Chemical synthetic biology 9780470713976, 0470713976

Citation preview

Cell Culture Engineering

Related Titles Luisi, P.P. (ed.)

Yoshida, T. (ed.)

Chemical Synthetic Biology

Applied Bioengineering

2011 Print ISBN: 978-0-470-71397-6

2017 Print ISBN: 978-3-527-34075-0

Further Volumes of the “Advanced Biotechnology” Series:

Nielsen, J., Hohmann, S. (eds.)

Systems Biology

Published:

2017 Print ISBN: 978-3-527-33558-9

Villadsen, J. (ed.)

Smolke, C.

Fundamental Bioengineering

Synthetic Biology

2016 Print ISBN: 978-3-527-33674-6

Parts, Devices, and Applications

Love, J. Ch. (ed.)

Micro- and Nanosystems for Biotechnology 2016 Print ISBN: 978-3-527-33281-6

2018 Print ISBN: 978-3-527-33075-1

Chang, H. N

Emerging Areas in Bioengineering 2018 Print ISBN: 978-3-527-34088-0

Wittmann, Ch., Liao, J.C. (eds.)

Industrial Biotechnology Microorganisms (2 Volumes) 2017 Print ISBN: 978-3-527-34179-5

Planned: Zhao, H.

Protein Engineering Tools and Applications

Wittmann, Ch., Liao, J.C. (eds.)

Industrial Biotechnology Products and Processes 2017 Print ISBN: 978-3-527-34181-8

2020 Print ISBN: 978-3-527-34470-3

Cell Culture Engineering Recombinant Protein Production

Edited by Gyun Min Lee Helene Faustrup Kildegaard

Volume Editors

KAIST Department of Biological Sciences 373-1, Kusong-Dong, Yusong-Gu 305-701 Daejon South Korea

All books published by Wiley-VCH are carefully produced. Nevertheless, authors, editors, and publisher do not warrant the information contained in these books, including this book, to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate.

Dr. Helene Faustrup Kildegaard

Library of Congress Card No.:

Technical University of Denmark The Novo Nordisk Foundation Center for Biosustainability Kgs. Lyngby, 2800 Denmark

applied for

Professor Gyun Min Lee

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

Series Editors Sang Yup Lee

Bibliographic information published by the Deutsche Nationalbibliothek

KAIST 373-1; Guseong-Dong 291 Daehak-ro,Yuseong-gu 305-701 Daejon South Korea

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at .

Prof. Dr. Jens Nielsen

© 2020 Wiley-VCH Verlag GmbH & Co. KGaA, Boschstr. 12, 69469 Weinheim, Germany

Chalmers University Department of Chemical and Biol.Eng Kemivägen 10 Department of Chemical and Biol.Eng 412 96 Göteborg Sweden Gregory Stephanopoulos

Massachus. Inst. of Technology Department of Chemical Engineering Massachusetts Ave 77 United States Cover: Culture Flasks in microbiological

laboratory / science photo, fotolia

All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Print ISBN: 978-3-527-34334-8 ePDF ISBN: 978-3-527-81138-0 ePub ISBN: 978-3-527-81140-3 oBook ISBN: 978-3-527-81141-0 Cover Design Adam-Design, Weinheim,

Germany Typesetting SPi Global, Chennai, India Printing and Binding

Printed on acid-free paper 10 9 8 7 6 5 4 3 2 1

v

Contents About the Series Editors xvii

1

1

Platform Technology for Therapeutic Protein Production Tae Kwang Ha, Jae Seong Lee, and Gyun Min Lee

1.1 1.2 1.2.1 1.2.2 1.3 1.3.1 1.3.2 1.3.3 1.3.4 1.3.4.1 1.3.4.2 1.4 1.4.1 1.4.2 1.4.3 1.4.4 1.4.5 1.5 1.5.1 1.5.2 1.6 1.6.1 1.6.2 1.6.3

Introduction 1 Overall Trend Analysis 3 Mammalian Cell Lines 3 Brief Introduction of Advances and Techniques 5 General Guidelines for Recombinant Cell Line Development 6 Host Selection 6 Expression Vector 7 Transfection/Selection 7 Clone Selection 8 Primary Parameters During Clone Selection 8 Clone Screening Technologies 9 Process Development 9 Media Development 10 Culture Environment 10 Culture Mode (Operation) 10 Scale-up and Single-Use Bioreactor 11 Quality Analysis 12 Downstream Process Development 12 Purification 12 Quality by Design (QbD) 13 Trends in Platform Technology Development 14 Rational Strategies for Cell Line and Process Development 14 Hybrid Culture Mode and Continuous System 15 Recombinant Human Cell Line Development for Therapeutic Protein Production 16 Conclusion 17 Acknowledgment 17 Conflict of Interest 17 References 17

1.7

vi

Contents

23

2

Cell Line Development for Therapeutic Protein Production Soo Min Noh, Seunghyeon Shin, and Gyun Min Lee

2.1 2.2 2.2.1 2.2.2 2.2.3 2.3 2.3.1 2.3.2

Introduction 23 Mammalian Host Cell Lines for Therapeutic Protein Production 25 CHO Cell Lines 25 Human Cell Lines 26 Other Mammalian Cell Lines 27 Development of Recombinant CHO Cell Lines 27 Expression Systems for CHO Cells 28 Cell Line Development Process Using CHO Cells Based on Random Integration 28 Vector Construction 29 Transfection and Selection 30 Gene Amplification 30 Clone Selection 31 Cell Line Development Process Using CHO Cells Based On Site-Specific Integration 32 Development of Recombinant Human Cell Lines 34 Necessity for Human Cell Lines 34 Stable Cell Line Development Process Using Human Cell Lines 35 Important Consideration for Cell Line Development 36 Clonality 36 Stability 36 Quality of Therapeutic Proteins 37 Conclusion 38 References 38

2.3.2.1 2.3.2.2 2.3.2.3 2.3.2.4 2.3.3 2.4 2.4.1 2.4.2 2.5 2.5.1 2.5.2 2.5.3 2.6

3

Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells 49 Joo-Hyoung Lee, Henning G. Hansen, Sun-Hye Park, Jong-Ho Park, and Yeon-Gu Kim

3.1 3.2 3.2.1 3.2.2 3.2.3 3.2.4 3.3 3.3.1 3.3.2 3.3.3 3.4 3.4.1 3.4.2 3.4.3 3.5 3.5.1

Introduction 49 Gene Delivery: Transient Transfection Methods 50 Calcium Phosphate-Based Transient Transfection 50 Electroporation 51 Polyethylenimine-Based Transient Transfection 52 Liposome-Based Transient Transfection 52 Expression Vectors 53 Expression Vector Composition and Preparation 53 Episomal Replication 53 Coexpression Strategies 54 Mammalian Cell Lines 54 HEK293 Cell-Based TGE Platforms 55 CHO Cell-Based TGE Platforms 56 TGE Platforms Using Other Cell Lines 58 Cell Culture Strategies 58 Culture Media for TGE 58

Contents

3.5.2 3.5.3 3.5.4 3.6 3.7

Optimization of Cell Culture Processes for TGE 59 qp -Enhancing Factors in TGE-Based Culture Processes 59 Culture Longevity-Enhancing Factors in TGE-Based Culture Processes 59 Large-Scale TGE-Based Protein Production 60 Concluding Remarks 62 References 62

4

Enhancing Product and Bioprocess Attributes Using Genome-Scale Models of CHO Metabolism 73 Shangzhong Li, Anne Richelle, and Nathan E. Lewis

4.1 4.1.1 4.1.2 4.1.2.1 4.1.2.2

Introduction 73 Cell Line Optimization 73 CHO Genome 75 Development of Genomic Resources of CHO 75 Development of Transcriptomics and Proteomics Resources of CHO 75 Genome-Scale Metabolic Model 76 What Is a Genome-Scale Metabolic Model 76 Reconstruction of GEMs 77 Knowledge-Based Construction 77 Draft Reconstruction 77 Curation of the Reconstruction 77 Conversion to a Computational Format 79 Model Validation and Evaluation 79 GEM Application 80 Common Usage and Prediction Capacities of Genome-Scale Models 82 GEMs as a Platform for Omics Data Integration, Linking Genotype to Phenotype 83 Predicting Nutrient Consumption and Controlling Phenotype 84 Enhancing Protein Production and Bioprocesses 85 Case Studies 86 Conclusion 86 Acknowledgments 88 References 88

4.2 4.2.1 4.2.2 4.2.2.1 4.2.2.2 4.2.2.3 4.2.2.4 4.2.2.5 4.3 4.3.1 4.3.2 4.3.3 4.3.4 4.3.5 4.4

5

Genome Variation, the Epigenome and Cellular Phenotypes 97 Martina Baumann, Gerald Klanert, Sabine Vcelar, Marcus Weinguny, Nicolas Marx, and Nicole Borth

5.1

Phenotypic Instability in the Context of Mammalian Production Cell Lines 97 Genomic Instability 99 Epigenetics 101 DNA Methylation 102

5.2 5.3 5.3.1

vii

viii

Contents

5.3.2 5.3.3 5.3.4 5.4 5.5 5.5.1 5.5.1.1 5.5.1.2 5.5.2 5.5.2.1 5.5.2.2 5.6

Histone Modifications 102 Downstream Effectors 104 Noncoding RNAs 104 Control of CHO Cell Phenotype by the Epigenome 105 Manipulating the Epigenome 107 Global Epigenetic Modification 107 Manipulating Global DNA Methylation 107 Manipulating Global Histone Acetylation 108 Targeted Epigenetic Modification 109 Targeted Histone Modification 110 Targeted DNA Methylation 112 Conclusion and Outlook 113 References 114

6

Adaption of Generic Metabolic Models to Specific Cell Lines for Improved Modeling of Biopharmaceutical Production and Prediction of Processes 127 Calmels Cyrielle, Chintan Joshi, Nathan E. Lewis, Malphettes Laetitia, and Mikael R. Andersen

6.1 6.1.1 6.1.2 6.1.2.1 6.1.2.2 6.1.2.3 6.1.2.4 6.1.2.5 6.1.2.6 6.2

Introduction 127 Constraint-Based Models 127 Limitations of Flux Balance Analysis 131 Thermodynamically Infeasible Cycles 131 Genetic Regulation 131 Limitation of Intracellular Space 132 Multiple States in the Solution 132 Biological Objective Function 133 Kinetics and Metabolite Concentrations 133 Main Source of Optimization Issues with Large Genome-Scale Models: Thermodynamically Infeasible Cycles 134 Definition of Thermodynamically Infeasible Fluxes 134 Loops Involving External Exchange Reactions 134 Reversible Passive Transporters from Major Facilitator Superfamily (MFS) 135 Reversible Passive Antiporters from Amino Acid-Polyamine-organoCation (APC) Superfamily 136 Na+ -linked Transporters 136 Transport via Proton Symport 137 Tools to Identify Thermodynamically Infeasible Cycles 138 Visualizing Fluxes on a Network Map 138 Algorithms Developed 138 Methods Available to Remove Thermodynamically Infeasible Cycles 139 Manual Curation 139 Software and Algorithms Developed for the Removal of Thermodynamically Infeasible Loops from Flux Distributions 140 Consideration of Additional Biological Cellular Constraints 144

6.2.1 6.2.2 6.2.2.1 6.2.2.2 6.2.2.3 6.2.2.4 6.2.3 6.2.3.1 6.2.3.2 6.2.4 6.2.4.1 6.2.4.2 6.3

Contents

6.3.1 6.3.1.1 6.3.1.2 6.3.2 6.3.2.1 6.3.2.2 6.3.2.3 6.3.2.4 6.3.3 6.3.3.1 6.3.3.2 6.4

Genetic Regulation 144 Advantages of Considering Gene Regulation in Genome-Scale Modeling 144 Methods Developed to Take into Account a Feedback of FBA on the Regulatory Network 145 Context Specificity 146 What Are Context-Specific Models (CSMs)? 146 Methods and Algorithms Developed to Reconstruct Context-Specific Models (CSMs) 146 Performance of CSMs 148 Cautions About CSMs 149 Molecular Crowding 150 Consequences on the Predictions 150 Methods Developed to Account for a Total Enzymatic Capacity into the FBA Framework 151 Conclusion 152 References 153

7

Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing 163 Kok Siong Ang, Jongkwang Hong, Meiyappan Lakshmanan, and Dong-Yup Lee

7.1 7.2 7.2.1 7.2.1.1 7.2.1.2 7.2.1.3 7.2.1.4 7.2.2 7.2.2.1 7.2.2.2 7.2.2.3 7.2.2.4 7.3 7.3.1 7.3.2 7.3.3 7.4

Introduction 163 High-Throughput Omics Technologies 165 Sequencing-Based Omics Technologies 165 Historical Developments of Nucleotide Sequencing Techniques 165 Genome Sequencing of CHO Cells 166 Transcriptomics of CHO Cells 167 Epigenomics of CHO Cells 168 Mass Spectrometry-Based Omics Technologies 168 Mass Spectrometry Techniques 168 Proteomics of CHO Cells 170 Metabolomics/Lipidomics of CHO Cells 171 Glycomics of CHO Cells 172 Current CHO Multi-omics Applications 172 Bioprocess Optimization 174 Cell Line Characterization 174 Engineering Target Identification 176 Future Prospects 177 References 178

8

CRISPR Toolbox for Mammalian Cell Engineering 185 Daria Sergeeva, Karen Julie la Cour Karottki, Jae Seong Lee, and Helene Faustrup Kildegaard

8.1 8.2 8.3 8.3.1

Introduction 185 Mechanism of CRISPR/Cas9 Genome Editing 186 Variants of CRISPR-RNA-guided Endonucleases 187 Diversity of CRISPR/Cas Systems 187

ix

x

Contents

8.3.2 8.4 8.4.1 8.4.2 8.5 8.5.1 8.5.1.1 8.5.1.2 8.5.2 8.5.2.1 8.5.2.2 8.5.3 8.6 8.7 8.8

Engineered Cas9 Variants 188 Experimental Design for CRISPR-mediated Genome Editing 188 Target Site Selection and Design of gRNAs 189 Delivery of CRISPR/Cas9 Components 191 Development of CRISPR/Cas9 Tools 192 CRISPR/Cas9-mediated Gene Editing 192 Gene Knockout 192 Site-Specific Gene Integration 194 CRISPR/Cas9-mediated Genome Modification 195 Transcriptional Regulation 195 Epigenetic Modification 196 RNA Targeting 196 Genome-Scale CRISPR Screening 197 Applications of CRISPR/Cas9 for CHO Cell Engineering 197 Conclusion 199 Acknowledgment 200 References 200

9

CHO Cell Engineering for Improved Process Performance and Product Quality 207 Simon Fischer and Kerstin Otte

9.1 9.2 9.2.1 9.2.2 9.2.3 9.3 9.3.1 9.3.2 9.3.3

CHO Cell Engineering 207 Methods in Cell Line Engineering 208 Overexpression of Engineering Genes 208 Gene Knockout 209 Noncoding RNA-mediated Gene Silencing 209 Applications of Cell Line Engineering Approaches in CHO Cells 211 Enhancing Recombinant Protein Production 211 Repression of Cell Death and Acceleration of Growth 221 Modulation of Posttranslational Modifications to Improve Protein Quality 227 Conclusions 233 References 234

9.4

10

Metabolite Profiling of Mammalian Cells 251 Claire E. Gaffney, Alan J. Dickson, and Mark Elvin

10.1

Value of Metabolic Data for the Enhancement of Recombinant Protein Production 251 Technologies Used in the Generation of Metabolic Data Sets 252 Targeted and Untargeted Metabolic Analysis 253 Analytical Technologies Used in the Generation of Metabolite Profiles 253 Nuclear Magnetic Resonance 254 Mass Spectrometry 255 Metabolite Sample Preparation 256 Extracellular Sample Preparation 257 Quenching of Intracellular Metabolite Samples 257

10.2 10.2.1 10.2.2 10.2.2.1 10.2.2.2 10.2.3 10.2.3.1 10.2.3.2

Contents

10.2.3.3 10.2.3.4 10.3 10.3.1 10.3.2 10.3.3 10.4 10.4.1 10.4.2 10.4.3 10.4.4 10.4.5 10.5

Metabolite Extraction from Quenched Cells 257 Metabolic Flux Analysis 257 Approaches for Metabolic Data Analysis 257 Data Processing 258 Data Analysis 258 Data Interpretation and Integration 260 Implementation of Metabolic Data in Bioprocessing 261 Relationship Between Growth Phase and Metabolism 261 Identification of Metabolic Indicators Associated with High Cell-Specific Productivity 263 Utilizing Metabolic Data to Improve Biomass and Recombinant Protein Yield 263 Utilizing Metabolic Understanding to Improve Product Quality 265 Cell Line Engineering to Redirect Metabolic Pathways 265 Future Perspectives 266 Acknowledgments 267 References 267

11

Current Considerations and Future Advances in Chemically Defined Medium Development for the Production of Protein Therapeutics in CHO Cells 279 Wai Lam W. Ling

11.1 11.2 11.2.1 11.2.2 11.2.3 11.2.4 11.3 11.3.1

Introduction 279 Traditional Approach to Medium Development 279 Cell Line Selection 279 Design and Optimization 280 Process Consideration 282 Additional Considerations in Medium Development 284 Future Perspectives for Medium Development 284 Systems Biology and Synthetic Biology 284 Acknowledgment 288 Conflict of Interest 288 References 288

12

Host Cell Proteins During Biomanufacturing 295 Jong Youn Baik, Jing Guo, and Kelvin H. Lee

12.1 12.2 12.2.1 12.2.2 12.2.3 12.3 12.3.1 12.3.2 12.3.3 12.4

Introduction 295 Removal of HCP Impurities 295 Antibody Product 296 Non-antibody Protein Product 297 Difficult-to-Remove HCPs 298 Impacts of Residual HCPs 298 Drug Efficacy, Quality, and Shelf Life 298 Immunogenicity 299 Biological Activity 299 HCP Detection and Monitoring Methods 300

xi

xii

Contents

12.4.1 12.4.2 12.5 12.5.1 12.5.2 12.5.3 12.6

Anti-HCP Antiserum and Enzyme-Linked Immunosorbent Assay (ELISA) 300 Proteomics Approaches as Orthogonal Methods 302 Efforts for HCP Control 302 Upstream Efforts 303 Downstream Efforts 304 HCP Risk Assessment in CHO Cells 305 Future Directions 305 Acknowledgments 306 References 306

13

Mammalian Fed-batch Cell Culture for Biopharmaceuticals 313 William C. Yang

13.1 13.2 13.2.1 13.2.2 13.2.3 13.2.4 13.3 13.3.1 13.3.2 13.3.3 13.4 13.5 13.5.1 13.5.2 13.6 13.6.1 13.6.2 13.7 13.8 13.8.1 13.8.2 13.8.3 13.8.4 13.9 13.9.1 13.9.2 13.10 13.10.1 13.10.2 13.11

Introduction 313 Objectives of Cell Culture Process Development 314 Yield and Product Quality 314 Glycosylation 314 Charge Heterogeneity 315 Aggregation 316 Cells and Cell Culture Formats 316 Adherent Cells 316 Suspended Cells 316 Batch Cultures 317 Fed-batch Cultures 317 Cell Culture Media 319 Basal Media 319 Feed Media 320 Feeding Strategies 321 Metabolite Based 321 Respiration Based 323 Feed Media Design 323 Process Variable Design 325 Temperature 325 pH and pCO2 325 Dissolved Oxygen 326 Culture Duration 327 Cell Culture Supplements 327 Yield 328 Glycosylation 328 New and Emerging Technologies 329 Analytical Technologies 329 Bioreactor Technologies 331 Future Directions 332 References 333

Contents

14

Continuous Biomanufacturing 347 Sadettin S. Ozturk

14.1 14.2 14.2.1 14.2.2 14.2.2.1 14.2.2.2 14.2.3 14.3 14.3.1 14.3.2 14.3.3 14.3.4 14.4 14.4.1 14.4.1.1 14.4.1.2 14.4.1.3 14.4.1.4 14.4.1.5 14.4.1.6 14.5 14.5.1 14.5.2 14.5.3 14.5.3.1 14.5.3.2 14.5.3.3 14.6 14.7

Introduction 347 Continuous Upstream (Cell Culture) Processes 347 Continuous Culture without Cell Retention (Chemostat) 348 Continuous Culture with Cell Retention (Perfusion) 348 Cell Retention by Immobilization or Entrapment 349 Cell Retention by Cell Retention Device 350 Semicontinuous Culture 351 Advantages of Continuous Perfusion 351 Higher Volumetric Productivities 351 Better Utilization of Biomanufacturing Facilities 352 Better Product Quality and Consistency 352 Scale-up and Commercial Production 353 Cell Retention Systems for Continuous Perfusion 354 Cell Retention Devices 354 Filtration-Based Devices 354 Spin Filters 355 Continuous Centrifugation 356 Settler 356 BioSep Device 357 Hydrocyclones 358 Operation and Control of Continuous Perfusion Bioreactors 358 Feed and Harvest Flow and Volume Control 358 Circulation or Return Pump 359 Control of Perfusion Rate and Cell Density 359 Cell Build-up Phase 359 Production Phase 360 Cell Bleed or Purge 360 Current Status of Continuous Perfusion 360 Conclusions 362 Acknowledgment 362 References 363

15

Process Analytical Technology and Quality by Design for Animal Cell Culture 365 Hae-Woo Lee, Hemlata Bhatia, Seo-Young Park, Mark-Henry Kamga, Thomas Reimonn, Sha Sha, Zhuangrong Huang, Shaun Galbraith, Huolong Liu, and Seongkyu Yoon

15.1 15.2 15.3 15.3.1 15.3.2 15.3.3

PAT and QbD – US FDA’s Regulatory Initiatives 365 PAT and QbD – Challenges 365 PAT and QbD Implementations 366 NIR Spectroscopy 366 Mid-Infrared (MIR) Spectroscopy 367 Raman Spectroscopy 367

xiii

xiv

Contents

15.3.4 15.3.5 15.3.6 15.3.7 15.4 15.4.1

15.4.2 15.4.3 15.4.4 15.4.5 15.4.6 15.4.7

15.4.8

15.5

Fluorescence Spectroscopy 368 Chromatographic Techniques 368 Other Useful Techniques 369 Data Analysis and Modeling Tools 369 Case Studies 370 Estimation of Raw Material Performance in Mammalian Cell Culture Using Near-Infrared Spectra Combined with Chemometrics Approaches 370 Design Space Exploration for Control of Critical Quality Attributes of mAb 372 Quantification of Protein Mixture in Chromatographic Separation Using Multiwavelength UV Spectra 372 Characterization of Mammalian Cell Culture Raw Materials by Combining Spectroscopy and Chemometrics 374 Effect of Amino Acid Supplementation on Titer and Glycosylation Distribution in Hybridoma Cell Cultures 375 Metabolic Responses and Pathway Changes of Mammalian Cells Under Different Culture Conditions with Media Supplementations 377 Estimation and Control of N-Linked Glycoform Profiles of Monoclonal Antibody with Extracellular Metabolites and Two-Step Intracellular Models 378 Quantitative Intracellular Flux Modeling and Applications in Biotherapeutic Development and Production Using CHO Cell Cultures 381 Conclusion 383 References 383

16

Development and Qualification of a Cell Culture Scale-Down Model 391 Sarwat Khattak and Valerie Pferdeort

16.1 16.1.1 16.2 16.2.1 16.2.2 16.2.3 16.2.4 16.2.5 16.3 16.3.1 16.3.2 16.3.2.1 16.3.2.2 16.3.2.3 16.3.2.4 16.3.2.5 16.3.3

Purpose of the Scale-Down Model 391 Development Challenges 391 Types of Scale-Down Models 392 Power/Volume (P/V ) and Air velocity 392 Oxygen Transfer Coefficient (k L a) 392 Gas Entrance Velocity (GEV) 393 Oxygen Transfer Rate (OTR) 393 Model Refinement Workflow 395 Evaluation of a Scale-Down Model 395 Univariate Analysis 395 Multivariate Analysis 396 Statistical Background 396 Qualification Data Set 396 Observation Level Analysis 397 Batch-Level Analysis 397 Scores Contribution Plots 398 Equivalence Testing 399

Contents

16.3.3.1 16.3.3.2 16.3.3.3 16.4

Statistical Background 399 Considerations for Evaluation and Test Data Sets 399 Types of Analysis Outcomes 400 Conclusions and Perspectives 401 References 402 Index 407

xv

xvii

About the Series Editors Sang Yup Lee is a distinguished Professor at the Department of Chemical and Biomolecular Engineering at the Korea Advanced Institute of Science and Technology. At present, Prof. Lee is the Director of the Center for Systems and Synthetic Biotechnology, Director of the BioProcess Engineering Research Center, and Director of the Bioinformatics Research Center. He has published more than 500 journal papers, 64 books, and book chapters and has more than 580 patents (either registered or applied) to his credit. He has received numerous awards, including the National Order of Merit, the Merck Metabolic Engineering Award, the ACS Marvin Johnson Award, Charles Thom Award, Amgen Biochemical Engineering Award, Elmer Gaden Award, POSCO TJ Park Prize, and HoAm Prize. He is a fellow of American Association for the Advancement of Science, the American Academy of Microbiology, American Institute of Chemical Engineers, Society for Industrial Microbiology and Biotechnology, American Institute of Medical and Biological Engineering, the World Academy of Science, the Korean Academy of Science and Technology, and the National Academy of Engineering of Korea. He is also a foreign member of National Academy of Engineering, USA. In addition, he is an honorary professor of the University of Queensland (Australia), the Chinese Academy of Sciences, Wuhan University (China), Hubei University of Technology (China), and Beijing University of Chemical Technology (China) and an advisory professor of the Shanghai Jiaotong University (China). Apart from his academic associations, Prof. Lee is the editor-in-chief of the Biotechnology Journal and is also contributing to numerous other journals as associate editor and board member. Prof. Lee is serving as a member of Presidential Advisory Committee on Science and Technology (South Korea).

xviii

About the Series Editors

Jens Nielsen is a professor and director of Chalmers University of Technology (Sweden) since 2008. He obtained an MSc degree in Chemical Engineering and a PhD degree (1989) in Biochemical Engineering from the Technical University of Denmark (DTU), and after that, established his independent research group and was appointed a full-time professor there in 1998. He was a Fulbright visiting professor at MIT in 1995–1996. At DTU, he founded and directed the Center for Microbial Biotechnology. Prof. Nielsen has published more than 350 research papers and coauthored more than 40 books, and he is the inventor of more than 50 patents. He has founded several companies that have raised more than 20 million in venture capital. He has received numerous Danish and international awards and is a member of the Academy of Technical Sciences (Denmark), the National Academy of Engineering (USA), the Royal Danish Academy of Science and Letters, the American Institute for Medical and Biological Engineering, and the Royal Swedish Academy of Engineering Sciences. Gregory Stephanopoulos is the W.H. Dow Professor of Chemical Engineering at the Massachusetts Institute of Technology (MIT, USA) and Director of the MIT Metabolic Engineering Laboratory. He is also an instructor of bioengineering at Harvard Medical School (since 1997). He received his BS degree from the National Technical University of Athens and PhD from the University of Minnesota (USA). He has coauthored about 400 research papers and 50 patents, along with the first textbook on metabolic engineering. He has been recognized by numerous awards from the American Institute of Chemical Engineers (AIChE) (Wilhelm, Walker and Founders awards), American Chemical Society (ACS), Society of Industrial Microbiology (SIM), BIO (Washington Carver Award), the John Fritz Medal of the American Association of Engineering Societies, and others. In 2003, he was elected as a member of the National Academy of Engineering (USA) and in 2014 president of AIChE.

1

1 Platform Technology for Therapeutic Protein Production Tae Kwang Ha 1,∗ , Jae Seong Lee 1,2,∗ , and Gyun Min Lee 1,3 1 Technical University of Denmark, The Novo Nordisk Foundation Center for Biosustainability, Kemitorvet, 2800 Kgs. Lyngby, Denmark 2 Ajou University, Department of Molecular Science and Technology, 206 Worldcup-ro, Yeongtong-gu, 16499 Suwon, Republic of Korea 3 KAIST, Department of Biological Sciences, 291 Daehak-ro, Yuseong-gu, Daejeon 305-701, Republic of Korea

1.1 Introduction

®

In 1987, the human tissue plasminogen activator (trade name: Activase ) was the first therapeutic protein produced in Chinese hamster ovary (CHO) cells to receive US Food and Drug Administration (FDA) approval, which triggered the emergence of mammalian cell culture for production of biopharmaceuticals [1]. Therapeutic proteins are effective drugs for many diseases including diabetes, rheumatoid arthritis, clotting disorders, and cancers because of their highly specific functions with reduced side effects and no immune response [2, 3]. With the increasing number of therapeutic proteins, the biopharmaceutical market has expanded dramatically over the past few decades. The global market value of therapeutic proteins reached $140 billions in 2014, and AbbVie’s Humira (adalimumab), one of the profitable drugs in the biopharmaceutical industry, generated worldwide sales of $13.9 billions in 2015 [4, 5]. From 2011 to 2015, 40 novel therapeutic proteins were approved by the FDA, and nearly 70% of therapeutic proteins are produced in mammalian cells, particularly CHO cells, because of their capability for humanlike post-translation modification (PTM) including glycosylation and protein folding [6]. Notably, 7 out of 10 top-selling blockbuster therapeutic proteins were produced in mammalian cells in 2015 (Table 1.1), and this trend of the prominence of mammalian manufacturing platforms over microbial manufacturing platforms will continue with the steady increase in the proportion of complex molecules in the pipeline at both the qualitative and quantitative levels [5]. Therapeutic protein production, however, requires time-consuming and complicated processes. In a mammalian manufacturing platform of therapeutic proteins that includes the cloning of a target gene into an appropriate expression vector, the selection of a suitable host cell line for the target product, and final

®

*

Tae Kwang Ha and Jae Seong Lee are contributed equally to this work.

Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

Table 1.1 The 10 top-selling therapeutic proteins in 2015. Approved/ expiry (EU, US)

2015 Sales ($ millions)

Host

2003, 2002/ 2018, 2016

13 932.00

CHO

Amgen, Pfizer, Takeda Pharmaceuticals

2000, 1998/2015, 2028

9008.70

CHO

Non-Hodgkin’s lymphoma

Biogen-IDEC, Roche

1998, 1997/ 2013, 2016

7395.00

CHO

rh insulin analog

Diabetes mellitus

Sanofi

2000, 2000/ 2014, 2014

7095.40

E. coli

Humanized mAb

Anti-VEGF

Metastatic colorectal cancer, glioblastoma, metastatic renal carcinoma

Roche/ Genentech

2005, 2004/ 2019, 2017

7014.20

CHO

Herceptin (trastuzumab)

humanized mAb

Anit-HER2

Breast cancer, gastric cancer

Roche/ Genentech

2000, 1998/ 2014, 2019

6862.60

CHO

7

Remicade (infliximab)

Chimeric mAb

Anti-TNF

Crohn’s disease

J&J, Merck & Mitsubishi Tanabe Pharma

1999, 1998/ 2015, 2018

6826.10

Sp2/0

8

Neulasta (pegfilgrastim)

Peptide

PEGylated rh G-CSF

Chemotherapyinduced neutropenia

Amgen

2002, 2002/ 2015, 2014

4715.10b)

E. coli

9

Eylea (aflibercept)

Fusion protein (receptor – IgG Fc)

Anti-VEGF

Neovascular (wet) age-related macular degeneration

Regeneron, Bayer

2012, 2011/ 2020, 2021

4089.00b)

CHO

10

Lucentis (ranibizumab)

Humanized IgG fragment

Anti-VEGF

Neovascular (wet) age-related macular degeneration

Roche/ Genentech, Novartis

2007, 2006/ 2016, 2016

3580.00b)

E. coli

Product (active ingredient)

Product category

Targeta)

Therapeutic indication

Company

1

Humira (adalimumab)

Human mAb

Anti-TNF

Rheumatoid arthritis

AbbVie & Eisai

2

Enbrel (etanercept)

Fusion protein (receptor – IgG fragment)

Anti-TNF

Rheumatoid arthritis

3

Rituxan/ Mabthera (rituximab)

Chimeric mAb

Anti-CD20

4

Lantus (insulin glargine)

Peptide

5

Avastin (bevacizumab)

6

Ranking

a) In the case of peptide products, other general names of products, not generic and trade names, are described. b) Full-year 2015 financial reports of Amgen, Regeneron, Bayer, Roche/Genentech, and Novartis. Source: Adapted from Morrison 2016 [4] and Walsh 2014 [5].

1.2 Overall Trend Analysis

processing for commercialization, many resources are required to ensure quality control at every step [6]. Furthermore, the mammalian cell culture that involves CHO cells is considered to be difficult because of low yield, complexity, price of media, and obstacles to optimization of culture conditions. Traditionally, various parameters in the production processes have had to be independently optimized for each target product because of clonal variability and product dependency. The effect of each parameter, such as the type of the host cell line, expression vector design, screening and selection methods, media composition, feed media, and culture conditions, including temperature, pH, and agitation speed, on protein productivity and product quality is highly dependent on the specific cell lines [7, 8]. Along with the technical advances in the upstream process development, specific productivity of over 20 pg/cell/day and product titer of over 10 g/l have been reached in many cases in the biopharmaceutical industry [8, 9]. The improvement of specific productivity and final yield has been achieved not only through expression vector and clone selection methods but also through the enhancement of commercial culture media and optimization of operational conditions. Today, the focus in mammalian cell culture process development has changed from higher productivity to proper and consistent quality with higher productivity at all developmental stages and at large scales [10]. In the following sections, we provide a general overview of platform technology for therapeutic protein production that has been commonly used in mammalian cell culture. Because of the complexity and diversity of the field, there is limited room to cover all the details in this chapter. Rather, we include references for more detailed information, and we devote special attention to general guidelines and considerations for bioprocess development. Then, we introduce the trends in platform technology development that are applied recently in this field (Figure 1.1).

1.2 Overall Trend Analysis 1.2.1

Mammalian Cell Lines

Recombinant therapeutic proteins are mainly produced in mammalian host cell lines, including NS0 murine myeloma, CHO, and human embryonic kidney (HEK) 293 cells. Humans and other mammals share a closer evolutionary lineage compared to microorganisms such as Escherichia coli (E. coli), which means that mammalian cells are suitable for the generation of complex and highly valuable humanlike proteins [11, 12]. Murine NS0 cells were initially used in the production of therapeutic antibodies in the biopharmaceutical industry. NS0 cells lack endogenous glutamine synthetase (GS) enzyme activity, which makes them suitable for the use of the GS/methionine sulfoximine (MSX) amplification system. Although high

3

4

1 Platform Technology for Therapeutic Protein Production

Upstream process

Cell line development Host selection

Specific productivity Culture longevity Product quality

CHOK1, DG44, DXB11, etc. Growth rate, viable cell density, specific productivity, ER/mitochondrial capacity

Transfection/ selection

Expression vector

Transfection method: Calcium phosphate, electroporation, lipofection, retroviral transfection Antibiotics selection GS/MSX, DHFR/MTX amplification

Promoter: CMV, EF1α, SV40, etc. IRES, 2A peptide Selection marker Epigenetic elements: S/MARs, UCOEs, STAR, etc .

Clone selection

Parameters for selection Product titer/specific productivity Integral of viable cell density Product quality Cell line robustness

Screening technologies FACS Semi-solid matrices based system High throughput and automation

Process development Media Essential nutrients Nutrient metabolism Composition and concentration Feed media optimization Feeding strategy

Culture environment and scale-up

Culture mode

Temperature, pH Dissolved O2 and CO2 Osmolality Agitation speed Gas flow Bioreactor type (SUB)

Batch culture Fed batch culture Perfusion culture Concentrated fed batch culture Hybrid fed batch culture

Quality analysis Glycosylation Aggregation Disulfide bond formation …

Downstream process Yield and productivity Purity and capacity Speed

Purification Chromatography

Protein-A Cataion exchange chromatography Mimetic resin of Protein-A Continuous method

QbD

Non chromatography

Considerations

Membrane based system Phase partitioning

Cost effective Higher flow rate Shorter residence time Longer life cycle Larger capacity and high througput screening Size, shape, charge of target protein Sensitivity, components

Scientific, systematic, and comprehensive approach for drug development Consistent and intended product quality Integrative and continuous processing Small-scale and parallel facilities

Figure 1.1 Optimization parameters in upstream and downstream process.

1.2 Overall Trend Analysis

antibody productivity has been achieved in GS-NS0 cells, N-glycolylneuraminic acid-bound proteins produced from NS0 cells led to an immunogenicity concern in humans. Therefore, NS0 cells have limited use in therapeutic protein production today [8, 13]. Human cell lines including HEK293 have the ability to produce proteins mostly like natural human products, which is their main advantage over other expression systems. Recently, several therapeutic proteins produced from HEK293 cells have been approved by the FDA or the European Medicines Agency (EMA). A major concern with the use of human cell lines is low productivity and the risk of viral infection [14]. For these reasons, CHO cells are the most predominantly used mammalian host cell lines in the production of various therapeutic proteins, including monoclonal antibodies (mAbs), cytokines, and fusion proteins. Nearly 70% of all recombinant therapeutic proteins produced today are made in CHO cells because of several key advantages over other host cells, such as safety regarding human pathogenic viruses, ease of growth in a large-scale suspension culture, and the ability to express humanlike proteins along with humanlike PTMs. Furthermore, CHO cells have strong gene amplification systems such as dihydrofolate reductase (DHFR)/methotrexate (MTX) and GS/MSX to improve protein production, and various genetic manipulation strategies have been developed to improve protein production and product quality [15, 16]. 1.2.2

Brief Introduction of Advances and Techniques

With the expansion of biopharmaceutical markets, the improvement of mammalian cell lines is a key challenge to meet the higher demand for therapeutic proteins. Because the biopharmaceutical industry pursues inexpensive and high-yield manufacturing processes to maximize production yields at low cost, several strategies have been developed and implemented in recombinant mammalian cell line generation and cell culture processes. Strategies to improve therapeutic protein production in mammalian cells can be divided into two major categories: (i) increasing cell mass and (ii) increasing specific productivity. Increasing cell mass through a fast growth rate, maximum viable cell density, and/or longer culture duration has been achieved by process and media optimizations and genetic manipulation of several pathways involved in proliferation, apoptosis, autophagy, and cellular metabolism [13, 15]. Improving specific productivity by genetic manipulation has also been successfully implemented with CHO cells, such as the engineering of secretion, chaperone, cell cycle, transcription, and translation-related genes [15]. Protein quality, including glycosylation, is also a critical factor to determine the efficacy and stability of therapeutic proteins, which has been improved by the optimization of culture conditions, media, and feeding strategies and genetic manipulation of genes related to protein folding and glycosylation [17]. In 2011, the CHO-K1 genome was first sequenced by Xu et al., enabling more detailed and accurate bioinformatics analyses [18]. Previously, researchers had to infer genome information from other published mammalian genomes such as mouse and human genomes, raising inaccuracy issues regarding the CHO

5

6

1 Platform Technology for Therapeutic Protein Production

genome sequence. Since 2011, draft genomes of not only Chinese hamster but also six CHO cell lines including CHO-K1, DG44, and CHO-S have been sequenced and published [19]. Additionally, transcription start sites, expressed gene profiles, miRNA profiles, and secreted and host cell proteomes were serially identified and published, which has facilitated better understanding of CHO cells and substantially supported research efforts in cellular engineering [19, 20].

1.3 General Guidelines for Recombinant Cell Line Development The production of therapeutic proteins can be achieved by either a transient or stable gene expression system in mammalian cells. Given the fact that stable gene expression remains the preferred choice for the large-scale production of therapeutic proteins in the biopharmaceutical industry, we emphasize stable cell line generation in this section. A typical process of recombinant cell line development (CLD) for high-level therapeutic protein production includes the introduction of the exogenous gene of interest (GOI) into host cell lines in the form of an expression vector and the selection of stable and high-producing clones. Selected high producers are further evaluated in downstream processes with regard to a sustainable high production level with proper product quality within the acceptable range. There have been significant advances in host cell selection and engineering and expression vector engineering toward increased productivity and robust clone selection, as described in detail in the following sections. 1.3.1

Host Selection

Within selected mammalian expression systems among various manufacturing platforms, genetic and phenotypic diversity exists. Chinese hamster ovary cells have a family of cell lines, referred to as K1, DG44, DXB11, CHO-Toronto, CHOpro3-, and CHO-S, with distinct genomic backgrounds and physiological diversity (reviewed in [21]). Genetically divergent host cell lines have shown phenotypic differences with regard to growth rate, viable cell density, specific production rate, and ER/mitochondrial capacity [22, 23]. To date, DHFR-deficient CHO host cell lines, CHO-DXB11 and CHO-DG44, and CHO-K1 cell lines have been preferred for the production of therapeutic proteins in the industry because of their well-established gene amplification systems, referred to as DHFR and GS systems, respectively. The individual host cell line itself exhibits phenotypic heterogeneity within the cell population, i.e. clonal variation, which can be derived from either the inherent genomic plasticity of immortalized mammalian cell lines or nongenetic functional diversity such as stochastic gene expression [24, 25]. The heritable nature of functional properties including the specific growth rate and surface glycan content emphasizes that it is possible to screen the host cell population for the isolation of clonal derivatives with desirable attributes for biomanufacturing [26].

1.3 General Guidelines for Recombinant Cell Line Development

1.3.2

Expression Vector

The GOI encoding model proteins in the expression vector is delivered into the host cells. Mammalian expression vectors typically contain separate gene expression cassettes – one for expression in mammalian cells and the other for plasmid replication in bacteria. Within the cassettes for mammalian gene expression, selectable marker gene(s) and target product gene(s) are driven by promoters/enhancers such as cytomegalovirus (CMV), elongation factor α (EF1α), or simian virus 40 (SV40) promoter and terminated by 3′ polyadenylation signal sequences such as SV40 or bovine/human growth hormone polyadenylation sequence [8]. Based on this basic vector configuration, expression vector engineering has led to increased productivity and stability of production cell lines through modulation of the transcriptional activity of either GOI or marker genes (reviewed in [2]). The coexpression of GOI and marker genes via internal ribosome entry site (IRES) elements or self-cleaving 2A peptides allows for the selection of producers devoid of false positive survivors without expressing GOI [2, 27]. Additionally, selection marker attenuation through the use of a weak promoter, deoptimization of the marker gene, or insertion of mRNA/protein destabilizing elements weakens the selection marker, resulting in high selection stringency and the selection of high producers [2, 28]. Some cis-acting DNA regulatory elements have added value in vector engineering because of augmented attention directed to the epigenetic regulation of GOI in CLD (reviewed in [29]). The inclusion of epigenetic elements including scaffold/matrix attachment regions (S/MARs), ubiquitously acting chromatin opening elements (UCOEs), and the stabilizing and antirepressor (STAR) element can not only promote gene expression by remodeling the chromatin landscape so that it is favorable to high transgene expression but also allow for stable expression in long-term cultures because of its antisilencing effect [2, 15, 29]. Recent advances in promoter engineering efforts include either modifying natural promoters, e.g. mutation of methylation-prone CpG sites or insertion of the methylation-resistant core CpG element for enhanced stability [30, 31], or constructing synthetic promoters through the bottom-up assembly of several sequence elements such as transcription factor regulatory elements (TFREs) to core promoters. This effort may drive the tailored control of recombinant gene transcription for the next generation of mammalian cell factories (reviewed in [32]). 1.3.3

Transfection/Selection

The choice of how to introduce vector DNA into mammalian cells, i.e. transfection, is usually determined by its efficiency and toxicity. In contrast with the transient gene expression, in which many factors, including cost effectiveness vs. transfection efficiency and the cytotoxicity of the transfection reagent, must be considered for efficient large-scale transfection, the stable gene expression system allows an easy choice of transfection methods, as it merely requires small-scale transfection in one shot followed by selection of the transfected population. Among several transfection methods, such as calcium phosphate, electroporation, lipofection, and retroviral transfection, nonviral gene transfer

7

8

1 Platform Technology for Therapeutic Protein Production

methods, such as lipofection and electroporation, are commonly used to generate stable cell lines. Selectable marker genes in the expression vector enable the selection of the transfected population and the subsequent subpopulation that harbors stably integrated GOI in the chromosome, owing to cell growth and survival advantages upon the expression of marker genes in the presence of selection pressure. A variety of metabolic and antibiotic selectable markers are used, where a single or double selection approach can be applied [8]. The metabolic selection system, including GS and DHFR systems coupled with MSX and MTX addition, respectively, is frequently used in the biopharmaceutical industry. These systems exploit the complementation of glutamine or nucleoside precursor (hypoxanthine and thymidine) auxotrophy by transfected GS or DHFR encoding genes. In combination with the removal of glutamine or nucleoside precursors in the media, the addition of GS and DHFR inhibitor (MSX and MTX, respectively) not only improves the selection stringency but can also result in increased productivity through one step or stepwise-increased gene amplification [16]. 1.3.4

Clone Selection

The selection process generates stable cell lines, more accurately, a stable pool of cells harboring GOI at random locations in the genome. Randomly integrated transgenes confer highly variable expression levels, possibly because of the different chromosomal context of integration sites, transgene copy-number variation, and disruption of the genome by gene amplification. It necessitates considerable screening effort to select production clones with the desired clone attributes (see Section 1.4.1). Several clone screening methods have been developed to accelerate CLD while increasing the predictability of clone assessments and lowering the number of clones to be assessed (see Section 1.4.2). Typical selection strategies start with a few hundred clones and end with a small number (∼10–20) of candidate production cell lines through each assessment stage [33]. 1.3.4.1

Primary Parameters During Clone Selection

The recombinant clone screening process is aimed at the isolation of the “right” candidate production clones through the evaluation of several features early in CLD for the large-scale production of therapeutic proteins in stirred tank bioreactors. Such parameters include (i) a high product yield, i.e. a product titer that is a function of high specific productivity (q; above 20 pg/cell/day to meet industrial demand) and/or the time integral of viable cell concentration (IVCC), (ii) cell line stability, which refers to maintaining the production capability over an extended period during subculture and scale-up, (iii) the desired product quality (e.g. glycosylation, proteolytic processing, molecular integrity, and aggregation) that meets predefined criteria with high consistency and comparability, (iv) cell line robustness, including acceptable cell growth with high viability and the preferred metabolism, such as low lactate synthesis, that fits the final large-scale production process [8, 16, 33].

1.4 Process Development

1.3.4.2

Clone Screening Technologies

When it comes to screening a large number of clones, significant progress has been made in clone screening technologies, pursuing efficient high-throughput screening methods, apart from the traditional time-consuming and laborintensive limiting dilution cloning. Most high-throughput methods rely on both automation of the cloning step and capture of the product secreted by the clones. Fluorescence-activated cell sorting (FACS) and semisolid matrix-based systems such as ClonePix FLTM or CellCelectorTM allow the rapid and high-throughput isolation of high-producing cell lines with a high level of confidence in “clonality” (reviewed in [2, 28]). These are fluorescence-based systems, necessitating conversion of the amount of secreted recombinant protein into a fluorescent signal. FACS, which was adopted originally to detect fluorescent cells, can be used to isolate high producers when combined with a labeling strategy: (i) capturing the secreted target protein on the cell surface or in close proximity to the individual clonal population stained with a fluorescent antibody or (ii) measuring the expression of surrogate reporters genetically linked to the GOIs, which include fluorescent proteins or surface marker proteins labeled with a fluorescent antibody [2, 28]. The use of semisolid media enables to limit the diffusion of secreted proteins while supporting cell growth and thereby facilitates the isolation of high-producing cell lines in a high-throughput manner when coupled with automated detection and clone picking [34, 35]. A complete automation system from clone selection to cell culture provides the highest throughput for the isolation of high-producing cell lines by employing the aforementioned screening methods and robotic systems [36, 37]. Despite the adoption of effective productivity screening technologies in the early stage of CLD, there is still the challenge of the performance consistency of candidate clones that have been adapted from static to suspension growth (in the case of the use of semisolid media) and scaled up to a large volume, manufacturing relevant production platform (see Section 1.4) [38]. Various scale-down models utilizing miniaturized systems with analysis capabilities have been developed in an attempt to simulate the large-scale performance of clones and to streamline the CLD [28].

1.4 Process Development Over the past few decades, a more than 100-fold improvement of titers in mammalian cells has been achieved by advances in CLD and selection techniques as well as the optimization of media and culture processes. Over 10 g/l of antibody concentrations in the fed-batch or perfusion process has been reported in many cases. Although the selection of the most suitable clone for the stable production of therapeutic proteins is one of the most important steps in the upstream process, cell performance, including productivity, product quality, and metabolic profiles, depends strongly on cell culture conditions such as the media, environmental parameters, culture mode, and scale-up processes [7, 8].

9

10

1 Platform Technology for Therapeutic Protein Production

1.4.1

Media Development

Early mammalian cell culture media contained bovine serum and animal-derived raw materials that were a complex mixture of unknown components. Because of safety concerns, serum-free media containing non-animal-derived hydrolysates such as soy, wheat, and yeast instead of animal-derived materials were developed and commercially available in the 1970s and 1980s [39, 40]. Nowadays, to avoid lot-to-lot variation, fully chemically defined media without any unknown components have been developed and implemented in small-scale as well as in large-scale culture processes [41, 42]. Media are a critical factor for improving cell growth, productivity, and product quality, so the optimization of culture media is necessarily considered in the early stage of CLD because of the clonal variation of metabolism, nutrient consumption, and interactions with components among production cell lines. A traditional approach for media optimization is based on the titration of individual components, but it is labor-intensive and time-consuming to evaluate the effect of numerous components in the media. To reduce the experimental efforts, a combination of statistical design of experiment (DoE) approaches with a high-throughput scale-down method is commonly used in industrial processes. DoE is useful not only in media optimization steps but also in the development of feed media for fed-batch culture [7, 8]. 1.4.2

Culture Environment

Optimization of culture environmental parameters, such as culture temperature, pH, agitation speed, dissolved oxygen and carbon dioxide, gas flow rate, osmolality, and more, is also required for a high yield of therapeutic proteins with reliable product quality. As with culture media, these parameters must be optimized for the specific production cell line because the effect of each parameter on culture performance, productivity, and product quality varies significantly from clone to clone [43, 44]. For example, culture temperature is the most commonly and easily adjustable culture parameter. To extend culture longevity and improve productivity, culture temperature is often lowered from 37 ∘ C to 30–35 ∘ C at 48–72 hours post inoculation, depending on the production cell line [45]. The effect of each parameter on culture performance and product quality has been well summarized previously [8]. A typical stirred tank bioreactor is equipped with temperature, agitation, pH, dissolved gas, and sometimes osmolality controllers. A traditional method of culture environment optimization was performed based on the control of individual parameters in a bench-scale bioreactor. Nowadays, a combination of statistical DoE approaches with a multiparallel microscale bioreactor system such as Ambr can unlock the bottleneck in process development.

®

1.4.3

Culture Mode (Operation)

The mode of mammalian cell culture is classified as a batch, fed-batch, or continuous (chemostat and perfusion) culture based on the mode of feeding the bioreactors. In the batch culture, the bioreactor is fed only once at the beginning of

1.4 Process Development

the culture with a media containing all nutrients and no more feeding, except for oxygen, after that. It is convenient to set up and maintain, and it is relatively safe against contamination. However, the culture duration is relatively short because of limiting nutrients or the accumulation of toxic by-products, resulting in relatively low productivity. As an alternative, fed-batch and perfusion cultures have been commonly used in industrial scales. In the fed-batch culture, a fresh volume of selected nutrients that are depleted during cell culture is added to the bioreactor to improve cell growth, culture longevity, and productivity. To improve the efficiency of fed-batch culture, components of the feed media and feeding strategies must be optimized. For example, glutamine is an essential component as a main nitrogen source as well as an energy source in mammalian cell cultures. During the culture, glutamine is metabolized into ammonia, which is known to reduce cell growth, protein production, and product quality. The accumulation of ammonia was significantly reduced in fed-batch cultures of CHO cells where glutamine concentration was maintained at a low level by feeding with the necessary amount of glutamine [46]. Currently, fed-batch cultures are most widely used for the large-scale commercial production of therapeutic antibodies [47]. In perfusion culture, cells are maintained at a much higher concentration over even months by feeding fresh media and simultaneously removing spent media while keeping cells in the bioreactors using cell retention devices. Perfusion culture has some drawbacks, such as complex and expensive equipment, risk of contamination, and regulatory uncertainties. Nevertheless, perfusion culture is used for low titer or unstable products such as recombinant blood clotting factors and enzymes because of the short retention time of the product in the bioreactor. Numerous biopharmaceutical companies have started to use the perfusion technology along with disposable equipment and cell retention devices such as alternating tangential flow (ATF). Using perfusion culture, they can achieve much higher cell density and product yield than with fed-batch culture, achieving considerable cost savings [48, 49]. 1.4.4

Scale-up and Single-Use Bioreactor

The mammalian cell culture process is usually performed in bench-scale bioreactors (1–2 L) and then scaled up to larger bioreactors (10 000–20 000 L) for commercial production purposes [50]. The aim of scale-up is to produce larger quantities of therapeutic proteins with equivalent product quality. Process scale-up, however, remains a challenging task because of difficulties in maintaining agitation efficiency, avoiding hydrodynamic shear and bubble stress, efficient oxygen and carbon dioxide transfer, etc. Therefore, a systematic approach for improving scale-up activities is necessary [50]. Today, single-use bioreactor systems are being increasingly used in mammalian cell culture as a new trend. As the product titers in mammalian cell cultures have been increased significantly over the past decade, a traditional bioreactor over 10 000 L may not be necessary in manufacturing therapeutic proteins. The scale of single-use bioreactors reaches up to 2000–2500 L. They have the advantages of lower investment and operational costs, flexibility, higher process replication,

11

12

1 Platform Technology for Therapeutic Protein Production

and reduced contamination rates compared to traditional bioreactors. However, there is still some work to be done to improve and optimize single-use bioreactor systems, particularly mixing and aeration. In addition, the amount of disposable materials generated in a single-use bioreactor system is a concern [7]. 1.4.5

Quality Analysis

Maintaining consistent and comparable product quality is one of the most important and challenging parts of therapeutic protein production because product quality significantly affects the safety and efficacy of the drugs. Measurement of the safety and efficacy of drugs is subdivided into biological activity, pharmacokinetics, pharmacodynamics, immunogenicity, and overall safety/toxicity. Quality attributes have also been subdivided into product-related impurities containing aggregation, fragmentation, glycosylation, disulfide bond formation, oxidation, deamidation, C- and N-terminal modifications, and so on, as well as process-related impurities containing DNA, host cell proteins, and raw materials. These quality attributes are highly affected by the cell line, culture media, and process conditions, and these factors must be optimized for specific cell lines and products [51, 52]. Among the several quality attributes, glycosylation is an important factor determining the quality of the therapeutic proteins. Glycosylation is easily affected by upstream process parameters such as the host cell type, glucose level, glutamine level, cell viability, culture temperature, and pH. Various approaches including host cell engineering and process development based on high-throughput and DoE methods have been applied to achieve the desired product quality [17].

1.5 Downstream Process Development Dramatic improvement in the productivity in large-scale processes has shifted the bottleneck from production to the purification step in therapeutic protein production. A key challenge in the purification step is the development of efficient and cost-effective systems with higher yield and purity. Many recent advances have been achieved in downstream process through the implementation of a high-throughput process, improved platform technologies, and unit operations based on quality by design (QbD) and DoE experimental optimization. In particular, QbD, a new concept for regulatory needs, has resulted in a noticeable change in the perspective on the development of downstream processing strategies in the biopharmaceutical industry. 1.5.1

Purification

Traditionally, two major methods, chromatographic and nonchromatographic separation, have been used for protein purification. The chromatographic method includes affinity, ion exchange (IEX), hydrophobic interaction (HIC), size exclusion, and mixed mode chromatography [7]. The most common affinity

1.5 Downstream Process Development

chromatographic process is protein-A method, which has been used for the capture and purification of mAb for over a decade. The protein-A resin has a dynamic binding capacity ranging from 15 to 100 g mAb/l with a high flow rate. However, the protein-A method has some drawbacks, such as resin leaching, nonspecific binding of impurities, including host cell proteins and DNA, and high price [53, 54]. Cation exchange (CEX) chromatography and mimetic resin of protein-A have been applied as an alternative to the traditional protein-A method. Subsequently, IEX and HIC are frequently used to purify non-mAb target proteins that are not tagged with a purification motif or to improve the purity of mAb because they have higher resolution in differentiating among related protein variants. Although these separation methods are much more cost-effective than protein-A, they suffer from limited capacity and elution issues because of the high affinity between the displacer and resins. As alternative methods, several optimization strategies for the resins, elution conditions, and operation modes were tested based on the DoE and modeling approaches [53, 55]. Recently, continuous chromatography methods such as multicolumn countercurrent solvent gradient purification have been implemented for the purification of recombinant streptokinase, mAb, and antibody fragments, along with cost savings and better productivity [7, 56–58]. Nonchromatographic separation includes a membrane-based system and phase partitioning; it is an alternative method to reduce or exclude chromatographic operations in the downstream process [59, 60]. The membrane-based system, which depends on the size, shape, and/or charge of the target proteins, has the advantages of low cost and ease of scaling up. The phase partitioning method, which is based on mixing two aqueous solutions of structurally different components, has the advantages of low cost, implementation of high-throughput screening, and combination of concentration and purification in a single step with a large scale. In the biopharmaceutical industry, the membrane-based system has been studied with regard to high permeability, capacity, sterility, and purification of large biomolecules, while the phase partitioning method has been studied to deal with issues of reduction and sensitivity caused by the complex interaction of multiple components and the feed stream variability [7, 56, 57]. 1.5.2

Quality by Design (QbD)

With an evolved understanding of the interactions between process parameters and product quality in mammalian cell cultures, a new concept, QbD, has been implemented in the biopharmaceutical industry. The aim of biopharmaceutical development is to design a quality product to meet patient needs and to consistently deliver the intended product performance. Because developmental strategies for therapeutic proteins differ from company to company and from product to product, more systematic approaches, such as the integration of prior knowledge, the relationship between a process and the quality attributes of the product using DoE, and quality risk management, have become necessary to enhance the desired quality of the product and help regulatory agencies to understand the strategies of a company. The QbD concept pursues a more

13

14

1 Platform Technology for Therapeutic Protein Production

scientific, systematic, and comprehensive approach to discovering, developing, and manufacturing pharmaceutical products. If the quality of the product is measured only at the final stage of manufacturing, it would be inefficient. Product quality should be monitored throughout the manufacturing process by implementing process design. For the successful implementation of QbD concepts, cooperation across a multitude of company teams from R&D to manufacturing to quality assurance and regulatory affairs is required. In 2003, the International Conference on Harmonization (ICH) of Technical Requirements for Registration of Pharmaceuticals for Human Use created a vision for a future pharmaceutical quality system that includes QbD concepts, and the FDA has started assessing the implementation and effectiveness of process design from development to manufacturing [61–63].

1.6 Trends in Platform Technology Development 1.6.1

Rational Strategies for Cell Line and Process Development

There exists a strong desire to understand molecular mechanisms underlying high productivity and protein quality in mammalian cells. With the recent emergence of both Chinese hamster and several CHO cell line genome sequences, various efforts have accelerated the development of next-generation CHO cell factories, which can be categorized into two areas: novel target/marker discovery and targeted approaches to CLD (Figure 1.2). Multiomic approaches including transcriptomics, proteomics, metabolomics, and more recent lipidomics data sets provide insights into physiological differences across production hosts and clones while suggesting potential engineering targets associated with desired attributes [64, 65]. Recently developed genome-scale models of CHO Omics study

Recombinant mammalian cells

Stable and high productivity

Improved host cells

Cellular markers Modeling

Rapid and Predictable CLD

TI platform

Hot spots Genome-wide screening

Continuous system

Media

Bioreactor (Perfusion/hybrid)

Harvest and Virus inactivation

Purification

Filtration and polishing

Drug substance

Figure 1.2 Schematic representation of trends in platform technology development encompassing the concepts from new CLD technology to continuous system.

1.6 Trends in Platform Technology Development

metabolism have demonstrated the integration of high-throughput omics data sets and are capable of simulating experimentally observed phenotypes [66]. To identify a larger set of novel engineering targets, genome-wide screening methods employing RNA interference (RNAi) or genome editing tools such as a regularly interspaced short palindromic repeat (CRISPR)/CRISPR-associated protein 9 (Cas9) system can also be implemented [67, 68]. Such advances in systems biology approaches will result in the translation of knowledge gained to improve CLD and bioprocessing in mammalian cells through targeted engineering strategies. Candidate cellular markers identified from high producers or altered bioprocessing may serve as not only cell engineering targets but also readouts for the assessment of process development. The preselection of host cells or engineered host cells harboring desired marker attributes can expedite CLD with little effort for process development. Additionally, these markers will facilitate the generation of process analytical technology with the implementation of real-time monitoring and data analysis. Accumulation of the information from the iterative bioprocess can be translated for better process understanding and process control, ultimately leading to the successful settlement of QbD concepts. Combined with improved host cells, targeted integration (TI)-based CLD offers new possibilities to shorten the CLD process with highly predictable gene expression. First-generation TI tools are based on recombinase-mediated cassette exchange (RMCE) systems, among which Cre/Lox and Flp/FRT systems have been widely adapted in CLD [2, 15, 16, 37]. The prerequisite platform cell lines are made through the traditional CLD process where reporter genes flanked by recombinase targeting sequences are integrated in highly transcribed chromosomal loci, the so-called hot spots. Afterward, the introduction of recombinase and targeting vectors with GOIs exchanges the reporter genes for GOIs, thereby allowing the use of hot spots for expressing therapeutic proteins. The advent of genome editing tools has enabled the more direct integration of GOIs into designated genomic sites in mammalian cells [68, 69]. The introduction of site-specific DNA double-strand breaks facilitates the integration of expression cassettes at precise locations by major DNA repair mechanisms, nonhomologous end joining (NHEJ), and homology-directed repair (HDR) [68]. The TI of transgenes allows stable and reproducible transgene expression between clones, suggesting the potential use of such TI platform technologies to overcome the limitations of clonal variation during CLD [69]. In the same context of target discovery, the systems biology-aided identification or prediction of hot spots will facilitate the implementation of this new CLD technology, which may allow the construction of high and stable production cell lines in a short time. 1.6.2

Hybrid Culture Mode and Continuous System

The fed-batch culture of mammalian cells for therapeutic protein production has been dominantly used in the biopharmaceutical industry for more than two

15

16

1 Platform Technology for Therapeutic Protein Production

decades. However, fed-batch cultures also have some drawbacks, such as the accumulation of by-products and increased osmolality during the cultures. In addition, therapeutic proteins that have a growth inhibitory effect or are subject to degradation in culture conditions cannot be produced in fed-batch cultures because of the prolonged culture duration. Therefore, there is increasing interest in using the perfusion culture process in the biopharmaceutical industry. With the trend of increased use of smaller equipment, an intensive and simple continuous process can provide operational flexibility and support process development, production for clinical study, and commercial manufacturing [70] (Figure 1.2). Furthermore, the low residence time of therapeutic proteins in a continuous system that includes perfusion culture facilitates the production of not only unstable proteins but also stable proteins such as mAbs [70, 71]. To advance this continuous system, stable cell lines with high productivity and the improvement of media performance are highly required. In addition, further development in downstream process is needed to meet commercial expectations [72]. Recently, a new culture method, referred to as a hybrid perfusion/fed-batch process, was developed to take advantage of both perfusion and fed-batch modes. The cell culture starts with the perfusion process for a few days to provide high cell density and then the operational mode changes to a fed-batch process for the remaining time. This process has shown a significant increase in productivity with short-duration cultures and low cost [73]. 1.6.3 Recombinant Human Cell Line Development for Therapeutic Protein Production Even though CHO cells have been dominantly used in therapeutic protein production because of the ability of humanlike glycosylated protein production, antigenic glycans such as N-glycolylneuraminic acid and α-galactose that are not presented in human-derived proteins are synthesized in CHO cells. The presence of antigenic glycan structures may result in increased immunogenicity, reduced efficacy, and altered pharmacokinetics in humans. In addition, some cases have been reported in which therapeutic proteins such as human interferon and recombinant factor VIII produced from CHO cells showed lower activity than those produced from HEK293 cells [74]. Therefore, the production of therapeutic proteins in human cell lines is expanding, and the FDA and EMA recently approved five drugs produced from HEK293 cells [14]. Two major concerns regarding the use of human cell lines are the risk of viral infection and low productivity. The current manufacturing process using human cells has multiple viral inactivation and clearance steps that may provide more effective viral clearance than CHO cells. However, the low productivity of human cells is still a concern. To overcome this issue, the implementation of a gene amplification system may be an efficient option. The recent TI-based CLD can also be applied to human cell lines [14, 74]. Because therapeutic protein production using stable human cell lines is a beginning step, more experience and research will be needed.

References

1.7 Conclusion Health care systems are facing tremendous costs associated with the increasing demand for therapeutic proteins to address unmet medical needs. The increasing demand for therapeutic proteins has been a driving force for the development of platform technologies that can be applied for a variety of products in the same way. Successful platforms help to streamline upstream and downstream process development, enhance predictability and efficiency during CLD and manufacturing, and accelerate time lines to deliver high-value recombinant therapeutics. On the basis of the established platform, the process of fine tuning in the area of cell line engineering, such as manipulation of PTMs, changes in media composition and culture parameters will provide greater flexibility with less resource and effort expenditure in bioprocess development to produce diverse product lines from stable, easy-to-express proteins to labile, difficult-to-express proteins. Integrating emerging trends in CLD and the process control tool box, including real-time process monitoring and control, automation, and scale-down single-use bioreactors, promises the advent of continuous cell culture bioprocessing, which will lead to a decrease in infrastructure with greater cost efficiency and high productivity with consistent product quality. The advancements in robust platforms will facilitate biopharmaceutical drug discovery and development and contribute to disease treatment through the high accessibility of therapeutic proteins.

Acknowledgment The authors thank the Novo Nordisk Foundation and Danish Council for Independent Research – Technology and Production Sciences (FTP) for funding.

Conflict of Interest The authors declare no conflict of interest.

References 1 Jayapal, K., Wlaschin, K., Hu, W., and Yap, G. (2007). Recombinant protein

therapeutics from CHO cells-20 years and counting. Chem. Eng. Prog. 103: 40–47. 2 Lai, T., Yang, Y., and Ng, S.K. (2013). Advances in mammalian cell line development technologies for recombinant protein production. Pharmaceuticals (Basel) 6 (5): 579–603. 3 Wurm, F.M. (2004). Production of recombinant protein therapeutics in cultivated mammalian cells. Nat. Biotechnol. 22 (11): 1393–1398.

17

18

1 Platform Technology for Therapeutic Protein Production

4 Morrison, C. and Lähteenmäki, R. (2016). Public biotech in 2015 – the num-

bers. Nat. Biotechnol. 34 (7): 709–715. 5 Walsh, G. (2014). Biopharmaceutical benchmarks 2014. Nat. Biotechnol. 32

(10): 992–1000. 6 Wells, E. and Robinson, A.S. (2017). Cellular engineering for therapeutic pro-

7

8 9

10 11

12 13 14

15

16

17 18

19

20

21

tein production: product quality, host modification, and process improvement. Biotechnol. J. 12 (1): https://doi.org/10.1002/biot.201600105. Gronemeyer, P., Ditz, R., and Strube, J. (2014). Trends in upstream and downstream process development for antibody manufacturing. Bioengineering 1: 188–212. Li, F., Vijayasankaran, N., Shen, A.Y. et al. (2010). Cell culture processes for monoclonal antibody production. MAbs 2 (5): 466–479. Butler, M. and Meneses-Acosta, A. (2012). Recent advances in technology supporting biopharmaceutical production from mammalian cells. Appl. Microbiol. Biotechnol. 96 (4): 885–894. Kelley, B. (2009). Industrialization of mAb production technology: the bioprocessing industry at a crossroads. MAbs 1 (5): 443–452. Huang, Y.M., Hu, W., Rustandi, E. et al. (2010). Maximizing productivity of CHO cell-based fed-batch culture using chemically defined media conditions and typical manufacturing equipment. Biotechnol. Progr. 26 (5): 1400–1410. Birch, J.R. and Racher, A.J. (2006). Antibody production. Adv. Drug Delivery Rev. 58 (5–6): 671–685. Kunert, R. and Reinhart, D. (2016). Advances in recombinant antibody manufacturing. Appl. Microbiol. Biotechnol. 100 (8): 3451–3461. Dumont, J., Euwart, D., Mei, B. et al. (2016). Human cell lines for biopharmaceutical manufacturing: history, status, and future perspectives. Crit. Rev. Biotechnol. 36 (6): 1110–1122. Kim, J.Y., Kim, Y.G., and Lee, G.M. (2012). CHO cells in biotechnology for production of recombinant proteins: current state and further potential. Appl. Microbiol. Biotechnol. 93 (3): 917–930. Noh, S.M., Sathyamurthy, M., and Lee, G.M. (2013). Development of recombinant Chinese hamster ovary cell lines for therapeutic protein production. Curr. Opin. Chem. Eng. 2: 391–397. Hossler, P., Khattak, S.F., and Li, Z.J. (2009). Optimal and consistent protein glycosylation in mammalian cell culture. Glycobiology 19 (9): 936–949. Xu, X., Nagarajan, H., Lewis, N.E. et al. (2011). The genomic sequence of the Chinese hamster ovary (CHO)-K1 cell line. Nat. Biotechnol. 31 (8): 759–765. Lewis, N.E., Liu, X., Li, Y. et al. (2013). Genomic landscapes of Chinese hamster ovary cell lines as revealed by the Cricetulus griseus draft genome. Nat. Biotechnol. 31 (8): 759–765. Fischer, S., Handrick, R., and Otte, K. (2015). The art of CHO cell engineering: a comprehensive retrospect and future perspectives. Biotechnol. Adv. 33 (8): 1878–1896. Wurm, F. (2013). CHO quasispecies—implications for manufacturing processes. Processes 1: 296–311.

References

22 Golabgir, A., Gutierrez, J.M., Hefzi, H. et al. (2016). Quantitative feature

23

24

25 26

27 28

29

30 31

32 33

34

35

36 37 38

extraction from the Chinese hamster ovary bioprocess bibliome using a novel meta-analysis workflow. Biotechnol. Adv. 34 (5): 621–633. Hu, Z., Guo, D., Yip, S.S.M. et al. (2013). Chinese hamster ovary K1 host cell enables stable cell line development for antibody molecules which are difficult to express in DUXB11-derived dihydrofolate reductase deficient host cell. Biotechnol. Progr. 29 (4): 980–985. Frye, C., Deshpande, R., Estes, S. et al. (2016). Industry view on the relative importance of “clonality” of biopharmaceutical-producing cell lines. Biologicals 44 (2): 117–122. Sanchez, A. and Golding, I. (2013). Genetic determinants and cellular constraints in noisy gene expression. Science 342 (6163): 1188–1193. Davies, S.L., Lovelady, C.S., Grainger, R.K. et al. (2013). Functional heterogeneity and heritability in CHO cell populations. Biotechnol. Bioeng. 110 (1): 260–274. González, M., Martín-Ruíz, I., Jiménez, S. et al. (2011). Generation of stable Drosophila cell lines using multicistronic vectors. Sci. Rep. 1: 75. Priola, J.J., Calzadilla, N., Baumann, M. et al. (2016). High-throughput screening and selection of mammalian cells for enhanced protein production. Biotechnol. J. 11 (7): 853–865. Harraghy, N., Calabrese, D., Fisch, I. et al. (2015). Epigenetic regulatory elements: recent advances in understanding their mode of action and use for recombinant protein production in mammalian cells. Biotechnol. J. 10 (7): 967–978. Moritz, B., Becker, P.B., and Göpfert, U. (2015). CMV promoter mutants with a reduced propensity to productivity loss in CHO cells. Sci. Rep. 5: 16952. Mariati, Yeo, J.H., Koh, E.Y. et al. (2014). Insertion of core CpG island element into human CMV promoter for enhancing recombinant protein expression stability in CHO cells. Biotechnol. Progr. 30 (3): 523–534. Brown, A.J. and James, D.C. (2016). Precision control of recombinant gene transcription for CHO cell synthetic biology. Biotechnol. Adv. 34 (5): 492–503. Porter, A.J., Racher, A.J., Preziosi, R., and Dickson, A.J. (2010). Strategies for selecting recombinant CHO cell lines for cGMP manufacturing: improving the efficiency of cell line generation. Biotechnol. Progr. 26 (5): 1455–1464. Caron, A.W., Nicolas, C., Gaillet, B. et al. (2009). Fluorescent labeling in semi-solid medium for selection of mammalian cells secreting high-levels of recombinant proteins. BMC Biotechnol. 9: 42. Hou, J.J., Hughes, B.S., Smede, M. et al. (2014). High-throughput ClonePix FL analysis of mAb-expressing clones using the UCOE expression system. New Biotechnol. 31 (3): 214–220. Lindgren, K., Salmén, A., Lundgren, M. et al. (2009). Automation of cell line development. Cytotechnology 59 (1): 1–10. Estes, S. and Melville, M. (2014). Mammalian cell line developments in speed and efficiency. Adv. Biochem. Eng. Biotechnol. 139: 11–33. Porter, A.J., Dickson, A.J., and Racher, A.J. (2010). Strategies for selecting recombinant CHO cell lines for cGMP manufacturing: realizing the potential in bioreactors. Biotechnol. Progr. 26 (5): 1446–1454.

19

20

1 Platform Technology for Therapeutic Protein Production

39 Fletcher, T. (2005). Designing culture media for recombinant protein produc-

tion: a rational approach. Bioprocess Int. 3: 30–36. 40 Kim, S.H. and Lee, G.M. (2009). Development of serum-free medium sup-

41

42

43

44

45

46

47

48

49

50 51

52

53

plemented with hydrolysates for the production of therapeutic antibodies in CHO cell cultures using design of experiments. Appl. Microbiol. Biotechnol. 83 (4): 639–648. Rouiller, Y., Périlleux, A., Collet, N. et al. (2013). A high-throughput media design approach for high performance mammalian fed-batch cultures. MAbs 5 (3): 501–511. Kishishita, S., Katayama, S., Kodaira, K. et al. (2015). Optimization of chemically defined feed media for monoclonal antibody production in Chinese hamster ovary cells. J. Biosci. Bioeng. 120 (1): 78–84. Butler, M. (2005). Animal cell cultures: recent achievements and perspectives in the production of biopharmaceuticals. Appl. Microbiol. Biotechnol. 68 (3): 283–291. Gawlitzek, M., Estacio, M., Fürch, T., and Kiss, R. (2009). Identification of cell culture conditions to control N-glycosylation site-occupancy of recombinant glycoproteins expressed in CHO cells. Biotechnol. Bioeng. 103 (6): 1164–1175. Yoon, S.K., Song, J.Y., and Lee, G.M. (2003). Effect of low culture temperature on specific productivity, transcription level, and heterogeneity of erythropoietin in Chinese hamster ovary cells. Biotechnol. Bioeng. 82 (3): 289–298. Wong, D.C.F., Wong, K.T.K., Goh, L.T. et al. (2005). Impact of dynamic online fed-batch strategies on metabolism, productivity and N-glycosylation quality in CHO cell cultures. Biotechnol. Bioeng. 89 (2): 164–177. Pollock, J., Coffman, J., Ho, S.V., and Farid, S.S. (2017). Integrated continuous bioprocessing: economic, operational, and environmental feasibility for clinical and commercial antibody manufacture. Biotechnol. Progr. https://doi.org/10 .1002/btpr.2492. Lehr, B. and Lyons, D. (2016). Perfusion in the 21st Century. Quality, flexibility, and cost savings are driving use of perfusion technology in biosimilars manufacturing. Biopharm. Int. 29 (8): 24–26. Tapia, F., Vázquez-Ramírez, D., Genzel, Y., and Reichl, U. (2016). Bioreactors for high cell density and continuous multi-stage cultivations: options for process intensification in cell culture-based viral vaccine production. Appl. Microbiol. Biotechnol. 100 (5): 2121–2132. Heath, C. and Kiss, R. (2007). Cell culture process development: advances in process engineering. Biotechnol. Progr. 23 (1): 46–51. Eon-Duval, A., Broly, H., and Gleixner, R. (2012). Quality attributes of recombinant therapeutic proteins: an assessment of impact on safety and efficacy as part of a quality by design development approach. Biotechnol. Progr. 28 (3): 608–622. Torkashvand, F. and Vaziri, B. (2017). Main quality attributes of monoclonal antibodies and effect of cell culture components. Iran Biomed. J. 21 (3): 131–141. Liu, H.F., Ma, J., Winter, C., and Bayer, R. (2010). Recovery and purification process development for monoclonal antibody production. MAbs 2 (5): 480–499.

References

54 Tarrant, R.D., Velez-Suberbie, M.L., Tait, A.S. et al. (2012). Host cell pro-

55

56 57

58

59 60

61

62 63

64

65

66

67

68

69

tein adsorption characteristics during protein A chromatography. Biotechnol. Progr. 28 (4): 1037–1044. Ghose, S., Hubbard, B., and Cramer, S.M. (2006). Evaluation and comparison of alternatives to protein A chromatography mimetic and hydrophobic charge induction chromatographic stationary phases. J. Chromatogr. A 1122 (1–2): 144–152. Cramer, S.M. and Holstein, M.A. (2011). Downstream bioprocessing: recent advances and future promise. Curr. Opin. Chem. Eng. 1 (1): 27–37. Saraswat, M., Musante, L., Ravidá, A. et al. (2013). Preparative purification of recombinant proteins: current status and future trends. Biomed. Res. Int. 2013: 312709. Steinebach, F., Ulmer, N., Decker, L. et al. (2017). Experimental design of a twin-column countercurrent gradient purification process. J. Chromatogr. A 1492: 19–26. Fröhlich, H., Villian, L., Melzner, D., and Strube, J. (2012). Membrane technology in bioprocess science. Chem. Ing. Tech. 84: 905–917. Giese, G., Myrold, A., Gorrell, J., and Persson, J. (2013). Purification of antibodies by precipitating impurities using polyethylene glycol to enable a two chromatography step process. J. Chromatogr. B 938: 14–21. Brunner, M., Fricke, J., Kroll, P., and Herwig, C. (2017). Investigation of the interactions of critical scale-up parameters (pH, pO2 and pCO2 ) on CHO batch performance and critical quality attributes. Bioprocess. Biosyst. Eng. 40 (2): 251–263. Rathore, A.S. and Winkle, H. (2009). Quality by design for biopharmaceuticals. Nat. Biotechnol. 27 (1): 26–34. Sommeregger, W., Sissolak, B., Kandra, K. et al. (2017). Quality by control: towards model predictive control of mammalian cell culture bioprocesses. Biotechnol. J. https://doi.org/10.1002/biot.201600546. Kildegaard, H.F., Baycin-Hizal, D., Lewis, N.E., and Betenbaugh, M.J. (2013). The emerging CHO systems biology era: harnessing the ’omics revolution for biotechnology. Curr. Opin. Biotechnol. 24 (6): 1102–1107. Zhang, Y., Baycin-Hizal, D., Kumar, A. et al. (2017). High-throughput lipidomic and transcriptomic analysis to compare SP2/0, CHO, and HEK-293 mammalian cell lines. Anal. Chem. 89 (3): 1477–1485. Hefzi, H., Ang, K.S., Hanscho, M. et al. (2016). A consensus genome-scale reconstruction of Chinese hamster ovary cell metabolism. Cell Syst. 3 (5): 434–443. Xiao, S., Chen, Y.C., Buehler, E. et al. (2016). Genome-scale RNA interference screen identifies antizyme 1 (OAZ1) as a target for improvement of recombinant protein production in mammalian cells. Biotechnol. Bioeng. 113 (11): 2403–2415. Lee, J.S., Grav, L.M., Lewis, N.E., and Faustrup Kildegaard, H. (2015). CRISPR/Cas9-mediated genome engineering of CHO cell factories: application and perspectives. Biotechnol. J. 10 (7): 979–994. Lee, J.S., Kallehauge, T.B., Pedersen, L.E., and Kildegaard, H.F. (2015). Site-specific integration in CHO cells mediated by CRISPR/Cas9 and homology-directed DNA repair pathway. Sci. Rep. 5: 8572.

21

22

1 Platform Technology for Therapeutic Protein Production

70 Croughan, M.S., Konstantinov, K.B., and Cooney, C. (2015). The future of

71

72 73

74

industrial bioprocessing: batch or continuous? Biotechnol. Bioeng. 112 (4): 648–651. Daszkowski, T. (2013). Continuous processing in biotech production: an alternative to a modern single use, batch, facility? Integrated Continuous Biomanufacturing Conference (20–24 October 2013), Barcelona, Spain. Konstantinov, K. and Cooney, C. (2015). White paper on continuous bioprocessing. J. Pharm. Sci. 104 (3): 813–820. Hiller, G.W., Ovalle, A.M., Gagnon, M.P. et al. (2017). Cell-controlled hybrid perfusion fed-batch CHO cell process provides significant productivity improvement over conventional fed-batch cultures. Biotechnol. Bioeng. 114 (7): 1438–1447. Gugliotta, A., Ceaglio, N., Raud, B. et al. (2017). Glycosylation and antiproliferative activity of hyperglycosylated IFN-α2 potentiate HEK293 cells as biofactories. Eur. J. Pharm. Biopharm. 112: 119–131.

23

2 Cell Line Development for Therapeutic Protein Production Soo Min Noh, Seunghyeon Shin, and Gyun Min Lee KAIST, Department of Biological Sciences, 291 Daehak-ro, Yuseong-gu, 34141 Daejeon, Republic of Korea

2.1 Introduction Biopharmaceuticals are any pharmaceutical drugs manufactured from a biological source, and they include nucleic acids, blood, vaccines, and therapeutic proteins such as monoclonal antibodies (mAbs) and erythropoietin (EPO). Since the first approval of recombinant human insulin (Humulin, Eli Lilly) as a recombinant therapeutic protein in 1982, the size of the biopharmaceutical market has been growing rapidly. The number of approved biopharmaceuticals is more than 300 in the USA and the European Union by 2016, and many more are in the process of being approved [1]. Therapeutic proteins are mostly produced by microorganisms (microbial systems) and mammalian cells (mammalian expression systems), and in rare cases by insect cells and plant cells. Microbial systems have the advantages of low cost, short timeline, easy control, and high productivity. The first approved recombinant human insulin was made from Escherichia coli. However, many other subsequent targets for therapeutic proteins were revealed to be more complex, and prokaryotes cannot express a large complex protein with multiple subunits, cofactors, and eukaryotic posttranslational modifications. Posttranslational modifications such as glycosylation and disulfide bond formation are very important for biological function, stability, and pharmacokinetics of the products [2]. Yeast expression systems (e.g. Saccharomyces cerevisiae and Pichia pastoris), which also achieve rapid cell growth and high protein yields, can produce some proteins that cannot be produced from E. coli because of some folding and glycosylation problems [3]. However, yeasts still cannot modify proteins with human glycosylation structures. Plant cells can be maintained in simple synthetic media, and they can synthesize complex multimeric proteins and glycoproteins with greater similarity to human counterparts [4]. However, plant cells do not necessarily produce proteins with the same 3D structure as that of humans. Furthermore, the glycosylation pattern occurring in the late Golgi apparatus is different between plants and mammals. Insect cells are mostly used for the development of viruslike particles and vaccines. They produce trimmed N-glycan precursors that do not develop further into terminal galactose and/or sialic acid residues [5]. Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

24

2 Cell Line Development for Therapeutic Protein Production

Insect

Yeast

Plant

Mannose GlcNAc Galactose Sialic acid (NGNA) Sialic acid (NANA) Fucose Xylose

Hamster (CHO)

Mouse

Animal

Human

Figure 2.1 N-linked glycosylation patterns in different organisms. Although proteins produced from yeast, insect cells, or plant cells lack galactose and sialic acid in their N-linked glycosylation pattern, proteins produced from animal cells show quite a similar pattern to those produced from human cells.

Mammalian expression systems are, therefore, preferred for the production of many complex therapeutic proteins despite the fact that they are slower and more expensive than the other systems. Although a microbial system requires cell lysis and subsequent protein refolding, most proteins can be secreted with proper folding in mammalian cells. Unlike yeast, insect, and plants, mammalian cells have a glycosylation pattern that is highly similar to that of humans (Figure 2.1). To produce therapeutic proteins from mammalian cells, an immortalized cell line is necessary. The cells should grow for prolonged periods to produce a large amount of therapeutic protein. There are several types of mammalian host cell lines used for the production of therapeutic proteins. Among them, Chinese hamster ovary (CHO) cells have been most widely used for the commercial production of biopharmaceuticals since 1987. Murine myeloma cell lines such as Sp2/0-Ag14 and NS0 have also been used but to a much lesser extent than CHO cells. Recently, there has been increasing interest in human cell lines such as human embryonic kidney (HEK293), human fibrosarcoma (HT-1080), and PER.C6. The host mammalian cell lines are transfected with the expression vector(s) with a gene of interest (GOI). The protein expression can be transient or stable (Figure 2.2). In the transient expression system, transfected cells are cultivated until the end of the production phase and GOI does not necessarily integrate into the genome of the cells. In the stable expression system, the cells undergo a selection phase after transfection. The GOI is integrated into the chromosome of the surviving cells, and stable cell lines with high productivity are finally selected. Although the process of stable cell line generation is time-consuming and labor-intensive, it is a common practice to produce a large amount of therapeutic protein with consistent product quality.

2.2 Mammalian Host Cell Lines for Therapeutic Protein Production

MTX or MSX Transfection Selection Transfection Production

Cloning

Expression

(a)

(b)

Production

Expression

Figure 2.2 Comparison between (a) transient expression system and (b) stable expression system. Transient expression system produces target proteins within a relatively short timeline. The stable expression system requires a relatively longer timeline, but it can stably produce a larger amount of the target protein.

