthesized in the cytosol, IPN synthase cyclizes it to form IPN. The IPN synthase behaves as a soluble enzyme and colocalization of this enzyme in the cytosol with ACVS has been confirmed by electron microscopy experiments. In the third (and last) step of the biosynthesis of penicillin, the L-a-aminoadipyl side chain of IPN is replaced by a more hydrophobic side chain by the IAT, giving rise to hydrophobic penicillins. IATs from P. chrysogenum and A. nidulans contain a functional peroxisomal targeting sequence PTS1 at the C-terminal end (ARL and ANI, respectively). Electron microscopy immunodetection has shown that P. chrysogenum IAT is located inside peroxisomes (microbodies). In addition, the transport of IAT inside the peroxisomal matrix is not dependent on the processing state of the protein, since the unprocessable mutant protein IATC103S is correctly targeted to peroxisomes, although it is not active.10 The hydrophobic side-chain precursors phenylacetic acid (for benzylpenicillin) or phenoxyacetic acid (for phenoxymethylpenicillin) have to be previously activated as thioesters with CoA (phenylacetyl-CoA or phenoxymethyl-CoA) by aryl-CoA ligases to be incorporated into the 6-APA molecule by the IAT. Two aryl-CoA ligases were identified in P. chrysogenum, both containing a peroxisomal targeting signal PTS1 at the C-terminal end. The first one, encoded by the phl gene, activates phenylacetic acid and contributes to penicillin production.13 The second one, encoded by the phlB gene, was initially reported to function as another phenylacetylCoA ligase. However, recent studies reported that the phlB gene (renamed aclA gene) encodes a broad substrate-specificity

Penicillins and Cephalosporins

289

plasma membrane cell wall

mitochondrion vacuole

αAA

Cys α AA α AA Cys Val

cytosol

ACV ACVS

IPNS

IPN PAA-CoA nucleus

PAL

IAT

PenG

PAA

PenG ???

microbody

septum

Figure 4 Compartmentalization of the penicillin biosynthetic pathway in Penicillium chrysogenum. aAA, L-a-aminoadipic acid; ACVS, ACV synthetase; Cys, L-cysteine; IAT, IPN acyltransferase; IPNS, IPN synthase; PAA, phenylacetic acid; PAA-CoA, phenylacetyl-CoA; PenG, benzylpenicillin; Val, L-valine.

acyl-coenzyme A ligase involved in the activation of adipic acid, which is a side chain precursor for cephalosporins containing adipic acid instead of a-aminoadipic acid.21 The localization of these two aryl-CoA ligases in peroxisomes has been confirmed recently by physical isolation of those organelles and mass spectrometry identification of the peroxisomal proteins.22 The peroxisomal colocalization of IAT and aryl-CoA ligase indicates that the last two enzymes of the penicillin pathway form a peroxisomal functional complex, pointing to this organelle as a key compartment for the biosynthesis of hydrophobic penicillins. Peroxisomes (microbodies) are organelles surrounded by a single membrane, ranging in diameter between 200 and 800 nm. Several oxidative reactions take place in the peroxisomal matrix, such as the b-oxidation of fatty acids. The microbody luminal pH has been estimated to be slightly alkaline in P. chrysogenum, which is the pH range optimal for the aryl-CoA ligase and IAT. Peroxisomes are not absolutely essential for penicillin production, since a mutant strain of A. nidulans lacking functional peroxisomes was still able to produce penicillin. However, there is a positive correlation between microbody abundance and penicillin production, an effect that is likely related to an increased transport of penicillin and/or its precursors across the microbody membrane. In fact, the increase in the number of peroxisomes has also been observed in penicillin high-producing strains of P. chrysogenum.17 Since IPN is synthesized in the cytosol and is the substrate for the peroxisomal IAT, transport events through the membrane of this organelle are very important. Because of the hydrophilic nature of IPN, the uptake of this precursor from the cytosol is likely to occur through specific carriers. An active IPN transport system must be present in the peroxisomal membrane to assure an adequate pool of IPN inside microbodies. In addition, the hydrophobic penicillins synthesized by the IAT have to be transported out of the microbody, first to the cytosol and then outside the fungal cell. Since penicillin is produced in very large amounts by overproducing strains, which accumulate high extracellular levels, secretion through simple diffusion mechanisms is very unlikely. It has been reported that penicillin secretion is sensitive to verapamil, an antagonist of multidrug transporters,17 which suggests that secretion is an active process that involves this type of transporters. Several ABC transporters have been identified in P. chrysogenum and some of them had increased expression in one industrial strain.17 However, the role of these transporters in the secretion of penicillin is still

290

Penicillins and Cephalosporins

unclear. This may indicate the presence of additional secretion pathways implemented in the penicillin-overproducing strains that were absent (or inefficient) in the low-producing wild-type strains. Another mechanism that has been discussed is the secretion of penicillin through vesicles after pexophagy. During the late stage of the culture, peroxisomes are integrated into vacuoles by the pexophagy phenomenon. If the peroxisomes are integrated into vacuoles, the benzylpenicillin synthesized in the peroxisomal matrix would be transferred to vacuoles and might be later secreted out of the cells. Although fusion of the vacuoles to the plasma membrane by an exocytosis process is possible, there is no evidence to suggest that this might be a major mechanism of penicillin secretion.10

3.24.3.4

Compartmentalization of the Cephalosporin Biosynthetic Pathway

There is also evidence for the compartmentalization of the cephalosporin biosynthesis pathway in A. chrysogenum. This is an example of a complex secondary metabolism pathway where several internal transporters are involved (Fig. 5). In A. chrysogenum, it was reported that the enzymes of the cephalosporin biosynthesis pathway, namely, ACVS, IPN synthase, expandase/hydroxylase, and DAC acetyltransferase, have a cytosolic location.10 However, the central step of the cephalosporin biosynthetic pathway (conversion of IPN to penicillin N by the two-component CefD1–CefD2 epimerization system) seems to be located in microbodies. This observation is based on the presence of canonical peroxisomal targeting sequences in these two proteins.10 Moreover, the optimum pH for the in vitro conversion of IPN into penicillin N in A. chrysogenum cell-free extracts was 7.0, which is coincident with the estimated pH of the peroxisomal lumen. The peroxisomal location of the epimerization system is also supported by the fact that CefD1 and CefD2 homologues of P. chrysogenum have been identified in the peroxisome matrix by mass spectrometry.22 This implies the presence of specific transport systems for precursors and intermediates across the peroxisomal membrane. As was indicated for P. chrysogenum, an active IPN transport system must be present in the peroxisomal membrane to ensure an adequate pool of IPN inside microbodies. In addition, since the

plasma membrane cell wall

mitochondrion

vacuole

α AA

Cys α AA

α AA Cys Val

ACV ACVS

IPNS

IPN

CefT

microbody

IPN

CefD1 nucleus

CPC

???

CefD2 cytosol

CPC

CefM

DACAT

DAC

PenN

PenN

DAOC E-H

CefT

PenN

E-H

septum

Figure 5 Compartmentalization of the cephalosporin C biosynthetic pathway in Acremonium chrysogenum. aAA, L-a-aminoadipic acid; ACVS, ACV synthetase; CPC, cephalosporin C; Cys, L-cysteine; DAC, deacetylcephalosporin C; DACAT, DAC acetyltransferase; DAOC, deacetoxycephalosporin C; E-H, expandase/hydroxylase; IPNS, IPN synthase; PenN, penicillin N; Val, L-valine.

Penicillins and Cephalosporins

291

expandase/hydroxylase and DAC acetyltransferase activities are located in the cytosol, penicillin N has to be transported out of the peroxisomal matrix. One key transporter that is involved in the secretion of penicillin N from the microbody lumen to the cytosol has been recently identified. This transporter is encoded by the cefM gene, which is located in the early cephalosporin cluster (chromosome VII) downstream of the cefD1 gene. The cefM gene encodes a drug efflux pump protein from the Family 3 of major facilitator superfamily (MFS) class of membrane proteins that is located on microbodies. Therefore, after epimerization of IPN to penicillin N inside the microbodies, the latter compound is transported by means of the CefM carrier to the cytosol, where the two last enzymes of the cephalosporin pathway synthesize cephalosporin C.10 Secretion of intermediates of the cephalosporin C biosynthetic pathway to the culture medium appears to be performed by the CefT transporter, which is a multidrug efflux pump protein from the same MFS class as the CefM protein. The CefT protein, which is encoded by the cefT gene, has 12 transmembrane spanners (TMSs) and contains all characteristic motifs of the drug: Hþ antiporter 12-TMS group of the MFS proteins. Although this transporter participates in the secretion of cephalosporin C, it is not the main transporter for this antibiotic. Recently, heterologous expression of the cefT gene in the cephalosporin producer P. chrysogenum TA98, a strain carrying the cephalosporin biosynthesis genes, revealed that the CefT protein is functional in P. chrysogenum, acting as a hydrophilic b-lactam transporter involved in the secretion of hydrophilic b-lactams containing the a-aminoadipic acid side chain (IPN and penicillin N). In addition, when the cefT gene was expressed in the parental strain P. chrysogenum Wisconsin 54-1255, it resulted in an increased secretion of IPN and a drastic reduction of benzylpenicillin production.10 Similar results were obtained in a P. chrysogenum strain engineered to produce adipyl-7-amino-3-carbamoyloxymethyl-3-cephem-4-carboxylic acid (ad7-ACCCA). Expression of the A. chrysogenum cefT gene in the P. chrysogenum strain resulted in almost a twofold increase in cephalosporin (ad-7-ACCCA) production. CefT is correctly targeted to the P. chrysogenum plasma membrane, which indicates that this protein may have a similar location in A. chrysogenum.10,23

3.24.4

Biotechnological Implications in the Biosynthesis of Penicillins and Cephalosporins

The discovery of penicillin and cephalosporin was just the starting point for the modern chemotherapy era. The combination of industrial programs and biotechnology has led to the biosynthesis of new compounds and antibiotic overproduction.

3.24.4.1

Strategies Applied to the Production of Penicillins

Penicillin high-producing strains developed along industrial strain improvement programs and large-scale production of 6-APA are the basis for the biosynthesis of semisynthetic penicillins.

3.24.4.1.1

Industrial Strain Improvement and Genetic Engineering

Since the isolation of P. chrysogenum NRRL 1951 from an infected cantaloupe in 1943, this microorganism has been subjected to classical mutagenesis along industrial strain improvement programs. After several rounds of mutagenesis, the Wisconsin Q-176 strain was obtained. This strain is the original ancestor of the Wisconsin line of strains, which have given rise to penicillin highproducing strains such as the P2 strain of Panlabs (Taiwan), the DS04825 strain obtained at DSM (The Netherlands), or the ASP-78 and the E1 strains, both obtained at Antibioticos S.A. (Spain). These strains are the parents of those overproducer mutants currently used for the industrial production of penicillin, which reaches titers of more than 50 g l1 in fed-batch cultures. The mutagenesis undergone by the P. chrysogenum strains has introduced several important modifications, which have been partially characterized.17,24 The amplification of the genomic region that includes the three penicillin biosynthetic genes, pcbAB, pcbC, and penDE, is one of the well-characterized modifications. These genes are arranged in a single cluster located in a DNA region present as a single copy in the genome of the wild-type NRRL 1951 and Wisconsin 54-1255 strains (laboratory reference strain), but that is amplified in tandem repeats in penicillin-overproducing strains.16 The phlA gene encoding the phenylacetyl-CoA ligase and the ppt gene encoding the PPTase, which activates the ACVS, are not located in the amplified region.13,25 Many of the improved penicillin producers contain several copies of the amplified region, such as the E1, which contains 12–14 copies. The mechanism of gene amplification is intriguing, but it has been suggested that the conserved hexanucleotides located at the borders of the amplified region may be hot spots for site-specific recombination after mutagenesis.16 The homogentisate pathway has also undergone modifications by the industrial improvement programs. This pathway is used for the catabolism of phenylacetic acid (the side-chain precursor in the biosynthesis of benzylpenicillin) and it has been shown that it is diminished in Wisconsin 54-1255 and presumably in derived strains as well. This is due to modifications in a microsomal cytochrome P450 monooxygenase encoded by the pahA gene (Pc21g14280), which leads to a reduced degradation of phenylacetic acid and to penicillin overproduction.17 In addition to these genetic modifications, microbodies (peroxisomes) are also known to be more abundant in high-producing strains.17 These organelles are involved in the final steps of the penicillin pathway, since activation of the side-chain precursor and its addition to the penam core take place in the peroxisomal matrix. With the recent publication of the P. chrysogenum genome17 and proteome,24 some light has been shed on the molecular basis for improved productivity. Transcription of genes involved in the biosynthesis of amino acid precursors for penicillin biosynthesis, as

292

Penicillins and Cephalosporins

well as of genes encoding microbody proteins, was higher in high-producing strains.17 In addition, the increase in penicillin production along the industrial strain improvement program seemed to be also a consequence of complex metabolic reorganizations, pointing to oxidative stress responses as adaptation mechanisms for penicillin overproduction24. As has been indicated so far, the rounds of classical mutagenesis introduced random modifications that led to the development of penicillin-overproducing strains. In addition to this, the advances in genetics and molecular biology have allowed researchers to engineer P. chrysogenum strains in order to introduce punctual modifications with a positive effect on the penicillin titers. The increase in penicillin production has been related to the overexpression of different genes, such as the ppt gene encoding PPTase,25 or the laeA gene encoding the secondary metabolism global regulator PcLaeA.19