In this chapter, mammalian cell lines used for therapeutic protein production (especially focusing on CHO and human cell lines) will be introduced with their brief history and characteristics. Then, the cell line development process of CHO and human cells will be reviewed. Some important considerations during the cell line development process will also be discussed.

2.2 Mammalian Host Cell Lines for Therapeutic Protein Production 2.2.1

CHO Cell Lines

CHO cells have been used most widely for the commercial production of therapeutic proteins since the approval of the first CHO-derived tissue plasminogen activator (tPA) in 1987. Currently, more than half of the top 10 selling drugs, including both chemical and biological, are produced in CHO cells [6]. Such popularity is explained by several traits that CHO cells possess (i) the ability to grow in a serum-free suspension culture for large-scale production of therapeutic proteins, (ii) amplification systems that enable gene amplification leading to higher productivity, (iii) posttranslational modification capability that makes the quality of CHO-derived glycoproteins compatible with that of humans, and (iv) resistance to human viral infections because of the lack of genes responsible for human viral entry. In addition, CHO cells have been demonstrated as safe hosts for the past three decades, which makes the approval from regulatory agencies easier. CHO cells were first established from the ovary of a Chinese hamster and immortalized by Theodore Puck in 1957 [7]. There are several CHO-derived

25

26

2 Cell Line Development for Therapeutic Protein Production

cell lines such as CHO-K1, CHO-S, DXB11, and DG44. The CHO-K1 cell line was cloned from the original CHO cells and has been used in many laboratories for decades, and the first genome sequence of CHO cells was published using CHO-K1 [8]. The CHO-S cell line was derived from the CHO-Toronto cell line (a sister cell line of CHO-K1) and adapted to grow in suspension culture. The DXB11 cell line was generated from the CHO-K1 cell line for the metabolic studies relating to cancer chemotherapy by deleting the dihydrofolate reductase (DHFR) activity, resulting in a deletion of one locus for DHFR and a missense mutation of the other DHFR locus [9]. The DXB11 cell line became the starting point of the DHFR system and became the host cell line for the production of human tPA. Because there was a detectable rate of reversion to DHFR activity in the DXB11 cell line, DG44 cell line was generated from the Toronto cell line with full deletion of the two DHFR loci on chromosome 2 [10]. 2.2.2

Human Cell Lines

Several human cell lines are used for the production of therapeutic proteins, and they can produce therapeutic proteins that are most similar to the proteins naturally synthesized by humans. One concern with the use of a human cell line is the increased risk of transfer of adventitious agents because of lack of a species barrier [11]. HEK293 cell lines overexpressing the adenovirus E1A and E1B genes were generated in 1977 by the transfection of HEK cells with adenovirus type 5 DNA [12] and have been used for transgene expression since then. Because of their ease of cultivation and transfection, HEK293 cell lines have been the most widely used host cell lines for transient expression. Recently, several therapeutic proteins produced in the HEK293 cell lines such as recombinant factor VIII Fc fusion protein (rFVIIIFc) and dulaglutide have been approved by the Food and Drug Administration (FDA) or the European Medicines Agency (EMA) [13]. The eHEK293T and HEK293E cell lines were generated by stable transfection of the parental line with the simian virus 40 (SV40) large T antigen gene and the Epstein–Barr virus nuclear antigen 1 (EBNA1) gene, respectively, to support the episomal replication and maintenance of the transfected plasmid DNAs [14, 15]. The HEK293H and HEK293F cell lines were clonally isolated for fast growth in serum-free medium, high transfection efficiency, and high level of protein production. One disadvantage of the HEK293 cell lines is the formation of large aggregates in suspension cultures. Therefore, to solve the aggregation problem, a hybrid of the kidney and B cell line (HKB-11) was developed by fusion of the HEK293S cell line with the human suspension cell line, 2B8 (a Burkitt’s lymphoma derivative) [16]. Although the HKB-11 cell line showed several characteristics suitable for producing therapeutic proteins, it showed unstable expression of EBNA1 during long-term cultivation because of a loss of EBV genome. To overcome the instability of the EBNA1 genome, the F2N78 cell line was established by the fusion of HEK293 cells with Namalwa cells [17]. In the F2N78 cell line, the EBV genome was inserted into the chromosome rather than existing as an episome. The PER.C6 cell line was created by immortalizing human embryonic retinoblasts with the E1 gene of adenovirus [18]. The production platform of

2.3 Development of Recombinant CHO Cell Lines

PER.C6 showing high productivity was established by Crucell and DSM biologics [19]. The PER.C6 cell line has been used for the production of classical vaccines as well as therapeutic proteins [20]. CEVEC’s amniocyte production (CAP) system was developed by immortalization of primary human amniocytes using adenoviral genetic E1/pIX functions [21]. In the CAP system, amniocytes were cotransfected with the E1-expressing plasmid containing the E1A-, E1B-, and pIX-function and the plasmid expressing the target protein. Only transfected cells can grow and divide while nontransfected cells cease to grow at an early passage. This method is novel in the way that it uses a nontumorigenic cell line, and the selection process does not require antibiotics. 2.2.3

Other Mammalian Cell Lines

Although not as popular as CHO cell lines and human cell lines, murine myeloma cell lines such as Sp2/0-Ag14 and NS0 have been used as host cell lines for the production of therapeutic proteins. The Sp2/0-Ag14 cell line is a nonIg-secreting cell line derived from a fusion of a BALB/c mouse spleen and the mouse myeloma cell line [22]. NS0 was cloned to establish a nonsecreting cell line and has been well studied in combination with the glutamine synthetase (GS) expression system [23]. However, murine cell lines express considerably higher levels of N-glycolylneuraminic acid (NGNA) and Galα1-3Galβ1-GlcNAc-R (α-Gal) compared to hamster cell lines. There was a case of severe anaphylactic reaction to the nonhuman glycan epitopes of an Sp2/0-derived antibody in patients [24]. Additionally, the baby hamster kidney (BHK21) cell line has been used in the production of some coagulation factors such as factor VIIa and factor VIII [25]. In addition, the COS cell line (fibroblast-like cell line derived from monkey kidney tissues) has been used mostly for transient expression [26].

2.3 Development of Recombinant CHO Cell Lines CHO cells are used for both transient and stable expression systems. A transient expression system is usually adopted when a large number of various proteins are prepared in a quick manner. The advantages of a transient expression system include a simpler expression vector with no selective marker, a shorter timeline for therapeutic protein production, applicability to a wide range of host cell lines, and suitability to multiple processing at the same time [27]. However, several drawbacks still limit the utility of transient expression systems for large-scale production of therapeutic proteins because (i) the transient expression experiments require a large quantity of the transfection reagent and plasmid DNA that are economically burdensome and (ii) the protein yield from a transient system is dependent on the transfection efficiency, which may vary among the experiments. In addition, although the volumetric productivity in transient expression systems has been improved lately [28–30], it is still lower than that in stable gene expression systems. A stable expression system, on the other hand, requires a longer timeline because of the selection of the transfected cell pool followed by gene

27

28

2 Cell Line Development for Therapeutic Protein Production

amplification and/or clone selection processes. However, GOI is integrated into the chromosome of the host cells during the cell line generation process. Once stable clones with high productivity are established, those clones show consistent and high production of the target protein. High producing clones can be cultivated repeatedly under optimal culture conditions leading to maximum yield. Therefore, to produce a large amount of therapeutic proteins with consistent product quality, stable expression systems are inevitably used in the biopharmaceutical industry. 2.3.1

Expression Systems for CHO Cells

To establish a stable recombinant CHO cell line with high productivity, two expression systems are dominantly used: the DHFR system and the GS system. DHFR catalyzes the conversion of folic acid to tetrahydrofolate, which is required for biosynthetic pathways that produce glycine, purine, and thymidylic acid. The DHFR system can be applied to mutant strains of CHO cells such as DXB11 and DG44, in which the dhfr gene was either mutated or deleted. To survive, such auxotrophic cell lines require a culture medium containing glycine, hypoxanthine, and thymidine (GHT). In a DHFR system, a GOI is transfected into host cells with the dhfr gene either in the same expression vector or in a different expression vector. The transfected cells undergo selection in the media without GHT, and the surviving cell pool will have the GOI together with the dhfr gene in their genomes. The inserted genes can be amplified with the use of MTX, a DHFR inhibitor, and the DHFR system is frequently used because of its high efficiency of gene amplification. The GS system uses the gs gene as a selectable marker. GS catalyzes the conversion of ammonia and glutamate into glutamine, and MSX inhibits the activity of the GS protein. Although the DHFR system can be used only with DHFR-deficient cell lines, the GS system can be used with cell lines having an endogenous gs gene. NS0 murine myeloma cells are auxotrophic for glutamine and can be selected in glutamine-free medium using the GS system. In contrast, CHO cells possess an endogenous gs gene and can be selected with the concurrent addition of MSX at a low level [31, 32]. The CHO-K1 cell line was first used for the GS system [33]. To improve the efficiency of the cell line generation, a GS knockout cell line was also developed [34]. Compared to the DHFR system that requires a long timeline for gene amplification through the gradual increment of MTX, the GS system can achieve sufficient expression level with a single round of selection and amplification, thereby shortening the total timeline required for cell line generation [23, 33]. Furthermore, the GS system mitigates ammonia accumulation in culture medium because the overexpressed GS protein catalyzes the conversion of glutamate and ammonia into glutamine [35]. 2.3.2 Cell Line Development Process Using CHO Cells Based on Random Integration Both the DHFR and GS systems share a concept that a selection gene (dhfr or gs) is transfected with a GOI, providing a selective advantage to the cells in the selective

2.3 Development of Recombinant CHO Cell Lines

Vector construction ne tion ge Selec or GS) R (DHF

Transfection

GOI

Clone selection

Pool selection and gene amplification

Suspension adaptation Expansion and evaluation

Cell line characterization and process development Cell banking

Figure 2.3 Schematic representation of cell line development process. Expression vector(s) with selection gene and GOI is transfected to host cells. The transfected host cells undergo pool selection process and, if necessary, repeated gene amplification process. Single-cell-derived clones are selected from the pool and clones are expanded and adapted to grow in suspension culture. Final cell lines are characterized and undergo process development to achieve maximum production.

medium. A schematic representation of the development of a stable recombinant cell line is shown in Figure 2.3. 2.3.2.1

Vector Construction

The first step of the cell line development process is the molecular cloning of a GOI in a mammalian expression vector. In most cases, a selection gene (dhfr

29

30

2 Cell Line Development for Therapeutic Protein Production

or gs) is in the same expression vector but sometimes can be in a separate vector because they can be linked upon gene transfer by intracellular ligation or cointegration at the same site [36]. Antibiotics such as zeocin, geneticin (G418), hygromycin B, blasticidin, and puromycin can be used for dual selection and a relevant resistance gene can be inserted into the expression vector. Protein expression can be improved through vector design by providing the appropriate promoter, signal peptide, and codon optimization [37]. To increase the probability of obtaining a high-level producer, the selection gene is driven from a weak promoter, and the GOI is driven from a strong promoter such as the CMV promoter. Several cis-acting elements might also affect cellular expression. For example, the Kozak sequence has a role in translation initiation [38], and the viral posttranscriptional regulatory sequence from the woodchuck hepatitis virus (WPRE) enhances transgene expression when placed downstream of an open reading frame [39]. Moreover, signal peptides can improve the secretion efficiency of a therapeutic protein, and the optimization of signal peptides has been well studied [40]. Codon optimization can improve the expression level of GOI because various organisms show a preference for certain codons [41]. In CHO cells, protein expression was improved by codon optimization [42, 43]. 2.3.2.2

Transfection and Selection

The expression vector(s) are then transfected into host cells. Many transfection methods for mammalian cells have been developed, which include biological, chemical, and physical methods [44]. The biological method usually involves virus-mediated transfection (transduction), and it is highly efficient and easy to use. However, because of the risk of immunogenicity and the requirement for higher levels of safety containment, a nonviral gene transfer is rather preferred for manufacturing therapeutic proteins. A chemical method was the first method to be used for mammalian cells, and it uses a cationic polymer, calcium phosphate, or cationic lipid. The chemical method has the merits of less induction of the immune system, and there are many commercially available products with good efficiency [45]. A physical method has been developed most recently, and it includes direct microinjection, electroporation, and laser irradiation. Electroporation is the most widely used method because it can easily and rapidly transfect a large number of cells once the optimal conditions for transfection are determined [44]. The transfected cells, 24–48 hours after the transfection, are transferred to a selection medium to undergo pool selection. In the case of DHFR system, the selection medium will be deficient in GHT [46]. In the case of GS system, the selection medium will be deficient in glutamine, and 25–50 μM of MSX may be added to suppress the endogenous GS protein of the host cells [34]. The resulting pool of cells surviving in the selection medium will have the dhfr or gs gene and GOI that are randomly integrated into their genomes at various regions. Each cell will exhibit different levels of protein expression because the number of transgene(s) and the surrounding gene sequence of the integration sites vary. 2.3.2.3

Gene Amplification

When using the DHFR system, it is common to amplify the GOI during the cell line development process. Cells develop resistance to the increased level of MTX

2.3 Development of Recombinant CHO Cell Lines

by amplifying the dhfr gene. Because the size of the amplification unit is much larger than that of the dhfr gene (130–3000 kilobases), the GOI is coamplified together with the dhfr gene [47]. The initial pools are exposed to a low concentration of MTX so that only the cells containing at least several copies of the dhfr gene and GOI may survive. Such a procedure is repeated using stepwise increasing concentrations of MTX because a single-step exposure to a high concentration of MTX may result in MTX-resistant DHFR mutants or cells with altered MTX transport properties [48]. This amplification process is often a bottleneck during the cell line development process because it requires several months to reach the high MTX level. The gene copy number of the GOI increases with the increased level of MTX, but there exists a saturation limit after which the gene copy number does not increase any more even with a higher level of MTX [49]. When using the GS system, the timeline for gene amplification is relatively short (Figure 2.4). A single round of amplification (100–500 μM) is sufficient to achieve efficient expression, and further rounds do not seem to result in higher product titers [33]. To minimize the time required for cell line generation, no amplification is even used. In most cases, a low level of MSX (25–50 μM) used during the first selection after transfection is sufficient enough to achieve a desired level of protein production [50]. 2.3.2.4

Clone Selection

Another time-consuming step in the cell line development process is the selection of clones that exhibit the best productivity, desired product quality, and

Transfection

Selection with no MTX or low level of MTX

Gene amplification (gradual increment of MTX)

Cloning DHFR deficient CHO

(a) Time line

Transfection

Single round selection with high level of MSX

Cloning Usually CHO-K1

(b)

Figure 2.4 Comparison between (a) DHFR system and (b) GS system. DHFR system requires repeated gene amplification process with a gradual increment of MTX and it makes the cell line development timeline longer. GS system does not require repeated amplification process and the cell line development timeline is shorter.

31

32

2 Cell Line Development for Therapeutic Protein Production

good growth profile. A traditional way of clone selection is “limiting dilution” in which cells are transferred to multiwell plates such that on an average only one cell is present per two to three wells and forms a colony of a single cell. Each clone is then scaled up and evaluated for its growth, productivity, and product quality. Because the traditional methods are time-consuming and limited by the number of clones that can be screened, alternative high-throughput methods (i.e. fluorescence-activated cell sorting (FACS), the ClonePixTM system (Genetix), and the Cell XpressTM system (Cyntellect)) have been developed and are widely used in industry. FACS is the most widely used method for high-throughput screening. After staining cells with antibodies conjugated with fluorescent molecules, high producing clones can be sorted out based on the fluorescence intensity of cells by FACS [51–53]. The ClonePix system is an automated colony picker that enables screening a large number of clones in a semisolid medium. Therapeutic proteins secreted from the colony are captured by antibodies conjugated with fluorescent molecules, and the colonies with high fluorescent intensity are then picked [54, 55]. The Cell Xpress system identifies high producing clones in a semisolid medium by live cell imaging with a detection reagent, and non- or low-producing clones are eliminated by laser-induced apoptosis [56]. In the case of the DHFR system, there exist two different strategies for clone selection according to the clone selection timing in the gene amplification process. The first strategy is selection based on individual clones in which cell cloning is performed once at the time of the first selection of parental clones and once again after obtaining high-producer clones through gene amplification. The second strategy is a selection based on cell pools in which cell cloning is performed only once at the final selection stage after gene amplification is completed. Although the first strategy is labor-intensive and requires a longer time, it has been shown to be more efficient in providing higher producers [57]. After the isolation of the clones with high productivity, their stability in regard to productivity is evaluated by subculturing them for a couple of months. Culture conditions such as temperature, pH, dissolved oxygen tension (DOT), and osmolality can affect the productivity of the clones [58–64]. Therefore, the culture conditions of selected clones are optimized in a small-scale bioreactor with a working volume of 1–2 l using cell culture media used for large-scale cultures. Nowadays, a microscale cell culture system that uses cost-effective, disposable, multiple micro-bioreactors and is controlled by an automated workstation is also widely used for culture process optimization in industry. 2.3.3 Cell Line Development Process Using CHO Cells Based On Site-Specific Integration Random integration of a GOI has some drawbacks such as variation in productivity and a laborious selection process. To overcome these drawbacks, site-specific integration of a GOI has emerged as an alternative method of transgene integration. By adopting a targeted integration method, the GOI can be inserted into a site that shows a higher expression level of the therapeutic protein and stability on a long-term basis. For site-specific integration of a GOI, the information

2.3 Development of Recombinant CHO Cell Lines

on a high-expressing site, known as a hot spot, is first needed [65]. Although the exact mechanism explaining the different levels of gene expression at different integration sites is still under study, several mechanisms are considered as being responsible, such as cis-regulatory elements, DNA and histone modifications, and chromosome structure [66–68]. A recombinase-mediated cassette exchange (RMCE) method is capable of targeted integration using site-specific recombinases and distinct recombinaserecognition sequences. After specific sites are targeted by random integration of heterologous genes flanked by recombinase-recognition sequences, those sites are confirmed by different levels of gene expression. Then, the sites that show a higher level of expression are exchanged with the GOI by the corresponding recombinase. The Cre/loxP system, consisting of the Cre recombinase and the loxP sequence, was first used in the CHO-DG44 cell line to obtain high-producing clones selected by measuring the GFP expression level [69]. This system has been applied to different CHO cell lines in modified forms, including accumulative integration, mutated loxP sequences, and epigenetic modifier elements [70–72]. Another example of RMCE is the Flp/FRT system, and it used the Flp recombinase and the FRT sequence in the CHO DG-44 cell line producing EPO to screen for a stable EPO producer [73]. Finally, PhiC31 integrase was exploited in the CHO-S cell line producing a humanized IgG1 antibody to reduce the possibility of a reversible reaction in other systems (Cre/loxP and Flp/FRT) by irreversible integration of the GOI into the genome [74]. One of the drawbacks of the targeted integration methods by a recombinase-mediated system is that they need to prepare the tagged sites containing reporter sequences flanked by FRT or loxP before GOI integration into specific sites of the host cells [75]. An endonuclease-mediated targeted integration method has also been used for site-specific integration by introducing a DNA double-strand break (DSB) and then being repaired by DNA damage repair pathways. There are three major kinds of targetable nucleases: zinc-finger nuclease (ZFN), transcription activator-like effector nuclease (TALEN), and clustered regularly interspaced short palindromic repeats (CRISPRs)/CRISPR-associated protein (Cas) RNA-guided nuclease [76]. These nucleases have a DNA-binding domain or guide RNA and cleave the targeted sequence by the Fok I cleavage domain or the Cas9 protein. Each zinc-finger motif interacts with a 3-bp DNA sequence, and two ZFNs need to form a dimer to function properly [77]. TALENs also contain the Fok I domain as a nuclease domain but use another type of DNA-binding domain, called transcription activator-like effectors (TALEs) [78]. Unlike ZFNs, TALENs can recognize DNA sequences by pairing one TALE repeat with one base, which guarantees a higher specificity and lower off-targeting compared to ZFNs [79]. It was demonstrated that ZFN- and TALEN-mediated targeted integration is applicable in mammalian cells by introducing the IgG gene into the FUT8 locus in the CHO-K1 cell line [80]. In addition to the ZFNs and TALENs, CRISPR/Cas9 has recently emerged as a genomic engineering tool. It is composed of a single-chain guide RNA targeting specific DNA sequence and the Cas9 protein acting as an active nuclease in the presence of the guide RNA [81]. The CRISPR/Cas9 system has many advantages, including ease of design and preparation, lower

33

34

2 Cell Line Development for Therapeutic Protein Production

cytotoxicity, higher efficiency, and lower cost compared to the other two methods mentioned above [82–85]. CRISPR/Cas9-mediated gene integration has been attempted in human cells by means of CRISPR/Cas9-mediated precise integration into a target chromosome (PITCh) system dependent on the microhomology-mediated end-joining (MMEJ) mechanism [86]. In CHO cells, CRISPR/Cas9 system can integrate a GOI into specific sites by exploiting the homology-directed repair (HDR) pathway [87]. Transposons are mobile genetic elements found in various organisms and classified by their transition mechanism. DNA transposons are composed of inverted repeats, directed repeats, DNA-binding domain, nuclear localization signal (NLS), and catalytic amino acids [88]. DNA transposons need a specific transposase enzyme that is encoded between inverted terminal repeats (ITRs) to transfer by a mechanism of copy-and-paste, and several kinds of DNA transposon have been used as a genetic engineering tool in many studies [89]. Two vectors, a donor plasmid with a GOI flanked by ITRs and a helper plasmid with a transposase sequence, are necessary for transgene integration [90]. Transposition activities of Sleeping Beauty, Tol2, Mos1, and PiggyBac transposons were measured in different mammalian cell lines including CHO cells [91]. Although the transposon-mediated gene integration method has no intended target specificity, some transposons have the tendency to integrate into specific regions, such as transcriptional start sites (TSSs), CpG islands, and DNase I hypersensitive sites [92]. It was observed that the PiggyBac transposon has a preference for transposition into specific sites including the TTAA sequence as well as TSSs and CpG islands in primary human T cells [93]. PiggyBac transposon-mediated integration was proved as an efficient tool and showed higher productivity and stability compared to the random integration method due to insertion into actively transcribed regions [94]. Recently, cell pools generated by the PiggyBac transposon yielded high productivity up to 7.6 g/l [95]. Furthermore, the product quality of the cell pools generated by the PiggyBac transposon was comparable to the control, and the higher productivity was reproducible in 36 l bioreactor [96].

2.4 Development of Recombinant Human Cell Lines 2.4.1

Necessity for Human Cell Lines

Although CHO cells mimic many of the human glycosylation patterns, proteins produced by CHO (and other nonhuman) cells sometimes show nonhuman glycan structures such as NGNA and α-Gal. Because humans have circulating antibodies against NGNA and α-Gal, proteins produced from a nonhuman cell line can be highly immunogenic and/or rapidly cleared from circulation [97]. In addition, CHO and other nonhuman cells lack some sugar-transferring enzymes such as α2-6 sialyltransferase and α1-3/4 fucosyltransferase [98]. There are several cases in which human cells must be used as a host cell line. Drotrecogin alfa, a recombinant activated protein C showing antithrombotic,

2.4 Development of Recombinant Human Cell Lines

anti-inflammatory, and profibrinolytic properties, is not produced in CHO cells with adequate efficiency because CHO cells have less capacity for γ-carboxylation than that of human cell lines [99]. Dimeric cartilage matrix protein-angiopoietin-1 (CMP-Ang1), a potential growth factor for therapeutic angiogenesis and vascular stabilization, requires specific N-glycosylation. Dimeric CMP-Ang1 produced by CHO cells showed no activity while that produced by HEK293 cells showed reproducible activity [100]. Therefore, the necessity of human cell lines for the production of therapeutic proteins cannot be disregarded, and efforts in using human cell lines should be pursued. Human cell lines, especially the HEK293 cell line, are widely used for transient expression systems. In general, the production yield of HEK293 cell lines is relatively higher than that of CHO cell lines when therapeutic proteins are transiently expressed [101]. However, large-scale production of therapeutic proteins using a transient expression system requires a large amount of purified DNA and a large number of cells to be prepared on the day before transfection, possibly causing economical and technical problems. Another major concern with transient expression systems is the batch-to-batch variability in regard to protein yield and quality, especially protein glycosylation, although there are studies showing little batch-to-batch variation [102]. 2.4.2

Stable Cell Line Development Process Using Human Cell Lines

Human cell lines are also used for establishing stable cell lines but in fewer cases than that of CHO cell lines. Basically, the protocol for stable cell line development for human cell lines is similar to that used for CHO cell lines. In most cases for human cell lines, however, only antibiotics are used for the selection, and no gene amplification is available. Accordingly, the specific productivity of human stable cell lines is relatively low compared to stable CHO cell lines. Although not as much as CHO cell lines, hundreds of milligrams per liter of therapeutic proteins were stably produced in human cells [51, 103]. A stable HEK293 cell line producing human recombinant IFNα2b was established in the presence of blasticidin, and the product titer exceeded 200 mg/l in a batch culture [103]. In addition, up to a 655 mg/l titer was achieved using the antibody-producing stable HEK293 cell line screened with high-throughput FACS [51]. To increase the productivity of HEK293 cell lines, the GS-mediated gene amplification system was applied to HEK293 cells. However, a high level of expression of endogenous gs gene in HEK293 cells resulted in elevated resistance to MSX and therefore hampered the GS-mediated selection and gene amplification by MSX [104]. A stable F2N78 cell line producing IgG was established using the puromycin gene as a selection marker, and it showed an antibody production up to 340 mg/l in fed-batch cultures [105]. A PER.C6 cell line producing an antibody was established in the presence of neomycin (G418), and it showed a high level of antibody production (300–500 mg/l) in a batch culture [19]. With the modified perfusion/fed-batch system, yields of alpha-1-antitrypsin (A1PI) over 2.5 g/l and of human IgM antibodies up to 2 g/l were achieved [106, 107].

35

36

2 Cell Line Development for Therapeutic Protein Production

The CAP system of CEVEC does not require the use of antibiotic selection. Cotransfection of the primary human amniocytes with a plasmid expressing adenoviral E1 functions and a plasmid containing a GOI resulted in stable cell lines expressing a fully glycosylated and sialylated protein. In an optimized fed-batch culture, 200–250 mg/l of C1-inhibitor was produced [108].

2.5 Important Consideration for Cell Line Development In addition to the productivity, there are several factors that should be considered during the cell line development process, and these include clonality, stability, and product quality. 2.5.1

Clonality

Once cells are transfected with expression vectors, clonal variations occur because of the differences in the chromosomal context of the randomly integrated genes and/or a disruption of the endogenous genes caused during the gene integration and amplification process. Clones, even if they are derived from the same parental cells, react differently to culture conditions such as culture temperature, osmolality, and media additives [109–112]. Such clonal variation, therefore, should be considered in clone selection during the cell line development process. Recently, regulatory agencies such as the FDA and EMA have required “proof of clonality” for the approval of biopharmaceuticals, which is an important issue especially in the industry [113]. Clonality means that the cell population is arisen from a “single cell,” which is often assured by several rounds of limiting dilution. Such a regulatory request was established to ensure the purity and homogeneity of the product. To confirm the clonality, visual monitoring of plates is usually done using automatic image devices such as CloneSelectTM Imager (Molecular Devices) and Cell MetricTM imager (Solentim). Using such devices, 96-well plates are automatically imaged to identify the formation of colonies derived from a single cell. However, the viewpoint of the industry on the importance of clonality seems to be different from that of regulatory agencies [114]. Because the host cells used for the manufacture of biopharmaceuticals are immortalized cell populations that have escaped from normal control of cell division, the rate of genetic drift and chromosomal instability are inevitably very high [115]. Even after single cell cloning, genomic heterogeneity caused by inherent DNA replication errors, error-prone DNA repair process, and Darwinian selection occurs during cell growth and expansion. Therefore, although strict cloning is conducted during the cell line development process, the heterogeneity of the producing cells cannot be avoided and more focus should be on the consistency of the manufacturing process and on the quality of the product rather than clonality itself. 2.5.2

Stability

The productivity of recombinant cells often declines as cell age increases during prolonged culture. Unstable clones are not suitable for large-scale production

2.5 Important Consideration for Cell Line Development

because cell age inevitably increases during the scale-up process. Thus, cell line stability is an important factor to be considered in clone selection. Although the underlying causes and the precise molecular mechanisms behind cell line instability have yet to be fully elucidated, possible reasons are loss of gene copy number, transcriptional regulation, and posttranscriptional regulation. The decrease in protein productivity is in many cases caused by a loss in gene copy numbers [49, 116–118]. In some cases, the gene copy number did not change, and only the mRNA level was decreased [119, 120]. Methylation-induced transcriptional silencing has been pointed out as one of the epigenetic reasons for production instability [121, 122]. In contrast, there was a case in which antibody productivity decreased without any decrease in both gene copy number and mRNA level. In this case, decreased cell growth was responsible for the decreased antibody yield [123]. In the case of a CHO cell line with DHFR-mediated gene amplification, there are a number of reports that have shown instability of production during long-term culture in the absence of MTX [49, 116, 124]. Production instability was observed even in the presence of MTX but at a lesser extent than in the absence of MTX. In addition, there is a report that MTX did not show any benefit in product stability while the addition of hypoxanthine and thymidine demonstrated better maintenance of genetic stability [125]. Compared with the DHFR expression system, less information is available on the production stability of rCHO cells with GS-mediated gene amplification. Humanized antibody productivity of GS-CHO cells decreased during long-term culture regardless of the presence of MSX [126]. In contrast, antibody productivity of GS-CHO cells was kept constant in the presence of MSX [32]. Conflicting reports on the production stability are likely due to clonal variations. The clones derived from the same cell line generation process showed different production stabilities [122, 127]. Generation of clones with improved production stability facilitates large-scale production of therapeutic proteins. The location of a GOI in the chromosome may affect the production stability. It was reported that clones with amplified genes located near the telomeric regions are more stable than those located in other regions [128]. In addition, some chromosomes were revealed to be genetically more conservative among different CHO cell lines [129]. Such information at the chromosomal level will enable identification of potential target sites for site-specific integration to generate producing cell lines with improved production stability. Furthermore, gene engineering of CHO host cells would also lead to the isolation of more clones with improved production stability as observed with the knockout of the gene Fam60A in CHO-K1 [130]. 2.5.3

Quality of Therapeutic Proteins

Maintaining consistent and comparable product quality is one of the most important and challenging issues in the production of therapeutic proteins. During the cell line development process, product quality attributes vary significantly among the clones. Therefore, it is important to select the clones producing the therapeutic protein with the desired quality using various analytical assays and quality

37

38

2 Cell Line Development for Therapeutic Protein Production

assessment criteria. In particular, for the production of biosimilar candidates, the extent of comparability to original drugs, which is the most important concern during the approval process, is an important criterion to select the clones. To determine whether the quality of a biosimilar candidate sits within the “goalposts” of acceptable features, state-of-the-art liquid chromatography and mass spectrometry technologies are used to compare intact masses, protein sequences, posttranslational modifications, and microvariants [131]. Methods based on chromatographic or electrophoretic separation such as size-exclusion chromatography (SEC), high-performance liquid chromatography (HPLC), and isoelectric focusing (IEF) have been used traditionally to assess the quality attributes such as molecule integrity, aggregation, glycosylation, and charge heterogeneity. Recently, rapid and high-throughput analysis of sequence variants as well as glycosylation patterns during the early stage of cell line development process has become available [132–134].

2.6 Conclusion Mammalian cells are preferred for the production of complex therapeutic proteins. Among them, CHO cells have been dominantly used in the industry for the production of therapeutic proteins including antibodies. To a lesser extent, human cells such as HEK293, HT-1080, and PER.C6 have also been used for commercial production of therapeutic proteins. Currently, the generation of recombinant cell lines is the most time-consuming step for developing the process for commercial production of therapeutic proteins. However, recent progress in gene editing technology such as CRISPR/Cas9 will make cell line generation more efficient and faster. As the demand for antibodies and other therapeutic proteins continue to increase, the popularity of mammalian cells, particularly CHO cells, is likely to persist.

References 1 Biopharma (2017). Biopharmaceutical products in the U.S. and European

2 3 4

5

markets. http://www.biopharma.com/approvals.html (accessed 22 April 2019). Walsh, G. and Jefferis, R. (2006). Post-translational modifications in the context of therapeutic proteins. Nat. Biotechnol. 24 (10): 1241–1252. Gerngross, T.U. (2004). Advances in the production of human therapeutic proteins in yeasts and filamentous fungi. Nat. Biotechnol. 22 (12): 1589. Hellwig, S., Drossard, J., Twyman, R.M., and Fischer, R. (2004). Plant cell cultures for the production of recombinant proteins. Nat. Biotechnol. 22 (11): 1415–1422. Kost, T.A., Condreay, J.P., and Jarvis, D.L. (2005). Baculovirus as versatile vectors for protein expression in insect and mammalian cells. Nat. Biotechnol. 23 (5): 567–575.

References

6 Philippidis, A. (2017). The top 15 best-selling drugs of 2016. http://www

7 8

9

10

11

12

13

14

15

16

17

18

19 20

21

.genengnews.com/the-lists/the-top-15-best-selling-drugs-of-2016/77900868 (accessed 22 April 2019). Puck, T.T. (1957). The genetics of somatic mammalian cells. Adv. Biol. Med. Phys. 5: 75–101. Xu, X., Nagarajan, H., Lewis, N.E. et al. (2011). The genomic sequence of the Chinese hamster ovary (CHO)-K1 cell line. Nat. Biotechnol. 29 (8): 735–741. Urlaub, G. and Chasin, L.A. (1980). Isolation of Chinese hamster cell mutants deficient in dihydrofolate reductase activity. Proc. Natl. Acad. Sci. U.S.A. 77 (7): 4216–4220. Urlaub, G., Kas, E., Carothers, A.M., and Chasin, L.A. (1983). Deletion of the diploid dihydrofolate reductase locus from cultured mammalian cells. Cell 33 (2): 405–412. Swiech, K., Picanco-Castro, V., and Covas, D.T. (2012). Human cells: new platform for recombinant therapeutic protein production. Protein Expression Purif. 84 (1): 147–153. Graham, F.L., Smiley, J., Russell, W.C., and Nairn, R. (1977). Characteristics of a human cell line transformed by DNA from human adenovirus type-5. J. Gen. Virol. 36 (1): 59–72. Dumont, J., Euwart, D., Mei, B. et al. (2016). Human cell lines for biopharmaceutical manufacturing: history, status, and future perspectives. Crit. Rev. Biotechnol. 36 (6): 1110–1122. Dubridge, R.B., Tang, P., Hsia, H.C. et al. (1987). Analysis of mutation in human-cells by using an Epstein-Barr-virus shuttle system. Mol. Cell. Biol. 7 (1): 379–387. Young, J.M., Cheadle, C., Foulke, J.S. Jr., et al. (1988). Utilization of an Epstein-Barr virus replicon as a eukaryotic expression vector. Gene 62 (2): 171–185. Cho, M.S., Yee, H., and Chan, S. (2002). Establishment of a human somatic hybrid cell line for recombinant protein production. J. Biomed. Sci. 9 (6 Pt. 2): 631–638. Lee, H.J., Chang, M., Kim, J.M. et al. (2013). Application of a new human cell line, F2N78, in the transient and stable production of recombinant therapeutics. Biotechnol. Progr. 29 (2): 432–440. Fallaux, F.J., Bout, A., Van der Velde, I. et al. (1998). New helper cells and matched early region 1-deleted adenovirus vectors prevent generation of replication-competent adenoviruses. Hum. Gene Ther. 9 (13): 1909–1917. Jones, D., Kroos, N., Anema, R. et al. (2003). High-level expression of recombinant IgG in the human cell line PER.C6. Biotechnol. Progr. 19 (1): 163–168. Yallop, C., Crowley, J., Cote, J. et al. (2008). PER.C6 cells for the manufacture of biopharmaceutical proteins. In: Modern Biopharmaceuticals (ed. J. Knablein), 779–807. Weinheim: Wiley-VCH. Schiedner, G., Hertel, S., Bialek, C. et al. (2008). Efficient and reproducible generation of high-expressing, stable human cell lines without need for antibiotic selection. BMC Biotechnol. 8: 13.

®

39

40

2 Cell Line Development for Therapeutic Protein Production

22 Shulman, M., Wilde, C.D., and Kohler, G. (1978). A better cell line for mak-

ing hybridomas secreting specific antibodies. Nature 276 (5685): 269–270. 23 Barnes, L.M., Bentley, C.M., and Dickson, A.J. (2000). Advances in animal

24

25

26 27

28

29

30

31

32

33

34

35 36

cell recombinant protein production: GS-NS0 expression system. Cytotechnology 32 (2): 109–123. Chung, C.H., Mirakhur, B., Chan, E. et al. (2008). Cetuximab-induced anaphylaxis and IgE specific for galactose-alpha-1,3-galactose. N. Engl. J. Med. 358 (11): 1109–1117. Soukharev, S., Hammond, D., Ananyeva, N.M. et al. (2002). Expression of factor VIII in recombinant and transgenic systems. Blood Cells Mol. Dis. 28 (2): 234–248. Blasey, H.D., Aubry, J.P., Mazzei, G.J., and Bernard, A.R. (1995). Large scale transient expression with COS cells. Cytotechnology 18 (3): 183–192. Wurm, F. and Bernard, A. (1999). Large-scale transient expression in mammalian cells for recombinant protein production. Curr. Opin. Biotechnol. 10 (2): 156–159. Daramola, O., Stevenson, J., Dean, G. et al. (2014). A high-yielding CHO transient system: coexpression of genes encoding EBNA-1 and GS enhances transient protein expression. Biotechnol. Progr. 30 (1): 132–141. Rajendra, Y., Hougland, M.D., Alam, R. et al. (2015). A high cell density transient transfection system for therapeutic protein expression based on a CHO GS-knockout cell line: process development and product quality assessment. Biotechnol. Bioeng. 112 (5): 977–986. Steger, K., Brady, J., Wang, W. et al. (2015). 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. Screening 20 (4): 545–551. Bebbington, C.R., Renner, G., Thomson, S. et al. (1992). High-level expression of a recombinant antibody from myeloma cells using a glutamine synthetase gene as an amplifiable selectable marker. Nat. Biotechnol. 10 (2): 169–175. Brown, M.E., Renner, G., Field, R.P., and Hassell, T. (1992). Process development for the production of recombinant antibodies using the glutamine synthetase (GS) system. Cytotechnology 9 (1–3): 231–236. Cockett, M.I., Bebbington, C.R., and Yarranton, G.T. (1990). High-level expression of tissue inhibitor of metalloproteinases in Chinese hamster ovary cells using glutamine-synthetase gene amplification. Nat. Biotechnol. 8 (7): 662–667. Fan, L.C., Kadura, I., Krebs, L.E. et al. (2012). Improving the efficiency of CHO cell line generation using glutamine synthetase gene knockout cells. Biotechnol. Bioeng. 109 (4): 1007–1015. Wurm, F.M. (2004). Production of recombinant protein therapeutics in cultivated mammalian cells. Nat. Biotechnol. 22 (11): 1393–1398. Chen, C. and Chasin, L.A. (1998). Cointegration of DNA molecules introduced into mammalian cells by electroporation. Somat. Cell Mol. Genet. 24 (4): 249–256.

References

37 Jalah, R., Rosati, M., Kulkarni, V. et al. (2007). Efficient systemic expres-

38

39

40

41 42

43

44 45 46

47

48

49

50

51

sion of bioactive IL-15 in mice upon delivery of optimized DNA expression plasmids. DNA Cell Biol. 26 (12): 827–840. Kozak, M. (1987). At least six nucleotides preceding the AUG initiator codon enhance translation in mammalian cells. J. Mol. Biol. 196 (4): 947–950. Zufferey, R., Donello, J.E., Trono, D., and Hope, T.J. (1999). Woodchuck hepatitis virus posttranscriptional regulatory element enhances expression of transgenes delivered by retroviral vectors. J. Virol. 73 (4): 2886–2892. Kober, L., Zehe, C., and Bode, J. (2013). Optimized signal peptides for the development of high expressing CHO cell lines. Biotechnol. Bioeng. 110 (4): 1164–1173. Gustafsson, C., Govindarajan, S., and Minshull, J. (2004). Codon bias and heterologous protein expression. Trends Biotechnol. 22 (7): 346–353. Hung, F., Deng, L., Ravnikar, P. et al. (2010). mRNA stability and antibody production in CHO cells: improvement through gene optimization. Biotechnol. J. 5 (4): 393–401. Chung, B.K., Yusufi, F.N., Mariati et al. (2013). Enhanced expression of codon optimized interferon gamma in CHO cells. J. Biotechnol. 167 (3): 326–333. Kim, T.K. and Eberwine, J.H. (2010). Mammalian cell transfection: the present and the future. Anal. Bioanal. Chem. 397 (8): 3173–3178. Nayerossadat, N., Maedeh, T., and Ali, P.A. (2012). Viral and nonviral delivery systems for gene delivery. Adv. Biomed. Res. 1: 27. Jayapal, K.P., Wlaschin, K.F., Hu, W.S., and Yap, M.G. (2007). Recombinant protein therapeutics from CHO cells—20 years and counting. Chem. Eng. Prog. 103: 40–47. Milbrandt, J.D., Heintz, N.H., White, W.C. et al. (1981). Methotrexateresistant Chinese-hamster ovary cells have amplified a 135-kilobase-pair region that includes the dihydrofolate-reductase gene. Proc. Natl. Acad. Sci. U.S.A. 78 (10): 6043–6047. Assaraf, Y.G. and Schimke, R.T. (1987). Identification of methotrexate transport deficiency in mammalian-cells using fluoresceinated methotrexate and flow-cytometry. Proc. Natl. Acad. Sci. U.S.A. 84 (20): 7154–7158. Kim, S.J., Kim, N.S., Ryu, C.J. et al. (1998). Characterization of chimeric antibody producing CHO cells in the course of dihydrofolate reductase-mediated gene amplification and their stability in the absence of selective pressure. Biotechnol. Bioeng. 58 (1): 73–84. Kalwy, S., Porter, A., Watson, C., and Gay, R. (2004). Rapid selection of high yielding GS-CHO cell lines using the GS expression system in a protein-free, fully chemically defined, animal component-free process. Proceedings of The 227th ACS National Meeting (March 28–April 1 2004), Anaheim, CA. Song, M., Raphaelli, K., Jones, M.L. et al. (2011). Clonal selection of high producing, stably transfected HEK293 cell lines utilizing modified, high-throughput FACS screening. J. Chem. Technol. Biotechnol. 86 (7): 935–941.

41

42

2 Cell Line Development for Therapeutic Protein Production

52 Borth, N., Zeyda, M., Kunert, R., and Katinger, H. (2000). Efficient selec-

53

54

55

56

57

58

59

60

61

62

63

64

tion of high-producing subclones during gene amplification of recombinant Chinese hamster ovary cells by flow cytometry and cell sorting. Biotechnol. Bioeng. 71 (4): 266–273. DeMaria, C.T., Cairns, V., Schwarz, C. et al. (2007). Accelerated clone selection for recombinant CHO cells using a FACS-based high-throughput screen. Biotechnol. Progr. 23 (2): 465–472. Nakamura, T. and Omasa, T. (2015). Optimization of cell line development in the GS-CHO expression system using a high-throughput, single cell-based clone selection system. J. Biosci. Bioeng. 120 (3): 323–329. Dharshanan, S., Chong, H., Hung, C.S. et al. (2011). Rapid automated selection of mammalian cell line secreting high level of humanized monoclonal antibody using Clone Pix FL system and the correlation between exterior median intensity and antibody productivity. Electron. J. Biotechnol. 14 (2). Cresswell, J.R., Lin, N., Richardson, G.A., and Kayser, K.J. (2009). Cell XpressTM applications in development and characterization of biopharmaceutical recombinant protein producing cell lines. In: Cell Line Development (ed. M. Al-Rubeai), 109–126. Netherlands, Dordrecht: Springer. Jun, S.C., Kim, M.S., Baik, J.Y. et al. (2005). Selection strategies for the establishment of recombinant Chinese hamster ovary cell line with dihydrofolate reductase-mediated gene amplification. Appl. Microbiol. Biotechnol. 69 (2): 162–169. Yoon, S.K., Kim, S.H., and Lee, G.M. (2003). Effect of low culture temperature on specific productivity and transcription level of anti-4-1BB antibody in recombinant Chinese hamster ovary cells. Biotechnol. Progr. 19 (4): 1383–1386. Yoon, S.K., Choi, S.L., Song, J.Y., and Lee, G.M. (2005). Effect of culture pH on erythropoietin production by Chinese hamster ovary cells grown in suspension at 32.5 and 37.0 degrees C. Biotechnol. Bioeng. 89 (3): 345–356. Oguchi, S., Saito, H., Tsukahara, M., and Tsumura, H. (2006). pH Condition in temperature shift cultivation enhances cell longevity and specific hMab productivity in CHO culture. Cytotechnology 52 (3): 199–207. Becerra, S., Berrios, J., Osses, N., and Altamirano, C. (2012). Exploring the effect of mild hypothermia on CHO cell productivity. Biochem. Eng. J. 60: 1–8. Restelli, V., Wang, M.D., Huzel, N. et al. (2006). The effect of dissolved oxygen on the production and the glycosylation profile of recombinant human erythropoietin produced from CHO cells. Biotechnol. Bioeng. 94 (3): 481–494. Zhu, M.M., Goyal, A., Rank, D.L. et al. (2005). Effects of elevated pCO2 and osmolality on growth of CHO cells and production of antibody-fusion protein B1: a case study. Biotechnol. Progr. 21 (1): 70–77. Han, Y.K., Koo, T.Y., and Lee, G.M. (2009). Enhanced interferon-beta production by CHO cells through elevated osmolality and reduced culture temperature. Biotechnol. Progr. 25 (5): 1440–1447.

References

65 Kim, J.Y., Kim, Y.G., and Lee, G.M. (2012). CHO cells in biotechnology for

66 67 68 69

70

71

72

73

74

75

76 77

78

79 80

81

production of recombinant proteins: current state and further potential. Appl. Microbiol. Biotechnol. 93 (3): 917–930. Wang, Z., Wei, G.H., Liu, D.P., and Liang, C.C. (2007). Unravelling the world of cis-regulatory elements. Med. Biol. Eng. Comput. 45 (8): 709–718. Kim, J.K., Samaranayake, M., and Pradhan, S. (2009). Epigenetic mechanisms in mammals. Cell. Mol. Life Sci. 66 (4): 596–612. Shen, Y., Yue, F., McCleary, D.F. et al. (2012). A map of the cis-regulatory sequences in the mouse genome. Nature 488 (7409): 116–120. Kito, M., Itami, S., Fukano, Y. et al. (2002). Construction of engineered CHO strains for high-level production of recombinant proteins. Appl. Microbiol. Biotechnol. 60 (4): 442–448. Kameyama, Y., Kawabe, Y., Ito, A., and Kamihira, M. (2010). An accumulative site-specific gene integration system using Cre recombinase-mediated cassette exchange. Biotechnol. Bioeng. 105 (6): 1106–1114. Inao, T., Kawabe, Y., Yamashiro, T. et al. (2015). Improved transgene integration into the Chinese hamster ovary cell genome using the Cre-loxP system. J. Biosci. Bioeng. 120 (1): 99–106. Kawabe, Y., Inao, T., Komatsu, S. et al. (2017). Improved recombinant antibody production by CHO cells using a production enhancer DNA element with repeated transgene integration at a predetermined chromosomal site. J. Biosci. Bioeng. 123 (3): 390–397. Kim, M.S. and Lee, G.M. (2008). Use of Flp-mediated cassette exchange in the development of a CHO cell line stably producing erythropoietin. J. Microbiol. Biotechnol. 18 (7): 1342–1351. Ahmadi, M., Mahboudi, F., Akbari Eidgahi, M.R. et al. (2016). Evaluating the efficiency of phiC31 integrase-mediated monoclonal antibody expression in CHO cells. Biotechnol. Progr. 32 (6): 1570–1576. Hacker, D.L. and Balasubramanian, S. (2016). Recombinant protein production from stable mammalian cell lines and pools. Curr. Opin. Struct. Biol. 38: 129–136. Carroll, D. (2014). Genome engineering with targetable nucleases. Annu. Rev. Biochem. 83: 409–439. Bitinaite, J., Wah, D.A., Aggarwal, A.K., and Schildkraut, I. (1998). Fok I dimerization is required for DNA cleavage. Proc. Natl. Acad. Sci. U.S.A. 95 (18): 10570–10575. Christian, M., Cermak, T., Doyle, E.L. et al. (2010). Targeting DNA double-strand breaks with TAL effector nucleases. Genetics 186 (2): 757–761. Boch, J., Scholze, H., Schornack, S. et al. (2009). Breaking the code of DNA binding specificity of TAL-type III effectors. Science 326 (5959): 1509–1512. Cristea, S., Freyvert, Y., Santiago, Y. et al. (2013). In vivo cleavage of transgene donors promotes nuclease-mediated targeted integration. Biotechnol. Bioeng. 110 (3): 871–880. Jinek, M., Chylinski, K., Fonfara, I. et al. (2012). A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity. Science 337 (6096): 816–821.

43

44

2 Cell Line Development for Therapeutic Protein Production

82 Pennisi, E. (2013). The CRISPR craze. Science 341 (6148): 833–836. 83 Kim, H. and Kim, J.S. (2014). A guide to genome engineering with pro-

grammable nucleases. Nat. Rev. Genet. 15 (5): 321–334. 84 Chen, K. and Gao, C. (2014). Targeted genome modification technolo-

85 86

87

88 89 90

91

92

93

94

95

96

97

98

gies and their applications in crop improvements. Plant Cell Rep. 33 (4): 575–583. Falahi, F., Sgro, A., and Blancafort, P. (2015). Epigenome engineering in cancer: fairytale or a realistic path to the clinic? Front. Oncol. 5: 22. Nakade, S., Tsubota, T., Sakane, Y. et al. (2014). Microhomology-mediated end-joining-dependent integration of donor DNA in cells and animals using TALENs and CRISPR/Cas9. Nat. Commun. 5: 5560. Lee, J.S., Kallehauge, T.B., Pedersen, L.E., and Kildegaard, H.F. (2015). Site-specific integration in CHO cells mediated by CRISPR/Cas9 and homology-directed DNA repair pathway. Sci. Rep. 5: 8572. Miskey, C., Izsvák, Z., Kawakami, K., and Ivics, Z. (2005). DNA transposons in vertebrate functional genomics. Cell. Mol. Life Sci. 62 (6): 629–641. Ivics, Z., Li, M.A., Mátés, L. et al. (2009). Transposon-mediated genome manipulation in vertebrates. Nat. Methods 6 (6): 415–422. Fonager, J., Franke-Fayard, B.M., Adams, J.H. et al. (2011). Development of the piggyBac transposable system for Plasmodium berghei and its application for random mutagenesis in malaria parasites. BMC Genomics 12 (155): 1–17. Wu, S.C., Meir, Y.J., Coates, C.J. et al. (2006). piggyBac is a flexible and highly active transposon as compared to Sleeping beauty, Tol2, and Mos1 in mammalian cells. Proc. Natl. Acad. Sci. U.S.A. 103 (41): 15008–15013. Huang, X., Guo, H., Tammana, S. et al. (2010). Gene transfer efficiency and genome-wide integration profiling of Sleeping Beauty, Tol2, and PiggyBac transposons in human primary T cells. Mol. Ther. 18 (10): 1803–1813. Galvan, D.L., Nakazawa, Y., Kaja, A. et al. (2009). Genome-wide mapping of PiggyBac transposon integrations in primary human T cells. J. Immunother. 32 (8): 837–844. Matasci, M., Baldi, L., Hacker, D.L., and Wurm, F.M. (2011). The PiggyBac transposon enhances the frequency of CHO stable cell line generation and yields recombinant lines with superior productivity and stability. Biotechnol. Bioeng. 108 (9): 2141–2150. Rajendra, Y., Peery, R.B., and Barnard, G.C. (2016). Generation of stable Chinese hamster ovary pools yielding antibody titers of up to 7.6 g/L using the piggyBac transposon system. Biotechnol. Progr. 32 (5): 1301–1307. Rajendra, Y., Balasubramanian, S., Peery, R.B. et al. (2017). Bioreactor scale up and protein product quality characterization of PiggyBac transposon derived CHO pools. Biotechnol. Progr. 33 (2): 534–540. Ghaderi, D., Zhang, M., Hurtado-Ziola, N., and Varki, A. (2012). Production platforms for biotherapeutic glycoproteins. Occurrence, impact, and challenges of non-human sialylation. Biotechnol. Genet. Eng. Rev. 28: 147–175. Grabenhorst, E., Schlenke, P., Pohl, S. et al. (1999). Genetic engineering of recombinant glycoproteins and the glycosylation pathway in mammalian host cells. Glycoconjugate J. 16 (2): 81–97.

References

99 Berkner, K.L. (1993). Expression of recombinant vitamin K-dependent

100

101

102

103

104

105

106

107 108

109

110

111

112

113

proteins in mammalian cells: factors IX and VII. Methods Enzymol. 222: 450–477. Oh, N., Kim, K., Kim, S.J. et al. (2015). A designed angiopoietin-1 variant, dimeric CMP-Ang1 activates Tie2 and stimulates angiogenesis and vascular stabilization in N-glycan dependent manner. Sci. Rep. 5: 15291. Hacker, D.L., Kiseljak, D., Rajendra, Y. et al. (2013). Polyethyleneimine-based transient gene expression processes for suspension-adapted HEK-293E and CHO-DG44 cells. Protein Expression Purif. 92 (1): 67–76. Nallet, S., Fornelli, L., Schmitt, S. et al. (2012). Glycan variability on a recombinant IgG antibody transiently produced in HEK-293E cells. New Biotechnol. 29 (4): 471–476. Loignon, M., Perret, S., Kelly, J. et al. (2008). Stable high volumetric production of glycosylated human recombinant IFNalpha2b in HEK293 cells. BMC Biotechnol. 8: 65. Yu, D.Y., Noh, S.M., and Lee, G.M. (2016). Limitations to the development of recombinant human embryonic kidney 293E cells using glutamine synthetase-mediated gene amplification: methionine sulfoximine resistance. J. Biotechnol. 231: 136–140. Seo, J.S., Min, B.S., Kwon, Y.B. et al. (2016). Characteristics of human cell line, F2N78, for the production of recombinant antibody in fed-batch and perfusion cultures. J. Biosci. Bioeng. 121 (3): 317–324. Ross, D., Brown, T., Harper, R. et al. (2012). Production and characterization of a novel human recombinant alpha-1-antitrypsin in PER.C6 cells. J. Biotechnol. 162 (2–3): 262–273. Tchoudakova, A., Hensel, F., Murillo, A. et al. (2009). High level expression of functional human IgMs in human PER.C6 cells. MAbs 1 (2): 163–171. Essers, R., Kewes, H., and Schiedner, G. (2011). Improving volumetric productivity of a stable human CAP cell line by bioprocess optimization. BMC Proc. 5 (Suppl. 8): P66. Kim, N.S., Kim, S.J., and Lee, G.M. (1998). Clonal variability within dihydrofolate reductase-mediated gene amplified Chinese hamster ovary cells: stability in the absence of selective pressure. Biotechnol. Bioeng. 60 (6): 679–688. Barnes, L.M., Moy, N., and Dickson, A.J. (2006). Phenotypic variation during cloning procedures: analysis of the growth behavior of clonal cell lines. Biotechnol. Bioeng. 94 (3): 530–537. Dahodwala, H., Nowey, M., Mitina, T., and Sharfstein, S.T. (2012). Effects of clonal variation on growth, metabolism, and productivity in response to trophic factor stimulation: a study of Chinese hamster ovary cells producing a recombinant monoclonal antibody. Cytotechnology 64 (1): 27–41. Kim, H.S. and Lee, G.M. (2007). Differences in optimal pH and temperature for cell growth and antibody production between two Chinese hamster ovary clones derived from the same parental clone. J. Microbiol. Biotechnol. 17 (5): 712–720. Evans, K., Albanetti, T., Venkat, R. et al. (2015). Assurance of monoclonality in one round of cloning through cell sorting for single cell deposition coupled with high resolution cell imaging. Biotechnol. Progr. 31 (5): 1172–1178.

45

46

2 Cell Line Development for Therapeutic Protein Production

114 Frye, C., Deshpande, R., Estes, S. et al. (2016). Industry view on the rel-

115 116

117

118

119

120

121

122

123

124

125

126

127

ative importance of “clonality” of biopharmaceutical-producing cell lines. Biologicals 44 (2): 117–122. Wurm, F. and Wurm, M. (2017). Cloning of CHO cells, productivity and genetic stability—a discussion. Processes 5 (2): 20. Fann, C.H., Guirgis, F., Chen, G. et al. (2000). Limitations to the amplification and stability of human tissue-type plasminogen activator expression by Chinese hamster ovary cells. Biotechnol. Bioeng. 69 (2): 204–212. Hammill, L., Welles, J., and Carson, G.R. (2000). The gel microdrop secretion assay: Identification of a low productivity subpopulation arising during the production of human antibody in CHO cells. Cytotechnology 34 (1): 27–37. Beckmann, T.F., Kramer, O., Klausing, S. et al. (2012). Effects of high passage cultivation on CHO cells: a global analysis. Appl. Microbiol. Biotechnol. 94 (3): 659–671. Chusainow, J., Yang, Y.S., Yeo, Y.H.M. et al. (2009). A study of monoclonal antibody-producing CHO cell lines: what makes a stable high producer? Biotechnol. Bioeng. 102 (4): 1182–1196. Li, H.W., Chen, K.M., Wang, Z. et al. (2016). Genetic analysis of the clonal stability of Chinese hamster ovary cells for recombinant protein production. Mol. Biosyst. 12 (1): 102–109. Yang, Y.S., Mariati, Chusainow, J., and Yap, M.G.S. (2010). DNA methylation contributes to loss in productivity of monoclonal antibody-producing CHO cell lines. J. Biotechnol. 147 (3–4): 180–185. Kim, M., O’Callaghan, P.M., Droms, K.A., and James, D.C. (2011). A mechanistic understanding of production instability in CHO cell lines expressing recombinant monoclonal antibodies. Biotechnol. Bioeng. 108 (10): 2434–2446. Bailey, L.A., Hatton, D., Field, R., and Dickson, A.J. (2012). Determination of Chinese hamster ovary cell line stability and recombinant antibody expression during long-term culture. Biotechnol. Bioeng. 109 (8): 2093–2103. Pallavicini, M.G., Deteresa, P.S., Rosette, C. et al. (1990). Effects of methotrexate on transfected DNA stability in mammalian-cells. Mol. Cell. Biol. 10 (1): 401–404. Chen, F., Fan, L., Wang, J.Q. et al. (2012). Insight into the roles of hypoxanthine and thydimine on cultivating antibody-producing CHO cells: cell growth, antibody production and long-term stability. Appl. Microbiol. Biotechnol. 93 (1): 169–178. Jun, S.C., Kim, M.S., Hong, H.J., and Lee, G.M. (2006). Limitations to the development of humanized antibody producing Chinese hamster ovary cells using glutamine synthetase-mediated gene amplification. Biotechnol. Progr. 22 (3): 770–780. Dorai, H., Corisdeo, S., Ellis, D. et al. (2012). Early prediction of instability of Chinese hamster ovary cell lines expressing recombinant antibodies and antibody-fusion proteins. Biotechnol. Bioeng. 109 (4): 1016–1030.

References

128 Yoshikawa, T., Nakanishi, F., Itami, S. et al. (2000). Evaluation of stable and

129

130

131

132

133

134

highly productive gene amplified CHO cell line based on the location of amplified genes. Cytotechnology 33 (1–3): 37–46. Cao, Y.H., Kimura, S., Itoi, T. et al. (2012). Construction of BAC-based physical map and analysis of chromosome rearrangement in chinese hamster ovary cell lines. Biotechnol. Bioeng. 109 (6): 1357–1367. Ritter, A., Nuciforo, S., Schulze, A. et al. (2017). Fam60A plays a role for production stabilities of recombinant CHO cell lines. Biotechnol. Bioeng. 114 (3): 701–704. Xie, H., Chakraborty, A., Ahn, J. et al. (2010). Rapid comparison of a candidate biosimilar to an innovator monoclonal antibody with advanced liquid chromatography and mass spectrometry technologies. MAbs 2 (4): 379–394. Primack, J., Flynn, G.C., and Pan, H. (2011). A high-throughput microchip-based glycan screening assay for antibody cell culture samples. Electrophoresis 32 (10): 1129–1132. Markely, L.R., Ong, B.T., Hoi, K.M. et al. (2010). A high-throughput method for quantification of glycoprotein sialylation. Anal. Biochem. 407 (1): 128–133. Zeck, A., Regula, J.T., Larraillet, V. et al. (2012). Low level sequence variant analysis of recombinant proteins: an optimized approach. PLoS One 7 (7): e40328.

47

49

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells Joo-Hyoung Lee 1,2 , Henning G. Hansen 3 , Sun-Hye Park 1,4 , Jong-Ho Park 1,2 , and Yeon-Gu Kim 1,4 1 Korea Research Institute of Bioscience and Biotechnology (KRIBB), Biotherapeutics Translational Research Center, 125 Gwahak-ro, Yuseong-gu, KRIBB : 34141 Daejeon, Republic of Korea 2 Korea Advanced Institute of Science and Technology (KAIST), Department of Biological Sciences, 335 Gwahak-ro, Yuseong-gu, KAIST : 34141 Daejeon, Republic of Korea 3 Technical University of Denmark, The Novo Nordisk Foundation Center for Biosustainability, Building 220, Kemitorvet, DK-2800 Kgs. Lyngby, Denmark 4 Korea University of Science and Technology (UST), Department of Bioprocess Engineering, 217 Gajeong-ro, Yuseong-gu, UST : 34113 Daejeon, Republic of Korea

3.1 Introduction In 1986, a human tissue plasminogen activator (Genentech), derived from recombinant mammalian cells, was approved as the first therapeutic protein [1]. Subsequently, recombinant mammalian cells became the workhorse of the biopharmaceutical industry for therapeutic glycoprotein production [2]. Currently, recombinant Chinese hamster ovary (CHO) cells are the most frequently used cells for therapeutic glycoprotein production. Since 1986, protein titers for CHO cell culture have increased more than 100-fold, reaching 10 g/l in the past two decades [3]. This increase in foreign protein production is primarily attributable to the establishment of stable cell lines with high protein productivity as well as optimization of cell line-specific media [3, 4]. To screen for the functionality and manufacturability of a multitude of therapeutic protein candidates, a rapid, robust, and cost-effective production platform is required. Despite stable gene expression being predominant in the large-scale commercial production of recombinant proteins, stable cell line construction is time-consuming and labor-intensive [5]. As a cost-efficient and higher throughput alternative, recombinant proteins are transiently produced primarily in human embryonic kidney 293 (HEK293) and CHO cells [6]. The principle of transient gene expression (TGE) is that transgene-harboring plasmids transfected into cell nuclei are not stably integrated into the genome but rather transcribed as nonintegrated plasmids. This type of expression is transient, as the introduced plasmids are not replicated and get diluted over

Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

50

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

time as a consequence of cell division and degradation. TGE has proven to be an efficient production platform for producing milligram to gram quantities of protein within a few days to two weeks [7]. Recombinant proteins produced by TGE can be utilized for studying the manufacturability, toxicity, stability, specificity, and activity of the product. In this chapter, we describe previous and current strategies for improving product titers in TGE-based protein production platforms in mammalian cells.

3.2 Gene Delivery: Transient Transfection Methods In TGE-based protein production platforms, optimizing the efficiency of gene delivery into mammalian cells is vital because this parameter directly correlates with specific protein productivity (qp ). There are two general methods of gene delivery: viral- and non-viral-based. Viral-based gene delivery (transduction) is available for a wide range of mammalian cells, with typically higher gene delivery efficiency than that of non-viral-based methods. However, viral-based gene delivery is associated with a significant biosafety risk and requires a relatively complicated protocol, including viral manipulation and viral particle removal to achieve high-yield therapeutic protein production [8, 9]. Although viral-based gene delivery involves vector elements being packaged into virus particles and transduced into cells, non-viral-based gene delivery is primarily based on altering the physiochemical properties and environment of the plasmid to favor its cellular uptake [10]. Although non-viral-based gene delivery has low gene delivery efficiency, it has numerous advantages compared to viral-based gene delivery, including cost-efficient and high-throughput plasmid construction, amplification, and purification; inexpensive transfection methods; and low viral contamination risk [9, 11, 12]. Thus, non-viral-based gene delivery is typically used to evaluate the performance of new candidate therapeutic proteins in a rapid and high-throughput manner. The following section describes the most commonly used non-viral-based transfection methods for mammalian cells. Table 3.1 shows examples of methods used for TGE-based recombinant protein production. 3.2.1

Calcium Phosphate-Based Transient Transfection

In the early 1970s, calcium phosphate transfection was one of the first transfection methods to facilitate non-viral-based gene delivery [37]. The mechanism of calcium phosphate-based transfection involves the formation of complexes between positively charged calcium phosphate and negatively charged nucleic acids, which are then internalized cells through endocytosis and phagocytosis [38]. Although the reagents are inexpensive and easily available, calcium phosphate-based transfection is prone to highly variable transfection efficiency because of its sensitivity to small change in pH. In addition, the buffer used can be

3.2 Gene Delivery: Transient Transfection Methods

Table 3.1 Transient transfection reagents commonly used for TGE-based recombinant protein production. Manufacturer/ distributor

Transfection methods

Specific characteristics

References

Branched PEI

Sigma

PEI

ND

[13]

PEI, linear (MW 25 000)

Polysciences

PEI

ND

[14–25]

PEI-MAX (MW 40 000)

Polysciences

PEI

ND

[20, 26, 27]

PEIpro

Polyplus

PEI

Large-scale TGE

[20]

Lipofectamine 2000

Life Technologies

Cationic lipid

ND

[26, 28–34]

Lipofectamine 3000

Life Technologies

Cationic lipid

ND

[32]

DMRIE-C reagent

Life Technologies

Cationic lipid

ND

[34]

293fectin

Life Technologies

Cationic lipid

ND

[35]

FreeStyle MAX Reagent

Life Technologies

Cationic lipid

Large-scale transient transfection

[6, 26, 36]

ExpiFectamine 293

Life Technologies

Cationic lipid

Optimized for Expi293F cells

[36]

ExpiFectamine CHO

Life Technologies

Cationic lipid

Optimized for ExpiCHO-S cells

[36]

FuGENE-HD

Promega

Nonliposomal formulation

Optimized for transient transfection

[32]

Name

Abbreviations: ND, not described; DMRIE-C, 1,2-dimyristyloxypropyl-3-dimethyl-hydroxy ethyl ammonium bromide/cholesterol.

cytotoxic and therefore must be removed by inconvenient centrifugation-based washing of the suspended cells [39]. 3.2.2

Electroporation

In electroporation-based transfection, an electrical pulse is applied to create temporary pores in the cell membrane through which plasmids can enter the cells [40]. Consequently, electroporation does not involve plasmid delivery vehicles (reagents), which allows for transfected cell culture without the presence of undesired foreign materials and process-derived impurities. One of the main disadvantages of cuvette-based electroporation platforms is the lack of scalability. Cytotoxicity resulting from a high-voltage pulse and unstable membranes can also be a significant drawback. Thus, the optimization of electroporation conditions is vital for transient transfection efficiency and cell viability. For transient

51

52

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

transfection in mammalian cells, several kinds of electroporation systems have been used: MaxCyte STX Scalable Transfection System (MaxCyte) [41], Bio-Rad Gene Pulser (Bio-Rad) [42], Amaxa Nucleofector I device (Lonza) [43], and Neon Transfection System (Life Technologies/Invitrogen) [44]. A recent study reported that flow electroporation partially resolves the issue of scalability and optimization of electroporation conditions to minimize cytotoxic effect while maintaining high transfection efficiency [41]. Consequently, electroporation is one of the efficient TGE strategies presently used for protein production.

3.2.3

Polyethylenimine-Based Transient Transfection

Polyethylenimine (PEI) is one of the most widely used non-viral-based transient transfection methods for large-scale protein production in mammalian cells. PEI can be synthesized with various molecular weights as either branched or linear structures, resulting in positively charged particles termed polyplexes [45, 46]. PEI binds nucleic acids, forming PEI/nucleic acid complexes, which are then internalized through endocytosis and deposited in the cytoplasm [47]. Despite low cost and efficient gene delivery in the use of PEI, PEI-based transient transfection requires correct balance between adequate reagent and DNA amount for reducing detrimental effect on the cell viability such as cytotoxicity, cell dysfunction, and apoptosis [46, 47]. PEI-based transient transfection was initially optimized in HEK293 cells with a focus on obtaining high reproducibility with low cytotoxic effects and high transfection efficiency [14–20]. Subsequently, PEI-based TGE methods have been developed for CHO cells using linear PEI [19, 21–25], branched PEI [13], and PEI-MAX [20, 26, 27]. A few attempts have been made to use cationic polymers as alternatives to PEI, including diethanolamine, chitosan, poly-l-lysine, and linear β-cyclodextrin-containing polymers [48–52]. However, because of the low transfection efficiency of these polymers compared to PEI, they have not been widely used. 3.2.4

Liposome-Based Transient Transfection

Liposome-based transfection, also termed as lipofection, was first reported in 1987 and has been used extensively since then for delivering genes into mammalian cells [53]. The mechanism of liposome-based transfection involves association of the cationic lipids with nucleic acids, forming lipoflexes that are then internalized by endocytosis, resulting in import into the cytoplasm and nucleus [54]. Liposome-based transfection reagents are available for a wide range of mammalian cells. Thus, optimization for a cell line of interest is typically not needed. In general, liposome-based transfection reagents are relatively expensive compared to those for PEI-based transfection. Moreover, liposome-based transfection is typically associated with a substantial accumulation of particles in the cytoplasm, increasing their cytotoxic effect [55]. However, under optimal conditions, liposome-based transfection provides scalability, high transfection efficiency, and high protein productivity. Consequently, liposome-based transfection is currently one of the most efficient methods for the TGE-based

3.3 Expression Vectors

production of therapeutic proteins [7], and several liposome-based transfection reagents are commercially available for HEK293 cell- and CHO cell-based TGE platforms (Table 3.1).

3.3 Expression Vectors 3.3.1

Expression Vector Composition and Preparation

The design as well as the amplification and purification of the expression vector to be used for TGE-based recombinant protein production in mammalian cells affect the qp . For example, the number of plasmids transfected into the cell as well as the number of transcripts expressed per plasmid both depend on the vector composition. For all reagent-based transfection methods, the transfection efficiency depends on the amount of vector as well as the ratio of vector to transfection reagent. Equally important, the transfection efficiency depends on the purity and vector integrity of the plasmid preparation. In addition, reducing the vector size (length) has been shown to improve cellular and nuclear uptake during transient transfection [56]. As a result, removing unnecessary elements such as eukaryotic selection markers increases the efficiency of gene delivery [57]. Transcriptional and translational rates as well as mRNA stability mainly depend on regulatory DNA elements up- and downstream of the coding sequence. To ensure high transcriptional rates, a strong viral promoter, such as the human or mouse cytomegalovirus (CMV)-derived promoter, is commonly used for TGE-based protein production [58–62]. Alternatively, strong nonviral promoters have also been successfully used, including those derived from human elongation factor-1α (EF-1α), chicken α-actin, mouse phosphoglycerate kinase (PGK), and human ubiquitin C [58, 63]. Protein production in TGE-based platforms can also be increased by optimizing polyadenylation signals [58] and posttranscriptional regulatory elements to increase mRNA stability, nuclear export rates, and translation rates [16, 64–66]. When upscaling a TGE-based protein production platform, the plasmid preparation yield becomes a cost issue. Typically, transfection methods require about 1 mg of plasmid per liter of cell culture. To ensure high plasmid yields in Escherichia coli preparations, the vectors used for TGE-based protein production generally have a high copy number origin of replication, i.e. an ultra-high copy-number variant of the pUC19 origin of replication [7]. 3.3.2

Episomal Replication

In conventional TGE-based protein production platform, nonreplicating plasmids (episomes) are eventually lost through cell division and/or degradation. As an alternative, the transgene can be expressed from replication-proficient episomes, leading to long-term episomal persistence in mammalian cells [67, 68]. The most common examples include Simian virus 40 (SV40)-derived SV40 large T antigen (SV40LT), Epstein–Barr virus (EBV)-derived EBV nuclear antigen-1

53

54

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

(EBNA-1), and polyomavirus large T antigen (PyLT), used in combination with each of their viral origins [69]. Thus, these elements allow the episome to reside and replicate when the transfected cells pass through the S phase of the cell cycle [59, 70]. The SV40-based episomal replication system was the first reported virus-based replication-proficient system in mammalian cells [71]. During replication of the SV40 genome, the SV40LT is responsible for unwinding double-stranded DNA and initiating replication through direct binding to the SV40 origin of replication (SV40 ori) [72]. When an SV40 ori-harboring plasmid is transfected into an SV40LT-positive mammalian host cell line, the plasmid can undergo SV40LT-dependent episomal replication. Similarly, the replication components of EBV, including the origin of replication (oriP) and EBNA-1, enable episomal replication of the transfected plasmid in mammalian cells [73]. Finally, the polyoma virus Py origin of replication (Pyori) and PyLT are also used as a replication-proficient episomal expression system in mammalian cells [59, 74, 75]. All three of these episomal replication systems have been used for TGE-based protein production in mammalian cells [14, 76]. In addition, a number of episomal replication-proficient cell lines for TGE-based protein production have been reported, including HEK293 cells expressing a mutant version of the SV40LT (HEK293T cells), HEK293 cells expressing EBNA-1 (HEK293E cells), and CHO-K1 cells expressing PyLT (CHO-T cells) [28, 77–80]. 3.3.3

Coexpression Strategies

Transient coexpression of an effector gene and transgene encoding the recombinant protein products can be used to improve qp as well as volumetric productivity in mammalian cells [81]. Commonly used strategies included the coexpression of growth factor, human cell cycle regulatory proteins, antiapoptosis-associated proteins, unfolded protein response-related genes, inhibitors of histone deacetylases, and DNA methyltransferases [82–85]. In the following paragraph, nonexhaustive examples are described to touch upon the applicability of employing coexpression strategies. Transient coexpression of the fibroblast growth factor (FGF) gene together with an antibody-encoding transgene in both HEK293E and CHO-DG44 cells is reported to yield increases in antibody titers of 80% and 25%, respectively [86]. Moreover, transient coexpression of the p21 and p27 cell cycle inhibitors together with an antibody-encoding transgene in FreeStyle 293-F cells increased the antibody titer from approximately 40–100% by arresting cells during G1 phase [35]. Finally, transient coexpression of the antiapoptotic gene Bcl-xL together with an Fc-fusion protein-encoding transgene in CHO-S and HEK293E cells increased the product titer by approximately 100% by protecting cells from early apoptosis during cultivation [24].

3.4 Mammalian Cell Lines Therapeutic glycoproteins are generally produced using mammalian cell lines because these cells generate humanlike glycosylation [78]. For TGE-based protein

3.4 Mammalian Cell Lines

production, HEK293 and CHO cells are the preferred host cell lines [87]. These two host cell lines, along with alternatives, are described in this section. 3.4.1

HEK293 Cell-Based TGE Platforms

The human host cell line HEK293 is widely used for large-scale TGE-based protein production because of its easy adaptation to suspension cultivation in serum-free conditions, high transfection efficiency and qp , and the introduction of viral elements conferring episomal replication early [88, 89]. In an attempt to overcome the limitations of HEK293 cells in TGE, engineered HEK293 cells have been successfully established for transient systems over the past few decades [77, 78] (Table 3.2). Examples include 293N3S (adapted to suspension growth) [102], 293S (adapted to serum-free conditions) [103], HEK293E (stable expression of EBNA-1) [14], and HEK293T (stable expression of SV40LT) [93]. In recent work by Lee et al. (2017), EBNA-1-amplified HEK293 cells were successfully developed by employing dihydrofolate reductase (dhfr)-mediated gene amplification, thus showing the potential of the HEK293 cell-based TGE platform for therapeutic protein production [98]. Additionally, an N-acetylglucosaminyltransferase-I negative HEK293S mutant cell line was constructed for the production of restricted and homogeneous N-glycan formation [96]. More recently, the biopharmaceutical industry as well Table 3.2 HEK293-derived cells commonly used for TGE-based recombinant protein production.

Name

Characteristics

HEK293E

Culture conditions

Culture medium

References

Suspension Stably expressing EBNA-1 (ATCC CRL-10852 and Invitrogen)

EX-CELL 293 and Freestyle 293

[14, 88, 90]

293-6E

Stably expressing a truncated EBNA-1

Suspension

Serum-free F17 and [17, 91, 92] 293 Freestyle

HEK293T

Stably expressing SV40LT (ATCC CRL-1573/11268)

Adherent

DMEM with 8%–10% FBS

[93–95]

HEK293S GnTI−

Lacks GnTI activity (ATCC CRL-3022)

Adherent and suspension

DMEM with 10% FBS and FreeStyle F17

[96, 97]

EBNA-1-amplified Stably amplifying HEK293 EBNA-1

Suspension

Expi293 expression [98] medium

HKB-11

Hybrid cell line of HEK293 and B-cells

Suspension

MICT7.0 (Bayer proprietary medium)

FreeStyle 293-F

Derivative of wild-type Suspension HEK293 cells

FreeStyle 293 [6] expression medium

Expi293F

Derivative of FreeStyle293-F

Expi293 expression [36, 100, 101] medium

Suspension

[99]

55

56

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

as academic groups have started using nonengineered, commercially available HEK293-derived cell lines such as FreeStyle 293-F (Life Technologies) and Expi293F (Life Technologies), with product titers up to 1.2 g/l four days after transfection [36]. In contrast to CHO cells, large-scale protein production from stable HEK293 cell lines is not presently a preferred option because the efficiency of gene amplification for foreign protein production is insufficient [104]. Moreover, recombinant proteins produced in HEK293 cells might comprise posttranslational modifications that cannot be carried out by CHO cells [105]. Thus, the characteristics of a protein produced in HEK293 cells are not necessarily the same as that produced in CHO cells. Consequently, the use of CHO cell-based TGE protein production platforms may be preferable to maintain a consistent host cell type throughout the investigation and development phases. 3.4.2

CHO Cell-Based TGE Platforms

CHO cells are the primary workhorses for large-scale stable expression of human glycoproteins in the biopharmaceutical market [106]. Initially, TGE-based protein production in CHO cells was hampered by low transfection efficiency and volumetric productivity [104]. Extensive efforts have been undertaken to improve TGE-based protein production in CHO cells [89, 104]. For example, the ExpiCHO-S cell line (Life Technologies) was selected for high levels of transfection efficiency, productivity, and growth properties to maximize protein expression levels [36]. To improve the characteristics of CHO cells for TGE-based protein production, several cell engineering strategies have been performed to increase qp and volumetric productivity (Figure 3.1) (Table 3.3). As described in Section 3.3.2,

CHO cell engineering Wild-type CHO cells

Transient gene expression Engineered CHO cells

Engineered CHO cells

Expression vector with gene of interest

Selection & expansion Overexpression or knockout

Antiapoptosis Bcl-xL Proapoptosis Bax Bak

Protein expression mTOR Metabolism GS Glycosylation FUT8

Episomal replication EBNA-1 PyLT Secretion XBP-1 XBP-1S

Figure 3.1 Representative illustration of CHO cell engineering in TGE.

High production of recombinant protein

3.4 Mammalian Cell Lines

Table 3.3 CHO-derived cells commonly used for TGE-based recombinant protein production. Culture conditions

Culture medium

References

Combines GS system and FUT8 knockout

Suspension

UltraCHO

[23]

CHOEBNALT 85

Stably coexpressing murine PyLT and EBNA-1

Suspension

Mixture of CD CHO and 293 SFMII

[80, 107]

CHO-3E7

Stably expressing HSV VP16 and truncated EBNA-1

Suspension

Freestyle CHO expression medium

[108, 109]

Freestyle CHO-S

Derivative of CHO-S cells

Suspension

Freestyle CHO expression medium

[6]

ExpiCHO-S

Derivative of CHO-S cells

Suspension

ExpiCHO expression medium

[36]

CHO-T

Stably expressing mouse PyLT

Suspension

DMEM/F12 with 10% FBS and Ex-Cell 302

[76]

EB-GS22

Stably coexpressing EBNA-1 and GS

Suspension

CD-CHO or MedImmune proprietary medium

[104]

CHO-DG44

Overexpressing Bcl-xL

Adherent and suspension

CHOM (proprietary serum-free medium) and CHO-S-SFMII

[89]

DKO

Double knockout of Bax and Bak

Suspension

Ham’s F12-based medium

[110]

CHO-DG44

Overexpressing XBP-1

Suspension

Mixture of CD CHO and HyQ PF-CHO

[111]

CHOS-XE

Stably expressing XBP1 and ERO1-Lα

Adherent and suspension

CD-CHO with 10% FCS and CD-CHO

[29]

CHO-Ab

Co-overexpressing PDI and ERO1L

Adherent

IMDM with 10% dFBS

[112]

CHO-S

Triple knockout of FUT8, Bak, and Bax

Suspension

CD CHO medium

[113]

CHO-K1-ID2

Stably expressing mTOR

Adherent

ChoMaster HTS medium with 0–1% FCS

[114]

Name

Characteristics

Potelligent CHOK1SV

CHO cells have been engineered for episomal replication, giving rise to a threefold increase in protein production [76, 104, 108]. The transient expression of a therapeutic human fusion protein in CHO cells stably expressing the antiapoptotic protein Bcl-xL yielded a 3.4-fold increase in protein titer. This increase in titer was attributed to an increase in the maximum viable cell density and culture longevity [89]. In addition, a combined knockout of the proapoptotic factors Bax and Bak was found to increase the antibody titers with up to fourfold upon transient expression because of an increase in viability and a decrease

57

58

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

in apoptosis [110, 113]. CHO cells have been engineered to stably express the unfolded protein response transcriptional regulator, the spliced form of X-box binding protein 1 (XBP-1S), disulfide oxidase, and endoplasmic reticulum oxidoreductase 1α (ERO1-Lα). Using these cell lines for TGE-based protein production increased the product titers to 2.5- to 6.2-fold [29, 111]. Moreover, stable overexpression of the mammalian target of rapamycin (mTOR) in CHO cells effectively increased antibody titers fourfold upon transient expression of the antibody-encoding gene [114]. 3.4.3

TGE Platforms Using Other Cell Lines

Although HEK293 and CHO cells are the most frequently used host cell lines for TGE-based protein production, other mammalian cell-based TGE platforms are also available. For example, the HKB-11 cell line, which is a hybrid of HEK293 and B-cell lymphoma cells, harbors EBV and is readily adapted to suspension growth [99]. Upon transient transfection with the Tat/TAR transactivation system of HIV-I in the expression vector EBNA/oriP, HKB11 cells are reported to increase their production of interleukin-2SA (IL-2 mutein) 18-fold [99]. Furthermore, human-derived cells such as CAP-T (stably expressing SV4LT), simian-derived Vero, and COS are attractive potential host cell lines for transient protein production [5, 115, 116].

3.5 Cell Culture Strategies 3.5.1

Culture Media for TGE

Serum, the supernatant of clotted blood, is commonly used in mammalian cell culture media because of its high content of embryonic growth factor. However, the use of serum in mammalian cell culture for therapeutic protein production is limited owing to the risk of pathogen transmission as well as its high protein content [117], which can interfere with product purification. Serum-free media have been developed specifically for cultivation of HEK293 or CHO cells because they are the major host cell lines used to produce recombinant proteins. However, it was reported that the components of serum-free media that foster growth in mammalian cells, such as phosphate, dextran sulfate, and iron chelators, can interfere with the formation of DNA:transfection reagent complexes and inhibit transfection [118, 119]. In this light, ideal serum-free media for TGE-based protein production should support high cell density and increase transfection efficiency. The development of commercially available media for HEK293 and CHO cells markedly improved TGE-based protein production. Commercial transfection media suitable for HEK293 cells include EX-CELL 293 Serum-Free Medium (SAFC Biosciences) [15, 120, 121], FreeStyle 293 Expression Medium (Life Technologies) [16, 122], and Expi293 Expression Medium (Life Technologies) [36]. Those for CHO cells include CHO-S-SFMII (Invitrogen/Gibco) [123], UltraCHO medium (Cambrex) [23], ProCHO5 medium (Lonza) [121, 124–126], CD-CHO medium (Life Technologies/Invitrogen) [104, 127], and ExpiCHO Expression Medium (Life Technologies) [36].

3.5 Cell Culture Strategies

3.5.2

Optimization of Cell Culture Processes for TGE

Several cell culture process strategies related to the efficiency of TGE-based recombinant protein production have been investigated in the pursuit of qp -enhancing conditions and to improve cell growth. Increasing the transfection efficiency and accelerating gene expression by modulation of the transcriptional machinery leads to an increase in qp . Additionally, increasing the cell growth rate and extending the culture time through optimization of culture conditions increases the integral viable cell density, thereby increasing volumetric productivity. Detailed strategies are explained in the following section. 3.5.3

qp -Enhancing Factors in TGE-Based Culture Processes

qp -Enhancing factors are commonly applied in transfected cell cultures to increase the titers of therapeutic proteins. As described in Section 3.3.1, optimized delivery of DNA into the host cell nucleus is important to achieve high transfection efficiencies and high qp . Nocodazole and hydroxyurea, known as antimitotic reagents, arrest cells in the G2/M phase and the G1 phase, respectively. In the presence of these antimitotic reagents, the transfection efficiency and the transient production of immunoglobulin G (IgG) were improved twofold in CHO-S cells [123]. Nocodazole may act to maintain CHO cells in a transfectable state, possibly through disrupting mitotic spindle formation, which decreases the nuclear membrane integrity [123]. Another study indicates that dimethyl sulfoxide (DMSO) can be used to improve electroporation efficiency in mammalian cells [128]. Lithium acetate (LiAc) is commonly used in yeast and bacteria cultures to increase transformation efficiency [129, 130]. DMSO and LiAc are thought to increase the cellular uptake of DNA by altering the cell membrane and/or wall permeability. Both of these supplements are reported to significantly increase transient expression levels (threefold) in CHO cells [23]. Gene expression is regulated by histone acetylation. Hyperacetylation of histones results in more relaxed chromatin structures [131, 132], promoting an increase in transcription rates. Several histone deacetylation inhibitors (HDACis) have been reported to improve recombinant protein yields in transiently and stably transfected mammalian cells [133, 134]. For example, sodium butyrate (NaBu) and valproic acid (VPA) have been found to increase TGE levels and qp [42, 122, 126, 135, 136]. Several studies report that NaBu treatment increases recombinant antibody production up to 2.2-fold in TGE-based platforms in CHO cells [135]. Similarly, VPA treatment has been shown to increase titers in transiently transfected HEK293E cells (up to fourfold) and CHO-DG44 (1.5-fold) [136]. 3.5.4 Culture Longevity-Enhancing Factors in TGE-Based Culture Processes Lowering the temperature of transiently transfected cells (hypothermia) prolongs the culture longevity [137, 138]. The effects of hypothermia include an accumulation of cells in G1 phase, a reduction in cellular metabolism, and an increase

59

60

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

in the steady-state level of transgene expression [125]. These changes result in a decrease in the specific growth rate (𝜇) with a concomitant increase in qp and sustained high cell viability in transiently transfected mammalian cell cultures. When introduced at the appropriate time, hypothermia can be used to increase volumetric productivity. Mild hypothermic condition at 31 ∘ C are reported to increase antibody titers up to threefold for TGE-based protein production in CHO cells [125]. As with hypothermia, hyperosmotic pressure can be induced, by adding salts or sugar to the medium, to increase recombinant protein production. However, elevated osmolality suppresses cell growth and induces apoptosis [139–141]. Hyperosmolality is known to increase qp in TGE-based CHO cell cultures, giving rise to a fourfold increase in recombinant antibody titer [124]. Enzymatic digests of proteins (hydrolysates) are widely used as protein supplements to replace serum in mammalian cell cultures. Supplementation with hydrolysates in a TGE-based protein production system increased the volumetric productivity of human placental secreted alkaline phosphatase (SEAP) two to four times in HEK293E cells [142, 143]. However, the hydrolysate constituents (phosphate and dextran sulfate) have been demonstrated to hinder transfection [118]. For this reason, hydrolysates are preferably added after transfection. LongR3 -IGF, an analog of insulin-like growth factor (IGF)-I, has been modified specifically to promote cell survival and growth for recombinant protein production in mammalian cells [144]. Mild hypothermia combined with LongR3 -IGF synergistically increased recombinant IgG monoclonal antibody titers 11-fold in transiently transfected CHO cells [13].

3.6 Large-Scale TGE-Based Protein Production Large-scale TGE-based protein production using mammalian cells is a powerful technology for the rapid generation of milligram to gram quantities of recombinant proteins for use in biochemical and preclinical studies. Examples of published TGE-based protein production platforms in HEK293 and CHO cells exceeding 2 l cultures are shown in Table 3.4. With the exception of two studies performed with calcium phosphate-based transfection and electroporation, most reported studies on large-scale TGE-based protein production have used PEI-based transient transfection [122]. To the knowledge of the authors, there are currently no reports describing cell cultures exceeding 2 l using liposomal transient transfection for protein production. The most widely used HEK293-derived host cell lines are those that express EBNA-1 (HEK293E and 293-6E cells). These cells are combined with PEI transfection for large-scale TGE-based protein production [16, 17, 142]. In contrast, a variety of CHO cell-derived hosts are applied for large-scale TGE-based protein production, including CHO-DG44 [21], CHO-S [149], CHOK1SV [23], CHOK1SV-GS-KO (glutamine synthetase knockout CHO cells) [150], and CHO-EBNA-GS cells (CHO cells with co-expression of EBNA-1 and glutamine synthetase) [104]. The current maximum yields reported for large-scale TGE-based protein production cultures over 2 l are 1.1 g/l for HEK293E cells [16] and 2 gl for CHO cells [104].

3.6 Large-Scale TGE-Based Protein Production

Table 3.4 Large-scale TGE-based recombinant protein production platforms used in HEK293 and CHO cells.

Cell line

Transfection method

Culture system

Scale (l) Product

Titer (mg/l) References

HEK293SFE

PEI

STR

10/14

Tie-2 and neuropilin-1 (ED)

ND

[142]

HEK293E

Calcium phosphate

STR

3

Human recombinant IgG

18

[60]

HEK293E

PEI

STR (helical ribbon)

3.5

Human placental SEAP

20

[14]

HEK293E

PEI

Schott glass round bottle

2

Recombinant IgG

1000

[16]

HEK293E

Calcium phosphate

STR

110

Human recombinant IgG

7.7

[145]

HEK293E

PEI

WAVE bioreactor

10

Various proteins 12

[90]

HEK293E

Calfection

STR

30

Human recombinant IgG

2.5

[146]

HEK293E

PEI

WAVE bioreactor and STR

10/100 Recombinant IgG

6.7

[147]

293-6E

PEI

STR

10

IL-17B and IL-15 and IL-11

ND

[92]

293-6E

PEI

WAVE bioreactor

5

Recombinant IgG

116

[17]

CHO-DG44

PEI

STR

13

Human recombinant IgG

6

[21]

CHO-DG44

PEI

Orbital shake bioreactors

30

Human recombinant IgG

60

[148]

CHO-S

PEI

WAVE bioreactor

20

Human recombinant IgG

9.4

[149]

CHO-S

Electroporation

Shake flask

2.8

Humanized IgG1

3500

[41]

CHOK1SV

PEI

WAVE bioreactor

8

Monoclonal antibody

66

[23]

CHO-EBNAGS

PEI

WAVE bioreactor

6

Recombinant IgG

2000

[104]

CHO K1SV GS KO

PEI

STR

2

Human recombinant IgG

249

[150]

Only studies using cultures with volumes >2 l are shown. Abbreviations: ND, not described; STR, stirred-tank bioreactor; SEAP, secreted alkaline phosphatase; ED, extracellular domains; IL, interleukin.

61

62

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

Large-scale TGE systems typically use stirred-tank bioreactors or WAVE bioreactors for relatively precise control of culture environment parameters such as gas flow, pH, and temperature. The use of disposable plastic bags as bioreactors is becoming more popular than conventional reusable bioreactors. The main advantages of single-use bioreactors include relatively low installation costs as well as ease of maintenance, validation, and handling [17, 104, 127]. Importantly, there are some drawbacks of large-scale TGE-based protein production platforms. For example, the amount of plasmid DNA needed is proportional to the cell culture volume. In addition, large-scale centrifugation steps are typically needed, as proteins are purified from batch cultures. Finally, there is a substantial contamination risk associated with large-scale transfections. Recently, automation of a large-scale TGE-based protein production platforms has been reported [97], ensuring robustness, reproducibility of transfection, and minimizing the risk of human errors.

3.7 Concluding Remarks The technological breakthrough of therapeutic protein production in HEK293 and CHO cell-based TGE platforms is the combined result of improving transfection methods, developing new cell lines, and optimizing cell culture processes. Differences in protein quality between TGE-based and stable cell line-based protein production are potentially curtailing the clinical applicability of TGE-produced proteins. The quality differences likely originate to some extent from the use of different host cells as well as from clonal variation. In addition, transfection is typically cytotoxic, giving rise to a significant decrease in viability with the concomitant lysis of dead cells into the medium. In this context, development of a “CHO cell only” platform for transient and stable protein production is a promising strategy for minimizing differences in protein quality. Furthermore, cell engineering efforts combined with TGE-based protein production platforms have the potential to meet the increasing demand of producing difficult-to-express proteins such as fusion protein-based biopharmaceuticals.

References 1 Kaufman, R.J., Wasley, L.C., Spiliotes, A.J. et al. (1985). Coamplification

and coexpression of human tissue-type plasminogen activator and murine dihydrofolate reductase sequences in Chinese hamster ovary cells. Mol. Cell. Biol. 5 (7): 1750–1759. 2 De Jesus, M. and Wurm, F.M. (2011). Manufacturing recombinant proteins in kg-ton quantities using animal cells in bioreactors. Eur. J. Pharm. Biopharm. 78 (2): 184–188. 3 Huang, Y.M., Hu, W., Rustandi, E. et al. (2010). Maximising productivity of CHO cell-based fed-batch culture using chemically defined media conditions and typical manufacturing equipment. Biotechnol. Progr. 26 (5): 1400–1410.

References

4 Hacker, D.L., De Jesus, M., and Wurm, F.M. (2009). 25 years of recombinant

5 6

7

8

9

10 11 12 13

14

15

16

17

18

proteins from reactor-grown cells—where do we go from here? Biotechnol. Adv. 27 (6): 1023–1027. Geisse, S. (2009). Reflections on more than 10 years of TGE approaches. Protein Expression Purif. 64 (2): 99–107. Liu, C., Dalby, B., Chen, W. et al. (2008). Transient transfection factors for high-level recombinant protein production in suspension cultured mammalian cells. Mol. Biotechnol. 39 (2): 141–153. Jäger, V., Büssow, K., and Schirrmann, T. (2015). Transient recombinant protein expression in mammalian cells. In: Animal Cell Culture (ed. M. Al-Rubeai), 27–64. Springer International Publishing. Polo, J.M., Belli, B.A., Driver, D.A. et al. (1999). Stable alphavirus packaging cell lines for Sindbis virus and Semliki Forest virus-derived vectors. Proc. Natl. Acad. Sci. U.S.A. 96 (8): 4598–4603. Hacker, D.L., Baldi, L., Adam, M., and Wurm, F.M. (2009). Transient gene expression in mammalian cells: promises and challenges for medical biotechnology. In: Encyclopedia of Industrial Biotechnology: Bioprocess, Bioseparation, and Cell Technology (ed. M.C. Flickinger), 1–8. Wiley. Nayerossadat, N., Maedeh, T., and Ali, P.A. (2012). Viral and nonviral delivery systems for gene delivery. Adv. Biomed. Res. 1 (1): 27. Ramamoorth, M. and Narvekar, A. (2015). Non viral vectors in gene therapy-an overview. J. Clin. Diagn. Res. 9 (1): GE01–GE06. Zuber, G., Dauty, E., Nothisen, M. et al. (2001). Towards synthetic viruses. Adv. Drug Delivery Rev. 52 (3): 245–253. Galbraith, D.J., Tait, A.S., Racher, A.J. et al. (2006). Control of culture environment for improved polyethylenimine-mediated transient production of recombinant monoclonal antibodies by CHO cells. Biotechnol. Progr. 22 (3): 753–762. Durocher, Y., Perret, S., and Kamen, A. (2002). High-level and high-throughput recombinant protein production by transient transfection of suspension-growing human 293-EBNA1 cells. Nucleic Acids Res. 30 (2): e9. Backliwal, G., Hildinger, M., Hasija, V., and Wurm, F.M. (2008). High-density transfection with HEK-293 cells allows doubling of transient titers and removes need for a priori DNA complex formation with PEI. Biotechnol. Bioeng. 99 (3): 721–727. Backliwal, G., Hildinger, M., Chenuet, S. et al. (2008). 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. 36 (15): e96. Raymond, C., Tom, R., Perret, S. et al. (2011). A simplified polyethylenimine-mediated transfection process for large-scale and high-throughput applications. Methods 55 (1): 44–51. Cervera, L., Gutiérrez-Granados, S., Berrow, N.S. et al. (2015). Extended gene expression by medium exchange and repeated transient transfection for recombinant protein production enhancement. Biotechnol. Bioeng. 112 (5): 934–946.

63

64

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

19 Rajendra, Y., Kiseljak, D., Baldi, L. et al. (2015). Transcriptional and

20

21 22

23 24

25

26

27

28

29

30

31 32

33

34

post-transcriptional limitations of high-yielding, PEI-mediated transient transfection with CHO and HEK-293E cells. Biotechnol. Progr. 31 (2): 541–549. Delafosse, L., Xu, P., and Durocher, Y. (2016). Comparative study of polyethylenimines for transient gene expression in mammalian HEK293 and CHO cells. J. Biotechnol. 227: 103–111. Derouazi, M., Girard, P., Van Tilborgh, F. et al. (2004). Serum-free large-scale transient transfection of CHO cells. Biotechnol. Bioeng. 87 (4): 537–545. Muller, N., Derouazi, M., Van Tilborgh, F. et al. (2007). Scalable transient gene expression in Chinese hamster ovary cells in instrumented and non-instrumented cultivation systems. Biotechnol. Lett. 29 (5): 703–711. Ye, J., Kober, V., Tellers, M. et al. (2009). High-level protein expression in scalable CHO transient transfection. Biotechnol. Bioeng. 103 (3): 542–551. Zustiak, M.P., Jose, L., Xie, Y. et al. (2014). Enhanced transient recombinant protein production in CHO cells through the co-transfection of the product gene with Bcl-xL. Biotechnol. J. 9 (9): 1164–1174. Rajendra, Y., Balasubramanian, S., Kiseljak, D. et al. (2015). Enhanced plasmid DNA utilization in transiently transfected CHO–DG44 cells in the presence of polar solvents. Biotechnol. Progr. 31 (6): 1571–1578. Codamo, J., Hou, J.J.C., Hughes, B.S. et al. (2011). Efficient mAb production in CHO cells incorporating PEI-mediated transfection, mild hypothermia and the co-expression of XBP-1. J. Chem. Technol. Biotechnol. 86 (7): 923–934. Rajendra, Y., Hougland, M.D., Schmitt, M.G., and Barnard, G.C. (2015). Transcriptional and post-transcriptional targeting for enhanced transient gene expression in CHO cells. Biotechnol. Lett. 37 (12): 2379–2386. Codamo, J., Munro, T.P., Hughes, B.S. et al. (2011). Enhanced CHO cell-based transient gene expression with the epi-CHO expression system. Mol. Biotechnol. 48 (2): 109–115. Cain, K., Peters, S., Hailu, H. et al. (2013). A CHO cell line engineered to express XBP1 and ERO1-Lα has increased levels of transient protein expression. Biotechnol. Progr. 29 (3): 697–706. Maurisse, R., De Semir, D., Emamekhoo, H. et al. (2010). Comparative transfection of DNA into primary and transformed mammalian cells from different lineages. BMC Biotechnol. 10 (1): 9. Lin, C.Y., Huang, Z., Wen, W. et al. (2015). Enhancing protein expression in HEK-293 cells by lowering culture temperature. PLoS One 10 (4): e0123562. Ooi, A., Wong, A., Esau, L. et al. (2016). A guide to transient expression of membrane proteins in HEK-293 cells for functional characterization. Front. Physiol. 7: 300. Rosser, M.P., Xia, W., Hartsell, S. et al. (2005). Transient transfection of CHO-K1-S using serum-free medium in suspension: a rapid mammalian protein expression system. Protein Expression Purif. 40 (2): 237–243. Reisinger, H., Sevcsik, E., Vorauer-Uhl, K. et al. (2007). Serum-free transfection of CHO-cells with tailor-made unilamellar vesicles. Cytotechnology 54 (3): 157–168.

References

35 Vink, T., Oudshoorn-Dickmann, M., Roza, M. et al. (2014). A simple, robust

36

37 38 39 40 41

42

43

44

45

46

47

48

49 50 51

and highly efficient transient expression system for producing antibodies. Methods 65 (1): 5–10. Jain, N.K., Barkowski-Clark, S., Altman, R. et al. (2017). A high density CHO-S transient transfection system: comparison of ExpiCHO and Expi293. Protein Expression Purif. 134: 38–46. Graham, F.L. and van der Eb, A.J. (1973). A new technique for the assay of infectivity of human adenovirus 5 DNA. Virology 52 (2): 456–467. Kim, T.K. and Eberwine, J.H. (2010). Mammalian cell transfection: the present and the future. Anal. Bioanal. Chem. 397 (8): 3173–3178. Jordan, M. and Wurm, F.M. (2004). Transfection of adherent and suspended cells by calcium phosphate. Methods 33 (2): 136–143. Inoue, T. and Krumlauf, R. (2001). An impulse to the brain—using in vivo electroporation. Nat. Neurosci. 4: 1156–1158. Steger, K., Brady, J., Wang, W. et al. (2015). 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. Screening 20 (4): 545–551. Parham, J.H., Iannone, M.A., Overton, L.K., and Hutchins, J.T. (1998). Optimization of transient gene expression in mammalian cells and potential for scale-up using flow electroporation. Cytotechnology 28 (1–3): 147–155. Krammer, F., Pontiller, J., Tauer, C. et al. (2010). Evaluation of the influenza A replicon for transient expression of recombinant proteins in mammalian cells. PLoS One 5 (10): e13265. Abdul Halim, N.S., Fakiruddin, K.S., Ali, S.A., and Yahaya, B.H. (2014). A comparative study of non-viral gene delivery techniques to human adipose-derived mesenchymal stem cell. Int. J. Mol. Sci. 15 (9): 15044–15060. Boussif, O., Lezoualc’h, F., Zanta, M.A. et al. (1995). A versatile vector for gene and oligonucleotide transfer into cells in culture and in vivo: polyethylenimine. Proc. Natl. Acad. Sci. U.S.A. 92 (16): 7297–7301. Thomas, M. and Klibanov, A.M. (2003). Non-viral gene therapy: polycation-mediated DNA delivery. Appl. Microbiol. Biotechnol. 62 (1): 27–34. Moghimi, S.M., Symonds, P., Murray, J.C. et al. (2005). A two-stage(ethylenimine)-mediated cytotoxicity: implications for gene transfer/therapy. Mol. Ther. 11 (6): 990–995. Kusumoto, K.I., Akao, T., Mizuki, E., and Nakamura, O. (2006). Gene transfer effects on various cationic amphiphiles in CHO cells. Cytotechnology 51 (2): 57–66. Dang, J.M. and Leong, K.W. (2006). Natural polymers for gene delivery and tissue engineering. Adv. Drug Delivery Rev. 58 (4): 487–499. Jiang, H.L., Kim, T.H., Kim, Y.K. et al. (2008). Efficient gene delivery using chitosan–polyethylenimine hybrid systems. Biomed. Mater. 3 (2): e025013. Corsi, K., Chellat, F., Yahia, L.H., and Fernandes, J.C. (2003). Mesenchymal stem cells, MG63 and HEK293 transfection using chitosan-DNA nanoparticles. Biomaterials 24 (7): 1255–1264.

65

66

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

52 Gonzalez, H., Hwang, S.J., and Davis, M.E. (1999). New class of polymers

53

54

55 56

57

58 59

60

61

62

63 64

65 66

67 68 69

for the delivery of macromolecular therapeutics. Bioconjugate Chem. 10 (6): 1068–1074. Felgner, P.L., Gadek, T.R., Holm, M. et al. (1987). Lipofection: a highly efficient, lipid-mediated DNA-transfection procedure. Proc. Natl. Acad. Sci. U.S.A. 84 (21): 7413–7417. Parker, A.L., Newman, C., Briggs, S. et al. (2003). Nonviral gene delivery: techniques and implications for molecular medicine. Expert Rev. Mol. Med. 5 (22): 1–15. Jacobsen, L.B., Calvin, S.A., Colvin, K.E., and Wright, M. (2004). FuGENE 6 transfection reagent: the gentle power. Methods 33 (2): 104–112. Yin, W., Xiang, P., and Li, Q. (2005). Investigations of the effect of DNA size in transient transfection assay using dual luciferase system. Anal. Biochem. 346 (2): 289–294. Oliveira, P.H. and Mairhofer, J. (2013). Marker-free plasmids for biotechnological applications–implications and perspectives. Trends Biotechnol. 31 (9): 539–547. Makrides, S.C. (1999). Components of vectors for gene transfer and expression in mammalian cells. Protein Expression Purif. 17 (2): 183–202. Van Craenenbroeck, K., Vanhoenacker, P., and Haegeman, G. (2000). Episomal vectors for gene expression in mammalian cells. FEBS J. 267 (18): 5665–5678. Meissner, P., Pick, H., Kulangara, A. et al. (2001). Transient gene expression: recombinant protein production with suspension-adapted HEK293-EBNA cells. Biotechnol. Bioeng. 75 (2): 197–203. Xia, W., Bringmann, P., McClary, J. et al. (2006). High levels of protein expression using different mammalian CMV promoters in several cell lines. Protein Expression Purif. 45 (1): 115–124. Lee, J., Lau, J., Chong, G., and Chao, S.H. (2007). Cell-specific effects of human cytomegalovirus unique region on recombinant protein expression. Biotechnol. Lett. 29 (12): 1797–1802. Smale, S.T. and Kadonaga, J.T. (2003). The RNA polymerase II core promoter. Annu. Rev. Biochem. 72 (1): 449–479. Klein, R., Ruttkowski, B., Knapp, E. et al. (2006). WPRE-mediated enhancement of gene expression is promoter and cell line specific. Gene 372: 153–161. Le Hir, H., Nott, A., and Moore, M.J. (2003). How introns influence and enhance eukaryotic gene expression. Trends Biochem. Sci. 28 (4): 215–220. Matsumoto, K., Wassarman, K.M., and Wolffe, A.P. (1998). Nuclear history of a pre-mRNA determines the translational activity of cytoplasmic mRNA. EMBO J. 17 (7): 2107–2121. Vos, J.M. (1998). Mammalian artificial chromosomes as tools for gene therapy. Curr. Opin. Genet. Dev. 8 (3): 351–359. Vos, J.M. (1999). Therapeutic mammalian artificial episomal chromosomes. Curr. Opin. Mol. Ther. 1 (2): 204–215. Längle-Rouault, F., Patzel, V., Benavente, A. et al. (1998). Up to 100-fold increase of apparent gene expression in the presence of Epstein–Barr virus

References

70

71

72 73

74

75

76

77

78 79

80

81

82

83

84

oriP sequences and EBNA1: implications of the nuclear import of plasmids. J. Virol. 72 (7): 6181–6185. Young, J.M., Cheadle, C., Foulke, J.S. et al. (1988). Utilization of an Epstein–Barr virus replicon as a eukaryotic expression vector. Gene 62 (2): 171–185. Lufino, M.M., Edser, P.A., and Wade-Martins, R. (2008). Advances in high-capacity extrachromosomal vector technology: episomal maintenance, vector delivery, and transgene expression. Mol. Ther. 16 (9): 1525–1538. Bullock, P.A. and Simmons, D.T. (1997). The initiation of simian virus 40 DNA replication in vitro. Crit. Rev. Biochem. Mol. Biol. 32 (6): 503–568. Lindner, S.E. and Sugden, B. (2007). The plasmid replicon of Epstein–Barr virus: mechanistic insights into efficient, licensed, extrachromosomal replication in human cells. Plasmid 58 (1): 1–12. Yates, J.L., Warren, N., and Sugden, B. (1985). Stable replication of plasmids derived from Epstein–Barr virus in various mammalian cells. Nature 313 (6005): 812–815. Rawlins, D.R., Milman, G., Hayward, S.D. et al. (1985). Sequence-specific DNA binding of the Epstein–Barr virus nuclear antigen (EBNA-1) to clustered sites in the plasmid maintenance region. Cell 42 (3): 859–868. Kunaparaju, R., Liao, M., and Sunstrom, N.A. (2005). Epi-CHO, an episomal expression system for recombinant protein production in CHO cells. Biotechnol. Bioeng. 91 (6): 670–677. Pham, P.L., Kamen, A., and Durocher, Y. (2006). Large-scale transfection of mammalian cells for the fast production of recombinant protein. Mol. Biotechnol. 34 (2): 225–237. Durocher, Y. and Butler, M. (2009). Expression systems for therapeutic glycoprotein production. Curr. Opin. Biotechnol. 20 (6): 700–707. DuBridge, R.B., Tang, P., Hsia, H.C. et al. (1987). Analysis of mutation in human cells by using an Epstein–Barr virus shuttle system. Mol. Cell. Biol. 7 (1): 379–387. Silla, T., Tagen, I., Geimanen, J. et al. (2006). Vectors, cell lines and their use in obtaining extended episomal maintenance replication of hybrid plasmids and expression of gene products. US Patent 7,790,446, filed 10 February 2006 and issued 7 September 2010. Hansen, H.G., Pristovšek, N., Kildegaard, H.F., and Lee, G.M. (2017). Improving the secretory capacity of Chinese hamster ovary cells by ectopic expression of effector genes: lessons learned and future directions. Biotechnol. Adv. 35 (1): 64–76. Fischer, S., Handrick, R., and Otte, K. (2015). The art of CHO cell engineering: a comprehensive retro and future perspectives. Biotechnol. Adv. 33 (8): 1878–1896. Johari, Y.B., Estes, S.D., Alves, C.S. et al. (2015). Integrated cell and process engineering for improved transient production of a “difficult-to-express” fusion protein by CHO cells. Biotechnol. Bioeng. 112 (12): 2527–2542. Hansen, H.G., Nilsson, C.N., Lund, A.M. et al. (2015). Versatile microscale screening platform for improving recombinant protein productivity in Chinese hamster ovary cells. Sci. Rep. 5: e18016.

67

68

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

85 Pybus, L.P., Dean, G., West, N.R. et al. (2014). Model-directed engineering

86

87 88

89

90

91

92

93

94 95

96

97 98

99

of “difficult-to-express” monoclonal antibody production by Chinese hamster ovary cells. Biotechnol. Bioeng. 111 (2): 372–385. Backliwal, G., Hildinger, M., Chenuet, S. et al. (2008). Coexpression of acidic fibroblast growth factor enhances specific productivity and antibody titers in transiently transfected HEK293 cells. New Biotechnol. 25 (2): 162–166. Zhu, J. (2012). Mammalian cell protein expression for biopharmaceutical production. Biotechnol. Adv. 30 (5): 1158–1170. Baldi, L., Hacker, D.L., Adam, M., and Wurm, F.M. (2007). Recombinant protein production by large-scale transient gene expression in mammalian cells: state of the art and future perspectives. Biotechnol. Lett. 29 (5): 677–684. Majors, B.S., Betenbaugh, M.J., Pederson, N.E., and Chiang, G.G. (2008). Enhancement of transient gene expression and culture viability using Chinese hamster ovary cells overexpressing Bcl-xL. Biotechnol. Bioeng. 101 (3): 567–578. Geisse, S. and Henke, M. (2005). Large-scale transient transfection of mammalian cells: a newly emerging attractive option for recombinant protein production. J. Struct. Funct. Genom. 6 (2): 165–170. Loignon, M., Perret, S., Kelly, J. et al. (2008). Stable high volumetric production of glycosylated human recombinant IFNalpha2b in HEK293 cells. BMC Biotechnol. 8 (1): 65. Carter, J., Zhang, J., Dang, T.L. et al. (2010). Fusion partners can increase the expression of recombinant interleukins via transient transfection in 2936E cells. Protein Sci. 19 (2): 357–362. Pear, W.S., Nolan, G.P., Scott, M.L., and Baltimore, D. (1993). Production of high-titer helper-free retroviruses by transient transfection. Proc. Natl. Acad. Sci. U.S.A. 90 (18): 8392–8396. Chang, V.T., Crispin, M., Aricescu, A.R. et al. (2007). Glycoprotein structural genomics: solving the glycosylation problem. Structure 15 (3): 267–273. Li, J., Menzel, C., Meier, D. et al. (2007). A comparative study of different vector designs for the mammalian expression of recombinant IgG antibodies. J. Immunol. Methods 318 (1): 113–124. Reeves, P.J., Callewaert, N., Contreras, R., and Khorana, H.G. (2002). Structure and function in rhodopsin: high-level expression of rhodopsin with restricted and homogeneous N-glycosylation by a tetracycline-inducible N-acetylglucosaminyltransferase I-negative HEK293S stable mammalian cell line. Proc. Natl. Acad. Sci. U.S.A. 99 (21): 13419–13424. Zhao, Y., Bishop, B., Clay, J.E. et al. (2011). Automation of large scale transient protein expression in mammalian cells. J. Struct. Biol. 175 (2): 209–215. Lee, J.H., Lim, S.M., Park, S.H. et al. (2017). Investigation of relationship between EBNA-1 expression level and specific foreign protein productivity in transient gene expression of HEK293 cells. Process Biochem. 55: 182–186. Cho, M.S., Yee, H., Brown, C. et al. (2003). Versatile expression system for rapid and stable production of recombinant proteins. Biotechnol. Progr. 19 (1): 229–232.

References

100 Chiou, H.C., Vasu, S., Liu, C.Y. et al. (2014). Scalable transient protein

101

102 103

104

105

106

107

108

109

110

111

112

113

expression. In: Animal Cell Biotechnology, Methods and Protocols (ed. R. Pörtner), 35–55. Springer Science+Business Media, LLC. Vasu, S., Chiou, H., Rogers, J. et al. (2013). High yield transient expression in mammalian cells using unique pairing of high density growth and transfection medium and expression enhancers. US Patent 20,130,316,400, filed 2 May 2013 and issued 28 November 2013. Graham, F.L. (1987). Growth of 293 cells in suspension culture. J. Gen. Virol. 68 (Pt. 3): 937–940. Garnier, A., Côté, J., Nadeau, I. et al. (1994). Scale-up of the adenovirus expression system for the production of recombinant protein in human 293S cells. Cytotechnology 15 (1–3): 145–155. Daramola, O., Stevenson, J., Dean, G. et al. (2014). A high-yielding CHO transient system: coexpression of genes encoding EBNA-1 and GS enhances transient protein expression. Biotechnol. Progr. 30 (1): 132–141. Yu, D.Y., Noh, S.M., and Lee, G.M. (2016). Limitations to the development of recombinant human embryonic kidney 293E cells using glutamine synthetase-mediated gene amplification: methionine sulfoximine resistance. J. Biotechnol. 231: 136–140. Noh, S.M., Sathyamurthy, M., and Lee, G.M. (2013). Development of recombinant Chinese hamster ovary cell lines for therapeutic protein production. Curr. Opin. Chem. Eng. 2 (4): 391–397. Silla, T., Tagen, I., Kalling, A. et al. (2010). Viral expression plasmids for production of proteins, antibodies, enzymes, virus-like particles and for use in cell-based assays. US Patent 8,377,653, filed 7 July 2010 and issued 19 February 2013. Durocher, Y. and Loignon, M. (2009). Process, vectors and engineered cell lines for enhanced large-scale transfection. US Patent 8,637,315, filed 9 March 2009 and issued 28 January 2014. Stoops, J., Byrd, S., and Hasegawa, H. (2012). Russell body inducing threshold depends on the variable domain sequences of individual human IgG clones and the cellular protein homeostasis. Biochim. Biophys. Acta, Mol. Cell Res. 1823 (10): 1643–1657. Macaraeg, N.F., Reilly, D.E., and Wong, A.W. (2013). Use of an anti-apoptotic CHO cell line for transient gene expression. Biotechnol. Progr. 29 (4): 1050–1058. Ku, S.C., Ng, D.T., Yap, M.G., and Chao, S.H. (2008). Effects of overexpression of X-box binding protein 1 on recombinant protein production in Chinese hamster ovary and NS0 myeloma cells. Biotechnol. Bioeng. 99 (1): 155–164. Mohan, C. and Lee, G.M. (2010). Effect of inducible co-overexpression of protein disulfide isomerase and endoplasmic reticulum oxidoreductase on the specific antibody productivity of recombinant Chinese hamster ovary cells. Biotechnol. Bioeng. 107 (2): 337–346. Grav, L.M., Lee, J.S., Gerling, S. et al. (2015). One-step generation of triple knockout CHO cell lines using CRISPR/Cas9 and fluorescent enrichment. Biotechnol. J. 10 (9): 1446–1456.

69

70

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

114 Dreesen, I.A. and Fussenegger, M. (2011). Ectopic expression of human

115

mTOR increases viability, robustness, cell size, proliferation, and antibody production of chinese hamster ovary cells. Biotechnol. Bioeng. 108 (4): 853–866. Fischer, S., Charara, N., Gerber, A. et al. (2012). Traneitn recombinant protein expression in a human amniocyte cell line: the CAP-T cell system. Biotechnol. Bioeng. 109 (9): 2250–2261. Fliedl, L. and Kaisermayer, C. (2011). Transient gene expression in HEK293 and vero cells immobilized on microcarriers. J. Biotechnol. 153 (1–2): 15–21. Grillberger, L., Kreil, T.R., Nasr, M., and Reiter, M. (2009). Emerging trends in plasma-free manufacturing of recombinant protein therapeutics expressed in mammalian cells. Biotechnol. J. 4 (2): 186–201. Geng, Z.H., Nudson, W., Davis, L. et al. (2007). Optimizing medium for transient transfection. In: Cell Technology for Cell Products (ed. R. Smith), 53–55. Dordrecht: Springer. Eberhardy, S.R., Radzniak, L., and Liu, Z. (2009). Iron (III) citrate inhibits polyethylenimine-mediated transient transfection of Chinese hamster ovary cells in serum-free medium. Cytotechnology 60 (1–3): 1–9. Rajendra, Y., Kiseljak, D., Manoli, S. et al. (2012). Role of non-specific DNA in reducing coding DNA requirement for transient gene expression with CHO and HEK-293E cells. Biotechnol. Bioeng. 109 (9): 2271–2278. Bollin, F., Dechavanne, V., and Chevalet, L. (2011). Design of experiment in CHO and HEK transient transfection condition optimization. Protein Expression Purif. 78 (1): 61–68. You, M., Liu, Y., Chen, Y. et al. (2013). Maximizing antibody production in suspension-cultured mammalian cells by the customized transient gene expression method. Biosci. Biotechnol. Biochem. 77 (6): 1207–1213. Tait, A.S., Brown, C.J., Galbraith, D.J. et al. (2004). Transient production of recombinant proteins by Chinese hamster ovary cells using polyethyleneimine/DNA complexes in combination with microtubule disrupting anti-mitotic agents. Biotechnol. Bioeng. 88 (6): 707–721. Zhang, X., Garcia, I.F., Baldi, L. et al. (2010). Hyperosmolarity enhances transient recombinant protein yield in Chinese hamster ovary cells. Biotechnol. Lett. 32 (11): 1587–1592. Wulhfard, S., Tissot, S., Bouchet, S. et al. (2008). Mild hypothermia improves transient gene expression yields several fold in Chinese hamster ovary cells. Biotechnol. Progr. 24 (2): 458–465. Wulhfard, S., Baldi, L., Hacker, D.L., and Wurm, F.M. (2010). Valproic acid enhances recombinant mRNA and protein levels in transiently transfected Chinese hamster ovary cells. J. Biotechnol. 148 (2): 128–132. Abbott, W.M., Middleton, B., Kartberg, F. et al. (2015). Optimisation of a simple method to transiently transfect a CHO cell line in high-throughput and at large scale. Protein Expression Purif. 116: 113–119.

®

116

117

118

119

120

121

122

123

124

125

126

127

References

128 Melkonyan, H., Sorg, C., and Klempt, M. (1996). Electroporation efficiency

129

130

131 132

133

134

135

136

137

138

139

140

141

in mammalian cells is increased by dimethyl sulfoxide (DMSO). Nucleic Acids Res. 24 (21): 4356–4357. Gietz, R.D. and Schiestl, R.H. (1991). Applications of high efficiency lithium acetate transformation of intact yeast cells using single-stranded nucleic acids as carrier. Yeast 7 (3): 253–263. Papagianni, M., Avramidis, N., and Filioussis, G. (2007). High efficiency electrotransformation of Lactococcus lactis spp. lactis cells pretreated with lithium acetate and dithiothreitol. BMC Biotechnol. 7 (1): 15. Ropero, S. and Esteller, M. (2007). The role of histone deacetylases (HDACs) in human cancer. Mol. Oncol. 1 (1): 19–25. Wade, P.A. (2001). Transcriptional control at regulatory checkpoints by histone deacetylases: molecular connections between cancer and chromatin. Hum. Mol. Genet. 10 (7): 693–698. Liu, C.H., Chu, I.M., and Hwang, S.M. (2001). Pentanoic acid, a novel protein synthesis stimulant for Chinese hamster ovary (CHO) cells. J. Biosci. Bioeng. 91 (1): 71–75. Allen, M.J., Boyce, J.P., Trentalange, M.T. et al. (2008). Identification of novel small molecule enhancers of protein production by cultured mammalian cells. Biotechnol. Bioeng. 100 (6): 1193–1204. Kim, W.H., Kim, M.S., Kim, Y.G., and Lee, G.M. (2012). Development of apoptosis-resistant CHO cell line expressing PyLT for the enhancement of transient antibody production. Process Biochem. 47 (12): 2557–2576. Backliwal, G., Hildinger, M., Kuettel, I. et al. (2008). Valproic acid: a viable alternative to sodium butyrate for enhancing protein expression in mammalian cell cultures. Biotechnol. Bioeng. 101 (1): 182–189. Weidemann, R., Ludwig, A., and Kretzmer, G. (1994). Low temperature cultivation – a step towards process optimization. Cytotechnology 15 (1): 111–116. Schlaeger, E.J. and Lundstrom, K. (1998). Effect of temperature on recombinant protein expression in Semliki Forest virus infected mammalian cell lines growing in serum-free suspension cultures. Cytotechnology 28 (1–3): 205–211. Ryu, J.S., Kim, T.K., Chung, J.Y., and Lee, G.M. (2000). Osmoprotective effect of glycine betaine on foreign protein production in hyperosmotic recombinant chinese hamster ovary cell cultures differs among cell lines. Biotechnol. Bioeng. 70 (2): 167–175. Kim, M.S., Kim, N.S., Sung, Y.H., and Lee, G.M. (2002). Biphasic culture strategy based on hyperosmotic pressure for improved humanized antibody production in Chinese hamster ovary cell culture. In Vitro Cell. Dev. Biol. Anim. 38 (6): 314–319. Lee, M.S., Kim, K.W., Kim, Y.H., and Lee, G.M. (2003). Proteome analysis of antibody-expressing CHO cells in response to hyperosmotics pressure. Biotechnol. Progr. 19 (6): 1734–1741.

71

72

3 Transient Gene Expression-Based Protein Production in Recombinant Mammalian Cells

142 Pham, P.L., Perret, S., Doan, H.C. et al. (2003). Large-scale transient trans-

143

144

145 146

147

148

149

150

fection of serum-free suspension-growing HEK293 EBNA1 cells: peptone additives improve cell growth and transfection efficiency. Biotechnol. Bioeng. 84 (3): 332–342. Pham, P.L., Perret, S., Cass, B. et al. (2005). Transient gene expression in HEK293 cells: peptone addition posttransfection improves recombinant protein synthesis. Biotechnol. Bioeng. 90 (3): 332–344. Voorhamme, D. and Yandell, C.A. (2006). LONG R3IGF-I as a more potent alternative to insulin in serum-free culture of HEK293 cells. Mol. Biotechnol. 34 (2): 201–204. Girard, P., Derouazi, M., Baumgartner, G. et al. (2002). 100-liter transient transfection. Cytotechnology 38 (1): 15–21. Lindell, J., Girard, P., Müller, N. et al. (2004). Calfection: a novel gene transfer method for suspension cells. Biochim. Biophys. Acta, Gene Struct. Expr. 1676 (2): 155–161. Tuvesson, O., Uhe, C., Rozkov, A., and Lüllau, E. (2008). Development of a generic transient transfection process at 100 L scale. Cytotechnology 56 (2): 123–136. Stettler, M., Zhang, X., Hacker, D.L. et al. (2007). Novel orbital shake bioreactors for transient production of CHO derived IgGs. Biotechnol. Progr. 23 (6): 1340–1346. Haldankar, R., Li, D., Saremi, Z. et al. (2006). Serum-free suspension large-scale transient transfection of CHO cells in WAVE bioreactors. Mol. Biotechnol. 34 (2): 191–199. Rajendra, Y., Hougland, M.D., Alam, R. et al. (2015). A high cell density transient transfection system for therapeutic protein expression based on a CHO GS-knockout cell line: process development and product quality assessment. Biotechnol. Bioeng. 112 (5): 977–986.

73

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models of CHO Metabolism ∗

∗

Shangzhong Li 1,2, , Anne Richelle 1,3, , and Nathan E. Lewis 1,2,3 1 Novo Nordisk Center for Biosustainability at the University of California, San Diego, 9500 Gilman Dr, La Jolla, CA 92093, USA 2 University of California, San Diego, Department of Bioengineering, 9500 Gilman Dr, La Jolla, CA 92093, USA 3 University of California, San Diego, Department of Pediatrics, 9500 Gilman Dr, La Jolla, CA 92093, USA

4.1 Introduction In 1987, the first Chinese hamster ovary (CHO)-derived human tissue plasminogen activator (tPA) was approved by the Food and Drug Administration (FDA), thus opening up a new era of manufacturing complex therapeutic proteins in mammalian cell cultures. Since then, CHO cells have remained the preferred workhorse for therapeutic protein production. Among all the approved biopharmaceutical proteins between 1982 and 2014, 35.5% are produced in CHO cells [1]. CHO cells have been preferred for various reasons including their plasticity and adaptability, which has made them easier to transfect recombinant genes and to adapt to serum-free suspension culture. Furthermore, the establishment of gene amplification systems, such as dihydrofolate reductase (DHFR)-mediated or glutamine synthetase (GS)-mediated gene amplification, has enabled the acquisition of higher transgene copy numbers and protein titers [2]. Finally, CHO cells are able to produce glycoproteins with human-compatible posttranslational modifications and thus generates safe and bioactive drugs [2]. 4.1.1

Cell Line Optimization

The general workflow for developing a recombinant CHO (rCHO) cell line is as follows. First, a vector containing a target transgene and selection marker is designed, cloned, and transfected into CHO cells. Then, the cells expressing the target transgene are selected and amplified using corresponding drugs. There are two widely used amplification methods: the DHFR and GS systems. Specifically, methotrexate (MTX) can be applied to DHFR-deficient CHO cells and methionine sulfoximine (MSX) can be applied to GS-deficient CHO cells. These drugs can be used to amplify the copy number of the transgenes if they are accompanied by replacement copies of DHFR or GS, respectively. To ensure clonality *

Shangzhong Li and Anne Richelle contributed equally.

Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

74

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

of the cells used for production, single-cell cloning, limiting dilution methods, or high-throughput cell screening systems such as fluorescence-activated cell sorting (FACS) can be performed [2, 3]. Clones are subsequently tested to ensure that they stably produce high titers of recombinant protein before large-scale production. The entire process can be time-consuming (6–12 months). To speed up the timeline and enhance the productivity, great efforts have been made over the past few decades to optimize different aspects of the process. • Vector optimization: People have tried multiple vector designs to enhance the production. Internal ribosome entry sites (IRESs) are used to put multiple genes into one transcript and activate their transcription using a single promoter. This enables fast clone generation, as shown previously for expressing monoclonal antibodies [4]. Expression can be further enhanced using matrix attachment regions (MARs) to take advantage of the sequence to which the nuclear matrix attaches to form a transcriptionally active chromatin structure [5]. • Environment optimization: The bioprocess environment, outside the cells, also affects production. For example, low culture temperature can reduce the growth rate and increase the productivity of CHO cells [6, 7]. Chemical compounds, such as sodium butyrate (NaBu), can improve the productivity [8]. These approaches can be combined to further improve the production [9]. • Engineering cell lines [10]: The previous optimization methods change the outside factors, but cell line engineering provides additional opportunities to control product quality and quantity. Such efforts have included approaches to increase the length of high-production phases. For example, the B-cell lymphoma 2 (BCL2) gene regulates cell death by inhibiting cell apoptosis. Overexpressing BCL2 successfully increased viable cell density (VCD), which means more cells are producing proteins [11]. Ammonia and lactate are two waste metabolites produced together with interest proteins by the cells. They have negative effect on cell growth and product quality. Therefore, engineering techniques have been used to reduce the expression of lactate dehydrogenase-A (LDH-A), which included an improvement on the mAb galactosylation by reducing the production of ammonia and lactate [12]. Through extensive optimization using the strategies above, CHO cell productivity has increased substantially over the past three decades [13]. However, many opportunities remain to enhance productivity and control product quality. Indeed, many opportunities for engineering CHO cells remain and we are just beginning to unravel the complexity of cellular pathways. Previous efforts aimed to optimize the performance of single factors, but an understanding of all cell pathways would enable the engineering of multiple connected subsystems. Here, we describe the CHO metabolic network as an example of one such system. The fundamental theory behind the optimization is enriching the metabolites that are good for production and reducing the metabolites that inhibit recombinant protein production. There are thousands of metabolites in the cells being

4.1 Introduction

produced and degraded at the same time. Thus, to address this complexity, we need to consider the whole system and completely delineate the effect of different factors on the system. 4.1.2

CHO Genome

Even though CHO cells have been studied for over half a century, our knowledge of their composition is still limited. However, recent studies are elucidating the inner workings of the cell. 4.1.2.1

Development of Genomic Resources of CHO

The first draft genome of CHO-K1 was published in 2011 [14]. This resource has enabled the use of a wide range of new technologies (e.g. gene editing) and high throughput omics technologies to study the cell lines systematically. For example, RNA-Seq and DNA-Seq data can now be mapped to the reference genome to identify mutations and quantify transcripts, and the sequence can also be used to identify peptides based on MS–MS fragment patterns. Thus, many new types of specific biological questions can be addressed [15]. Since the initial publication of the CHO-K1 genome, several subsequent sequencing efforts have aimed to provide additional insights into the structure and evolution of the CHO genomes. The instability of the CHO genome has resulted in many genome rearrangements and heterogeneity [16] in the cell lines derived from the original CHO cell line [17, 18]. To study the variability of the CHO cell lines and their relations with the original Chinese hamster, the genomes of several commonly used CHO cell lines and the Chinese hamster were sequenced in 2013 [19]. Many mutations identified in genes related to bioprocessing and apoptosis pathway were studied in detail using the assembled Chinese hamster genome. Concurrently, another effort was able to assign the genomic sequence to chromosomes by extracting chromosomes separately and sequencing them using Roche/454 and Illumina technologies [20]. Subsequently, CHO-DXB11, the first cell line used for large-scale protein production, was sequenced and provided further insights into gene copy numbers and relative stability of different chromosomes [21]. The recent development of third-generation sequencing platforms such as Pacific Biosciences and Oxford Nanopore sequencing promise to provide a more accurate reference genome sequence for CHO, as these technologies extend the read length to several thousand base pairs and more effectively fill in sequence gaps in repetitive regions. Indeed, this can be achieved using hybrid genome assembly techniques by combining next- and third-generation sequencing technologies [22, 23]. 4.1.2.2

Development of Transcriptomics and Proteomics Resources of CHO

Without other sources of evidence, genome assemblies are usually functionally annotated using homology methods. Studies on transcript and protein levels can directly provide experimental validation for the genome annotation. The first large-scale transcriptomic study on CHO built a coexpression network

75

76

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

using ∼300 microarray datasets [24]. Another study built a CHO transcript database using RNA-Seq from multiple CHO cell lines under different conditions. About 65561 transcripts were assembled by combining Roche/454 and Illumina sequencing technologies [25]. A more recent study compared transcriptomes of multiple CHO cell lines and Chinese hamster tissues and discovered the potential effect of variable gene expression levels on metabolism and glycosylation patterns [26]. It is anticipated that over the coming years, such technologies will become common tools used to explore the molecular basis of desirable cell traits in CHO cells. Proteomic efforts for CHO cells were initially performed in relatively small scales. For example, 2D gel electrophoresis was used to identify dozens or a few hundred proteins. With the advent of mature mass spectrometry technology and with the use of the genome sequence data, it is now possible to identify thousands of proteins in a single run. The first large-scale proteomic analysis was performed in 2012, in which more than 6000 proteins were successfully identified in CHO cells [27]. Further analysis identified 1015 proteins from the supernatant of CHO cell culture [28]. These efforts are greatly extending our knowledge about CHO cells and will continue to elucidate important information about the CHO host cell proteins.

4.2 Genome-Scale Metabolic Model With the thorough information from the genome, researchers can now study the cell lines using systematic approaches. One such example is the development and use of genome-scale metabolic models (GEMs) of CHO cells [29–31]. 4.2.1

What Is a Genome-Scale Metabolic Model

The availability of a genome sequence and annotation provide information about each gene and its associated protein. Furthermore, the annotation can include additional information, such as protein function and subcellular localization. This information enables the reconstruction of all metabolic pathways. Metabolism is a set of chemical reactions that are catalyzed by the enzymes in the living organisms [32]. Reactions can be connected by shared metabolites, thus forming a metabolic pathway. Pathways are categorized based on their biological functions and localization. Pathways are not isolated, rather they work together by interacting with each other to achieve complicated biological tasks to make the cell lines grow. Combining all the pathways results in the genome-scale metabolic network reconstruction. The reconstruction can then be transformed into a mathematical model, and algorithms can be employed to predict the response of a cell line to different metabolic conditions [33]. In summary, genome-scale metabolic network reconstructions contain all metabolic pathways of a cell and can be used with diverse methods [34] to connect a cell’s genotype to its phenotype. These resources help researchers to better understand what is really happening inside cells. Knowing more about the system enables more options for the use of exciting new tools to guide cell line engineering.

4.2 Genome-Scale Metabolic Model

4.2.2

Reconstruction of GEMs

To correctly predict metabolic genotype–phenotype interactions, we first need to have a high-quality model that best represents the real state of the cell system, which is developed through a well-established process (Figure 4.1). Two key parts of a reconstruction are metabolic reactions and their gene–protein-reaction (GPR) associations. GPRs indicate which genes and their associated enzymes catalyze the corresponding reaction. The steps to construct the GEMs are as follows [35, 36]: (1) Knowledge base construction with all known metabolic information for the organism (2) Draft reconstruction (3) Curation of the reconstruction (4) Conversion to computational format (5) Model evaluation and validation 4.2.2.1

Knowledge-Based Construction

The basic unit of a genome-scale metabolic network reconstruction is the reaction. The whole network reconstruction is composed of pathways, each of which is composed of a series of connected reactions. The key features of a reaction are metabolites, enzymes, its direction, and the compartment in which the reaction occurs. Thermodynamic [37] calculations or detailed biochemical assays define the direction of the reaction. Genome assembly and annotation efforts provide the remaining information. Multiple databases have been built to store this information at different levels. For example, NCBI [38] stores information about genes and their products. Brenda [39] stores information about enzymes for each reaction. KEGG [40] stores information about pathways in each organism. MetaCyc [41] provides information about both enzymes and pathways. A detailed list of databases has been reviewed previously [32]. 4.2.2.2

Draft Reconstruction

After compiling a list of reactions that exist in an organism of interest, the next step is to integrate GPRs with reactions. To facilitate the network reconstruction, a variety of algorithms have been developed to automate network reconstruction steps. Such algorithms can identify the homologs between an organism of interest and experimentally verified data from other organisms (or the same organism, if available). The algorithms will subsequently predict the GPRs based on other manually curated models or databases. Some widely used tools are RAVEN [42], Model SEED [43], AUTOGRAPH [44], IdentiCS [45], and GEM system [46]. Among them, RAVEN can assign predicted subcellular localization to the reaction. IdentiCS and GEM System can also predict new genes in the draft GEM. 4.2.2.3

Curation of the Reconstruction

Automatic reconstruction methods build reaction lists and assign GPRs based on enzyme homology. It is important to note that such network reconstructions

77

Network reconstruction

Mathematical representation

Metabolites

Output metabolites

–1

0

0

0

0

1

0

0

0

–1

0

0

0

0

0

1

0

0

0

0

0 –1 –1

0

0

1

0

1 –1 0 0 1 –1

1 0

0

0

0

0

0

0

0

0

0

0

0

0 –1

0

1

0

0

0

0

0

0

1

0

0

0

0

0 –1

Reactions (a) Genome

(b) Metabolic network

(c) Stoichiometric matrix

System representation

Input metabolites

Cellular System

Biomass components

(d) Metabolic model

Figure 4.1 General framework for reconstructing a genome-scale metabolic model. (a) Based on the genome sequence, a genome-scale reconstruction of the metabolism of an organism of interest. This is assembled from the organism-specific parts lists (e.g. genes, proteins, metabolites, and reactions). (b) The reconstruction can be represented graphically as a metabolic network. (c) The network can be mathematically converted in a stoichiometric matrix, which contains the stoichiometric coefficients for each metabolite (row) in each reaction (column). (d) A system representation (i.e. metabolic model) of the cell can be obtained by identifying which metabolites are consumed or secreted, as well as produce biomass components that are required for cell growth (e.g. ribosomes, proteins, lipids, and nucleic acids).

4.2 Genome-Scale Metabolic Model

will require careful manual curation, or they will likely be highly inaccurate. For example, efforts must be made to ensure that organism-specific pathways are complete, and thermodynamics must be accurately considered. In addition, GPRs need to be carefully curated, as they do not always have simple relationships of one gene to one reaction. For example, there are also many other types of GPRs: one enzyme can catalyze multiple reactions or multiple enzymes may need to form a complex to catalyze one reaction. Also, multiple different enzymes could catalyze one reaction, so having any of them would be sufficient to catalyze the reaction. The information for these types of GPRs are not always consistent between different organisms, which means there is a need to manually curate them to ensure that the network reconstruction is of high quality. The best evidence is experimental validation. However, not all enzyme activities have been verified. In such cases, less confident evidence from other organisms in public databases will have to be used and noted in the model annotation. The detailed types of evidence and their confidentiality are further explained (see Table 2 in Ref. [47]). 4.2.2.4

Conversion to a Computational Format

When a high-quality genome-scale network reconstruction is obtained, it can be converted into a mathematical format to build a GEM. All of the metabolites and their reactions can be represented by a stoichiometric matrix, in which each column represents a reaction and each row represents a metabolite. The integer values inside the matrix represent the stoichiometric coefficients of each reaction in which products have positive values and reactants have negative values. The ranges of feasible reaction rates can also be defined as flux bounds. The mathematical representation can then be used to further improve the model by running tests, such as ability to produce all the components that are required for cell growth, to validate mass balance of the total network, to identify the existence of thermodynamically infeasible loops [48], etc. This can be done using many software packages [49]. 4.2.2.5

Model Validation and Evaluation

After passing basic tests, the models are subjected to the final step of validating that the model can explain real biological phenotypes, preferably based on experimental data. For example, one can predict which gene knockouts are lethal and test these with experimental knockout data. Many algorithms have been developed to validate the model, as will be discussed in the following sections. A detailed 96-step protocol for metabolic network reconstruction and GEM development has been previously published [47]. Several metabolic network reconstructions have been developed for CHO cells, but many were built without CHO-specific genomic information, as such information was not previously available. The first GEM for CHO was derived from a mouse GEM [50, 51]. This model was valuable to study the effect of metabolic switch from lactate production to consumption [52]. Another study identified growth-limiting factors of CHO cell lines in fed-batch culture [53]. Finally, a community effort was recently completed in which a consensus GEM was developed

79

80

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

for CHO, which included the connection of the metabolic reactions to the genes associated with each reaction [29].

4.3 GEM Application Numerous modeling methods exist to describe and quantify the reactions occurring in a cell under specific environments. They range from deterministic kinetic models to stochastic and statistical models. The various methods continue to allow the integration of increasing amounts of experimental data available, thanks to the emergence of the omics technologies [54]. Although kinetic modeling has been widely used for small-scale metabolic models, these mechanistic models are for the moment not scalable and suitable for genome-wide approaches because of challenges in acquiring the required kinetic parameters (i.e. rate constants, enzymes, and intracellular metabolite concentrations). In addition, it has been difficult to develop scalable algorithms to overcome problems associated with such complex parametric systems (i.e. model nonlinearity, parameter identifiability, and computational tractability) [55]. However, many significant advances have been emerging in this space, and efforts toward the construction of genome-scale kinetic models have been recently reviewed [56]. Constraint-based modeling approaches have helped to overcome some of the limitations encountered with other modeling techniques. Specifically, the constraint-based method reduces the need for detailed kinetic parameters and is frequently less computationally intensive. These advantages have enabled the analysis of genome-scale metabolic networks. This approach is based on the concept that biological systems are constrained by physicochemical laws and by their genetics and environment. Having been first applied to metabolism in the mid-1980s, this methodology allows the analysis of general characteristics of metabolism by using only the knowledge about reaction stoichiometry in the metabolic network and assuming a quasi-steady-state of the associated mass balance equation for each intracellular metabolite. Consequently, the reactions in a genome-scale metabolic network reconstruction can be described by a system of linear algebraic equations whose solution represents the feasible phenotypic states of the cell given the network topology and under the constraints that limit the functional capability of the cell [54]. The constraint-based approach relies on the conversion of the network stoichiometry into a consistent mathematical format known as the stoichiometric matrix and on the identification of the major constraints acting simultaneously on and in the cell. For a genome-scale reconstruction involving m metabolites and n reactions, the constraint-based model will be defined as follows: Nv = 0

(4.1)

where N is the stoichiometric matrix of dimension m × n and v the vector of specific reaction rates (metabolic fluxes). Actually, this equation is the differential equation representing the mass balance of each metabolite involved

4.3 GEM Application

in the network expressed under balanced growth conditions (i.e. intracellular metabolites are neither accumulated nor diluted). The linear problem defined in Eq. (4.1) is generally underdetermined as metabolic networks usually have more reactions than metabolites. Therefore, there is no unique solution to the system of linear equations but rather a region of feasible solutions (space of admissible flux distribution v). Constraints can be defined for each reaction, such as upper and lower flux bounds for each reaction, equality, and/or inequality constraints. These can represent specific intra- and extracellular conditions (e.g. maximum enzyme capacity, reaction thermodynamics, and regulatory mechanisms). These can decrease the range of feasible solutions by shrinking the solution space to a region associated with the most biologically relevant flux distribution for the observed cell phenotype under the given conditions. The most common constraint introduced in metabolic model is related to the existence of a maximum capacity for some reaction fluxes: v ≤ vmax

(4.2)

where vmax is the vector of maximum capacities (i.e. upper bounds) associated with each flux included in vector v defined by Eq. (4.2). The analysis of these kinds of systems is based on convex theory [57] and are part of the “network-based pathway analysis” (NPA) of constraint-based methods, such as “elementary fluxes” and “extreme pathways,” which are used to identify systemic properties of a network [58]. Generally, the system defined by Eqs. (4.1) and (4.2) is solved, after constraining the model with the available experimental data (e.g. the uptake and secretion rate of key metabolites). Assuming the availability of some measured fluxes (vmes ), the metabolic network model defined in Eq. (4.1) becomes )( ) ( v N 0 =0 (4.3) 1 Ne −vmes where N e is the stoichiometric matrix associated with the measured metabolite concentrations, which are considered as unbalanced conditions contrarily to intracellular ones. 13 C tracers can also be used as experimental information to constrain flux solutions [59]. However, even with the addition of constraints and the use of experimental data, models generally remain underdetermined. A common methodology that helps to overcome this problem is the linear optimization technique flux balance analysis (FBA) [60]. FBA is based on the hypothesis that cells have evolved to be optimal with respect to specific cellular objectives (e.g. biomass growth, ATP yield, or productivity). Hence, the flux distribution (v) corresponding to the defined objective function Z (i.e. optimal cellular behavior expressed as a linear combination of fluxes of v, such as the production of biomass precursors or energy equivalents) is obtained by solving the following linear programming (LP) problem: )( ) ( v N 0 =0 (4.4) max Z s.t. 1 Ne −vmes The major drawback of this method is related to the dependence between the model predictions (i.e. optimal flux distribution) and the choice of the objective

81

82

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

function. Although maximizing biomass is the most widely used cost criterion, Schuetz et al. [61] clearly highlighted the importance of identifying relevant cellular objectives with respect to environmental conditions and cell lines considered. It is important to note that if the system (4.4) is underdetermined, while there is a unique solution of this linear optimization problem (maximum value of the cost function Z), generally there exist different equivalently optimal flux distributions (v) leading to this optimum (alternate optimal solutions). In this context, the study of Lee et al. [62] was one of the first to present a procedure allowing a rigorous determination of all possible combinations of steady-state reaction fluxes satisfying an underdetermined linear equation system. They used a recursive mixed-integer linear programming (MILP) algorithm to enumerate all multiple alternate optimal solutions of an FBA problem. However, the MILP algorithm is computationally expensive for genome-scale networks [63]. To this end, flux variability analysis (FVA) [64, 65] determines the range of admissible values for each flux instead of identifying all possible alternate optimal solutions. Specifically, FVA computes the minimum and maximum flux values possible for each reaction while ensuring the same objective capacity. Wiback et al. [66] introduced an equivalent concept, called the “alpha-spectrum,” which is related to the definition of the overall solution space of a linear equation system by using NPA methods (i.e. range of possible values for each elementary mode or extreme pathway activity). The FVA formulation of the system described by (4.3) is related to a double LP problem (i.e. a maximization and a subsequent minimization): ( )( ) vi,upper = max(vi ) N 0 v s.t. =0 ∀vi , i = 1, … , n Ne −vmes 1 vi,lower = min(vi ) (4.5) where n is the number of fluxes in v, vi,upper and vi,lower are, respectively, the upper and lower values of each flux vi satisfying the system of linear Eq. (4.3). Beyond FVA, a few hundred other constraint-based modeling algorithms have been developed and deployed [34]. Many of these methods have been implemented in the COBRA toolbox [67]. 4.3.1 Common Usage and Prediction Capacities of Genome-Scale Models In the past, our knowledge of cell metabolism was mainly achieved based on experimental practice through cell culture observations. The development of modeling approaches has opened up ways to systematize the investigation of the mechanisms underlying the functions of biological systems. Indeed, mathematical models are platforms to integrate and organize the increasing amount of information we have access to either at the level of metabolites involved in metabolic reactions or the enzymes catalyzing these reactions. These predictive tools can be used to investigate the cellular processes and their connection to phenotype acquisition through the simulation and analysis of metabolic responses of the cell in function of modifications observed in its environment.

4.3 GEM Application

As the full genome-scale reconstruction of CHO cell metabolism was only recently available [29], most of the computational studies performed up to date with this organism were done using small-scale models (i.e. models involving a reduced set of reactions principally related to the central metabolism: glycolysis, Krebs cycle, pentose phosphate pathway, and amino acid metabolism [52, 53, 68]). However, GEMs and associated constraint-based tools have been applied in cell engineering and biomedical applications to perform different predictions with a multitude of other organisms, allowing us to have a glimpse of future opportunities opening in the CHO context. Among them, the most common prediction capacities of GEMs are maximum theoretical yield (i.e. percentage of conversion of a substrate in a targeted molecule), the influence of gene addition or deletion (i.e. prediction of gene knockout mutant phenotypes), and analysis of up- and downregulation of targeted reactions depending on the environment to identify potential drug targets by evaluating the effect of potential disease treatments or to study the required modification to improve the production of a specific metabolite [69] or protein production [33].

4.3.2 GEMs as a Platform for Omics Data Integration, Linking Genotype to Phenotype GEMs can be used to link genes, enzymes, and metabolites through the use of GPRs and therefore allow researchers to explain the metabolic state of a cell, based on the expression of metabolic genes. With the knowledge captured by the GPRs, GEMs can be used to integrate diverse omics data types, including proteomics, transcriptomics, metabolomics, fluxomics, lipidomics, phenotype microarrays, and glycomics that are now emerging as experimental routines, thanks to the innovations in these fields [70, 71]. Despite being relatively new, omics technologies have rapidly emerged as valuable tools in the study and engineering of CHO cells [15, 72, 73]. Indeed, omics data can provide a more in-depth understanding of the molecular basis of the physiological differences among cell lines, as they provide insights into how cells respond to an environmental change (growth media, feeding strategies, and process conditions) and obtain a phenotype of interest [15, 30, 55, 74, 75]. Numerous computational approaches have been developed to integrate diverse types of omics data [76–79]. Among the constraint-based methods, context-specific extraction algorithms, such as mCADRE [80], iMAT [81], and GIMME [82] and many others have been developed [82, 83]. These model extraction methods (MEMs) use gene expression data to recapitulate the metabolism of an organism under a specific condition (e.g. specific cell line or tissue) by extracting the subset of reactions from the GEM that is active [83, 84]. These methods have proven valuable and been able to enhance the accuracy of model-predicted growth rates and gene essentiality. However, as currently no quantitative description of the GPR relationship exists in GEMs, the integration of gene expression data requires the use of strong assumptions to link the GPR expression and the metabolic reaction activity, which could lead to an over-simplification of the complex relation existing between fluxes,

83

84

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

enzymes, and genes [85]. Thus, further development of more precise methods would further improve the predictions of cell-line-specific models. Although multiple data types have been used to assess the influence of CHO cell genetics on the protein production capacity [24, 86–89], integration of these high-throughput data within a mathematical framework is only now emerging because of the novelty of the genomic sequences and the GEM for CHO cells. However, the further development of methods in computational systems biology will provide a unique opportunity to obtain a holistic and integrated picture of the gene expression and cell culture environment control on cell metabolism. This is enabling the shift from an empirical research to a more rational approach for bioprocess development and optimization. 4.3.3

Predicting Nutrient Consumption and Controlling Phenotype

For decades, random mutagenesis has been used to find cell factories with desired phenotypes. With the emergence of genetic tools, more targeted cell engineering strategies have been applied to control cellular attributes of interest. These engineering strategies have been mainly performed at a level of energy metabolism to reduce the accumulation of toxic by-products and/or increase metabolic efficiency [10]. Now, the emerging mathematical metabolic models can further guide cellular engineering. Indeed, GEM-based analysis may be used to predict potential targets for genetic modification to obtain beneficial phenotypes [33, 74, 90, 91]. With the guidance of computational approaches and the development of multiplex genome editing techniques [92–94], new concepts of combinatorial [2] and dynamic [95, 96] engineering are now emerging. They will allow researchers to move toward the design and implementation of more complex genetic modifications, accounting for the multifaceted interaction of the diverse cellular processes occurring simultaneously within a cell. To date, most improvements in mammalian cell cultures have been achieved by empirical media and bioprocess optimization. The current optimization practice is often intended to determine ideal feeding strategies in culture media over time, along with process parameters. This is done by screening different media and process conditions and analysis using statistical design-of-experiment strategies. The use of mechanistic mathematical models allows one to predict cell metabolic behavior under different media compositions and process conditions, which in turn can be used to objectively determine the optimal operating conditions with respect to the desired cell phenotype and related production criteria. Moreover, computational approaches allow the identification of nonintuitive genetic interactions that in vivo experiments alone cannot provide. These new hypotheses can guide experimental design and therefore facilitate the improvement of bioprocess performances [33, 55, 56]. CHO cell production processes are usually associated with feeding of glucose and glutamine at high concentrations, and this can lead to excessive lactate and ammonia secretion. The accumulation of these by-products inhibits cell growth and is often associated with a lower productivity. Strategies such as rational

4.3 GEM Application

supplementation of substrates, use of alternative carbon sources, and metabolic engineering of cell lines have been used to overcome the accumulation of toxic by-products (i.e. lactate and ammonia) [10]. However, we still do not fully understand the cellular mechanisms underlying these events and how they are connected to process conditions. To this end, several studies have successfully used mathematical models to predict the phenotypic differences occurring due to variations in growth media, feeding strategies, and operating conditions and doing so rationally guide the culture process design [52, 53, 97–99]. This emphasizes the usefulness of computational approaches to address pertinent questions on how to optimally conduct CHO cultures for protein production. In this context, the CHO GEM will be a valuable tool by providing a holistic understanding of the cellular basis of the key metabolic events for both growth and productivity.

4.3.4

Enhancing Protein Production and Bioprocesses

In past decades, CHO cells have been used to produce numerous protein-based drugs that treat complex human diseases (e.g. inflammatory disorders, cancer, and infectious diseases) [1]. However, it is still unclear how to obtain the ideal CHO cell line [100, 101], as we lack knowledge of the interconnections between cell pathways, physiological functions, and process conditions and their influence on the acquisition of known quality attributes [10]. In the industrial context of protein production, high volumetric productivity and product titers are required to obtain affordable protein therapeutics. Most improvements in this perspective have been obtained by empirical bioprocess optimization and random mutagenesis [102, 103]. Despite successful efforts in these fields, several protein-based therapeutics remain difficult to be expressed at commercially enabling productivities and/or appropriate product quality. However, the availability of high-throughput data and the emergence of genome editing tools provide novel opportunities for targeted genome engineering of the host cell. Also, efforts to further enhance drug production will be facilitated as the molecular basis of these processes are studied and linked to protein production. GEMs and associated modeling techniques are starting points to elucidate and quantify the inherent variability of CHO cell metabolism and protein production. Furthermore, recent developments in constraint-based modeling can now account simultaneously for metabolism, gene expression, and protein synthesis. These models have been developed in microbes to explore how protein abundances affect cell physiology and protein synthesis capacity [104–107]. Future efforts will extend these strategies to include pathways specific to protein processing and posttranslational modifications (e.g. glycosylation [107, 108]) relevant to mammalian cells. These evolving computational approaches will be instrumental to identify new process conditions and cell engineering strategies that control product quantity and quality attributes such as glycosylation. This will enable to identify the safest, most affordable, and efficacious protein therapeutics [10, 33, 55, 109, 110].

85

86

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

4.3.5

Case Studies

GEMs of CHO cell metabolism have been used for a handful of applications in the recent past (Figure 4.2). In 1999, Nyberg et al. [68] performed the first study using a modeling approach to analyze the utilization of peptide-derived amino acids by CHO cells during glucose-limited continuous cultures with varying dilution rates. They constructed a simplified metabolic network containing 33 reactions related to central carbon metabolism and assessed the metabolic significance of accounting for peptides-derived amino acids. They showed that the knowledge of the amount of amino acids liberated from peptides was essential to properly account for metabolite uptake and secretion and to ultimately match the experimental data. Although this first study relies on a simplistic model, it already highlighted the usefulness of mechanistic models for unraveling the relationship between metabolism and nutrient status of the cellular environment. Over the years, an increasing amount of experimental data has become available because of innovations of analytical techniques used to collect and process experimental data. These data have allowed researchers to go deeper into the characterization of CHO cell physiology. In this context, Selvarasu et al. [53] presented an integrated framework combining fed-batch culture data, metabolomics, and in silico metabolic network modeling to assess the pathways related to growth limitation. They constructed a metabolic network involving 1540 enzymatic reactions based on a genome-scale mouse metabolic model, cDNA annotation, and metabolomics data and used it to determine the internal metabolic behaviors attaining physiological changes during growth and nongrowth phases. Their strategy enabled the identification of major growth-limiting factors including the oxidative stress and depletion of lipid metabolites. Recently, Yusufi et al. [30] have integrated multi-omics profiling (i.e. genomic, transcriptomic, metabolomic, and glycomic levels) with the CHO GEM to systematically explore the impact of cell line engineering on CHO cell biology and protein production. Doing so, they demonstrated that global cellular adaptations occur at multiple levels (DNA damage and repair, mitochondrial energy, and membrane lipid metabolism) following transgene integration in the recombinant cell line. This study enables the global assessment of biological traits of the producing CHO cell line and represents a significant step toward the elucidation of the interconnection between the multiple layers of cellular hierarchy, from genomic to metabolic. These integrated systems biology approaches are key tools to leverage the cellular basis for high productivity and improve mammalian therapeutic production factories.

4.4 Conclusion As demonstrated here, the recent efforts to sequence the Chinese hamster genome have enabled researchers to obtain, for the first time, a comprehensive view of the metabolic pathways in CHO cells. Furthermore, advances in genome-scale modeling now provide a toolbox for the use of modeling

Intensity

Substrates Elution time

Metabolome

Gene

Products

Intensity

Sample

Transcriptome

m/z

Proteome (a)

(b)

(c)

Figure 4.2 Example uses and prediction capabilities of genome-scale models. (a) GEMs can be used as a platform for omics data integration (e.g. metabolomic, proteomic, and transcriptomic data). (b) GEMs allow the prediction of nutrient consumption in specific conditions and doing so can be used to aid in media optimization or to control cell phenotypes. (c) GEMs aid in efforts to enhance protein production and optimize bioprocesses.

88

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

techniques for data analysis, thus potentially guiding efforts to optimize cell culture bioprocesses, identify open genomic loci for efficient transgene integration [111] and guide cell line engineering to enhance protein production.

Acknowledgments This work was supported by funding from funding from the NIGMS (R35 GM119850) and the Novo Nordisk Foundation through the Center for Biosustainability at the Technical University of Denmark (NNF10CC1016517), and A.R. was funded by a Lilly Innovation Fellowship Award.

References 1 Walsh, G. (2014). Biopharmaceutical benchmarks 2014. Nat. Biotechnol. 32:

992–1000. 2 Kim, J.Y., Kim, Y.-G., and Lee, G.M. (2012). CHO cells in biotechnology for

3

4

5

6

7

8

9

10

production of recombinant proteins: current state and further potential. Appl. Microbiol. Biotechnol. 93: 917–930. Lee, J.S., Park, J.H., Ha, T.K. et al. (2018). Revealing key determinants of clonal variation in transgene expression in recombinant CHO cells using targeted genome editing. ACS Synth. Biol. 7 (12): 2867–2878. https://doi.org/ 10.1021/acssynbio.8b00290. Ho, S.C.L., Bardor, M., Feng, H. et al. (2012). IRES-mediated Tricistronic vectors for enhancing generation of high monoclonal antibody expressing CHO cell lines. J. Biotechnol. 157: 130–139. Kim, J.-M., Kim, J.-S., Park, D.-H. et al. (2004). Improved recombinant gene expression in CHO cells using matrix attachment regions. J. Biotechnol. 107: 95–105. Clark, K.J.R., Chaplin, F.W.R., and Harcum, S.W. (2004). Temperature effects on product-quality-related enzymes in batch CHO cell cultures producing recombinant tPA. Biotechnol. Prog. 20: 1888–1892. Furukawa, K. and Ohsuye, K. (1998). Effect of culture temperature on a recombinant CHO cell line producing a C-terminal 𝛼-amidating enzyme. Cytotechnology 26: 153–164. Sung, Y.H., Song, Y.J., Lim, S.W. et al. (2004). Effect of sodium butyrate on the production, heterogeneity and biological activity of human thrombopoietin by recombinant Chinese hamster ovary cells. J. Biotechnol. 112: 323–335. Chen, F., Kou, T., Fan, L. et al. (2011). The combined effect of sodium butyrate and low culture temperature on the production, sialylation, and biological activity of an antibody produced in CHO cells. Biotechnol. Bioprocess Eng. 16: 1157–1165. Richelle, A. and Lewis, N.E. (2017). Improvements in protein production in mammalian cells from targeted metabolic engineering. Curr. Opin. Syst. Biol. 6: 1–6.

References

11 Simpson, N.H. and Al-Rubeai, M. (1997). Overexpression of the

12

13

14 15

16

17

18 19

20 21

22

23

24

25

apoptosis suppressor gene, bcl-2, in hybridoma cells prolongs viability and increases productivity in batch culture. In: Animal Cell Technology (eds. M.J.T. Carrondo, B. Griffiths and J.L.P. Moreira), 613–620. Dordrecht: Springer. Noh, S.M., Park, J.H., Lim, M.S. et al. (2016). Reduction of ammonia and lactate through the coupling of glutamine synthetase selection and downregulation of lactate dehydrogenase-A in CHO cells. Appl. Microbiol. Biotechnol. 101: 1035–1045. Jayapal, K.P., Wlaschin, K.F., Hu, W., and Yap, M.G. (2007). Recombinant protein therapeutics from CHO cells – 20 years and counting. Chem. Eng. Prog. 103: 40. Xu, X., Nagarajan, H., Lewis, N.E. et al. (2011). The genomic sequence of the Chinese hamster ovary (CHO)-K1 cell line. Nat. Biotechnol. 29: 735–741. Kildegaard, H.F., Baycin-Hizal, D., Lewis, N.E., and Betenbaugh, M.J. (2013). The emerging CHO systems biology era: harnessing the ’omics revolution for biotechnology. Curr. Opin. Biotechnol. 24: 1102–1107. Cao, Y., Kimura, S., Itoi, T. et al. (2012). Construction of BAC-based physical map and analysis of chromosome rearrangement in Chinese hamster ovary cell lines. Biotechnol. Bioeng. 109: 1357–1367. Puck, T.T. (1958). Genetics of somatic mammalian cells: III. Long-term cultivation of euploid cells from human and animal subjects. J. Exp. Med. 108: 945–956. Wurm, F.M. and Hacker, D. (2011). First CHO genome. Nat. Biotechnol. 29: 718–720. Lewis, N.E., Liu, X., Li, Y. et al. (2013). Genomic landscapes of Chinese hamster ovary cell lines as revealed by the Cricetulus griseus draft genome. Nat. Biotechnol. 31: 759–765. Brinkrolf, K., Rupp, O., Laux, H. et al. (2013). Chinese hamster genome sequenced from sorted chromosomes. Nat. Biotechnol. 31: 694–695. Kaas, C.S., Kristensen, C., Betenbaugh, M.J., and Andersen, M.R. (2015). Sequencing the CHO DXB11 genome reveals regional variations in genomic stability and haploidy. BMC Genomics 16: 160. Mostovoy, Y., Levy-Sakin, M., Lam, J. et al. (2016). A hybrid approach for de novo human genome sequence assembly and phasing. Nat. Methods 13: 587–590. Koren, S., Schatz, M.C., Walenz, B.P. et al. (2012). Hybrid error correction and de novo assembly of single-molecule sequencing reads. Nat. Biotechnol. 30: 693–700. Clarke, C., Doolan, P., Barron, N. et al. (2011). Large scale microarray profiling and coexpression network analysis of CHO cells identifies transcriptional modules associated with growth and productivity. J. Biotechnol. 155: 350–359. Rupp, O., Becker, J., Brinkrolf, K. et al. (2014). Construction of a public CHO cell line transcript database using versatile bioinformatics analysis pipelines. PLoS One 9: e85568.

89

90

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

26 Vishwanathan, N., Yongky, A., Johnson, K.C. et al. (2015). Global insights

27 28

29

30

31

32

33

34

35 36

37 38 39 40 41

42

into the Chinese hamster and CHO cell transcriptomes. Biotechnol. Bioeng. 112: 965–976. Baycin-Hizal, D., Tabb, D.L., Chaerkady, R. et al. (2012). Proteomic analysis of Chinese hamster ovary cells. J. Proteome Res. 11: 5265–5276. Kumar, A., Baycin-Hizal, D., Wolozny, D. et al. (2015). Elucidation of the CHO super-ome (CHO-SO) by proteoinformatics. J. Proteome Res. 14: 4687–4703. Hefzi, H., Ang, K.S., Hanscho, M. et al. (2016). A consensus genome-scale reconstruction of Chinese hamster ovary cell metabolism. Cell Syst. 3: 434–443.e8. Yusufi, F.N.K., Lakshmanan, M., Ho, Y.S. et al. (2017). Mammalian systems biotechnology reveals global cellular adaptations in a recombinant CHO cell line. Cell Syst. 4: 530–542.e6. Rejc, Ž., Magdevska, L., Tršeliˇc, T. et al. (2017). Computational modelling of genome-scale metabolic networks and its application to CHO cell cultures. Comput. Biol. Med. 88: 150–160. Baart, G.J.E. and Martens, D.E. (2012). Genome-scale metabolic models: reconstruction and analysis. In: Methods in Molecular BiologyNeisseria Meningitidis, Methods in Molecular Biology (Methods and Protocols), vol. 799 (ed. M. Christodoulides), 107–126. Humana Press. Gutierrez, J.M. and Lewis, N.E. (2015). Optimizing eukaryotic cell hosts for protein production through systems biotechnology and genome-scale modeling. Biotechnol. J. 10: 939–949. Lewis, N.E., Nagarajan, H., and Palsson, B.O. (2012). Constraining the metabolic genotype-phenotype relationship using a phylogeny of in silico methods. Nat. Rev. Microbiol. 10: 291–305. Feist, A.M., Herrgård, M.J., Thiele, I. et al. (2009). Reconstruction of biochemical networks in microorganisms. Nat. Rev. Microbiol. 7: 129–143. Lewis, N.E., Jamshidi, N., Thiele, I., and Palsson, B.Ø. (2009). Metabolic systems biology. In: Encyclopedia of Complexity and Systems Science (ed. R. Meyers). New York, NY: Springer. Flamholz, A., Noor, E., Bar-Even, A., and Milo, R. (2012). eQuilibrator – the biochemical thermodynamics calculator. Nucleic Acids Res. 40: D770–D775. NCBI Resource Coordinators (2017). Database resources of the National Center for Biotechnology Information. Nucleic Acids Res. 45: D12–D17. Scheer, M., Grote, A., Chang, A. et al. (2011). BRENDA, the enzyme information system in 2011. Nucleic Acids Res. 39: D670–D676. Kanehisa, M. (2000). KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 28: 27–30. Caspi, R., Billington, R., Ferrer, L. et al. (2016). The MetaCyc database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databases. Nucleic Acids Res. 44: D471–D480. Agren, R., Liu, L., Shoaie, S. et al. (2013). The RAVEN toolbox and its use for generating a genome-scale metabolic model for Penicillium chrysogenum. PLoS Comput. Biol. 9: e1002980.

References

43 Devoid, S., Overbeek, R., DeJongh, M. et al. (2013). Automated genome

44 45

46

47 48 49 50

51

52

53

54

55

56

57 58

annotation and metabolic model reconstruction in the SEED and Model SEED. In: Methods in Molecular Biology Systems Metabolic Engineering, Methods in Molecular Biology (Methods and Protocols), vol. 985 (ed. H. Alper), 17–45. Totowa, NJ: Humana Press. Notebaart, R.A., van Enckevort, F.H.J., Francke, C. et al. (2006). Accelerating the reconstruction of genome-scale metabolic networks. BMC Bioinf. 7: 296. Sun, J. and Zeng, A.-P. (2004). IdentiCS – identification of coding sequence and in silico reconstruction of the metabolic network directly from unannotated low-coverage bacterial genome sequence. BMC Bioinf. 5: 112. Arakawa, K., Yamada, Y., Shinoda, K. et al. (2006). GEM System: automatic prototyping of cell-wide metabolic pathway models from genomes. BMC Bioinf. 7: 168. Thiele, I. and Palsson, B.Ø. (2010). A protocol for generating a high-quality genome-scale metabolic reconstruction. Nat. Protoc. 5: 93–121. Noor, E., Lewis, N.E., and Milo, R. (2012). A proof for loop-law constraints in stoichiometric metabolic networks. BMC Syst. Biol. 6: 140. Lakshmanan, M., Koh, G., Chung, B.K.S., and Lee, D.-Y. (2014). Software applications for flux balance analysis. Briefings Bioinf. 15: 108–122. Quek, L.-E. and Nielsen, L.K. (2008). On the reconstruction of the Mus musculus genome-scale metabolic network model. In: Genome Informatics 2008, Proceedings of the 19th International Conference on Genome Informatics, Gold Coast, Queensland, Australia, 1–3 December 2008, vol. 21 (eds. J. Arthur and S.-K. Ng), 89–100. Selvarasu, S., Karimi, I.A., Ghim, G.-H., and Lee, D.-Y. (2010). Genome-scale modeling and in silico analysis of mouse cell metabolic network. Mol. Biosyst. 6: 152–161. Martínez, V.S., Dietmair, S., Quek, L.-E. et al. (2013). Flux balance analysis of CHO cells before and after a metabolic switch from lactate production to consumption. Biotechnol. Bioeng. 110: 660–666. Selvarasu, S., Ho, Y.S., Chong, W.P.K. et al. (2012). Combined in silico modeling and metabolomics analysis to characterize fed-batch CHO cell culture. Biotechnol. Bioeng. 109: 1415–1429. Bordbar, A., Monk, J.M., King, Z.A., and Palsson, B.O. (2014). Constraint-based models predict metabolic and associated cellular functions. Nat. Rev. Genet. 15: 107–120. Galleguillos, S.N., Ruckerbauer, D., Gerstl, M.P. et al. (2017). What can mathematical modelling say about CHO metabolism and protein glycosylation? Comput. Struct. Biotechnol. J. 15: 212–221. Srinivasan, S., Cluett, W.R., and Mahadevan, R. (2015). Constructing kinetic models of metabolism at genome-scales: a review. Biotechnol. J. 10: 1345–1359. Pryce, J.D. (1973). R. Tyrell Rockafellar, Convex Analysis (Princeton University Press, 1970), xviii 451 pp. Proc. Edinb. Math. Soc. 18: 339. Papin, J.A., Stelling, J., Price, N.D. et al. (2004). Comparison of networkbased pathway analysis methods. Trends Biotechnol. 22: 400–405.

91

92

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

59 Niklas, J. and Heinzle, E. (2011). Metabolic flux analysis in systems biol-

60 61

62

63

64

65

66

67

68

69

70

71 72

73

74

ogy of mammalian cells. In: Genomics and Systems Biology of Mammalian Cell Culture, Advances in Biochemical Engineering Biotechnology, vol. 127 (eds. W. Hu and A.P. Zeng), 109–132. Berlin, Heidelberg: Springer-Verlag. Orth, J.D., Thiele, I., and Palsson, B.Ø. (2010). What is flux balance analysis? Nat. Biotechnol. 28: 245–248. Schuetz, R., Kuepfer, L., and Sauer, U. (2007). Systematic evaluation of objective functions for predicting intracellular fluxes in Escherichia coli. Mol. Syst. Biol. 3: 119. Lee, S., Phalakornkule, C., Domach, M.M., and Grossmann, I.E. (2000). Recursive MILP model for finding all the alternate optima in LP models for metabolic networks. Comput. Chem. Eng. 24: 711–716. Reed, J.L. and Palsson, B.Ø. (2004). Genome-scale in silico models of E. coli have multiple equivalent phenotypic states: assessment of correlated reaction subsets that comprise network states. Genome Res. 14: 1797–1805. Burgard, A.P., Vaidyaraman, S., and Maranas, C.D. (2001). Minimal reaction sets for Escherichia coli metabolism under different growth requirements and uptake environments. Biotechnol. Prog. 17: 791–797. Mahadevan, R. and Schilling, C.H. (2003). The effects of alternate optimal solutions in constraint-based genome-scale metabolic models. Metab. Eng. 5: 264–276. Wiback, S.J., Mahadevan, R., and Palsson, B.Ø. (2003). Reconstructing metabolic flux vectors from extreme pathways: defining the alpha-spectrum. J. Theor. Biol. 224: 313–324. Heirendt, L., Arreckx, S., Pfau, T. et al. (2019). Creation and analysis of biochemical constraint-based models: the COBRA Toolbox v3.0. Nat. Protoc. 14: 639–702. Nyberg, G.B., Robert Balcarcel, R., Follstad, B.D. et al. (1999). Metabolism of peptide amino acids by Chinese hamster ovary cells grown in a complex medium. Biotechnol. Bioeng. 62: 324–335. Maia, P., Rocha, M., and Rocha, I. (2015). In silico constraint-based strain optimization methods: the quest for optimal cell factories. Microbiol. Mol. Biol. Rev. 80: 45–67. Lewis, N.E., Cho, B.-K., Knight, E.M., and Palsson, B.O. (2009). Gene expression profiling and the use of genome-scale in silico models of Escherichia coli for analysis: providing context for content. J. Bacteriol. 191: 3437–3444. Joyce, A.R. and Palsson, B.Ø. (2006). The model organism as a system: integrating “omics” data sets. Nat. Rev. Mol. Cell Biol. 7: 198–210. Lewis, A.M., Abu-Absi, N.R., Borys, M.C., and Li, Z.J. (2015). The use of ‘Omics technology to rationally improve industrial mammalian cell line performance. Biotechnol. Bioeng. 113: 26–38. Vivek-Ananth, R.P. and Samal, A. (2016). Advances in the integration of transcriptional regulatory information into genome-scale metabolic models. Biosystems 147: 1–10. Zhang, C. and Hua, Q. (2015). Applications of genome-scale metabolic models in biotechnology and systems medicine. Front. Physiol. 6: 413.