3.24.4.1.2

Production of Semisynthetic Penicillins

The strategy followed to produce specific penicillins the positive influence of which is well known since the beginning of industrial penicillin production consisted of feeding penicillin fermentations with side-chain precursors. The most important penicillins obtained by this method are benzylpenicillin or penicillin G (phenylacetate as side chain) and phenoxymethylpenicillin or penicillin V (phenoxyacetate as side chain), which are called biosynthetic penicillins. If no side-chain precursor is added to a submerged fermentation, the microorganism will produce several different natural penicillins, such as penicillin F and K, which contain 3-hexenoic acid and octanoic acid as side chains, respectively. Therefore, phenylacetic acid must be supplemented in excess to promote the biosynthesis of benzylpenicillin rather than other natural penicillins. Phenylacetic and phenoxyacetic acids are industrially used as side-chain precursors for the production of benzylpenicillin and phenoxymethylpenicillin in submerged fermentations. Purification of penicillins is done by a straightforward crystallization adding potassium acetate. These penicillins constitute the precursors of semisynthetic penicillins and cephalosporins, since chemical or enzymatic release of the side chain gives rise to the 6-APA structural core. The development of semisynthetic penicillins began in the late 1950s after the isolation of 6-APA from fermented broths.5 Industrial production of 6-APA is currently based upon the enzymatic deacylation of either or both biosynthetic penicillins, since the method of chemical deacylation of these penicillins to yield 6-APA is polluting and nonprofitable. The traditional chemical synthesis of 6-APA consisted of a one-pot deacylation of the fermentation product penicillin G and began around 1970. This procedure, which required hazardous chemicals and solvents, remained in use for 15–20 years until it was largely replaced by the enzymatic hydrolysis of penicillin G or V, which is economically viable only when immobilized biocatalysts are used. Although penicillin V shows higher stability in aqueous solutions at lower pH during extraction from the fermented broth, penicillin G is the molecule of choice in the manufacture of 6-APA and 85% of 6-APA produced worldwide is from penicillin G. Penicillin G acylase and penicillin V acylase are the enzymes used in this process, since they release the benzyl and phenoxy moieties, respectively, in a very efficient way.26 Penicillin acylases are both intracellular and extracellular enzymes mainly obtained from recombinant strains of Escherichia coli (for penicillin G acylase) or Fusarium oxysporum (for penicillin V acylase). The large-scale production of these proteins is achieved by nutrient-controlled metabolism and by genotypic changes of the producing microorganisms. Isolation of mutant microbial strains with enhanced constitutive production, and cloning and expression of these enzymes in different hosts for large-scale enzyme production constitute the major aims for current research.26 Once the 6-APA core is obtained, different side chains are incorporated, thus giving rise to semisynthetic penicillins, which can be grouped into five categories: antistaphylococcal penicillins, aminopenicillins, carboxypenicillins, ureidopenicillins, and b-lactamase-resistant penicillins.27 The wide variety of semisynthetic penicillins currently available is the result of the effort carried out to improve bioavailability, antibacterial spectrum, stability, tolerance, and effectiveness against a wide range of organisms, including most streptococcal and staphylococcal species, aerobic Gram-negative organisms, and many anaerobic organisms. Antistaphylococcal penicillins are moderately effective against pneumococci and streptococci and highly active against most strains of staphylococci. Members of this group are cloxacillin, dicloxacillin, methicillin, nafcillin, and oxacillin. Aminopenicillins, represented by ampicillin and amoxicillin, are active against most aerobic Gram-positive cocci (Staphylococcus aureus is usually resistant) and anaerobic Gram-positive bacteria. Ampicillin is also effective against some aerobic Gram-negative bacilli such as E. coli, Proteus mirabilis, and Haemophilus influenzae. Carboxypenicillins have, in general, a similar spectrum of activity as that of ampicillin but are also active against Pseudomonas aeruginosa. Members of this family of antibiotics are carbenicillin and ticarcillin, the latter 4 times more potent than carbenicillin. Ureidopenicillins have a wider spectrum of activity compared to the carboxypenicillins. Piperacillin and mezlocillin have greater activity against Gram-negative enteric organisms and also provide coverage against most anaerobes and enterococcus, but are not effective against S. aureus and some strains of E. coli and Klebsiella pneumoniae. Some of these semisynthetic penicillins are given in combination with b-lactamase inhibitors, such as the combination amoxicillin/clavulanic acid or ampicillin/sulbactam, to improve their efficacy.

3.24.4.2

Production of Cephalosporins and Genetic Engineering of P. chrysogenum and A. chrysogenum

There are two main groups of cephalosporin antibiotics: the first one is derived from synthetic penicillins (G or V) and the second one from cephalosporin C.

Penicillins and Cephalosporins 3.24.4.2.1

293

Penicillin-Derived Cephalosporins

The penicillin-derived cephalosporins are mainly based on 7-aminodeacetoxycephalosporanic acid (7-ADCA), which is produced after the chemical expansion of the benzylpenicillin ring (yielding phenylacetyl-7-ADCA) followed by an enzymatic deacylation that removes the aromatic side chain.28 Chemical ring expansion is an expensive and polluting multistep process, instead of which production of 7-ADCA through other processes is desirable. An alternative method to produce 7-ADCA was achieved through the genetic engineering of P. chrysogenum, which cannot naturally produce cephalosporins. The high capacity of P. chrysogenum to produce antibiotics, together with the higher costs derived from A. chrysogenum fermentations, has promoted the production of cephalosporanic ring-derived antibiotics in P. chrysogenum. This microorganism has been genetically modified to express different combinations of the cephalosporin biosynthetic genes obtained from different cephalosporin producer microorganisms. By introduction of the cefE gene from S. clavuligerus or the cefEF gene of A. chrysogenum, P. chrysogenum was able to expand the penicillin thiazolidine ring to a six-membered dihydrothiazine ring. Feeding adipic acid as a side-chain precursor resulted in the production of adipyl-6-APA, which rapidly expanded to adipyl-7-ADCA. This strategy also leads to the production of adipyl 7-aminocephalosporanic acid (adipyl-7-ACA) if the acetyltransferase gene (cefG) is also introduced. Following a similar approach, the cefEF gene from A. chrysogenum and the cmcH gene of S. clavuligerus (encoding a carbamoyl transferase) were expressed in P. chrysogenum, giving rise to the production of ad7-ACCCA, which is an interesting semisynthetic cephalosporin precursor from the stability point of view23. The previous biological routes for the production of 7-ADCA have been developed in the penicillin producer fungus P. chrysogenum, basically because early attempts to produce 7-ADCA in A. chrysogenum after inactivation of the cefEF gene were unsuccessful. However, the production of 7-ADCA in A. chrysogenum was achieved through another approach.29 This strategy was based upon the overexpression of the S. clavuligerus cefE gene in an industrial strain of A. chrysogenum previously disrupted in the cefEF gene. This strain accumulates DAOC, which is the starting material for 7-ADCA production through two enzymatic steps. During the first step, DAOC is transformed into ketoadipyl-7-ADCA by a D-amino acid oxidase (DAO). This compound spontaneously reacts with the hydrogen peroxide produced in this reaction yielding glutaryl-7-ADCA. In the second step, glutaryl-7ADCA is further hydrolyzed to 7-ADCA by means of a glutaryl acylase (GLA). The advantage of this system is the efficiency of the process from both industrial and environmental points of view, since dangerous and expensive chemical steps are avoided.