References

75 Farrell, A., McLoughlin, N., Milne, J.J. et al. (2014). Application of multi-

76

77

78 79

80

81 82 83

84

85

86

87

88

89

90

omics techniques for bioprocess design and optimization in Chinese hamster ovary cells. J. Proteome Res. 13: 3144–3159. García Martín, H., Martín, H.G., Kumar, V.S. et al. (2015). A method to constrain genome-scale models with 13 C labeling data. PLoS Comput. Biol. 11: e1004363. Aurich, M.K., Fleming, R.M.T., and Thiele, I. (2016). MetaboTools: a comprehensive toolbox for analysis of genome-scale metabolic models. Front. Physiol. 7: 327. Berger, B., Peng, J., and Singh, M. (2013). Computational solutions for omics data. Nat. Rev. Genet. 14: 333–346. Bersanelli, M., Mosca, E., Remondini, D. et al. (2016). Methods for the integration of multi-omics data: mathematical aspects. BMC Bioinf. 17 (Suppl. 2): 15. Wang, Y., Eddy, J.A., and Price, N.D. (2012). Reconstruction of genome-scale metabolic models for 126 human tissues using mCADRE. BMC Syst. Biol. 6: 153. Zur, H., Ruppin, E., and Shlomi, T. (2010). iMAT: an integrative metabolic analysis tool. Bioinformatics 26: 3140–3142. Becker, S.A. and Palsson, B.O. (2008). Context-specific metabolic networks are consistent with experiments. PLoS Comput. Biol. 4: e1000082. Opdam, S., Richelle, A., Kellman, B. et al. (2017). A systematic evaluation of methods for tailoring genome-scale metabolic models. Cell Syst. 4: 318–329.e6. Richelle, A., Chiang, A.W.T., Kuo, C.-C., and Lewis, N.E. (2019). Increasing consensus of context-specific metabolic models by integrating data-inferred cell functions. PLoS Comput. Biol. 15 (4): e1006867. https://doi.org/10.1371/ journal.pcbi.1006867. Machado, D. and Herrgård, M. (2014). Systematic evaluation of methods for integration of transcriptomic data into constraint-based models of metabolism. PLoS Comput. Biol. 10: e1003580. Yee, J.C., Gerdtzen, Z.P., and Hu, W.-S. (2009). Comparative transcriptome analysis to unveil genes affecting recombinant protein productivity in mammalian cells. Biotechnol. Bioeng. 102: 246–263. Ghorbaniaghdam, A., Chen, J., Henry, O., and Jolicoeur, M. (2014). Analyzing clonal variation of monoclonal antibody-producing CHO cell lines using an in silico metabolomic platform. PLoS One 9: e90832. Carlage, T., Hincapie, M., Zang, L. et al. (2009). Proteomic profiling of a high-producing Chinese hamster ovary cell culture. Anal. Chem. 81: 7357–7362. Maccani, A., Hackl, M., Leitner, C. et al. (2014). Identification of microRNAs specific for high producer CHO cell lines using steady-state cultivation. Appl. Microbiol. Biotechnol. 98: 7535–7548. Kim, B., Kim, W.J., Kim, D.I., and Lee, S.Y. (2014). Applications of genome-scale metabolic network model in metabolic engineering. J. Ind. Microbiol. Biotechnol. 42: 339–348.

93

94

4 Enhancing Product and Bioprocess Attributes Using Genome-Scale Models

91 Long, M.R., Ong, W.K., and Reed, J.L. (2015). Computational methods in

metabolic engineering for strain design. Curr. Opin. Biotechnol. 34: 135–141. 92 Yang, Z., Wang, S., Halim, A. et al. (2015). Engineered CHO cells for

93

94

95

96

97

98

99

100

101 102 103

104

105

production of diverse, homogeneous glycoproteins. Nat. Biotechnol. 33: 842–844. Grav, L.M., Lee, J.S., Gerling, S. et al. (2015). One-step generation of triple knockout CHO cell lines using CRISPR/Cas9 and fluorescent enrichment. Biotechnol. J. 10: 1446–1456. Fischer, S., Handrick, R., Aschrafi, A., and Otte, K. (2015). Unveiling the principle of microRNA-mediated redundancy in cellular pathway regulation. RNA Biol. 12: 238–247. Brockman, I.M. and Prather, K.L.J. (2015). Dynamic metabolic engineering: new strategies for developing responsive cell factories. Biotechnol. J. 10: 1360–1369. Le, H., Vishwanathan, N., Kantardjieff, A. et al. (2013). Dynamic gene expression for metabolic engineering of mammalian cells in culture. Metab. Eng. 20: 212–220. Zhou, M., Crawford, Y., Ng, D. et al. (2011). Decreasing lactate level and increasing antibody production in Chinese hamster ovary cells (CHO) by reducing the expression of lactate dehydrogenase and pyruvate dehydrogenase kinases. J. Biotechnol. 153: 27–34. Selvarasu, S., Kim, D.Y., Karimi, I.A., and Lee, D.-Y. (2010). Combined data preprocessing and multivariate statistical analysis characterizes fed-batch culture of mouse hybridoma cells for rational medium design. J. Biotechnol. 150: 94–100. Carinhas, N., Duarte, T.M., Barreiro, L.C. et al. (2013). Metabolic signatures of GS-CHO cell clones associated with butyrate treatment and culture phase transition. Biotechnol. Bioeng. 110: 3244–3257. Charaniya, S., Karypis, G., and Hu, W.-S. (2009). Mining transcriptome data for function-trait relationship of hyper productivity of recombinant antibody. Biotechnol. Bioeng. 102: 1654–1669. Dietmair, S., Nielsen, L.K., and Timmins, N.E. (2011). Engineering a mammalian super producer. J. Chem. Technol. Biotechnol. 86: 905–914. Browne, S.M. and Al-Rubeai, M. (2007). Selection methods for high-producing mammalian cell lines. Trends Biotechnol. 25: 425–432. Castro, P.M., Hayter, P.M., Ison, A.P., and Bull, A.T. (1992). Application of a statistical design to the optimization of culture medium for recombinant interferon-gamma production by Chinese hamster ovary cells. Appl. Microbiol. Biotechnol. 38: 84–90. Yang, L., Yurkovich, J.T., Lloyd, C.J. et al. (2016). Principles of proteome allocation are revealed using proteomic data and genome-scale models. Sci. Rep. 6: 36734. King, Z.A., Lloyd, C.J., Feist, A.M., and Palsson, B.O. (2015). Next-generation genome-scale models for metabolic engineering. Curr. Opin. Biotechnol. 35: 23–29.

References

106 Liu, J.K., O’Brien, E.J., Lerman, J.A. et al. (2014). Reconstruction and model-

107 108

109

110

111

ing protein translocation and compartmentalization in Escherichia coli at the genome-scale. BMC Syst. Biol. 8: 110. Feizi, A., Österlund, T., Petranovic, D. et al. (2013). Genome-scale modeling of the protein secretory machinery in yeast. PLoS One 8: e63284. Spahn, P.N., Hansen, A.H., Hansen, H.G. et al. (2016). A Markov chain model for N-linked protein glycosylation – towards a low-parameter tool for model-driven glycoengineering. Metab. Eng. 33: 52–66. Sha, S., Agarabi, C., Brorson, K. et al. (2016). N-glycosylation design and control of therapeutic monoclonal antibodies. Trends Biotechnol. 34: 835–846. Spahn, P.N., Hansen, A.H., Kol, S. et al. (2017). Predictive glycoengineering of biosimilars using a Markov chain glycosylation model. Biotechnol. J. 12: 1600489. Pristovšek, N., Nallapareddy, S., Grav, L.M. et al. (2019). Systematic evaluation of site-specific recombinant gene expression for programmable mammalian cell engineering. ACS Synth. Biol. 8 (4): 758–774. https://doi .org/10.1021/acssynbio.8b00453.

95

97

5 Genome Variation, the Epigenome and Cellular Phenotypes Martina Baumann 1 , Gerald Klanert 1 , Sabine Vcelar 1,2 , Marcus Weinguny 1,2 , Nicolas Marx 1,2 , and Nicole Borth 1,2 1

Austrian Centre of Industrial Biotechnology (ACIB), Muthgasse 11, Vienna 1190, Austria University of Natural Resources and Life Sciences (BOKU), Department of Biotechnology, Muthgasse 18, Vienna 1190, Austria 2

5.1 Phenotypic Instability in the Context of Mammalian Production Cell Lines For the production of large, complex-structured therapeutic proteins requiring human-like posttranslational modifications (PTMs), mammalian cells, in particular Chinese hamster ovary (CHO) cells, are the production host of choice. The most fundamental reason for the importance and widespread use of CHO cells in the biopharmaceutical industry is their adaptability, which is based on both genomic and phenotypic heterogeneity and allows for evolution toward desired phenotypes, including growth under a variety of culture conditions as well as efficient production of proteins. The plasticity of CHO cell lines can be exploited to isolate clonally derived populations with improved physiological or process-relevant properties [1]; however, as a severe downside, it also results in genomic and phenotypic instability, which leads to substantial variation and unpredictable stability of expression. Genetic instability has been described in many studies as an inherent feature of CHO, as well as of other immortalized, fast growing mammalian cell lines, which are characterized by mutations, singe-nucleotide polymorphisms (SNPs), translocations, and chromosomal rearrangements [2, 3]. For CHO production cell lines, the most important feature with respect to instability is the decrease in recombinant protein productivity or varied host cell protein expression during extended culture time [4–8]. Such instability not only impacts cell productivity, but can also have adverse effects on downstream purification and on the consistency of product quality, affecting biological function and pharmacokinetics. It is therefore crucially connected to safety and efficacy of the therapeutic proteins [9]. Development of production cell lines generally relies on stable, random integration of the transgene sequence into the genome of the host cell, often followed by chemical selection and gene amplification [10, 11], resulting in variations in gene dosage and chromosomal context of integrated copies within the transfectant Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

98

5 Genome Variation, the Epigenome and Cellular Phenotypes

pool, ranging from nonproducing to high-producing cells. Accordingly, in order to minimize heterogeneity, cell pools are subcloned to generate populations originating from a single cell. As these clones derived from the same parental cell pool may differ significantly in phenotype and process performance, several thousand clones have to be screened to identify the best producers [12, 13]. However, the homogeneity aimed for by subcloning is transient, as new variants appear on a continuous basis, such that re-subcloning a subclone will again generate subclones with varying properties [12]. The variation observed in this publication (with “genetically identical” single-cell-derived clones) is surprisingly high, with coefficients of variation of up to 70%, approaching the level of variation within transfectant pools. Results suggest that a significant fraction of total variation arises from phenotypic differences resulting from expression fluctuations in individual cells over time [12, 14], caused by subtle differences in the regulation of cellular pathways rather than by “hard-fact” mutations of a gene. Taking advantage of this cell-to-cell variation, Pichler et al. [15] selected for naturally occurring CHO variants that possess an improved cellular machinery for antibody production via three rounds of transient transfection and sorting for 1% top producers. Obtained subclones achieved a threefold higher specific productivity after transient transfection, which could be maintained for approximately three months, indicating that these changes in productivity were regulatory, and not mutational. Again, interclonal variation, although it significantly increases the workload required for cell line development, is an asset as long as the objective is to find an outstanding high-producing subclone; however, it becomes a problem once such a subclone has been found and needs to be maintained as a stable and reproducible producer cell line. In addition to clonal variation which may be observed in many facets of phenotypic behavior, including growth, final cell density, and metabolism, the specific changes that are observed with respect to recombinant protein productivity are of high industrial relevance. In the context of industrial biopharmaceutical production, stability is typically defined as reproducible protein yield and quality from a given cell line over extended periods of time, from thawing of the frozen master cell bank to expansion of the cells for large-scale production. As a high number of cell doublings is required to reach an appropriate scale, the minimal accepted period for proof of stability is three months, whereas for approved therapeutic proteins, six months are the standard for stability studies [16]. This type of instability is unpredictable and varies from clone to clone; however, reports from industry describe the occurrence of 50–70% unstable clones among those routinely tested. Such high failure rates generate increased efforts during the screening process as more clones need to be tested and consequently result in higher manpower requirements, higher costs, and delays in time to market [13]. Underlying causes and kinetics for instability or loss of cell-specific productivity (qP ) are not fully elucidated yet, but may be connected to the increased stress and the allocation of resources to recombinant protein production [17]. Two primary mechanisms for loss of productivity were described in a number of studies [7]: (i) loss of recombinant gene copies [18–20] and (ii) gene silencing, presumably via epigenetic mechanisms such as DNA methylation or histone modifications [21–24]. Production instability is also influenced by the gene

5.2 Genomic Instability

expression system used. For example, cell lines generated with the dihydrofolate reductase (DHFR) gene amplification system typically have a higher number (up to ∼500 copies) of tandem repeats of the recombinant genes compared to the GS (glutamine synthetase) system and thus are more prone to repeat-induced gene silencing via DNA methylation [20, 25–27]. It can be assumed that cells that reduce their transgene expression have a physiological advantage and thus will outgrow producer cells, leading to a steady increase in non- or low producers. For years, the plasticity of CHO phenotypes was described as being related to “genomic instability,” for lack of more comprehensive understanding of the underlying causes. In view of recent sequencing efforts and studies that investigate the impact of the epigenome on cell line behavior, it is time to review our current knowledge on both genomic variation and epigenetic regulation in CHO cells.

5.2 Genomic Instability Genetic instability is a relatively uncontrollable and unpredictable phenomenon that was found in a number of studies to be neither host specific nor product specific [6]. Some mutations will be preclusive as they impact vital gene functionality, while others are carried along or may even be expanded within a population if they lead to a growth advantage. The high rate of replication in immortalized cell lines such as CHO leads to an increase in error rate during replication, which may be further enhanced by mutations that are frequently found in cancer cell lines, such as a loss of p53 function, which was also described for CHO [28]. Genomic instability can fall into several categories: (i) mutations: this includes SNPs that may affect protein function as well as short insertions/deletions (indels) that may lead to frameshifts or loss of function. Either of the above may also occur in regulatory, noncoding regions of the genome where they may lead to alterations in gene expression. (ii) Larger structural variants: these include deletions, duplications, or translocations of larger sequences within the genome, which may again lead to altered gene expression patterns or loss of function of an entire gene. (iii) Chromosomal aberrations: these may include numerical aneuploidies such as loss of an entire chromosome or trisomy, or rearrangements of chromosomes where entire segments of chromosomes are joined together into a new pattern [29]. Such rearrangements are a characteristic feature of cancer cells and have also been described in CHO cells. It is likely that such rearrangements or mutations occur randomly at every division; in fact, they are not exclusive to cancer or immortalized cells but seem to be connected to high division rates in culture, as they were also observed in normal healthy stem cells that were maintained in an artificial environment [30–32]. Assessment of genomic stability is difficult and elaborate. The techniques available include cytogenetic methods that analyze the number of chromosomes per cell and any larger scale rearrangements by chromosome counting, banding, or painting (Figure 5.1) [33–35]. Genotyping methods that have been used for multilocus genomic fingerprinting include amplified fragment length

99

100

5 Genome Variation, the Epigenome and Cellular Phenotypes

Chr 1 Chr 2 Chr 3 Chr 4 Chr 5

(a)

Chr 6 Chr 7 Chr 8 Chr 9 Chr 10 Chr X

(b)

Figure 5.1 Example of multicolor fluorescence in situ hybridization (mFISH) of Chinese hamster lung fibroblast and CHO-K1 cells using painting probes specific for the Chinese hamster chromosomes. Each chromosome is distinguished by a different color. (a) Chromosome pattern of an lung fibroblast cell (the presumptive rearrangement on chromosome 6 is due to a faulty probe that was sorted from an already rearranged cell line). (b) Chromosome pattern of a CHO-K1 cell showing diverse rearrangements and aneuploidies.

polymorphism (AFLP), random amplified polymorphic DNAs (RAPDs), and inter simple sequence repeats (ISSRs). These are PCR based and aim to detect dominant marker changes in the genome [36]. They can be set up for any organism and can answer diverse research questions, such as phylogenetic relationships, genetic differentiation and diversity, molecular marker, genomic mapping, and intraspecific variation [37]. However, they work well primarily for homogenous populations and are difficult to standardize for immortalized cell lines, which, due to their heterogeneity, provide no clear pattern that can be compared or followed. It is no coincidence that, despite the need for methods to identify unique cell lines and production clones, no such method has so far been published for CHO cells. All of the above methods are able to detect large-scale rearrangements such as translocation, deletion, or amplification of genomic sequences. The only method available to analyze small mutations and variants on a whole genome basis is next-generation sequencing (NGS). Here, both genome-wide analysis of small mutations such as SNPs and indels and of structural variants such as translocations is possible, and because of the high read counts that are obtained, it is also possible to analyze for population heterogeneity [38, 39]. In these studies, a high number of SNPs and other small mutations were observed in CHO cells. Feichtinger et al. investigated the genetic heterogeneity of different related CHO cell lines, all originating from the serum-dependent CHO-K1 cell line that underwent evolutionary pressures including adaptation to growth in suspension, to different media formulations, and to prolonged culture times. The study revealed that genetic variation is high on all levels, including SNPs, indels, and structural variants (duplications, deletions, inversions, and translocations). It is not necessarily linked to the process of adaption or selection but also simply occurs during culture, with new variants appearing and disappearing continuously. The CHO genome seems to undergo continuous and random

5.3 Epigenetics

rearrangements, most prominently translocations, confirming previous reports of chromosomal aberrations and irregularities in chromosome number [40–42]. Interestingly, the number of variants, their distribution, and frequency of the analyzed subclone is not significantly different from that of a cell pool, eliminating the initial genomic homogeneity of a single cell. This corresponds to the findings of other studies on subclones using cytogenetic methods, showing that the progeny quickly diversifies into a population with karyotype distributions similar to that of a parental cell line [43, 44]. Thus, while subcloning may be required to select for a specific integration site or behavior that needs to be protected from outgrowth by faster growing cells, the common expectation that subclones are genomically homogenous needs to be questioned for a rapidly growing, immortalized cell line such as CHO. On top of these sequence variants, several studies have shown that a large number of chromosomal rearrangements occur in CHO cells and that the number of chromosomes of each cell line and variant can vary. Each cell line typically has a predominant most frequent number of chromosomes that may range from 18 to 22, but within each population, typically 50% of the individual cells will have different chromosome counts, thus contributing to the heterogeneity of a population [42].

5.3 Epigenetics As already outlined above, the most important impact of genomic variation is not necessarily the change of protein sequences or a total loss of function, but a change in the expression pattern of genes that may lead to subtle changes in behavior and a variety of phenotypes. Apart from translocations or mutations in regulatory regions, such changes in gene expression are typically achieved by a set of mechanisms that leaves the genome sequence unaltered, but leads to changes in chromatin structure and thus in accessibility and transcription rate of individual genes. These mechanisms are summarized under the term epigenetics. Epigenetics plays a crucial role in expression heterogeneity and describes inheritable gene expression alterations that do not entail a change in the DNA sequence itself. Differences in gene expression are required for development and differentiation of multicellular organisms, where all cells bear the same genetic code, but fulfill distinct functions. Epigenetic regulation effects substantial cellular processes such as the formation of euchromatin and heterochromatin [45], genomic imprinting [46], and X-chromosome inactivation [47]. Also, environmental influences modulate epigenetic changes [48], which are involved in the development of certain chronic diseases [49]. These changes comprise covalent postsynthetic modifications of DNA and the histones associated with it, nucleosome remodeling, and the exchange of histone variants. Although DNA is mostly modified by the conversion of cytosine to 5-methylcytosine (5mC) at CpG sites (cytosine nucleotide followed by a guanine nucleotide, separated by a phosphate group) or by the methylation of the sixth position of adenine [50], with the latter also being present in RNA molecules [51], histone modifications cover a broader spectrum of functional group additions in both the globular domains and N-terminal tails

101

102

5 Genome Variation, the Epigenome and Cellular Phenotypes

of core and linker histones [52]. The number of identified modifications is still expanding [53], including acetylation, methylation, phosphorylation, glycosylation, sumoylation, ubiquitination, biotinylation, citrullination, ADP-ribosylation, and many more, with acetylation and methylation being the best studied ones for epigenetic regulation of gene expression. In the following, the most frequent epigenetic mechanisms are described. 5.3.1

DNA Methylation

Methylation of cytosines is by far the best characterized DNA modification. If present in promoter regions, it leads to repression of the respective gene [54] by sterically hindering transcription factor binding [55]. On the other hand, actively transcribed gene sequences are typically fully methylated to enable efficient transcription [56, 57]. The addition of a methyl group to cytosine is catalyzed by a group of enzymes called DNA methyltransferases (DNMTs) [58]. Three enzymatically active DNMTs have been identified so far, DNMT1, DNMT3a, and DNMT3b. DNMT1 is responsible for maintaining the existing methylation state of DNA by methylating the nonmethylated DNA strand during DNA replication and ensuring inheritance of the existing pattern by daughter cells. Ubiquitin-like with PHD and RING finger domains 1 (UHRF1) recognizes and binds hemimethylated DNA and guides DNMT1 to methylate the second strand [59]. DNMT3a and b, which are activated by DNMT3l, a catalytically inactive protein that interacts with the catalytic domains of DNMT3a/b, perform de novo methylation [60] and thus are able to change the gene expression pattern of a cell during development or adaptation. Demethylation of DNA can occur either passively because of a lack of maintenance methylation during DNA replication or actively with the help of the ten-eleven translocation dioxygenases (TET1–3). These enzymes oxidize 5mC to 5-hydroxy-methylcytosine, 5-formylcytosine, and further to 5-carboxylcytosine, with the latter two being recognized and excised by thymine-DNA glycosylase (TDG) followed by the base excision repair pathway [61]. Thus, DNA methylation provides the cell with a mechanism to inherit and maintain a given gene expression pattern and at the same time provides the ability to change and modify this pattern as required. It is thus an important mechanism that enables adaptation to altered culture conditions and also serves as the long term memory of cells. 5.3.2

Histone Modifications

Modification of histone tails represents another layer of epigenetic regulation. Histones are basic proteins that play an important role in organizing the DNA into smaller units, called nucleosomes (Figure 5.2). The nucleosome core particle consists of 147 base pairs of DNA wrapped around a histone octamer, which consists of four core histones (H2A, H2B, H3, and H4), and the linker histone H1, fixing the DNA to the core. Nucleosomes are connected to each other by linker DNA [62]. The location of the nucleosomes determines the accessibility of DNA elements and can be shifted along the DNA (nucleosomal positioning) [63]. Different variants exist for most histones, which are sometimes associated with distinct chromatin structures and function [64]. Core histones are composed of a

5.3 Epigenetics

Chromosome

Chromatin structure

Me

Ac

H2A H2B H3 H4

H1

Me

Nucleosome structure

Figure 5.2 Schematic structure of a nucleosome. The nucleosome is a subunit of the chromatin structure consisting of DNA wrapped around histone octamers (two copies each of the core histones H2A, H2B, H3, and H4) and the linker histone H1, fixing the DNA to the core. N-terminal tails of the core histones protrude from the nucleosome surface and are susceptible to posttranslational modifications, such as acetylation (Ac) or methylation (Me), influencing the accessibility of the DNA.

globular domain and an N-terminal tail that protrudes from the nucleosome core surface and is thus more susceptible to PTMs, which are added and removed enzymatically to generate the so-called histone code [65]. The histone code is highly dynamic as PTMs are reversible, the histone turnover rate is high [66], and histone tails can be clipped [67]. The most widely studied PTM of histones is acetylation, especially at lysine residues present on all four core histones. The addition of acetyl groups is executed by histone acetyltransferases (HATs). It reduces the positive net charge

103

104

5 Genome Variation, the Epigenome and Cellular Phenotypes

of histones and thus the interaction between histones and the negatively charged DNA. This leads to higher accessibility of the DNA to transcription factors and therefore promotes active transcription. The counterparts for these enzymes are histone deacetylases (HDACs), which remove acetyl groups, generating a tighter interaction between histones and DNA. Apart from acetylation, enzymatic methylation of arginine and lysine residues can both up- and downregulate transcription depending on their position on the histone and the number of methyl adducts added (up to three at lysine and up to two at arginine) [68]. 5.3.3

Downstream Effectors

In addition to the abovementioned cis mechanisms of DNA and histone modifications (structural changes due to functional group additions), these adducts may also exert certain trans effects. The so-called “reader” proteins can bind to modified histones via distinct domains [69]. For example, heterochromatin protein 1 binds to trimethylated lysine 9 of histone H3 (H3K9) with the help of its chromodomain, thereby associates with the nucleosome, and initiates the formation of heterochromatin [70]. Reader proteins of DNA methylation [71] often act as transcriptional repressors by recruiting HDACs, demonstrating the highly interconnected landscape of epigenetic regulation [72]. Readers not only recognize the modification they are binding to, but are often also selective for the surrounding sequence context [69], increasing their specificity. 5.3.4

Noncoding RNAs

Long noncoding RNAs (lncRNAs) can guide chromatin modifying enzymes to sequence-specific locations in a cis- and trans-acting mechanism. For example, the lncRNA HOX Transcript Antisense Intergenic RNA (HOTAIR) is acting on hundreds of DNA locations to propagate histone modifications [73]. Also, the inactivation of the second female X-chromosome is initiated by the expression of the lncRNA Xist [47]. Besides lncRNAs, there are several groups of short noncoding RNAs showing impact on the epigenetic profile, with microRNAs (miRNAs), small-interfering RNAs (siRNAs), and piwi-interfering RNAs (piRNAs) being the most prominent ones. miRNAs exhibit posttranscriptional gene silencing (PTGS) by directing mRNA cleavage or translational inhibition. A group of miRNAs (epi-miRNAs) directly targets enzymes involved in epigenetic modifications, thereby influencing the epigenetic state of the cells [74]. siRNAs are also mostly known for executing PTGS but can also target DNA sequences, especially promoters, to change gene expression [75, 76]. piRNAs are crucial for silencing retrotransposons and repetitive sequences [77]. As mentioned above, all of these mechanisms are highly interconnected and should not be considered as isolated processes. Noncoding RNAs typically exert their mechanism by providing targets for chromatin-state modifying enzymes that either change histone marks or alter DNA methylation. Thus, they are the initiator; however, for persistent changes, the cell needs to permanently imprint the DNA via methylation.

5.4 Control of CHO Cell Phenotype by the Epigenome

5.4 Control of CHO Cell Phenotype by the Epigenome More detailed studies on the whole genome DNA methylation of CHO cells [78] provide a global view of the CHO cell methylome as a linkage between the genome and transcriptome. In Feichtinger et al. [38], it was found that the overall DNA methylation pattern of CHO cells remained fairly constant if cells are maintained under constant culture conditions, consistent with its function as inheritable information that determines the transcriptome pattern of cells. During adaptation to altered culture conditions or after selection of specific phenotypes, however, the pattern changed dramatically, indicating a change in transcriptome to enable survival under the new conditions. It can either be that a subpopulation of cells with the desired gene expression survives the adaptation/selection or that cells gradually change their expression pattern. Although the global DNA methylation remains fairly unchanged during batch culture, histone modifications are continuously adapted to sense changes in the environment [38], which is also reflected by the change in epigenetic factor abundance [79]. Following up the study from Feichtinger et al., Hernandez et al. reported that, besides the presence or absence of DNA methylation in the promoter regions and the presence of transcription- and promoter-specific histone marks, the interaction of long-non coding RNAs with coding genes regulate expression levels [80]. lncRNAs can form RNA-DNA-DNA triplexes, predominantly in promoter and enhancer regions of coding genes, thus changing their transcription levels. During a 9-day batch cultivation, genes showing the highest variation in expression levels also showed a high number of potential triplex forming interactions with differentially expressed (DE) lncRNAs, indicating an association of DE lncRNAs and the temporal regulation of coding genes. Specific attention was paid to the impact of epigenetic marks of viral promoters that typically control expression of the recombinant protein. Viral promoters are especially prone to silencing in mammalian cells as they are recognized as artificial and are not integrated into the cellular regulatory network that responds to the state of cells [81–84]. Silencing of recombinant genes over time, reflected by decreased mRNA levels [21, 85], results in most cases from a loss of active transcription marks around the promoter [86–91], which is often, but not always [85, 92], accompanied by DNA methylation [7, 24, 93] and the establishment of transcription repressing histone marks. A possibility to prevent methylation and to enhance cell line stability is the use of DNA regulatory elements in promoter design. This has some advantages over chemical treatment described later in this chapter: (i) no concentration-dependent cytotoxic effect will affect the producing cell line, (ii) there are no additional costs for media additives (i.e. the necessary chemicals), (iii) the enhancing effect is independent of the chemical – removing the chemical will not decrease the stability or productivity of the cell line, and (iv) safer handling – some of the described chemicals are known mutagens. As described earlier, CpG dinucleotides are necessary to silence genes via DNA methylation. The CpGs are identified by DNMTs, methylated, and “turned

105

106

5 Genome Variation, the Epigenome and Cellular Phenotypes

off.” To counter this, a promoter deprived of any CpG islands was designed using a mouse cytomegalovirus (CMV) enhancer, the human elongation factor 1 core promoter, and a synthetic intron at the 5′ UTR, with the intention to enhance stability and protein expression levels [94]. The CpG-free promoter slightly improved the expression level and the number of cells (10% more) still expressing the gene of interest (GOI) after eight weeks. However, the authors stated that the CpG-free promoter is better suited for the selection process than for long-term stability – the better results after eight weeks are because of higher initial expression levels and not due to improvement in stability. Taking a closer look at the cause for the instability of both promoter variants, the study shows that in both cases histone modifications are the governing silencing method of the cell. Similar to Spencer et al. [92], the authors concluded that histone modifications may be a better biomarker for long-term stability of transfected genes than DNA methylation. Other studies by the same group [95, 96] found similar evidence that DNA methylation may not be as important for transgene stability as previously thought. Originally, the studies investigated the insertion of the core CpG island element from the hamster adenine phosphoribosyltransferase, which was reported to prevent methylation, into the human CMV promoter. However, this effect could unfortunately not be confirmed in CHO-K1. They were not able to see significant differences in methylation levels between wild-type and transfected samples, although transfected cells were able to maintain expression levels. However, the authors assume that their choice of DNA methylation measurement may be too crude to find differences between wild-type and recombinant CHO-K1 cell lines. Moreover, they argue that the core element may have some complex mechanism involving location, orientation, and copy number, which needs the right combination to enhance the stability. These recurring reports of dismissing DNA methylation for long-term stability show that the exact mechanism of epigenetic silencing is still elusive and that much more effort has to be put into elucidating the exact epigenetic machinery. Another study took a closer look at the CpG islands of human cytomegalovirus promoter (hCMV) and searched for frequently methylated CpG islands in recombinant CHO cell lines, identifying C at positions −508, −179, and −41 upstream of the transcription start site [97]. Point mutations of C to G in each of these and subsequent stable expression of the altered promoter in CHO-K1 resulted in enhanced stability for C-179 and C-41 mutants. Interestingly, the authors did not see any further improvement using both mutations together, speculating that the methylation pattern of the CMV promoter could be more important than the level of methylation. In view of these diverging results, it may be possible that the driving force for instability is the stress of overproduction, as mentioned before. Thus, if a promoter is engineered to avoid methylation, cells will find another way of downregulating recombinant protein production, which would explain the contradicting results to some extent: if there are several options available, in each case investigated, one of these mechanisms will be found, whichever was first at hand to release the stress for the cell.

5.5 Manipulating the Epigenome

Another DNA element shown to prevent DNA methylation is the so-called ubiquitous chromatin opening element (UCOE). This element was reported to prevent DNA methylation of lentiviral vectors in mouse cells [98] and to improve stability of expression in CHO [99–102]. Another possibility is to use matrix attachment regions (MARs) in the promoter design, which allow the cells to generate an enhanced histone acetylation region and in turn a better DNA accessibility [103]. The incorporation of MARs into the promoter was already used to generate higher producing and stable CHO cell lines [104–107]. A good overview of epigenetic regulatory elements is found in Harraghy et al. [108]. A study comparing different DNA regulatory elements found that UCOEs had the most beneficial effect on long-term stability in CHO-K1 [109]. Measuring the DNA methylation of the used promoters, promoters with UCOE were able to prevent DNA methylation for 120 generations. MAR elements did increase the titer as well, but not as highly as UCOE. The authors speculated that only a subset of MAR elements is working in CHO [110].

5.5 Manipulating the Epigenome 5.5.1 5.5.1.1

Global Epigenetic Modification Manipulating Global DNA Methylation

Besides the insertion of epigenetic regulatory elements into promoter and vector design to enhance cell line performance, the global epigenetic pattern can be altered via chemical treatment. To generate global DNA hypomethylation, two drugs were applied in CHO so far, 5-Azacytidine (5-Aza) and 5-Aza-2′ -deoxycytidine (5-Aza-DAC). Both are approved by the US – FDA and the European Medicines Agency (EMA) for treatment of myeloid malignancies [111]. They are cytosine analogs, inserted directly into the cell’s DNA during replication, where they act as traps for DNMTs. Normally, the methyltransferase binds covalently to the carbon 6 of the cytosine ring and is successively released via β-elimination of the covalent bond. This release is blocked by 5-Aza and DAC, trapping the enzyme and leading to its degradation by the proteasome [112]. However, these molecules affect only the maintenance methyltransferase DNMT1, as ubiquitin tagging for proteasomal degradation requires the KEN-box, which is not present in the de novo DNMTs 3a and 3b [113]. Thus, application of these drugs results in overall demethylation of cells, which increases with each subsequent cell division. Importantly, these demethylating effects only occur at low concentrations, whereas higher concentrations lead to cytotoxicity [114]. Both chemicals were used to investigate demethylation events in CHO. Spencer et al. [92], administered DAC to CHO-K1 cells, to investigate the silencing of CMV promoters. The demethylating effect of DAC was not able to improve expression levels of the GOI, leading to the conclusion that histone modifications are mainly responsible for the silencing of CMV – at least during the first weeks after subcloning.

107

108

5 Genome Variation, the Epigenome and Cellular Phenotypes

5-Aza and DAC were also used to investigate the productivity loss of antibodies in CHO-DG44 cells. Several studies reported that long-term productivity loss in the absence of selective pressure is due to increases in methylation of the hCMV promoter [21] and the administration of DAC was able to partially reverse this [24]. However, productivities could not be fully restored, possibly because of incomplete demethylation in the DAC-treated cells. Increasing the concentration of DAC resulted in higher death rates and did not improve the demethylation results. In a similar study, 5-Aza instead of DAC was able to reverse the loss in productivity [7] and the use of increasing concentrations resulted in higher productivity restoration, but also had adverse effects on viability and growth. Besides these studies that looked specifically at demethylation of the recombinant promoter, 5-Aza has also been used to change the expression pattern of endogenous genes to activate silenced promoters required for cholesterol synthesis of the auxotrophic murine NS0 cells [115]. The authors were able to show that important genes of the cholesterol synthesis were upregulated after 5-Aza treatment compared to the original NS0 cells without addition of 5-Aza. Besides addition of nucleoside analogs such as 5-Aza and DAC, chemicals influencing nucleotide biosynthesis can help promote transcriptome changes. In a study by Vishwanathan et al. [116], it was shown that part of the success of cell line development using methotrexate (MTX) in the DHFR gene amplification system is based on alterations in transcriptome pattern irrespective of increases in gene copy numbers. These transcriptome changes enable cells to handle production stress better and thus result in better production clones being isolated. 5.5.1.2

Manipulating Global Histone Acetylation

Another way to influence the global epigenome is by manipulation of histone tail modifications. Because of its important role in active gene transcription, histone acetylation is the main target here. The most frequently used histone deacetylase inhibitor (HDACi) is sodium butyrate (NaBu), which has been studied extensively as an additive to bioprocesses that boosts productivity. NaBu is able to inhibit all HDAC class I and II enzymes so that ongoing acetyltransferase activity results in a hyperacetylated genome. In addition, several “off”-targets, such as transcription factors, are also affected. Addition of NaBu results in growth arrest through dephosphorylation of the retinoblastoma (Rb) protein and the upregulation of cyclin-dependent kinase inhibitors p16INK-4A and p21Cip1 . Part of the increase in productivity is thus simply due to the growth arrest and the availability of more energy to be channeled into protein production rather than growth. One mechanism responsible for the dephosphorylation of Rb may be due to reduced amounts of cyclin D and E during NaBu treatment [117], although several pathways are likely to be affected [118–121]. NaBu is used as a medium additive to increase the production of antibodies and other relevant pharmaceutical compounds in recombinant CHO cells [122–128]. As a side effect, an increase in apoptotic cells was observed in a dose-specific manner. Moreover, these production-enhancing effects seem to be clone specific where clones with lower productivity benefit more from NaBu treatment than high-producing clones. The transcript level for the recombinant product is

5.5 Manipulating the Epigenome

increased upon addition of NaBu, indicating that the mode of action is indeed a higher accessibility of the genome for transcription [126, 129]. An additional study by Wippermann et al. [121] showed similar results in butyrate-treated CHO-DP12 cells, which also demonstrated the interplay between histone modifications and DNA methylation by correlating alterations in DNA methylation with changes in gene expression. Transcriptome analysis of recombinant CHO showed that the transcription of apoptosis and cell-cycle-specific genes is upregulated, along with genes involved in protein folding and vesicle transport [130, 131]. The higher expression of apoptosis and cell cycle genes is not surprising, as HDACi is also investigated for possible anticancer treatment [132, 133]. Of interest are the higher expression levels in protein folding and vesicle transport, both limiting bottlenecks for high-producing cell lines. For industrial applications, the chemical has some major disadvantages, however. First, because of the required millimolar concentration, the cost for large-scale processes is prohibitive, and second, the increased occurrence of apoptosis, oxidative stress, and unfolded protein response may lead to changes in product quality, such as aggregation [134]. To avoid some of the negative side effects of NaBu, other, more efficient, and cheaper alternatives were investigated. A promising alternative to NaBu, both with respect to reduced side effects and cost, was found to be valproic acid [127, 135, 136]. An even cheaper alternative is valeric acid, which does not promote apoptosis as strongly as NaBu [137, 138]. As aggregation of product was still an issue with these alternatives, miRNA-2861, a specific inhibitor of HDAC-5, was proposed as an alternative, which was able to boost productivity without product aggregation or reduced growth [139]. Another study reports that combinations of HDACi (NaBu, Trichostatin A) and DNMTi (DAC) may enhance the productivity of cell lines [140]; however, it was also shown that optimal concentrations and combinations may vary between cell lines. The highest synergistic effects were accomplished using one or both HDACi in combination with DAC, yet again indicating that DNA methylation and histone deacetylation are highly connected. This conclusion was also made by Raynal et al. [141], using two different drugs (HDACi and DNMTi) to reverse the activity of a hypermethylated (silenced) promoter. After administration of both drugs together, the cells were able to stably express the GOI for six months. Interestingly, after HDACi treatment alone, the gene was reactivated (for two weeks), although DNA methylation was not altered, demonstrating that histone acetylation may overcome silencing by DNA hypermethylation. This observation let the authors refine the current scientific consensus, arguing that DNA methylation serves as a molecular mark, not necessarily permanently suppressing gene expression, but rather acting as a memory signal for gene silencing. 5.5.2

Targeted Epigenetic Modification

The detailed interrogation of epigenetic effects on gene expression and phenotype requires more fine-tuned and sophisticated approaches than the random global approaches described so far. Targeted modification of genomic loci enables to directly link the cause and effect of differential gene expression while avoiding

109

110

5 Genome Variation, the Epigenome and Cellular Phenotypes

many off-target effects caused by the global derangement of the cell’s regulatory network encountered in random genetic engineering of a cell line. To design tools for such targeted modification, typically effector domains that add or remove chemical groups to or from the genomic target site are directed to the region of choice using sequence-specific DNA binding tools, most notably zinc finger proteins (ZFPs), transcription activator-like effectors (TALEs), and RNA-guided engineered nucleases (RGENs). ZFPs were already employed in the late 1990s [142–144], whereas TALEs were discovered nearly 10 years later [145, 146] and RGENs shortly thereafter [147]. The different targeting technologies each harbor various advantages and drawbacks, which have been reviewed in detail [148–151]. Briefly, ZFPs and TALEs are DNA-binding proteins that, dependent on the amino acid composition of their binding domains, can specifically target appropriate DNA sequences. By fusing them with endonuclease domains or effector domains, both genome and epigenome editing have been successfully accomplished [152–155]. However, this requires de novo protein design for each new target sequence and thus results in high costs and prolonged timelines. On the other hand, the binding mechanism of RGENs such as clustered regular interspaced palindromic repeats (CRISPRs)-CRISPR-associated protein 9 (Cas9) is based on base-pairing between the RNA entity and the DNA target strand and the unvarying recognition of a protospacer adjacent motif (PAM) by the Cas9 protein. For CRISPR-Cas9, the nuclease has been engineered to yield a dead-Cas9 (dCas9) without nuclease activity, but preserving the targeting mechanism [156]. Thus, dCas9 in conjunction with a guide RNA (gRNA) serves as a shuttle, carrying any cargo in the form of fused effector domains to the intended genomic sites. 5.5.2.1

Targeted Histone Modification

For modification of gene expression by targeting histone marks, several modulators have already been utilized, including HATs (p300) [157], histone demethylases (LSD1) [158, 159], and histone methyltransferases (G9A, SUV39H1) [160]. By depositing or removing a distinct chemical group at histone tails, these effectors promise to allow specific and direct modulation of the surrounding genomic landscape and furthermore to determine the causality of (some) epigenetic events. The underlying principle of targeted activation by histone modifiers is the formation of euchromatin and the consequential accessibility by transcription factors and RNA polymerases. Additionally, several enzymes as the adenovirus early region 1A binding protein p300 do not only deposit chemical groups at histone tails but also affect other transcriptional activators [161]. Because of the limited number of studies available so far, current results are yet inconclusive. The first study published, in which a dCas9-p300 fusion construct was directed against the myogenic differentiation (MYOD) promoter, the octamer-binding transcription factor 4 (OCT4) promoter and the Interleukin 1 receptor antagonist (IL1RN) promoter showed an up to ninefold stronger activation compared to the transcriptional activator dCas9-VP64 [157]. On the other hand, in 2016, a comprehensive study comparing different dCas9 activator species reported comparable expression levels to either one of the before mentioned systems [162].

5.5 Manipulating the Epigenome

The authors point out that other systems such as SunTag-VP64 (SuperNova tagging system that can recruit multiple copies of VP64 to the target DNA site) are more potent activators; however, no clear conclusion on the reason for these diverging results are given. A possible explanation may be that the same single guide RNAs (sgRNAs) were used for comparability of the construct’s activation capability at the same genomic locus. Thus, the different acting range of each fusion protein, which is dependent on its flexibility and of course on whether it can interact with the respective regulatory element, was not taken into account. Additionally, no comparative study has been conducted so far characterizing features of highly potent sgRNAs and the activity of the corresponding construct. Nonetheless, dCas9-p300 marks an important step toward the completion of an epigenetic tool kit as it permits a more direct investigation and correlation of specific histone marks and their effects on gene expression. Furthermore, dCas9-p300 has been used in combination with dCas9-Krüppel associated box (KRAB) (transcriptional repressor) to identify functional regulatory elements in the human genome in a high-throughput approach [163]. The lysine demethylase LSD1 was adopted as the p300 counterpart for targeted transcriptional repression. Histone modifications such as H3K4me3 or H4K20me are linked to gene activation and Pol II binding [164]. The targeted removal of these marks is therefore opening the opportunity to specifically inactivate genes bearing these specific epigenetic states and thus evaluate the significance of their chromatin state. Targeting TALE-LSD1 [159] and dCas9-LSD1 [158] to the enhancer of the stem cell leukemia locus in K562 cells or the distal enhancer of OCT4 in mouse embryonic stem cells, respectively, led to active H3K4 demethylation and silencing of the respective gene. Furthermore, it was shown that dCas9-LSD1 would act specifically only on enhancers, thus enabling a more precise and detailed characterization of enhancers. Similarly, methylation of H3K9 by euchromatic histone-lysine N-methyltransferase 2 (EHMT2), also called G9a, is associated with gene repression [165]. Targeting a dCas9-G9a fusion protein to the promoter of a vascular endothelial growth factor A (VEGF-A) expressing HEK293T cell line resulted in twofold repression [166]. This is in accordance with earlier studies of ZF-G9a targeted downregulation of VEGF-A [160]. Directed histone deacetylation, commonly linked to gene repression, has also been achieved by fusion of histone deacetylase 3 (HDAC3) to dCas9 [167], although generating only a moderate downregulation. TALE-Sin3a fusion constructs on the other hand showed a threefold reduction of Grm2 expression in mouse primary cortical neurons [168]. As the paired amphipathic helix protein Sin3a is only interacting with HDAC3 [169], thus indirectly acts upon histone modifications, and exhibits a cross talk with a plethora of different transcriptional repressors [170], a direct comparison cannot be made. However, it highlights the importance of choosing the right tool when applying targeted histone modification in mammalian cells. Some systems, e.g. TALE-Sin3a or dCas9-p300, could be considered as brute force technologies to increase or decrease gene expression, whereas others enable the delicate and detailed characterization of genomic elements.

111

112

5 Genome Variation, the Epigenome and Cellular Phenotypes

5.5.2.2

Targeted DNA Methylation

Although identification of CpG methylation and its connection to gene expression were understood relatively early, the key enzymes involved in the process remained unresolved for a long time. Only in 1999, the de novo DNA methyltransferases DNMT3a and DNMT3b were identified as essential for mammalian CpG methylation [171] and the TET enzyme family found to be responsible for DNA demethylation in 2011 [172]. Thus, targeted DNA methylation and demethylation in mammalian cells is a comparably young approach [154, 173, 174]. The correlation between methylation of CpG islands in promoter regions and gene expression has been recently shown in CHO cells [38, 80], hence rendering targeted DNA methylation or demethylation tools an interesting alternative for cell line engineering in industrially relevant cell lines. The utility of ZFP-DNMTs was demonstrated when DNMT3a was targeted against SRY (sex determining region Y)-box 2 (SOX2) and Maspin [175]. The efficacy was not very high, however, and most importantly, the downregulation of the targeted gene was hereditary over multiple cell generations. Consequently and opposed to other repression tools, targeted DNA methylation – because of its inherent nature of heredity by maintenance DNMTs – could possibly even be used for long-term therapeutic approaches in vivo [176]. In 2016, the first dCas9-based targeted DNA methylation tool was published by Vojta et al. [177]. Dependent on the sgRNAs used, up to 60% induction of CpG methylation could be achieved. The authors further assessed their tool’s activity radius defining a peak of methylation activity at around 27 bp from the PAM sequence. As reported in other studies, coexpression of several sgRNAs resulted in synergistic effects and in the highest increase in methylation and downregulation of the associated gene (more than 50%). However, opposed to the ZFP-DNMT3a construct, methylation patterns were not inherited. This is likely due to the different delivery methods of the methylation tools. Although the SOX2/Maspin study used viral transduction and hence stably integrating their construct into the host genome, the dCas9-DNMT3A study employed lipofection. In contrast to the dCas9-DNMT3A approach, a new study using a dCas9-DNMT3A-DNMT3L fusion protein revealed that multiplexing is not increasing the efficiency of induced methylation [178]. A reason for the different conclusions might be distinct transfection strategies using different molar ratios of sgRNA:dCas9-construct. Nonetheless, methylation efficiency of the dCas9-DNMT3A-DNMT3L fusion construct was significantly higher (four- to fivefold) compared to DNMT3a alone, which is to be expected as DNMT3L has been reported to stimulate de novo activity of DNMT3A [60]. Also in the same study, Stepper et al. reported a 50% decrease in gene expression, further underlining the usefulness of DNMT-directed DNA methylation for gene regulation [178]. DNA demethylation has been a long sought for mechanism to revert gene silencing in mammalian cells. In 2013, two groups showed independently that targeting TET enzymes to promoter regions by either TALEs or ZFPs resulted in DNA demethylation and activation of the corresponding genes. The TALE-TET1 fusion proteins induced demethylation by >15% in HeLa and HEK293 cells and could upregulate expression of selected genes by up to 14-fold [179]. However,

5.6 Conclusion and Outlook

the applicability of TALE systems for DNA demethylation is limited as TALE binding has been reported to be methylation sensitive and thus confines the application range [180]. The ZFP-TET study could show an up to twofold induction of expression of the targeted gene, but much lower demethylation levels (induction of 3000×) of mtDNA obtained from Chinese hamster and 22 CHO cell lines [25]. Such sequencing depth showed significant heteroplasmy across CHO cell lines, suggesting that differences can exist across clones leading to corresponding variations in production batches. 7.2.1.3

Transcriptomics of CHO Cells

The first efforts in CHO transcriptome sequencing used the Sanger method to sequence expression sequence tags (ESTs) derived from CHO cDNA libraries [1], thereby constructing the first CHO-specific microarray with more than 4000 ESTs. The coverage increased to over 28 000 unigenes through continued sequencing efforts using cDNA libraries constructed with CHO cell lines and Chinese hamster tissue [26]. 454 and Illumina sequencing were later introduced to increase the coverage of the CHO transcriptome [27]. In 2010, Birzele et al. demonstrated the use of Illumina sequencing and a proprietary bioinformatics pipeline to assemble and annotate a CHO transcriptome data set de novo [28]. In 2011, Becker et al. released the first public CHO transcriptome, which was sequenced using the 454 technology [29]. The gene coverage was improved with further sequencing and a hybrid assembly pipeline using both de novo and reference assemblies [30]. In the recent years, microRNAs (miRNA) have garnered attention as a means to engineer CHO cell lines. Hackl et al. deep sequenced small RNA fractions using Illumina sequencing to sequence CHO miRNA from multiple cell lines under different cultivation conditions [31]. In addition to miRNAs that mapped to the mouse genome, novel miRNAs were predicted using machine learning techniques, giving rise to 365 mature miRNA sequences. Similarly, Johnson et al. used Illumina sequencing and homology matching with human, mouse, and rat to identify 350 mature miRNA sequences [32]. Following the publication of the CHO-K1 genome in 2011, the reported miRNA sequences from both earlier works were mapped into the CHO genome to determine the respective miRNA gene loci and pre-miRNA sequences [33]. This is required to mimic endogenous miRNA expression for cell engineering. A total of 212 genomic loci and 319 mature miRNAs were confirmed. The list of CHO miRNAs was expanded to 378 mature miRNA sequences more recently [34]. This number is significantly smaller than other mammals such as human (2588 mature miRNAs) and mouse (1982 mature miRNAs), which may suggest that some miRNAs remain undiscovered [35]. At present, the focus is on identifying relevant miRNA targets to increase recombinant production [36, 37].

167

168

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

7.2.1.4

Epigenomics of CHO Cells

Epigenetic modifications such as histone modifications and DNA methylation have long been shown to influence gene expression [38]. Noncoding RNAs (ncRNAs) are also implicated in modifying gene expression [39]. For CHO cells, epigenomics is still in its infancy. Epigenetic modifications in CHO cells have been observed during cell cultures and across batches [23]. It has also been correlated with recombinant gene expression [40]. These observations highlight its importance for bioproduction and present opportunities for cell line engineering. To detect DNA methylation, bisulfite sequencing can be employed [41]. In this method, bisulfite converts cytosines to uracils, while leaving methylated cytosines unchanged. By sequencing the treated DNA sample with an untreated sample, methylation sites can be found. Wippermann et al. produced the DNA methylation map of CHO DP-12 cell line using bisulfite sequencing with Illumina [42]. For detecting DNA protein interactions, chromatin immunoprecipitation and sequencing (ChIP-seq) was used. In ChIP-seq, the DNA binding proteins are cross-linked to their attached DNA sites and the DNA fragmented. The DNA protein complexes are then immunoprecipitated with protein-specific antibodies and collected for cross-link reversal. The released DNA fragments are sequenced using high-throughput sequencers such as Illumina. Feichtinger et al. employed bisulfite sequencing and ChIP-seq with Illumina to observe CHO epigenomic changes during the time course of a culture, as well as adaption to different media conditions [23]. They observed minor changes in DNA methylation between different growth phases but greater modifications in response to changes in culture conditions. Interestingly, histone modifications showed continuous changes as the cell lines were subject to changes in conditions and selection pressures. 7.2.2 7.2.2.1

Mass Spectrometry-Based Omics Technologies Mass Spectrometry Techniques

Mass spectrometry is an experimental technique most popular in omics fields of study, such as proteomics, metabolomics, lipidomics, and glycomics [43]. In its simplest form, a mass spectrometer is composed of three main components. The first part is the ionizer, where the analyte is ionized and sent to the mass analyzer. The second component, mass analyzer, separates the ions by their mass-to-charge ratio (m/z) and sends them to the final component, the detector that measures the impacting ions and records the abundance at each detected m/z. The resulting mass spectra are then analyzed using computational tools to determine molecular identities and even quantitation. Early ionizers such as electron impact (EI) and chemical ionizers (CIs) could only be used to analyze biomolecules below 1000 Da. Newer ionization techniques such as fast atom bombardment (FAB) and plasma desorption (PD) raised the mass limit, but it took the inventions of electrospray ionization (ESI) and matrix-assisted laser desorption/ionization (MALDI) in the 1980s to revolutionize the use of mass spectrometry in biological studies. Both ESI and MALDI offer virtually unlimited mass range, assuring their popularity in studies on macromolecules. A refinement of ESI, nano-ESI, reduces

7.2 High-Throughput Omics Technologies

the flow requirements by 3 orders of magnitude, making it highly efficient for biological samples. There are a number of different mass analyzers in use, which can be further combined to create even more configurations to meet specific needs. One commonly used mass analyzer is the time-of-flight (TOF), where analyte molecules are imparted with the same kinetic energy using an electric field. Because of the different m/z ratios, heavier ions travel slower, thus resulting in separation. The m/z ratio can then be determined based on the travel time (hence TOF). Another common mass analyzer is the quadrupole analyzer, which consists of four parallel rods with alternating electric currents. Only ions in a selected small m/z range will travel straight through the center while the remainder will oscillate and impact the rods. By modifying the voltages, ions of different m/z ratios can be selected for detection. Some mass analyzers operate on the principle of ion trapping. The quadrupole ion trap (QIT) first traps analyte ions of specific m/z ratios in an oscillating electric field generated by quadrupoles, followed by controlled release toward the mass detector. A variant reduces the trapping field from 3D to 2D and is referred to as a linear ion trap (LIT). The Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer uses a magnetic field trap (the ICR cell) to contain analyte ions and force them into their respective cyclotron rotational radius. As ions, their movement past electrodes will induce signals, which are recorded and the data are extracted using Fourier transform. FT-ICRs are able to achieve far higher mass resolution than other mass analyzers at the cost of data acquisition speeds. A related design, the Orbitrap, uses static electrostatic fields to trap analyte ions in an orbit around an electrode and force them into axial oscillations. Similar to FT-ICRs, the ions are detected by inducing signals when moving past electrodes, followed by data processing with Fourier transform. Tandem mass spectrometry, or MS/MS, combines multiple mass analyzers to select ions for fragmentation and further analysis. The aim is to observe the resulting fragments to identify the precursor ions. A commonly seen hybrid is the triple quadrupole or “QqQ” mass spectrometer, where three quadrupole units are used in series. The first quadrupole is used to select precursor ions of specific m/z ratios to be fragmented by collisions in the second quadrupole cell, followed by the scanning of daughter ions with the third quadrupole. This configuration is particularly suited for targeted experiments with high sensitivity. Another common configuration is the QTOF, where the third quadrupole is replaced by a TOF. This offers increased mass resolution because of the TOF component. Other tandem designs include TOF-TOF, LIT-ICR, and LIT-Orbitrap. Mass spectrometers are commonly paired with liquid chromatography (LC), or LC–MS, for various omics studies. The commonly used LC systems are high-performance liquid chromatography (HPLC) and ultra-high-performance liquid chromatography (UPLC). LC offers chemical separation capabilities that are orthogonal to the physical separation by mass in MS. Commonly used HPLCs include reverse-phase, ion-exchange, and size-exclusion. The time taken for an ion to traverse the column (retention time) can also be used to differentiate between isomers, including stereoisomers. Most LC–MS configurations employ the ESI/nano-ESI ion source because of the ease of direct linkage. To some

169

170

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

extent, LC can be used by collecting the elution at different time intervals to obtain discrete samples to be used with a MALDI source, but is comparatively less popular. The ESI and MALDI ion sources enabled MS to be used to study macromolecules, but the explosion in genomic information from sequencing technologies greatly boosted MS-based studies of proteins. MS proteomic studies can be divided into two types: top-down and bottom-up. In top-down, whole proteins are sent into the MS for analysis. This approach is suited for identifying different proteoforms and posttranslational modifications (PTM)s, as whole proteins are analyzed. Protein identities are established based on their molecular weights (including PTMs). However, this approach is mainly hampered by difficulties in protein solubility, MS ionization, and chromatographic separation [44]. By contrast, the bottom-up approach currently dominates proteomic studies. In bottom-up, complex protein samples can be first fractionated using gel-based electrophoresis [45]. Selected sample fractions are then digested using highly specific enzymes such as trypsin, followed by LC–MS or LC–MS/MS analysis. In the shotgun approach or discovery approach, all peptide ions are scanned. To identify peptides and hence proteins, the detected ions are then matched against a list of expected peptides, a process called peptide mass fingerprinting. This list is generated by performing “in silico digestion” on a database of expected proteins to generate expected peptides. Because of peptides having different ionization efficiencies, not all peptides are detected. Signals of low-abundance peptide ions may be drowned out by high-abundance ions. As more peptides from a protein are detected, the greater the confidence in confirming protein identity. If MS/MS is employed, the second MS can aid peptide identification through characteristic fragments. Available database search tools include MASCOT, SEQUEST, and X!Tandem [46–48]. 7.2.2.2

Proteomics of CHO Cells

Before the CHO genome sequence publication, only a limited number of CHO protein sequences were available, which necessitates searches against more comprehensive protein databases of human, mouse, and rat to annotate collected MS data [49–51]. Alternatively, proteins were identified on 2D gels with annotated proteome maps [52]. The resulting coverage of the CHO proteome was poor, with approximately 500 proteins identified using surrogate sequences [50, 53]. Using the CHO genome sequence, Baycin-Hizal et al. expanded the coverage to 6164 unique proteins using multiple LC separation techniques and shotgun proteomics [10]. To perform absolute protein quantitation with MS, absolute quantification (AQUA) and quantification concatamers (QconCAT) are available methods. However, they require the synthesis of isotopic internal standards, driving up cost and labor requirements. Thus, relative quantitation with isotopic labeling methods is more popular, such as isotope-coded affinity tag reagents (ICAT), stable isotope labeling with amino acids (SILAC), and isobaric tags for relative and absolute quantitation (iTRAQ) [54]. In particular, SILAC is a popular method for cell cultures. In CHO proteome and recombinant product profiling, iTRAQ and SILAC methods have been successfully employed [26, 49, 55]. However, such isotopic label methods still involve additional costs. Alternative to isotopic

7.2 High-Throughput Omics Technologies

labels, there are two label-free methods, peptide signal intensity, and peptide fragments spectral counting [56]. Such label-free methods are considered only semiquantitative, but spectral counting has been gaining popularity. Reasons for this popularity include applicability to the entire list of detected peptides and no additional experimental cost. Spectral counting has been applied to CHO proteome studies [10, 50]. Selected reaction monitoring (SRM) is a targeted approach using MS/MS, where only a specified list of masses are selected at the first MS stage for fragmentation and analysis. As only specific masses are selected and analyzed, SRM can achieve better sensitivity at the expense of reduced peptide coverage [57]. SRM is often employed with other isotopic labeling methods, or label-free methods, for quantitation purposes. To counter the limited coverage of SRM, sequential window acquisition of theoretical mass spectra (SWATH) MS was developed using fast, high-resolution MS/MS machines [58]. At the first MS stage, selected windows repeatedly cycled through for precursor ions to be sent for fragmentation and analysis. For proteomics applications, the method typically focuses on the range of 400–1200 m/z where most peptides fall into, divided into 25 m/z windows. So far, SWATH MS is rarely employed in CHO proteome studies [59]. 7.2.2.3

Metabolomics/Lipidomics of CHO Cells

In metabolomics, nuclear magnetic resonance (NMR) and MS are the most common methods [60]. Although NMR has advantages of being quantitative and lack of sample preparation requirements, its disadvantages include lack of sensitivity and higher upfront and running costs. These contribute to the current dominance of MS in metabolomics studies. MS-based metabolomics share many similarities with MS proteomics. For example, MS is commonly paired with chromatography methods such as LC or GC. Metabolite MS peaks are identified by matching against known metabolite ion masses. Tandem MS can be employed to differentiate isomers through characteristic fragmentations, although open spectral libraries are limited, and fragmentation patterns are known to differ across MS machines. SRM and SWATH techniques can also be applied to metabolomics study. Unlike proteomics, the identification of metabolites remains the biggest challenge because of the large number of isomers, adducts, and fragments, which grossly inflate the number of putative identities. Different classes of metabolites also show different fragment patterns, preventing us from deriving generalized rules. These difficulties significantly limit the number of metabolites identifiable, namely those with available standards and chromatographic retention times to aid identification. To quantify a metabolite, a calibration curve has to be created by spiking in an isotope-labeled synthetic analog of known concentration. MS-based metabolomics has been used to profile CHO intracellular and extracellular metabolome [61–64]. Currently, the metabolite coverage is mostly concentrated in the major metabolites of glycolysis pathway, tricarboxylic acid (TCA) cycle, amino acids, and nucleosides. Nevertheless, the current coverage is useful in gaining insights into CHO metabolism. To tackle this, various computational tools and open spectral libraries are being developed to improve the metabolite identification process and expand metabolite coverage [65–67].

171

172

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

Lipidomics, as a subset of metabolomics, is a relatively recent omics field that has seen significant advances [68]. Like metabolomics, the study of lipids using LC–MS presents similar challenges in identifying detected metabolites. Hence, there is a growing interest in developing in silico tools toward solving this problem [69–71]. Currently, lipidomic studies into mammalian cells have highlighted significant differences in lipid profiles between SP2/0, baby hamster kidney (BHK), and CHO cell lines. Lipidome differences between wild-type and high-producer CHO-K1 cell lines have also been elucidated [7]. As lipids have been implicated in protein production and secretory capacities, more CHO lipidome studies should be explored in the future. 7.2.2.4

Glycomics of CHO Cells

As a major recombinant biotherapeutics producer, the glycoforms produced by CHO cells are of great importance [72]. Although glycans are found on host cell proteins and lipids, the focus of CHO glycomics studies is on glycosylation of recombinant proteins. In particular, N-glycans are of great importance to product efficacy and safety. Hence, there are efforts to reduce N-glycan heterogeneity, and even engineering optimized N-glycan structures [73]. LC and LC–MS are popular tools employed in glycan analysis [74]. Glycoproteins can be analyzed as a whole, or reduced to heavy chains and light chains (for antibodies), or only released glycans enzymatically cleaved from the protein. Glycan profiling of intact glycoproteins or glycopeptides typically employ LC–MS or capillary electrophoresis (CE) in place of LC. The heterogeneity of N-glycans complicates whole glycoproteins analysis, while glycopeptide isomers are difficult to distinguish. Analysis of free glycans is comparatively simpler but glycan cleavage reactions may be incomplete, resulting in unreleased glycans. Free glycans can be analyzed using chromatographic methods including MS, CE, or hydrophilic interaction liquid chromatography (HILIC) [75]. To analyze the N-glycan structure, enzymes are used to sequentially remove terminal monosaccharides, followed by analysis with chromatography. MS paired with chromatography offers a high-throughput workflow with high sensitivity and without the requiring sequential monosaccharide removal, but this approach has difficulties distinguishing isomers. Currently, numerous studies with CHO cells employ MS-based glycan profiling [76–78]. For glycan quantitation, isotopic tags with MS may be employed [74]. Recently, lectin-based microarrays have been developed to offer an alternative approach to glycan profiling. It offers a relatively simple procedure that works directly with glycoproteins. However, current lectin microarrays using plant-derived lectins can suffer from poor specificity [73].

7.3 Current CHO Multi-omics Applications With increasingly successful high-throughput technologies in different omics, it is natural to combine different omics data to achieve greater insights into cell culture behavior and guide cellular engineering. To date, there have been a number

7.3 Current CHO Multi-omics Applications

Table 7.1 Studies of CHO cells combining more than two omics data. Study aim

Type of omics

Subjects for study

References

Bioprocess characterization

Transcriptome Proteome

Low temperature vs. normal temperature

[79]

Cell line characterization

Transcriptome Proteome

High-productivity CHO cells vs. low-productivity CHO cells

[49]

Culture characterization

Transcriptome Proteome

NaBu treatment vs. normal condition

[80]

Cell line characterization

Transcriptome Proteome

High growth rate CHO cells vs. low growth rate CHO cells

[51]

Bioprocess/culture characterization

Transcriptome Proteome

Low-temperature and NaBu treatment vs. normal condition

[81]

Cell line characterization

Transcriptome Proteome

17 different clones exhibiting different phenotypes

[82]

Culture characterization

Proteomics Glyco-profiling

Glucose starvation and culture duration

[83]

Evaluation of genome database

Genome Transcriptome

Evaluation of public genomic references for RNA-seq data

[84]

Cell line characterization

Transcriptome Metabolome

CHO-K1 WT vs. EPO-producing clone

[85]

Cell line characterization

Transcriptome Proteome

High-productivity CHO cells vs. low productivity CHO cells

[86] [59]

Cell line characterization

Epigenome Transcriptome

DNA methylation mapping for mAb producing clone

[42]

Synthetic codon design

Genome Transcriptome Translatome Proteome

Previous databases

[87]

Cell line characterization

Metabolome Proteome

CHO-DXB11 WT vs. mAb producing clone

[88]

Bioprocess characterization

Transcriptome Metabolome

Reduced oxygen level in the process to control sialylation

[89]

Culture characterization

Transcriptome Metabolomics

Hyperosmolality vs. normal osmolality

[90]

Cell line characterization

Genome Transcriptome

CHO-K1 WT vs. XylT2 mutant

[91]

Cell line characterization

Genome Transcriptome Metabolome Glycome Lipidome

CHO-KI WT and mAb producing clone

[7]

Cell line characterization

Transcriptome Lipidome

SP2/0, CHO, and HEK-293F

[9]

173

174

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

of published studies that combine two or more omics data sets to study CHO cells. These studies are summarized in Table 7.1. 7.3.1

Bioprocess Optimization

Low-temperature culture is a widely used technique in CHO cell cultures to increase recombinant protein-specific productivity. Baik et al. combined transcriptome and proteome data to study erythropoietin (EPO)-producing CHO cell cultures at two different temperatures (33 and 37 ∘ C) [79]. They were able to identify differentially expressed proteins that were associated with lower growth rate as well as higher productivity of low-culture temperature. In another study, high- and low-productivity clones were compared using transcriptome and proteome data [49]. The authors found a lack of correlation between mRNA and protein levels. However, regulation of genes/proteins in both data sets concurs in terms of functional categorization. Interestingly, the differentially expressed genes and proteins include enzymes involved in chromatin modification. Multi-omics profiling can also be harnessed to scrutinize the effects of manufacturing bioprocess conditions on cell line phenotypes such as product glycosylation patterns. The cause of batch-to-batch product sialylation variability in CHO cell culture at manufacturing scale was investigated by combining transcriptome with metabolome analysis in a scaled down pilot plant system [89]. They demonstrated that the sialylation variations resulted from the oxidative stress and shifted glucose metabolism due to low oxygen availability. This was correlated well with the need for reduced dissolved oxygen set point to reproduce the phenotypic profiles at manufacturing scale. The authors proposed a biological mechanism linking sialylation variations with oxidative stress due to large-scale manufacturing. 7.3.2

Cell Line Characterization

Recently, Zhang et al. profiled transcriptome and lipidome to characterize the differences between three mammalian cell lines (SP2/0, CHO, and HEK-293) [9]. Transcriptome data from Illumina RNA-seq and lipid profiling high-performance thin layer chromatography (HP-TLC) and MS were used in the study. HEK cells were found to contain higher levels of lyso phosphatidylethanolamine (LPE) and lyso phosphatidylcholine (LPC) than SP2/0 and CHO, while a lower level of sphingomyelin was found in SP2/0 cells than others. The transcriptome data were useful in identifying differentially regulated genes within the respective lipid pathways that can contribute to the observed differences in lipid composition. More recently, a large-scale multi-omics profiling was applied to characterize differences between the wild-type CHO-K1 and high producer [7]. In this study, genomic, transcriptomic, metabolomic, lipidomic, and glycomic data were integrated using the CHO genome-scale model to elucidate genotypic and phenotypic differences between host and producer cells (Figure 7.2). The authors observed extensive genomic deletions and rearrangements that result in gene copy-number changes, especially at the antibody gene integration

DNA damage repair Genomics Gene copy number variations 4X

2X

Xrcc1

Tep1

Cell structure

Cers2

Cytoskeleton lipids

Mitochondria Ndufa3

Neil3

Apoptosis survival

Mtus1

Poln

0.5X

Transcriptomics Differential gene expression Up

Ercc1

Energy

Ercc1

Xrcc1

Ndufa3

Neil3 Tep1

Mtus1

Cers2

Galc

Nd3

Pla2g2a

Pcyt1b

Ext1

Poln

Chpf

Chst12

Dwn Oxidative phosphorylation

Metabolomics Metabolite profiling Up

NAD+

FAD

GSH

GSSG

Ceramide metabolism SM GC

Cer CerP

Phosphatidylcholine synthesis LysoPC

PhCho

Cho

Dwn

Productivity traits emerge at genome level No major changes in N-glycosylation

Physiological characteristics

Longer chain fatty acids More favorable membrane structure

Targets for vector integration sites

Figure 7.2 Multi-omic data integration using CHO genome-scale model to unravel the global cellular changes in CHO-K1 and mAb producing clone, SH87.

176

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

sites. Mitochondrial gene copy numbers were also found to be increased, presumably because of a metabolic burden increase from recombinant protein expression. Altered gene expression profiles, in particular DNA damage repair, lipid metabolism, and metabolic genes, were also found in the producer cells. Interestingly, lipidomic analysis revealed that producer cells have higher levels of long-chain lipid species, which are hypothesized to be linked to protein transport and secretion. These altered phenotypes in the producer cell line show how the cells reshape their intracellular pathways to handle the metabolic requirements of recombinant protein production. Such a systematic integration of different omics data, aided by a high-quality metabolic model, successfully demonstrated an effective multi-omics framework for holistic characterization of CHO cells. 7.3.3

Engineering Target Identification

In addition to the characterization of cells, another major goal of omics-based studies is to identify cellular engineering targets to improve cell line performance, especially in terms of growth, resistance to apoptosis, product secretion, and product quality. A few successful examples of engineering target identification through multi-omics profiling are discussed below. Doolan et al. employed microarray transcriptomics and MS proteomics to analyze differences between high- and low-growth CHO cells to identify engineering targets [51]. Differentially expressed genes and proteins were identified and a priority list of 21 targets was generated by overlap comparison. siRNA sequences were designed and used for knockdown studies with five growth-related genes. In particular, knock down of the Valosin-containing protein was found to have the greatest negative impact on growth and viability. Other knockdown genes such as HSPB1 and ENO1 have negative and positive impacts, respectively. In a recent study, transcriptome and metabolome analysis were combined to analyze the biological response to hyperosmotic stress in CHO cells [90]. This study aimed to identify targets genes to increase specific productivity in hyperosmotic culture conditions. In a prior study, hyperosmolarity was correlated with increased ATP availability. Therefore, the authors sought to understand the mechanisms behind the hyperosmotic phenotype and hence engineering targets to increase ATP formation, suggesting a number of proteins and miRNA targets. One of the identified miRNAs has also been highlighted to be most promising based on the correlation of target gene with specific productivity, which has been experimentally confirmed independently [92, 93]. In another multi-omics study, MS methods (iTRAQ and SWATH) were used to identify differentially expressed genes between low-producer and high-producer clones [86]. Interestingly, the SWATH method gave a much larger number of differentially regulated genes. These results were checked using RNA-seq, with 63% of genes being identified as differentially regulated with both omics. By gene enrichment, a number of differentially regulated pathways were identified in the high producer, including upregulated glutathione biosynthesis and downregulated DNA replication. Furthermore, a high level of glutathione in the high producer was measured by metabolomics analysis, suggesting the importance of

7.4 Future Prospects

reactive oxygen species (ROS). In a follow-up study, CHO cells were engineered to overexpress two targets (GCLC and GCLM) in GSH synthesis pathway, each gene in stably transfected to separate clones [59]. The cell lines overexpressing GCLM showed 70% higher antibody production in transient transfections, confirming the beneficial role of decreasing ROS in cellular productivity.

7.4 Future Prospects With high-throughput omics technologies generating large volumes of data in CHO bioprocessing, the next key step is to perform effectively integrative analysis of the data; different omics data sets present various modular data of cells at different levels of cellular hierarchy. Therefore, utilizing all available omics data in a combined manner is necessary to build a more comprehensive picture, as such linking the genotype to phenotype systematically. For example, complementary information from different omics can increase confidence in the analysis, while unique information can also be inferred from individual omics. However, the integration of diverse multiple-omics data sets is highly challenging because of the differences in experimental methodology used and the scales of the high-throughput data generated [94]. As a result, such diverse data cannot be easily normalized or standardized using a single pipeline, thus requiring sophisticated computational approaches to integrate them. Multi-omics data integration is not restricted to bioproduction problems; in fact, they are now widely employed in cancer research [95]. With the availability of several large-scale multi-omics databases such as the cancer genome atlas (TCGA) and cancer cell line encyclopedia (CCLE), which contain experimental data on gene copy numbers, expression levels, and DNA methylation, recently, a wide range of statistical and deep learning tools have been developed to handle such data from disparate sources [96, 97]. These tools can link genotypes with phenotypes to discover engineering targets, as well as correlate cellular responses with environmental changes. Broadly, the multi-omic data integration approaches can be categorized into two classes based on the mathematical approaches used: statistical methods and network-based modeling. The statistical methods such as multivariate comparison of individual omics data can reveal their regularities or irregularities [98]. For example, such analysis of transcriptomics and proteomics data can provide various information about the existence of posttranscriptional mechanisms for the given condition. Similarly, correlation and regression-based metrics can also be utilized to assess the relationship between two different omics data sets either in original scale or in transformed coordinates. It should be noted that such statistical methods can be either supervised or unsupervised depending on the availability of corresponding phenotypic data; if we have the outcomes of two conditions being assessed, e.g. wild type vs. producer, then we can use appropriate techniques that consider the known differences as well. Network-based methods, on the other hand, combine the omics data onto existing biological networks such as metabolism, signaling, and protein–protein

177

178

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

interactions, which contain various functional entities such as genes, proteins, and metabolites. It then assesses the enrichment of particular pathways or cellular processes across different omics. Biological networks from databases such as Kyoto Encyclopedia of Genes and Genomes (KEGG) and MetaCyc are typically used for such purposes. Furthermore, network-based modeling approaches are also used to link heterogeneous omics data. Recently, the CHO genome-scale model has been reconstructed, serving as a valuable resource that links genes, metabolic reactions, and metabolic pathways [4]. This model can be used as a scaffold to link the transcriptomic, metabolomic, lipidomic, and glycomic data sets as it has been successfully demonstrated in a recent publication (Figure 7.2) [7]. At the present moment, the CHO genome-scale model encompasses only biological entities corresponding to metabolism. Therefore, a more comprehensive model incorporating additional layers of cellular mechanisms including signaling pathways and transcriptional regulation can be even more useful. However, given the complexity of mammalian cells, this presents a formidable but exciting challenge.

References 1 Jayapal, K.P., Wlaschin, K.F., Hu, W.S., and Yap, M.G.S. (2007). Recombinant

2 3 4

5

6 7

8

9

10

protein therapeutics from CHO cells – 20 years and counting. Chem. Eng. Prog. 103: 40–47. Xu, X., Nagarajan, H., Lewis, N.E. et al. (2011). The genomic sequence of the Chinese hamster ovary (CHO)-K1 cell line. Nat. Biotechnol. 29: 735–741. Wurm, F.M. and Hacker, D. (2011). First CHO genome. Nat. Biotechnol. 29: 718–720. Hefzi, H., Ang, K.S., Hanscho, M. et al. (2016). A consensus genome-scale reconstruction of Chinese hamster ovary cell metabolism. Cell Syst. 3: 434–443.e438. Möller, J. and Pörtner, R. (2017). Model-based design of process strategies for cell culture bioprocesses: state of the art and new perspectives. In: New Insights into Cell Culture Technology, Chapter 05 (ed. S.J.T. Gowder), 157–172. Rijeka: InTech. Lee, S.Y., Lee, D.Y., and Kim, T.Y. (2005). Systems biotechnology for strain improvement. Trends Biotechnol. 23: 349–358. Yusufi, F.N.K., Lakshmanan, M., Ho, Y.S. et al. (2017). Mammalian systems biotechnology reveals global cellular adaptations in a recombinant CHO cell line. Cell Syst. 4: 530–542.e536. Lewis, N.E., Liu, X., Li, Y. et al. (2013). Genomic landscapes of Chinese hamster ovary cell lines as revealed by the Cricetulus griseus draft genome. Nat. Biotechnol. 31: 759–765. Zhang, Y., Baycin-Hizal, D., Kumar, A. et al. (2017). High-throughput lipidomic and transcriptomic analysis to compare SP2/0, CHO, and HEK-293 mammalian cell lines. Anal. Chem. 89: 1477–1485. Baycin-Hizal, D., Tabb, D.L., Chaerkady, R. et al. (2012). Proteomic analysis of Chinese hamster ovary cells. J. Proteome Res. 11: 5265–5276.

References

11 Heather, J.M. and Chain, B. (2016). The sequence of sequencers: the history

of sequencing DNA. Genomics 107: 1–8. 12 Lander, E.S., Linton, L.M., Birren, B. et al. (2001). Initial sequencing and anal-

ysis of the human genome. Nature 409: 860–921. 13 Venter, J.C., Adams, M.D., Myers, E.W. et al. (2001). The sequence of the

human genome. Science 291: 1304–1351. 14 Escalona, M., Rocha, S., and Posada, D. (2016). A comparison of tools for the

15 16

17 18 19

20

21

22 23

24

25

26

27

28

simulation of genomic next-generation sequencing data. Nat. Rev. Genet. 17: 459–469. McCarthy, A. (2010). Third generation DNA sequencing: pacific biosciences’ single molecule real time technology. Chem. Biol. 17: 675–676. Zhao, S., Fung-Leung, W.P., Bittner, A. et al. (2014). Comparison of RNA-Seq and microarray in transcriptome profiling of activated T cells. PloS One 9: e78644. Li, B. and Dewey, C.N. (2011). RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinf. 12: 323. Dobin, A. and Gingeras, T.R. (2015). Mapping RNA-seq reads with STAR. Curr. Protoc. Bioinf. 51: 11.14.1–11.14.19. Bi, Y. and Davuluri, R.V. (2013). NPEBseq: nonparametric empirical Bayesian-based procedure for differential expression analysis of RNA-seq data. BMC Bioinf. 14: 262. Kaas, C.S., Kristensen, C., Betenbaugh, M.J., and Andersen, M.R. (2015). Sequencing the CHO DXB11 genome reveals regional variations in genomic stability and haploidy. BMC Genom. 16: 160. Derouazi, M., Martinet, D., Besuchet Schmutz, N. et al. (2006). Genetic characterization of CHO production host DG44 and derivative recombinant cell lines. Biochem. Biophys. Res. Commun. 340: 1069–1077. Brinkrolf, K., Rupp, O., Laux, H. et al. (2013). Chinese hamster genome sequenced from sorted chromosomes. Nat. Biotechnol. 31: 694–695. Feichtinger, J., Hernandez, I., Fischer, C. et al. (2016). Comprehensive genome and epigenome characterization of CHO cells in response to evolutionary pressures and over time. Biotechnol. Bioeng. 113: 2241–2253. Partridge, M.A., Davidson, M.M., and Hei, T.K. (2007). The complete nucleotide sequence of Chinese hamster (Cricetulus griseus) mitochondrial DNA. DNA Seq. 18: 341–346. Kelly, P.S., Clarke, C., Costello, A. et al. (2017). Ultra-deep next generation mitochondrial genome sequencing reveals widespread heteroplasmy in Chinese hamster ovary cells. Metab. Eng. 41: 11–22. Kantardjieff, A., Nissom, P.M., Chuah, S.H. et al. (2009). Developing genomic platforms for Chinese hamster ovary cells. Biotechnol. Adv. 27: 1028–1035. Jacob, N.M., Kantardjieff, A., Yusufi, F.N. et al. (2010). Reaching the depth of the Chinese hamster ovary cell transcriptome. Biotechnol. Bioeng. 105: 1002–1009. Birzele, F., Schaub, J., Rust, W. et al. (2010). Into the unknown: expression profiling without genome sequence information in CHO by next generation sequencing. Nucleic Acids Res. 38: 3999–4010.

179

180

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

29 Becker, J., Hackl, M., Rupp, O. et al. (2011). Unraveling the Chinese hamster

30

31

32 33

34

35

36

37

38 39 40

41 42

43 44 45

ovary cell line transcriptome by next-generation sequencing. J. Biotechnol. 156: 227–235. Rupp, O., Becker, J., Brinkrolf, K. et al. (2014). Construction of a public CHO cell line transcript database using versatile bioinformatics analysis pipelines. PloS One 9: e85568. Hackl, M., Jakobi, T., Blom, J. et al. (2011). Next-generation sequencing of the Chinese hamster ovary microRNA transcriptome: identification, annotation and profiling of microRNAs as targets for cellular engineering. J. Biotechnol. 153: 62–75. Johnson, K.C., Jacob, N.M., Nissom, P.M. et al. (2011). Conserved microRNAs in Chinese hamster ovary cell lines. Biotechnol. Bioeng. 108: 475–480. Hackl, M., Jadhav, V., Jakobi, T. et al. (2012). Computational identification of microRNA gene loci and precursor microRNA sequences in CHO cell lines. J. Biotechnol. 158: 151–155. Diendorfer, A.B., Hackl, M., Klanert, G. et al. (2015). Annotation of additional evolutionary conserved microRNAs in CHO cells from updated genomic data. Biotechnol. Bioeng. 112: 1488–1493. Fischer, S., Marquart, K.F., Pieper, L.A. et al. (2017). miRNA engineering of CHO cells facilitates production of difficult-to-express proteins and increases success in cell line development. Biotechnol. Bioeng. 114: 1495–1510. Jadhav, V., Hackl, M., Bort, J.A. et al. (2012). A screening method to assess biological effects of microRNA overexpression in Chinese hamster ovary cells. Biotechnol. Bioeng. 109: 1376–1385. Fischer, S., Handrick, R., and Otte, K. (2015). The art of CHO cell engineering: a comprehensive retrospect and future perspectives. Biotechnol. Adv. 33: 1878–1896. Gibney, E.R. and Nolan, C.M. (2010). Epigenetics and gene expression. Heredity 105: 4–13. Peschansky, V.J. and Wahlestedt, C. (2014). Non-coding RNAs as direct and indirect modulators of epigenetic regulation. Epigenetics 9: 3–12. Veith, N., Ziehr, H., MacLeod, R.A., and Reamon-Buettner, S.M. (2016). Mechanisms underlying epigenetic and transcriptional heterogeneity in Chinese hamster ovary (CHO) cell lines. BMC Biotechnol. 16: 6. Li, Y. and Tollefsbol, T.O. (2011). DNA methylation detection: bisulfite genomic sequencing analysis. Methods Mol. Biol. 791: 11–21. Wippermann, A., Rupp, O., Brinkrolf, K. et al. (2015). The DNA methylation landscape of Chinese hamster ovary (CHO) DP-12 cells. J. Biotechnol. 199: 38–46. Di Girolamo, F., Lante, I., Muraca, M., and Putignani, L. (2013). The role of mass spectrometry in the "Omics" Era. Curr. Org. Chem. 17: 2891–2905. Gregorich, Z.R. and Ge, Y. (2014). Top-down proteomics in health and disease: challenges and opportunities. Proteomics 14: 1195–1210. Zhang, Y., Fonslow, B.R., Shan, B. et al. (2013). Protein analysis by shotgun/bottom-up proteomics. Chem. Rev. 113: 2343–2394.

References

46 Perkins, D.N., Pappin, D.J., Creasy, D.M., and Cottrell, J.S. (1999).

47

48 49

50

51

52 53

54

55

56 57 58

59

60

Probability-based protein identification by searching sequence databases using mass spectrometry data. Electrophoresis 20: 3551–3567. Eng, J.K., McCormack, A.L., and Yates, J.R. (1994). An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. J. Am. Soc. Mass Spectrom. 5: 976–989. Craig, R. and Beavis, R.C. (2004). TANDEM: matching proteins with tandem mass spectra. Bioinformatics 20: 1466–1467. Nissom, P.M., Sanny, A., Kok, Y.J. et al. (2006). Transcriptome and proteome profiling to understanding the biology of high productivity CHO cells. Mol. Biotechnol. 34: 125–140. Carlage, T., Hincapie, M., Zang, L. et al. (2009). Proteomic profiling of a high-producing Chinese hamster ovary cell culture. Anal. Chem. 81: 7357–7362. Doolan, P., Meleady, P., Barron, N. et al. (2010). Microarray and proteomics expression profiling identifies several candidates, including the valosin-containing protein (VCP), involved in regulating high cellular growth rate in production CHO cell lines. Biotechnol. Bioeng. 106: 42–56. Kuystermans, D., Dunn, M.J., and Al-Rubeai, M. (2010). A proteomic study of cMyc improvement of CHO culture. BMC Biotechnol. 10: 25. Ahram, M., Strittmatter, E.F., Monroe, M.E. et al. (2005). Identification of shed proteins from Chinese hamster ovary cells: application of statistical confidence using human and mouse protein databases. Proteomics 5: 1815–1826. Chahrour, O., Cobice, D., and Malone, J. (2015). Stable isotope labelling methods in mass spectrometry-based quantitative proteomics. J. Pharm. Biomed. Anal. 113: 2–20. Carlage, T., Kshirsagar, R., Zang, L. et al. (2012). Analysis of dynamic changes in the proteome of a Bcl-XL overexpressing Chinese hamster ovary cell culture during exponential and stationary phases. Biotechnol. Progr. 28: 814–823. Lundgren, D.H., Hwang, S.I., Wu, L., and Han, D.K. (2010). Role of spectral counting in quantitative proteomics. Exp. Rev. Proteomics 7: 39–53. Lange, V., Picotti, P., Domon, B., and Aebersold, R. (2008). Selected reaction monitoring for quantitative proteomics: a tutorial. Mol. Syst. Biol. 4: 222. Gillet, L.C., Navarro, P., Tate, S. et al. (2012). Targeted data extraction of the MS/MS spectra generated by data-independent acquisition: a new concept for consistent and accurate proteome analysis. Mol. Cell. Proteomics 11: O111.016717. Orellana, C.A., Marcellin, E., Gray, P.P., and Nielsen, L.K. (2017). Overexpression of the regulatory subunit of glutamate-cysteine ligase enhances monoclonal antibody production in CHO cells. Biotechnol. Bioeng. 114: 1825–1836. Emwas, A.H. (2015). The strengths and weaknesses of NMR spectroscopy and mass spectrometry with particular focus on metabolomics research. Methods Mol. Biol. 1277: 161–193.

181

182

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

61 Chong, W.P., Yusufi, F.N., Lee, D.Y. et al. (2011). Metabolomics-based iden-

62

63

64

65

66 67

68

69

70 71

72 73 74

75

76

77

tification of apoptosis-inducing metabolites in recombinant fed-batch CHO culture media. J. Biotechnol. 151: 218–224. Selvarasu, S., Ho, Y.S., Chong, W.P. et al. (2012). Combined in silico modeling and metabolomics analysis to characterize fed-batch CHO cell culture. Biotechnol. Bioeng. 109: 1415–1429. Mohmad-Saberi, S.E., Hashim, Y.Z., Mel, M. et al. (2013). Metabolomics profiling of extracellular metabolites in CHO-K1 cells cultured in different types of growth media. Cytotechnology 65: 577–586. Dietmair, S., Hodson, M.P., Quek, L.E. et al. (2012). Metabolite profiling of CHO cells with different growth characteristics. Biotechnol. Bioeng. 109: 1404–1414. Zhou, B., Xiao, J.F., and Ressom, H.W. (2013). Prioritization of putative metabolite identifications in LC–MS/MS experiments using a computational pipeline. Proteomics 13: 248–260. Shahaf, N., Rogachev, I., Heinig, U. et al. (2016). The WEIZMASS spectral library for high-confidence metabolite identification. Nat. Commun. 7: 12423. Lee, T.S., Ho, Y.S., Yeo, H.C. et al. (2013). Precursor mass prediction by clustering ionization products in LC–MS-based metabolomics. Metabolomics 9: 1301–1310. Cajka, T. and Fiehn, O. (2014). Comprehensive analysis of lipids in biological systems by liquid chromatography-mass spectrometry. TRAC, Trends Anal. Chem. 61: 192–206. Yeo, H.C., Chung, B.K.-S., Chong, W. et al. (2015). A genetic algorithm-based approach for pre-processing metabolomics and lipidomics LC–MS data. Metabolomics 12: 5. Kind, T., Liu, K.H., Lee, D.Y. et al. (2013). LipidBlast in silico tandem mass spectrometry database for lipid identification. Nat. Methods 10: 755–758. Kyle, J.E., Crowell, K.L., Casey, C.P. et al. (2017). LIQUID: an-open source software for identifying lipids in LC–MS/MS-based lipidomics data. Bioinformatics 33: 1744–1746. Sha, S., Agarabi, C., Brorson, K. et al. (2016). N-glycosylation design and control of therapeutic monoclonal antibodies. Trends Biotechnol. 34: 835–846. Zhang, L., Luo, S., and Zhang, B. (2016). Glycan analysis of therapeutic glycoproteins. mAbs 8: 205–215. Yang, X., Wang, Z., Guo, L. et al. (2017). Proteome-wide analysis of N-glycosylation stoichiometry using SWATH technology. J. Proteome Res. 16: 3830–3840. Doherty, M., McManus, C.A., Duke, R., and Rudd, P.M. (2012). High-throughput quantitative N-glycan analysis of glycoproteins. Methods Mol. Biol. 899: 293–313. Tep, S., Hincapie, M., and Hancock, W.S. (2012). The characterization and quantitation of glycomic changes in CHO cells during a bioreactor campaign. Biotechnol. Bioeng. 109: 3007–3017. Bennun, S.V., Yarema, K.J., Betenbaugh, M.J., and Krambeck, F.J. (2013). Integration of the transcriptome and glycome for identification of glycan cell signatures. PLoS Comput. Biol. 9: e1002813.

References

78 North, S.J., Huang, H.H., Sundaram, S. et al. (2010). Glycomics profiling of

79

80

81

82 83

84

85

86

87

88

89 90

91

92

Chinese hamster ovary cell glycosylation mutants reveals N-glycans of a novel size and complexity. J. Biol. Chem. 285: 5759–5775. Baik, J.Y., Lee, M.S., An, S.R. et al. (2006). Initial transcriptome and proteome analyses of low culture temperature-induced expression in CHO cells producing erythropoietin. Biotechnol. Bioeng. 93: 361–371. Yee, J.C., de Leon Gatti, M., Philp, R.J. et al. (2008). Genomic and proteomic exploration of CHO and hybridoma cells under sodium butyrate treatment. Biotechnol. Bioeng. 99: 1186–1204. Kantardjieff, A., Jacob, N.M., Yee, J.C. et al. (2010). Transcriptome and proteome analysis of Chinese hamster ovary cells under low temperature and butyrate treatment. J. Biotechnol. 145: 143–159. Kang, S., Ren, D., Xiao, G. et al. (2014). Cell line profiling to improve monoclonal antibody production. Biotechnol. Bioeng. 111: 748–760. Fan, Y., Jimenez Del Val, I., Muller, C. et al. (2015). A multi-pronged investigation into the effect of glucose starvation and culture duration on fed-batch CHO cell culture. Biotechnol. Bioeng. 112: 2172–2184. Lakshmanan, M., Yu, K., Koduru, L., and Lee, D.Y. (2015). In silico modeldriven cofactor engineering strategies for improving the overall NADP(H) turnover in microbial cell factories. J. Ind. Microbiol. Biotechnol. 42: 1401–1414. Ley, D., Seresht, A.K., Engmark, M. et al. (2015). Multi-omic profiling of EPO-producing Chinese hamster ovary cell panel reveals metabolic adaptation to heterologous protein production. Biotechnol. Bioeng. 112: 2373–2387. Orellana, C.A., Marcellin, E., Schulz, B.L. et al. (2015). High-antibodyproducing Chinese hamster ovary cells up-regulate intracellular protein transport and glutathione synthesis. J. Proteome Res. 14: 609–618. Ang, K.S., Kyriakopoulos, S., Li, W., and Lee, D.Y. (2016). Multi-omics data driven analysis establishes reference codon biases for synthetic gene design in microbial and mammalian cells. Methods 102: 26–35. Blondeel, E.J.M., Ho, R., Schulze, S. et al. (2016). An omics approach to rational feed: enhancing growth in CHO cultures with NMR metabolomics and 2D-DIGE proteomics. J. Biotechnol. 234: 127–138. Lewis, A.M., Croughan, W.D., Aranibar, N. et al. (2016). Understanding and controlling sialylation in a CHO Fc-fusion process. PloS One 11: e0157111. Pfizenmaier, J., Junghans, L., Teleki, A., and Takors, R. (2016). Hyperosmotic stimulus study discloses benefits in ATP supply and reveals miRNA/mRNA targets to improve recombinant protein production of CHO cells. Biotechnol. J. 11: 1037–1047. van Wijk, X.M., Dohrmann, S., Hallstrom, B.M. et al. (2017). Whole-genome sequencing of invasion-resistant cells identifies laminin alpha2 as a host factor for bacterial invasion. mBio 8 (1): , e02128-16. Strotbek, M., Florin, L., Koenitzer, J. et al. (2013). Stable microRNA expression enhances therapeutic antibody productivity of Chinese hamster ovary cells. Metab. Eng. 20: 157–166.

183

184

7 Toward Integrated Multi-omics Analysis for Improving CHO Cell Bioprocessing

93 Sarver, A.L., Li, L., and Subramanian, S. (2010). MicroRNA miR-183

94 95 96 97

98

functions as an oncogene by targeting the transcription factor EGR1 and promoting tumor cell migration. Cancer Res. 70: 9570–9580. Palsson, B. and Zengler, K. (2010). The challenges of integrating multi-omic data sets. Nat. Chem. Biol. 6: 787–789. Buescher, J.M. and Driggers, E.M. (2016). Integration of omics: more than the sum of its parts. Cancer Metab. 4: 4. Huang, S., Chaudhary, K., and Garmire, L.X. (2017). More is better: recent progress in multi-omics data integration methods. Front. Genet. 8: 84. Bersanelli, M., Mosca, E., Remondini, D. et al. (2016). Methods for the integration of multi-omics data: mathematical aspects. BMC Bioinf. 17 (Suppl. 2): 15. Cavill, R., Jennen, D., Kleinjans, J., and Briede, J.J. (2016). Transcriptomic and metabolomic data integration. Briefings Bioinf. 17: 891–901.

185

8 CRISPR Toolbox for Mammalian Cell Engineering Daria Sergeeva 1 , Karen Julie la Cour Karottki 1 , Jae Seong Lee 1,2 , and Helene Faustrup Kildegaard 1 1 The Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Building 220, Kemitorvet, 2800 Kgs. Lyngby, Denmark 2 Department of Molecular Science and Technology, Ajou University, 206 Worldcup-ro, Yeongtong-gu, Suwon 16499, Republic of Korea

8.1 Introduction The conversion of bacterial CRISPR/Cas9 (clustered regularly interspaced short palindromic repeats/CRISPR-associated protein 9) immune system into a simple and versatile genome editing tool has revolutionized biological research. CRISPR/Cas9 has been rapidly adapted for a vast range of applications in diverse organisms. Especially, CRISPR technology has transformed the engineering of mammalian cells, providing tools for precise and efficient genome manipulations including gene knockout and knock-in, transcriptional activation and repression, and epigenetic modifications. Nowadays, CRISPR has been implemented in many research groups to study cell biology, establish human disease models, develop new therapeutic methods, and build complex synthetic gene circuits in mammalian cells [1–4]. Moreover, functional studies have been facilitated by genome-scale CRISPR screens, where knockout or transcriptional modulation approaches were used to elucidate gene functions [5]. Advances in CRISPR/Cas9 offer new opportunities in biotechnology toward the development of cell factories producing chemicals and drugs. The biotechnological potential of CRISPR/Cas9 has been demonstrated by metabolic engineering of microbial cell factories (bacterial and yeasts cells) for bio-based production of chemicals and fuels [6]. In the case of mammalian cell factories, the potential of CRISPR/Cas9 can be illustrated by optimizing Chinese hamster ovary (CHO) cells, the most commonly used host cells for production of therapeutic glycoproteins. This chapter will describe recent developments of CRISPR technology for mammalian cell engineering and will discuss how the technology can be applied in CHO cell engineering toward improved production of biopharmaceuticals.

Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

186

8 CRISPR Toolbox for Mammalian Cell Engineering

8.2 Mechanism of CRISPR/Cas9 Genome Editing CRISPR/Cas systems have evolved in prokaryotes as a defense mechanism against foreign genetic elements such as viruses [7]. The most investigated CRISPR/Cas9 system (class 2 type II; described in Section 8.3.1) is composed of three main components: Cas9 endonuclease, CRISPR RNA (crRNA), and trans-activating crRNA (tracrRNA) (Figure 8.1). crRNAs are processed from CRISPR arrays that are clusters of repeat sequences interspaced by variable sequences (spacers) homologous to foreign genetic elements (protospacers). Transcribed crRNA and tracrRNA hybridize to a RNA duplex known as guide RNA (gRNA) that binds to the Cas9 protein and form an active ribonucleoprotein

Cas9

tracrRNA

crRNA

PAM

DSB

NHEJ

No template Indels

HDR

MMEJ

+ Template with homology arms dsDNA

HR

+ Template with microhomology

ssDNA

SSTR

Gene disruption

Gene integration or disruption Precise gene integration or correction

Figure 8.1 CRISPR/Cas9-mediated genome editing. gRNA-directed Cas9 induces a doublestranded break (DSB) in DNA, which can then be repaired by cellular DNA repair mechanisms. Error-prone NHEJ-mediated repair can introduce indels of variable length at the site of the DSB, resulting in gene disruption. HDR-mediated repair can lead to precise gene integration or correction when double-stranded or single-stranded DNA donor template with homology arms is provided. Alternatively, gene integration or disruption can result from MMEJ-mediated repair with an assistance of short homologous sequences in a donor template.

8.3 Variants of CRISPR-RNA-guided Endonucleases

(RNP) complex. With the guidance of the gRNA, this complex searches for complementary sequences in foreign DNA and directs its cleavage by two nuclease domains of Cas9. Each nuclease domain cleaves one strand of DNA; hence, Cas9 activity leads to a double-stranded break (DSB) of DNA [8]. The breakthrough in the field happened, when it was shown that the bacterial CRISPR/Cas9 system could be used for genome editing in eukaryotic cells, especially in mammalian cells [9, 10]. To simplify the system, the crRNA:tracrRNA duplex was fused into a chimeric single guide RNA (sgRNA) [9]. gRNA binds to target DNA through an approximately 20-nucletide (nt) region that is adjacent to the protospacer-adjacent motif (PAM), which is recognized by the PAM-interacting domain of Cas9. Thus, by customizing a 20-nt region of the gRNA to pair with the DNA sequence of interest, Cas9 can be targeted to any genomic locus containing a PAM sequence, making it an easily programmable platform for genome editing. DSBs generated by Cas9 activate the intrinsic cellular DNA repair mechanisms in mammalian cells, such as non-homologous end joining (NHEJ) and homology-directed repair (HDR). These DNA repair pathways predominate at different cell cycle phases and recruit various molecular factors to perform the repair [11]. Mammalian cells may generate random insertion/deletion mutations (indels) at the site of DSB via NHEJ, leading to the potential disruption of genes and functional gene knockout. In contrast, HDR allows precise targeted gene integration, gene replacement, or correction. HDR can be exploited in the presence of a double-stranded DNA (dsDNA) or single-stranded DNA (ssDNA) template with homologous regions spanning the DSB. Depending on the DNA used as a template, homologous recombination (HR) or single-stranded template repair (SSTR) occurs [12]. Besides NHEJ and HDR, an alternative repair mechanism called microhomology-mediated end joining (MMEJ) can occur at the site of DSB. MMEJ requires short homologous sequences (5–20 nt) for DSB repair and harnesses HDR- and NHEJ-independent DNA repair machinery. All these naturally occurring DNA repair pathways are utilized in CRISPR/Cas9-mediated genome editing for different genome engineering purposes.

8.3 Variants of CRISPR-RNA-guided Endonucleases 8.3.1

Diversity of CRISPR/Cas Systems

CRISPR/Cas systems display a wide evolutionary diversity in bacteria and archaea. Based on the differences in their components, CRISPR/Cas systems have been divided into two classes: class 1 systems (types I, III, and IV) that rely on multi-subunit protein complexes and class 2 systems (types II, V, and VI) that utilize single-effector proteins [7]. The widely studied DNA-targeting CRISPR/Cas9 system belongs to class 2 type II and comprises the single-effector protein Cas9, which contains RuvC and HNH nuclease domains. The most commonly used Cas9 has been adapted from Streptococcus pyogenes (SpCas9). The SpCas9-mediated DNA recognition requires a 20-nt target complementary sequence in the crRNA and 5′ -NGG-3′ PAM in the

187

188

8 CRISPR Toolbox for Mammalian Cell Engineering

target DNA. SpCas9 gene has a relatively large size (4.2 kb), which can hamper its efficient delivery into mammalian cells. Smaller Cas9 orthologs with different PAM requirements have been described and developed as genome editing endonucleases, for example, Cas9 from Staphylococcus aureus (SaCas9), which requires a 5′ -NNGRRT-3′ PAM sequence [13]. Another CRISPR system – class 2 type V CRISPR/Cpf1 (also known as Cas12a) expands the options for mammalian genome editing [14]. Cpf1 appears to have increased specificity and facilitates targeting of AT-rich sequences [15, 16]. Containing only one RuvC endonuclease domain, Cpf1 creates a staggered DNA DSB and mediates DNA cleavage by recognition of a short 5′ -TTTN-3′ PAM. In contrast to Cas9s, Cpf1 is guided by a single short crRNA and does not require a tracrRNA. As both Cpf1 and its gRNAs are smaller than SpCas9 counterparts, it overcomes some limitations of CRISPR delivery into mammalian cells. Moreover, Cpf1 has an ability to process its own crRNA, which can be used to simplify multiplexed genome editing [17]. In contrast to type II and type V, recently characterized class 2 type VI effector proteins Cas13 mediate single-stranded RNA (ssRNA) cleavage [18]. Cas13 binding is determined by a crRNA with a 28 nt sequence complementary to RNA protospacer, which should be flanked by A, U, or C. The discovery of RNA-targeted CRISPR systems opens the door for the development of new RNA editing tools for mammalian cells (described in Section 8.5.3). 8.3.2

Engineered Cas9 Variants

Since the first applications of CRISPR/Cas9 as a genome editing tool, engineered Cas9 variants have appeared. One of the first developments was mutating one of the two nuclease domains in Cas9, creating a “nickase” Cas9 (nCas9) [10, 19]. nCas9 cleaves only one strand of DNA leading to nicks, which are predominantly repaired with higher fidelity than DSBs. Inactivation of both nuclease domains of Cas9 results in “dead” Cas9 (dCas9), which lacks nuclease activity but retains RNA-guided DNA-binding ability [19]. Fusions of dCas9 with various effector domains mediate site-specific transcriptional or epigenetic regulation without cleaving target DNA (described in Section 8.5.2). To improve Cas9 specificity and alter PAM recognition, different protein engineering efforts have been performed. Cas9s with enhanced specificity were constructed by mutating residues, forming unspecific bonds with DNA strand [20, 21]. Using structural information and molecular evolution, PAM specificities of SpCas9 and SaCas9 were expanded for broadening the Cas9 targeting range [22, 23]. Furthermore, to allow inducible control of genome modifications, Cas9 was split into two fragments, which can be brought back together upon chemical or light induction [24].

8.4 Experimental Design for CRISPR-mediated Genome Editing Overall, the generation of a mammalian cell line with CRISPR/Cas9-mediated modification, for example, cell lines with gene knockouts, can be achieved within

8.4 Experimental Design for CRISPR-mediated Genome Editing

four weeks [25]. The process includes the selection of target sites, construction of reagents, transfection of Cas9 and gRNA, single-cell isolation, expansion of clones, and analysis of modifications (Figure 8.2). The initial design of experiments for CRISPR-mediated genome editing requires several considerations at the following steps: target site selection, design of gRNA, and choice of delivery methods for CRISPR/Cas9 components. If the goal of experiment is targeted gene integration or gene correction, DNA repair templates in the form of either dsDNA or ssDNA should be designed. After the delivery of Cas9, gRNA, and, if necessary, donor template, cells can be subcloned by fluorescence-activated cell sorting (FACS) or limiting dilution. An advantage of FACS is the possibility of enriching transfected cells, when applying fluorescent markers in CRISPR/Cas9 components, leading to an increase of genome editing events. Modifications introduced upon CRISPR/Cas9-mediated genome editing can be monitored by different assays. In the case of gene knockouts, mismatch-recognizing nucleases (Surveyor nuclease or T7 endonuclease I) or indel detection by amplicon analysis (IDAA) can be applied for the detection of indels [26]. Genomic changes can also be analyzed by DNA sequencing (Sanger sequencing or next-generation sequencing). These methods will confirm the indel length and position and will help to select clones with desired genomic modifications. For verification of gene knockout, it is also recommended to conduct functional confirmation of gene disruption, for example, by qPCR, western blot, and functional assays, to verify the loss of gene function. 8.4.1

Target Site Selection and Design of gRNAs

There are two main considerations in the selection of target site for genome editing: the presence of the PAM sequence in the vicinity of the targeted genomic site and the minimization of potential off-target activity. In the context of large mammalian genomes, SpCas9 may bind and cleave off-target sites with relatively high rates [27]. Therefore, methods to measure and enhance Cas9 specificity and improve computational tools for gRNA design are of great interest. The methods to improve Cas9 specificity include modification of gRNA (shortening of spacers to 17–18 nt or chemical modification of gRNA), discovery of Cas9 orthologs with higher specificity, Cas9 protein engineering, limitation of the time Cas9 is present in the nucleus, and development of better predictive models for the design of gRNA [28]. Several deep sequencing-based protocols, such as GUIDE-seq, Digenome-seq, Circle-seq, and Site-seq, have been applied for genome-wide measuring Cas9 off-target activity in human cell lines [29–32]. They confirmed that sequence homology alone is not fully predictive for Cas9 off-target sites. Thus, more advanced rules for gRNA design need further investigation. Present computational tools for gRNA design can calculate gRNA specificity based on in silico prediction of off-target sites on the basis of sequence homology and in vivo and in vitro assessment of Cas9 specificity [28]. There are many online tools for the design of gRNA, which can be easily used to select the target sequence with the highest predicted editing specificity. These tools differ in the number of genomes and Cas9 orthologs supported, type

189

~1 day

PAM

PAM

PAM

PAM

sgRNA

Cas9-GFP

Target ~4 days

Vector construction

gRNA design

8 CRISPR Toolbox for Mammalian Cell Engineering

U6 sgRNA plasmid

Cas9

CMV

sgRNA scaffold

Cas9 plasmid

GFP

Donor DNA (optional)

sgRNA Transfection ~2 days

Transfection and selection

Cas9

FACS Single cell sort GFP positive

Waste

Transfected single cells Genomic

~30 days

Clonal expansion and verification

190

Expansion

Verification Functional MW WT KO

Cell banking

Figure 8.2 Timeline of CRISPR/Cas9 genome editing experiment. First, gRNA for a specific target is designed using in silico tools. Second, the target sequence is cloned into a sgRNA expression plasmid. Third, sgRNA and Cas9 plasmids and optional DNA repair template are transfected into cells. Then, transfected cells are enriched by FACS, clonally expanded, and verified to derive cell lines with desired modifications.

8.4 Experimental Design for CRISPR-mediated Genome Editing

of gRNA design, and algorithms used to score gRNA specificity. Examples of online tools, which can be applied to the design of sgRNA for CRISPR editing in CHO cells, include CRISPy (http://staff.biosustain.dtu.dk/laeb/crispy/) [33] and Benchling (https://benchling.com/crispr/). As the efficiency of gRNA may vary for different genomic sites depending on many factors, including the chromatin state [34], it is recommended to design at least two gRNAs for each target locus and test their efficiency. 8.4.2

Delivery of CRISPR/Cas9 Components

Depending on the desired genome editing application, there are various ways in which CRISPR/Cas9 components can be introduced into the cell. gRNA and Cas9 can be delivered as DNA plasmids, RNA, or RNP complex. For short-term experiments in cell lines, the preferable method of knocking out or editing a gene by CRISPR/Cas9 is transient transfection, as a constitutive expression of CRISPR components is unwanted once the desired genome modification has occurred. For some applications such as genome-wide CRISPR screens, it is often desirable to make a stable cell line expressing Cas9 and gRNAs using viral transduction combined with selection to ensure genome editing in all cells and even dispersal of the many different gRNAs. There are many mammalian expression vectors available that encode Cas9 variants and gRNA (see https://www.addgene.org). For gRNA expression, RNA polymerase III U6 promoter is typically used. The promoter driving the Cas9 expression in mammalian cells can be constitutive (e.g. cytomegalovirus [CMV] or elongation factor 1 alpha [EF1-α] promoters) or inducible. The plasmid may also contain a reporter gene (e.g. GFP) or selection marker (e.g. antibiotic resistance gene) to facilitate screening, enrichment, and selection of transfected cells as mentioned above. The plasmid-based delivery of CRISPR/Cas9 has been widely used because of distinct benefits of this system, such as low-cost and ready-to-use transfection methods. However, there are numerous drawbacks to plasmid delivery, for example, potential risk of random integration of plasmid DNA into a host genome and high off-target cleavage. Although prolonged expression of CRISPR components increase on-target editing events over time, this also increases off-target editing. To improve the efficiency and specificity of CRISPR/Cas9, it was proposed to deliver preassembled gRNA:Cas9 RNPs directly into the cell [35]. To form RNP, purified Cas9 is assembled together with synthetic gRNA (sgRNA or crRNA:trRNA complex) and then RNP is delivered into the cell via electroporation or lipid-mediated transfection. RNP cleaves DNA within several hours after the delivery and is rapidly degraded in the cell. Such fast action and a short duration of the Cas9 presence in the nucleus increase the efficiency and reduce the off-target effects [36, 37]. Certain cell types, including primary cells and stem cells, can be difficult to transfect via the nonviral methods described above. Also, for CRISPR screening (described in Section 8.6), it is important to obtain an even dispersal of gRNA integration over a pool of cells. In these cases, it is advantageous to carry out viral delivery, in which CRISPR components are encapsulated by a viral vector

191

192

8 CRISPR Toolbox for Mammalian Cell Engineering

and enters the cells through transduction [38]. The most extensively studied viral vectors include retroviruses, adenoviruses, and adeno-associated viruses (AAVs). Some common drawbacks of using viral delivery include safety concerns, limitations of insert size, and less ease-of-use compared with plasmid-based transfection. Over the years, however, improved vector construction, viral particle production, transduction efficiency, and general safety have refined the system [39, 40].

8.5 Development of CRISPR/Cas9 Tools Since the first publications of CRISPR/Cas9 application in mammalian cells, diverse CRISPR-based tools have been developed for gene editing, gene regulation, epigenetic modification, genome imaging, and CRISPR-based chromatin immunoprecipitation. Besides DNA targeting, CRISPR can be applied for RNA knockdown and modification (Figure 8.3). The following sections will cover recent advances in CRISPR/Cas tool development and their applications in mammalian cells. 8.5.1 8.5.1.1

CRISPR/Cas9-mediated Gene Editing Gene Knockout

Gene disruption was the first application of CRISPR/Cas9 in mammalian cells and remains the most exploitable method for functional knockout of the target genes. The easiest way to create a knockout cell line is to introduce two components into the cell: gRNA specific to the target sequence and Cas9. DSBs induced by CRISPR/Cas9 are preferentially repaired by error-prone NHEJ, leading to indels. These mutations can cause a frame shift in the coding regions of genes that disrupts their proper translation and results in a functional knockout. When using Cas9 for gene knockout through the creation of indels, it is most common to target an early exon in the coding sequence or functional domain to disrupt as much of the protein as possible. Aside from NHEJ repair of CRISPR/Cas9-mediated DSBs, HDR can be applied to generate gene knockouts by introducing a premature stop codon. Codelivery of a DNA template bearing a stop codon together with CRISPR/Cas9 components can lead to increased knockout efficiency [43]. The introduction of multiple gRNAs along with a common Cas9 protein can lead to simultaneous modification of multiple target sequences located at the same or different genes, referred to as multiplexing. Multiplex genome engineering via CRISPR/Cas9 can be achieved by co-transfection of multiple gRNA plasmids, a single vector with gRNA arrays, or multiple RNP complexes [9, 10, 36]. Also, a multiplexing method using Cpf1 has been published, utilizing the special feature of Cpf1 to process its own crRNA array [17]. Simultaneous introduction of DSBs in different genomic loci can give rise to different types of genome modifications: multiallelic and multigene modifications, large deletions, or chromosomal rearrangements. Targeting multiple genes by a multiplexing strategy

8.5 Development of CRISPR/Cas9 Tools

Gene editing

Gene regulation

Gene knockout

Transcription repression Transcriptional machinery

Cas9

Repressor

dCas9

TSS

Gene knock-in

Transcription activation Activator dCas9

Cas9

TSS

+ Donor DNA

(a)

Transcriptional machinery

(b) Epigenetic modification

Genome imaging Fluorescent protein dCas9

DNA methylation Methyltransferase

dCas9

me me

(d) Chromatin immunoprecipitation Ab

Demethylase dCas9

me me

Affinity tag dCas9 Histone modification Acetyltransferase dCas9

(c)

(e) RNA knockdown

(f)

DNA-binding proteins

DNA

ac ac

Cas13

RNA binding and/or modification

dCas13

Figure 8.3 CRISPR toolbox for mammalian cell engineering. (a) CRISPR-mediated gene editing is based on Cas9 nuclease activity, where repair of Cas9-introduced DSBs can lead to gene knockout or knock-in. (b) Fusion of catalytically inactive dCas9 to transcriptional activation or repression domains can mediate transcriptional regulation by alteration of transcriptional machinery binding to transcription start site (TSS). (c) Fusion of dCas9 to DNA or chromatin modification domains enable epigenetic changes of target sites. (d) dCas9 can be fused to fluorescent domains for imaging of genomic loci [41]. (e) An affinity-tagged dCas9 can be used in chromatin immunoprecipitation assays to study protein interactions at specific genomic site [42]. (f ) CRISPR/Cas13 can be used for RNA knockdown, binding, and modification in mammalian cells.

193

194

8 CRISPR Toolbox for Mammalian Cell Engineering

accelerates development of cell lines with complex genome edits, for example, multiple knockouts. Up to seven genomic loci were simultaneously modified in human cell line by using the all-in-one plasmid with gRNA expression array and Cas9 [44]. Besides multigene knockout, concurrent delivery of gRNAs designed to target two different genomic sites can result in targeted deletion of genomic segments ranging from several hundred base pairs to 1 Mbp [45]. This paired gRNAs strategy can be used for the purpose of gene knockout, when several exons in the gene are removed [46]. It was also shown that DNA cleavage by Cas9 at two genomic loci can result in chromosomal rearrangement and expression of fusion transcripts [47]. As was mentioned earlier, wild-type Cas9 is known to recognize and cleave many off-target sites. To minimize off-target mutagenesis, a double-nicking strategy can be used to introduce DSBs and disrupt the gene by nCas9 and a pair of offset gRNAs [48, 49]. Single-stranded nicks on the opposite strands of the target DNA lead to a production of composite DSBs, which are then repaired by NHEJ and result in indel formation or large genomic deletions. Because of the doubled number of bases that need to be specifically recognized by two nCas9s, this strategy increases the specificity of genome editing. 8.5.1.2

Site-Specific Gene Integration

CRISPR/Cas9 has facilitated site-specific integration of DNA by taking advantage of native DNA repair mechanisms [50]. HDR of CRISPR/Cas9-introduced DSBs near the targeted genomic loci can lead to the introduction of desired mutations or insertion of genes when the proper DNA template is presented. Single-nucleotide substitutions or modification of short sequences (e.g. insertion of tags or codon mutations) can be introduced using short oligonucleotide templates, called single-stranded oligodeoxynucleotides (ssODNs) [25]. The design of ssODNs requires addition of homology arms, whose length can vary from short (30–40 nt) to long (70–90 nt) [51]. Although ssODN-mediated integration was shown to be more efficient than double-stranded donor integration, ssODNs have a narrow insert length because of the limitation of single-stranded oligo synthesis. Alternatively, single-stranded DNA template can be provided in the form of recombinant AAV vectors, but their ssDNA length is also limited to 4.5 kb [52]. Therefore, for the integration of large DNA cassettes, dsDNA vectors with long homology arms (around 1 kb each) are used as donor template. Co-delivery of donor plasmids with CRISPR/Cas9 components was applied for the site-specific integration of transgenes encoding fluorescent markers, antibiotic resistance genes, recombinant proteins, and landing pads for recombinase-mediated cassette exchange [9, 53, 54]. The drawback of HDR-mediated strategy is that HDR appears at low frequency in many mammalian cell lines, leading to a low rate of integration events. The efficiency of HDR-mediated genome editing can be increased by chemical treatment, transient modulation of DNA repair proteins, or cell cycle synchronization [55, 56]. These approaches resulted in significant improvements in total integration efficiency but were shown to be cell-type specific (and context dependent) [57]. Other targeted knock-in strategies rely on homology-independent integration. An advantage of this strategy is the utilization of more frequent DNA repair

8.5 Development of CRISPR/Cas9 Tools

mechanisms such as NHEJ and simple construction of donor vectors because there is no requirement for long homology arms. Although NHEJ repair is error prone, it appears at a higher rate than HDR, which can be beneficial when increased targeted integration efficiency is needed. It was shown that CRISPR/Cas9-induced NHEJ can mediate site-specific knock-in more efficiently than HDR-based strategy [58]. The same trend was demonstrated by homology-independent targeted integration (HITI) method that relies on the simultaneous introduction of DSBs at genomic loci of interest and a donor vector by the gRNA [59]. MMEJ-mediated repair can also facilitate knock-in through an introduction of a donor with microhomology sequences, which was proved by PITCh method [60]. More sophisticated genome engineering can be done by multiplexed targeted integration of genes at multiple alleles and/or loci. As a proof of concept, multigenic homology-directed targeted integration of transgenes at different loci was performed in human hematopoietic stem and progenitor cells by using Cas9 RNP and AAVs [61]. In the future, this approach can be applied for the introduction of genes of multi-subunit protein complexes, building multigene pathways, and revealing functional gene networks in mammalian cells. 8.5.2

CRISPR/Cas9-mediated Genome Modification

With CRISPR gaining foothold as a powerful genome editing method, many have started repurposing the system for RNA-guided site-specific DNA modifications rather than cleavage. The catalytically inactive version of Cas9 (dCas9) combined with different effector molecules has been used for various applications. The following section will focus on the development of CRISPR/dCas9 technology for downregulation and upregulation of genes in mammalian cells via transcriptional and epigenetic modifications. 8.5.2.1

Transcriptional Regulation

The first mention of CRISPR being used as a transcriptional regulator was in a study by Qi, L. et al. in 2013. They engineered a catalytically inactive Cas9 and observed that, with coexpression of gRNA, it could sterically hinder the RNA polymerase from binding and elongating, leading to considerable repression of specific transcription in bacteria, a tool they coined CRISPR interference (CRISPRi) [62]. The effect in eukaryotic cells, however, was moderate, likely because of the complexity of transcriptional regulation. Follow-up studies have shown that fusing dCas9 with repressing regulatory domains, for example, Krüppel-associated box (KRAB), could yield increased transcriptional repression. RNA-sequencing analysis confirmed that CRISPRi knockdown is highly specific. Thus, dCas9-KRAB can be applied for efficient and targeted repression of multiple endogenous genes in mammalian cells [63]. In line with CRISPRi, it has been explored whether fusing dCas9 with activating regulatory elements could increase the expression of targeted genes, referred to as CRISPR activation (CRISPRa). Initial experiments have shown that fusing well-known transcription activators, such as VP16 and the p65 activation domain to dCas9 and coexpressing with gRNA, increases the expression of targeted genes

195

196

8 CRISPR Toolbox for Mammalian Cell Engineering

[63]. Several different combinations of gRNAs and activation effectors have since been tested across various cell types in various settings. In 2016 Chavez, A. et al. compared the effectiveness of the so-called second-generation activators including VP64, VPR (VP64-p65-Rta), SAM (synergetic activation mediator), and SunTag in human, mouse, and fly cell lines. The study was the first comprehensive side-by-side evaluation of their relative potential and revealed that SAM, SunTag, and VPR were consistently superior at increasing expression across cell lines [64]. Recently, dCpf1-based transcriptional activators were developed, which allows synergistic tuning of expression of endogenous genes, leveraging the multiplex capability of Cpf1 [65]. CRISPRi and CRISPRa have since been applied to and exerted regulatory effects in cells from a variety of organisms with varied success rates. For better reproducibility of CRISPR-based regulation of different genes, evidence suggests that careful design of the gRNA in relation to the transcription start site (TSS) as well as nucleosome occupancy of the target site is of great importance [34, 64, 66]. 8.5.2.2

Epigenetic Modification

The role of epigenetic modifications in transcription, genome stability, and nuclear organization has solidified over the last decades. Even though numerous discoveries, such as involvement of histone modification, chromatin remodeling, and DNA methylation, have surfaced, technological limitations have obstructed insights into the precise mechanistic level [67–69]. Fusing dCas9 with epigenetic modifiers presents a promising tool for future epigenetic research. Addition of methylation modulators to dCas9 and targeting it to an unmethylated or methylated promoter region can result in repression or activation of the targeted gene, allowing for functional studies of epigenetic regulation [70]. In the same manner, fusing dCas9 with an acetyltransferase can catalyze acetylation of histones transforming chromatin to a more relaxed state and in response initiate transcriptional activation [71]. 8.5.3

RNA Targeting

Although current CRISPR/Cas systems have been applied mostly for DNA editing, the recent discovery of RNA-targeting CRISPR/Cas13 systems provides a starting point for expanding the CRISPR toolbox for RNA manipulation. Upon binding to a complementary RNA target, Cas13 engages RNase activity both in vitro and in vivo [18]. This RNA cleavage activity of Cas13 was used for the development of in vitro methods for nucleic acid detection and specific RNA knockdown in bacterial and mammalian cells [72–74]. Thus, Cas13 provides the basis for improved tools for controlling cellular processes at the transcript level in mammalian cells. CRISPR/Cas13 system opens new possibilities in the study of RNA function by targeted RNA binding and modification. Substitutions in the catalytic domain of Cas13 converted it into inactive programmable RNA-binding protein dCas13 [18]. dCas13 could be fused to effector domains with different functions to edit RNA sequence, enhance or inhibit translation, alter the splicing or visualize RNA trafficking, and localization [73–75].

8.7 Applications of CRISPR/Cas9 for CHO Cell Engineering

8.6 Genome-Scale CRISPR Screening In addition to the aforementioned applications of CRISPR/Cas9 for gene editing and regulation, the technique and said tools can be extended to large forward genetic screens that can be used not only to interrogate gene function but also to identify novel targets for treatment of disease or engineering industrial mammalian cells. In contrast to reverse genetic engineering, where known genes are modulated and the resulting phenotype is studied, the approach taken in forward genetic screening includes changing numerous genes at a pool level, applying a phenotypic selection and subsequently identifying the responsible genes. Traditionally, these screens have relied on random mutagenesis and isolating individuals with an interesting phenotype [76–78]. As the mutation is initially unknown, identification of causality is a very lengthy and difficult process and represents one of the main weaknesses of this system. With the advancement of RNA interference (RNAi) came a welcomed alternative approach to forward genetic screening. RNAi reagents degrading known sequences of mRNA replaced random mutagenesis, and the identification of causal mutations was hugely simplified [79]. However, the incomplete knockdown of targeted genes and substantial off-target effects have restrained the extent to which RNAi screens can be used. The advancement of CRISPR technologies presents a novel approach to attempt to solve these issues and increase the versatility for the next generation of forward screening methods. Consequently, the first records of CRISPR used as a screening tool in mammalian cells were published in 2014 by Shalem et al. [80] and Wang et al. [81]. CRISPR tools, such as CRISPRi and CRISPRa, have later been included in the screening setup, with activation/gain-of-function screens showing highest potential [5, 82, 83].

8.7 Applications of CRISPR/Cas9 for CHO Cell Engineering The major objectives of mammalian cell engineering toward industrial production of therapeutic proteins are high productivity and product quality. To overcome cellular limitations in growth, productivity, and post-translational modifications (PTMs), various strategies have been applied in CHO cell engineering, such as overexpression, knock-in, knockout, or post-transcriptional silencing of genes [84]. CRISPR/Cas9 offers a new engineering tool for a broad field of facile-targeted genome engineering of CHO cells, described in detail in the following section. In terms of programmable genome engineering, CRISPR/Cas9 was not the first tool to be applied. Before the era of CRISPR/Cas9, zinc finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs) have been developed and used to generate DSBs and engineer mammalian cells. There were successful examples of CHO genome engineering by ZFN and TALENs for knockout of genes involved in glycosylation and metabolism as well as targeted gene integration [85–91]. In a short time, more simple and efficient CRISPR/Cas9

197

198

8 CRISPR Toolbox for Mammalian Cell Engineering

system has emerged as a fascinating alternative, which paved the way for accelerated genome editing for improved recombinant protein production together with advances in CHO genome sequencing and annotation [92]. The first demonstration of CRISPR/Cas9 activity in CHO cells was published in 2014 when two genes involved in glycosylation (C1GALT1-specific chaperone 1 [COSMC] and α-1,6-fucosyltransferase [FUT8]) were disrupted with the indels frequency up to 47% in a pool of cells [33]. Later, the enrichment of cells carrying GFP-Cas9 plasmid and multiplexing knockout strategies were developed [93]. Simultaneous introduction of three gRNA plasmids and fluorescent enrichment yielded 59% of clones harboring indels in FUT8, BAX, and BAK genes. Generated knockout cell lines showed improved resistance to apoptosis and disrupted fucosylation activity, relevant for prolonged cell cultivation and production of nonfucosylated therapeutic proteins. More recently, a novel CRISPR/Cpf1 system was used for multiplexed gene knockout in CHO cells, where Cpf1 showed parallel and comparable efficiency to Cas9 efficiency of genome cleavage [94]. In this study, the multiplexed knockout strategy was based on a pair of gRNAs, expressed as a crRNA array. Introduction of two DSBs by paired gRNAs enables full deletion of genes and regulatory regions and has an advantage in predictable loss of function in contrast to indel formation strategy. As the successful application of CRISPR-mediated gene editing, various CHO engineering strategies have been developed in an attempt to improve product quality and to expedite cell line development for high productivity. A characteristic example of such genome engineering efforts is glycoengineering. Desired glycosylation profiles of therapeutic proteins are the crucial property of product quality attributes for increased stability and efficacy of proteins. CRISPR-mediated knockout of FUT8 allows production of monoclonal antibodies lacking the core fucose on N-glycans, which induce stronger antibody-dependent cell-mediated cytotoxicity (ADCC) [93, 95, 96]. To produce α-2,6-sialylated antibodies with superior effector functions, two α-2,3-sialyltransferases ST3GAL4 and ST3GAL6 were consecutively disrupted by CRISPR/Cas9 and α-2,6-sialyltransferase ST6GAL1 was overexpressed in CHO cells that lack the expression of α-2,6-sialyltransferase but only express α-2,3-sialyltransferases [97]. As the glycosylation pathway and genes involved in this PTM are highly characterized, it makes it possible to predict optimal CRISPR/Cas9 engineering strategies to obtain a specific glycosylation profile with desired properties [98]. Further advanced CRISPR/Cas9 glycoengineering can provide cell lines with homogeneous glycosylation of therapeutic proteins, resolving the challenge of product heterogeneity during glycoprotein production. Besides different glycan structures, heterogeneity of monoclonal antibodies is also caused by different C-terminal lysine levels. C-terminal lysine on antibody heavy chains is cleaved by carboxypeptidase D (CpD). To maintain consistent C-terminal lysine levels, CpD was knocked out by paired-gRNA-mediated deletion of CpD exons, leading to the production of homogeneous proteins [46]. Impurities, such as host cell proteins (HCPs), present another challenge in the production of biopharmaceuticals as they can cause an unpredictable immunogenic response and impair product quality and stability. Variable HCPs

8.8 Conclusion

secreted from viable cells and released from dead cells were identified, which represent appropriate targets to be removed from cell culture and purification process by CRISPR/Cas9 technology [99]. As an example, lipoprotein lipase (LPL) that has the ability to degrade antibody formulations was disrupted by CRISPR/Cas9 [100]. CRISPR/Cas9 engineering has expanded its potential for efficient cell line development. To engineer CHO cells for the rapid adaptation to a suspension culture, RNA sequencing was exploited to identify genes differentially expressed during the adaptation process. Then, two of identified downregulated genes, Igfbp4 and AqpI, were disrupted by CRISPR/Cas9 RNPs, leading to reduced adaptation time [101]. CRISPR/Cas9 knockout of frequently used metabolic selection marker, glutamine synthetase (GS), was reported in CHO cells to facilitate CHO cell line development [102]. Moreover, CRISPRi was recently applied in CHO cells to enhance coamplification of another essential selection marker, dihydrofolate reductase (DHFR), and a gene of interest (GOI). Transcriptional repression of dhfr by dCas9-KRAB resulted in selecting clones with increased productivity [103]. In addition to such cell engineering efforts, targeted gene integration provides a new opportunity for cell line development. Conventional methods of CHO cell line development for production of biopharmaceuticals are based on a random integration of transgenes. These methods are time-consuming and labor-intensive, which yields cell lines exhibiting a wide range of expression, growth, and stability characteristics. Targeted integration of the transgene into high transcriptional active sites in the genome (“hot spots”) would be an ideal solution for acceleration of cell line development for the production of recombinant proteins. The first application of CRISPR-mediated targeted integration of transgenes was demonstrated in CHO cells that employed HDR for insertion of large gene cassettes encoding recombinant proteins [104]. The efficiency of targeted integration after the drug or lectin enrichment varied between 7% and 28% depending on the target locus. Fluorescent enrichment of genome-edited cells was further applied to avoid usage of lectin and antibiotic selection [53]. It resulted in a threefold increase of cells with HDR-mediated integration relative to nonenriched samples, with ∼7% frequency of successful integration. The same HDR-mediated approach was also used to integrate the Bxb1 and Flp/FRT recombinase target sites flanking fluorescent marker and thymidine kinase (“landing pads”) into a defined locus in CHO cells with 27% efficiency [54]. These cell lines were subsequently used for recombinase-mediated cassette exchange with antibody-encoding donor plasmids for streamlining the antibody development process. The NHEJ-mediated targeted integration in CHO was also reported, although its efficiency was considerably low (0.45%) [105].

8.8 Conclusion CRISPR has proven itself a highly valuable RNA-guided genome engineering technique. Despite the great potential already demonstrated, there is still room

199

200

8 CRISPR Toolbox for Mammalian Cell Engineering

for improvements in the current CRISPR system. The issue of off-target effects is one of the major concerns, especially in fields such as gene therapy, where random disruptions are non-negotiable. In line with this, the effect of CRISPR on chromosomal rearrangements and genome stability has yet to be uncovered. Furthermore, some aspects of the technique still have to be improved; for example, the knock-in efficiency as well as insert size in targeted integration. As discussed in this chapter, some CRISPR issues regarding off-target effects, endonuclease size, and type of target nucleic acid seems to have been alleviated with the development of next-generation RNA-guided endonucleases such as Cpf1 and Cas13. Further mining for novel CRISPR systems in bacterial and archaeal genomes will aid in establishing a catalog of nucleases and CRISPR systems that can potentially accommodate all applications. Beyond technical issues of the CRISPR technology, the identification of target sites for genome engineering is another challenge, particularly in CHO cells. This issue can be resolved by improved annotation of CHO genome and revealing of epigenetic landscape, which will help to refine the prediction of gRNA target sites. Understanding of complex cellular interactions by analysis of omics data sets and their integration with the genome-scale model of CHO cell metabolism and the secretory network can provide novel targets for the rational engineering of CHO cells. Together with a better understanding of CHO biology, CRISPR/Cas9 technology has the promise to resolve current bottlenecks in biopharmaceutical production, such as protein folding, secretion, and PTMs, and create cell lines with superior capacity for therapeutic protein production. Overall, advanced cell line engineering using genome engineering tools will help to generate efficient mammalian cell factories, providing patients with new therapeutics of high quality.

Acknowledgment The Novo Nordisk Foundation (NNF10CC1016517 and NNF16CC0020908) kindly supported this work. The authors declare no conflict of interest.

References 1 Wang, F. and Qi, L.S. (2016). Applications of CRISPR genome engineering in

cell biology. Trends Cell Biol. 26: 875–888. 2 Xiong, X., Chen, M., Lim, W.A. et al. (2016). CRISPR/Cas9 for human

genome engineering and disease research. Annu. Rev. Genomics Hum. Genet. 17: 131–154. 3 Jusiak, B., Cleto, S., Perez-Piñera, P., and Lu, T.K. (2016). Engineering synthetic gene circuits in living cells with CRISPR technology. Trends Biotechnol. 34: 535–547. 4 Porteus, M. (2016). Genome editing: a new approach to human therapeutics. Annu. Rev. Pharmacol. Toxicol. 56: 163–190. 5 Shalem, O., Sanjana, N.E., and Zhang, F. (2015). High-throughput functional genomics using CRISPR-Cas9. Nat. Rev. Genet. 16: 299–311.

References

6 Jakoˇci¯ unas, T., Jensen, M.K., and Keasling, J.D. (2016). CRISPR/Cas9

advances engineering of microbial cell factories. Metab. Eng. 34: 44–59. 7 Wright, A.V., Nuñez, J.K., and Doudna, J.A. (2016). Biology and applications

8 9 10 11

12

13 14

15 16

17

18

19

20

21

22 23

of CRISPR systems: harnessing nature’s toolbox for genome engineering. Cell 164: 29–44. Wang, H., La Russa, M., and Qi, L.S. (2016). CRISPR/Cas9 in genome editing and beyond. Annu. Rev. Biochem. 85: 227–264. Mali, P., Yang, L., Esvelt, K.M. et al. (2013). RNA-guided human genome engineering via Cas9. Science 339: 823–826. Cong, L., Ran, F.A., Cox, D. et al. (2013). Multiplex genome engineering using CRISPR/Cas systems. Science 339: 819–823. Ceccaldi, R., Rondinelli, B., and D’Andrea, A.D. (2016). Repair pathway choices and consequences at the double-strand break. Trends Cell Biol. 26: 52–64. Bhargava, R., Onyango, D.O., and Stark, J.M. (2016). Regulation of singlestrand annealing and its role in genome maintenance. Trends Genet. 32: 566–575. Ran, F.A., Cong, L., Yan, W.X. et al. (2015). In vivo genome editing using Staphylococcus aureus Cas9. Nature 520: 186–191. Zetsche, B., Gootenberg, J.S., Abudayyeh, O.O. et al. (2015). Cpf1 is a single RNA-guided endonuclease of a class 2 CRISPR-Cas system. Cell 163: 759–771. Kim, D., Kim, J., Hur, J.K. et al. (2016). Genome-wide analysis reveals specificities of Cpf1 endonucleases in human cells. Nat. Biotechnol. 34: 863–868. Kleinstiver, B.P., Tsai, S.Q., Prew, M.S. et al. (2016). Genome-wide specificities of CRISPR-Cas Cpf1 nucleases in human cells. Nat. Biotechnol. 34: 869–874. Zetsche, B., Heidenreich, M., Mohanraju, P. et al. (2017). Multiplex gene editing by CRISPR-Cpf1 using a single crRNA array. Nat. Biotechnol. 35: 31–34. Abudayyeh, O.O., Gootenberg, J.S., Konermann, S. et al. (2016). C2c2 is a single-component programmable RNA-guided RNA-targeting CRISPR effector. Science 353: aaf5573. Jinek, M., Chylinski, K., Fonfara, I. et al. (2012). A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity. Science 337: 816–821. Kleinstiver, B.P., Pattanayak, V., Prew, M.S. et al. (2016). High-fidelity CRISPR-Cas9 nucleases with no detectable genome-wide off-target effects. Nature 529: 490–495. Slaymaker, I.M., Gao, L., Zetsche, B. et al. (2015). Rationally engineered Cas9 nucleases with improved specificity. Science https://doi.org/10.1126/ science.aad5227. Kleinstiver, B.P., Prew, M.S., Tsai, S.Q. et al. (2015). Engineered CRISPR-Cas9 nucleases with altered PAM specificities. Nature 523: 481–485. Kleinstiver, B.P., Prew, M.S., Tsai, S.Q. et al. (2015). Broadening the targeting range of Staphylococcus aureus CRISPR-Cas9 by modifying PAM recognition. Nat. Biotechnol. 33: 1293–1298.

201

202

8 CRISPR Toolbox for Mammalian Cell Engineering

24 Nuñez, J.K., Harrington, L.B., and Doudna, J.A. (2016). Chemical and bio-

25 26

27

28 29

30 31

32 33

34 35

36

37

38 39 40 41

physical modulation of Cas9 for tunable genome engineering. ACS Chem. Biol. 11: 681–688. Ran, F.A., Hsu, P.D., Wright, J. et al. (2013). Genome engineering using the CRISPR-Cas9 system. Nat. Protoc. 8: 2281–2308. Lonowski, L.A., Narimatsu, Y., Riaz, A. et al. (2017). Genome editing using FACS enrichment of nuclease-expressing cells and indel detection by amplicon analysis. Nat. Protoc. 12: 581–603. Fu, Y., Foden, J.A., Khayter, C. et al. (2013). High-frequency off-target mutagenesis induced by CRISPR-Cas nucleases in human cells. Nat. Biotechnol. 31: 822–826. Tycko, J., Myer, V.E., and Hsu, P.D. (2016). Methods for optimizing CRISPR-Cas9 genome editing specificity. Mol. Cell 63: 355–370. Tsai, S.Q., Zheng, Z., Nguyen, N.T. et al. (2015). GUIDE-seq enables genome-wide profiling of off-target cleavage by CRISPR-Cas nucleases. Nat. Biotechnol. 33: 187–197. Kim, D., Bae, S., Park, J. et al. (2015). Digenome-seq: genome-wide profiling of CRISPR-Cas9 off-target effects in human cells. Nat. Methods 12: 237–243. Tsai, S.Q., Nguyen, N.T., Malagon-Lopez, J. et al. (2017). CIRCLE-seq: a highly sensitive in vitro screen for genome-wide CRISPR-Cas9 nuclease off-targets. Nat. Methods 14: 607–614. Cameron, P., Fuller, C.K., Donohoue, P.D. et al. (2017). Mapping the genomic landscape of CRISPR-Cas9 cleavage. Nat. Methods 14: 600–606. Ronda, C., Pedersen, L.E., Hansen, H.G. et al. (2014). Accelerating genome editing in CHO cells using CRISPR Cas9 and CRISPy, a web-based target finding tool. Biotechnol. Bioeng. 111: 1604–1616. Horlbeck, M.A., Witkowsky, L.B., Guglielmi, B. et al. (2016). Nucleosomes impede Cas9 access to DNA in vivo and in vitro. eLife 5: e12677. Kim, S., Kim, D., Cho, S.W. et al. (2014). Highly efficient RNA-guided genome editing in human cells via delivery of purified Cas9 ribonucleoproteins. Genome Res. 24: 1012–1019. Liang, X., Potter, J., Kumar, S. et al. (2015). Rapid and highly efficient mammalian cell engineering via Cas9 protein transfection. J. Biotechnol. 208: 44–53. Liang, X., Potter, J., Kumar, S. et al. (2017). Enhanced CRISPR/Cas9mediated precise genome editing by improved design and delivery of gRNA, Cas9 nuclease, and donor DNA. J. Biotechnol. 241: 136–146. Cai, Y. and Mikkelsen, J.G. (2016). Lentiviral delivery of proteins for genome engineering. Curr. Gene Ther. 16: 194–206. Nayerossadat, N., Maedeh, T., and Ali, P.A. (2012). Viral and nonviral delivery systems for gene delivery. Adv. Biomed. Res. 1: 27. Yin, H., Kauffman, K.J., and Anderson, D.G. (2017). Delivery technologies for genome editing. Nat. Rev. Drug Discovery 16: 387–399. Chen, B., Gilbert, L.A., Cimini, B.A. et al. (2013). Dynamic imaging of genomic loci in living human cells by an optimized CRISPR/Cas system. Cell 155: 1479–1491.

References

42 Fujita, T. and Fujii, H. (2013). Efficient isolation of specific genomic regions

43 44

45

46

47 48

49

50 51

52 53

54

55

56

57

and identification of associated proteins by engineered DNA-binding molecule-mediated chromatin immunoprecipitation (enChIP) using CRISPR. Biochem. Biophys. Res. Commun. 439: 132–136. Li, K., Wang, G., Andersen, T. et al. (2014). Optimization of genome engineering approaches with the CRISPR/Cas9 system. PLoS One 9: e105779. Sakuma, T., Nishikawa, A., Kume, S. et al. (2014). Multiplex genome engineering in human cells using all-in-one CRISPR/Cas9 vector system. Sci. Rep. 4: https://doi.org/10.1038/srep05400. He, Z., Proudfoot, C., Mileham, A.J. et al. (2015). Highly efficient targeted chromosome deletions using CRISPR/Cas9. Biotechnol. Bioeng. 112: 1060–1064. Hu, Z., Zhang, H., Haley, B. et al. (2016). Carboxypeptidase D is the only enzyme responsible for antibody C-terminal lysine cleavage in Chinese hamster ovary (CHO) cells. Biotechnol. Bioeng. 113: 2100–2106. Choi, P.S. and Meyerson, M. (2014). Targeted genomic rearrangements using CRISPR/Cas technology. Nat. Commun. 5: 3728. Mali, P., Aach, J., Stranges, P.B. et al. (2013). CAS9 transcriptional activators for target specificity screening and paired nickases for cooperative genome engineering. Nat. Biotechnol. 31: 833–838. Ran, F.A., Hsu, P.D., Lin, C.-Y. et al. (2013). Double nicking by RNA-guided CRISPR Cas9 for enhanced genome editing specificity. Cell 154: 1380–1389. Liu, Z., Liang, Y., Ang, E.L., and Zhao, H. (2017). A new era of genome integration-simply cut and paste! ACS Synth. Biol. 6: 601–609. Richardson, C.D., Ray, G.J., DeWitt, M.A. et al. (2016). Enhancing homology-directed genome editing by catalytically active and inactive CRISPR-Cas9 using asymmetric donor DNA. Nat. Biotechnol. 34: 339–344. Kaulich, M. and Dowdy, S.F. (2015). Combining CRISPR/Cas9 and rAAV templates for efficient gene editing. Nucleic Acid Ther. 25: 287–296. Lee, J.S., Grav, L.M., Pedersen, L.E. et al. (2016). Accelerated homologydirected targeted integration of transgenes in Chinese hamster ovary cells via CRISPR/Cas9 and fluorescent enrichment. Biotechnol. Bioeng. 113: 2518–2523. Inniss, M.C., Bandara, K., Jusiak, B. et al. (2017). A novel Bxb1 integrase RMCE system for high fidelity site-specific integration of mAb expression cassette in CHO cells. Biotechnol. Bioeng. 114: 1837–1846. Chu, V.T., Weber, T., Wefers, B. et al. (2015). Increasing the efficiency of homology-directed repair for CRISPR-Cas9-induced precise gene editing in mammalian cells. Nat. Biotechnol. 33: 543–548. Lin, S., Staahl, B.T., Alla, R.K., and Doudna, J.A. Enhanced homologydirected human genome engineering by controlled timing of CRISPR/Cas9 delivery. eLife 3: e04766. Song, J., Yang, D., Xu, J. et al. (2016). RS-1 enhances CRISPR/Cas9- and TALEN-mediated knock-in efficiency. Nat. Commun. 7: https://doi.org/10 .1038/ncomms10548.

203

204

8 CRISPR Toolbox for Mammalian Cell Engineering

58 He, X., Tan, C., Wang, F. et al. (2016). Knock-in of large reporter genes in

59

60

61

62

63

64 65

66

67 68 69 70 71

72 73 74 75 76

human cells via CRISPR/Cas9-induced homology-dependent and independent DNA repair. Nucleic Acids Res. 44: e85. Suzuki, K., Tsunekawa, Y., Hernandez-Benitez, R. et al. (2016). In vivo genome editing via CRISPR/Cas9 mediated homology-independent targeted integration. Nature 540: 144–149. Sakuma, T., Nakade, S., Sakane, Y. et al. (2016). MMEJ-assisted gene knock-in using TALENs and CRISPR-Cas9 with the PITCh systems. Nat. Protoc. 11: 118–133. Bak, R.O., Dever, D.P., Reinisch, A. et al. (2017). Multiplexed genetic engineering of human hematopoietic stem and progenitor cells using CRISPR/Cas9 and AAV6. eLife 6: e27873. Qi, L.S., Larson, M.H., Gilbert, L.A. et al. (2013). Repurposing CRISPR as an RNA-guided platform for sequence-specific control of gene expression. Cell 152: 1173–1183. Gilbert, L.A., Larson, M.H., Morsut, L. et al. (2013). CRISPR-mediated modular RNA-guided regulation of transcription in eukaryotes. Cell 154: 442–451. Chavez, A., Tuttle, M., Pruitt, B.W. et al. (2016). Comparison of Cas9 activators in multiple species. Nat. Methods 13: 563–567. Tak, Y.E., Kleinstiver, B.P., Nuñez, J.K. et al. (2017). Inducible and multiplex gene regulation using CRISPR–Cpf1-based transcription factors. Nat. Methods 14: 1163–1166. Horlbeck, M.A., Gilbert, L.A., Villalta, J.E. et al. (2016). Compact and highly active next-generation libraries for CRISPR-mediated gene repression and activation. eLife 5: e19760. Weinhold, B. (2006). Epigenetics: the science of change. Environ. Health Perspect. 114: A160–A167. Pfister, S.X. and Ashworth, A. (2017). Marked for death: targeting epigenetic changes in cancer. Nat. Rev. Drug Discovery 16: 241–263. Enríquez, P. (2016). CRISPR-mediated epigenome editing. Yale J. Biol. Med. 89: 471–486. Liu, X.S., Wu, H., Ji, X. et al. (2016). Editing DNA methylation in the mammalian genome. Cell 167: 233–47.e17. Hilton, I.B., D’Ippolito, A.M., Vockley, C.M. et al. (2015). Epigenome editing by a CRISPR/Cas9-based acetyltransferase activates genes from promoters and enhancers. Nat. Biotechnol. 33: 510–517. Gootenberg, J.S., Abudayyeh, O.O., Lee, J.W. et al. (2017). Nucleic acid detection with CRISPR-Cas13a/C2c2. Science 356: 438–442. Abudayyeh, O.O., Gootenberg, J.S., Essletzbichler, P. et al. (2017). RNA targeting with CRISPR-Cas13. Nature 550: 280–284. Cox, D.B.T., Gootenberg, J.S., Abudayyeh, O.O. et al. (2017). RNA editing with CRISPR-Cas13. Science 358: 1019–1027. Mackay, J.P., Font, J., and Segal, D.J. (2011). The prospects for designer single-stranded RNA-binding proteins. Nat. Struct. Mol. Biol. 18: 256–261. Patton, E.E. and Zon, L.I. (2001). The art and design of genetic screens: zebrafish. Nat. Rev. Genet. 2: 956–966.

References

77 St Johnston, D. (2002). The art and design of genetic screens: Drosophila

melanogaster. Nat. Rev. Genet. 3: 176–188. 78 Jorgensen, E.M. and Mango, S.E. (2002). The art and design of genetic

screens: Caenorhabditis elegans. Nat. Rev. Genet. 3: 356–369. 79 Boutros, M. and Ahringer, J. (2008). The art and design of genetic screens:

RNA interference. Nat. Rev. Genet. 9: 554–566. 80 Shalem, O., Sanjana, N.E., Hartenian, E. et al. (2014). Genome-scale

CRISPR-Cas9 knockout screening in human cells. Science 343: 84–87. 81 Wang, T., Wei, J.J., Sabatini, D.M., and Lander, E.S. (2014). Genetic screens

in human cells using the CRISPR-Cas9 system. Science 343: 80–84. 82 Gilbert, L.A., Horlbeck, M.A., Adamson, B. et al. (2014). Genome-scale

83

84

85

86

87

88

89

90

91

92

93

CRISPR-mediated control of gene repression and activation. Cell 159: 647–661. Joung, J., Konermann, S., Gootenberg, J.S. et al. (2017). Genome-scale CRISPR-Cas9 knockout and transcriptional activation screening. Nat. Protoc. 12: 828–863. Fischer, S., Handrick, R., and Otte, K. (2015). The art of CHO cell engineering: a comprehensive retrospect and future perspectives. Biotechnol. Adv. 33: 1878–1896. Yang, Z., Wang, S., Halim, A. et al. (2015). Engineered CHO cells for production of diverse, homogeneous glycoproteins. Nat. Biotechnol. 33: 842–844. Fan, L., Kadura, I., Krebs, L.E. et al. (2012). Improving the efficiency of CHO cell line generation using glutamine synthetase gene knockout cells. Biotechnol. Bioeng. 109: 1007–1015. Santiago, Y., Chan, E., Liu, P.-Q. et al. (2008). Targeted gene knockout in mammalian cells by using engineered zinc-finger nucleases. Proc. Natl. Acad. Sci. U.S.A. 105: 5809–5814. Cost, G.J., Freyvert, Y., Vafiadis, A. et al. (2010). BAK and BAX deletion using zinc-finger nucleases yields apoptosis-resistant CHO cells. Biotechnol. Bioeng. 105: 330–340. Orlando, S.J., Santiago, Y., DeKelver, R.C. et al. (2010). Zinc-finger nuclease-driven targeted integration into mammalian genomes using donors with limited chromosomal homology. Nucleic Acids Res. 38: e152. Sakuma, T., Takenaga, M., Kawabe, Y. et al. (2015). Homologous recombination-independent large gene cassette knock-in in CHO cells using TALEN and MMEJ-directed donor plasmids. Int. J. Mol. Sci. 16: 23849–23866. Cristea, S., Freyvert, Y., Santiago, Y. et al. (2013). In vivo cleavage of transgene donors promotes nuclease-mediated targeted integration. Biotechnol. Bioeng. 110: 871–880. Lee, J.S., Grav, L.M., Lewis, N.E., and Faustrup Kildegaard, H. (2015). CRISPR/Cas9-mediated genome engineering of CHO cell factories: application and perspectives. Biotechnol. J. 10: 979–994. Grav, L.M., Lee, J.S., Gerling, S. et al. (2015). One-step generation of triple knockout CHO cell lines using CRISPR/Cas9 and fluorescent enrichment. Biotechnol. J. 10: 1446–1456.

205

206

8 CRISPR Toolbox for Mammalian Cell Engineering

94 Schmieder, V., Bydlinski, N., Strasser, R. et al. (2017). Enhanced genome

95

96

97

98

99

100

101

102

103

104

105

editing tools for multi-gene deletion knock-out approaches using paired CRISPR sgRNAs in CHO cells. Biotechnol. J. 13: 1700211. Sun, T., Li, C., Han, L. et al. (2015). Functional knockout of FUT8 in Chinese hamster ovary cells using CRISPR/Cas9 to produce a defucosylated antibody. Eng. Life Sci. 15: 660–666. Zong, H., Han, L., Ding, K. et al. (2017). Producing defucosylated antibodies with enhanced in vitro antibody-dependent cellular cytotoxicity via FUT8 knockout CHO-S cells. Eng. Life Sci. 17: 801–808. Chung, C., Wang, Q., Yang, S. et al. (2017). Integrated genome and protein editing swaps α-2,6 sialylation for α-2,3 sialic acid on recombinant antibodies from CHO. Biotechnol. J. 12: 1600502. Spahn, P.N., Hansen, A.H., Hansen, H.G. et al. (2016). A Markov chain model for N-linked protein glycosylation – towards a low-parameter tool for model-driven glycoengineering. Metab. Eng. 33: 52–66. Park, J.H., Jin, J.H., Lim, M.S. et al. (2017). Proteomic analysis of host cell protein dynamics in the culture supernatants of antibody-producing CHO cells. Sci. Rep. 7: 44246. Chiu, J., Valente, K.N., Levy, N.E. et al. (2017). Knockout of a difficult-toremove CHO host cell protein, lipoprotein lipase, for improved polysorbate stability in monoclonal antibody formulations. Biotechnol. Bioeng. 114: 1006–1015. Lee, N., Shin, J., Park, J.H. et al. (2016). Targeted gene deletion using DNA-free RNA-guided Cas9 nuclease accelerates adaptation of CHO cells to suspension culture. ACS Synth. Biol. 5: 1211–1219. Grav, L.M., la Cour Karottki, K.J., Lee, J.S., and Kildegaard, H.F. (2017). Application of CRISPR/Cas9 genome editing to improve recombinant protein production in CHO cells. Methods Mol. Biol. (Clifton, NJ) 1603: 101–118. Shen, C.-C., Sung, L.-Y., Lin, S.-Y. et al. (2017). Enhancing protein production yield from Chinese hamster ovary cells by CRISPR interference. ACS Synth. Biol. 6: 1509–1519. Lee, J.S., Kallehauge, T.B., Pedersen, L.E., and Kildegaard, H.F. (2015). Site-specific integration in CHO cells mediated by CRISPR/Cas9 and homology-directed DNA repair pathway. Sci. Rep. 5: 8572. Bachu, R., Bergareche, I., and Chasin, L.A. (2015). CRISPR-Cas targeted plasmid integration into mammalian cells via non-homologous end joining. Biotechnol. Bioeng. 112: 2154–2162.

207

9 CHO Cell Engineering for Improved Process Performance and Product Quality Simon Fischer 1 and Kerstin Otte 2 1 Boehringer Ingelheim Pharma GmbH & Co.KG, Cell Line Development CMB, Bioprocess & Analytical Development, Birkendorfer Straße 65, 88397 Biberach an der Riss, Germany 2 University of Applied Sciences Biberach, Institute of Applied Biotechnology, Hubertus-Liebrecht-Strasse 35, 88400 Biberach, Germany

9.1 CHO Cell Engineering Although Chinese hamster ovary (CHO) cells have been successfully employed as a manufacturing host cell system for more than 30 years, these cell lines still suffer from naturally occurring limitations with regard to growth rates and recombinant protein production capacity [1]. The molecular basis for these limitations may be based on the fact that this cell type has not evolved to exhibit superior growth and recombinant protein production properties in large-scale stirred tank bioreactors, but rather to accomplish its task as a fibroblast cell in an ovary tissue. Nonetheless, today, an entire industry is primarily relying on CHO cells as a manufacturing host system for therapeutic protein production [2, 3]. Historically, the most important modified CHO cell lines, which eventually paved the way for an economical utilization of CHO cells for biopharmaceutical manufacturing, were the different dihydrofolate reductase (DHFR)-deficient CHO sublines named DXB11 and DG44, respectively [4, 5]. They mark the starting point of the commercial exploitation of CHO cells in biotechnology in the mid-1980s [6]. Several years later, a more effective metabolic selection system was introduced, which was based on the glutamine synthetase (GS) enzyme that can be inhibited by methionine sulfoximine (MSX), enabling a more stringent selection and thus generation of high-expressing recombinant CHO cell lines [7, 8]. In the meanwhile, the number of GS-deficient CHO host cell lines, besides the first CHO-K1SV [9], has expanded considerably and several biopharmaceutical companies are already using these cell lines for commercial manufacturing. Both DHFR- and GS-deficient CHO cell lines can be selected for stable transfectants in growth media lacking hypoxanthine/thymidine and l-glutamine, respectively, if the cells were previously transfected with an expression plasmid encoding a transgene in combination with a functional DHFR or GS gene copy.

Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

208

9 CHO Cell Engineering for Improved Process Performance and Product Quality

Monoclonal antibodies (mAbs) are still the most frequently developed class of biologics [10]. However, the number of multispecific formats as well as highly potent fusion proteins (in the following sections, referred to as “multispecifics”) have also increased dramatically over the past few years [11, 12]. Certainly, these molecules have never undergone an evolutionary process and thus represent new territories for production cells in terms of translation, intracellular processing, folding, and secretion. Thus, it seems little surprising that these novel and complex therapeutics often turn out to be difficult-to-express for CHO cells [13–15]. Although the state-of-the-art industrial cell line development (CLD) workflows frequently deliver clonal cell lines exceeding product yields of 5 g/l, also standard IgGs were found to be challenging to be expressed by CHO cells [16, 17], e.g. if the molecule design is suboptimal. For instance, if the primary amino acid sequence of an IgG exhibits aggregation-prone regions on the surface or very unusual amino acid residues at certain positions, compared to a wide range of known IgG sequences, it can also have a dramatic influence productivity [18–20]. Consequently, there is an urgent demand to steadily improve industrial CHO host cell lines in order to be prepared for future bioprocessing challenges. In addition, health care systems are already facing tremendous costs associated with the increasing demand for therapeutic proteins to address unmet medical needs [21]. Hence, industrial manufacturing processes of biopharmaceuticals are highly dependent on nonexpensive and high-yielding production platforms in order to maximize production yields but also to reduce associated costs. Therefore, host cell engineering of CHO cells represents a valuable strategy to overcome limitations in the production of biologics. There are different opportunities to counteract limitations of mammalian cell factories. In the following chapters, we have summarized the most relevant cell line and state-of-the-art engineering techniques currently applied for CHO cell engineering including overexpression or knockout of genes, as well as the usage of noncoding RNAs. Furthermore, we provide an overview on applications of cell line engineering approaches in CHO cells to enhance recombinant protein production, repress cell death and accelerate growth, and modulate posttranslational modifications (PTMs).

9.2 Methods in Cell Line Engineering 9.2.1

Overexpression of Engineering Genes

After the identification of beneficial genes for the production of biopharmaceutical proteins, the overexpression of these genes is one of the promising strategies to improve the performance of mammalian production cell lines. This technique has frequently been exploited during the past 25 years using both transient and stable overexpression strategies. To achieve overexpression of beneficial genes, the usually codon-optimized complementary DNA (cDNA) lacking any intronic sequences is isolated and cloned into a mammalian expression vector [22]. The plasmid DNA (pDNA) is subsequently delivered into the cells by transfection preferentially via electroporation or lipofection. Transfected

9.2 Methods in Cell Line Engineering

cells are then subjected to antibiotic selection pressure to generate cell pools with the plasmid DNA stably integrated into their genome. The expression of the gene of interest (GOI) is often driven by strong viral or cellular promoter and enhancer sequences to ensure high expression levels [23], while the selective gene is normally controlled by weak promoters to increase the overall expression level [24]. After the selection process, the resulting cell culture represents a heterogeneously mixed pool of cells showing various extent of transgene overexpressions. This procedure, however, results in phenotypic differences between individual cells, and thus, cell lines derived from a single progenitor cell subsequently have to be established (process is called “single cell cloning”) to obtain a homogenous host cell line exhibiting a strong and stable engineered phenotype. 9.2.2

Gene Knockout

Instead of overexpressing beneficial GOIs to improve production characteristics of CHO cells, there is the possibility to knockout disadvantageous genes to engineer host cell lines [25]. For the stable deletion of genes from the genome, several methods are available including chemical- or radiation-induced random mutagenesis in addition to a variety of precise genome editing techniques. However, targeted genome engineering with high specificity has become a preferentially used methodology to random mutagenesis, especially from a regulatory point of view. In this conjunction, the current state-of-the-art technologies mainly comprise the use of zinc-finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), meganucleases, or the recently introduced clustered regularly interspaced short palindromic repeats (CRISPRs/Cas9) system [26–29]. The CRISPR/Cas9 technology has recently entered the field of CHO cell engineering because of advantages such as the ease of use of the methodology as well as less time-consuming, and much more cost-effective procedures compared to the more sophisticated and expensive alternatives. Hence, in the context of rational design of mammalian cell factories, this novel methodology doubtlessly has the potential to revolutionize current cell line optimization strategies [30] and is of particular interest for multiplexed CHO cell engineering approaches where several genes can be rapidly knocked out. 9.2.3

Noncoding RNA-mediated Gene Silencing

Instead of deleting a gene from the genome of a production cell, disadvantageous genes may be silenced. This can be achieved by RNA interference (RNAi), and as the gene is silenced, this method is also known as knockdown of gene expression. RNAi was originally discovered in in Caenorhabditis elegans (C. elegans) [31], and since then, gene silencing using small double-stranded RNAs (dsRNAs), which are also termed small-interfering RNAs (siRNAs), has become a frequently applied technology in cell engineering. SiRNAs are small double-stranded RNA molecules of 20–25 base pairs in length and exhibit complete sequence complementarity to their target messenger RNA (mRNA) [32]. After exogenous delivery of siRNAs by directly introducing small dsRNA into the cytoplasm of

209

210

9 CHO Cell Engineering for Improved Process Performance and Product Quality

a cell or by expression from small hairpin RNA (shRNA) containing vectors, the RNA molecules are cleaved by the RNase-III enzyme DICER and loaded onto an Argonaute-2 (AGO2) protein [33, 34], which is the only member of the AGO family exhibiting slicer activity [34]. The resulting RNA-induced silencing complex (RISC) is established in the cytoplasm [35] and binding to the target mRNA leads to immediate mRNA degradation [36]. The thermodynamic stability at the 5′ -terminus of the dsRNA determines which strand will be favored as the guide strand for binding to the mRNA target [37]. Although siRNAs are considered artificially designed molecules for targeted gene silencing, recent studies demonstrated the presence of naturally occurring siRNAs in eukaryotes. Endogenous siRNAs can be derived from transposons, repetitive sequences, long stem loop structures, or sense–antisense transcripts [38–41]. Notably, the high specificity of an siRNA limits its application for multiplexed knockdown of a larger number of engineering genes and thus for modulating several cellular pathways in parallel. A second species of noncoding RNA molecules are microRNAs (miRNAs), which have recently entered the field of CHO cell engineering. MiRNAs are 19–25 nucleotides in length and bind to the 3′ untranslated region (3′ UTR) of a target gene by imperfect base pairing. Although the seed sequence, usually composed of nucleotides 2–8, binds as a perfect match, the remaining nucleotides only partially bind by imperfect base pairing and thereby allow for the targeting of multiple genes [42, 43]. Therefore, these endogenous small RNAs are capable of regulating entire cellular pathways [44, 45], and their use as engineering tools may enable the modulation of entire signaling pathways. This may improve phenotypic outcome because of the fact that changes in cellular phenotypes are most likely not the result of altering the expression of an individual gene but rather of a plethora of genes involved in the same or different pathways. To substantiate this hypothesis, recent studies discovered that large numbers of miRNAs can actually regulate multiple different cellular pathways concomitantly in order to keep the cell in homeostasis [46]. These properties in addition to the fact that overexpression of a noncoding miRNA does not add any additional translational burden to the host cell make miRNAs attractive molecular tools for next-generation host cell engineering. In order to functionally analyze a large number of miRNAs in CHO cells, high-content functional miRNA screening approaches [47], as well as miRNome profiling studies, helped to unravel novel target molecules to be used for CHO cell engineering [48–50]. First identified in 1993 as a critical regulator of development in the nematode C. elegans [51], miRNAs have been demonstrated to play essential roles in the regulation of virtually all cellular process in metazoans as a fine-tuner of gene expression [42]. Their important role in cellular regulation is emphasized by the exceptionally high evolutionary conservation of sequences among species [52, 53]. Of note, about 50% of all miRNA loci in mammals are in close proximity to other miRNAs [34], enabling the generation of miRNA clusters, which are transcribed from a single polycistronic transcription unit [54]. The fine-tuning of the expression of many different target genes is achieved by the imperfect nature of target recognition of miRNAs and thereby lowering target specificity of an individual miRNA to its mRNA targets [36, 55, 56]. MiRNAs with

9.3 Applications of Cell Line Engineering Approaches in CHO Cells

identical seed sequences are grouped into families [43]; however, miRNAs from the same seed family can have surprisingly different roles in vivo. The function of miRNAs highly depends on the composition of the cellular transcriptome, which impedes a clear functional classification for individual miRNAs [47, 57]. The lack of genomic sequence information before 2011 substantially hindered miRNA research in CHO cells [58], and different strategies for functional miRNA analysis had been pursued such as transient transfections of either miRNA mimics or artificial expression vectors encoding chimeric miRNA precursor hairpins [59, 60]. However, chimeric miRNA expression vectors were shown to be inferior to vectors coding for the endogenous Cricetulus griseus miRNA sequence [47, 61]. Taking advantage of next-generation RNA sequencing technology, 307 mature miRNAs and 200 pre-miR sequences were initially found in total RNA samples of different CHO cell lines cultivated under various culture conditions [62, 63]. These miRNA sequences were subsequently annotated as cgr-miRNAs in miRBase [64, 65]. Recently, thorough in silico re-analysis of the CHO genome revealed the presence of an additional 71 mature miRNAs in CHO cells [66].

9.3 Applications of Cell Line Engineering Approaches in CHO Cells 9.3.1

Enhancing Recombinant Protein Production

The challenge of improving recombinant protein expression and thereby overall yield from culturing of production cells can be met through various molecular approaches of cell line engineering and as well by targeting a variety of cellular processes or signal transduction pathways. These may include gene transcription, protein translation/modification, unfolded protein response (UPR), ubiquitination/proteasomal degradation, metabolism, intracellular trafficking, cytoskeleton dynamics, or secretion/exocytosis. In the following chapter, we will discuss a number of examples demonstrating the breadth of used approaches. A comprehensive list of cell engineering studies aiming at improving recombinant protein production in CHO cells is provided in Table 9.1. Cellular metabolism is one of the most interesting processes to be addressed while aiming to improve culture performance of CHO production cells. Tailoring the metabolic activity of CHO production cells by overexpressing specific genes, which affected cellular metabolism, was performed about two decades ago to increase culture longevity and product yields. The forced expression of vitreoscilla hemoglobin (VHb) in CHO cells reported by Pendse and Bailey led to a 40–100% increase in human tissue plasminogen activator (tPA) productivity [85]. Several years later, nutrient consumption and accumulation of toxic byproducts in the cell culture medium were addressed by other groups to contribute to expand the list of metabolic engineering genes [86, 87, 89–94]. In addition, most of the mentioned studies also demonstrated increased product yields as a result of improved media utilization apart from optimizing the metabolic activity of engineered CHO cells. Recently, overexpressing glutamate-cysteine ligase modifier subunit (GCLM) was shown to improve specific productivity, titer, and the frequency of generating high-producing CHO clones by 70% [76].