3.24.4.2.2

Cephalosporin C-Derived Cephalosporins

The cephalosporin C obtained as a secondary metabolite from A. chrysogenum fermentations shows some advantages, such as the resistance to staphylococcal penicillinases or selective toxicity, but the weak antibacterial activity of this antibiotic promoted the use of this antibiotic as precursor for the synthesis of semisynthetic cephalosporins derived from 7-ACA. Current strains used for the production of cephalosporin C are derived from the Brotzu‘s initial isolate and have been obtained by random mutagenesis through strain improvement programs, reaching the titers of 20–25 g l1. In contrast to penicillin highproducing strains, cephalosporin industrial strains contain only one copy of the cephalosporin biosynthetic genes. Genetic manipulation of some genes involved in the biosynthesis and transport of cephalosporin C or its intermediates has been successfully carried out to improve productivity or to direct the fermentation to the synthesis of new products. Examples of genetic manipulation with a positive effect on cephalosporin C biosynthesis are the overexpression of the cefT gene or the cefG gene.28 The feasibility of introducing new biosynthetic capabilities into A. chrysogenum through the combination of fungal and bacterial genes was proved through one biological procedure to obtain 7-ACA directly by fermentation. This method consisted of introducing the 7-ACA biosynthetic operon in A. chrysogenum. This operon is composed of the genes encoding DAO from Fusarium solani and GLA from Pseudomonas diminuta and allowed the strain to convert cephalosporin C into 7-ACA and two side products, 7-ADCA and 7-aminodeacetylcephalosporanic acid (7-DAC). Although the strain produced detectable levels of 7-ACA, these levels were not significant for commercial purposes.29 Other attempts to increase cephalosporin C production in A. chrysogenum have been focused on aerobic metabolism, since it is well known that oxygen regulates the biosynthesis of cephalosporin. The improvement of A. chrysogenum aerobic metabolism was achieved through the overexpression of the Vitreoscilla gene encoding the oxygenbinding heme protein hemoglobin, which led to significant higher yields of cephalosporin C30. The cephalosporin C produced by A. chrysogenum has to be purified through a complex and expensive process including several chromatographic steps. In the first step, biomass and antibiotic-containing broth are separated by filtration or centrifugation. The filtered broth is passed through a couple of hydrophobic interaction chromatography columns to remove proteins, peptides, salts, and other impurities, including DAC and DAOC. The first column (scavenger) is filled with an adsorber resin (e.g., Diaion HP20) and allows the binding of hydrophobic colored compounds with a minimal adsorption of cephalosporin C. The second column (adsorber) is filled with a hydrophobic resin such as Sepabeads SP700, which binds cephalosporin C. Elution of the antibiotic is achieved by a pH change and further purification is carried out through ion exchange chromatography, which removes the remaining color. After this purification process, cephalosporin C can be isolated as either the dried sodium or potassium salt. Alternatively, cephalosporin C in solution can be directly converted to yield 7-ACA, advanced 70 intermediates or advanced 30 intermediates.28 Cephalosporin C can be converted to 7-ACA by either a chemical or an enzymatic process, which removes the 7-aminoadipyl side chain. The original chemical cleavage has been superseded and improved by processes using phosphoric pentachloride, and although the yield and product quality produced by this method are excellent, the need for organic solvents and the production of toxic wastes have gradually led to the replacement of the chemical cleavage by the environmentally safer enzymatic cleavage.28 The industrial conversion of cephalosporin C into 7-ACA is currently achieved through a two-step process that employs the enzymes

294

Penicillins and Cephalosporins

H N

H2N O

H COOH

S O

S

H2N OCOCH3

N

OCOCH3

N

O

COOH

COOH

CPC

7-ACA HOOC-(CH2)3-COOH

O2

GLA

DAO H2O2.NH3 H N

O COOH

O

H N

S O

OCOCH3

N

HOOC

O

S O

OCOCH3

N COOH

COOH

KA-7-ACA

GL-7-ACA Nonenzymatic decarboxylation

H2O2

CO2

Figure 6 Two-step enzymatic conversion of cephalosporin C into 7-aminocephalosporanic acid (7-ACA). CPC, cephalosporin C; DAO, D-amino acid oxidase; GLA, glutaryl acylase; GL-7-ACA, glutaryl-7-ACA; KA-7-ACA, keto-adipyl-7-ACA.

DAO and GLA (Fig. 6). In the first step, cephalosporin C is oxidatively deaminated to keto-adipyl-7-ACA (KA-7-ACA) by means of the DAO, and the peroxide released in this reaction induces the spontaneous oxidative decarboxylation to glutaryl-7-ACA (GL-7ACA). During the second step, this compound is hydrolyzed to 7-ACA and glutarate by the GLA. Since GLA also accepts KA-7-ACA as substrate and the reaction is strongly inhibited by GL-7-ACA, another process for the enzymatic conversion of cephalosporin C into 7-ACA combines the enzymatic cleavage of cephalosporin C by DAO with catalase and GLA in a single reaction vessel. DAO catalyzes the oxidative deamination to KA-7-ACA and catalase removes all the peroxide generated during the reaction, thus avoiding the formation of GL-7-ACA. Under these conditions, GLA will convert KA-7-ACA into 7-ACA and ketoglutarate.

3.24.4.2.3

Semisynthetic Cephalosporins

Commercial cephalosporins are all semisynthetic and are derived from 7-ACA, 7-ADCA, 7-DAC, or the corresponding nuclei of the cephamycins. The addition of a new side chain at position 70 or alteration of the 30 side chain will lead to advanced intermediates with modified antibacterial spectrum, b-lactamase stability, and pharmacokinetic properties. The medically useful cephalosporins are categorized as first-, second-, third-, or fourth-generation products depending on their spectrum and resistance to enzymatic degradation.5 First-generation cephalosporins are moderate-spectrum agents generally susceptible to b-lactamases, but are not as effective against anaerobic microorganisms as the penicillins. Although they are most effective against aerobic bacteria such as penicillinase-producing, methicillin-susceptible staphylococci and streptococci, they are not the drugs of choice for such infections. They are also effective against several Gram-negative bacteria such as E. coli, Proteus, Klebsiella, Salmonella, Shigella, and Enterobacter species, but have no activity against Bacteroides fragilis, enterococci, methicillin-resistant staphylococci, Pseudomonas, Acinetobacter, Enterobacter, indole-positive Proteus, or Serratia. Members of this family of antibiotics are cephalothin, cephaloridine, cephapirin, cefazolin, cephalexin, cephradine, and cefadroxil. Second-generation cephalosporins are slightly less effective against Gram-positive organisms, but have a greater Gram-negative spectrum and are more effective against Gram-negative organisms such as Klebsiella, E. coli, and Proteus species. They are ineffective against P. aeruginosa, Actinobacter species, and a good number of obligate anaerobes. The second-generation cephalosporins are relatively more resistant to b-lactamases, but are characterized by poor penetration of the blood–brain barrier. Examples of this generation of cephalosporins are cefamandole, cefotiam, cefaclor, cefuroxime, ceforanide, cefonicid, and cefprozil. In addition,

Penicillins and Cephalosporins

295

cefmetazole, cefotetan, and cefoxitin have antianaeorobic activity, although they belong to the cephamycin group of b-lactam antibiotics. Third-generation cephalosporins are usually highly resistant to b-lactamases, have increased activity against Gram-negative aerobic organisms, and have a broad spectrum of activity, being effective against Proteus vulgaris, Enterobacter species, Citrobacter species, Haemophilus species, Neisseria species, and Moraxella species. However, these drugs exhibit only moderate activity against Gram-positive bacteria (in particular, those members available in an oral formulation, and those with antipseudomonal activity) and are inferior in activity against staphylococci, although they are generally effective against penicillin-resistant Streptococcus pneumoniae. Some third-generation cephalosporins are able to cross the blood–brain barrier and penetrate the central nervous system, making them effective in therapy for bacterial meningitis caused by susceptible bacteria. Members of this group of antibiotics are ceftiofur, ceftriaxone, cefsulodin, cefotaxime, cefoperazone, ceforanide, ceftazidime, cefpodoxime, cefixime, ceftibuten, cefdinir, and ceftizoxime. The fourth-generation cephalosporins feature extended-spectrum activity and chemical characteristics that may lead to reduced development of resistance by Gram-negative organisms. They also have a greater resistance to b-lactamases than the thirdgeneration cephalosporins. Many can cross the blood–brain barrier and are effective in meningitis. They are also used against P. aeruginosa. Cefclidine, cefepime, cefluprenam, cefoselis, cefozopran, cefpirome, and cefquinome are members of this generation of cephalosporins. Some of these semisynthetic cephalosporins are given in combination with b-lactamase inhibitors, such as the combination cefoperazone/sulbactam (Sulperazone) to improve their efficacy. The development of new semisynthetic cephalosporins in the pharmacology programs is under way and new generations of cephalosporins will be available in the future. Efforts are being focused to increase spectrum, b-lactamase stability, and bioavailability.