211

Table 9.1 CHO cell engineering approaches for enhanced recombinant protein production in CHO cell lines. Cellular pathway

Origin of engineering gene

Protein synthesis

Target gene

Engineered phenotype

References

Not indicated (overexpression)

Protein disulfide isomerase (PDI)

Increased productivity of monoclonal antibodies

Borth et al. [67]

CHO (overexpression)

ERp57 (an isoform of PDI)

2.1-Fold increase in specific thrombopoietin (TPO) productivity without decreasing cell growth

Hwang et al. [68]

CHO (overexpression)

Calnexin (CNX) and calreticulin (CRT)

1.9-Fold increase in specific TPO productivity without negatively influencing cell growth and biological activity of the recombinant TPO

Chung et al. [69]

Bovine (overexpression)

Nonphosphorylatable version of the eukaryotic translation initiation factor 2 α (eIF2α)

Enhanced transient expression of recombinant proteins

Underhill et al. [70]

Artificial (overexpression)

Artificial zinc finger protein transcription factor (ZFP-TF)

10-Fold increase in IgG titer

Kwon et al. [71]

CHO (overexpression)

Activating transcription factor 4 (ATF4)

Increase in human antithrombin III (AT-III) titer

Ohya et al. [72]

CHO (overexpression)

Growth arrest and DNA damage-inducible protein 34 (GADD34)

40% increase in human AT-III titer

Omasa et al. [73]

Human (overexpression)

Mammalian target of rapamycin (mTOR)

Increased cell growth, viability, apoptosis resistance, and specific productivity of glycoproteins

Dreesen and Fussenegger [74]

Human (overexpression)

Heat-shock 70 kDa protein 5 (BIP), activating transcription factor 6C (ATF6C), and X-box binding protein 1 (XBP1)

Increased expression of difficult-to-express monoclonal antibodies and reduced cell growth

Pybus et al. [17]

CHO (overexpression)

Ying Yang 1 (YY1)

Increased production of several product genes (SEAP, VEGF165, and IgG including rituximab)

Tastanova et al. [75]

CHO (overexpression)

Glutamate-cysteine ligase modifier subunit (GCLM)

Increased specific productivity, mAb titer, and frequency of high producer clones by 70%

Orellana et al. [76]

CHO (knockout)

Dihydrofolate reductase (DHFR)

Transgene amplification for increased protein production

Urlaub et al. [5] Page and Sydenham [77]

CHO (knockout)

Glutamine synthetase (GS)

Transgene amplification for increased protein production

Sanders et al. [8] Cockett et al. [7]

CHO (knockout)

DHFR

Rapid establishment of DHFR−/− cells within one month

Santiago et al. [29]

CHO (knockout)

GS

Increased selection efficiency of high-producing CHO cells

Fan et al. [9]

CHO (knockdown)

DHFR

>100% increase in specific IgG productivity and 30% improved stability of transgene expression

Wu et al. [78]

CHO (knockout)

Insulin-like growth factor 1 receptor (IGF1R)

Increased production of a difficult-to-express protein (IGF1)

Romand et al. [79]

CHO (knockout)

FAM60A

Increased expression stability of mAb production clones

Ritter et al. [80]

CHO (knockout)

C12orf35

Several fold increased mAb productivity of stable pools and clones

Ritter et al. [81]

CHO (knockout)

Activating transcription factor 6 β (ATF6β)

Improved mAb productivity

Pieper et al. [82]

CHO (knockout)

Ceramide synthase 2 (CerS2) and Rab1 GAP Tbc domain family member 20 (Tbc1D20)

Improved mAb productivity

Pieper et al. [83]

CHO (knockdown)

Breast cancer 1 (BRCA1)

Increased mAb productivity of up to 5.3-fold

Matsuyama et al. [84]

Table 9.1 (Continued) Cellular pathway

Origin of engineering gene

Metabolism

Target gene

Engineered phenotype

References

Vitreoscilla (overexpression)

Vitreoscilla hemoglobin (VHb)

40–100% increase in specific human tPA productivity

Pendse and Bailey [85]

Rat (overexpression)

Carbamoyl phosphate synthetase I (CPS I) and ornithine transcarbamoylase (OTC)

25–33% Decreased accumulation of ammonium and 15–30% increased cell growth

Park et al. [86]

Yeast (overexpression)

Pyruvate carboxylase 2 (PYC2)

35% Decrease in lactate production and twofold increase in product titer

Fogolin et al. [87]

Human (overexpression)

Pyruvate carboxylase (PC)

Increased viability because of 21–39% decreased lactate production

Kim and Lee [88]

Mouse (overexpression)

Glucose transporter protein 5 (GLUT5)

Less lactate production and higher cell densities in fructose fed-batch processes

Wlaschin and Hu [89]

Mouse (overexpression)

GLUT5

Less lactate production, increased growth rate, prolonged culture duration, and higher product titer

Le et al. [90]

CHO (overexpression)

Malate dehydrogenase II (MDH2)

Increased intracellular levels of ATP and NADH led to an 1.9-fold improvement in integral viable cell number

Chong et al. [91]

CHO (overexpression)

Taurine transporter (TAUT)

Improved viability and increased IgG titer

Tabuchi et al. [92]

CHO (overexpression)

TAUT and alanine aminotransferase 1 (ALT1)

Higher IgG yield in shorter cultivation time

Tabuchi and Sugiyama [93]

CHO (knockout)

Lactate dehydrogenase A (LDHA)

45–79% Reduced lactate concentrations and diminished glucose consumption

Kim and Lee [94]

CHO (knockout)

LDHA

Diminished medium acidification because of decreased lactate production leading to less apoptosis

Jeong et al. [95]

Secretion

CHO (knockout)

Enolase 1 (ENO1)

Increase in viable cell density

Doolan et al. [96]

CHO (knockout)

LDHA and pyruvate dehydrogenase kinase (PDHK)

68–90% Increase in IgG titer

Zhou et al. [97]

CHO (knockdown)

Knockdown of LDHA combined with BCL2 overexpression

Improved culture longevity because of decreased lactate production and increased apoptosis resistance

Jeong et al. [98]

Human (overexpression)

X-box binding protein 1 (XBP1)

Higher endoplasmic reticulum content and increase in product titer

Tigges and Fussenegger [99] Becker et al. [100]

Human (overexpression)

Spliced form of XBP-1 (XBP1s)

Fourfold increase in specific IgG productivity

Ku et al. [101] Gulis et al. [102]

Human (overexpression)

Suppressor of loss of YPT1 protein 1 (SLY1) and syntaxin binding protein 3 (MUNC18C)

15-Fold increase in IgG production

Peng and Fussenegger [103]

Human (overexpression)

Tricystronic expression of SLY1, MUNC18C, and XBP1

20-Fold increase in IgG production

Peng and Fussenegger [103]

Human (overexpression)

Ceramide transfer protein (CERT)

Increase in specific productivity of human serum albumin (HSA) and monoclonal antibodies

Florin et al. [104]

Human (overexpression)

Mutant form of CERT (S132A)

35% Increase in specific t-PA productivity

Rahimpour et al. [105]

Human (overexpression)

Synaptosome-associated protein of 23 kDa (SNAP-23) and vesicle-associated membrane protein 8 (VAMP8)

Increase in SEAP productivity by enhanced secretory capacity

Peng et al. [106]

Human (overexpression)

Human signaling receptor protein 14 (SRP14)

Improved secretion and production of difficult-to-express proteins

Le Fourn et al. [107]

Table 9.1 (Continued) Cellular pathway

Origin of engineering gene

Cell cycle

microRNA

Target gene

Engineered phenotype

References

Human (overexpression)

Cyclin-dependent kinase inhibitor 1A (p21CIP1 ) and CCAAT/enhancer-binding protein 𝛼 (C/EBP-𝛼)

Growth arrest and 10- to 15-fold increase in specific SEAP productivity

Fussenegger et al. [108]

Human (overexpression)

Tricystronic expression of p21CIP1 , C/EBP-𝛼, and BCL-xL

Growth arrest and 30-fold increase in specific SEAP productivity

Fussenegger et al. [108] Astley et al. [109]

Human (overexpression)

Cyclin-dependent kinase inhibitor 1B (p27KIP1 )

Increased specific SEAP productivity

Mazur et al. [110]

Human (overexpression)

p21CIP1

Fourfold increase in IgG production

Bi et al. [111]

Human (overexpression)

Myelocytomatosis oncogene (C-MYC)

>70% Increase in maximal cell density without additional supply of nutrients

Kuystermans and Al-Rubeai [112]

CHO (knockout)

Ataxia telangiectasia and Rad3 related (ATR)

Fourfold increase in specific IgG productivity and threefold improved IgG titer

Lee et al. [113]

CHO (inhibition)

miR-7a-5p (inhibition)

Reduced growth and enhanced SEAP productivity

Barron et al. [102] Meleady et al. [56] Sanchez et al. [114]

CHO (overexpression)

miR-30a, c, d, e

Enhanced mAb and SEAP productivity

Fischer et al. [47]

CHO (overexpression)

miR-2861

Enhanced mAb and SEAP productivity

Fischer et al. [115]

CHO (overexpression)

miR-17-5p

Enhanced growth and EPO-Fc productivity

Jadhav et al. [59] Clarke et al. [116] Jadhav et al. [117] Loh et al. [118]

CHO (overexpression)

miR-19b

Enhanced mAb productivity

Loh et al. [118] Clarke et al. [116]

CHO (overexpression)

miR-20a

Enhanced mAb productivity

Loh et al. [118] Clarke et al. [116]

CHO (overexpression)

miR-17-92a

Enhanced mAb and EPO-Fc productivity

Jadhav et al. [117] Loh et al. [118]

CHO (overexpression)

miR-92a

Enhanced mAb productivity

Loh et al. [119]

Human (overexpression)

miR-1287

Enhanced mAb productivity

Strotbek et al. [60]

Human (overexpression)

mitosRNA-1978

Enhanced mAb productivity

Strotbek et al. [60]

CHO (inhibition)

miR-34a

Enhanced SEAP productivity

Kelly et al. [120]

CHO (overexpression)

miR-483-3p

Enhanced mAb and rAAV productivity

Emmerling et al. [121]

CHO (inhibition)

miR-23

Enhanced SEAP productivity

Kelly et al. [122]

CHO (overexpression)

miR-143-3p

Enhanced productivity of difficult-to-express proteins

Schoellhorn et al. [123]

Human (overexpression)

miR-557

Twofold increase in difficult-to-express mAb in fed-batch cultivation

Strotbek et al. [60] Fischer et al. [124]

218

9 CHO Cell Engineering for Improved Process Performance and Product Quality

Moreover, the cellular protein synthesis machinery has been exploited by stable genomic introduction of genes promoting protein production to increase the yield of recombinant proteins (r-proteins) using CHO cell cultures. Overexpression of specific transcription factors such as ZFP-TF, ATF4, GADD34, or more recently YY1 was shown to substantially boost volumetric yields of various r-proteins of up to 10-fold compared to parental cells [71–73, 75]. In addition, forced ectopic expression of mammalian target of rapamycin (mTOR), which is another key protein critically involved in protein synthesis, substantially increased overall culture performance of recombinant CHO cells leading to increased cell growth, viability, apoptosis resistance, and specific productivity [74]. Pybus et al. reported that co-overexpression of heat shock 70 kDa protein 5 (BIP), activating transcription factor 6C (ATF6C), and X-box binding protein 1 (XBP1) could increase the productivity of difficult-to-express mAbs in CHO cells [17]. This approach is especially of interest, as the demand for complex and difficult-to-express therapeutic proteins increases. Hence, strategies have to be developed to provide sufficient amount of clinical-grade material for clinical studies and market supply. Recently, Pieper et al. found that the stable knockdown of activating transcription factor 6 β (ATF6β), a repressor of the prosurvival and UPR promoting factor ATF6α, significantly improved antibody titer and viable cell density (VCD) in CHO-IgG cells under fed-batch conditions [82]. This was associated with an elevated expression of the UPR genes glucose-regulated protein 78 (GRP78), homocysteine-inducible ER protein with ubiquitin-like domain 1 (HERPUD1), and CCAAT/enhancer-binding protein homologous protein (CHOP) [82]. Another study on overcoming challenges of difficult-to-express proteins reported that the production of insulin-like growth factor 1 (IGF1) using CHO cells turned out to be difficult because of an activation of the endogenous IGF-1 receptor (IGF1R) on the production cell line by IGF1 itself. Elevated production of IGF1 resulted in growth retardation and low IGF1 product titers [79]. In an elegant approach, Romand et al. knocked out cgr-IGF1R in the CHO production cell line using ZFN technology to block the activated inhibitory signaling induced by the recombinantly expressed IGF1 and hence could increase IGF1 productivity by sevenfold [79]. Another promising strategy to increase productivity is the engineering of the secretory capacity of the CHO production cell. Le Fourn et al. overexpressed the human signaling receptor protein 14 (SRP14) to engineer the secretory capacity of CHO cells, which successfully improved product yields [107]. Further genetic engineering efforts to increase secretion yielded several exciting concepts describing the exploitation of various genes involved in the secretory pathway. Overexpression of protein disulfide isomerase (PDI), suppressor of loss of YPT1 protein 1 (SLY1), syntaxin binding protein 3 (MUNC18C), X-box binding protein 1 (XBP1), ceramide transfer protein (CERT), synaptosome-associated protein of 23 kDa (SNAP23), vesicle-associated membrane protein 8 (VAMP8), or combinations thereof could be demonstrated to increase product titer of recombinant proteins [67, 99, 101, 103–106]. A different approach to engineer the secretory pathway was reported by Pieper et al., where ectopic expression of a human mitochondrial genome-encoded small RNA (mitosRNA-1978)

9.3 Applications of Cell Line Engineering Approaches in CHO Cells

in an IgG expressing CHO cell line strongly improved specific productivity and the combined stable knockdown of two mitosRNA-1978 target genes, ceramide synthase 2 (CerS2) and the Rab1 GAP Tbc domain family member 20 (Tbc1D20), resulted in dramatically increased antibody production in CHO-IgG cells accompanied by enhanced cell growth [83]. A dominant issue in bioprocessing using mammalian expression systems is a continuous acidification of the culture medium because of the generation of lactic acid as a result of pyruvate conversion by the lactate dehydrogenase (LDH) [125]. As a result of this, lactate-mediated decrease in culture pH cell growth is impeded. Strategies to avoid oxidation of pyruvate to lactate comprise the repression of LDH and different strategies have been applied to achieve this goal. Using siRNAs directed against LDHA resulted in a decrease of LDHA levels below 11–25% of residual enzyme activity and a decline of lactate levels below 21% without impairing cell proliferation and productivity [88]. The simultaneous siRNA-mediated knockdown of LDH and pyruvate dehydrogenase kinase (PDHK) activity resulted in reduced lactate concentrations and increased volumetric mAb productivities [97]. Jeong et al. took advantage of using antisense mRNA for specific gene knockdown and showed that in CHO cells constitutively expressing LDH antisense mRNA erased LDHA activity successfully and diminished acidosis mediated apoptosis in CHO cells [95]. Notably, a recent study revealed that a complete knockout of LDH is lethal in CHO cells [126], a fact that always has to be taken into account if complete gene knockout strategies are envisioned to induce particular cell phenotypes using precise genome editing. Influencing cytoskeleton dynamics can lead to improved phenotypes of pharmaceutical production cells since cell division, intracellular trafficking, cell stability, and secretion might be optimized by genetic engineering [127]. A number of studies investigated differences between high- and low-producing mammalian cell factories and have identified cytoskeleton genes such as vimentin and annexin to be downregulated in high-producing cell lines, while other cytoskeleton genes were found to be upregulated [128–132]. Another key regulator protein of the actin cytoskeleton, cofilin-1 (CFL1) was identified to be highly downregulated when cell-specific secreted alkaline phosphatase (SEAP) productivity increased [133], its transient siRNA-mediated knockdown in CHO cells led to enhanced recombinant protein productivity by up to 80%, and stable downregulation led to a 65% and 47% increase in cellular SEAP and tPA productivity, respectively [134]. Although the modulation of cytoskeleton dynamics appears to be auspicious for improving CHO cell behavior, it still remains a neglected field for genetic engineering of mammalian production cell factories. Recently, it was reported that the lack of a telomeric region of chromosome 8 correlates with increased productivities and higher production stabilities of mAb producing CHO cell lines [81]. Further studies indicated that the knockout of the gene Fam60A, which is located within this telomeric region on Chr8 in CHO-K1a cells, leads to the isolation of significantly more clones exhibiting higher protein production stabilities of mAbs during long-term cultivation [80], and disruption of a second gene within this region, C12orf35, leads to increased productivities in recombinant CHO cell lines [135].

219

220

9 CHO Cell Engineering for Improved Process Performance and Product Quality

The extraordinarily high potential of miRNAs as a powerful cell engineering tool has recently been highlighted by various experimental setups to improve productivity of CHO cell lines. Barron et al. discovered that introduction of miR-7a-5p mimics in CHO cells decelerated cellular proliferation rates but increased cell-specific SEAP productivity [136]. Conversely, stable inhibition of miR-7a resulted by a different group in an increase in culture longevity and therefore elevated volumetric SEAP expression levels [114]. In addition, stable overexpression of members of the miR-17-92a cluster enhanced cell-specific mAb productivity, while enforced expression of the entire miR-17-92a cluster did not improve cellular performance [118]. These results are in line with findings by Jadhav et al., who discovered that stable overexpression of miR-17-5p increases recombinant protein expression in CHO cells by approximately threefold [117]. A recently published follow-up study by Loh et al. unraveled that overexpression of miR-92a resulted in enhanced recombinant protein production in CHO cells primarily because of modulation of cholesterol biosynthesis and increased Golgi volume, which in turn increased protein secretion [119]. Targeting the well-characterized tumor suppressor miRNA, miR-34a, by stable depletion using a miRNA sponge decoy vector, the overall product yield in both fed-batch and small-scale clonal batch CHO production cultures was enhanced by approximately twofold, despite having a negative impact on cell growth [120]. A transient screening performed by Strotbek et al. identified nine miRNAs to increase mAb titer. Two of these were human encoded, hsa-miR-557 and hsa-miR-1287, which could be demonstrated to positively affect growthand cell-specific productivity after stable overexpression in CHO cells [60]. A demonstration of the huge potential of the miRNA technology, especially with regard to difficult-to-express biologics, was recently provided by Fischer et al., who engineered CHO host cells by stable introduction of miR-557 [124]. MiRNA-engineered CHO cells were tested against control host cells (lacking the proproductive miR-557), and compared in two independent CLD campaigns using an easy as well as a difficult-to-express antibody [124]. Constitutive overexpression of miR-557 increased overall CLD performance and clonally derived final production cell lines exhibited an average increase in mAb titer of twofold in representative fed-batch cultivation processes, without having any detectable effects on product quality [124]. An unexpectedly huge number of miRNAs influencing diverse biotechnologically relevant cellular functions such as protein production, cell growth, apoptosis, and necrosis were identified in an even more comprehensive study [47]. Out of these, the entire miR-30 family contributed to enhanced recombinant protein expression in CHO cells, which could eventually be ascribed to an increase in cell-specific productivity (miR-30c, miR-30e) or cell growth (miR-30a) by stable miR-30 overexpression [47]. It later turned out that the proproductive effects of the miR-30 family in CHO cells was partly mediated by the regulation of genes involved in the ubiquitin pathway such as the S-phase kinase associated protein 2 (SKP2) [137]. In addition, miR-2861 was identified in the large-scale miRNA screen as a novel miRNA in CHO cells and could be presented as an enhancer of protein production by modulating HDAC5 without negatively influencing product quality attributes [115]. Furthermore, modulation

9.3 Applications of Cell Line Engineering Approaches in CHO Cells

of oxidative metabolism and mitochondrial activity by stable sequestration of miR-23 expression in CHO cells resulted in an average threefold increase in specific SEAP productivity by modulating LETM1 and IDH1 [122]. Emmerling et al. presented miR-483 as a species and product-independent enhancer of cellular productivity in mammalian cells, which enhanced recombinant mAb as well as adeno-associated virus production in CHO and HeLa cells, respectively [121]. Recently, miR-143 was reported to improve production of difficult-to-express proteins by targeting MAPK7 [123]. Given the substantial increase in CHO cell engineering studies using miRNAs, the authors believe that all these examples are just the beginning of a larger number of miRNA species, which will soon become available for next-generation CHO cell engineering. 9.3.2

Repression of Cell Death and Acceleration of Growth

A typical observation toward the end of a manufacturing process is that the viability of the grown cell culture decreases because of an increasing percentage of dying cells in the bioreactor. A dying cell culture population usually consists of a mixture of cells suffering from various forms of cell death including apoptosis, necrosis, and autophagy [138, 139]. A comprehensive list of cell engineering studies aiming at the modulation of cell death or cell proliferation in CHO cells is provided in Table 9.2. The programmed cell death, which is also called apoptosis, represents a well-characterized cellular pathway because of its fundamental importance in the development of cancer where this form of cell death has been lost by tumor cells [164]. Apoptosis is an essential cellular process and can be modulated by changing the expression of key proteins of the different apoptotic pathways, the intrinsic and extrinsic apoptotic pathways. Apoptosis is induced by nutrient depleted (spent) media or by the accumulation of toxic metabolites toward the end of a cultivation process [140, 141, 143, 144, 146, 154, 156]. In this context, a major problem is that dying cell cultures directly contribute to a loss of product quality attributes such as mAb fragmentation or decreased maturation of N-glycans at the Fc part of the antibody. This loss of product integrity occurs as proteases and other enzymes are released from nonviable cells and accumulate in the bioreactor at the end of the bioprocess [165]. Another issue connected with increased cell death is a dramatic loss of cell culture harvest performance because of markedly reduced filterability properties of the cell culture supernatant. When it comes to the development of a commercial production process, harvest performance is of critical importance for a successful implementation and scale-up of a bioprocess. Hence, a delayed onset of cell death during bioprocessing would be highly desirable to increase harvest viabilities, culture longevity, and ultimately product yields [166, 167]. Ectopic expression of antiapoptotic genes has become one of the mostly favored strategies to overcome apoptosis in CHO manufacturing cell lines. Overexpression of antiapoptotic genes such as B-cell lymphoma XL (BCL-XL), B-cell lymphoma 2 (BCL2), apoptosis and caspase activation inhibitor (AVEN), Fas apoptotic inhibitory molecule (FAIM), X-linked inhibitor of apoptosis (XIAP), or myeloid cell leukemia 1 (MCL1) could be demonstrated to prolong cultivation

221

Table 9.2 CHO cell engineering approaches for enhanced apoptosis resistance and cell growth in CHO cell lines. Cellular pathway

Apoptosis

Origin of engineering gene

Target gene

Engineered phenotype

References

Human (overexpression)

B-cell lymphoma 2 (BCL2)

75% Increase in maximum viable cell density

Tey et al. [140]

Human (overexpression)

BCL2 and BCL2-like 1 (BCL-xL)

Improved viability and enhanced apoptosis resistance

Mastrangelo et al. [141] Meents et al. [142]

Human (overexpression)

X-linked inhibitor of apoptosis (XIAP)

Improved apoptosis resistance

Sauerwald et al. [143]

Human (overexpression)

Apoptosis, caspase inhibitor (AVEN) and BCL-xL

Improved apoptosis resistance

Figueroa et al. [144]

Human (overexpression)

BCL-xL

88% Increase in viability and enhances IgG titer by >2-fold

Chiang and Sisk [145]

Human (overexpression)

Myelocytomatosis oncogene (c-myc) and BCL2

Improved apoptosis resistance and increased viable cell density

Infandi and Al-Rubeai (2005)

CHO (overexpression)

Fas apoptotic inhibitory molecule (FAIM)

Improved apoptosis resistance leading to 80% Wong et al. [146] increased VCD and 2.5-fold enhanced interferon 𝛾 (IFN𝛾) titer

Bombyx mori silkworm hemolymph (overexpression)

Apoptosis-inhibiting 30K protein (30Kc6)

Increased viable cell density and enhanced erythropoietin (EPO) titer of up to 10-fold

Choi et al. [147]

Human (overexpression)

Telomerase reverse transcriptase (TERT)

Increased apoptosis resistance resulting in higher viable cell density

Crea et al. [148]

Human and adenoviral (overexpression)

AVEN and control protein E1B 19K (E1B-19K)

Increased viable cell density and viability; 50% Figueroa et al. [149] enhanced IgG titer

Mouse (overexpression)

Murine double-mutant 2 (MDM2)

Increased apoptosis resistance

Arden et al. [150]

Human (overexpression)

E2F transcription factor 1 (E2F1)

20% increased viable cell density and improved proliferation

Majors et al. [151]

Human and adenoviral (overexpression)

AVEN, E1B-19K and a mutant form of XIAP (EAX197)

60% Increase in viable cell density and 80% enhanced IgG titer

Dorai et al. [152]

CHO (overexpression)

Heat-shock proteins 27 and 70 (HSP27 and HSP70)

Extended culture longevity and 2.5-fold increase in IFN𝛾 titer

Lee et al. [153]

Human (overexpression)

Myeliod cell leukemia 1 (MCL1)

Improved viability and 20–35% increased IgG Majors et al. [154] titer

Human (overexpression)

RAC-𝛼 serine/threonine-protein kinase (AKT1)

Delayed onset of apoptosis and autophagy during batch culture

Hwang and Lee [155]

Human (overexpression)

Mutated form of BCL-xL (Asp29Asn variant)

Improved apoptosis resistance

Majors et al. [156]

CHO (overexpression)

Chinese hamster heat-shock protein 27 (HSP27)

2.2-Fold higher peak VCD, sustained viability, Tan et al. [157] and 2.3-fold increase in mAb titer

CHO (knockout)

BCL2-associated X protein (BAX) and BCL2-antagonist/killer (BAK)

Increased apoptosis resistance because of inhibition of caspase activation leading up to fivefold increase in IgG titer

Cost et al. [158]

CHO (knockout)

BAX, BAK, and FUT8

Prolonged culture longevity because of diminished apoptosis; engineered clones produced afucosylated IgGs

Grav et al. [159]

CHO (knockdown)

Caspase 3

Extended culture longevity by more than two days in batch cultures

Kim and Lee (2002)

CHO (knockdown)

Caspase 3 and 7

Enhanced cell viability and 55% increase in hTPO titer

Sung et al. [160]

CHO (knockdown)

Caspase 8 and 9

Enhanced viability in batch and fed-batch cultivations

Yun et al. (2007)

CHO (knockdown)

Alpha-1,3/1,6-mannosyltransferase (ALG2), Requiem (REQ), Fas (TNFRSF6)-associated via death domain (FADD), and Fas apoptotic inhibitory molecule (FAIM)

Elevated cell density and culture longevity; 1.2- to 2.5-fold increase in IFN𝛾 titer

Wong et al. [146]

Table 9.2 (Continued) Cellular pathway

Autophagy and apoptosis

Proliferation

miRNA

Origin of engineering gene

Target gene

Engineered phenotype

References

CHO (knockdown)

Bax and Bak

Enhanced cell viability and 35% increase in IFN𝛾 titer

Lim et al. [161]

CHO (overexpression)

BCL-xL

Delayed onset of autophagy and apoptosis

Kim et al. [34]

Human (BCL-2) and CHO BCL-2 and Beclin-1 (Beclin-1) (overexpression)

Extended culture longevity and higher viable cell density because of decreased apoptosis and autophagy

Lee et al. [162]

Human (overexpression)

Cyclin-dependent kinase like 3 (CDKL3) and cytochrome c oxidase subunit (COX15)

Increased maximum viable cell density

Jaluria et al. [163]

CHO (overexpression)

Valosin-containing protein (VCP)

Increased cell proliferation and viability

Doolan et al. [96]

CHO (knockdown)

Cofilin (CFL1)

65% (SEAP) and 47% (tPA) increase in specific productivity

Hammond and Lee [134]

Human (overexpression)

miR-557

Enhanced cell growth

Strotbek et al. [60] Fischer et al. [124]

CHO (overexpression)

miR-30a

Enhanced cell growth

Fischer et al. [47, 115]

9.3 Applications of Cell Line Engineering Approaches in CHO Cells

processes by inhibiting apoptosis [140–144, 146]. Of note, a recent study demonstrated that overexpression of heat-shock protein 27 (HSP27) in CHO cells delayed the activation of caspases and hence apoptosis, which finally led to increased monoclonal antibody yields in fed-batch cultivation process [157]. Inhibition of proapoptotic genes can also be exploited to constitutively increase cell line performance of CHO manufacturing cell lines. Permanent genomic knockout of BCL2-associated X protein (BAX) and BCL2-antagonist/killer (BAK) expression in CHO production cells via ZFN improved resistance to apoptosis by decreasing the activation of caspases and thus enhanced therapeutic protein production by up to fivefold compared to controls [158]. In order to achieve an efficient knockdown of target gene siRNAs are often applied to specifically repress gene function in order to achieve improved apoptosis resistance, modified glycosylation pattern, increased metabolic efficiency, or simply enhanced recombinant protein production [88, 97, 125, 134, 146, 160, 161, 168–182]. Stable siRNA-mediated knockdown of BAX and BAK had been reported to increase culture longevity of CHO cell processes at nutrient-depleted or high osmolytic cultivation conditions, which finally led to improved cell growth and final interferon 𝛾 (IFN𝛾) titer [161]. Another antiapoptotic cell engineering strategy comprised repression of caspase-3 (CASP3) and -7 (CASP7), which were both silenced in parallel resulting in an increased performance of siRNA-engineered CHO cells during sodium butyrate (NaBu) treatment [160]. Sodium butyrate is an unspecific inhibitor of histone deacetylases (HDACs) and treatment of CHO cells with NaBu has shown to increase cell-specific productivity but also to induce apoptosis. In the presence of 1 mM sodium butyrate, knockdown of CASP3 and CASP7 increased cell viability and prolonged the cultivation process finally resulting in a substantial increase in human thrombopoietin (hTPO) yield. In another study, Wong et al. performed transcriptomics studies in CHO cells, which identified four early apoptosis signaling genes to be differentially expressed at the end of the cultivation process [146]. In this study, two proapoptotic genes (REQUIEM and ALG2) were stably inhibited by siRNAs, and higher apoptosis resistance, increased cell growth, and thus enhanced final IFNγ product yield were observed in engineered cell lines [146]. A recent study showed that stable knockdown of breast cancer 1 (BRCA1) gene in CHO cells using shRNAs increased mAb productivity up to 5.3-fold as well as production stability compared to control cells [183]. Although mock control cells exhibited decreased H3K4 methylation levels after long-term cultivation, H3K4 methylation levels remained unaffected in stable BRCA1 downregulated cell lines [183]. Besides apoptosis, autophagy and necrosis contribute to cell death in bioprocesses. Autophagy is triggered under starvation conditions when nutrient supply is limited at the end of a CHO production process [184]. Autophagy is a process that enables mammalian cells to generate energy by degradation of their own organelles and other cellular constituents [185]. Necrosis (or necroptosis) is defined as a form of cell death where the cell loses its membrane integrity because of mechanical or osmotic stress. CHO cell engineering approaches employing overexpression of the serine–threonine protein kinase 1 (AKT1) or the combination of BCL2 and Beclin-1 (BECN1), a component

225

226

9 CHO Cell Engineering for Improved Process Performance and Product Quality

of the phosphatidylinositol-3-kinase (PI3K) complex, have successfully been applied to simultaneously target both autophagy and apoptosis pathways in order to keep cell viability at a high level and to prolong cultivation times [155, 162]. Interestingly, forced expression of a core autophagy pathway genes called autophagy-related protein 9A (ATG9A) in CHO cells could not improve susceptibility to autophagy [186]. However, as overexpression of ATG9A has only been investigated in a single CHO cell line, it might be possible that this particular cell line did not react but other CHO cell lines may be improved. Given that autophagy plays an essential role in CHO cells during bioprocessing, it seems surprising that this cellular process has been rather neglected by CHO cell researchers in the past. In the opinion of the authors, autophagy represents a promising cellular mechanism to target in CHO cells in order to enhance bioprocess performance in fed-batch cultivation processes. However, we still need to understand autophagy in more detail and how it exactly works in CHO cells cultivated in serum-free media. Improving cell division rates and proliferation to accelerate growth phases in bioprocesses has also been frequently addressed by cell line engineering studies. As cell cultures usually show decreased cell-specific productivity during exponential growth, shortening the period for biomass production would add considerable value. As a result, production phases would be extended and product yields likely will increase. Doolan et al. demonstrated that overexpression of the valosin-containing protein (VCP) results in enhanced cell proliferation and viability of recombinant CHO cell lines [96]. Other cell line engineering approaches were targeting genes involved in cell cycle progression to improve growth rates of CHO production cells. For instance, co-overexpression of the cyclin-dependent kinase like 3 (CDKL3) and cytochrome-c oxidase subunit 15 (COX15) proteins resulted in an increased maximum VCD [163]. Kuystermans and Al-Rubeai stably introduced the myelocytomatosis oncogene (c-MYC) into CHO cells and increased maximum VCD by 70% without additional nutrient supply [112]. Surprisingly, even a blockage of cell cycle progression might be useful to enhance bioprocess performance, as overexpression of cell cycle inhibitors such as p21CIP1 or p27KIP1 increased cell-specific productivity in CHO cells at the expense of accompanying cell cycle arrest [108, 110, 111, 187]. To specifically accelerate cell growth and proliferation, cell cycle checkpoint kinases constitute interesting targets to be addressed by host cell engineering as these regulatory proteins control cell cycle progression and proliferation [188]. Knockdown of the cell cycle checkpoint kinase ataxia telangiectasia and Rad3 related (ATR) in mAb-producing CHO cells facilitated a more rapid establishment of high-producing cell clones during methotrexate (MTX)-mediated gene amplification compared to control cells. Cell cycle checkpoint regulators such as ATR induce cell cycle arrest as a consequence of MTX-mediated DNA replication stress in treated CHO cells [189]. In this conjunction, the observed accelerated cellular recovery was explained by the fact that ATR inhibition abrogated MTX-mediated blockage of the cell cycle. Apoptosis engineering also presents a suitable topic to be addressed using the novel miRNA technology in order to enhance process robustness because of increased harvest viabilities and resistance to stressful cultivation conditions in

9.3 Applications of Cell Line Engineering Approaches in CHO Cells

a bioreactor [190]. The first functional investigation of the potential of miRNAs to protect CHO cells from apoptotic cell death as presented by Druz et al. using, miR-466h. The authors discovered miR-466h to induce apoptosis in CHO cells during starvation in nutrient-depleted culture medium [191, 192]. ShRNA-mediated repression of miR-466h in CHO-SEAP cells increased the integral of VCD and culture longevity because of inhibition of apoptosis [193]. In addition, Druz et al. could further demonstrate that knockdown of miR-466h led to increased expression of the five antiapoptotic genes Bcl-2-like protein 2 (BCL2L2), Dolichyl-diphosphooligosaccharide—protein glycosyltransferase subunit (DAD1), baculoviral IAP repeat-containing protein 6 (BIRC6), signal transducer and activator of transcription 5A (STAT5A), and smoothened (SMO). All of these genes actually harbor partial complementary binding sites for miR-466h within the 3′ UTR of their transcripts, suggesting that these genes might serve as direct targets of miR-466 in CHO cells [193]. Conserved gene and microRNA function might be exploited to develop novel cell engineering strategies for CHO cells. For instance, miR-34a has previously been demonstrated to induce apoptosis in human and murine cells [194, 195]. MiR-34 actually represents the first miRNA to be investigated for its potential as a cancer therapeutic in clinical trials. Based on the proapoptotic function of miR-34, Kelly et al. stably sequestered miR-34a in CHO-SEAP cells in order to protect cells from apoptosis. Unexpectedly, engineered cells lacking miR-34a exhibited enhanced cell-specific productivity (see also Section 9.3.1), but no increased resistance to apoptosis [120]. These results show that constitutive repression of miRNAs in CHO cells might not always result in the expected phenotype. Furthermore, because of a functional redundancy among miRNAs in mammalian cells [46], it is very likely that other proapoptotic miRNAs present on the CHO genome might balance out a loss of a single miRNA in a specific cellular pathway. Consequently, gain-of-function approaches using miRNA overexpression potentially represents a better strategy toward improved CHO cell phenotypes compared to a miRNA knockout or knockdown. 9.3.3 Modulation of Posttranslational Modifications to Improve Protein Quality Although therapeutic proteins produced by CHO cells are predominantly used for application in humans and are generally considered to be safe, they only exhibit human-like but not human-identical PTMs [196–198]. Numerous cell line engineering approaches have aimed at modulating PTMs and product quality of recombinant proteins expressed in CHO cells and a comprehensive list of engineering studies is provided in Table 9.3. Among PTMs, glycosylation is of particular relevance because the glycosylation state critically affects the properties of recombinant proteins such as serum half-life, stability, immunogenicity, and functionality in the human body [177, 224, 225]. Control and optimization of protein N-glycosylation present on many recombinant glycoproteins are therefore crucial and alteration in the glycan composition can be used to modulate the production quality of a recombinant biotherapeutic. In addition, improper N-glycosylation of

227

Table 9.3 CHO cell engineering approaches for improved product quality in CHO cell lines.

Cellular pathway

Origin of engineering gene

Target gene

Engineered phenotype

References

Glycosylation

Human (overexpression)

Alpha 2,6 sialyltransferase (ST6GAL)

Expression of partly 𝛼2,6-sialylated recombinant proteins

Lee et al. [199] Zhang et al. [200] Bragonzi et al. [201]

Rat (overexpression)

ST6GAL

Expression of 𝛼2,6-sialylated recombinant human tissue plasminogen activator (tPA)

Minch et al. [202]

Bovine (GnT-IV ) and human (GnT-V ) (overexpression)

𝛼-1,3-d-mannoside 𝛽 1,4 N-acetylglucosaminyltransferase (GnT-IV) and 𝛼-1,6 d-mannoside 𝛽-1,6 N-acetylglucosaminyltransferase (GnT-V)

56.2% Increase in tetra-antennary sugar chains on recombinant IFN𝛾

Fukuta et al. [203]

Mouse (ST3GAL), rat (ST6GAL) and human (GnT-V ) (overexpression)

ST3GAL, ST6GAL, and GnT-V

Increase in the extent of sialylation of human recombinant IFN𝛾 of up to 80%

Fukuta et al. [204]

Rat (overexpression)

ST6GAL

Improved sialylation and therapeutic activity of recombinant IgG3

Jassal et al. [205]

Human (overexpression)

𝛼-2,3 sialyltransferase (ST3GAL) and 𝛽 1,4 galactosyltransferase (GalT)

Expression of homogeneously distributed and >90% sialylated glycoproteins

Weikert et al. [206]

Human (overexpression)

ST6GAL and GalT

Increased sialylation level of recombinant human EPO

Jeong et al. [98]

CHO (overexpression)

CMP-sialic acid transporter (CMP-SAT)

4–16% Increased sialylation of human IFN𝛾

Wong et al. [207]

Human (overexpression)

CMP-sialic acid synthetase (CMP-SAS), CMP-SAT, and ST3GAL

Further enhanced sialylation of recombinant human EPO

Jeong et al. [208]

CHO (CMP-SAT) and human (ST3GAL) (overexpression)

Mutant uridine diphosphate-N-acetyl glucosamine 2-epimerase (GNE), CMP-SAT, and ST3GAL

>10-Fold increase in CMP-sialic acid concentration leading to a 32% increase in sialylation of human recombinant EPO

Son et al. [209]

Rat (overexpression)

𝛽-1,4 N-acetylglucosaminyltransferase III (GnT-III)

Expression of antibodies with increased bisecting glycan chains that resulted in a 20-fold lower antibody dosage with high ADCC

Davies et al. [210]

Rat (GnT-III) and human (ManII) (overexpression)

GnT-III and Golgi 𝛼-mannosidase II (ManII)

Expression of nonfucosylated antibodies possessing N-glycans of the complex type

Ferrara et al. [211]

Human (overexpression)

𝛽-1,6 N-acetylglucosaminyl-transferase (C2GnT)

Redirection of the O-glycosylation pathway in CHO cells

Prati et al. [212]

CHO (knockout)

𝛼-1,6-fucosyltransferase (FUT8)

Production of completely nonfucosylated antibodies resulting in 100-fold enhanced antibody-dependent cellular cytotoxicity (ADCC)

Yamane-Ohnuki et al. [213]

CHO (knockout)

FUT8

Production of completely nonfucosylated antibodies with enhanced ADCC

Malphettes et al. [214]

CHO (knockout)

FUT8

Genomic knockout of FUT8 with simultaneous integration of an antibody expression cassette

Cristea et al. [215]

CHO (knockout)

N-acetylglucosaminyl-transferase 1 (MGAT1)

Production of recombinant proteins with Man5 as the predominant N-linked glycosylation species

Zhong et al. [216] Sealover et al. [217]

CHO (knockout)

GDP-fucose transporter (SLC35C1) and CMP-sialic acid transporter (SLC35A1)

Production of recombinant antibodies lacking both fucose and sialic acid to increase ADCC

Zhang et al. [218]

CHO (knockout)

MGAT4A, 4B, and 5

Expression of EPO with almost homogeneous biantennary N-glycans with a minor amount of poly-N-acetyl-lactosamine (poly-LacNAc)

Yang et al. [219]

CHO (knockout)

MGAT4A/4B/5 and 𝛽-1,4 galactosyltransferase 1 (B4GALT1)

>90% Reduction in galactosylation of recombinant EPO

Yang et al. [219]

CHO (knockout)

𝛽-1,3 N-acetylglucosaminyltransferase 2 (B3GNT2)

Expression of recombinant EPO lacking poly-LacNAc

Yang et al. [219]

Table 9.3 (Continued)

Cellular pathway

Origin of engineering gene

Target gene

Engineered phenotype

References

CHO (knockout)

ST3GAL4/6 and MGAT4A/4B/5

Expression of recombinant EPO with heterogeneous tetra-antennary N-glycans without sialylation

Yang et al. [219]

CHO (knockout)

FUT8 and B4GALT1

Expression of an IgG1 with homogenous biantennary N-glycans without fucose and almost no galactose

Yang et al. [219]

CHO (knockdown)

Sialidase (NEU2)

60% Decrease in sialidase activity led to increased sialic acid content in IFN𝛾

Ferrari et al. (1998) Ngantung et al. [176]

CHO (knockdown)

Sialidases NEU1 and 3

98% Decrease in sialidase activity led to 26–33% increase in sialic acid content of IFN𝛾

Zhang et al. [182]

CHO (knockdown)

𝛼-1,6-fucosyltransferase (FUT8)

Reduction in core fucose by 60–88% resulted in 100-fold improved ADCC of the produced IgG

Mori et al. [175] Beuger et al. [168]

CHO (knockdown)

GDP-fucose 4,6-dehydratase (GMD)

Production of 100% defucosylated recombinant antibodies if culture medium lacks l-fucose

Kanda et al. [172]

CHO (knockdown)

FUT8 and GMD

Production of fully nonfucosylated antibodies with improved ADCC

Imai-Nishiya et al. [220]

CHO (knockdown)

GDP-fucose transporter (GFT)

10–40% Increase in defucosylated AT-III

Omasa et al. [73]

Posttranslational modifications

CHO (knockout)

Peptidylglycine 𝛼-amidating monooxygenase (PAM)

Reduction in C-terminal amidated species of recombinant monoclonal antibodies

Skulj et al. [221]

Drug product stability

CHO (knockout)

Lipoprotein lipase (LPL)

Removal of LPL host cell protein decreased polysorbate-20 and -80 degradation and increased drug product stability

Chiu et al. [222]

Viral resistance

CHO (knockout)

CMP sialic acid transporter (SLC35A1)

Increased resistance to MVM virus infection

Mascarenhas et al. [223]

9.3 Applications of Cell Line Engineering Approaches in CHO Cells

pattern of recombinant antibodies can lead to severe immunological reactions [169]. Therefore, interfering with genes involved within the glycosylation pathways might be a promising strategy to generate CHO cells that will express recombinant proteins with tailored glycosylation patterns. Genomic knockout of N-acetylglucosaminyltransferase (MGAT1) in CHO cells could be shown to enable production of glycoproteins with Man5 as the predominant N-linked glycosylation species [216, 217]. An impressive example toward the development of novel designer CHO cells capable of producing glycoproteins with predefined glycopatterns was a detailed ZFN-assisted knockout screening of 19 different glycosylation genes in recombinant CHO cells [219]. The authors nicely demonstrated a variety of knockout combinations to achieve a desired glycosylation profile on recombinant proteins by combinatorial genome editing approaches, thereby highlighting future potential of this genetic engineering technology [219]. Further engineering efforts comprised the combinatorial introduction of genes involved in different parts of the N-glycosylation pathway to steadily improve N-glycan structures of recombinant proteins for therapeutic use. These included different N-acetylglucosaminyltransferases (GnT-III, IV, and V), uridine diphosphate-N-acetyl glucosamine 2-epimerase (GNE), CMP-sialic acid synthetase (CMP-SAS), and Golgi α-mannosidase II (ManII) [98, 203, 206–211]. It has been reported that monoclonal antibodies lacking core fucose induce stronger antibody-dependent cell-mediated cytotoxicity (ADCC) and at lower antibody doses [181]. As monoclonal antibodies represent by far the most important class of therapeutic glycoproteins being manufactured, research toward development of strategies to produce low or nonfucosylated Mab glycans has been intensified. Modulation of α-1,6-fucosyltransferase (FUT8), an enzyme that catalyzes the transfer of the core fucose to the Fc part of an antibody, has been addressed by numerous groups. Homologous recombination-mediated stable genomic knockout of FUT8 completely erased FUT8 activity in CHO cells [213]. Similarly, ZFN-based genome editing leads to stable excision of the FUT8 gene from the genome [214] and complete abolishment of fucosylation on the Fc domain of IgG. To demonstrate the functionality of a combined knockout/knockin strategy in CHO cells, a FUT8 knockout was performed simultaneously with the introduction of an antibody expression cassette [215]. Other methods to reduce the fucosylation level of antibody glycans were also examined. Mori et al. reported that an 80% knockdown of FUT8 mRNA by anti-FUT8 siRNA resulted in >60% reduction of mAb fucosylation, leading to a 100-fold improved ADCC [175]. Furthermore, constitutive knockdown of FUT8 resulted in 88% defucosylated anti-IGF-1 receptor antibody without negative effect on productivity [168]. Loss-of-function studies involving other genes of the fucosylation pathway revealed that attenuation of GDP-mannose 4,6-dehydratase (GMD) expression was superior to FUT8 knockdown and yielded entirely nonfucosylated mAbs [172]. Finally, a combined knockdown of both FUT8 and GMD also resulted in fully defucosylated mAbs exhibiting enhanced ADCC [220]. The first report that employed the novel CRISPR/Cas9 tool for precise genome editing technique in CHO cells described the genomic disruption of two genes at a time involved in the O- and N-glycosylation pathways, core 1 β3GalT specific molecular chaperone (COSMC) and FUT8

231

232

9 CHO Cell Engineering for Improved Process Performance and Product Quality

[226]. Subsequently, Grav et al. demonstrated that CRISPR/Cas9 can be used to rapidly generate multiple knockout phenotypes in a single approach, enabling the simultaneous disruption of FUT8, BAX, and BAK without noticeable off-target effects [159]. An alternative approach used ZFNs, TALENs, and the CRISPR-Cas9 to inactivate the GDP-fucose transporter (SLC35C1) in CHO cells resulting antibodies lacking core fucose [227]. Another important PTM is sialylation. Desialylated serum glycoproteins have significantly lower circulatory half-lives as compared to their sialylated counterparts [169]. Thus, improving sialylation is an important factor for the manufacturing of therapeutic proteins. In contrast to human and mouse cells, CHO cells were shown to lack expression of α-2,6-sialyltransferase and only express 𝛼-2,3-sialyltransferase [228]. Consequently, CHO cells inherently cannot produce glycoproteins with similar terminal sialic acid content as compared to human cell lines [229, 230]. Already in 1989, a CHO cell line that stably overexpressed 𝛽 galactoside 𝛼-2,6-sialyltransferase (ST6GAL) was generated, which was able to secrete recombinant proteins additionally harboring 𝛼2,6-sialylated glycan residues [199]. Other groups later confirmed that overexpression of ST6GAL of various species (human, mouse, and rat) could indeed increase sialylation content on N-glycans of recombinant proteins produced in genetically engineered CHO cells [200–202, 204, 205]. Recently, a combinatorial approach of overexpression of 𝛼-2,6 sialyltransferase and disruption of ST3GAL4 and ST3GAL6 genes by CRISPR/Cas9 was performed to minimize the 𝛼-2,3 sialylation. This leads indeed to increased 𝛼-2,6 sialylation level relative to 𝛼-2,3 sialylation [231]. Also, transient approaches have been tested where the expression of ST6Gal1 in CHO and HEK293 513 cells increased the 𝛼-2,6 sialylation of an antibody [232] and transient coexpression of ST6Gal1 and β4GalT1 significantly enhanced antibody sialylation [232]. In a combinatorial approach, an engineered CHO cell line stably expressing the human 𝛼2,3-sialyltransferase (ST3Gal3), the rat-mutated GNE/MNK-R263L-R266Q glycosylation enzymes, and the Chinese hamster CMP-sialic acid transporter was successful in increasing tetra-sialylation of recombinant human erythropoietin (EPO) [209]. The combinatorial knockout of two different key carbohydrate transporters for GDP-fucose (SLC35C1) and CMP-sialic acid (SLC35A1) using ZFN technology yielded an engineered CHO cell line that produces afucosylated and asialylated glycoproteins [218, 233]. Applying the opposite rationale, glycoprotein sialylation levels could be increased by inhibition of sialidases. The knockdown of different CHO sialidases by shRNA or siRNA improved the sialylation level of recombinant proteins, particularly for the sialidases NEU2 and NEU3 [176, 182]. Recently, the co-overexpression of the N-acetylglucosaminyltransferases I (MGAT1) and IV (MGAT4) was reported to improve sialylation of albumin-EPO. The sialic acid content of the recombinant protein was highest in cells with excess MGAT4 gene, and these cells showed a higher tri- and tetra-antennary structure than control cells [234]. Finally, although the modulation of N-linked glycosylation pattern on recombinant proteins has gained major attention during the last decades, modulation of O-glycosylation has also been studied in CHO cells [212]. In addition, a recent study has found some interesting new features of the O-glycoproteome in liver cells [235]. However, it still needs to be elucidated if the findings made

9.4 Conclusions

by Schjoldager et al. can be confirmed in CHO cells as well. Furthermore, as alterations in N-glycosylation still seem to have a more dramatic impact on the therapeutic potential and safety profile of recombinant proteins, further efforts regarding modulation of O-glycosylation will be less expected. The presence of C-terminal amidated species of mAbs produced by CHO cells is another PTM of amino acid residues influencing the variability of monoclonal antibodies and thereby their quality [236–238]. The ZFN-mediated genomic knockout of the peptidylglycine 𝛼-amidating monooxygenase in CHO cells led to a significant reduction in C-terminal amidated mAbs, resulting in a more homogeneous product [221]. Heterogeneity of C-terminal lysine levels is often observed in therapeutic monoclonal antibodies. A recent study showed that the CRISPR/Cas9 knockout of carboxypeptidase D (CpD) completely abolished C-terminal lysine cleavage [239]. In the past, the influence of the production cell line and its cellular entities on drug product stability had rather been a neglected field of investigation. A surprising influence of host cell protein (HCP) content on polysorbate degradation was recently presented by Chiu et al. [222]. In this study, the authors demonstrated that an individual HCP, lipoprotein lipase (LPL), which was completely copurified with the therapeutic protein, led to product instability in the final drug product formulation because of polysorbate degradation [222]. Genomic knockout of LPL in CHO cells using CRISPR and TALENs, however, could markedly improve drug product stability because of the absence of polysorbate degradation of material produced by engineered CHO cell lines [222]. Viral safety is another very important factor for biopharmaceutical process development. In the past, it was shown that viruses such as the parvovirus minute virus of mice (MVM) can successfully infect and replicate in CHO cell cultures and thus represents a vital financial and regulatory threat for CHO-based production processes, as occurrence of any viral contamination can oblige the manufacturer to shut down an entire production facility for months [240]. Consequently, efforts are currently aiming at increasing regulatory safety of CHO manufacturing cell lines by host cell engineering in order to reduce the risk of viral contaminations. As the presence of sialic acids on cell surface receptors play a critical role in viral entry mechanisms of MVM viruses [241, 242], Mascarenhas et al. knocked out a sialic acid transporter (SLC35A1) using ZFN in CHO host cells [223]. Knockout cell lines showed almost entirely asialylated cell surface proteins [223]. As a result, SLC35A1 knockout cells were completely resistant to MVM infection [223]. Of note, however, if sialylation is an important PTM of a therapeutic protein such as EPO, these engineered CHO host cell lines lacking the ability to sialylate were of course not the vital option to be employed for commercial production.

9.4 Conclusions Isolated more than 60 years ago, various sublines of the original CHO cell line have found a solid place both within the biopharmaceutical industry and in academic research laboratories all over the world. There are certainly numerous key factors that have eventually contributed to the great success of CHO cells

233

234

9 CHO Cell Engineering for Improved Process Performance and Product Quality

to become the most frequently employed mammalian expression system for therapeutic proteins. However, the biopharmaceutical industry is still facing tremendous challenges, e.g. (i) an increasing number of complex biological formats that need to be produced at similar amounts as classical monoclonal antibodies, (ii) the requirement for tailored product characteristics (biosimilars), (iii) reductions in health care system budgets, and (iv) an overall elevated number of drug candidates that puts enormous pressure on the global development and manufacturing capacity for biologics. These and additional factors eventually require regular and systematic “updates” of CHO host cell lines. Although we are still far away from holistically understanding the complexity of CHO cells, the abovementioned cell line engineering approaches have valuably contributed to substantial improvements of CHO manufacturing cell lines over the past decades. However, there is still room for improvement! As current CHO cell lines express a plethora of endogenous genes that are presumably not required or even disadvantageous to advanced recombinant protein production and growth in chemically defined media in stirred tank bioreactors, new cell line engineering technologies such as precise genome editing or the use of noncoding RNA-mediated pathway engineering will pave the way for generation of even more advanced CHO cell lines. Advanced host cell lines in combination with other critical cornerstones such as a high-performing vector system, a rapid cell line generation process, an advanced and stable cell culture media platform, as well as a scalable process finally forms the basis for successful and cost-effective biopharmaceutical drug developments. Therefore, consideration of all these elements will be the key toward successful and cost-effective biological drug development in the future.

References 1 Kunert, R. and Reinhart, D. (2016). Advances in recombinant antibody man-

ufacturing. Appl. Microbiol. Biotechnol. 100: 3451–3461. 2 Lalonde, M.E. and Durocher, Y. (2017). Therapeutic glycoprotein production

in mammalian cells. J. Biotechnol. 251: 128–140. 3 Wells, E. and Robinson, A.S. (2017). Cellular engineering for therapeu-

4

5

6 7

tic protein production: product quality, host modification, and process improvement. Biotechnol. J. 12 (1): https://doi.org/10.1002/biot.201600105. Urlaub, G. and Chasin, L.A. (1980). Isolation of Chinese hamster cell mutants deficient in dihydrofolate reductase activity. Proc. Natl. Acad. Sci. U.S.A. 77: 4216–4220. Urlaub, G., Kas, E., Carothers, A.M., and Chasin, L.A. (1983). Deletion of the diploid dihydrofolate reductase locus from cultured mammalian cells. Cell 33: 405–412. Wurm, F.M. and Hacker, D. (2011). First CHO genome. Nat. Biotechnol. 29: 718–720. Cockett, M.I., Bebbington, C.R., and Yarranton, G.T. (1990). High level expression of tissue inhibitor of metalloproteinases in Chinese hamster

References

8

9

10 11

12

13

14

15

16

17

18

19 20

21

22

ovary cells using glutamine synthetase gene amplification. Biotechnology 8: 662–667. Sanders, P.G., Hussein, A., Coggins, L., and Wilson, R. (1987). Gene amplification: the Chinese hamster glutamine synthetase gene. Dev. Biol. Stand. 66: 55–63. Fan, L., Kadura, I., Krebs, L.E. et al. (2012). Improving the efficiency of CHO cell line generation using glutamine synthetase gene knockout cells. Biotechnol. Bioeng. 109: 1007–1015. Shukla, A.A., Wolfe, L.S., Mostafa, S.S., and Norman, C. (2017). Evolving trends in mAb production processes. Bioeng. Transl. Med. 2: 58–69. Jost, C. and Pluckthun, A. (2014). Engineered proteins with desired specificity: DARPins, other alternative scaffolds and bispecific IgGs. Curr. Opin. Struct. Biol. 27: 102–112. Spiess, C., Zhai, Q., and Carter, P.J. (2015). Alternative molecular formats and therapeutic applications for bispecific antibodies. Mol. Immunol. 67: 95–106. Johari, Y.B., Estes, S.D., Alves, C.S. et al. (2015). Integrated cell and process engineering for improved transient production of a “difficult-to-express” fusion protein by CHO cells. Biotechnol. Bioeng. 112: 2527–2542. Lee, G.W., Fecko, J.K., Yen, A. et al. (2007). Improving the expression of a soluble receptor:Fc fusion protein in CHO cells by coexpression with the receptor ligand. Cell Technol. Cell Prod. 29–39. Mathias, S., Fischer, S., Handrick, R. et al. (2018). Visualisation of intracellular production bottlenecks in suspension-adapted CHO cells producing complex biopharmaceuticals using fluorescence microscopy. J. Biotechnol. 271: 47–55. Bentley, K.J., Gewert, R., and Harris, W.J. (1998). Differential efficiency of expression of humanized antibodies in transient transfected mammalian cells. Hybridoma 17: 559–567. Pybus, L.P., Dean, G., West, N.R. et al. (2014). Model-directed engineering of “difficult-to-express” monoclonal antibody production by Chinese hamster ovary cells. Biotechnol. Bioeng. 111: 372–385. Mason, M., Sweeney, B., Cain, K. et al. (2012). Identifying bottlenecks in transient and stable production of recombinant monoclonal-antibody sequence variants in Chinese hamster ovary cells. Biotechnol. Progr. 28: 846–855. Seeliger, D., Schulz, P., Litzenburger, T. et al. (2015). Boosting antibody developability through rational sequence optimization. mAbs 7: 505–515. van der Kant, R., Karow-Zwick, A.R., Van Durme, J. et al. (2017). Prediction and reduction of the aggregation of monoclonal antibodies. J. Mol. Biol. 429: 1244–1261. Tartari, F., Santoni, M., Pistelli, M., and Berardi, R. (2017). Healthcare cost of HER2-positive and negative breast tumors in the United States (2012–2035). Cancer Treat. Rev. 60: 12–17. Makrides, S.C. (1999). Components of vectors for gene transfer and expression in mammalian cells. Protein Expression Purif. 17: 183–202.

235

236

9 CHO Cell Engineering for Improved Process Performance and Product Quality

23 Gopalkrishnan, R.V., Christiansen, K.A., Goldstein, N.I. et al. (1999). Use

24 25

26

27

28 29

30

31

32 33 34 35

36 37 38 39 40

of the human EF-1alpha promoter for expression can significantly increase success in establishing stable cell lines with consistent expression: a study using the tetracycline-inducible system in human cancer cells. Nucleic Acids Res. 27: 4775–4782. Wurm, F.M. (2004). Production of recombinant protein therapeutics in cultivated mammalian cells. Nat. Biotechnol. 22: 1393–1398. Kramer, O., Klausing, S., and Noll, T. (2010). Methods in mammalian cell line engineering: from random mutagenesis to sequence-specific approaches. Appl. Microbiol. Biotechnol. 88: 425–436. Cho, S.W., Kim, S., Kim, J.M., and Kim, J.S. (2013). Targeted genome engineering in human cells with the Cas9 RNA-guided endonuclease. Nat. Biotechnol. 31: 230–232. Galetto, R., Duchateau, P., and Paques, F. (2009). Targeted approaches for gene therapy and the emergence of engineered meganucleases. Expert Opin. Biol. Ther. 9: 1289–1303. Miller, J.C., Tan, S., Qiao, G. et al. (2011). A TALE nuclease architecture for efficient genome editing. Nat. Biotechnol. 29: 143–148. Santiago, Y., Chan, E., Liu, P.Q. et al. (2008). Targeted gene knockout in mammalian cells by using engineered zinc-finger nucleases. Proc. Natl. Acad. Sci. U.S.A. 105: 5809–5814. Lee, J.S., Grav, L.M., Lewis, N.E., and Faustrup Kildegaard, H. (2015). CRISPR/Cas9-mediated genome engineering of CHO cell factories: application and perspectives. Biotechnol. J. 10: 979–994. Fire, A., Xu, S., Montgomery, M.K. et al. (1998). Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 391: 806–811. Hamilton, A.J. and Baulcombe, D.C. (1999). A species of small antisense RNA in posttranscriptional gene silencing in plants. Science 286: 950–952. Bagasra, O. and Prilliman, K.R. (2004). RNA interference: the molecular immune system. J. Mol. Histol. 35: 545–553. Kim, V.N., Han, J., and Siomi, M.C. (2009). Biogenesis of small RNAs in animals. Nat. Rev. Mol. Cell Biol. 10: 126–139. Forstemann, K., Horwich, M.D., Wee, L. et al. (2007). Drosophila microRNAs are sorted into functionally distinct argonaute complexes after production by DICER-1. Cell 130: 287–297. John, B., Enright, A.J., Aravin, A. et al. (2004). Human MicroRNA targets. PLoS Biol. 2: e363. Tomari, Y. and Zamore, P.D. (2005). Perspective: machines for RNAi. Genes Dev. 19: 517–529. Golden, D.E., Gerbasi, V.R., and Sontheimer, E.J. (2008). An inside job for siRNAs. Mol. Cell 31: 309–312. Lippman, Z. and Martienssen, R. (2004). The role of RNA interference in heterochromatic silencing. Nature 431: 364–370. Okamura, K., Balla, S., Martin, R. et al. (2008). Two distinct mechanisms generate endogenous siRNAs from bidirectional transcription in Drosophila melanogaster. Nat. Struct. Mol. Biol. 15: 581–590.

References

41 Werner, A., Cockell, S., Falconer, J. et al. (2014). Contribution of natural

42 43 44 45

46

47

48

49

50

51

52

53

54 55 56

57

antisense transcription to an endogenous siRNA signature in human cells. BMC Genomics 15: 19. Bartel, D.P. (2009). MicroRNAs: target recognition and regulatory functions. Cell 136: 215–233. Gurtan, A.M. and Sharp, P.A. (2013). The role of miRNAs in regulating gene expression networks. J. Mol. Biol. 425 (19): 3582–3600. Barron, N., Sanchez, N., Kelly, P., and Clynes, M. (2011). MicroRNAs: tiny targets for engineering CHO cell phenotypes? Biotechnol. Lett. 33: 11–21. Hackl, M., Borth, N., and Grillari, J. (2012). miRNAs–pathway engineering of CHO cell factories that avoids translational burdening. Trends Biotechnol. 30: 405–406. Fischer, S., Handrick, R., Aschrafi, A., and Otte, K. (2015). Unveiling the principle of microRNA-mediated redundancy in cellular pathway regulation. RNA Biol. 12: 238–247. Fischer, S., Buck, T., Wagner, A. et al. (2014). A functional high-content miRNA screen identifies miR-30 family to boost recombinant protein production in CHO cells. Biotechnol. J. 9: 1279–1292. Gammell, P., Barron, N., Kumar, N., and Clynes, M. (2007). Initial identification of low temperature and culture stage induction of miRNA expression in suspension CHO-K1 cells. J. Biotechnol. 130: 213–218. Hammond, S., Swanberg, J.C., Polson, S.W., and Lee, K.H. (2012). Profiling conserved microRNA expression in recombinant CHO cell lines using Illumina sequencing. Biotechnol. Bioeng. 109: 1371–1375. Maccani, A., Hackl, M., Leitner, C. et al. (2014). Identification of microRNAs specific for high producer CHO cell lines using steady-state cultivation. Appl. Microbiol. Biotechnol. 98 (17): 7535–7548. Lee, R.C., Feinbaum, R.L., and Ambros, V. (1993). The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell 75: 843–854. Ibanez-Ventoso, C., Vora, M., and Driscoll, M. (2008). Sequence relationships among C. elegans, D. melanogaster and human microRNAs highlight the extensive conservation of microRNAs in biology. PLoS One 3: e2818. Pasquinelli, A.E., Reinhart, B.J., Slack, F. et al. (2000). Conservation of the sequence and temporal expression of let-7 heterochronic regulatory RNA. Nature 408: 86–89. Lee, Y., Jeon, K., Lee, J.T. et al. (2002). MicroRNA maturation: stepwise processing and subcellular localization. EMBO J. 21: 4663–4670. Arvey, A., Larsson, E., Sander, C. et al. (2010). Target mRNA abundance dilutes microRNA and siRNA activity. Mol. Syst. Biol. 6: 363. Meleady, P., Gallagher, M., Clarke, C. et al. (2012). Impact of miR-7 over-expression on the proteome of Chinese hamster ovary cells. J. Biotechnol. 160: 251–262. Ventura, A., Young, A.G., Winslow, M.M. et al. (2008). Targeted deletion reveals essential and overlapping functions of the miR-17 through 92 family of miRNA clusters. Cell 132: 875–886.

237

238

9 CHO Cell Engineering for Improved Process Performance and Product Quality

58 Bratkovic, T., Glavan, G., Strukelj, B. et al. (2012). Exploiting microRNAs for

cell engineering and therapy. Biotechnol. Adv. 30: 753–765. 59 Jadhav, V., Hackl, M., Bort, J.A. et al. (2012). A screening method to assess

60

61

62

63 64

65

66

67

68

69

70

71

72

biological effects of microRNA overexpression in Chinese hamster ovary cells. Biotechnol. Bioeng. 109: 1376–1385. Strotbek, M., Florin, L., Koenitzer, J. et al. (2013). Stable microRNA expression enhances therapeutic antibody productivity of Chinese hamster ovary cells. Metab. Eng. 20: 157–166. Klanert, G., Jadhav, V., Chanoumidou, K. et al. (2014). Endogenous microRNA clusters outperform chimeric sequence clusters in Chinese hamster ovary cells. Biotechnol. J. 9: 538–544. Hackl, M., Jakobi, T., Blom, J. et al. (2011). Next-generation sequencing of the Chinese hamster ovary microRNA transcriptome: identification, annotation and profiling of microRNAs as targets for cellular engineering. J. Biotechnol. 153: 62–75. Johnson, K.C., Jacob, N.M., Nissom, P.M. et al. (2011). Conserved microRNAs in Chinese hamster ovary cell lines. Biotechnol. Bioeng. 108: 475–480. Griffiths-Jones, S., Grocock, R.J., van Dongen, S. et al. (2006). miRBase: microRNA sequences, targets and gene nomenclature. Nucleic Acids Res. 34: D140–D144. Kozomara, A. and Griffiths-Jones, S. (2014). miRBase: annotating high confidence microRNAs using deep sequencing data. Nucleic Acids Res. 42: D68–D73. Diendorfer, A.B., Hackl, M., Klanert, G. et al. (2015). Annotation of additional evolutionary conserved microRNAs in CHO cells from updated genomic data. Biotechnol. Bioeng. 112 (7): 1488–1493. Borth, N., Mattanovich, D., Kunert, R., and Katinger, H. (2005). Effect of increased expression of protein disulfide isomerase and heavy chain binding protein on antibody secretion in a recombinant CHO cell line. Biotechnol. Progr. 21: 106–111. Hwang, S.O., Chung, J.Y., and Lee, G.M. (2003). Effect of doxycycline-regulated ERp57 expression on specific thrombopoietin productivity of recombinant CHO cells. Biotechnol Prog. 19 (1): 179–184. Chung, J.Y., Lim, S.W., Hong, Y.J. et al. (2004). Effect of doxycycline-regulated calnexin and calreticulin expression on specific thrombopoietin productivity of recombinant Chinese hamster ovary cells. Biotechnol Bioeng. 85 (5): 539–546. Underhill, M.F., Coley, C., Birch, J.R. et al. (2003). Engineering mRNA translation initiation to enhance transient gene expression in chinese hamster ovary cells. Biotechnol Prog. 19 (1): 121–129. Kwon, R.J., Kim, S.K., Lee, S.I. et al. (2006). Artificial transcription factors increase production of recombinant antibodies in Chinese hamster ovary cells. Biotechnol. Lett. 28: 9–15. Ohya, T., Hayashi, T., Kiyama, E. et al. (2008). Improved production of recombinant human antithrombin III in Chinese hamster ovary cells by ATF4 overexpression. Biotechnol. Bioeng. 100: 317–324.

References

73 Omasa, T., Takami, T., Ohya, T. et al. (2008). Overexpression of GADD34

74

75

76

77

78

79

80

81

82

83

84

85

86

enhances production of recombinant human antithrombin III in Chinese hamster ovary cells. J. Biosci. Bioeng. 106: 568–573. Dreesen, I.A. and Fussenegger, M. (2011). Ectopic expression of human mTOR increases viability, robustness, cell size, proliferation, and antibody production of chinese hamster ovary cells. Biotechnol. Bioeng. 108: 853–866. Tastanova, A., Schulz, A., Folcher, M. et al. (2016). Overexpression of YY1 increases the protein production in mammalian cells. J. Biotechnol. 219: 72–85. Orellana, C.A., Marcellin, E., Gray, P.P., and Nielsen, L.K. (2017). Overexpression of the regulatory subunit of glutamate-cysteine ligase enhances monoclonal antibody production in CHO cells. Biotechnol. Bioeng. 114: 1825–1836. Page, M.J. and Sydenham, M.A. (1991). High level expression of the humanized monoclonal antibody Campath-1H in Chinese hamster ovary cells. Biotechnology (N Y). 9 (1): 64–68. Wu, S.C., Hong, W.W., and Liu, J.H. (2008). Short hairpin RNA targeted to dihydrofolate reductase enhances the immunoglobulin G expression in gene-amplified stable Chinese hamster ovary cells. Vaccine. 26 (38): 4969–4974. https://doi.org/10.1016/j.vaccine.2008.06.081. Romand, S., Jostock, T., Fornaro, M. et al. (2016). Improving expression of recombinant human IGF-1 using IGF-1R knockout CHO cell lines. Biotechnol. Bioeng. 113: 1094–1101. Ritter, A., Nuciforo, S., Schulze, A. et al. (2017). Fam60A plays a role for production stabilities of recombinant CHO cell lines. Biotechnol. Bioeng. 114: 701–704. Ritter, A., Voedisch, B., Wienberg, J. et al. (2016). Deletion of a telomeric region on chromosome 8 correlates with higher productivity and stability of CHO cell lines. Biotechnol. Bioeng. 113: 1084–1093. Pieper, L.A., Strotbek, M., Wenger, T. et al. (2017). ATF6beta-based fine-tuning of the unfolded protein response enhances therapeutic antibody productivity of Chinese hamster ovary cells. Biotechnol. Bioeng. 114: 1310–1318. Pieper, L.A., Strotbek, M., Wenger, T. et al. (2017). Secretory pathway optimization of CHO producer cells by co-engineering of the mitosRNA-1978 target genes CerS2 and Tbc1D20. Metab. Eng. 40: 69–79. Matsuyama, R., Yamano, N., Kawamura, N., and Omasa, T. (2017). Lengthening of high-yield production levels of monoclonal antibody-producing Chinese hamster ovary cells by downregulation of breast cancer 1. J Biosci Bioeng. 123 (3): 382–389. https://doi.org/10.1016/j.jbiosc.2016.09.006. Pendse, G.J. and Bailey, J.E. (1994). Effect of Vitreoscilla hemoglobin expression on growth and specific tissue plasminogen activator productivity in recombinant Chinese hamster ovary cells. Biotechnol. Bioeng. 44: 1367–1370. Park, H., Kim, I.H., Kim, I.Y. et al. (2000). Expression of carbamoyl phosphate synthetase I and ornithine transcarbamoylase genes in Chinese hamster ovary DHFR-cells decreases accumulation of ammonium ion in culture media. J. Biotechnol. 81: 129–140.

239

240

9 CHO Cell Engineering for Improved Process Performance and Product Quality

87 Fogolin, M.B., Wagner, R., Etcheverrigaray, M., and Kratje, R. (2004). Impact

88

89

90

91

92

93

94

95

96

97

98

99

of temperature reduction and expression of yeast pyruvate carboxylase on hGM-CSF-producing CHO cells. J. Biotechnol. 109: 179–191. Kim, S.H. and Lee, G.M. (2007). Down-regulation of lactate dehydrogenase-A by siRNAs for reduced lactic acid formation of Chinese hamster ovary cells producing thrombopoietin. Appl. Microbiol. Biotechnol. 74: 152–159. Wlaschin, K.F. and Hu, W.S. (2007). Engineering cell metabolism for high-density cell culture via manipulation of sugar transport. J. Biotechnol. 131: 168–176. Le, H., Vishwanathan, N., Kantardjieff, A. et al. (2013). Dynamic gene expression for metabolic engineering of mammalian cells in culture. Metab. Eng. 20: 212–220. Chong, W.P., Reddy, S.G., Yusufi, F.N. et al. (2010). Metabolomics-driven approach for the improvement of Chinese hamster ovary cell growth: overexpression of malate dehydrogenase II. J. Biotechnol. 147: 116–121. Tabuchi, H., Sugiyama, T., Tanaka, S., and Tainaka, S. (2010). Overexpression of taurine transporter in Chinese hamster ovary cells can enhance cell viability and product yield, while promoting glutamine consumption. Biotechnol. Bioeng. 107: 998–1003. Tabuchi, H. and Sugiyama, T. (2013). Cooverexpression of alanine aminotransferase 1 in Chinese hamster ovary cells overexpressing taurine transporter further stimulates metabolism and enhances product yield. Biotechnol. Bioeng. 110: 2208–2215. Kim, S.H. and Lee, G.M. (2007). Functional expression of human pyruvate carboxylase for reduced lactic acid formation of Chinese hamster ovary cells (DG44). Appl. Microbiol. Biotechnol. 76: 659–665. Jeong, D., Kim, T.S., Lee, J.W. et al. (2001). Blocking of acidosis-mediated apoptosis by a reduction of lactate dehydrogenase activity through antisense mRNA expression. Biochem. Biophys. Res. Commun. 289: 1141–1149. Doolan, P., Meleady, P., Barron, N. et al. (2010). Microarray and proteomics expression profiling identifies several candidates, including the valosin-containing protein (VCP), involved in regulating high cellular growth rate in production CHO cell lines. Biotechnol. Bioeng. 106: 42–56. Zhou, M., Crawford, Y., Ng, D. et al. (2011). Decreasing lactate level and increasing antibody production in Chinese Hamster Ovary cells (CHO) by reducing the expression of lactate dehydrogenase and pyruvate dehydrogenase kinases. J. Biotechnol. 153: 27–34. Jeong, Y.T., Choi, O., Lim, H.R. et al. (2008). Enhanced sialylation of recombinant erythropoietin in CHO cells by human glycosyltransferase expression. J. Microbiol. Biotechnol. 18: 1945–1952. Tigges, M. and Fussenegger, M. (2006). Xbp1-based engineering of secretory capacity enhances the productivity of Chinese hamster ovary cells. Metab. Eng. 8: 264–272.

References

100 Becker, E., Florin, L., Pfizenmaier, K., and Kaufmann, H. (2008). An XBP-1

101

102

103

104

105

106

107

108

109

110

111

112

113

dependent bottle-neck in production of IgG subtype antibodies in chemically defined serum-free Chinese hamster ovary (CHO) fed-batch processes. J Biotechnol. 135 (2): 217–223. https://doi.org/10.1016/j.jbiotec.2008.03.008. Ku, S.C., Ng, D.T., Yap, M.G., and Chao, S.H. (2008). Effects of overexpression of X-box binding protein 1 on recombinant protein production in Chinese hamster ovary and NS0 myeloma cells. Biotechnol. Bioeng. 99: 155–164. Gulis, G., Simi, K.C., de Toledo, R.R. et al. (2014). Optimization of heterologous protein production in Chinese hamster ovary cells under overexpression of spliced form of human X-box binding protein. BMC Biotechnol. 14: 26. https://doi.org/10.1186/1472-6750-14-26. Peng, R.W. and Fussenegger, M. (2009). Molecular engineering of exocytic vesicle traffic enhances the productivity of Chinese hamster ovary cells. Biotechnol. Bioeng. 102: 1170–1181. Florin, L., Pegel, A., Becker, E. et al. (2009). Heterologous expression of the lipid transfer protein CERT increases therapeutic protein productivity of mammalian cells. J. Biotechnol. 141: 84–90. Rahimpour, A., Vaziri, B., Moazzami, R. et al. (2013). Engineering the cellular protein secretory pathway for enhancement of recombinant tissue plasminogen activator expression in Chinese hamster ovary cells: effects of CERT and XBP1s genes. J. Microbiol. Biotechnol. 23: 1116–1122. Peng, R.W., Abellan, E., and Fussenegger, M. (2011). Differential effect of exocytic SNAREs on the production of recombinant proteins in mammalian cells. Biotechnol. Bioeng. 108: 611–620. Le Fourn, V., Girod, P.A., Buceta, M. et al. (2014). CHO cell engineering to prevent polypeptide aggregation and improve therapeutic protein secretion. Metab. Eng. 21: 91–102. Fussenegger, M., Schlatter, S., Datwyler, D. et al. (1998). Controlled proliferation by multigene metabolic engineering enhances the productivity of Chinese hamster ovary cells. Nat. Biotechnol. 16: 468–472. Astley, K., Naciri, M., Racher, A., and Al-Rubeai, M. (2007). The role of p21cip1 in adaptation of CHO cells to suspension and protein-free culture. J Biotechnol. 130 (3): 282–290. Mazur, X., Fussenegger, M., Renner, W.A., and Bailey, J.E. (1998). Higher productivity of growth-arrested Chinese hamster ovary cells expressing the cyclin-dependent kinase inhibitor p27. Biotechnol. Progr. 14: 705–713. Bi, J.X., Shuttleworth, J., and Al-Rubeai, M. (2004). Uncoupling of cell growth and proliferation results in enhancement of productivity in p21CIP1 -arrested CHO cells. Biotechnol. Bioeng. 85: 741–749. Kuystermans, D. and Al-Rubeai, M. (2009). cMyc increases cell number through uncoupling of cell division from cell size in CHO cells. BMC Biotechnol. 9: 76. Lee, J.S., Ha, T.K., Park, J.H., and Lee, G.M. (2013). Anti-cell death engineering of CHO cells: co-overexpression of Bcl-2 for apoptosis inhibition, Beclin-1 for autophagy induction. Biotechnol Bioeng. 110 (8): 2195–2207. https://doi.org/10.1002/bit.24879.

241

242

9 CHO Cell Engineering for Improved Process Performance and Product Quality

114 Sanchez, N., Kelly, P., Gallagher, C. et al. (2013). CHO cell culture longevity

115

116

117

118

119

120

121

122

123

124

125

126

127

and recombinant protein yield are enhanced by depletion of miR-7 activity via sponge decoy vectors. Biotechnol. J. 9 (3): 396–404. Fischer, S., Paul, A., Wagner, A. et al. (2015). miR-2861 as novel HDAC5 inhibitor in CHO cells enhances productivity while maintaining product quality. Biotechnol. Bioeng. 112 (10): 2142–2153. Clarke, C., Henry, M., Doolan, P., et al. (2012). Integrated miRNA, mRNA and proteinexpression analysis reveals the role of post-transcriptional regulation incontrolling CHO cell growth rate. BMCGenomics. 13:656. doi:10 .1186/1471-2164-13--656. Jadhav, V., Hackl, M., Klanert, G. et al. (2014). Stable overexpression of miR-17 enhances recombinant protein production of CHO cells. J. Biotechnol. 175: 38–44. Loh, W.P., Loo, B., Zhou, L. et al. (2014). Overexpression of microRNAs enhances recombinant protein production in Chinese hamster ovary cells. Biotechnol. J. 9 (9): 1140–1151. Loh, W.P., Yang, Y., and Lam, K.P. (2017). miR-92a enhances recombinant protein productivity in CHO cells by increasing intracellular cholesterol levels. Biotechnol. J. 12 (4): https://doi.org/10.1002/biot.201600488. Kelly, P.S., Gallagher, C., Clynes, M., and Barron, N. (2015). Conserved microRNA function as a basis for Chinese hamster ovary cell engineering. Biotechnol. Lett. 37 (4): 787–798. Emmerling, V.V., Fischer, S., Stiefel, F. et al. (2016). Temperature-sensitive miR-483 is a conserved regulator of recombinant protein and viral vector production in mammalian cells. Biotechnol. Bioeng. 113 (4): 830–841. Kelly, P.S., Breen, L., Gallagher, C. et al. (2015). Re-programming CHO cell metabolism using miR-23 tips the balance towards a highly productive phenotype. Biotechnol. J. 10: 1029–1040. Schoellhorn, M., Fischer, S., Wagner, A. et al. (2017). miR-143 targets MAPK7 in CHO cells and induces a hyperproductive phenotype to enhance production of difficult-to-express proteins. Biotechnol. Progr. 33 (4): 1046–1058. Schoellhorn, M., Fischer, S., Wagner, A. et al. (2017). miR-143 targets MAPK7 in CHO cells and induces a hyperproductive phenotype to enhance production of difficult-to-express proteins. Biotechnol Prog. 33 (4): 1046–1058. https://doi.org/10.1002/btpr.2475. Neermann, J. and Wagner, R. (1996). Comparative analysis of glucose and glutamine metabolism in transformed mammalian cell lines, insect and primary liver cells. J. Cell. Physiol. 166: 152–169. Yip, S.S., Zhou, M., Joly, J. et al. (2014). Complete knockout of the lactate dehydrogenase A gene is lethal in pyruvate dehydrogenase kinase 1, 2, 3 down-regulated CHO cells. Mol. Biotechnol. 56 (9): 833–838. Kim, J.Y., Kim, Y.G., Han, Y.K. et al. (2011). Proteomic understanding of intracellular responses of recombinant Chinese hamster ovary cells cultivated in serum-free medium supplemented with hydrolysates. Appl. Microbiol. Biotechnol. 89: 1917–1928.

References

128 Carlage, T., Hincapie, M., Zang, L. et al. (2009). Proteomic profiling of

129

130

131

132

133

134

135

136

137

138

139

140

141

142

a high-producing Chinese hamster ovary cell culture. Anal. Chem. 81: 7357–7362. Charaniya, S., Karypis, G., and Hu, W.S. (2009). Mining transcriptome data for function-trait relationship of hyper productivity of recombinant antibody. Biotechnol. Bioeng. 102: 1654–1669. Fomina-Yadlin, D., Mujacic, M., Maggiora, K. et al. (2015). Transcriptome analysis of a CHO cell line expressing a recombinant therapeutic protein treated with inducers of protein expression. J. Biotechnol. 212: 106–115. Kantardjieff, A., Jacob, N.M., Yee, J.C. et al. (2010). Transcriptome and proteome analysis of Chinese hamster ovary cells under low temperature and butyrate treatment. J. Biotechnol. 145: 143–159. Yee, J.C., Gerdtzen, Z.P., and Hu, W.S. (2009). Comparative transcriptome analysis to unveil genes affecting recombinant protein productivity in mammalian cells. Biotechnol. Bioeng. 102: 246–263. Hayduk, E.J. and Lee, K.H. (2005). Cytochalasin D can improve heterologous protein productivity in adherent Chinese hamster ovary cells. Biotechnol. Bioeng. 90: 354–364. Hammond, S. and Lee, K.H. (2012). RNA interference of cofilin in Chinese hamster ovary cells improves recombinant protein productivity. Biotechnol. Bioeng. 109: 528–535. Ritter, A., Rauschert, T., Oertli, M. et al. (2016). Disruption of the gene C12orf35 leads to increased productivities in recombinant CHO cell lines. Biotechnol. Bioeng. 113: 2433–2442. Barron, N., Kumar, N., Sanchez, N. et al. (2011). Engineering CHO cell growth and recombinant protein productivity by overexpression of miR-7. J. Biotechnol. 151: 204–211. Fischer, S., Handrick, R., and Otte, K. (2015). The art of CHO cell engineering: a comprehensive retrospect and future perspectives. Biotechnol. Adv. 33: 1878–1896. Al-Rubeai, M., Singh, R.P., Goldman, M.H., and Emery, A.N. (1995). Death mechanisms of animal cells in conditions of intensive agitation. Biotechnol. Bioeng. 45: 463–472. Rangan, S., Kamal, S., Konorov, S.O. et al. (2018). Types of cell death and apoptotic stages in Chinese Hamster Ovary cells distinguished by Raman spectroscopy. Biotechnol. Bioeng. 115: 401–412. Tey, B.T., Singh, R.P., Piredda, L. et al. (2000). Influence of BCL-2 on cell death during the cultivation of a Chinese hamster ovary cell line expressing a chimeric antibody. Biotechnol. Bioeng. 68: 31–43. Mastrangelo, A.J., Hardwick, J.M., Bex, F., and Betenbaugh, M.J. (2000). Part I. BCL-2 and BCL-XL limit apoptosis upon infection with alphavirus vectors. Biotechnol. Bioeng. 67: 544–554. Meents, H., Enenkel, B., Eppenberger, H.M. et al. (2002). Impact of coexpression and coamplification of sICAM and antiapoptosis determinants bcl-2/bcl-x(L) on productivity, cell survival, and mitochondria number in CHO-DG44 grown in suspension and serum-free media. Biotechnol Bioeng. 80 (6): 706–716.

243

244

9 CHO Cell Engineering for Improved Process Performance and Product Quality

143 Sauerwald, T.M., Betenbaugh, M.J., and Oyler, G.A. (2002). Inhibiting

144

145

146

147

148

149

150

151

152

153

154

155

156

apoptosis in mammalian cell culture using the caspase inhibitor XIAP and deletion mutants. Biotechnol. Bioeng. 77: 704–716. Figueroa, B. Jr., Chen, S., Oyler, G.A. et al. (2004). AVEN and BCL-XL enhance protection against apoptosis for mammalian cells exposed to various culture conditions. Biotechnol. Bioeng. 85: 589–600. Chiang, G.G. and Sisk, W.P. (2005). Bcl-x(L) mediates increased production of humanized monoclonal antibodies in Chinese hamster ovary cells. Biotechnol Bioeng. 91 (7): 779–792. Wong, D.C., Wong, K.T., Nissom, P.M. et al. (2006). Targeting early apoptotic genes in batch and fed-batch CHO cell cultures. Biotechnol. Bioeng. 95: 350–361. Choi, S.S., Rhee, W.J., Kim, E.J., and Park, T.H. (2006). Enhancement of recombinant protein production in Chinese hamster ovary cells through anti-apoptosis engineering using 30Kc6 gene. Biotechnol Bioeng. 95 (3): 459–467. Crea, F., Sarti, D., Falciani, F., and Al-Rubeai, M. (2006). Over-expression of hTERT in CHO K1 results in decreased apoptosis and reduced serum dependency. J Biotechnol. 121 (2): 109–123. Figueroa, B. Jr., Ailor, E., Osborne, D. 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. Biotechnol Bioeng. 97 (4): 877–892. Arden, N., Majors, B.S., Ahn, S.H. et al. (2007). Inhibiting the apoptosis pathway using MDM2 in mammalian cell cultures. Biotechnol Bioeng. 97 (3): 601–614. Majors, B.S., Arden, N., Oyler, G.A. et al. (2008). E2F-1 overexpression increases viable cell density in batch cultures of Chinese hamster ovary cells. J Biotechnol. 138 (3–4): 103–106. Dorai, H., Kyung, Y.S., Ellis, D. et al. (2009). Expression of anti-apoptosis genes alters lactate metabolism of Chinese Hamster Ovary cells in culture. Biotechnol Bioeng. 103 (3): 592–608. https://doi.org/10.1002/bit.22269. Lee, Y.Y., Wong, K.T., Tan, J. et al. (2009). Overexpression of heat shock proteins (HSPs) in CHO cells for extended culture viability and improved recombinant protein production. J Biotechnol. 143 (1): 34–43. https://doi.org/ 10.1016/j.jbiotec.2009.05.013. Majors, B.S., Betenbaugh, M.J., Pederson, N.E., and Chiang, G.G. (2009). Mcl-1 overexpression leads to higher viabilities and increased production of humanized monoclonal antibody in Chinese hamster ovary cells. Biotechnol. Progr. 25: 1161–1168. Hwang, S.O. and Lee, G.M. (2009). Effect of Akt overexpression on programmed cell death in antibody-producing Chinese hamster ovary cells. J. Biotechnol. 139: 89–94. Majors, B.S., Chiang, G.G., Pederson, N.E., and Betenbaugh, M.J. (2012). Directed evolution of mammalian anti-apoptosis proteins by somatic hypermutation. Protein Eng. Des. Sel. 25: 27–38.