3.24.5

Future Outlook

Knowledge of the molecular basis leading to increased titers of antibiotics is still scarce, although ‘omics’ studies are currently providing new clues on this subject. Research will be focused on the transport of biosynthetic intermediates, secretion of the final products, the delicate interplay between primary and secondary metabolism, the interactions of the components involved in b-lactam biosynthesis, the regulatory circuits and regulatory proteins, and signal transduction pathways. These aspects will help to understand the possible physiological and ecological role of b-lactam antibiotics in the producing fungi. In addition, the application of this knowledge and the use of biotechnology will lead to a further increase in the b-lactam titers and to the development and marketing of novel related and improved versions of the existing b-lactam antibiotics.

See Also: 1.12 Cell Preservation Technology; 2.30 Biofilters; 3.23 Antibiotics: The Miracle Menaced; 3.27 Plant Secondary Metabolites; 3.52 Metabolic Engineering.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

Katz, E.; Demain, A. L. The Peptide Antibiotics of Bacillus: Chemistry, Biogenesis, and Possible Functions. Bacteriol. Rev. 1977, 41, 449–474. Kong, K. F.; Schneper, L.; Mathee, K. Beta-lactam Antibiotics: From Antibiosis to Resistance and Bacteriology. APMIS 2010, 118, 1–36. Fleming, A. On the Antibacterial Action of Cultures of a Penicillium with Special Reference to Their Use in the Isolation of B. influenza. Br. J. Exp. Pathol. 1929, 10, 226–236. Clutterbuck, P. W.; Lovell, R.; Raistrick, H. Studies in the Biochemistry of the Microorganisms. XXVI. The Formation from Glucose by Members of the Penicillium chrysogenum Species of a Pigment, an Alkali Soluble Protein and Penicillin. The Antibacterial Substance of Fleming. Biochem. J. 1932, 26, 1907–1918. Demain, A. L.; Elander, R. P. The Beta-lactam Antibiotics: Past, Present, and Future. Antonie Van Leeuwenhoek 1999, 75, 5–19. Brotzu, G. Richerche su di un nova antibiotico. Lavori dell‘Istituto d‘Igiene di Cagliari 1948, 1, 1–11. Hamilton-Miller, J. M. Sir Edward Abraham‘s Contribution to the Development of the Cephalosporins: A Reassessment. Int. J. Antimicrob. Agents 2000, 15, 179–184. Brakhage, A. A.; Thön, M.; Spröte, P.; et al. Aspects on Evolution of Fungal Beta-lactam Biosynthesis Gene Clusters and Recruitment of Trans-acting Factors. Phytochemistry 2009, 70, 1801–1811. Liras, P. Biosynthesis and Molecular Genetics of Cephamycins. Cephamycins Produced by Actinomycetes. Antonie Van Leeuwenhoek 1999, 75, 109–124. Martín, J. F.; Ullán, R. V.; García-Estrada, C. Regulation and Compartmentalization of Beta-lactam Biosynthesis. Microbial Biotechnol. 2010, 3, 285–299. Aharonowitz, Y.; Cohen, G.; Martín, J. F. Penicillin and Cephalosporin Biosynthetic Genes: Structure, Regulation, and Evolution. Annu. Rev. Microbiol. 1992, 46, 461–495. Whiteman, P. A.; Abraham, E. P.; Baldwin, J. E.; et al. Acyl Coenzyme A: 6-aminopenicillanic Acid Acyltransferase from Penicillium chrysogenum and Aspergillus nidulans. FEBS (Fed. Eur. Biochem. Soc.) Lett. 1990, 262, 342–344. Lamas-Maceiras, M.; Vaca, I.; Rodríguez, E.; et al. Amplification and Disruption of the Phenylacetyl-CoA Ligase Gene of Penicillium chrysogenum Encoding an Aryl-capping Enzyme that Supplies Phenylacetic Acid to the Isopenicillin N Acyltransferase. Biochem. J. 2006, 395, 147–155. García-Estrada, C.; Vaca, I.; Ullán, R. V.; et al. Molecular Characterization of a Fungal Gene Paralogue of the Penicillin PenDE Gene of Penicillium chrysogenum. BMC Microbiol. 2009, 9, 104. Ullán, R. V.; Casqueiro, J.; Bañuelos, O.; et al. A Novel Epimerization System in Fungal Secondary Metabolism Involved in the Conversion of Isopenicillin N into Penicillin N in Acremonium Chrysogenum. J. Biol. Chem. 2002, 277, 46216–46225. Fierro, F.; García-Estrada, C.; Castillo, N. I.; et al. Transcriptional and Bioinformatic Analysis of the 56.8 kb DNA Region Amplified in Tandem Repeats Containing the Penicillin Gene Cluster in Penicillium chrysogenum. Fungal Genet. Biol. 2006, 43, 618–629. van den Berg, M. A.; Albang, R.; Albermann, K.; et al. Genome Sequencing and Analysis of the Filamentous Fungus Penicillium chrysogenum. Nat. Biotechnol. 2008, 26, 1161–1168.

296

Penicillins and Cephalosporins

18. Martín, J. F. Molecular Control of Expression of Penicillin Biosynthesis Genes in Fungi: Regulatory Proteins Interact with a Bi-directional Promoter Region. J. Bacteriol. 2000, 182, 2355–2362. 19. Kosalková, K.; García-Estrada, C.; Ullán, R. V.; et al. The Global Regulator LaeA Controls Penicillin Biosynthesis, Pigmentation and Sporulation, but Not Roquefortine C Synthesis in Penicillium chrysogenum. Biochimie 2009, 91, 214–225. 20. Shwab, E. K.; Keller, N. P. Regulation of Secondary Metabolite Production in Filamentous Ascomycetes. Mycol. Res. 2008, 112, 225–230. 21. Koetsier, M. J.; Gombert, A. K.; Fekken, S.; et al. The Penicillium chrysogenum AclA Gene Encodes a Broad-substrate-specificity Acyl-coenzyme a Ligase Involved in Activation of Adipic Acid, a Side-chain Precursor for Cephem Antibiotics. Fungal Genet. Biol. 2010, 47, 33–42. 22. Kiel, J. A.; van den Berg, M. A.; Fusetti, F.; et al. Matching the Proteome to the Genome: The Microbody of Penicillin-producing Penicillium chrysogenum Cells. Funct. Integr. Genom. 2009, 9, 167–184. 23. Harris, D. M.; Westerlaken, I.; Schipper, D.; et al. Engineering of Penicillium chrysogenum for Fermentative Production of a Novel Carbamoylated Cephem Antibiotic Precursor. Metab. Eng. 2009, 11, 125–137. 24. Jami, M. S.; Barreiro, C.; García-Estrada, C.; Martín, J. F. Proteome Analysis of the Penicillin Producer Penicillium chrysogenum: Characterization of Protein Changes during the Industrial Strain Improvement. Mol. Cell. Proteomics 2010, 9, 1182–1198. 25. García-Estrada, C.; Ullán, R. V.; Velasco-Conde, T.; et al. Post-translational Enzyme Modification by the Phosphopantetheinyl Transferase Is Required for Lysine and Penicillin Biosynthesis but Not for Roquefortine or Fatty Acid Formation in Penicillium chrysogenum. Biochem. J. 2008, 415, 317–324. 26. Arroyo, M.; de la Mata, I.; Acebal, C.; Castillón, M. P. Biotechnological Applications of Penicillin Acylases: State-of-the-art. Appl. Microbiol. Biotechnol. 2003, 60, 507–514. 27. Oshiro, B. T. The Semisynthetic Penicillins. Prim. Care Update OB/GYNS 1999, 6, 56–60. 28. Barber, M. S.; Giesecke, U.; Reichert, A.; Minas, W. Industrial Enzymatic Production of Cephalosporin-based Beta-lactams. Adv. Biochem. Eng. Biotechnol. 2004, 88, 179–215. 29. Velasco, J.; Adrio, J. L.; Moreno, M. A.; et al. Environmentally Safe Production of 7-aminodeacetoxycephalosporanic Acid (7-ADCA) Using Recombinant Strains of Acremonium chrysogenum. Nat. Biotechnol. 2000, 18, 857–861. 30. DeModena, J. A.; Gutiérrez, S.; Velasco, J.; et al. The Production of Cephalosporin C by Acremonium chrysogenum Is Improved by the Intracellular Expression of a Bacterial Hemoglobin. Biotechnology 1993, 11, 926–929.

Relevant Websites http://www.abc.net.au/science/slab/florey/story.htm – ABC. http://www.antibioticos-sa.com/2201.html – antibioticos. http://chemicalland21.com/lifescience/phar/6-AMINOPENICILLANIC%20ACID.htm – Chemicalland21.com. http://www.dsm.com/en_US/html/dai/intermediateshomenew.htm – DSM. http://www.dsm.com/en_US/html/dai/semisyntheticpenicillinshomenew.htm – DSM. http://www.dsm.com/en_US/html/dai/semi-synthcef-new.htm – DSM. http://www.freepatentsonline.com/4251442.html – freepatentsonline. http://www.freepatentsonline.com/7339055.html – freepatentsonline. http://www.iit.it/en/drug-discovery-and-development/bio-brotzu.html – Italian Institute of Technology. http://nobelprize.org/nobel_prizes/medicine/laureates/1945/chain-bio.html – Nobelprize.org. http://nobelprize.org/nobel_prizes/medicine/laureates/1945/fleming-bio.html – Nobelprize.org. http://nobelprize.org/nobel_prizes/medicine/laureates/1945/florey.html – Nobelprize.org. http://osdir.com/patents/Organic-compounds/Process-preparation-cephalosporin-intermediate-manufacture-cephalosporin-compounds-06919449.html – osdir. http://botit.botany.wisc.edu/toms_fungi/nov2003.htm – University of Wisconsin plant teaching collection. http://en.wikipedia.org/wiki/Penicillin – Wikipedia. http://en.wikipedia.org/wiki/Cephalosporin – Wikipedia. http://www.biotopics.co.uk/microbes/penici.html – www.BioTopics.co.uk.

3.25

Tetracyclines and Tetracycline Derivatives

ML Nelson, Paratek Pharmaceuticals, Inc., Boston, MA, United States SB Levy, Paratek Pharmaceuticals, Inc., Boston, MA, United States; and Tufts University School of Medicine, Boston, MA, United States © 2011 Elsevier B.V. All rights reserved. This is a reprint of M.L. Nelson, S.B. Levy, 3.25 - Tetracyclines and Tetracycline Derivatives, Editor: Murray Moo-Young, Comprehensive Biotechnology (Second Edition), Academic Press, 2011, Pages 269-283.

3.25.1 3.25.2 3.25.3 3.25.3.1 3.25.3.1.1 3.25.4 3.25.4.1 3.25.4.2 3.25.5 3.25.5.1 3.25.5.1.1 3.25.5.1.2 3.25.5.2 3.25.5.2.1 3.25.5.2.2 3.25.5.3 3.25.5.3.1 3.25.5.3.2 3.25.5.3.3 3.25.5.3.4 3.25.5.4 3.25.5.4.1 3.25.6 3.25.6.1 3.25.6.2 3.25.6.3 3.25.7 3.25.8 3.25.9 3.25.10 References

Introduction and Scope Tetracycline Generations and Origins First-Generation Tetracyclines The Tetracycline Biosynthesis Pathway Tetracyclines Derived From Other Actinomycetes and Microorganisms Antibacterial Uses of the Tetracyclines Antiparasitic Uses of the Tetracyclines Minimal Structural Requirements for Antibacterial Activity in Tetracycline Derivatives First- and Second-Generation Tetracyclines and Their Semisynthetic Modifications Chemistry of the A Ring and Antibacterial Activity A-Ring C2 Modifications of the Tetracyclines Position C4 Derivatives and Dynamics of the 4-Dimethylamino Group Chemistry of the B Ring and Antibacterial Activity Lower Periphery Derivatives and Antibacterial Activity Position C5 Derivatives of 5-OH Tetracyclines Chemistry of the C Ring and Antibacterial Activity The Anhydrotetracyclines The Synthesis and Antibacterial Activity of Methacycline and Second-Generation Tetracyclines C13-alkylthio Bacterial Tetracycline Efflux Protein Inhibitors Synthesis and Antibacterial Activity of Doxycycline Chemistry of the D Ring and Antibacterial Activity Minocycline Semisynthesis Semisynthesis of Third-Generation Tetracyclines: Derivatives of Minocycline, Sancycline, and Doxycycline Semisynthesis of the Glycylcyclines and Tygacil The 9-Aminomethylcyclines and PTK 0796 Transition Metal-Catalyzed Reactions of the Tetracyclines Tetracycline Antibacterial Quantitative Structure–Activity Relationships Antibacterial and General Chemical Properties of the Tetracyclines: Uptake and Membrane Activity Mechanism of Action and Antibacterial Activity Conclusions

297 299 299 299 300 301 301 301 302 302 302 303 303 303 304 304 304 304 305 305 306 306 307 308 308 309 309 309 309 310 310

Glossary

Minimum inhibitory concentration (MIC) Concentration, usually in mg ml 1, of a compound that inhibits the growth of a bacterial culture, usually via turbidimetric methods. Structure–activity relationships The description of how small changes in the molecular structure of a molecule change the activity and therapeutic profile within a family of molecules.

3.25.1

Introduction and Scope

Although there are many antibiotic-producing microorganisms of commercial importance, a soil actinomycete studied in the 1940s started a scientific front that continues to grow in size and scope to this day for the treatment of bacterial diseases. The organism, Streptomyces aureofaciens, discovered by Duggar in 1948,1 produced an unknown compound with unprecedented antibacterial activity because it inhibited both Gram-negative and Gram-positive bacteria.2 More importantly, this novel and mysterious compound was potent against rickettsial diseases, such as Rocky Mountain spotted fever, which was at that time incurable.

Comprehensive Biotechnology, 3rd edition, Volume 3

https://doi.org/10.1016/B978-0-444-64046-8.00167-1

297

298

Tetracyclines and Tetracycline Derivatives

Alongside chloramphenicol, discovered in the same year, gave it ‘wonder’ drug status and helped coin the term ‘broad-spectrum antibiotic’. Even before the chemical structure was fully elucidated, this compound was approved by the Food and Drug Adminstration (FDA) and brought to market as Aureomycin and used with success against other untreatable diseases such as typhoid fever, typhus, and common infections caused by invasive Streptococcus pneumoniae and b-hemolytic streptococci. By 1953, the initial chemical structure of Aureomycin and other fermentation-derived compounds was structurally elucidated and the term ‘tetracycline’ was used to describe the presence of a linearly arranged naphthacene ring system (1) (Fig. 1), while the initial compound described by Duggar was assigned structure (2) and named chlortetracycline.1 Soon, other derivatives and fermentation products were isolated from Streptomyces species and studied, and the compounds oxytetracycline (3) and tetracycline (4) (Fig. 2) became the focus of much research and clinical use around the globe. These three first-generation tetracycline natural products have set the stage for their further chemical evolution and biological evaluation that is the subject of this article. For over 60 years, the tetracyclines have been clinically useful for the treatment of microbial infections, attributed to their ability to inhibit the growth of both Gram-positive and Gram-negative bacteria. More recently and within the past 30 years, they have also been found to possess activity against degenerative states in mammals including humans, particularly processes characterized by inflammation and neurodegenerative diseases.3 The dual activity in both prokaryotic and eukaryotic cells has led to the division of tetracyclines into two groups based upon activity against microorganisms, and the family is now routinely broken into antibiotic and nonantibiotic tetracyclines based upon their inherent antibacterial activity. Chemically, the family is composed of natural products derived from fermentation, semisynthetic derivatives and their subsequent generations, and total synthesis derivatives, a field which began in the 1950s and is now revived and progressing due to the research and chemical methods developed by Myers et al.4 The representative numbers in each subfamily vary, where the semisynthetic group is the most numerous followed by smaller numbers of derivatives found in the total synthesis and natural product families, respectively.