References

157 Tan, J.G., Lee, Y.Y., Wang, T. et al. (2015). Heat shock protein 27 overex-

158

159

160

161

162

163

164 165

166 167 168

169

170

171

172

pression in CHO cells modulates apoptosis pathways and delays activation of caspases to improve recombinant monoclonal antibody titre in fed-batch bioreactors. Biotechnol. J. 10: 790–800. Cost, G.J., Freyvert, Y., Vafiadis, A. et al. (2010). BAK and BAX deletion using zinc-finger nucleases yields apoptosis-resistant CHO cells. Biotechnol. Bioeng. 105: 330–340. Grav, L.M., Lee, J.S., Gerling, S. et al. (2015). One-step generation of triple knockout CHO cell lines using CRISPR/Cas9 and fluorescent enrichment. Biotechnol. J. 10: 1446–1456. Sung, Y.H., Lee, J.S., Park, S.H. et al. (2007). Influence of co-down-regulation of caspase-3 and caspase-7 by siRNAs on sodium butyrate-induced apoptotic cell death of Chinese hamster ovary cells producing thrombopoietin. Metab. Eng. 9: 452–464. Lim, S.F., Chuan, K.H., Liu, S. et al. (2006). RNAi suppression of Bax and Bak enhances viability in fed-batch cultures of CHO cells. Metab. Eng. 8: 509–522. Lee, J.S., Ha, T.K., Park, J.H., and Lee, G.M. (2013). Anti-cell death engineering of CHO cells: co-overexpression of BCL-2 for apoptosis inhibition, Beclin-1 for autophagy induction. Biotechnol. Bioeng. 110: 2195–2207. Jaluria, P., Betenbaugh, M., Konstantopoulos, K., and Shiloach, J. (2007). Enhancement of cell proliferation in various mammalian cell lines by gene insertion of a cyclin-dependent kinase homolog. BMC Biotechnol. 7: 71. Blagosklonny, M.V. (2005). Molecular theory of cancer. Cancer Biol. Ther. 4: 621–627. Kaneko, Y., Sato, R., and Aoyagi, H. (2010). Changes in the quality of antibodies produced by Chinese hamster ovary cells during the death phase of cell culture. J. Biosci. Bioeng. 109: 281–287. Arden, N. and Betenbaugh, M.J. (2004). Life and death in mammalian cell culture: strategies for apoptosis inhibition. Trends Biotechnol. 22: 174–180. Simon, L. and Karim, M.N. (2002). Control of starvation-induced apoptosis in Chinese hamster ovary cell cultures. Biotechnol. Bioeng. 78: 645–657. Beuger, V., Kunkele, K.P., Koll, H. et al. (2009). Short-hairpin-RNA-mediated silencing of fucosyltransferase 8 in Chinese-hamster ovary cells for the production of antibodies with enhanced antibody immune effector function. Biotechnol. Appl. Biochem. 53: 31–37. Ghaderi, D., Zhang, M., Hurtado-Ziola, N., and Varki, A. (2012). Production platforms for biotherapeutic glycoproteins. Occurrence, impact, and challenges of non-human sialylation. Biotechnol. Genet. Eng. Rev. 28: 147–175. Hong, W.W. and Wu, S.C. (2007). A novel RNA silencing vector to improve antigen expression and stability in Chinese hamster ovary cells. Vaccine 25: 4103–4111. Irani, N., Wirth, M., van Den Heuvel, J., and Wagner, R. (1999). Improvement of the primary metabolism of cell cultures by introducing a new cytoplasmic pyruvate carboxylase reaction. Biotechnol. Bioeng. 66: 238–246. Kanda, Y., Imai-Nishiya, H., Kuni-Kamochi, R. et al. (2007). Establishment of a GDP-mannose 4,6-dehydratase (GMD) knockout host cell line: a new

245

246

9 CHO Cell Engineering for Improved Process Performance and Product Quality

173 174

175

176

177 178

179

180

181 182

183

184

185 186

187

188

strategy for generating completely non-fucosylated recombinant therapeutics. J. Biotechnol. 130: 300–310. Kellems, R.E. (1991). Gene amplification in mammalian cells: strategies for protein production. Curr. Opin. Biotechnol. 2: 723–729. Koterba, K.L., Borgschulte, T., and Laird, M.W. (2012). Thioredoxin 1 is responsible for antibody disulfide reduction in CHO cell culture. J. Biotechnol. 157: 261–267. Mori, K., Kuni-Kamochi, R., Yamane-Ohnuki, N. et al. (2004). Engineering Chinese hamster ovary cells to maximize effector function of produced antibodies using FUT8 siRNA. Biotechnol. Bioeng. 88: 901–908. Ngantung, F.A., Miller, P.G., Brushett, F.R. et al. (2006). RNA interference of sialidase improves glycoprotein sialic acid content consistency. Biotechnol. Bioeng. 95: 106–119. Ohtsubo, K. and Marth, J.D. (2006). Glycosylation in cellular mechanisms of health and disease. Cell 126: 855–867. Pallavicini, M.G., DeTeresa, P.S., Rosette, C. et al. (1990). Effects of methotrexate on transfected DNA stability in mammalian cells. Mol. Cell. Biol. 10: 401–404. Patel, M.S. and Korotchkina, L.G. (2001). Regulation of mammalian pyruvate dehydrogenase complex by phosphorylation: complexity of multiple phosphorylation sites and kinases. Exp. Mol. Med. 33: 191–197. Wu, S.-C. (2009). RNA interference technology to improve recombinant protein production in Chinese hamster ovary cells. Biotechnol. Adv. 27: 417–422. Yamane-Ohnuki, N. and Satoh, M. (2009). Production of therapeutic antibodies with controlled fucosylation. MAbs 1: 230–236. Zhang, M., Koskie, K., Ross, J.S. et al. (2010). Enhancing glycoprotein sialylation by targeted gene silencing in mammalian cells. Biotechnol. Bioeng. 105: 1094–1105. Matsuyama, R., Yamano, N., Kawamura, N., and Omasa, T. (2017). Lengthening of high-yield production levels of monoclonal antibody-producing Chinese hamster ovary cells by downregulation of breast cancer 1. J. Biosci. Bioeng. 123: 382–389. Kim, J.Y., Kim, Y.G., and Lee, G.M. (2012). CHO cells in biotechnology for production of recombinant proteins: current state and further potential. Appl. Microbiol. Biotechnol. 93: 917–930. Levine, B. (2005). Eating oneself and uninvited guests: autophagy-related pathways in cellular defense. Cell. 120 (2): 159–162. Lee, J.S. and Lee, G.M. (2012). Estimation of autophagy pathway genes for autophagy induction: overexpression of Atg9A does not induce autophagy in recombinant Chinese hamster ovary cells. Biochem. Eng. J. 68: 221–226. Astley, K., Naciri, M., Racher, A., and Al-Rubeai, M. (2007). The role of p21CIP1 in adaptation of CHO cells to suspension and protein-free culture. J. Biotechnol. 130: 282–290. Walworth, N.C. (2000). Cell-cycle checkpoint kinases: checking in on the cell cycle. Curr. Opin. Cell Biol. 12: 697–704.

References

189 Lee, K.H., Onitsuka, M., Honda, K. et al. (2013). Rapid construction of

190 191

192

193

194 195

196

197 198

199

200

201

202

203

transgene-amplified CHO cell lines by cell cycle checkpoint engineering. Appl. Microbiol. Biotechnol. 97: 5731–5741. Muller, D., Katinger, H., and Grillari, J. (2008). MicroRNAs as targets for engineering of CHO cell factories. Trends Biotechnol. 26: 359–365. Druz, A., Betenbaugh, M., and Shiloach, J. (2012). Glucose depletion activates mmu-miR-466h-5p expression through oxidative stress and inhibition of histone deacetylation. Nucleic Acids Res. 40: 7291–7302. Druz, A., Chu, C., Majors, B. et al. (2011). A novel microRNA mmu-miR-466h affects apoptosis regulation in mammalian cells. Biotechnol. Bioeng. 108: 1651–1661. Druz, A., Son, Y.J., Betenbaugh, M., and Shiloach, J. (2013). Stable inhibition of mmu-miR-466h-5p improves apoptosis resistance and protein production in CHO cells. Metab. Eng. 16: 87–94. Cho, W.C. (2007). OncomiRs: the discovery and progress of microRNAs in cancers. Mol Cancer. 6: 60. Raver-Shapira, N., Marciano, E., Meiri, E. et al. (2007). Transcriptional activation of miR-34a contributes to p53-mediated apoptosis. Mol Cell. 26 (5): 731–743. Butler, M. and Spearman, M. (2014). The choice of mammalian cell host and possibilities for glycosylation engineering. Curr. Opin. Biotechnol. 30: 107–112. Hossler, P., Khattak, S.F., and Li, Z.J. (2009). Optimal and consistent protein glycosylation in mammalian cell culture. Glycobiology 19: 936–949. Wang, Q., Stuczynski, M., Gao, Y., and Betenbaugh, M.J. (2015). Strategies for engineering protein N-glycosylation pathways in mammalian cells. Methods Mol. Biol. 1321: 287–305. Lee, E.U., Roth, J., and Paulson, J.C. (1989). Alteration of terminal glycosylation sequences on N-linked oligosaccharides of Chinese hamster ovary cells by expression of beta-galactoside alpha 2,6-sialyltransferase. J. Biol. Chem. 264: 13848–13855. Zhang, X., Lok, S.H., and Kon, O.L. (1998). Stable expression of human alpha-2,6-sialyltransferase in Chinese hamster ovary cells: functional consequences for human erythropoietin expression and bioactivity. Biochim. Biophys. Acta 1425: 441–452. Bragonzi, A., Distefano, G., Buckberry, L.D. 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: 273–282. Minch, S.L., Kallio, P.T., and Bailey, J.E. (1995). Tissue plasminogen activator coexpressed in Chinese hamster ovary cells with alpha(2,6)-sialyltransferase contains NeuAc alpha(2,6)Gal beta(1,4)Glc-N-AcR linkages. Biotechnol. Progr. 11: 348–351. Fukuta, K., Abe, R., Yokomatsu, T. et al. (2000). Remodeling of sugar chain structures of human interferon-gamma. Glycobiology 10: 421–430.

247

248

9 CHO Cell Engineering for Improved Process Performance and Product Quality

204 Fukuta, K., Yokomatsu, T., Abe, R. et al. (2000). Genetic engineering of CHO

205

206

207

208

209

210

211

212

213

214

215

216

cells producing human interferon-gamma by transfection of sialyltransferases. Glycoconjugate J. 17: 895–904. Jassal, R., Jenkins, N., Charlwood, J. et al. (2001). Sialylation of human IgG-Fc carbohydrate by transfected rat alpha2,6-sialyltransferase. Biochem. Biophys. Res. Commun. 286: 243–249. Weikert, S., Papac, D., Briggs, J. et al. (1999). Engineering Chinese hamster ovary cells to maximize sialic acid content of recombinant glycoproteins. Nat. Biotechnol. 17: 1116–1121. Wong, N.S., Yap, M.G., and Wang, D.I. (2006). Enhancing recombinant glycoprotein sialylation through CMP-sialic acid transporter over expression in Chinese hamster ovary cells. Biotechnol. Bioeng. 93: 1005–1016. Jeong, Y.T., Choi, O., Son, Y.D. et al. (2009). Enhanced sialylation of recombinant erythropoietin in genetically engineered Chinese-hamster ovary cells. Biotechnol. Appl. Biochem. 52: 283–291. Son, Y.D., Jeong, Y.T., Park, S.Y., and Kim, J.H. (2011). Enhanced sialylation of recombinant human erythropoietin in Chinese hamster ovary cells by combinatorial engineering of selected genes. Glycobiology 21: 1019–1028. Davies, J., Jiang, L., Pan, L.Z. et al. (2001). Expression of GnTIII in a recombinant anti-CD20 CHO production cell line: expression of antibodies with altered glycoforms leads to an increase in ADCC through higher affinity for FC gamma RIII. Biotechnol. Bioeng. 74: 288–294. Ferrara, C., Brunker, P., Suter, T. 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: 851–861. Prati, E.G., Matasci, M., Suter, T.B. et al. (2000). Engineering of coordinated up- and down-regulation of two glycosyltransferases of the O-glycosylation pathway in Chinese hamster ovary (CHO) cells. Biotechnol. Bioeng. 68: 239–244. Yamane-Ohnuki, N., Kinoshita, S., Inoue-Urakubo, M. 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. Biotechnol. Bioeng. 87: 614–622. Malphettes, L., Freyvert, Y., Chang, J. et al. (2010). Highly efficient deletion of FUT8 in CHO cell lines using zinc-finger nucleases yields cells that produce completely nonfucosylated antibodies. Biotechnol. Bioeng. 106: 774–783. Cristea, S., Freyvert, Y., Santiago, Y. et al. (2013). In vivo cleavage of transgene donors promotes nuclease-mediated targeted integration. Biotechnol. Bioeng. 110: 871–880. Zhong, X., Cooley, C., Seth, N. et al. (2012). Engineering novel Lec1 glycosylation mutants in CHO-DUKX cells: molecular insights and effector modulation of N-acetylglucosaminyltransferase I. Biotechnol. Bioeng. 109: 1723–1734.

References

217 Sealover, N.R., Davis, A.M., Brooks, J.K. et al. (2013). Engineering Chinese

218

219

220

221

222

223

224

225

226

227

228

229

hamster ovary (CHO) cells for producing recombinant proteins with simple glycoforms by zinc-finger nuclease (ZFN)-mediated gene knockout of mannosyl (alpha-1,3-)-glycoprotein beta-1,2-N-acetylglucosaminyltransferase (Mgat1). J. Biotechnol. 167: 24–32. Zhang, P., Haryadi, R., Chan, K.F. et al. (2012). Identification of functional elements of the GDP-fucose transporter SLC35C1 using a novel Chinese hamster ovary mutant. Glycobiology 22: 897–911. Yang, Z., Wang, S., Halim, A. et al. (2015). Engineered CHO cells for production of diverse, homogeneous glycoproteins. Nat. Biotechnol. 33: 842–844. Imai-Nishiya, H., Mori, K., Inoue, M. et al. (2007). Double knockdown of alpha1,6-fucosyltransferase (FUT8) and GDP-mannose 4,6-dehydratase (GMD) in antibody-producing cells: a new strategy for generating fully non-fucosylated therapeutic antibodies with enhanced ADCC. BMC Biotechnol. 7: 84. Skulj, M., Pezdirec, D., Gaser, D. et al. (2014). Reduction in C-terminal amidated species of recombinant monoclonal antibodies by genetic modification of CHO cells. BMC Biotechnol. 14: 76. Chiu, J., Valente, K.N., Levy, N.E. et al. (2017). Knockout of a difficult-to-remove CHO host cell protein, lipoprotein lipase, for improved polysorbate stability in monoclonal antibody formulations. Biotechnol. Bioeng. 114: 1006–1015. Mascarenhas, J.X., Korokhov, N., Burger, L. et al. (2017). Genetic engineering of CHO cells for viral resistance to minute virus of mice. Biotechnol. Bioeng. 114: 576–588. Elliott, S., Lorenzini, T., Asher, S. et al. (2003). Enhancement of therapeutic protein in vivo activities through glycoengineering. Nat. Biotechnol. 21: 414–421. Sinclair, A.M. and Elliott, S. (2005). Glycoengineering: the effect of glycosylation on the properties of therapeutic proteins. J. Pharm. Sci. 94: 1626–1635. Ronda, C., Pedersen, L.E., Hansen, H.G. et al. (2014). Accelerating genome editing in CHO cells using CRISPR Cas9 and CRISPy, a web-based target finding tool. Biotechnol. Bioeng. 111: 1604–1616. Chan, K.F., Shahreel, W., Wan, C. et al. (2016). Inactivation of GDP-fucose transporter gene (SLC35C1) in CHO cells by ZFNs, TALENs and CRISPR-Cas9 for production of fucose-free antibodies. Biotechnol. J. 11: 399–414. Jenkins, N., Parekh, R.B., and James, D.C. (1996). Getting the glycosylation right: implications for the biotechnology industry. Nat. Biotechnol. 14: 975–981. Bork, K., Horstkorte, R., and Weidemann, W. (2009). Increasing the sialylation of therapeutic glycoproteins: the potential of the sialic acid biosynthetic pathway. J. Pharm. Sci. 98: 3499–3508.

249

250

9 CHO Cell Engineering for Improved Process Performance and Product Quality

230 Butler, M. (2005). Animal cell cultures: recent achievements and perspectives

231

232 233

234

235

236

237

238

239

240

241 242

in the production of biopharmaceuticals. Appl. Microbiol. Biotechnol. 68: 283–291. Chung, C.Y., Wang, Q., Yang, S. et al. (2017). Integrated genome and protein editing swaps alpha-2,6 sialylation for alpha-2,3 sialic acid on recombinant antibodies from CHO. Biotechnol. J. 12 (2): https://doi.org/10.1002/biot .201600502. Raymond, C., Robotham, A., Spearman, M. et al. (2015). Production of alpha2,6-sialylated IgG1 in CHO cells. mAbs 7: 571–583. Haryadi, R., Zhang, P., Chan, K.F., and Song, Z. (2013). CHO-gmt5, a novel CHO glycosylation mutant for producing afucosylated and asialylated recombinant antibodies. Bioengineered 4: 90–94. Cha, H.M., Lim, J.H., Yeon, J.H. et al. (2017). Co-overexpression of MGAT1 and MGAT4 in CHO cells for production of highly sialylated albumin-erythropoietin. Enzyme Microb. Technol. 103: 53–58. Schjoldager, K.T., Joshi, H.J., Kong, Y. et al. (2015). Deconstruction of O-glycosylation–GalNAc-T isoforms direct distinct subsets of the O-glycoproteome. EMBO Rep. 16: 1713–1722. Gramer, M.J. (2014). Product quality considerations for mammalian cell culture process development and manufacturing. Adv. Biochem. Eng. Biotechnol. 139: 123–166. Johnson, K.A., Paisley-Flango, K., Tangarone, B.S. et al. (2007). Cation exchange-HPLC and mass spectrometry reveal C-terminal amidation of an IgG1 heavy chain. Anal. Biochem. 360: 75–83. Tsubaki, M., Terashima, I., Kamata, K., and Koga, A. (2013). C-terminal modification of monoclonal antibody drugs: amidated species as a general product-related substance. Int. J. Biol. Macromol. 52: 139–147. Hu, Z., Zhang, H., Haley, B. et al. (2016). Carboxypeptidase D is the only enzyme responsible for antibody C-terminal lysine cleavage in Chinese hamster ovary (CHO) cells. Biotechnol. Bioeng. 113: 2100–2106. Moody, M., Alves, W., Varghese, J., and Khan, F. (2011). Mouse minute virus (MMV) contamination–a case study: detection, root cause determination, and corrective actions. PDA J. Pharm. Sci. Technol. 65: 580–588. Cotmore, S.F. and Tattersall, P. (2007). Parvoviral host range and cell entry mechanisms. Adv. Virus Res. 70: 183–232. Lopez-Bueno, A., Rubio, M.P., Bryant, N. et al. (2006). Host-selected amino acid changes at the sialic acid binding pocket of the parvovirus capsid modulate cell binding affinity and determine virulence. J. Virol. 80: 1563–1573.

251

10 Metabolite Profiling of Mammalian Cells Claire E. Gaffney, Alan J. Dickson, and Mark Elvin The University of Manchester, Manchester Institute of Biotechnology, School of Chemical Engineering and Analytical Sciences, Faculty of Science and Engineering, 131 Princess Street, John Garside Building, M1 7DN, Manchester, UK

How do we describe a cell and predict its functional status? In the world of “omics” approaches, a cell can be described by the complement of genes that may be expressed (genomics), the RNA profile that may have functional activity for protein synthesis or cellular regulation (transcriptome), the proteins that may have structural and functional roles (proteome), and from the metabolite content of the cell and its immediate environment (metabolome). As an integrative readout of cellular status, the metabolome can offer a powerful (and, potentially, noninvasive) indication of cell status, providing information of molecular events that determine, regulate, or limit specific cell functions including the capacity to support recombinant protein production. This chapter focuses on the manner in which the metabolite status of mammalian cells used to manufacture commercially valuable recombinant proteins can be measured and how this information has been (and can be) used to optimize recombinant protein production in mammalian cells. The metabolite status of cells is frequently described as being generated through metabolomics or metabolite profiling. Although both terms convey a similar message, they are not interchangeable, as will be described in the sections that follow.

10.1 Value of Metabolic Data for the Enhancement of Recombinant Protein Production Recombinant protein expression in host cells is an energy-dependent process requiring the orchestration of multiple pathways for the correct transcription, translation, modification, and secretion of the desired product. As biochemical processes provide the precursors required for protein synthesis, the study of these processes is highly significant in the context of bioprocessing. The term “metabolomics” was first introduced in 1998 [1] in relation to yeast biochemical reactions and presented a descriptive global analysis of all metabolites in a biological system. Metabolites (carbohydrate-, amino Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

10 Metabolite Profiling of Mammalian Cells

y

Metabolomics 100’s of metabolites

om ple xit sin gc rea Inc

252

Glycomics 32 types of sugar linkages Lipidomics 8 lipid classes + subclasses Proteomics 20 amino acids Epigenomics 9 modification types Genomics and transcriptomics 4 nucleotides

Figure 10.1 Increasing complexity of “omics” data sets.

acid-, and lipid-based) are intermediates in cellular biochemical processes, which are collectively referred to as the “metabolome” [1]. Metabolite profiling is the assessment of a group of metabolites that can provide a snapshot of cell physiology under normal culture conditions or in response to extrinsic and intrinsic perturbations [2]. Metabolite detection and quantitation can be achieved through enzymatic assays coupled to a photometric readout [3, 4], which provides data on single (or a few) metabolites or through global analysis of metabolites achieved through methodologies frequently based on mass spectrometry (MS) techniques (Section 10.2.2.2). Unlike the genome (built from four different nucleotides) or the proteome (built from 20 different amino acids), the metabolome consists of hundreds of different metabolites with diverse chemical properties (Figure 10.1), which adds complexity to how samples are prepared and how specific metabolites are analyzed and quantified (Sections 10.2 and 10.3). In the context of recombinant protein production in bioprocessing, metabolic data have been applied to the optimization of cell culture medium, the rational design of feeding regimens, and to identify targets for cell line engineering, all with the purpose of improving volumetric or cell-specific productivity or protein quality (Section 10.4).

10.2 Technologies Used in the Generation of Metabolic Data Sets Metabolic information can be generated from samples taken during mammalian cell culture, generating either extracellular (footprint) and/or intracellular

10.2 Technologies Used in the Generation of Metabolic Data Sets

(fingerprint) profiles. Footprint analysis is an assessment of culture medium (after the removal of cells) and such samples can be used to monitor the culture process by assessment of the utilization of medium components and the accumulation of waste products throughout culture. Fingerprint analysis requires the complete separation of medium from cells followed by rapid and complete quenching of cells to stop any further metabolite interconversions before analysis. Through the inclusion of radiolabeled isotopes in cell culture medium, it is possible to track the flux of metabolites into different biochemical pathways and end points. There are also commercial instruments (e.g. YSI Biochemical Analyser [5, 6] and Agilent Seahorse Analyser [7]) that can be used for specific metabolite identification (e.g. glucose, lactate, and glutamine) as well as specific spectrophotometry-based enzymatic assays (e.g. glucose [3], lactate [4], and ammonia [8]). This section describes some of the technologies that may be used to measure the mammalian cell metabolome and the techniques used to detect, identify, and quantitate metabolites. Data analysis methodology will be presented in Section 10.3. 10.2.1

Targeted and Untargeted Metabolic Analysis

Metabolomics is the study of cells by measurement of their metabolite profiles, which can be achieved by untargeted and/or targeted analysis. Untargeted metabolomic approaches measure the broadest range of metabolites present in a sample without prior knowledge of the metabolome. This results in complex data sets that normally require extensive computational tools to enable comprehensive identification of metabolites. Therefore, untargeted metabolomic approaches provide a more global view of the sample and can uncover novel insights into the metabolites in a sample – revealing information across a wide range of metabolite classes. Targeted metabolomics generally provides higher sensitivity and selectivity than the untargeted approach, but this method requires prior knowledge of metabolites likely to be present in samples. Targeted metabolic analysis can also validate and expand upon results generated from untargeted metabolic analysis. Each approach (through use of metabolite standards) can be used to measure exact concentrations of predetermined metabolites [9, 10]. 10.2.2 Analytical Technologies Used in the Generation of Metabolite Profiles Once the metabolic study is designed and the type of experiment to be performed has been selected, it is important to match the needs to the correct analytical platform. The metabolome consists of hundreds of different metabolites with diverse chemical properties, which means no single analytical approach or technique is able to capture the entire metabolome. Therefore, different, but complementary, techniques are utilized to analyze diverse chemical species. For mammalian cells, the three most commonly used approaches are nuclear magnetic resonance (NMR), liquid chromatography–mass spectrometry (LC–MS), and gas chromatography–mass spectrometry (GC–MS). Each of these has advantages and disadvantages (Table 10.1). Alternative and less commonly used platforms include direct injection-mass spectrometry (DI-MS) and capillary

253

254

10 Metabolite Profiling of Mammalian Cells

Table 10.1 A comparison of analytical techniques used in metabolomic studies. Technology

Advantages

Disadvantages

References

NMR spectroscopy

• Nondestructive. • No derivatization required. • Fast (2–3 min/sample) as no separation needed. • Quantitative. • High resolution • Detects most organic classes. • Can be used to determine metabolite structure.

• Relatively insensitive – not suited for low-abundance compounds. • High setup costs. • Cannot detect or identify salts or inorganic ions. • Requires larger sample volumes than LC– or GC–MS.

[11, 12]

LC–MS

• Superb sensitivity • No derivatization required. • Many modes of separation available (e.g. HPLC and UPLC). • Large sample capacity. • Detects most organic and some inorganic molecules. • Small sample volumes. • Can be used in metabolite imaging (MALDI or DESI).

• Destructive (cannot recover sample). • Not very quantitative. • Slow (15–30 min/sample). • Limited commercial libraries available. • Novel compound identification difficult. • Higher startup cost.

[13, 14]

GC–MS

• • • •

Robust mature technology. Quantitative and sensitive. Moderate startup cost. Many commercial libraries available. • Detects most organic and some inorganic molecules. • Good separation reproducibility. • Compatible with liquids and gases.

• Destructive (cannot recover sample). • Sample derivatization required. • Sample separation desirable. • Slow (30–40 min/sample). • Novel compound identification difficult.

[13, 14]

Abbreviations: NMR, Nuclear magnetic resonance; LC–MS, liquid chromatography–mass spectrometry; GC–MS, gas chromatography–mass spectrometry; HPLC, high-performance liquid chromatography; UPLC, ultraperformance liquid chromatography ; MALDI, matrix-assisted laser desorption/ionization; DESI, desorption electrospray ionization.

electrophoresis–mass spectrometry (CE–MS) [15–17]. High-throughput spectroscopic profiling techniques relying on Fourier transform infrared spectroscopy (FT-IR) [18, 19] and Raman spectroscopy [20] have also been used for metabolite determination. 10.2.2.1

Nuclear Magnetic Resonance

NMR can identify and quantify a wide range of organic compounds, but unlike MS, NMR is nondestructive, so samples can be used for further analyses. However, the analysis of NMR spectra of complex mixtures is a challenge because of signal overcrowding making the analysis difficult; thus, new advanced

10.2 Technologies Used in the Generation of Metabolic Data Sets

quantitative 2D NMR techniques have been developed that have led to improved resolution and more intense signals [21]. NMR has been extensively used for metabolomics and metabolic flux analyses (MFA) of mammalian cells [22–26]. A major limitation of NMR is relatively low sensitivity, making it inappropriate for the analysis of large numbers of low-abundance metabolites found inside cells or in the culture medium. However, with new advancements in NMR technology (e.g. 2D NMR correlation spectroscopy [COSY], 2D NMR nuclear Overhauser effect spectroscopy [NOESY], and 2D NMR diffusion ordered spectroscopy [DOESY]), there is potential for much greater use of NMR in metabolomic studies [27]. 10.2.2.2

Mass Spectrometry

MS has advantages of high sensitivity and reproducibility and measures the masses of metabolites and their fragments to facilitate metabolite identification. A biological sample, comprising a mixture of metabolites, is injected into the mass spectrometer directly or after a separation step (either using liquid or gas chromatography). DI-MS has been used for high-throughput metabolomics [28, 29]. For example, Fuhrer et al. [28] detected 200–300 metabolites per sample with 1400 samples processed within a day using direct injection time-of-flight mass spectrometry (TOF-MS). However, as thousands of ions will be generated, utilization of a chromatographic step before entry of metabolites into the mass spectrometer allows for greater selectivity and sensitivity and (by providing retention time identification) can further aid metabolite identification when compared against known spectral libraries. Liquid Chromatography–Mass Spectrometry LC–MS is an analytical technique that

combines the physical separation capabilities of liquid chromatography (e.g. high-performance liquid chromatography [HPLC] or ultraperformance liquid chromatography [UPLC]) with the mass analysis capabilities of MS. Although LC separates metabolite mixtures, the MS stage provides structural identity of the metabolites. LC–MS has been used in mammalian metabolomic studies because of high sensitivity and a wide range of chemical selectivity in relation to analyte polarity and molecular mass [6, 30–35]. Although GC–MS requires chemical derivatization of metabolites (to facilitate the analysis of nonvolatile compounds), LC–MS does not, and this presents an advantage for LC–MS for certain analyses. However, a substantial potential drawback for use of LC–MS (as a nontargeted profiling tool) is the lack of transferable mass spectral libraries.

Gas Chromatography–Mass Spectrometry GC–MS is an analytical technique that

combines the features of GC and MS to identify compounds within a biological sample and it is currently the most mature technology for metabolite profiling. The concept of automated GC–MS metabolic profiling was developed around 40 years ago [36–38] and later adopted as a major technology for metabolomics. Using a GC–MS approach, it is possible to simultaneously profile several hundred chemically diverse compounds that include amino acids, sugars, organic acids, aromatic amines, and fatty acids [39]. The main advantage of GC–MS over

255

256

10 Metabolite Profiling of Mammalian Cells

LC–MS is the availability of many different spectral libraries to aid identification of metabolites (e.g. Fiehn spectral library [40]). The main limitations of GC–MS are that samples must be volatile after derivatization and many compounds can produce common fragments, potentially giving multiple identifications (e.g. the derivatization of arginine produces a compound identical to that produced by the derivatization of ornithine [41]). In the bioprocessing industry, GC–MS has been used predominantly for metabolite profiling of mammalian cells [10, 32, 42–51], the applications of which will be discussed in Section 10.4. 10.2.3

Metabolite Sample Preparation

Sampling primarily depends on the type of cells, their format (whether they are adherent or in suspension), and the type of metabolites to be measured (e.g. extracellular and/or intracellular). The main experimental steps for generating metabolic samples are summarized in Figure 10.2. Many metabolites are extremely labile (e.g. ATP) or have very high turnover rates (e.g. glucose-6-phosphate). Therefore, it is important that cellular metabolism is stopped (quenched) immediately at the point of cell sampling [52]. The analysis of intracellular metabolites requires the subsequent extraction of metabolites from the quenched cell extracts. Additional preanalytical steps such as sample concentration, lyophilization, and storage may also be undertaken before the analytical stage. Extracellular metabolites (Suspension/adherent)

Intracellular metabolites (Suspension/adherent)

Culture medium

Isolate

Cells

Cold centrifugation Fast filtration

Quench

Addition of cold quenching solutions

Extract

Multiple freeze thaw cycles

Storage (evaporation or lyophilisation)

Resuspend in accordance with analytical technique

Analytical technique (NMR, LC–MS, GC–MS)

Figure 10.2 A summary of the main steps to generate extracellular and intracellular metabolite profiles from mammalian cells. Nuclear magnetic resonance (NMR); liquid chromatography–mass spectrometry (LC–MS); gas chromatography–mass spectrometry (GC–MS).

10.3 Approaches for Metabolic Data Analysis

10.2.3.1

Extracellular Sample Preparation

For extracellular metabolite analysis, culture medium is separated from cells by a “cold” centrifugation step [10, 53] or by filtration through specific molecular-weight cutoff filters [30, 54]. Both approaches are performed rapidly at a low temperature to minimize extraneous metabolic activity. Care should be taken to prevent against sample freezing at this stage to avoid cell damage [53]. After separation, resultant supernatants are either analyzed immediately or are lyophilized and stored for later analysis. After storage, lyophilized samples are reconstituted before analysis by the technology of choice (Section 10.2.2). 10.2.3.2

Quenching of Intracellular Metabolite Samples

Quenching of enzymatic activity or metabolism is an essential factor to ensure that metabolite profiles measured for extracellular and intracellular metabolites reflect the true state of the cell and is not a consequence of artefacts during isolation. Common quenching strategies in mammalian cell culture are based on low temperature and pH [10, 43, 46, 53] to minimize leakage of intracellular metabolites into the surrounding liquid during the quenching process. 10.2.3.3

Metabolite Extraction from Quenched Cells

Typically, an extraction protocol is tailored toward specific metabolite properties and/or the analytical technique of choice [55]. Extraction conditions that favor the preservation of one metabolite species may well be damaging to other metabolite species (e.g. valine identified after methanol and ethanol quenching/extraction is not detected when perchloric acid or methanol/chloroform is used as the quenching/extraction method [56]). This presents challenges to find the right balance between a wide analytical outcome for sample complexity and metabolite stability. In mammalian cells (especially Chinese hamster ovary [CHO] cells), several reports have commented on the applicability of different extraction methodologies to metabolomics [10, 43, 46, 53, 57, 58]. Ultimately, the extraction methodology should be adapted toward the required metabolite range and the specific analytical technique to be used (Table 10.1). 10.2.3.4

Metabolic Flux Analysis

Metabolic flux refers to the passage of a metabolite through various pathways over time. It is an effective method for analysis of the balance between metabolism in competing pathways and determination of the fate of metabolites. Two important technologies in this area are flux balance analysis (FBA) and 13 C-fluxomics. In FBA, metabolic fluxes are estimated using mathematical models, which take into account the stoichiometry of each reaction [59]. In 13 C-fluxomics, metabolic precursors are enriched with 13 C before introduction into the system. The incorporation of 13 C into metabolites can be measured (by MS or NMR) and metabolic fluxes can be estimated [60].

10.3 Approaches for Metabolic Data Analysis Metabolic experiments lead to the generation of complex data sets, which can include hundreds of metabolites. Analysis of metabolomic data is a multistep

257

258

10 Metabolite Profiling of Mammalian Cells

Data acquisition NMR

LC–MS

GC–MS

Spectral data

Data processing

Data analysis

Data interpretation

Univariate analysis

Enrichment analysis

Multivariate analysis

Pathway analysis

- Principle component analysis (PCA) - Cluster analysis (CA) - Partial least squares (PLS) analysis

Baseline correction Noise filtering

Metabolite ID

Peak detection

Spectral libraries/databases

Data integration Metabolite profiles

Peak alignment Normalization

Metabolite models

Deconvolution

Data matrix

Identified metabolites

other ‘omics datasets

Figure 10.3 Data analysis workflow for NMR and MS-based metabolomics.

process (Figure 10.3), which requires comprehensive evaluation using available bioinformatics and statistical tools (Table 10.2). 10.3.1

Data Processing

Once metabolomic data have been acquired (Section 10.2), the raw data (e.g. NMR spectra, LC–MS, and GC–MS spectra) are preprocessed to facilitate compound identification. Generally, this preprocessing involves steps such as baseline correction [88–91], noise reduction [92], retention time correction (in the case of MS data) [93], peak detection, integration and alignment [94–96], normalization [97, 98], and deconvolution [99]. This generates a data matrix that includes the detected peaks of all given metabolites. Different databases (such as the Human Metabolome Database [HMDB] [65–67] or the Metabolite and Tandem MS Database [METLIN] [62]) or spectral libraries (such as the Fiehn Library [40] or the National Institute of Standards and Technology [NIST] [81]) are then used to identify the metabolites from the different spectra. 10.3.2

Data Analysis

Typical statistical analyses of metabolomic data sets consist of two parts. Initially, different univariate (single-variable) [100, 101] and multivariate (multivariable) [102, 103] methods are used to generate an overview of the data sets and to identify metabolites that show significant changes. Then, data mining techniques are used to identify groups of functionally related metabolites [104]. Therefore, once a data matrix has been produced from the raw data files (Section 10.3.1), then subsequent steps usually involve different forms of statistical analysis (e.g. principle component analysis [PCA], cluster analysis [CA], and partial least

10.3 Approaches for Metabolic Data Analysis

Table 10.2 Spectral libraries, databases, datasets, web tools, and public repositories for MS-based metabolomic analysis. Name and references

Description

MMCD (Madison-Qingdao Metabolomics Consortium Database) [61]

• Contains information on over 20000 metabolites • The database is compatible with both NMR and MS data • Facilitates high-throughput metabolomic investigations

METLIN (Metabolite and Tandem MS Database) [62–64]

• Metabolomic database containing over 16000 metabolites • Data management system to assist in metabolite research and identification • Annotated list of metabolites and their mass, chemical formula, and structure available on the METLIN website (metlin.scripps.edu)

HMDB (Human Metabolome Database) [65–67]

• Combines quantitative chemical, physical, clinical, and biological data about thousands of endogenous human metabolites • Strong focus on quantitative and analytic information about metabolites • Online resource designed to link (i) chemical data, (ii) clinical data, and (iii) molecular biology/biochemistry data

MetaboLights [68–72]

• Stores raw experimental metadata from metabolomic studies (it is a cross-species and cross application database) • It is a general purpose cross-species and cross-application database in metabolomics • Facilitates the search of mass spectra from metabolites using GC–MS • Comprises mass spectra and retention time indices of pure reference substances and mass spectral tags (MSTs) of yet unidentified metabolites • A mass spectral and retention index library for comprehensive metabolic profiling • The current libraries comprise over 1000 identified metabolites

GMD (Golm Metabolome Database) [73] FiehnLib [40]

BinBase [74]

• An automated peak annotation database system developed for the analysis of GC–TOF–MS data derived from complex volatile mixtures • Developed to track and identify derivatized metabolites • The vocBinBase algorithm assigns the identity of compounds existing in the database

MS2T (MS/MS spectral tag) [75]

• Informs about nontargeted metabolic profiling analysis using LC–MS and provides structural information for detected peaks • Uses optimized methods for acquisition of tandem mass spectrometry (MS/MS) data – enabling high-throughput acquisition of metabolite profiling data

Metabolomic Workbench

• Public repository for metabolomic metadata and experimental data spanning 20 different species and experimental platforms (NMR and MS) • Metabolites characterized are linked to chemical structures (in the metabolite structure database) to facilitate comparative analysis

[76]

259

260

10 Metabolite Profiling of Mammalian Cells

Table 10.2 (Continued) Name and references

Description

MassBase

• A mass spectral tag archive for metabolomics • The MassBase database is fundamental for metabolomics, provides raw mass chromatograms from various biological samples, and text-file data processed from raw chromatograms • Manages and analyses LC–MS data • It is a programming library and tool collection integrated into workflow systems, such as KNIME, Galaxy, and WS-PGRADE • Software provides ready-to-use tools for analysis of proteomics and nontargeted metabolomic data

[77]

OpenMS [78, 79]

MassBank [80]

• Public repository of MS data for sharing among the scientific research community

NIST Libraries (National Institute of Standards and Technology) [81]

• Access to comprehensive, annotated MS reference collections from various organisms

AMDIS (Automated Mass spectral Deconvolution and Identification System) [82, 83]

• Software developed to identify low-abundant peaks in total ion current (TIC) chromatograms • Method that deconvolutes spectra, peak shape, and retention time from complex chromatograms and subsequently matches the obtained spectra to that of a reference library

Metab

• R package for high-throughput processing of metabolomic data analyzed by AMDIS • It performs statistical hypothesis test (t-test) and analysis of variance (ANOVA) – speeds up data mining process

[84] MetaboAnalyst [85–87]

• Web-based analytical pipeline for high-throughput metabolomic studies • Procedures for data processing, normalization, multivariate statistical analysis, and data annotation

squares [PLS] analysis). A number of software packages are available for the processing and analysis of metabolomic data (Table 10.2). This table is not exhaustive but gives applications commonly used in the mammalian field. 10.3.3

Data Interpretation and Integration

Individual metabolites can be analyzed by generation of profiles for abundance of each metabolite throughout culture (metabolite profiling) [2, 32, 45]. However, this type of interpretation provides a limited functional readout of cellular metabolism. To better understand the role of each metabolite, in the context of cellular functionality, information derived from metabolomic analyses has to be related to both biochemical causes and physiological consequences [105, 106]. This form of data interpretation is achieved through enrichment and pathway

10.4 Implementation of Metabolic Data in Bioprocessing

analyses [107], and these two types of analyses are performed using ad hoc software tools [108], which map metabolites to known biochemical pathways on the basis of information held in public databases such as the Kyoto Encyclopedia of Genes and Genomes (KEGG) [109]. Metabolic modeling integrates data from metabolite profiles, enzyme kinetics, reaction stoichiometries, and other associated cofactor reactions (e.g. REDOX state) using mathematical models to provide a global metabolomic perspective. Once metabolic pathways have been ascribed, further integration with other available “omics” data sets (e.g. genomic, transcriptomic, and proteomic) can be applied for a more comprehensive understanding of the cellular biology [110–112].

10.4 Implementation of Metabolic Data in Bioprocessing Metabolic data, whether acquired through metabolite profiling or metabolomic analyses, have been used to improve bioprocessing capacity of mammalian cells. Such data have generated metabolic understanding of molecular principles that underpin bioprocessing including metabolic pathway flux in mammalian expression systems and the relationship between metabolism and growth phases [54, 113, 114]. Furthermore, metabolic information has been used to rationalize design of chemically defined media [115, 116], optimize feeding regimens to boost productivity [6, 26, 45, 117–119], improve product quality [120–122], define metabolic markers of high productivity [2, 114, 123, 124], and, more recently, define targets for cell line engineering [5, 125–135]. The remainder of this section highlights developments in bioprocessing that have been achieved through metabolic studies. 10.4.1

Relationship Between Growth Phase and Metabolism

Cell lines used in mammalian bioprocessing (e.g. CHO, nonsecreting murine myeloma [NS0], human embryonic kidney [HEK], and baby hamster kidney [BHK]) demonstrate a metabolic profile that is reflective of cancer cell lines [136]. Cells in batch culture exhibit a period of exponential growth, followed by a static phase (no growth, but frequently associated with increased protein-specific productivity [48, 137]) and a decline phase. Each phase has unique metabolic traits (summarized in Figure 10.4). 13 C-fluxomics [123], FBA [137], and in silico models [54] have all proven to be valuable tools in ascribing these culture phase transitions to metabolic function. During the exponential phase of culture, mammalian cells exhibit a high glycolytic flux exemplified by high glucose consumption rates [48, 138]. Templeton et al. [48] observed (through 13 C-fluxomics) that 90% of glucose was utilized in glycolysis by CHO-DHFR (dihydrofolate reductase) cells during exponential growth, generating lactate, and alanine, which were excreted as overflow products. During this phase of culture, amino acids (glutamine and asparagine) represent the greatest source of carbon influx (via oxaloacetate) to fuel the tricarboxylic acid (TCA) cycle [47, 48, 139], with ammonia excretion as a waste product [54].

261

Cell density

Accumulation

10 Metabolite Profiling of Mammalian Cells

Lactate Ammonia Alanine Glycerol

Citrate Malate Fumarate *Lactate Glycerol

Glycerol *Lactate Citrate Malate Fumarate

Exponential

Static

Decline

High glycolytic flux

Oxidative phosphorylation

Scavenger reactions

Culture time Utilization

262

Glucose Glutamine Asparagine

Glucose *Lactate Pyruvate Asparagine Glutamine

Glucose *Lactate Alanine Serine Leucine Lysine

Figure 10.4 Summary of the metabolic traits associated with different culture phases of recombinant cell lines. *Lactate metabolism varies depending on cell clone and culture environment.

Through the combination of UPLC–MS and mathematical modeling, Selvarasu et al. [54] reported that pathways linked to glutathione and glycerophospholipid metabolism showed the greatest activity during exponential growth, attributed to maintenance of REDOX potential and lipid synthesis, respectively. The entry into static phase coincides with a decline in growth rate, which has been attributed to pathways related to anaplerotic TCA replenishment, maintenance of lipid synthesis, tetrahydrofolate conversions, and REDOX modulation [26]. During static phase, cells generally demonstrate a shift toward mitochondrial oxidative metabolism and an increased flux in the TCA cycle, which produces reactive oxygen species (ROS) as a by-product of oxidative phosphorylation [48, 54]. This transition to mitochondrial metabolism coincides with the accumulation of TCA cycle intermediates (e.g. citrate, malate, and fumarate) and concurrent depletion of amino acids (e.g. asparagine, aspartate, and glutamine) and pyruvate, a feature observed in several independent studies in CHO cells via different methodologies [30, 32, 45, 53, 117]. Increased activity of the pentose phosphate pathway has also been observed in the static phase of CHO cell culture, a response suggested to occur as an adaptive response to counteract ROS production [48]. A shift from lactate and alanine production to consumption (with conversion to pyruvate and subsequent entry into the TCA cycle) has been observed in CHO [114], HEK [140], and NS0 [141] cells. However, this observation is not universal (exhibiting variability with cell clone and environment) and, in some cases, this metabolic switch has been associated with cells that can achieve greater specific productivity (Section 10.4.2).

10.4 Implementation of Metabolic Data in Bioprocessing

Progression into, and through, the decline phase of culture has also been associated with “scavenger” metabolism. Once the major carbon sources have been depleted (e.g. glucose, pyruvate, glutamine, asparagine, and lactate), several studies have observed the conversion of amino acids (e.g. alanine, serine, leucine, and lysine) into pyruvate to further fuel the TCA cycle [45, 117]. 10.4.2 Identification of Metabolic Indicators Associated with High Cell-Specific Productivity Clonally derived recombinant cell lines have been observed to demonstrate significant differences in growth- and cell-specific productivity [138, 139, 142, 143], and metabolic analyses of these clonal variants have provided insights into the metabolic hallmarks of high and low productivity. As described above (Section 10.4.1), the static phase of culture (associated with the redirection of pathways away from cell growth and toward increased protein production) is associated with increased TCA cycle flux. Therefore, many of the metabolic markers associated with high productivity are related to improved mitochondrial metabolism, such as increased activity of TCA cycle enzymes (e.g. malate dehydrogenase [MDH]) [138] and elevated amounts of REDOX regulation [139]. Inversely, indicators of low productivity are related to carbon losses to this flux during glycolysis (e.g. lactate and glycerol). Figure 10.5 depicts the metabolic pathways that are associated with higher productivities. The most extensively studied indicator of high productivity is the lactate shift toward consumption upon entry into the static phase of culture, where it is converted to pyruvate [113, 114, 117, 141, 144]. Martinez et al. [137] determined through FBA that ATP production per carbon molecule was 5.9 times greater for CHO cells during the period of lactate consumption (compared to the period of lactate production) because of an increase in the proportion of pyruvate processing into the TCA cycle. The switch to net lactate consumption is usually observed in response to environmental status such as extracellular glucose depletion. However, the coconsumption of residual glucose (from culture medium) and lactate has been observed in highly productive mammalian clones (e.g. CHO, HEK, and NS0) exhibiting high oxidative capacity [114, 140, 141]. Luo et al. [114] observed that supplementation of the cell culture medium with copper, an essential cofactor in mitochondrial metabolism, was by itself sufficient to promote this metabolic shift in CHO-DHFR cultures. The final stage of recombinant protein expression is the secretion of the target product into the medium. A greater cellular capacity for secretion (because of the greater amount of endoplasmic reticulum [ER] associated membranes) has been linked to the extent of lipogenesis and lipid metabolic indicators exhibited by specific NS0 clones [124]. 10.4.3 Utilizing Metabolic Data to Improve Biomass and Recombinant Protein Yield Historically, cell culture medium relied on components such as animal-derived serum or plant-based hydrosylate feed, prone to batch-to-batch variation and

263

264

10 Metabolite Profiling of Mammalian Cells

Glucose

Glucose Glycerol

Threitol

Threitol

*Glycerol* Serine

Glycine

Nucleotide precursors

Glycine Serine

Pyruvate Alanine Lactate

*Lactate* Oxidative phosphorylation

Alanine Lipid synthesis *TCA cycle*

Citrate Malate Fumarate

Glutamine Asparagine Aspartate Glutamate

Figure 10.5 Summary of metabolic pathways that influence cell line productivity. Arrows indicate the direction of reactions or transportations. Dashed arrows represent multiple metabolic reactions between metabolites. Bidirectional arrows describe reversible reactions. Bold arrows represent metabolic reactions that have been associated with improved bioprocessing capacity. Starred metabolites in italics represent reactions that regenerate NAD+ /NAP+ . Boxes represent additional pathways that are related to higher productivity.

potential for contaminants such as prions, mycoplasma, and viruses, that could impact product quality and marketability. Most importantly, these components were ill-defined, meaning that optimization of medium and feed could only be achieved through brute-force iterative processes. Ma et al. [117] developed a chemically defined feed as a hydrosylate replacement, which was suitable for use in both NS0 and CHO cell cultures, by combining data from LC–MS and GC–MS analyses of intracellular and extracellular samples. They implemented a three-phase approach starting with a proprietary chemically defined feed consisting of over 90 components formulated to excess concentrations. Through subsequent refinement of components that showed no impact or minimal benefit to NS0 cell culture, the resultant feed provided a 33% increase in integral viable cell density (IVCD) and 62% increase in specific productivity associated with a shift to net lactate consumption. These growth and metabolic responses were subsequently observed for a CHO-GS (glutamine synthetase) platform using the same feed [117]. Another study using a CHO-GS system, expressing an IgG4 antibody, used GC–MS to provide metabolite profiling of intracellular and extracellular metabolites taken during batch culture in bioreactors to design a feed to replenish limiting nutrients before their depletion at growth phase transitions to static

10.4 Implementation of Metabolic Data in Bioprocessing

phase [45]. Addition of a simple nutrient feed (composed of glucose, pyruvate, asparagine, aspartate, and glutamate) during the mid-exponential phase of culture extended culture longevity and increased cell biomass by 35%, resulting in a 100% increase in antibody yield. Subsequent metabolite profiling indicated a shift in metabolism toward an improved oxidative capacity responsible for the higher titer observed. In a similar study using a CHO-DHFR system expressing a single-chain chimeric antibody, Blondeel et al. [26] used two-dimensional difference gel electrophoresis (2D-DIGE) proteomics coupled with NMR-based metabolomics to identify pathways that were limiting to cell growth and subsequently designed a nutrient feed comprising of eight metabolites that were depleted on the transition to stationary phase (Section 10.4.1), which improved maximal cell density by 75%. 10.4.4

Utilizing Metabolic Understanding to Improve Product Quality

Quality attributes, such as glycosylation profiles, are determined by the host cell type and the culture conditions. Glucose present in the medium (and subsequent interconversion to other sugars) provides the precursors for glycan biosynthesis (N-linked and O-linked) [145]. Murine host cells (e.g. CHO, NS0) generate their own distinct glycosylation profiles, different from those of human cell lines (notably in terms of sialic acid content) because of the differential expression of enzymes involved in nucleotide–sugar metabolism and for processing of glycan structures. These differences can affect product stability, efficacy, and immunogenicity [121, 146–148]. Metabolic data have been used to describe the relationship between metabolism and glycosylation and illustrate strategies to improve glycosylation profiles of biopharmaceuticals via manipulation of culture medium composition or genomic engineering. For example, undesirable galactosylation has been attributed to the availability of the UDP-Gal precursor, and availability of UDP-Gal has been modified through the addition of uridine, manganese, and galactose to CHO cell culture medium [120, 149, 150]. Similarly, Wong et al. [151] observed that glutamine, uridine, and glucosamine availability impacted on the synthesis of the (uridine diphosphate) UDP-GlcNAc in CHO cell cultures, which could be modulated through the supplementation of uridine and glucosamine to improve IFN-𝛾 sialylation. Burleigh et al. [122], using FBA, showed that glutamine concentration had a significant effect on the microheterogeneity of glycan structures on human chorionic gonadotropin (HCG) produced by CHO cells. Baker et al. [121] demonstrated that addition of glucosamine and uridine increased the sialic content of tissue inhibitor of metalloproteinase-1 (TIMP-1) derived from CHO and NS0-GS platforms by increasing the availability of UDP-N-acetylhexosamine, while the (cartilage matrix protein) CMP-sialic acid precursor was increased through N-acetylmannosamine supplementation. 10.4.5

Cell Line Engineering to Redirect Metabolic Pathways

Genetic engineering has been implemented in mammalian cell lines to improve metabolic efficiency and product-quality attributes. Engineering pathways to

265

266

10 Metabolite Profiling of Mammalian Cells

promote lactate consumption (Section 10.4.2) at the pyruvate node has been a key target for genetic manipulation. Yip et al. [125] used a combinatorial approach to knockdown/out the genes for pyruvate dehydrogenase kinase (pdk 1–3) and lactate dehydrogenase (ldha) enzymes, respectively. With this approach, they sought to redirect lactate metabolism toward increased flux of pyruvate into the TCA cycle, preventing pyruvate conversion to lactate in CHO cells. However, this approach was lethal. An alternative strategy is the overexpression of pyruvate carboxylase (PC), which in its endogenous form resides in the mitochondria and catalyzes the carboxylation of pyruvate to oxaloacetate, while the yeast analog (PYC2) is localized in the cytosol when expressed in mammalian cells. This target has been overexpressed in HEK [126–128], BHK [129, 130], and CHO cells [131–133] and in all cases promoted a shift to net lactate consumption (even in the presence of glucose [Section 10.4.3]) with a concomitant increase in medium pH, increased IVCD, and increased product titer. FBA confirmed that this strategy resulted in a higher flux of pyruvate to oxaloacetate in HEK cells [126] and Irani et al. [130] observed an increase in oxygen uptake and intracellular ATP pools in BHK cells, suggesting that increased recombinant protein production is related to increased flux through the TCA cycle. Enzymes in the TCA cycle are also attractive targets for genetic engineering as inefficiency in TCA cycling may negatively impact on ATP production, which is essential for protein expression. One study [6] used 13 C-fluxomics in combination with quantitative LC–MS to identify MDH isoform 2 as the rate-limiting enzyme in CHO cells expressing an IgG in fed-batch culture. Overexpression of mdh II in these cells resulted in a decrease in specific malate and lactate secretion rate and increased intracellular ATP and NADH, which increased cell density (1.3-fold) and mAb (monoclonal antibody) titer (1.2-fold). Product quality (e.g. glycosylation) can be manipulated through modulating culture environment (Section 10.4.4); however, it may also be achieved through genetic manipulation of glycosylation enzymes. This approach was exemplified in the work of Yang et al. [152], who used zinc finger nucleases (ZFNs) to perform single and combinatorial gene knockouts of enzymes relating to N-glycosylation pathways (N-glycan branching, galactosylation, poly-LacNAc elongation, terminal capping by sialylation, and core 𝛼6-fucosylation) to promote homogeneous N-glycosylation profiles for erythropoietin (EPO) and IgG proteins in CHO cells. Finally, although not a direct metabolic target, the manipulation of apoptotic pathways through the overexpression of a number of anti-apoptotic genes (bcl-2, bcl-xL, e1b-19k, aven, and xiap) has been shown to promote a shift in metabolism toward increased mitochondrial oxidation in CHO cells, resulting in slower glucose and glutamine consumption and decreased alanine production [5, 134, 135].

10.5 Future Perspectives Today, many chemically defined media and complementary feeds are commercially available for the culture of mammalian cells. Historically, many of these

References

were derived via iterative step changes from a fundamental basic understanding of the needs of a cell. However, the richness of data obtained from the analytical power of MS coupled with MFA and metabolic model development allows for a much enhanced dissection of linkages between specific metabolic events and bioprocess effectiveness (biomass attainment, product yield, and product quality). The precision of metabolite analysis offered by current analytical technologies provides opportunities to address many specific questions that relate to enhanced product manufacture. This includes adaptation of cells to process scale-up, application of cell processes to high-density continuous processes, and development of process analytical technologies (to allow monitoring of both amount and quality of recombinant products). In addition, future potential lies in the development of media and feeds that can enhance the production of specific recombinant products (e.g. to enhance the production of novel format/difficult to express products). Technologies are also developing to enable assessment of metabolomics at the single cell level with MS-based methods [153], including matrix-assisted laser desorption (MALDI) [154, 155] and desorption electrospray ionization (DESI) [156], showing the potential to understand heterogeneity in cell populations, a facet especially relevant to clonal variation in CHO cell populations. Furthermore, noninvasive techniques based on Raman approaches [157] offer great potential for cell screening and cell line development. The techniques exist now and the route map to application also exists. Exploitation of the full potential for enhanced bioprocesses and manufacture of valuable products awaits.

Acknowledgments The authors would like to acknowledge the financial support of the Biotechnology and Biosciences Research Council (BBSRC) of the United Kingdom that through specific research grants (BB/E005985/1, BB/M001164/1, BB/M01701X/1, and BB/N004000/1) enabled us to develop the technologies and understanding of the themes described in this chapter. We also appreciate the research work undertaken, and contributions made, by a number of gifted colleagues especially Dr. Alex Croxford, Dr. David Knight, Dr. Andrew Picken, and Dr. Chris Sellick.

References 1 Oliver, S.G., Winson, M.K., Kell, D.B., and Baganz, F. (1998). Systematic

functional analysis of the yeast genome. Trends Biotechnol. 16: 373–378. 2 Sellick, C.A., Croxford, A.S., Maqsood, A.R. et al. (2015). Metabolite pro-

filing of CHO cells: molecular reflections of bioprocessing effectiveness. Biotechnol. J. 10: 1434–1445. 3 Froesch, E.R., Renold, A.E., and McWilliams, N.B. (1956). Specific enzymatic determination of glucose in blood and urine using glucose oxidase. Diabetes 5: 1–6.

267

268

10 Metabolite Profiling of Mammalian Cells

4 Bergmeyer, H.U., Bergmeyer, J., and Grassl, M. (1983). Methods of Enzymatic

Analysis. Weinheim: Verlag Chemie. 5 Templeton, N., Lewis, A., Dorai, H. et al. (2014). The impact of

6

7

8 9 10

11 12 13 14 15

16 17

18

19

20

anti-apoptotic gene Bcl-2 Delta expression on CHO central metabolism. Metab. Eng. 25: 92–102. Chong, W.P.K., Reddy, S.G., Yusufi, F.N.K. et al. (2010). Metabolomics-driven approach for the improvement of Chinese hamster ovary cell growth: overexpression of malate dehydrogenase II. J. Biotechnol. 147: 116–121. Salabei, J.K., Gibb, A.A., and Hill, B.G. (2014). Comprehensive measurement of respiratory activity in permeabilized cells using extracellular flux analysis. Nat. Protoc. 9: 421–438. Fawcett, J.K. and Scott, J.E. (1960). A rapid and precise method for the determination of urea. J. Clin. Pathol. 13: 156–159. Ivanisevic, J., Elias, D., Deguchi, H. et al. (2015). Arteriovenous blood metabolomics: a readout of intra-tissue metabostasis. Sci. Rep. 5: 1257–1270. Evie, I.M., Dickson, A.J., and Elvin, M. (2017). Metabolite profiling of mammalian cell culture processes to evaluate cellular viability. In: Cell Viability Assays. Methods in Molecular Biology, vol. 1601 (eds. D. Gilbert and O. Friedrich). New York, NY: Humana Press. Wishart, D.S. (2008). Quantitative metabolomics using NMR. TrAC, Trends Anal. Chem. 27: 228–237. Wishart, D.S. (2011). Advances in metabolite identification. Bioanalysis 3: 1769–1782. Dunn, W.B., Bailey, N.J.C., and Johnson, H.E. (2005). Measuring the metabolome: current analytical technologies. Analyst 130: 606–625. Zhang, A.H., Sun, H., Wang, P. et al. (2012). Modern analytical techniques in metabolomics analysis. Analyst 137: 293–300. Kai, H., Kinoshita, K., Harada, H. et al. (2017). Establishment of a direct-injection electron ionization-mass spectrometry metabolomics method and its application to lichen profiling. Anal. Chem. 89: 6408–6414. Chung, B.K.-S. and Lee, D.-Y. (2012). Computational codon optimization of synthetic gene for protein expression. BMC Syst. Biol. 6: 134. Hirayama, A., Wakayama, M., and Soga, T. (2014). Metabolome analysis based on capillary electrophoresis-mass spectrometry. TrAC, Trends Anal. Chem. 61: 215–222. Kaderbhai, N.N., Broadhurst, D.I., Ellis, D.I. et al. (2003). Functional genomics via metabolic footprinting: monitoring metabolite secretion by Escherichia coli tryptophan metabolism mutants using FT-IR and direct injection electrospray mass spectrometry. Comp. Funct. Genomics 4: 376–391. Keymanesh, K., Hassan, M., and Soroush, D. (2009). Metabolome comparison of transgenic and non-transgenic rice by statistical analysis of FTIR and NMR spectra. Rice Sci. 16: 119–123. Ellis, D.I. and Goodacre, R. (2006). Metabolic fingerprinting in disease diagnosis: biomedical applications of infrared and Raman spectroscopy. Analyst 131: 875–885.

References

21 Khlifi, S., Mauve, C., Glab, N. et al. (2017). An easier analysis of complex

22

23

24

25

26

27 28

29

30

31

32

33

34

mixtures with highly resolved and sensitivity enhanced 2D quantitative NMR: application to tracking sugar phosphates in plants. Anal. Methods 9: 2328–2333. Bradley, S.A., Ouyang, A.L., Purdie, J. et al. (2010). Fermentanomics: monitoring mammalian cell cultures with NMR spectroscopy. J. Am. Chem. Soc. 132: 9531–9533. Goudar, C., Biener, R., Boisart, C. et al. (2010). Metabolic flux analysis of CHO cells in perfusion culture by metabolite balancing and 2D [13 C, 1 H] COSY NMR spectroscopy. Metab. Eng. 12: 138–149. Aranibar, N., Borys, M., Mackin, N.A. et al. (2011). NMR-based metabolomics of mammalian cell and tissue cultures. J. Biomol. NMR 49: 195–206. Wagstaff, J.L., Masterton, R.J., Povey, J.F. et al. (2013). 1 H NMR spectroscopy profiling of metabolic reprogramming of Chinese hamster ovary cells upon a temperature shift during culture. PLoS One 8: e77195. Blondeel, E.J.M., Ho, R., Schulze, S. et al. (2016). An omics approach to rational feed enhancing growth in CHO cultures with NMR metabolomics and 2D-DIGE proteomics. J. Biotechnol. 234: 127–138. Gowda, G.A.N. and Raftery, D. (2017). Recent advances in NMR-based metabolomics. Anal. Chem. 89: 490–510. Fuhrer, T., Heer, D., Begemann, B., and Zamboni, N. (2011). High-throughput, accurate mass metabolome profiling of cellular extracts by flow injection - time-of-flight mass spectrometry. Anal. Chem. 83: 7074–7080. Garcia-Flores, M., Juarez-Colunga, S., Garcia-Casarrubias, A. et al. (2015). Metabolic profiling of plant extracts using direct-injection electrospray ionization mass spectrometry allows for high-throughput phenotypic characterization according to genetic and environmental effects. J. Agric. Food. Chem. 63: 1042–1052. Chong, W.P.K., Goh, L.T., Reddy, S.G. et al. (2009). Metabolomics profiling of extracellular metabolites in recombinant Chinese hamster ovary fed-batch culture. Rapid Commun. Mass Spectrom. 23: 3763–3771. Chong, W.P.K., Yusufi, F.N.K., Lee, D.Y. et al. (2011). Metabolomics-based identification of apoptosis-inducing metabolites in recombinant fed-batch CHO culture media. J. Biotechnol. 151: 218–224. Dietmair, S., Hodson, M.P., Quek, L.-E. et al. (2012). Metabolite profiling of CHO cells with different growth characteristics. Biotechnol. Bioeng. 109: 1404–1414. Chong, W.P.K., Thng, S.H., Hiu, A.P. et al. (2012). LC–MS-based metabolic characterization of high monoclonal antibody-producing Chinese hamster ovary cells. Biotechnol. Bioeng. 109: 3103–3111. Floris, P., McGillicuddy, N., Albrecht, S. et al. (2017). Untargeted LC–MS/MS profiling of cell culture media formulations for evaluation of high temperature short time treatment effects. Anal. Chem. 89: 9953–9960.

269

270

10 Metabolite Profiling of Mammalian Cells

35 Hsu, H.H., Araki, M., Mochizuki, M. et al. (2017). A systematic approach

36 37

38

39

40

41

42 43

44

45

46

47 48

49

to timeseries metabolite profiling and RNA-seq analysis of Chinese hamster ovary cell culture. Sci. Rep. 7: 43518. Gates, S.C. and Sweeley, C.C. (1978). Quantitative metabolic profiling based on gas-chromatography. Clin. Chem. 24: 1663–1673. Gates, S.C., Dendramis, N., and Sweeley, C.C. (1978). Automated metabolic profiling of organic-acids in human urine. I. Description of methods. Clin. Chem. 24: 1674–1679. Gates, S.C., Sweeley, C.C., Krivit, W. et al. (1978). Automated metabolic profiling of organic-acids in human urine. II. Analysis of urine samples from healthy adults, sick children, and children with neuroblastoma. Clin. Chem. 24: 1680–1689. Roessner, U., Wagner, C., Kopka, J. et al. (2000). Simultaneous analysis of metabolites in potato tuber by gas chromatography-mass spectrometry. Plant J. 23: 131–142. Kind, T., Wohlgemuth, G., Lee, D.Y. et al. (2009). FiehnLib: mass spectral and retention index libraries for metabolomics based on quadrupole and time-of-flight gas chromatography/mass spectrometry. Anal. Chem. 81: 10038–10048. Halket, J.M., Waterman, D., Przyborowska, A.M. et al. (2005). Chemical derivatization and mass spectral libraries in metabolic profiling by GC/MS and LC/MS/MS. J. Exp. Bot. 56: 219–243. Sellick, C.A., Maqsood, A.R., Croxford, A.S. et al. (2009). Optimizing CHO cell culture conditions. Genet. Eng. Biotechnol. News 29: 56–57. Sellick, C.A., Hansen, R., Maqsood, A.R. et al. (2009). Effective quenching processes for physiologically valid metabolite profiling of suspension cultured mammalian cells. Anal. Chem. 81: 174–183. Sellick, C.A., Knight, D., Croxford, A.S. et al. (2010). Evaluation of extraction processes for intracellular metabolite profiling of mammalian cells: matching extraction approaches to cell type and metabolite targets. Metabolomics 6: 427–438. Sellick, C.A., Croxford, A.S., Maqsood, A.R. et al. (2011). Metabolite profiling of recombinant CHO cells: designing tailored feeding regimes that enhance recombinant antibody production. Biotechnol. Bioeng. 108: 3025–3031. Dietmair, S., Timmins, N.E., Gray, P.P. et al. (2010). Towards quantitative metabolomics of mammalian cells: development of a metabolite extraction protocol. Anal. Biochem. 404: 155–164. Dean, J. and Reddy, P. (2013). Metabolic analysis of antibody producing CHO cells in fed-batch production. Biotechnol. Bioeng. 110: 1735–1747. Templeton, N., Dean, J., Reddy, P., and Young, J.D. (2013). Peak antibody production is associated with increased oxidative metabolism in an industrially relevant fed-batch CHO cell culture. Biotechnol. Bioeng. 110: 2013–2024. Mohmad-Saberi, S.E., Hashim, Y., Mel, M. et al. (2013). Metabolomics profiling of extracellular metabolites in CHO-K1 cells cultured in different types of growth media. Cytotechnology 65: 577–586.

References

50 Le, H., Jerums, M., and Goudar, C.T. (2015). Characterization of intrinsic

51

52

53

54

55

56

57

58

59 60 61

62 63

64

65 66

variability in time-series metabolomic data of cultured mammalian cells. Biotechnol. Bioeng. 112: 2276–2283. Zuhanis, Y., Hashim, H.Y., Saberi, S.E.M. et al. (2016). Profiling of extracellular metabolites in CHO-K1 cells cultured in medium with different levels of serum and glutamine. Curr. Metabolomics 4: 56–62. Vuckovic, D. (2012). Current trends and challenges in sample preparation for global metabolomics using liquid chromatography-mass spectrometry. Anal. Bioanal. Chem. 403: 1523–1548. Sellick, C.A., Hansen, R., Stephens, G.M. et al. (2011). Metabolite extraction from suspension-cultured mammalian cells for global metabolite profiling. Nat. Protoc. 6: 1241–1249. Selvarasu, S., Ho, Y.S., Chong, W.P.K. et al. (2012). Combined in silico modeling and metabolomics analysis to characterize fed-batch CHO cell culture. Biotechnol. Bioeng. 109: 1415–1429. Alvarez-Sanchez, B., Priego-Capote, F., and de Castro, M.D.L. (2010). Metabolomics analysis II. Preparation of biological samples prior to detection. TrAC, Trends Anal. Chem. 29: 120–127. Winder, C.L., Dunn, W.B., Schuler, S. et al. (2008). Global metabolic profiling of Escherichia coli cultures: an evaluation of methods for quenching and extraction of intracellular metabolites. Anal. Chem. 80: 2939–2948. Bort, J.A.H., Shanmukam, V., Pabst, M. et al. (2014). Reduced quenching and extraction time for mammalian cells using filtration and syringe extraction. J. Biotechnol. 182: 97–103. Beltran, A., Suarez, M., Rodriguez, M.A. et al. (2012). Assessment of compatibility between extraction methods for NMR- and LC/MS-based metabolomics. Anal. Chem. 84: 5838–5844. Orth, J.D., Thiele, I., and Palsson, B.O. (2010). What is flux balance analysis? Nat. Biotechnol. 28: 245–248. Krömer, J., Quek, L.-E., and Nielsen, L. (2009). 13 C-fluxomics: a tool for measuring metabolic phenotypes. Austr. Biochem. 40: 17–20. Cui, Q., Lewis, I.A., Hegeman, A.D. et al. (2008). Metabolite identification via the Madison metabolomics consortium database. Nat. Biotechnol. 26: 162–164. Smith, C.A., O’Maille, G., Want, E.J. et al. (2005). METLIN - a metabolite mass spectral database. Ther. Drug Monit. 27: 747–751. Tautenhahn, R., Cho, K., Uritboonthai, W. et al. (2012). An accelerated workflow for untargeted metabolomics using the METLIN database. Nat. Biotechnol. 30: 826–828. Zhu, Z.J., Schultz, A.W., Wang, J.H. et al. (2013). Liquid chromatography quadrupole time-of-flight mass spectrometry characterization of metabolites guided by the METLIN database. Nat. Protoc. 8: 451–460. Wishart, D.S., Tzur, D., Knox, C. et al. (2007). HMDB: the human metabolome database. Nucleic Acids Res. 35: D521–D526. Wishart, D.S., Knox, C., Guo, A.C. et al. (2009). HMDB: a knowledgebase for the human metabolome. Nucleic Acids Res. 37: D603–D610.

271

272

10 Metabolite Profiling of Mammalian Cells

67 Wishart, D.S., Jewison, T., Guo, A.C. et al. (2013). HMDB 3.0-the human

metabolome database in 2013. Nucleic Acids Res. 41: D801–D807. 68 Steinbeck, C., Conesa, P., Haug, K. et al. (2012). MetaboLights: towards

69

70

71 72

73 74 75

76

77 78

79 80

81

82

83

a new COSMOS of metabolomics data management. Metabolomics 8: 757–760. Salek, R.M., Haug, K., and Steinbeck, C. (2013). Dissemination of metabolomics results: role of MetaboLights and COSMOS. Gigascience 2: 1–3. Haug, K., Salek, R.M., Conesa, P. et al. (2013). MetaboLights-an open-access general-purpose repository for metabolomics studies and associated meta-data. Nucleic Acids Res. 41: D781–D786. Kale, N.S., Haug, K., Conesa, P. et al. (2016). MetaboLights: an open-access database repository for metabolomics data. Curr. Protoc. Bioinf. 53: 1–18. Salek, R.M., Conesa, P., Cochrane, K. et al. (2017). Automated assembly of species metabolomes through data submission into a public repository. Gigascience 6: 1–4. Kopka, J., Schauer, N., Krueger, S. et al. (2005). [email protected]: the golm metabolome database. Bioinformatics 21: 1635–1638. Skogerson, K., Wohlgemuth, G., Barupal, D.K., and Fiehn, O. (2011). The volatile compound BinBase mass spectral database. BMC Bioinf. 12: 321. Matsuda, F., Yonekura-Sakakibara, K., Niida, R. et al. (2009). MS/MS spectral tag-based annotation of non-targeted profile of plant secondary metabolites. Plant J. 57: 555–577. Sud, M., Fahy, E., Cotter, D. et al. (2016). Metabolomics workbench: an international repository for metabolomics data and metadata, metabolite standards, protocols, tutorials and training, and analysis tools. Nucleic Acids Res. 44: D463–D470. Misra, B.B. and van der Hooft, J.J.J. (2016). Updates in metabolomics tools and resources: 2014–2015. Electrophoresis 37: 86–110. Rost, H.L., Sachsenberg, T., Aiche, S. et al. (2016). OpenMS: a flexible open-source software platform for mass spectrometry data analysis. Nat. Methods 13: 741–748. Pfeuffer, J., Sachsenberg, T., Alka, O. et al. (2017). OpenMS - a platform for reproducible analysis of mass spectrometry data. J. Biotechnol. 261: 142–148. Horai, H., Arita, M., Kanaya, S. et al. (2010). MassBank: a public repository for sharing mass spectral data for life sciences. J. Mass Spectrom. 45: 703–714. Babushok, V.I., Linstrom, P.J., Reed, J.J. et al. (2007). Development of a database of gas chromatographic retention properties of organic compounds. J. Chromatogr. A 1157: 414–421. Stein, S.E. (1999). An integrated method for spectrum extraction and compound identification from gas chromatography/mass spectrometry data. J. Am. Soc. Mass Spectrom. 10: 770–781. Meyer, M.R., Peters, F.T., and Maurer, H.H. (2010). Automated mass spectral deconvolution and identification system for GC–MS screening for drugs, poisons, and metabolites in urine. Clin. Chem. 56: 575–584.

References

84 Aggio, R., Villas-Boas, S.G., and Ruggiero, K. (2011). Metab: an R package

85

86

87

88

89

90 91

92

93

94

95

96 97 98 99

100

for high-throughput analysis of metabolomics data generated by GC–MS. Bioinformatics 27: 2316–2318. Xia, J.G., Psychogios, N., Young, N., and Wishart, D.S. (2009). MetaboAnalyst: a web server for metabolomic data analysis and interpretation. Nucleic Acids Res. 37: W652–W660. Xia, J.G., Mandal, R., Sinelnikov, I.V. et al. (2012). MetaboAnalyst 2.0-a comprehensive server for metabolomic data analysis. Nucleic Acids Res. 40: W127–W133. Xia, J.G., Sinelnikov, I.V., Han, B., and Wishart, D.S. (2015). MetaboAnalyst 3.0-making metabolomics more meaningful. Nucleic Acids Res. 43: W251–W257. Dietrich, W., Rüdel, C.H., and Neumann, M. (1991). Fast and precise automatic baseline correction of one- and two-dimensional NMR spectra. J. Magn. Reson. (1969) 91: 1–11. Smith, C.A., Want, E.J., O’Maille, G. et al. (2006). XCMS: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal. Chem. 78: 779–787. Xi, Y.X. and Rocke, D.M. (2008). Baseline correction for NMR spectroscopic metabolomics data analysis. BMC Bioinf. 9: 324. Zhang, Z.M., Chen, S., and Liang, Y.Z. (2010). Baseline correction using adaptive iteratively reweighted penalized least squares. Analyst 135: 1138–1146. Lommen, A., Gerssen, A., Oosterink, J.E. et al. (2011). Ultra-fast searching assists in evaluating sub-ppm mass accuracy enhancement in U-HPLC/Orbitrap MS data. Metabolomics 7: 15–24. Robinson, M.D., De Souza, D.P., Keen, W.W. et al. (2007). A dynamic programming approach for the alignment of signal peaks in multiple gas chromatography-mass spectrometry experiments. BMC Bioinf. 8: 419. Yang, C., He, Z.Y., and Yu, W.C. (2009). Comparison of public peak detection algorithms for MALDI mass spectrometry data analysis. BMC Bioinf. 10: 4. Xiao, C.N., Hao, F.H., Qin, X.R. et al. (2009). An optimized buffer system for NMR-based urinary metabonomics with effective pH control, chemical shift consistency and dilution minimization. Analyst 134: 916–925. Burton, L., Ivosev, G., Tate, S. et al. (2008). Instrumental and experimental effects in LC–MS-based metabolomics. J. Chromatogr. B 871: 227–235. Weljie, A.M., Newton, J., Mercier, P. et al. (2006). Targeted profiling: quantitative analysis of 1 H NMR metabolomics data. Anal. Chem. 78: 4430–4442. Rasmussen, L.G., Savorani, F., Larsen, T.M. et al. (2011). Standardization of factors that influence human urine metabolomics. Metabolomics 7: 71–83. Hao, J., Liebeke, M., Astle, W. et al. (2014). Bayesian deconvolution and quantification of metabolites in complex 1D NMR spectra using BATMAN. Nat. Protoc. 9: 1416–1427. Broadhurst, D.I. and Kell, D.B. (2006). Statistical strategies for avoiding false discoveries in metabolomics and related experiments. Metabolomics 2: 171–196.

273

274

10 Metabolite Profiling of Mammalian Cells

101 Vinaixa, M., Samino, S., Saez, I. et al. (2012). A guideline to univariate sta-

102

103 104

105

106

107

108 109 110 111 112 113

114

115

116

117

tistical analysis for LC/MS-based untargeted metabolomics-derived data. Metabolites 2: 775–795. Fonville, J.M., Richards, S.E., Barton, R.H. et al. (2010). The evolution of partial least squares models and related chemometric approaches in metabonomics and metabolic phenotyping. J. Chemom. 24: 636–649. Bro, R. and Smilde, A.K. (2014). Principal component analysis. Anal. Methods 6: 2812–2831. Sugimoto, M., Kawakami, M., Robert, M. et al. (2012). Bioinformatics tools for mass spectroscopy-based metabolomic data processing and analysis. Curr. Bioinf. 7: 96–108. Mehrotra, B. and Mendes, P. (2006). Bioinformatics approaches to integrate metabolomics and other systems biology data. In: Plant Metabolomics, Biotechnology in Agriculture and Forestry, vol. 57 (eds. K. Saito, R.A. Dixon and L. Willmitzer). Berlin, Heidelberg: Springer-Verlag. Castle, A.L., Fiehn, O., Kaddurah-Daouk, R., and Lindon, J.C. (2006). Metabolomics standards workshop and the development of international standards for reporting metabolomics experimental results. Briefings Bioinf. 7: 159–165. Xia, J.G. and Wishart, D.S. (2010). MSEA: a web-based tool to identify biologically meaningful patterns in quantitative metabolomic data. Nucleic Acids Res. 38: W71–W77. Xia, J.G. and Wishart, D.S. (2010). MetPA: a web-based metabolomics tool for pathway analysis and visualization. Bioinformatics 26: 2342–2344. Ogata, H., Goto, S., Sato, K. et al. (1999). KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 27: 29–34. Goble, C. and Stevens, R. (2008). State of the nation in data integration for bioinformatics. J. Biomed. Inf. 41: 687–693. Gomez-Cabrero, D., Abugessaisa, I., Maier, D. et al. (2014). Data integration in the era of omics: current and future challenges. BMC Syst. Biol. 8: I1. Cavill, R., Jennen, D., Kleinjans, J., and Briede, J.J. (2016). Transcriptomic and metabolomic data integration. Briefings Bioinf. 17: 891–901. Altamirano, C., Cairo, J.J., and Godia, F. (2001). Decoupling cell growth and product formation in Chinese hamster ovary cells through metabolic control. Biotechnol. Bioeng. 76: 351–360. Luo, J., Vijayasankaran, N., Autsen, J. et al. (2012). Comparative metabolite analysis to understand lactate metabolism shift in Chinese hamster ovary cell culture process. Biotechnol. Bioeng. 109: 146–156. Burky, J.E., Wesson, M.C., Young, A. et al. (2007). Protein-free fed-batch culture of non-GS NS0 cell lines for production of recombinant antibodies. Biotechnol. Bioeng. 96: 281–293. Pan, X., Streefland, M., Dalm, C. et al. (2017). Selection of chemically defined media for CHO cell fed-batch culture processes. Cytotechnology 69: 39–56. Ma, N.N., Ellet, J., Okediadi, C. et al. (2009). A single nutrient feed supports both chemically defined NS0 and CHO fed-batch processes: improved productivity and lactate metabolism. Biotechnol. Progr. 25: 1353–1363.

References

118 Zagari, F., Jordan, M., Stettler, M. et al. (2013). Lactate metabolism shift in

119

120

121

122

123

124

125

126 127

128

129

130

131

132

CHO cell culture: the role of mitochondrial oxidative activity. New Biotechnol. 30: 238–245. Zhang, L., Shen, H., and Zhang, Y.X. (2004). Fed-batch culture of hybridoma cells in serum-free medium using an optimized feeding strategy. J. Chem. Technol. Biotechnol. 79: 171–181. Grainger, R.K. and James, D.C. (2013). CHO cell line specific prediction and control of recombinant monoclonal antibody N-glycosylation. Biotechnol. Bioeng. 110: 2970–2983. Baker, K.N., Rendall, M.H., Hills, A.E. et al. (2001). Metabolic control of recombinant protein N-glycan processing in NS0 and CHO cells. Biotechnol. Bioeng. 73: 188–202. Burleigh, S.C., van de Laar, T., Stroop, C.J.M. et al. (2011). Synergizing metabolic flux analysis and nucleotide sugar metabolism to understand the control of glycosylation of recombinant protein in CHO cells. BMC Biotechnol. 11: 95. Templeton, N., Xu, S., Roush, D.J., and Chen, H. (2017). 13 C metabolic flux analysis identifies limitations to increasing specific productivity in fed-batch and perfusion. Metab. Eng. 44: 126–133. Khoo, S.H.G. and Al-Rubeai, M. (2009). Metabolic characterization of a hyper-productive state in an antibody producing NS0 myeloma cell line. Metab. Eng. 11: 199–211. Yip, S.S.M., Zhou, M.X., Joly, J. et al. (2014). Complete knockout of the lactate dehydrogenase A gene is lethal in pyruvate dehydrogenase kinase 1, 2, 3 down-regulated CHO cells. Mol. Biotechnol. 56: 833–838. Henry, O. and Durocher, Y. (2011). Enhanced glycoprotein production in HEK-293 cells expressing pyruvate carboxylase. Metab. Eng. 13: 499–507. Vallee, C., Durocher, Y., and Henry, O. (2014). Exploiting the metabolism of PYC expressing HEK293 cells in fed-batch cultures. J. Biotechnol. 169: 63–70. Elias, C.B., Carpentier, E., Durocher, Y. et al. (2003). Improving glucose and glutamine metabolism of human HEK293 and Trichoplusiani insect cells engineered to express a cytosolic pyruvate carboxylase enzyme. Biotechnol. Progr. 19: 90–97. Irani, N., Beccaria, A.J., and Wagner, R. (2002). Expression of recombinant cytoplasmic yeast pyruvate carboxylase for the improvement of the production of human erythropoietin by recombinant BHK-21 cells. J. Biotechnol. 93: 269–282. Irani, N., Wirth, M., van den Heuvel, J., and Wagner, R. (1999). Improvement of the primary metabolism of cell cultures by introducing a new cytoplasmic pyruvate carboxylase reaction. Biotechnol. Bioeng. 66: 238–246. Toussaint, C., Henry, O., and Durocher, Y. (2016). Metabolic engineering of CHO cells to alter lactate metabolism during fed-batch cultures. J. Biotechnol. 217: 122–131. Fogolin, M.B., Wagner, R., Etcheverrigaray, M., and Kratje, R. (2004). Impact of temperature reduction and expression of yeast pyruvate carboxylase on hGM-CSF-producing CHO cells. J. Biotechnol. 109: 179–191.