7

6

5

6a

8 D

9

C

4a

11a

11

A

OH D

2

C

B

12a

12

A

NH2

OH

1

OH O

OH O

1 Figure 1

N(CH3)2

OH

3

B

10a

10

CI

4

5a

O

2

Naphthacene and chlortetracycline.

First-generation tetracyclines OH N(CH3)2 OH OH

N(CH3)2

OH

CI

OH NH2 O

OH O

2 Second-generation tetracyclines

OH

OH O

O

N(CH3)2 OH

OH O

O

4

OH

N(CH3)2 OH

NH2 OH O

OH O

3

CI

N(CH3)2

OH

OH NH2

NH2

OH

OH O

NH2

OH

OH O

N(CH3)2

OH

NH2

OH

OH O

N(CH3)2

OH

OH

O

OH O

5

OH O

OH

O

OH O

6

OH O

O

7

Third-generation tetracyclines N(CH3)2

N(CH3)2 OH

O

NH

N H

NH2 OH

OH O

OH O 8

Figure 2

First-, second-, and third-generation tetracyclines.

O

N(CH3)2

N(CH3)2 OH

H N

NH2 OH O

OH

OH O 9

O

Tetracyclines and Tetracycline Derivatives

3.25.2

299

Tetracycline Generations and Origins

The first-generation natural products tetracycline (Fig. 2) are chlortetracycline (2), oxytetracycline (3), and tetracycline (4), while the second-generation semisynthetic derivatives minocycline (5) and doxycycline (6) have been studied and used clinically since the early 1970s and are still used widely. Demeclocycline (7) is a natural product derived from Streptomyces strain mutation and is a valuable starting material for other generations of tetracyclines. Only two third-generation tetracyclines have entered the clinic in the past 30 years, tigecycline (8), trade named Tygacil and approved by the FDA in 2005, while a more recent novel tetracycline, PTK-0796 (9), has reported favorable preclinical bioactivity in phase II human clinical trials, while progressing to phase III human trials.5

3.25.3

First-Generation Tetracyclines

Chlortetracycline (2) was discovered by bioprospecting for antibiotic-producing organisms and is the product of aerobic fermentation from the soil actinomycete S. aureofaciens A377 (NRRL 2209). The next described was oxytetracycline (3), designated Terramycin, isolated from S. rimosus S-3279 (NRRL 2234).6 Catalytic hydrogenation of chlortetracycline led to the antibiotic tetracycline (4), designated Tetracyn, and is known as the core structure for which the entire family of tetracyclines derives its name. Tetracyn was technically the first semisynthetic tetracycline derivative used clinically, although now it is produced solely via fermentation by biochemical mutants of Streptomyces and/or the manipulation of the fermentation medium.

3.25.3.1

The Tetracycline Biosynthesis Pathway

The biosynthesis of the tetracyclines has been studied primarily in Streptomyces spp. and the pathways of oxytetracycline and chlortetracycline production have been determined by chemically or biologically manipulating producing organisms, changing their metabolism while charting the pathways and enzymology leading to their production (Fig. 3). The sequence of biochemical reactions that direct their biosynthesis reveals that the polyketide backbone, ring foldings and closures, and exocyclic modifications that produce the naphthacene ring scaffold and its numerous chemical functional groups are conserved between species and can also occur in other genera, including Nocardia, Dactylosporangium, Actinomadura, and Penicillium spp., among other aerobicfermentative microorganisms.7 O SCoA + O

O

O

O

OH

O CO2H

C6

CoAS O

I O

O

O

OH

OH

II R=CONH2 H3C N CH3 OH B A R O

15

B O

B OH O

O

12a

OH

OH

CI

OH

OH O

OH 2

OH OH N(CH3)2 OH

OH O

OH

OH O 3

N(CH3)2 OH

OH

R O

4

O

R

12 N(CH3)2 OH

OH

OH O

First-generation tetracycline biosynthetic pathway.

OH O

R O

S. rimosus

Figure 3

A

13

OH

O 16

B

A OH

OH

OH

OH O

OH

OH

14

H3C N CH3 OH B A R

O

O

R O

OH OH 11

A OH

OH

OH

10

S. aureofaciens OH

OH

OH

OH

R

OH OH

NH2

OH

OH O

OH

R

R O

HO

OH

OH

R

R O

300

Tetracyclines and Tetracycline Derivatives

Although not every biosynthesis intermediate has been isolated, radiolabel incorporation experiments with malonate and acetate subunits indicate that the tetracycline ring system is formed through addition of malonylamyl-CoA and acetyl-CoA units (Fig. 3, I) to form a linear nona- or decaketide precursor II, followed by a concerted series of enzyme-mediated foldings, ring closures, and stereochemical transformations by type II polyketide synthases, forming the basic naphthacene ring system. Methylation at position C6 yields 6-methylpretetramide (10), a stable and isolable precursor, which is subsequently modified to the intermediate 4-hydroxy-6-methylpretetramide (11). Pretetramide substrates are further modified by hydroxylase and ketoreductase enzymes, yielding 4-hydroxy-12-deoxyanhydrotetracycline (12). The introduction of the C12a hydroxyl group by hydroxylases and ketoreductase modification induces the structurally planar pretetramide molecule to the common three-dimensional (3D) pharmacophore pattern associated with bioactive tetracyclines, where the A ring is now right angular compared to the BCD rings, and its antibacterial shape is formed (13). Amination of position C4 forms 4-aminoanhydrotetracycline (14), followed by dimethylation resulting in 4-dimethylaminoanhydrotetracycline (15), an intermediate that has not been isolated within the pathway and is thought to be toxic to the bearing organism. Hydroxylation at position C6 and the formation of an unsaturated center between positions C5a and C11a produce a pivotal intermediate, 6-hydroxy-5a(11a)-dehydrotetracycline (16). With producer strain S. rimosus, the unsaturated substrate is hydroxylated at position C5, producing oxytetracycline (3), while with S. aureofaciens hydroxylation does not occur, forming tetracycline (4). Further, halogenation via haloperoxidases produces C7 chlortetracycline (2). Tetracycline (4) can also be produced directly by S. aureofaciens mutants with depressed chlorination ability or by the addition of heterocyclic chlorination inhibitors.

3.25.3.1.1

Tetracyclines Derived From Other Actinomycetes and Microorganisms

Mutant strains of Streptomyces are now routinely used to produce commercially available tetracyclines, while less common natural tetracyclines, such as demeclocycline (6) and demecycline (17) (Fig. 4), are produced via induced biochemical mutations and the modification of key steps within its biosynthetic pathway.3 Demecycline and demeclocycline are produced by C6 methylationblocked mutants, while demeclocycline, which is used clinically, is obtained by pathway inhibition using the C6 methylation inhibitor sulfaguanidine. The presence of a C6 hydroxyl group and the C7 chlorine in demeclocycline has also made this compound a useful synthetic intermediate in the semisynthesis of second-generation tetracyclines related to sancycline (18), the chemically simplest tetracycline that still maintains antibacterial activity, and minocycline (5).

OH

N(CH3)2

N(CH3)2

OH

OH

OH

O

OH O

O

OH

O

OH O

O

N(CH3)2 OH

R

CI H3CO

OH

OH 20

O

CH3

OH

21 O

NH2 OH O

OCH3 NHOH

O

OH

O

O

22 O

NH2

OH

CI

N(CH3)2 OH

OH

H3CO

CH3 OH OH O

O

CI

OH O

23

OH

N(CH3)2 OH

H3CO

NH2 CH3

OH

O

OH

OCH3 NO2

NH

OH

OH

O

O

OH

OH O

24

OH O

O

25

OH

OH OH

OH

OH

H3CO

OH

OH

NH2

OH

O

O

26 Structurally diverse tetracycline natural products.

OH

H3CO

NH2

CH3 OH

Figure 4

R

N(CH3)2

OH

OH O

O

O

19

CH3

OH OH O

OH O

18

OH OH

H3CO

CH3 OH

OH

17

OH O

OH

NH2

NH2 OH

OH O

N(CH3)2

OH

CI

OH OH O 27

O

O

OH

OH O

O 28

O

O

Tetracyclines and Tetracycline Derivatives

301

The class Actinomycetes is composed of over 3000 distinct species of soil microorganisms in 40 distinct genera, many of which have been found to produce tetracycline natural products.7 Some Streptomyces species can produce novel and bioactive tetracyclines during biosynthesis by malonyl-CoA pathway changes, substituting the usual C2 carboxamide group with an acetyl moiety to yield 2-acetylchlortetracycline (19) and terramycin X (20), structurally unusual tetracyclines with antibacterial activity.3,6 Carbohydrate derivatives, dactylocyclines A (21) and B (22), are produced by Dactylosporangium sp. SC14051 and are chemically unique, one of the few tetracyclines found possessing an aminoglycoside substructure at position C6 and a C4a hydroxyl group, compared to the natural tetracyclines. Chelocardin (23), produced by Nocardia sulphurea (NRRL 2822), is composed of aromatized C and D rings and an acetyl group at position C2 with a diastereomeric C4 amino group, compared to the usual C4 amino group in the natural tetracyclines. All of these compounds were found to have antibacterial activity primarily against Gram-positive bacteria, while chelocardin maintained Gram-negative antibacterial activity. The tetracycline, Sch 33256 (24), was isolated from Actinomadura brunnea and resembled chlortetracycline in addition to a C2 N-methylcarboxamide and a C8 position methoxy group. Other 8methoxylated tetracyclines, such as Sch 34164 (25), possess a C4a hydroxyl group with the methyl substituent at the carboxamide absent. Tetracyclines have also been isolated from fungi, where hypomycetin (26) has been isolated and characterized from the mycophilic fungus Hypomyces aurantius. Another novel tetracycline has been isolated from Penicillium viridicatum, the mycotoxin viridicatumtoxin (27), and has considerable toxicity with concurrent increased lipophilicity. More recently, a tetracycline that has been found to be structurally and biologically unique was isolated during the fermentation of Aspergillus niger, BMS-192548 (28), revealing a tetracycline devoid of the C4 dimethylamino group while possessing a C4a hydroxyl group. This compound was inactive as an antibacterial agent but had unexpected central nervous system activity, binding to neuropeptide Y receptors in mammalian cells with the potential to modify numerous neural and physiological processes.

3.25.4

Antibacterial Uses of the Tetracyclines

In the past 20 years, research into the chemistry and biology of the tetracyclines has been accelerating and the focus of both private companies and academic institutions, with the goals of producing newer and more potent generations of tetracyclines, especially against antibiotic-resistant bacteria. Concurrently, tetracycline derivatives have also found new uses as adjuvants and synergists against bacterial and parasitic infections such as Helicobacter pylori, rickettsial and malarial infections, and more common drug-resistant hospital-acquired pathogens such as MRSA (methicillin-resistant S. aureus) and VRE (vancomycin-resistant Enterococcus).8 Tetracyclines also possess multifactorial and diverse, and in some cases unknown, modes of action against pathogenic targets, giving hope that chemotherapy using tetracycline analogs will help in the fight against antibiotic-sensitive and antibiotic-resistant microbes.

3.25.4.1

Antiparasitic Uses of the Tetracyclines

Against parasites and as a prime example of new antimicrobial targets, the use of doxycycline as a prophylaxis agent against the malarial parasite Plasmodium falciparum is indicated, and recently a mechanism or mode of action at the level of the apicoplast was demonstrated, the parasite equivalent of eukaryotic mitochondria.9 Other pathogens, including multicellular parasites that harbor Wolbachia symbiotic bacteria, are susceptible to the action of tetracyclines and are responsible for diseases such as river blindness and Chagas disease.10 The locus of action turns out to be within the Wolbachia bacteria, enfeebling the parasite with eventual eradication.

3.25.4.2

Minimal Structural Requirements for Antibacterial Activity in Tetracycline Derivatives

All antibiotic tetracyclines have a linearly arranged naphthacene ring (Fig. 5) in common, with oxygen and nitrogen containing functional groups strategically located along the designated lower peripheral region, the 2N region and the C3–C4 region. The lower

Figure 5

Regions of the tetracycline pharmacophore and modifiable positions.

302

Tetracyclines and Tetracycline Derivatives

peripheral region outlines a pharmacophore pattern of hydroxyl, keto–enol, and carbonyl groups spanning the C1–C10 positions, generating tautomeric structures and spatial arrangement responsible for the potency and range of biological activities against bacteria. Chemical modification along the lower peripheral region eliminates ribosome affinity and antibacterial activity, while chemical modification along the upper periphery at positions C5–C9, can also influence activity, depending upon position and nature of the chemical substitution, and derivatives at these positions have produced clinically significant antibiotics. The position C2 exocyclic carbonyl and the C3 keto-enolate group are also required for antimicrobial activity, while broadspectrum activity is dependent on the presence and stereochemical orientation of the C4 dimethylamino group. The 2N amide nitrogen can also be chemically modified, producing compounds of variable antibacterial activity.

3.25.5

First- and Second-Generation Tetracyclines and Their Semisynthetic Modifications

The first-generation tetracyclines have been semisynthetically modified by disjunctive or conjunctive approaches, by removal or subtraction of functional groups, or by addition of substituents at modifiable positions, and thousands are reported synthesized in the literature.11,12 Both methods of modification, either alone or combined, have led to improved potency in tetracycline antibacterial activity and the production of therapeutically and commercially valuable compounds. Chemical changes along the lower periphery hinder the formation of tautomeric and metal-binding substructures needed for antibacterial activity, while their structure–activity relationships (SARs) as antibacterial agents rely primarily on chemically modifiable positions along the upper peripheral region at, positions C4 through C9.

3.25.5.1

Chemistry of the A Ring and Antibacterial Activity

The A ring of the tetracycline antibiotic possesses five different functional groups simultaneously (Fig. 6). Position C1 possesses a carbonyl moiety, C2 an unsubstituted carboxamide group, C3 a keto–enol group with a tautomerizable C2 proton, and C4 a dimethylamino group naturally a below the plane of the ring system. At the C12a ring juncture, a tertiary hydroxyl group is located a to the ring plane and is one of the most crucial functional groups needed for maintaining its 3D structure and antibacterial activity. Semisynthetic removal of the C12a hydroxyl group or introduction of sterically hindered substi