275

276

10 Metabolite Profiling of Mammalian Cells

133 Gupta, S.K., Srivastava, S.K., Sharma, A. et al. (2017). Metabolic engineering

134

135

136 137

138

139

140

141

142

143

144

145 146

147

148

of CHO cells for the development of a robust protein production platform. PLoS One 12: e018455. Dorai, H., Kyung, Y.S., Ellis, D. et al. (2009). Expression of anti-apoptosis genes alters lactate metabolism of Chinese hamster ovary cells in culture. Biotechnol. Bioeng. 103: 592–608. Ha, T.K., Jeon, M.K., Yu, D.Y., and Lee, G.M. (2013). Effect of Bcl-x(L) overexpression on lactate metabolism in Chinese hamster ovary cells producing antibody. Biotechnol. Progr. 29: 1594–1598. Gogvadze, V., Orrenius, S., and Zhivotovsky, B. (2008). Mitochondria in cancer cells: what is so special about them? Trends Cell Biol. 18: 165–173. Martinez, V.S., Dietmair, S., Quek, L.E. et al. (2013). Flux balance analysis of CHO cells before and after a metabolic switch from lactate production to consumption. Biotechnol. Bioeng. 110: 660–666. Ghorbaniaghdam, A., Chen, J.K., Henry, O., and Jolicoeur, M. (2014). Analyzing clonal variation of monoclonal antibody-producing CHO cell lines using an in silico metabolomic platform. PLoS One 9: e90832. Carinhas, N., Duarte, T.M., Barreiro, L.C. et al. (2013). Metabolic signatures of GS-CHO cell clones associated with butyrate treatment and culture phase transition. Biotechnol. Bioeng. 110: 3244–3257. Liste-Calleja, L., Lecina, M., Lopez-Repullo, J. et al. (2015). Lactate and glucose concomitant consumption as a self-regulated pH detoxification mechanism in HEK293 cell cultures. Appl. Microbiol. Biotechnol. 99: 9951–9960. Mulukutla, B.C., Gramer, M., and Hu, W.S. (2012). On metabolic shift to lactate consumption in fed-batch culture of mammalian cells. Metab. Eng. 14: 138–149. Barnes, L.M., Moy, N., and Dickson, A.J. (2006). Phenotypic variation during cloning procedures: analysis of the growth behavior of clonal cell lines. Biotechnol. Bioeng. 94: 530–537. Chusainow, J., Yang, Y.S., Yeo, Y.H.M. et al. (2009). A study of monoclonal antibody-producing CHO cell lines: what makes a stable high producer? Biotechnol. Bioeng. 102: 1182–1196. Stansfield, S.H., Allen, E.E., Dinnis, D.M. et al. (2007). Dynamic analysis of GS-NS0 cells producing a recombinant monoclonal antibody during fed-batch culture. Biotechnol. Bioeng. 97: 410–424. Zhang, L., Luo, S., and Zhang, B.L. (2016). Glycan analysis of therapeutic glycoproteins. mAbs 8: 205–215. Cosgrave, E.F.J., Struwe, W.B., Hayes, J.M. et al. (2013). N-linked glycan structures of the human Fc gamma receptors produced in NS0 cells. J. Proteome Res. 12: 3721–3737. Brooks, S.A. (2004). Appropriate glycosylation of recombinant proteins for human use - implications of choice of expression system. Mol. Biotechnol. 28: 241–255. Sealover, N.R., Davis, A.M., Brooks, J.K. et al. (2013). Engineering Chinese hamster ovary (CHO) cells for producing recombinant proteins with simple glycoforms by zinc-finger nuclease (ZFN)-mediated gene knockout of

References

149

150

151

152

153

154

155

156

157

mannosyl (alpha-1,3-)-glycoprotein beta-1,2-N-acetylglucosaminyltransferase (Mgat1). J. Biotechnol. 167: 24–32. Liu, J.T., Wang, J., Fan, L. et al. (2015). Galactose supplementation enhance sialylation of recombinant Fc-fusion protein in CHO cell: an insight into the role of galactosylation in sialylation. World J. Microbiol. Biotechnol. 31: 1147–1156. Gramer, M.J., Eckblad, J.J., Donahue, R. et al. (2011). Modulation of antibody galactosylation through feeding of uridine, manganese chloride, and galactose. Biotechnol. Bioeng. 108: 1591–1602. Wong, D.C.F., Wong, K.T.K., Nissom, P.M. et al. (2006). Targeting early apoptotic genes in batch and fed-batch CHO cell cultures. Biotechnol. Bioeng. 95: 350–361. Yang, Z., Wang, S.J., Halim, A. et al. (2015). Engineered CHO cells for production of diverse, homogeneous glycoproteins. Nat. Biotechnol. 33: 842–845. Ibanez, A.J., Fagerer, S.R., Schmidt, A.M. et al. (2013). Mass spectrometry-based metabolomics of single yeast cells. Proc. Natl. Acad. Sci. U.S.A. 110: 8790–8794. Pirman, D.A., Efuet, E., Ding, X.P. et al. (2013). Changes in cancer cell metabolism revealed by direct sample analysis with MALDI mass spectrometry. PLoS One 8: e61379. Bhandari, D.R., Schott, M., Rompp, A. et al. (2015). Metabolite localization by atmospheric pressure high-resolution scanning microprobe matrix-assisted laser desorption/ionization mass spectrometry imaging in whole-body sections and individual organs of the rove beetle Paederus riparius. Anal. Bioanal. Chem. 407: 2189–2201. Wiseman, J.M., Ifa, D.R., Song, Q.Y., and Cooks, R.G. (2006). Tissue imaging at atmospheric pressure using desorption electrospray ionization (DESI) mass spectrometry. Angew. Chem. Int. Ed. 45: 7188–7192. Yue, S.H. and Cheng, J.X. (2016). Deciphering single cell metabolism by coherent Raman scattering microscopy. Curr. Opin. Chem. Biol. 33: 46–57.

277

279

11 Current Considerations and Future Advances in Chemically Defined Medium Development for the Production of Protein Therapeutics in CHO Cells Wai Lam W. Ling Biologics Process R&D, Merck Research Laboratories, Merck Sharp & Dohme Corp, Kenilworth, NJ, USA

11.1 Introduction Protein therapeutics continue to grow globally as potent medicines treating major disease areas of oncology, immunology, hematology, and others with market shares of over $200 billion in 2016 (http://www.prnewswire.com/news-releases/global-biologics-market-will-be-worth-us479-752-million-by-2024-globalindustry-analysis-size-share-growth-trends-and-forecast-2016—2024-tmr-5961 50181.html) and projects to be more than $450 billion by 2024 (http://www .transparencymarketresearch.com/global-biologics-market.html). Although monoclonal antibodies are the dominating biotherapeutics, recombinant proteins, bispecific antibodies, and antibody–drug conjugates also maintain steady growth. Biotherapeutics are produced in biological systems such as mammalian, microbial, and yeast cells, with Chinese hamster ovary (CHO) cells as the predominated expression system in use [1]. Protein-expressing CHO cells are propagated and scaled up in cell culture media, and the protein product is produced and secreted into the extracellular milieu. This overview will focus on medium development that supports cell growth and biotherapeutic production in CHO cells. Although the following points are also critical in establishing practical and robust cell culture media, these topics are out of scope for this chapter discussion: (i) areas that are related to patent protection and their impact on media implementation for commercial biotherapeutic manufacture, (ii) media components that may impact protein purification efficiency and final yield, and (iii) regulatory submission strategy (i.e. drug master files or DMFs).

11.2 Traditional Approach to Medium Development 11.2.1

Cell Line Selection

As significant work is needed to develop media, most biologics manufacturers tend to develop a complementary set of platform cell line and media for protein production [2, 3]. However, it is not uncommon that some manufacturers have Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

280

11 Chemically Defined Medium Development for the Production of Protein Therapeutics

multiple cell line systems for different modalities (e.g. monoclonal antibodies and bispecific antibodies) [4]. In such cases, it may be required to develop separate medium sets for different cell line/modality systems. Before chemically defined (CD) medium development, selection of model cell lines and the preparation of frozen research cell banks are needed to minimize variability [5] that may be linked to culture age during medium development. To develop media that are applicable for multiple cell lines, model cell lines expressing different molecules or different cell lines expressing the same molecule at various productivities should be employed. This strategy is to address potential diverse culture performance across different cell systems to ensure that the newly developed media would be able to support a variety of cell lines [6]. The first medium developed should support cell propagation (seed train). During cell line development (CLD), the protein expressing cell lines are typically propagated in a previously internal developed medium or a commercially available medium. To minimize additional adaptation into media that are utilized in process development and production, it is best that the same medium is used in CLD and process development to maximize work flow efficiency [7]. If the CLD medium differs from the process development or production media, the development target of CD medium for biotherapeutics manufacture should demonstrate comparable cell culture performance in doubling times, growth rates, and cell line stability to those obtained in the medium used during CLD. Subsequent optimization to enhance culture performance and productivity necessitate spent medium analysis to balance medium nutrient levels for maximizing cell growth and productivity, while concomitantly reducing toxic metabolite build-up and minimizing nutrient depletion. To initiate base medium development, one of the well-documented traditional CD formulations can be used as a starting point [8, 9]. Supplementation of nutrients should be considered for these early generation lean formulations. Medium component selection should be conducted using design of experiment (DoE) screening designs, followed by response surface design for formulation optimization [10, 11]. 11.2.2

Design and Optimization

DoE approaches have been used extensively in medium component screening and formulation optimization. A practical approach for most labs performing component screens is the utilization of DoE such as factorial, Plackett–Burman, or definitive designs for multiple component factors, followed by response surface designs to narrow the optimal concentration ranges. Experimental parameters to be monitored include cell growth such as changes in cell doubling times or growth rates, cell diameters, production titer or specific productivity, metabolic changes in glucose, lactate, ammonia, amino acid levels, and product quality attributes such as glycosylation, charge variant profiles, and aggregate levels. Parameters related to cell growth serve as indicators for seed train efficiency. Cell diameter measurements provide information on the status of cell cycle as well as on the lipid and membrane-associated protein composition and distribution in the intracellular organellar (i.e. endoplasmic reticulum, Golgi,

11.2 Traditional Approach to Medium Development

Medium development and optimization

Component adjustments based on metabolite usage

Gene of interest Cell line Process

Cell culture

• • • Cell count Metabolite analyses •

Cell density and viability Metabolic profiles Protein productivity Product quality

Titer analysis Quality analysis

Figure 11.1 Standard workflow for medium development.

and vesicular transport) and plasma membranes. Productivity measurements link productive state of the culture. Metabolite analyses show the degree of efficiency in nutrient usage. Product quality assessments give the final outcome of the material produced in the test conditions. Most labs are equipped with appropriate instrumentations to support these analyses. Iterative formulation modifications through component adjustments and experimental evaluations that include stoichiometric nutrient balancing would lead to workable medium formulations (Figure 11.1). Several methods have been well published to address nutrient balance [12, 13] and metabolic flux [14] for iterative medium development. This strategy offers a direct and practical approach that links medium development to measureable extracellular matrices, such as spent medium component concentrations and cell densities. For development of media for seed train expansion, parameters such as specific growth rate, cell viability, cell diameters, metabolic profiles, and passaging productivity should be monitored for extended cell propagation to ∼100 doublings (or PDL, population doubling level) to ensure that the medium can support cell line selection, GMP cell banking, and inoculum scale-up to commercial-scale bioreactors. In general, as cell lines age, doubling times tend to be shorter than their younger counterparts. Therefore, passaging productivity monitoring can be a way to assess whether intrinsic specific productivity has been reduced with cell age in the test media. Cell age assessment is used to assess stability of the cell lines. It is important to distinguish potential differences of cell line inherent instability from medium composition. Considerations for maintaining selective pressure in seed train media may be used to address this. For Chinese hamster ovary-dihydrofolate reductase (CHO-DHFR) systems, removal of hypoxanthine and thymidine (HT) and/or addition of methotrexate (MTX) have been used to maintain selection pressure. For Chinese hamster ovary-glutamine synthetase (CHO-GS) systems, removal of glutamine and/or addition of methionine sulfoximine (MSX) are involved in ensuring selection pressure and gene amplification. Table 11.1 shows medium component classifications with their demonstrated cellular functions that are typically present in CHO cell culture media, and their effective concentration levels are described in their respective references. Subsequent evaluation should include testing of the identified media in at least three cell lines expressing different biologic molecules of the same modalities (i.e. monoclonal antibodies against different targets). The rationale for multiple

281

282

11 Chemically Defined Medium Development for the Production of Protein Therapeutics

Table 11.1 Cellular functions of medium components. Component class

Function

References

Amino acids

Protein production and glycosylation Cell mass and cellular function

Fan et al. 2015 [15] Xing et al. 2011 [16] Gonzalez-Leal et al. 2011 [17] Hagrot et al. 2016 [18]

Vitamins

Genetic stability, cell growth, and proliferation

Arigony et al. 2013 [19] Ishaque and Al-Rubeai 2002 [20] Ducommun et al. 2001 [21]

Trace metals

Cell growth and cell cycle regulation Cellular enzyme activities

Arigony et al. 2013 [19] Yuk et al. 2014 [22]

Carbon sources

Cell growth and cellular metabolism Protein glycosylation

Altamirano et al. 2000 [23] Li et al. 2012 [24] Hossler et al. 2017 [25]

Proteins, growth factors, and peptides

Cell growth

Lim et al. 2013 [26] Kang et al. 2012 [27] Chun et al. 2003 [28] Vander Heiden et al. 2001 [29] Gospdarowicz and Moran 1976 [30]

Salts and buffers

Osmolality balance Cell growth and productivity pH control

Reinhart et al. 2015 [31] Lee and Koo 2010 [32] Zhu et al. 2005 [33]

Nucleotides and nucleosides

Cell proliferation Glycosylation process

Zhang et al. 2013 [34] Morrone et al. 2003 [35]

Lipids and fatty acids

Membrane structure and integrity

Ibarguren et al. 2014 [36]

Antioxidants

Reactive oxygen species (ROS) (O2 *-) and H2 O2 reduction

Halliwell 2014 [37] Luo et al. 2014 [38]

Polyamines

Cell proliferation Translational control

Mandal et al. 2013 [39] Landau et al. 2010 [40]

Shear protectants

Reduction of bubble-associated cell damage in sparged bioreactors

Chang et al. 2017 [41] Peng et al. 2014 [42] Hu et al. 2008 [43]

molecules is to ensure that the developed media would be sufficient in supporting protein-expressing cell lines from the same host cell system and to ensure their application across multiple biotherapeutic expressing cell lines. 11.2.3

Process Consideration

Decisions on cell lines and production processes drive the media development strategy. To accommodate multiple production processes, such as fed batch and

11.2 Traditional Approach to Medium Development

continuous perfusion, multiple medium types are necessary. One approach to support this multiple process strategy is the development of medium modules that support different aspects of process-specific requirements. The advantage is that once the medium modules are developed and supplies are procured on site, they can be flexibly mixed to support individualized process needs. For late-stage development, the modules can be consolidated to optimal levels to generate a single or fewer bulk supplies for commercial manufacturing facility. For traditional fed-batch production processes, two classes of media, basal and feed media, are typically used. Glucose, a carbon source, is typically added separately as needed and maintained as a target range for the culture duration. The compositions of the basal and feed medium formulations are designed to be complementary. The basal medium is utilized for seed train expansion from vial thawing through inoculation in the production bioreactor and is typically leaner in component composition than its feed medium counterpart. The objective for this basal medium is to maintain cell growth rate, culture doubling time, and high cell viability and to maintain the performance with cell age and ability to generate minimal toxic metabolites such as lactate and ammonia during cell propagation. The osmolality of the base medium is usually at 270–330 mOsm/kg [32]. The other class is the feed media that are used to support higher cell densities and productivities in the final seed bioreactor (N-1) and production (N) bioreactor. The feed media tend to be richer in components and have higher osmolality (>900 mOsm/kg) than the basal medium with the objective to increase cell density and productivity. Optimization of feed strategies as a part of cell culture process development may involve evaluating feed concentration, time of addition, bolus, or continuous addition, while balancing cellular metabolism, cell viability, and productivity. As optimal feed strategy and media requirements may vary between cell lines and biotherapeutics, a modular media development approach would provide a flexible process development solution. Perfusion processes, previously implemented for production of labile or unstable proteins [44], are being leveraged as a method to achieve a high cell density of ≥80 × 106 cells/ml and high product yield. Continuous perfusion processes can be run for months with product recovery occurring periodically or continuously throughout the duration. The media supporting this type of production strategy need to possess components similar to the basal media used for seed train expansion with augmentation of critical components from feed media used in fed-batch production to ensure maintenance of high cell density and improved cell-specific productivity [45]. The medium module development approach provides the flexibility for an organization to switch between different production processes. A combination of perfusion at N-1 stage utilizing cell retention devices (e.g. alternating tangential flow filtration [ATF] or tangential flow filtration [TFF] [46, 47]) coupled to the fed-batch production [48] is another option of production processes. There are advantages to the implementation of a short-term perfusion cell expansion of N-1 step coupled with a fed-batch production compared to a continuous perfusion process: (i) basal media used for cell expansion will likely be applicable for N-1 perfusion with minimal medium component adjustments; (ii) N-1 perfusion increases viable cell density with overall lower medium volume requirement than continuous perfusion processes; and (iii) short-term

283

284

11 Chemically Defined Medium Development for the Production of Protein Therapeutics

N-1 perfusion step of 5–7 days is operationally less challenging than continuous perfusion that can last for 30 days or more. However, fed-batch production media for high-density cultures are more challenging to develop as prolonging cell viability while improving specific productivity requires a balance of cellular metabolism controlled by media composition and feed strategy. Culture conditions can affect product quality. For example, controlled culture pH [49, 50], temperature shift [51–55], dissolved oxygen [56], and pCO2 [57] have been demonstrated to affect product quality parameters, such as glycosylation, aggregate, and charge variant profiles. Therefore, medium formulations, production process conditions, and bioreactor control should be considered collectively during overall process development. 11.2.4

Additional Considerations in Medium Development

Media that are prepared at pH extremes can be challenging for process operations as they requires fine control to minimize local pH differences during media addition to the bioreactor. Medium solubility, stability, and shelf life are also critical to media preparation and hold. As such, researchers have sought and explored alternative medium components to mitigate these concerns. For example, phospho-tyrosine disodium salt, which replaces l-tyrosine, has been demonstrated to have a six-month shelf life at as high as 12.4 mM in a pH neutral stock solution [58]. S-sulfocysteine, which replaces l-cysteine, has been demonstrated to have >3 month shelf life at 15 mM in a pH neutral solution [59]. However, specialized component sourcing and long-term supply strategy should be in place before including these components in the final medium formulations. In addition, riboflavin, vitamin B2 , is more sensitive to photodegradation in N2 -saturated media than O2 -saturated media, whereas pyridoxine, vitamin B6 , degrades in UV irradiation regardless of N2 - or O2 -saturated media [60]. Tryptophan is more sensitive to photodegradation in O2 -saturated media than N2 -saturated media. When exposed to light, tryptophan is oxidized to kynurenine and formic acid [61]. Therefore, storage conditions should be considered to minimize oxidation or degradation of these components in the media.

11.3 Future Perspectives for Medium Development 11.3.1

Systems Biology and Synthetic Biology

Systems biology has been applied to multiple areas of biological research to gain a better understanding of disease origins [62–66] and to identify potential new drug targets based on disease biology. Bioprocesses that utilize only one cell type, CHO cells, are simpler systems than human and animal systems. However, the underlying biochemistry and molecular biology are shared between whole animal and CHO cell systems. Therefore, leveraging the systems biology knowledge from animal systems and applying these strategic systems biological tools to bioprocess development can be valuable in gaining better understanding of cell culture performance and subsequently better control of the overall

11.3 Future Perspectives for Medium Development

Table 11.2 Systems biology strategy for bioprocess development. Systems biology approaches

Goals

Current methods [67, 68]

Genomics

Genome analysis (Conesa and Mortzavi 2014 [69])

DNA microarray Transitional structural chemogenomics

Transcriptomics

mRNA identification and quantification analysis (Da Hora Junior et al. 2012 [70])

Microarray Next generation sequencing

Proteomics

Protein identification and quantification analysis (Wang et al. 2016 [71])

Separation: • Flow cytometry • 1D or 2D electrophoresis • Differential image gel electrophoresis • Liquid chromatography Detection: • Mass spectrometry

Metabolomics

Metabolite identification and quantification analysis (Nielsen and Jewett 2007 [72])

Separation: • Capillary electrophoresis • Gas chromatography • Liquid chromatography Detection: • Mass spectrometry • Nuclear magnetic resonance

Phenomics

Organism-wide phenotype analysis (Houle et al. 2010 [73])

Data integration and relationship • Omics data • Biological networks

production processes (Table 11.2). More recently, in conjunction with in-process monitoring, biopharmaceutical developers have incorporated systems biology tools [74], such as genomics, transcriptomics, proteomics, and secretomics, to study the genetic, transcriptional, translational, or secretory functions in their production processes (Figure 11.2). For example, proteomic analysis of CHO cultured with high glucose was targeted for supporting genetic engineering of robust cell lines [75], transcriptomic differences of temperature-shifted CHO cell lines was used as a predictor for cell-specific productivity [76], and proteomic analyses of high- and low-producing CHO cell lines grown in bioreactors have a role in cell line optimization [77]. Efficient utilization of medium components for cell growth resulted in metabolic changes that can be assessed using metabolic flux analysis [14]. Through these profiles, predictive models may be built to help modify medium formulations that may regulate appropriate pathways for enhanced cell culture performance. The utilization of this holistic “omics” profiling approach requires establishing analytical tools to monitor and examine potential physiological and extracellular changes for baseline and modified processes. Large data sets resulting from this work flow require sufficient data

285

286

11 Chemically Defined Medium Development for the Production of Protein Therapeutics

Medium development and optimization

Component adjustments

Model building and prediction

Pathway profiling Transcriptomics

Gene of interest Cell line

Proteomics

Cell culture

Process

Metabolomics Other ‘omics… Product quality Cellular characterization

Figure 11.2 Systems biology approach for medium development.

management and bioinformatics capability to generate models to predict process performance as a result of media component or process changes. Instituting a systems biology workflow can be an arduous task, and the “omics” data analysis and interpretation are complex. Two major challenges are (i) establishing the structured platforms for acquiring and manipulating the data and (ii) analyzing the data from a systems perspective that takes into account the biological complexity. Acquiring the data requires a spectrum of analytical tools, which generate large amounts of data that must be efficiently stored and analyzed. Analyzing the data in a systems context requires skilled personnel with access to appropriate pathway alignment database and bioinformatics tools for biological networks, such as protein–protein interaction (PPI), regulatory, signaling, metabolic networks, and others. Utilization of bioinformatics methods and computational analysis are most critical to link large volumes of numeric data to the specific data that are relevant to regulation of biological cellular pathways. Currently, a growing list of technologies, techniques, and visualization tools are being developed to improve the efficiency in the overall biological system understanding [78]. Systems biology approaches hold promise to provide a better understanding of the effects of environmental factors such as nutrient components, toxic metabolites, medium pH, and bioreactor operating conditions that may have on cellular responses such as DNA replication, transcription, translation, or protein secretion (Figure 11.3). Omics profiles will provide information on efficiency of the cellular pathway, rate-limiting steps, and bottlenecks in protein production and release. When pathway inefficiencies can be identified, attenuation, reversal, or redesign of the pathways can be accomplished via synthetic biology approaches. Synthetic biology is the study and utilization of non-natural and disruptive technologies that involve designing, building, testing, and controlling (fine tuning) protein production processes. “Omics” knowledge obtained through

11.3 Future Perspectives for Medium Development

Nutrient levels pH

Temperature

Transfection conditions

Toxic metabolites

Metabolic state Energy content

Substrate levels

Cell age

Shear stress

DNA

mRNA

Protein

Replication

Transcription

Translation

Enzyme activity

Secretion Trafficking

Oxidative state

Signaling pathways

Ammonia

O2 CO2

Osmolality

Figure 11.3 Extracellular and intracellular effects on biologics production.

systems biology methods provides inputs for synthetic biology. To date, synthetic bioengineers have built synthetic elements or modules giving rise to specific cellular behaviors and changes in cellular performance. For example, in the genetic front, Kramer et al. developed a transgene switch that toggles the on–off expression of a model glycoprotein transfected in CHO cells [79], Bashor et al. engineered the Ste5 scaffold to control MAPK signaling pathway [80], Park et al. introduced silkworm hemolymph 30Kc6 gene [81] for a specific plasma lipoprotein to inhibit human cell apoptosis [82] and increase protein productivity [83], and Brown and James designed engineered promoters for gene transcription [84] in a CHO cell system, which were demonstrated to respond to changes in bioreactor conditions. In the metabolomics front, a metabolic profiling for amino acid changes and spent media analyses in cell culture were used to optimize media formulations and feed strategy [85, 86]. Metabolic flux analysis for two semisteady continuous CHO cultures [16] identified changes in amino acid levels, which led to the development of optimized media resulting in reduced lactate and ammonia, and improved viable cell density and productivity. Principal component analysis comparing different media has been employed to develop new feed strategies [87]. Cell culture medium components have been shown to link to genetic stability and cell viability of CHO-K1 cells [19]. Robitaille et al. developed an in silico kinetic model based on intracellular and extracellular metabolic fluxes and culture performance [88]. Transcriptional, translational, metabolic, and cell–cell communication control may be the result of synthetic biology efforts for an overall systems biology network. When these modules are integrated together into a whole system along with CD media, protein expression not only can be controlled but also can be fine-tuned to adjust to different aspects of cell culture behaviors and productivity.

287

288

11 Chemically Defined Medium Development for the Production of Protein Therapeutics

Similar to the module design for medium development to accommodate varying cell lines, production processes, and molecular modalities, module design in the synthetic biology area provides flexibility to optimize the production cell lines to meet specific production preferences. A critical goal in biologics manufacturing is to ensure that the different production processes not only result in improved productivity but also result in consistent product quality. Both of these outputs rely on a specific cellular and physiological status as well as environmental controls. Chemically defined medium development remains as a critical development activity for improved productivity and product quality for biologics manufacturing. A modular medium concept supports implementation of different cell lines, molecular modalities, and production processes. As the “omics” technology and computational efforts become prevalent in routine process development settings, systems biology and synthetic biology strategies will likely be incorporated into routine biologics development workflow to optimize CHO cell functions. Modular synthetic biology approaches could offer both flexible and targeted approaches for CHO cell engineering. The combination of both modular approaches, medium development and synthetic biology, is likely to minimize the complexity of current and future needs in biologics production.

Acknowledgment The author is grateful to William Buggele, Marina Goldfeld, Brian Kwan, Gregg Nyberg, and David Roush for critical review of this manuscript.

Conflict of Interest The author is an employee of Merck Sharp & Dohme Corp. and has no other interest to declare.

References 1 Ling, W.L.W. (2015). Development of protein-free medium for therapeutic

protein production in mammalian cells: recent advances and perspectives. Pharm. Bioprocess. 3: 215–226. https://doi.org/10.4155/PBP.15.8. 2 Kober, L., Zehe, C., and Bode, J. (2013). Optimized signal peptides for the development of high expressing CHO cell lines. Biotechnol. Bioeng. 110: 1164–1173. 3 Yuk, I.H., Nishihara, J., Walker, D. Jr., et al. (2015). More similar than different: host cell protein production using three null CHO cell lines. Biotechnol. Bioeng. 112: 2068–2083. 4 Lalonde, M.-E. and Durocher, Y. (2017). Therapeutic glycoprotein production in mammalian cells. J. Biotechnol. 251: 128–140. https://doi.org/10.1016/j .jbiotec.2017.04.028.

References

5 He, L. and Frye, C. (2016). A risk-based genetic characterization strategy for

6

7

8

9

10

11 12 13

14

15

16

17

18

19

recombinant CHO cell lines used for clinical and commercial applications. BioPharm Int. 29: 32–40. Pan, X., Streefland, M., Dalm, C. et al. (2017). Selection of chemically defined media for CHO cell fed-batch culture processes. Cytotechnology 69: 39–56. https://doi.org/10.1007/s10616-016-0036-5. Wagner, M., Arunakumari, A., Dai, X.-P. et al. (2009). The impact of cell culture medium on cell line and process development timelines and strategies. BioPharm Int. 2009 (4). Yang, Z. and Xiong, H.-R. (2012). Culture conditions and types of growth media for mammalian cells. In: Biomedical Tissue Culture, 1–16. InTech https://doi.org/10.5772/52301. Parampalli, A., Eskridge, K., Smith, L. et al. (2007). Development of Serum-free media in CHO-DG44 cells using a central composite statistical design. Cytotechnology 54: 57–68. https://doi.org/10.1007/s10616-007-9074-3. Ling, W.L.W., Bai, Y., Cheng, C. et al. (2015). Development and manufacturability assessment of chemically-defined medium for the production of protein therapeutics in CHO cells. Biotechnol. Progr. 31: 1163–1171. Jiang, Z., Droms, K., Geng, Z. et al. (2012). Fed-batch cell culture process optimization. BioProcess Int. 10: 40–45. Xie, L. and Wang, D.I.C. (1994). Stoichiometric analysis of animal cell growth and its application in medium design. Biotechnol. Bioeng. 43: 1164–1174. Xie, L. and Wang, D.I.C. (1994). Fed-batch cultivation of animal cells using different medium design concepts and feed strategies. Biotechnol. Bioeng. 43: 1175–1189. Follstad, B.D., Balcarcel, R.R., Sephanopoulos, G., and Wang, D.I.C. (1999). Metabolic flux analysis of hybridoma continuous culture steady state multiplicity. Biotechnol. Bioeng. 63 (6): 675–683. Fan, Y., Del Val, I.J., Muller, C. et al. (2015). Amino acid and glucose metabolism in fed-batch CHO cell culture affects antibody production and glycosylation. Biotechnol. Bioeng. 112: 521–535. https://doi.org/10.1002/bit .25450. Xing, Z., Kenty, B., Koyrakh, I. et al. (2011). Optimizing amino acid composition of CHO cell culture media for a fusion protein production. Process Biochem. 46: 1423–1429. https://doi.org/10.1016/j.procbio.2011.03.014. Gonzalez-Leal, I.J., Carrillo-Cocom, L.M., Ramirez-Medrano, A. et al. (2011). Use of a Plackett–Burman statistical design to determine the effect of selected amino acids on monoclonal antibody production in CHO cells. Biotechnol. Progr. 27: 1709–1717. https://doi.org/10.1002/btpr.674. Hagrot, E., Oddsdottir, H.A., Hosta, J.G. et al. (2016). Poly-pathway model, a novel approach to simulate multiple metabolic states by reaction network-based model – application to amino acid depletion in CHO cell culture. J. Biotechnol. 228: 37–49. https://doi.org/10.1016/j.jbiotec.2016.03.015. Arigony, A.L.V., De Oliverira, I.M., Machado, M. et al. (2013). The influence of micronutrients in cell culture: a reflection on viability and genomic stability. BioMed Res. Int. 2013: 1–23. https://doi.org/10.1155/2013/597282.

289

290

11 Chemically Defined Medium Development for the Production of Protein Therapeutics

20 Ishaque, A. and Al-Rubeai, M. (2002). Role of vitamins in determining apop-

21

22

23

24

25

26

27

28

29

30 31

32

33

tosis and extent of suppression by bcl-2 during hybridoma cell culture. Apoptosis 7: 231–239. Ducommun, P., Ruffieux, P.-A., von Stockar, U., and Marison, I. (2001). The role of vitamins and amino acids on hybridoma growth and monoclonal antibody production. Cytotechnol 37: 65–73. Yuk, I.H., Russell, S., Tang, Y. et al. (2014). Effects of copper on CHO cells: cellular requirements and product quality considerations. Biotechnol. Progr. 31 (1): 226–238. https://doi.org/10.1002/btpr.2004. Altamirano, C., Paredes, C., Cairo, J.J., and Godia, F. (2000). Improvement of CHO cell culture medium formulation: simultaneous substitution of glucose and glutamine. Biotechnol. Progr. 16: 69–75. Li, J., Wong, C.L., Vijayasankaran, N. et al. (2012). Feeding lactate for CHO cell culture processes: impact on culture metabolism and performance. Biotechnol. Bioeng. 109: 1173–1186. https://doi.org/10.1002/bit.24389. Hossler, P., Racicot, C., Chumsae, C. et al. (2017). Cell culture media supplementation of infrequently used sugars for the targeted shifting of protein glycosylation profiles. Biotechnol. Progr. 33: 511–522. https://doi.org/10.1002/ btpr.2429. Lim, U.M., Yap, M.G.S., Lim, Y.P. et al. (2013). Identification of autocrine growth factors secreted by CHO cells for applications in single-cell cloning media. J. Proteome Res. 12: 3496–3510. https://doi.org/10.1021/pr400352n. Kang, S., Mullen, J., Miranda, L.P., and Deshpande, R. (2012). Utilization of tyrosine- and histidine-containing dipeptides to enhance productivity and culture viability. Biotechnol. Bioeng. 109: 2286–2294. https://doi.org/10.1002/ bit.24507. Chun, C., Heineken, K., Szeto, D. et al. (2003). Application of factorial design to accelerate identification of CHO growth factor requirement. Biotechnol. Progr. 19: 52–57. Vander Heiden, M.G., Plas, D.R., Rathmell, J.C. et al. (2001). Growth factors can influence cell growth and survival through effects on glucose metabolism. Mol. Cell. Biol. 21: 5899–5912. https://doi.org/10.1128/MCB.21.17.5899–5912. 2001. Gospodarowicz, D. and Moran, J.S. (1976). Growth factors in mammalian cell culture. Annu. Rev. Biochem. 45: 531–558. Reinhart, D., Damjavovic, L., Kaisermayer, C., and Kunert, R. (2015). Benchmarking of commercially available CHO cell culture media for antibody production. Appl. Microbiol. Biotechnol. 99: 4645–4657. https://doi.org/10 .1007/s00253-015-6514-4. Lee, G.M. and Koo, J. (2010). Osmolarity effects, Chinese hamster ovary cell culture. In: Encyclopedia of Industrial Biotechnology: Bioprocess, Bioseparation, and Cell Technology, vol. 6 (ed. M.C. Flickinger), 3738–3745. Wiley. Zhu, M.M., Goyal, A., Rank, D.L. et al. (2005). Effects of elevated pCO2 and osmolality on growth of CHO cells and production of antibody-fusion protein B1: a case study. Biotechnol. Progr. 21: 70–71. https://doi.org/10.1021/ bp049815s.

References

34 Zhang, C., Liu, Z., Liu, X. et al. (2013). Targeted metabolic analysis of

35

36

37 38

39

40

41

42

43

44

45

46

47

nucleotides and identification of biomarkers associated with cancer in culture cell models. Acta Pharm. Sin. B 3: 254–262. https://doi.org/10.1016/j.apsb .2013.06.002. Morrone, F.B., Jacques-Silva, M.C., Horn, A.P. et al. (2003). Extracellular nucleotides and nucleosides induce profileration and increase nucleoside transport in human glioma cell lines. J. Neuro-Oncol. 64: 211–218. Ibarguren, M., Lopez, D.J., and Escriba, P.V. (2014). The effect of natural and synthetic fatty acids on membrane structure, microdomain organization, cellular functions and human health. Biochim. Biophys. Acta 1838: 1514–1528. https://doi.org/10.1016/j.bbamem.2013.12.021. Halliwell, B. (2014). Cell culture, oxidative stress, and antioxidants: avoiding pitfalls. Biomed. J. 37: 99–105. https://doi.org/10.4103/2319-4170.128725. Luo, L., Kawakatsu, M., Guo, C.-W. et al. Effects of antioxidants on the quality and genomic stability of induced pluripotent stem cells. Nature 4: 1–7. https://doi.org/10.1038/srep03779. Mandal, S., Mandal, A., Johansson, H.E. et al. (2013). Depletion of cellular polyamines, spermidine and spermine, causes a total arrest in translation and growth in mammalian. Proc. Natl. Acad. Sci. U.S.A. 110: 2169–2174. https:// doi.org/10.1073/pnas.1219002110. Landau, G., Bercovich, Z., Park, M.H., and Kahana, C. (2010). The role of polyamines in supporting growth of mammalian cells in mediated through their requirement for translational initiation and elongation. Journal of Biological Chemistry 285: 12474–12481. https://doi.org/10.1074/jbc.M110.106419. Chang, D., Fox, R., Hicks, E. et al. (2017). Investigation of interfacial properties of pure and mixed poloxamers for surfactant-mediated shear protection of mammalian cells. Colloids Surf., B 156: 358–365. https://doi.org/10.1016/j .colsurfb.2017.05.040. Peng, H., Hall, K.M., Clayton, B. et al. (2014). Development of small scale cell culture models for screening poloxamer 188 lot-to-lot variation. Biotechnol. Progr. 30: 1411–1418. https://doi.org/10.1002/btpr.1967. Hu, W., Rathman, J.J., and Chalmers, J.J. (2008). An investigation of small-molecule surfactants to potentially replace pluronic F-68 for reducing bubble-associated cell damage. Biotechnol. Bioeng. 101: 119–127. https:// doi.org/10.1002/bit.21872. Chotteau, V. (2015). Perfusion processes. In: Animal Cell Culture, Cell Engineering, vol. 9 (ed. M. Al-Rubeai), 407–443. Switzerland: Springer International Publishing https://doi.org/10.1007/978-3-319-10320-4_13. Xu, S., Gavin, J., Jiang, R., and Chen, H. (2017). Bioreactor productivity and media cost comparison for different intensified cell culture processes. Biotechnol. Progr. https://doi.org/10.1002/btpr.2415. Clincke, M.-F., Molleryd, C., Zhang, Y. et al. (2011). Study of a recombinant CHO cell line producing a monoclonal antibody by ATF or TFF external filter perfusion in a WAVE bioreactorTM . BMC Proc. 5: P105. Wang, S., Godfrey, S., Ravikrishnan, J. et al. (2017). Shear contributions to cell culture performance and product recovery in ATF and TFF perfusion

291

292

11 Chemically Defined Medium Development for the Production of Protein Therapeutics

48

49

50

51

52

53

54

55

56

57 58

59

60

systems. J. Biotechnol. 246: 52–60. https://doi.org/10.1016/j.jbiotec.2017.01 .020. Padawer, I., Ling, W.L.W., and Bai, Y. (2013). Case study: an accelerated 8-day monoclonal antibody production process based on high seeding densities. Biotechnol. Progr. 29: 829–832. https://doi.org/10.1002/btpr.1719. St. Amand, M.M., Radhakrishnan, D., Robinson, A.S., and Ogunnaike, B.A. (2014). Identification of manipulated variables for a glycosylation control strategy. Biotechnol. Bioeng. 111 (10): 1957–1970. https://doi.org/10.1002/bit .25251. Muthing, J., Kemminer, S.E., Conradt, H.S. et al. (2003). Effects of buffering conditions and culture pH on production rates and glycosylation of clinical phase I anti-melanoma mouse IgG3 monoclonal antibody R24. Biotechnol. Bioeng. 83: 321–334. Bollati-Fogolin, M., Forno, G., Nimtz, M. et al. (2005). Temperature reduction in cultures of hGM-CSF-expressing CHO cells: effect on productivity and product quality. Biotechnol. Progr. 21: 17–21. Clark, K.J., Chaplin, F.W., and Harcum, S.W. (2004). Temperature effects on product quality related enzymes in batch CHO cell cultures producing recombinant tPA. Biotechnol. Progr. 20: 1888–1892. Trummer, E., Fauland, K., Seidinger, S. et al. (2006). Process parameter shifting: part II. Biphasic cultivation – a tool for enhancing the volumetric productivity of batch processes using Epo-Fc expressing CHO cells. Biotechnol. Bioeng. 94 (6): 1045–1052. Kishishita, S., Nishikawa, T., Shinoda, Y. et al. (2015). Effect of temperature shift on levels of acidic charge variants in IgG monoclonal antibodies in Chinese hamster ovary cell culture. J. Biosci. Bioeng. 119 (6): 700–705. https://doi .org/10.1016/j.jbiosc.2014.10.028. Kaisermayer, C., Reinhart, D., Gili, A. et al. (2016). Biphasic cultivation strategy to avoid Epo-Fc aggregation and optimize protein expression. J. Biotechnol. 227: 3–9. https://doi.org/10.1016/j.jbiotec.2016.03.054. Restelli, V., Wang, M.-D., Huzel, N. et al. (2006). The effect of dissolved oxygen on the production and the glycosylation profile of recombinant human erythropoietin produced from CHO cells. Biotechnol. Bioeng. 94 (30): 481–494. https://doi.org/10.1002/bit.20875. Kimura, R. and Miller, W.M. (1997). Glycosylation of CHO-derived recombinant tPA produced under elevated pCO2 . Biotechnol. Progr. 13: 311–317. Zimmer, A., Mueller, R., Wehsling, M. et al. (2014). Improvement and simplification of fed-batch bioprocesses with a highly soluble phosphotyrosine sodium salt. J. Biotechnol. 186: 110–118. https://doi.org/10.1016/j.jbiotec.2015 .11.022. Hecklau, C., Pering, S., Seibel, R. et al. (2016). S-Sulfocysteine simplifies fed-batch processes and increases the CHO specific productivity via anti-oxidant activity. J. Biotechnol. 218: 53–63. https://doi.org/10.1016/j.jbiotec .2015.11.022. Meunier, S.M., Todorovic, B., Dare, E.V. et al. (2016). Impact of dissolved oxygen during UV-irradiation on the chemical composition and function of CHO cell culture media. PLoS One 11 (3): e0150957. https://doi.org/10.1371/journal .pone.0150957.

References

61 Hamdy, M.S., Scott, E.L., Carr, R.H., and Sanders, J.P.M. (2012). A novel pho-

62

63

64

65

66

67

68

69

70

71

72

73 74

tocatalytic conversion of tryptophan to kynurenine using black light as a light source. Catal. Lett. 142: 338–344. Attur, M.G., Dave, M.N., Tsunoyama, K. et al. (2002). “A system biology” approach to bioinformatics and functional genomics in complex human diseases: arthritis. Curr. Issues Mol. Biol. 4: 129–146. Stilwell, J.L., Guan, Y., Neve, R.M., and Gray, J.W. (2007). Systems biology in cancer research: genomics to cellomics. Methods Mol. Biol. 356: 353–365. Bhattacharya, S. and Mariani, T.J. (2013). Systems biology approaches to identify developmental bases for lung diseases. Pediatr. Res. 73: 514–522. https:// doi.org/10.1038/pr.2013.7. Uhlen, M., Hallstrom, B.M., Lindskog, C. et al. (2016). Transcriptomics resources of human tissues and organs. Mol. Syst. Biol. 12: 862. https://doi .org/10.15252/msb.20155865. Baker, M.H.A., Sarmidi, M.R., Cheng, K.-K. et al. (2015). Metabolomics the complementary field in systems biology; a review on obesity and type 2 diabetes. Mol. BioSyst. 11: 1742–1774. https://doi.org/10.1039/c5mb00158g. Gagna, C.E., Lambert, W.C., and Lambert, P.C. (2012). Advanced techniques for genomics and proteomics: transitional structural chemogenomics, chemoproteomics, pharmacogenomics, ad pharmacoproteomics. Am. Lab. 5: 2012. http://www.americanlaboratory.com/913-Technical-Articles/ 114244-Advanced-Techniques-for-Genomics-and-Proteomics-TransitionalStructural-Chemogenomics-Chemoproteomics-Pharmacogenomics-andPharmacoproteomics/. Horgan, R.P. and Kenny, L.C. (2011). SAC review ‘omic’ technologies: genomics, transcriptomics, proteomics and metabolomics. Obstet. Gynaecol. 13: 189–195. https://doi.org/10.1576/toag.13.3.189.27672. Conesa, A. and Mortazavi, A. (2014). The common ground of genomics and systems biology. BMC Syst. Biol. 8 (Suppl. 2): S1–S10. https://doi.org/10.1186/ 1752-0509-8-S2-S1. Da Hora Junior, B.T., De Faria Poloni, J., Lopes, M.A. et al. (2012). Transcriptomics and systems biology analysis in identification of specific pathways involved in cacao resistance and susceptibility to witches’ broom disease. Mol. BioSyst. 8: 1507–1519. Wang, H., Lan, P., and shen, R.F. (2016). Integration of transcriptomic and proteomic analysis towards understanding the systems biology of root hairs. Proteomics 16: 877–893. https://doi.org/10.1002/pmic.201500265. Nielsen, J. and Jewett, C. (2007). The role of metabolomics in systems biology. In: Metabolomics, Topics in Current Genetics, vol. 18 (eds. J. Nielsen and M.C. Jewett). Berlin, Heidelberg: Springer-Verlag https://doi.org/10.1007/ 4735_2007_0228. Houle, D., Govindaraju, D.R., and Omholt, S. (2010). Phenomics: the next challenge. Nature 11: 856–866. https://doi.org/10.1038/nrg2897. Park, J.H., Lee, S.Y., Kim, T.Y., and Kim, H.U. (2008). Application of systems biology for bioprocess development. Trends Biotechnol. 26: 404–412. https:// doi.org/10.1016/j.tibtech.2008.05.001.

293

294

11 Chemically Defined Medium Development for the Production of Protein Therapeutics

75 Liu, Z., Dai, S., Bones, J. et al. (2015). A quantitative proteomic analysis of

76

77

78

79

80

81

82 83

84

85

86

87

88

cellular responses to high glucose in Chinese hamster ovary cells. Biotechnol. Progr. 31: 1026–1038. https://doi.org/10.1002/btpr.2090. Clarke, C., Doolan, P., Barron, N. et al. (2011). Predicting cell-specific productivity from CHO gene expression. J. Biotechnol. 151: 159–165. https://doi.org/ 10.1016/j.jbiotec.2010.11.016. Dorai, H., Liu, S., Yao, X. et al. (2013). Proteomic analysis of bioreactor cultures of an antibody-expressing CHO-GS cell line that promotes high productivity. J. Proteomics Bioinf. 6: 98–108. https://doi.org/10.4172/jpb .1000268. Pavlopoulos, G.A., Malliarakis, D., Papanikolaou, N. et al. (2015). Visualizaing genome and systems biology: technologies, tools, implementation techniques and trends, past, present and future. GigaScience 4: 38. https://doi.org/10 .1186/s13742-015-0077-2. Kramer, B.P., Viretta, A.U., Baba, M.D.-E. et al. (2004). An engineered epigenetic transgene switch in mammalian cells. Nat. Biotechnol. 22: 867–870. https://doi.org/10.1038/nbt980. Bashor, C.J., Helman, N.C., and Yan s, L.W.A. (2005). Using engineered scaffold interactions to reshape MAP kinase pathway signaling dynamics. Science 319: 1539–1543. https://doi.org/10.1126/science.1151153. Park, J.G., Choi, S.S., and Park, T.H. (2007). Enhancement of cell growth and viability of CHO cells in serum-free media by 30Kc6 gene expression. Process Biochem. 42: 8–15. https://doi.org/10.1016/j.procbio.2006.04.010. Choi, S.S., Rhee, W.J., and Park, T.H. (2002). Inhibition of human cell apoptosis by silkworm hemolymph. Biotechnol. Progr. 18: 874–878. Choi, S.S., Phee, W.J., and Park, T.H. (2005). Beneficial effect of silkworm hemolymph on a CHO cell systems: inhibition of apoptosis and increase of EPO production. Biotechnol. Bioeng. 91: 793–800. Brown, A.J. and James, D.C. (2016). Precision control of recombinant gene transcription for CHO cell synthetic biology. Biotechnol. Adv. 34: 492–503. https://doi.org/10.1016/j.biotechadv.2015.12.012. Sellick, C.A., Croxford, A.S., Magsood, A.R. et al. (2011). Metabolic profiling of recombinant CHO cells: designing tailored feeding regimes that enhance recombinant antibody production. Biotechnol. Bioeng. 108: 3025–3031. https://doi.org/10.1002/bit.23269. Mohmad-Saberi, S.E., Hashim, Y.Z.H.-Y., Mel, M. et al. (2013). Metabolomics profiling of extracellular metabolites in CHO-K1 cells cultured in different types of growth media. Cytotechnology 65: 577–586. https://doi.org/10.1007/ s10616-012-9508-4. Kyriakopoulos, S. and Kontoravdi, C. (2014). A framework for the systematic design of fed-batch strategies in mammalian cell culture. Biotchnol. Bioeng. 111: 2466–2476. Robitaille, J., Chen, J., and Jolicoeur, M. (2015). A single dynamic metabolic model can describe mAb producing CHO cell batch and fed-batch cultures in different culture media. PLoS One 10 (9): e0136815. https://doi.org/10.1371/ journal.pone.0136815.

295

12 Host Cell Proteins During Biomanufacturing Jong Youn Baik 1,2,3 , Jing Guo 1,2,4 , and Kelvin H. Lee 1,2 1 University of Delaware, Department of Chemical and Biomolecular Engineering, 150 Academy Street, Newark, DE 19716, USA 2 University of Delaware, Delaware Biotechnology Institute, 15 Innovation Way, Newark, DE 19711, USA 3 Current Address: Inha University, Department of Biological Engineering, 100 Inha-ro, Michuhol-gu, Incheon 22212, South Korea 4 Current Address: Teva Pharmaceuticals, CMC Process Development Group, 145 Brandywine Pkwy, West Chester, PA 19380, USA

12.1 Introduction During biopharmaceutical manufacturing production culture, cells produce and secrete recombinant proteins into the culture medium along with many types of impurities. The International Council for Harmonisation (ICH) quality guideline (Q6B – specification) states that “process-related impurities encompass those that are derived from the manufacturing process, i.e. cell substrates (e.g. host cell proteins (HCPs) and host cell DNA)” and that “product-related impurities (e.g. precursors and certain degradation products) are molecular variants arising during manufacture and/or storage” [1]. Host cell-derived impurities include DNA, proteins, lipids, and metabolites, and they should be removed through downstream purification processes. Of the impurities listed above, HCPs can impose a challenge to downstream purification processes because they may exhibit purification-related properties similar to biopharmaceutical proteins. HCPs include both secreted proteins and intracellular proteins, as intracellular proteins can be released from dead cells during production cultures (e.g. fed-batch and perfusion cultures) or harvesting steps. The scope of this review will encompass current HCP removal processes, the impact of residual HCPs, as well as HCP detection, quantification, and monitoring methods during biomanufacturing processes. Strategies for effective HCP removal and future directions for HCP risk management are also discussed.

12.2 Removal of HCP Impurities Current HCP removal processes involve a series of bioseparation methods, such as centrifugation, filtration, chromatography, and precipitation. For antibody products, including monoclonal antibodies (mAbs) and Fc-fusion Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

296

12 Host Cell Proteins During Biomanufacturing

Polishing

Protein A Harvest cell culture fluid

Protein A eluate

Polishing column eluate

Figure 12.1 Simplified scheme for downstream purification process.

proteins, HCPs are typically removed through a platform purification process (Figure 12.1). For non-antibody products, the process to remove HCPs can vary, depending on the host cell type and product molecular properties. Although the majority of HCPs can be cleared through downstream purification processes, there are still some HCPs that are difficult to remove from the drug substance or drug product. 12.2.1

Antibody Product

As the molecular properties of mAbs and Fc-fusion proteins are highly conserved from product to product and they share more than 95% amino acid sequence homology in their fragment-crystallizable (Fc) regions, a platform process is often employed to remove HCPs and other impurities (Figure 12.1). The three major steps of the process include clarification of the harvested cell culture fluid (HCCF), an initial capture step with protein A chromatography, and a few subsequent polishing steps, often with ion exchange chromatography (IEX) and/or hydrophobic interaction chromatography (HIC) [2, 3]. The clarification of the HCCF consists of a centrifugation step to remove cells and larger cell debris, followed by a depth filtration step to remove small cell debris [3]. The depth filtration step is also effective in HCP clearance through a combination of electrostatic and hydrophobic adsorptive interactions between the depth filter and proteins [4, 5]. However, the majority of HCP removal is realized by the following capture step with protein A chromatography [6]. Protein A, a cell-wall-associated protein on the surface of the bacterium Staphylococcus aureus, has a high binding affinity to the Fc region of the mAb or Fc-fusion protein [7]. Additional characteristics, such as stability over a wide pH range [8] and the ability to maintain functional performance after repeated cleanings [9], also add to the functionality and effectiveness of protein A chromatography. Although the majority of HCPs are removed during protein A capture, additional orthogonal polishing steps are still necessary to further lower the total HCP concentration. There are two types of IEX: anion exchange chromatography (AEX) and cation exchange chromatography (CEX). AEX is usually operated in flow-through mode as mAbs typically have a net positive charge at neutral pH and do not bind to the resin, whereas HCP impurities adsorb to the

12.2 Removal of HCP Impurities

resin [10]. CEX is often operated in bind-and-elute mode, where the positively charged mAb binds to the resin and the impurities flow through the column. After loading, the mAb is eluted from the column at higher salt concentrations. It has been reported that CEX can reduce HCP levels from 300–400 ppm to approximately 10 ppm [11]. Besides IEX, HIC has also been adopted as a polishing step to remove HCP impurities. It is usually operated in bind-and-elute mode, and high salt concentration is used to load the protein in HIC to promote hydrophobic binding, whereas low salt concentration is used to elute the protein from the column [12]. Hunter et al. reported that one HIC unit operation could reduce HCP concentrations from 10 000 ppm to approximately 300 ppm [13].

12.2.2

Non-antibody Protein Product

Although the downstream processing of mAbs relies on a relatively rigid platform, the purification of other biopharmaceutical products (e.g. insulin, erythropoietin (EPO), and interleukins (ILs)) varies, depending on the host cells and drug product characteristics. As a therapeutic drug to treat diabetes, insulin is produced predominantly either in Escherichia coli or Saccharomyces cerevisiae [14]. In the E. coli production platform, insulin is intracellularly overexpressed and then solubilized and renatured to obtain fully functional proteins. Immunoglobulin G (IgG) sepharose affinity chromatography is used as a capture step, followed by preparative reverse-phase chromatography to finally recover the product [15]. In the S. cerevisiae production platform, insulin is secreted into the culture medium and CEX is used as a primary capture step, followed by a series of separation methods. Among them, ethanol precipitation selectively targets HCP removal by using an ethanol concentration at which HCPs precipitate while insulin remains soluble [16]. EPO, a growth factor for the treatment of anemia related to kidney disease, is primarily produced in mammalian cells. Several industrial-scale purification processes have been established to remove impurities from the EPO product. All of these methods involve clarification, primary capture chromatography, and subsequent polishing chromatography steps. The various molecular properties of EPO enable multiple choices of capture step. For example, Zanette et al. used phenylboronate agarose (PBA) to capture EPO based on the ability of PBA to form reversible complexes with 1,2-cis-diol-containing molecules [17]. On the other hand, blue sepharose affinity chromatography [18] and IEX [19] have also been applied as a capture step for EPO purification. ILs are a family of proteins that stimulate and regulate the cells involved in immunity and inflammation. Because of the variety of ILs [20], the expression system and purification process are highly specific depending on which IL is being produced. For example, IL-7 has been expressed in E. coli cells for a 1000 l fermentation scale and purified with a series of HIC and IEC columns [21]. On the other hand, IL-12 has been produced in mammalian cells and recovered in a single step with heparin sepharose affinity chromatography [22].

297

298

12 Host Cell Proteins During Biomanufacturing

12.2.3

Difficult-to-Remove HCPs

Although downstream processing of the product can significantly reduce total HCP levels to meet the criteria required by the Food and Drug Administration (FDA), the persistence of some difficult-to-remove HCPs continues to challenge the entire purification process and jeopardizes drug efficacy and quality. There are at least three routes by which HCPs can challenge downstream processing. The first route refers to HCPs with variable expression during extended cell culture, as the composition of HCPs generated during upstream processing has been shown to affect downstream purification [23]. A proteomics approach identified 92 extracellular HCPs from Chinese hamster ovary (CHO) cells exhibiting up to 48-fold changes in protein expression over 500 days of cell culture [24]. The second route is HCP association with the product, where “hitchhiker” HCPs bind to the product, especially an antibody product, and are carried along throughout the purification process. This HCP-mAb interaction is considered to be the primary cause of HCPs persisting through protein A chromatography; a batch chromatography binding study exhibited substantial differences in HCP profiles of the protein A eluate between null HCCF and mAb-containing HCCF [25]. Additionally, several groups have identified individual HCP compositions in the protein A eluate [26–29] with proteomics techniques, and the most commonly observed HCPs may be worthy of special consideration. The third route refers to co-elution of HCPs with the product, as some HCPs can bind to the chromatography resin ligands during the loading step and can be eluted together with the product. Co-elution has been shown to occur during polishing chromatography steps, including HIC and IEX [28, 30].

12.3 Impacts of Residual HCPs Despite all efforts to remove HCPs from drug substance, the complete elimination of HCPs remains a challenge. In addition, HCPs have the potential to negatively impact drug quality and efficacy, thereby establishing HCPs as a critical quality attribute (CQA). According to the ICH guidelines (Q8), “a CQA is a physical, chemical, biological, or microbiological property or characteristic that should be within an appropriate limit, range, or distribution to ensure the desired product quality” [31]. Moreover, residual HCPs in the final drug product can pose safety concerns to patients or reduce product shelf life by degrading the product or other components in the formulation. 12.3.1

Drug Efficacy, Quality, and Shelf Life

Either HCPs in the HCCF before downstream purification processes or residual HCPs in the drug substance or drug product after purification processes can affect drug efficacy, quality, and shelf life. As HCCF samples are collected near the end of a typical fed-batch culture when cell viability is about 70–80%, dead cells release HCPs into the culture medium, allowing various enzymes to catalyze metabolic reactions that can modify biochemical properties of recombinant

12.3 Impacts of Residual HCPs

proteins. For example, the amount of sialidase, an enzyme that removes terminal sialic acid residues from N-glycans, is known to increase in the culture medium during later days of batch or fed-batch cultures [32]. Although sialidases appear to be removed during the purification process, and no detectable amount of sialidase in the drug substance or product has been reported, they can catalyze the removal of terminal sialic residues from the N-glycans of product proteins (e.g. antithrombin) and decrease sialic acid levels before their removal through downstream purification [32]. N-glycosylated proteins without sialic acid residues expose terminal galactose residues, which facilitates binding to the asialoglycoprotein receptors expressed in the liver, causing faster protein clearance [33]. Another example is the cleavage of C-terminal lysine residues of IgG1 by carboxypeptidase D [34]. Charge variants of mAbs, resulting from lysine removal, may affect stability and biological activity [35]. Residual HCPs in the drug substance or product can also have adverse effects on drug efficacy and shelf life. Many studies have reported degradation of antibody products over time, resulting in a decrease in drug efficacy as well as an increase in risk due to potential immunogenicity against cleaved antibody fragments [36, 37]. Proteases, such as cathepsin D, have been identified and shown to contribute to antibody cleavage and degradation [36, 38]. Additionally, studies have reported that lipase classes, such as lipoprotein lipase and phospholipase B-like2, degrade lipid components, such as polysorbate 20/80, in the final drug formulation [39, 40]. As these lipid additives are used as a stabilizer, a decrease in the concentration of these components can lead to a shorter drug shelf life. 12.3.2

Immunogenicity

The foremost problem with residual HCPs as foreign (exogenous) proteins is that they can trigger an immune response in patients when the drug is administered. Immune responses to foreign proteins occur via T-cell-dependent pathways, whereby proteins are taken up, digested, and presented by antigen-presenting cells, then recognized by T-cells, followed by further activation and maturation of B cells expressing complementary antibodies. Although the scale of CHO HCP immunogenicity is smaller than that elicited from the HCPs of nonmammalian organisms such as E. coli or yeast [41], sequence differences between CHO and human proteins are substantial. A recent study reported that only 20% of the CHO proteome has higher than 90% sequence homology to human, whereas over 60% of the proteome has less than 50% sequence homology to human [42]. Indeed, two clinical trials have been canceled because of the adverse CHO HCP-associated immune responses in patients [43, 44]. Besides direct immunogenicity, HCPs have the potential to induce and augment antidrug antibodies or induce an immune response to an endogenous protein [37]. 12.3.3

Biological Activity

Beyond immunogenic issues, residual HCPs can have negative biological impacts on patients [45]. Although many HCPs are inactive in the drug substance or product, some HCPs can maintain their biological function and induce unintended

299

300

12 Host Cell Proteins During Biomanufacturing

activities. For example, Beatson et al. reported that transforming growth factor β1 (TGFβ1) proteins expressed in CHO are functional, carried through purification steps, and can act on human cells [46]. As TGFβ1 is a multifunctional cytokine with a highly conserved protein sequence, active TGFβ1 can impact a variety of cellular processes in patients, including cell growth, wound healing, apoptosis, and immunosuppression [47–49]. Moreover, it is also possible that other biologically active HCPs, such as cytokines and autocrine signaling factors, can be expressed in CHO cells under certain conditions [50], thereby causing other unintended clinical effects, such as hypersensitivity, toxicity, and cell signaling.

12.4 HCP Detection and Monitoring Methods There is no well-established pharmacological evaluation for an acceptable, or safe, range of HCPs, mainly because of the heterogeneity and variety of HCPs. For this reason, robust purification processes that can achieve the lowest or “undetectable” residual HCP amount are desirable whenever possible; therefore, the use of sensitive and appropriate HCP detection and monitoring methods is critical. According to the ICH guideline Q6B, “a sensitive assay, e.g. immunoassay, capable of detecting a wide range of protein impurities is generally utilized” for the detection of HCPs [1]. However, immunoassays have their own limitations, requiring orthogonal approaches to fully characterize and monitor HCPs in the drug substance and product (Table 12.1). 12.4.1 Anti-HCP Antiserum and Enzyme-Linked Immunosorbent Assay (ELISA) Currently, enzyme-linked immunosorbent assay (ELISA) is the industry “gold standard” for the detection of HCPs because of its high sensitivity (0.5–1 ng/ml, [51]), coverage, and throughput [37, 52]. The key component of this immunoassay are polyclonal antibodies against HCPs generated by “immunization with a preparation of production cells minus the product-coding gene, fusion partners, or other appropriate cell lines” [1]. Polyclonal antibodies are typically raised in animals such as rabbit, goat, or chicken. Although individual HCPs are not identified, polyclonal anti-HCP antibodies can capture most, and theoretically all, proteins in a given HCP pool. In brief, a typical ELISA follows a series of steps as described next (Figure 12.2). (1) An assay plate is coated with polyclonal anti-HCP antibodies. (2) Samples potentially containing HCPs are loaded into coated wells. (3) The bound HCPs are recognized by anti-HCP antibodies (primary antibody). (4) HCP–antibody detection is amplified with a biotin–avidin complex on primary antibodies or a secondary antibody conjugated with an enzyme, such as alkaline phosphatase or horseradish peroxidase, that catalyzes substrates into fluorescent signals. Although the immunoassay is a widely accepted HCP detection method and generic HCP detection ELISA kits are available, they do have certain limitations. First, because of the heterogeneous expression levels of individual HCPs and various binding affinities between

12.4 HCP Detection and Monitoring Methods

Table 12.1 Methods for HCP detection and quantification. Technique

Application

Strengths

Limitations

ELISA

Total HCP level quantitation

High sensitivity Simple procedures

No information about [52, 53] individual HCP identity No detection or biased binding to some HCPs

1D- or 2D-PAGE

Study changes in HCP expression patterns

Visual investigation of Low sensitivity protein isoforms and modifications

[4, 13, 54, 55]

DIGE (2D-PAGE with labeling)

Study changes in HCP expression patterns

Visual investigation of Low sensitivity protein isoforms and modifications Better normalization than 2D PAGE

[56, 57]

LC–MS

Identification of High sensitivity No information about individual HCPs Identification of HCPs protein isoforms and modifications

iTRAQ (LC–MS with labeling)

Identification and quantitation of individual HCPs

MRM

Quantitation of High sensitivity targeted Quantitative individual HCPs comparison of multiple HCPs across multiple samples

High sensitivity Identification of HCPs Quantitative comparison of the entire proteome between samples

Figure 12.2 Schematic of typical ELISA for HCP quantitation. Numbers in the parentheses represent the ELISA procedures described in Section 12.4.1.

No information about protein isoforms and modifications Limited number of samples per run (4-plex or 8-plex)

References

[6, 55, 60, 61]

[28]

No information about [40] protein isoforms and modifications Prior knowledge about target is necessary

Substrate

Avidin-conjugated enzyme

(4)

Biotin-conjugated primary Ab

(3)

HCP

(2)

Coating Ab

(1)

Plate

301

302

12 Host Cell Proteins During Biomanufacturing

individual proteins and antibodies, the overall HCP measurement can be biased (i.e. more antibodies against the most abundant HCPs), rather than reflecting the true HCP amount [37]. Secondly, some HCPs are not detected because antibodies are not necessarily made against all HCPs. For example, one HCP, glutathione S transferase-α (GST-α), in the drug substance was detected by capillary electrophoresis-sodium dodecyl sulfate (CE-SDS) but not by HCP ELISA [53]. Finally, variations in cell line, product, or bioprocessing can affect HCP profiles, leading to biased immunoassay measurements [37]. Therefore, the development of manufacturing process-specific assays should be initiated for products in the later phases (phase III or commercial) of the pipeline; however, they require a substantial amount of time and effort. These limitations emphasize the importance of orthogonal approaches, such as protein identification by electrophoresis and/or mass spectrometry. 12.4.2

Proteomics Approaches as Orthogonal Methods

As complementary methods to immunoassay, protein separation and visualization by one-dimensional (1D) and two-dimensional (2D) sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) have been utilized [4, 13, 54, 55]. HCP-containing samples are separated by molecular weight (1D) or by both molecular weight and isoelectric point (2D) and subsequently visualized by gel staining or by immunoblotting. In addition, 2D-differential in-gel electrophoresis (DIGE) allows comparison of individual protein amounts and reduces gel-to-gel variations by running multiple samples labeled with different fluorescent dyes in one gel [56, 57]. Although these methods can be combined with mass spectrometry for spot identification, they have relatively poor sensitivity (8–52 ng protein by Coomassie staining and 0.3–1 ng protein by SYPRO Ruby staining) and therefore are only effective for the characterization of abundant proteins [51, 58]. Non-gel-based proteomics approaches, such as liquid chromatography– tandem mass spectrometry (LC–MS/MS), have much better sensitivity (0.92–46.2 pg, converted from 1 to 50 fmol [51], based on a report that the average length of trypsin-digested peptide is 8.4 amino acids [59]) and throughput [6, 51, 55, 60, 61]. Notably, isobaric tags for relative and absolute quantitation (iTRAQ) and multiple reaction monitoring (MRM) have recently been adopted to detect and monitor HCPs. In these methods, enzyme-digested peptide samples (either labeled or unlabeled) are separated by liquid chromatography, followed by MS/MS analysis [62]. iTRAQ allows comparison of multiple samples (4-plex or 8-plex) in a single run by using different isobaric labeling tags, while MRM enables quantification of multiple target proteins across multiple samples by selecting precursor ions of interest during the first MS stage and by identifying the selected precursor ions during the second MS stage [28, 40].

12.5 Efforts for HCP Control Given that the impact of individual residual HCPs is not fully understood at present and that it is uncertain whether a 100 ppm range is acceptable, robust and

12.5 Efforts for HCP Control

Table 12.2 Approaches for HCP removal.

Upstream processing

Downstream processing

Stages

Approaches

References

Cell line development

Knockout of critical HCPs

[34, 40]

Cell culture

Sustaining high cell viability (e.g. adjustment of harvest time)

[64, 65]

Harvest clarification

Selecting depth filter with high HCP removal capacity

[4]

Protein A chromatography

Column wash to disrupt HCP–product interactions

[68, 69]

Polishing steps

Alternative operating conditions (e.g. pH gradient elution) New modes of chromatography (e.g. mixed mode chromatography)

[70] [71, 72]

effective removal of HCPs from the drug substance and drug product is imperative. In addition, to achieve effective removal or control of HCPs, it is of great importance to identify which factors affect HCP profiles during manufacturing processes. Many studies have implicated a variety of factors, such as product type, cell viability, culture conditions, and chromatography resins, in impacting the HCP profile in both upstream and downstream processes (Table 12.2). Additionally, approaches to assess and predict HCP-associated risks that can adversely affect product efficacy and quality are currently in development. 12.5.1

Upstream Efforts

Studies have shown that many variables in the upstream process, such as host cell line selection, cell age, products (amino acid variations between biopharmaceutical proteins), culture conditions, media and feeding supplements, and harvest time, can affect HCP profiles in the HCCF [2, 24, 56, 63]. Yet the most important culture parameter impacting the HCP profile is cell viability at harvest [55, 56]; the relative abundance of intracellular HCPs in the HCCF shows a substantial increase at the later days of a production culture (i.e. at low cell viability) because dead cells lyse and release intracellular components including proteins. Indeed, 2D-DIGE studies confirmed that cell viability exhibited greater impact on the HCP expression pattern than other culture variables, such as media components, feeding strategy, temperature shift, and different clones expressing the same product [56, 57]. Considering the impact of cell viability on HCP profiles, efforts have been made to sustain high viability during the culture until harvest, including overexpression of antiapoptotic genes to prevent programmed cell death, shift to a lower temperature, and adjustment of harvest time [64, 65]. However, it has also been shown that the amount of extracellular HCPs accumulates substantially (1–2 g/l on day 14) throughout the culture with high (>70%) cell viabilities [63], suggesting that sustaining high cell viability

303

304

12 Host Cell Proteins During Biomanufacturing

may not be sufficient to control HCPs. Moreover, as cell age-dependent changes in HCP expression patterns eventually challenge downstream purification processes, the ideal solution is to completely remove HCPs that are detrimental to the drug efficacy and/or difficult to remove. With the recent development of genome editing tools, knockout of critical HCPs has been reported. For example, carboxypeptidase D, an HCP responsible for C-terminal lysine cleavage of antibodies, was knocked out using clustered regularly interspaced short palindromic repeats (CRISPRs)/CRISPR-associated 9 (Cas9) technology, leading to the complete removal of C-terminal lysine heterogeneity [34]. In addition, lipoprotein lipase, which persisted through downstream purification processes and caused cleavage of lipid components in the final formulation, was knocked out using both transcription activator-like effector nucleases (TALENs) and CRISPR/Cas9 technology [40]. Although these examples demonstrate the applicability of genome editing tools for HCP removal, there are considerations when performing gene knockouts. Studies have reported that the knockout of one gene can result in the activation of other genes to restore phenotype (gene function), a phenomenon referred to as genetic compensation [66]. Therefore, if one attempts to knockout a gene encoding a particular enzyme, the gene expression level of other enzyme genes in the same family should be carefully examined. Additionally, it is important to choose target HCPs that are not essential to cell growth and survival, or protein production. 12.5.2

Downstream Efforts

Given the persistence of some difficult-to-remove HCPs during protein A and polishing chromatography steps, various downstream strategies for effective HCP removal have been proposed [58, 67]. For the protein A capture step, as mAb-HCP interactions were proven to be the primary cause of HCPs entering the protein A elution pool, disrupting these interactions allows HCPs to flow through the column and achieves better separation. Post-load washing, as an intermediate step between loading and elution, is currently a principal way to dissociate HCPs from the product. As HCPs can bind to the mAb through various mechanisms, including electrostatic interaction, hydrophobic interaction, and hydrogen bonding, different wash buffers have been used to disrupt this binding. For example, the work by Chollangi et al. [68] showed that a column wash with basic buffer (pH ≥ 8) can be effective in improving HCP removal by anionizing both mAbs and the majority of HCPs so that they exhibit repulsive interactions. Additionally, additives such as arginine, isopropanol, and sodium chloride were also shown to significantly reduce HCP levels in a protein A elution pool by disrupting one or several interaction mechanisms between HCPs and mAbs [68, 69]. For polishing steps, strategies were focused on optimizing the conventional chromatography methods (IEX and HIC), as well as developing new modes of chromatography. For example, during CEX, as an alternative to salt gradient elution, pH gradient elution has been applied to mAbs and was shown to remove substantial amounts of HCPs [70]. Recently, mixed mode chromatography is gaining popularity as a polishing step to clear HCPs [71]. The mixed mode resin can adsorb the proteins through more than one mode of

12.6 Future Directions

interaction, resulting in higher selectivity and specificity. For example, Capto Adhere, a particular mixed mode resin, has been reported to achieve 99% HCP clearance to achieve a final level below 10 ppm for a mAb product [72]. 12.5.3

HCP Risk Assessment in CHO Cells

Despite all the aforementioned efforts, it is impractical to evaluate the impact of each individual HCP and to fully control them because of insufficient knowledge regarding the CHO proteome. Although about 24 000 genes have been identified from the current Chinese hamster and CHO cell annotations [73], biological function and protein expression level of these genes remain largely unknown. Moreover, a single gene can be translated into multiple protein isoforms, and a single protein can be posttranslationally modified in many ways (e.g. glycosylation, phosphorylation, acetylation, sumoylation, and truncations), further expanding the already overwhelming number of potential HCPs [74]. Another challenge is that knowledge about the characteristics of individual HCPs is lacking; an abundant HCP is not necessarily a persistent HCP, nor a critical one (i.e. immunogenic). Therefore, risk assessment of HCPs is necessary to predict, evaluate, and control critical HCPs throughout bioprocessing. Proposed elements that determine the risk surrounding a particular HCP include (i) severity, (ii) detectability, and (iii) abundance of the HCP [41]. Severity refers to the potential of a HCP to impact patient health, such as immunogenicity and biological activity, and can be determined experimentally or computationally. For example, CHOPPI (CHO protein-predicted immunogenicity), an immunogenic risk prediction tool [42], has been developed to provide information about the potential presence and immunogenicity of CHO HCPs using the CHO proteome database and EpiMatrix, an in silico platform for epitope identification and prediction [75]. Detectability refers to how easily a particular HCP can be identified and quantified; while abundance refers to the amount of an HCP. These elements depend, in part, on the detection method employed. Assessment based on these elements would allow a reduction in HCP-associated risks, as well as, development of critical HCP-specific clearance methods.

12.6 Future Directions With encouragement from regulatory groups, the biopharmaceutical industry is moving toward continuous biomanufacturing paradigms because of the many anticipated advantages over fed-batch, including less lot-to-lot product quality variability, operational flexibility, cost effectiveness, and smaller environmental and operational footprints [76, 77]. Continuous biomanufacturing requires on-line or at-line analysis and monitoring of CQAs; therefore, prompt HCP detection and monitoring methods must be established to replace current off-line methods. Moreover, a major challenge to continuous biomanufacturing is cell line instability, which may result in unexpected changes that can affect CQAs. As changes in HCP expression patterns during long-term cultures (up to

305

306

12 Host Cell Proteins During Biomanufacturing

one year) have been reported [24], HCP profiles should also be monitored when assessing cell line instability. Lastly, new immunoassays must ensure the detection of a broader spectrum of HCPs early in the process development cycle (e.g. HCPs in HCCF rather than HCPs in the final drug product) such that the assays are able to readily identify and quantify changes in HCP expression patterns resulting from any process changes. Although immunoassays are stipulated in the regulatory guideline (ICH Q6B) and are most widely utilized in industry, there are no restrictions on the types of methods that can be used in assaying HCPs and employing orthogonal approaches is encouraged. An ideal HCP quantification method should be able to (i) identify the entire profile in a single run, (ii) detect trace amounts of HCPs, (iii) accommodate a wide quantification range, and (iv) specify and monitor individual HCPs.

Acknowledgments We are grateful for the financial support from the National Science Foundation (1539359, 1412365, 1624698, and 1736123).

References 1 ICH (1999). Q6B specifications: test procedures and acceptance criteria for

2

3 4

5

6

7

8

biotechnological/biological products. Guidance for Industry. https://www.ich .org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q6B/Step4/ Q6B_Guideline.pdf (accessed 19 April 2019). Shukla, A.A., Hubbard, B., Tressel, T. et al. (2007). Downstream processing of monoclonal antibodies – application of platform approaches. J. Chromatogr. B Analyt. Technol. Biomed. Life. Sci. 848: 28–39. Liu, H.F., Ma, J., Winter, C., and Bayer, R. (2010). Recovery and purification process development for monoclonal antibody production. mAbs 2: 480–499. Hogwood, C.E., Tait, A.S., Koloteva-Levine, N. et al. (2013). The dynamics of the CHO host cell protein profile during clarification and protein A capture in a platform antibody purification process. Biotechnol. Bioeng. 110: 240–251. Yigzaw, Y., Piper, R., Tran, M., and Shukla, A.A. (2006). Exploitation of the adsorptive properties of depth filters for host cell protein removal during monoclonal antibody purification. Biotechnol. Prog. 22: 288–296. Tarrant, R.D., Velez-Suberbie, M.L., Tait, A.S. et al. (2012). Host cell protein adsorption characteristics during protein A chromatography. Biotechnol. Prog. 28: 1037–1044. Hober, S., Nord, K., and Linhult, M. (2007). Protein A chromatography for antibody purification. J. Chromatogr. B Analyt. Technol. Biomed. Life. Sci. 848: 40–47. Huse, K., Bohme, H.J., and Scholz, G.H. (2002). Purification of antibodies by affinity chromatography. J. Biochem. Biophys. Methods 51: 217–231.

References

9 Hale, G., Drumm, A., Harrison, P., and Phillips, J. (1994). Repeated cleaning

10

11

12

13

14 15 16 17

18

19

20

21

22

23 24

of protein A affinity column with sodium hydroxide. J. Immunol. Methods 171: 15–21. Wang, C., Soice, N.P., Ramaswamy, S. et al. (2007). Cored anion-exchange chromatography media for antibody flow-through purification. J. Chromatogr. A 1155: 74–84. Xu, Z., Li, J., and Zhou, J.X. (2012). Process development for robust removal of aggregates using cation exchange chromatography in monoclonal antibody purification with implementation of quality by design. Prep. Biochem. Biotechnol. 42: 183–202. Chen, J., Tetrault, J., and Ley, A. (2008). Comparison of standard and new generation hydrophobic interaction chromatography resins in the monoclonal antibody purification process. J. Chromatogr. A 1177: 272–281. Hunter, A.K., Wang, X., Suda, E.J. et al. (2009). Separation of product associating E. coli host cell proteins OppA and DppA from recombinant apolipoprotein A-I(Milano) in an industrial HIC unit operation. Biotechnol. Prog. 25: 446–453. Baeshen, N.A., Baeshen, M.N., Sheikh, A. et al. (2014). Cell factories for insulin production. Microb. Cell Fact. 13: 141. Nilsson, J., Jonasson, P., Samuelsson, E. et al. (1996). Integrated production of human insulin and its C-peptide. J. Biotechnol. 48: 241–250. Mollerup, I., Jensen, S.W., Larsen, P. et al. (1996). Insulin purification. Encycl. Ind. Biotechnol. 992: 1219–1222. Zanette, D., Soffientini, A., Sottani, C., and Sarubbi, E. (2003). Evaluation of phenylboronate agarose for industrial-scale purification of erythropoietin from mammalian cell cultures. J. Biotechnol. 101: 275–287. Surabattula, R., Rao, K.R.S.S., and Polavarapu, R. (2011). An optimized process for expression, scale-up and purification of recombinant erythropoietin produced in Chinese hamster ovary cell culture. Res. Biotechnol. 2: 58–74. Hu, Y., Chen, S., Xu, M., and Zhang, S. (2004). An improved, inexpensive procedure for the large-scale purification of recombinant human erythropoietin. Biotechnol. Appl. Biochem. 40: 89–94. Akdis, M., Burgler, S., Crameri, R. et al. (2011). Interleukins, from 1 to 37, and interferon-gamma: receptors, functions, and roles in diseases. J. Allergy Clin. Immunol. 127: 701–721. e701–770. Ouellette, T., Destrau, S., Zhu, J. et al. (2003). Production and purification of refolded recombinant human IL-7 from inclusion bodies. Protein Expr. Purif. 30: 156–166. Jayanthi, S., Koppolu, B., Smith, S.G. et al. (2014). Efficient production and purification of recombinant human interleukin-12 (IL-12) overexpressed in mammalian cells without affinity tag. Protein Expr. Purif. 102: 76–84. Wu, P. (2013). Practical experiences integrating upstream and downstream processing. Am. Pharm. Rev. 16: 84–86. Valente, K.N., Lenhoff, A.M., and Lee, K.H. (2015). Expression of difficult-to-remove host cell protein impurities during extended Chinese hamster ovary cell culture and their impact on continuous bioprocessing. Biotechnol. Bioeng. 112: 1232–1242.

307

308

12 Host Cell Proteins During Biomanufacturing

25 Sisodiya, V.N., Lequieu, J., Rodriguez, M. et al. (2012). Studying host cell

26

27

28

29

30

31 32

33 34

35

36

37 38

39

40

protein interactions with monoclonal antibodies using high throughput protein A chromatography. Biotechnol. J. 7: 1233–1241. Levy, N.E., Valente, K.N., Choe, L.H. et al. (2014). Identification and characterization of host cell protein product-associated impurities in monoclonal antibody bioprocessing. Biotechnol. Bioeng. 111: 904–912. Zhang, Q., Goetze, A.M., Cui, H. et al. (2016). Characterization of the co-elution of host cell proteins with monoclonal antibodies during protein A purification. Biotechnol. Prog. 32: 708–717. Chiverton, L.M., Evans, C., Pandhal, J. et al. (2016). Quantitative definition and monitoring of the host cell protein proteome using iTRAQ – a study of an industrial mAb producing CHO-S cell line. Biotechnol. J. 11: 1014–1024. Aboulaich, N., Chung, W.K., Thompson, J.H. et al. (2014). A novel approach to monitor clearance of host cell proteins associated with monoclonal antibodies. Biotechnol. Prog. 30: 1114–1124. Levy, N.E., Valente, K.N., Lee, K.H., and Lenhoff, A.M. (2016). Host cell protein impurities in chromatographic polishing steps for monoclonal antibody purification. Biotechnol. Bioeng. 113: 1260–1272. ICH (2009). Q8(R2) pharmaceutical development. Guidance for Industry. Munzert, E., Mthing, J., Buntemeyer, H., and Lehmann, J. (1996). Sialidase activity in culture fluid of Chinese hamster ovary cells during batch culture and its effects on recombinant human antithrombin III integrity. Biotechnol. Prog. 12: 559–563. Weiss, P. and Ashwell, G. (1989). The asialoglycoprotein receptor: properties and modulation by ligand. Prog. Clin. Biol. Res. 300: 169–184. Hu, Z., Zhang, H., Haley, B. et al. (2016). Carboxypeptidase D is the only enzyme responsible for antibody C-terminal lysine cleavage in Chinese hamster ovary (CHO) cells. Biotechnol. Bioeng. 113: 2100–2106. Khawli, L.A., Goswami, S., Hutchinson, R. et al. (2010). Charge variants in IgG1: isolation, characterization, in vitro binding properties and pharmacokinetics in rats. mAbs 2: 613–624. Robert, F., Bierau, H., Rossi, M. et al. (2009). Degradation of an Fc-fusion recombinant protein by host cell proteases: identification of a CHO cathepsin D protease. Biotechnol. Bioeng. 104: 1132–1141. Wang, F.Q., Richardson, D., and Shameem, M. (2015). Host-cell protein measurement and control. Biopharm. Int. 28: 32–38. Gao, S.X., Zhang, Y., Stansberry-Perkins, K. et al. (2011). Fragmentation of a highly purified monoclonal antibody attributed to residual CHO cell protease activity. Biotechnol. Bioeng. 108: 977–982. Vanderlaan, M., Sandoval, W., Liu, P. et al. (2015). Hamster phospholipase B-like 2 (PLBL2): a host-cell protein impurity in therapeutic monoclonal antibodies derived from Chinese hamster ovary cells. Bioprocess Int. 13: 19–29. Chiu, J., Valente, K.N., Levy, N.E. et al. (2017). Knockout of a difficult-to-remove CHO host cell protein, lipoprotein lipase, for improved polysorbate stability in monoclonal antibody formulations. Biotechnol. Bioeng. 114: 1006–1015.

References

41 Bracewell, D.G., Francis, R., and Smales, C.M. (2015). The future of host cell

42

43

44

45

46

47 48 49

50

51

52 53

54

55

56

protein (HCP) identification during process development and manufacturing linked to a risk-based management for their control. Biotechnol. Bioeng. 112: 1727–1737. Bailey-Kellogg, C., Gutierrez, A.H., Moise, L. et al. (2014). CHOPPI: a web tool for the analysis of immunogenicity risk from host cell proteins in CHO-based protein production. Biotechnol. Bioeng. 111: 2170–2182. Gutierrez, A.H., Moise, L., and De Groot, A.S. (2012). Of [Hamsters] and men: a new perspective on host cell proteins. Hum. Vacc. Immunother. 8: 1172–1174. Hanania, N.A., Noonan, M., Corren, J. et al. (2015). Lebrikizumab in moderate-to-severe asthma: pooled data from two randomised placebo-controlled studies. Thorax 70: 748–756. Wang, X., Hunter, A.K., and Mozier, N.M. (2009). Host cell proteins in biologics development: identification, quantitation and risk assessment. Biotechnol. Bioeng. 103: 446–458. Beatson, R., Sproviero, D., Maher, J. et al. (2011). Transforming growth factor-beta1 is constitutively secreted by Chinese hamster ovary cells and is functional in human cells. Biotechnol. Bioeng. 108: 2759–2764. Hannon, G.J. and Beach, D. (1994). p15INK4B is a potential effector of TGF-beta-induced cell cycle arrest. Nature 371: 257–261. Li, M.O., Wan, Y.Y., Sanjabi, S. et al. (2006). Transforming growth factor-beta regulation of immune responses. Annu. Rev. Immunol. 24: 99–146. Nakao, A., Fujii, M., Matsumura, R. et al. (1999). Transient gene transfer and expression of Smad7 prevents bleomycin-induced lung fibrosis in mice. J. Clin. Invest. 104: 5–11. Pak, S.C., Hunt, S.M., Bridges, M.W. et al. (1996). Super-CHO-A cell line capable of autocrine growth under fully defined protein-free conditions. Cytotechnology 22: 139–146. Tscheliessnig, A.L., Konrath, J., Bates, R., and Jungbauer, A. (2013). Host cell protein analysis in therapeutic protein bioprocessing – methods and applications. Biotechnol. J. 8: 655–670. Eaton, L.C. (1995). Host cell contaminant protein assay development for recombinant biopharmaceuticals. J. Chromatogr. A 705: 105–114. Zhu-Shimoni, J., Yu, C., Nishihara, J. et al. (2014). Host cell protein testing by ELISAs and the use of orthogonal methods. Biotechnol. Bioeng. 111: 2367–2379. Krawitz, D.C., Forrest, W., Moreno, G.T. et al. (2006). Proteomic studies support the use of multi-product immunoassays to monitor host cell protein impurities. Proteomics 6: 94–110. Tait, A.S., Hogwood, C.E., Smales, C.M., and Bracewell, D.G. (2012). Host cell protein dynamics in the supernatant of a mAb producing CHO cell line. Biotechnol. Bioeng. 109: 971–982. Jin, M., Szapiel, N., Zhang, J. et al. (2010). Profiling of host cell proteins by two-dimensional difference gel electrophoresis (2D-DIGE): implications for downstream process development. Biotechnol. Bioeng. 105: 306–316.

309

310

12 Host Cell Proteins During Biomanufacturing

57 Grzeskowiak, J.K., Tscheliessnig, A., Toh, P.C. et al. (2009). 2-D DIGE to

58

59

60

61

62

63

64 65

66

67

68

69

70

71

72

expedite downstream process development for human monoclonal antibody purification. Protein Expr. Purif. 66: 58–65. Hogwood, C.E., Bracewell, D.G., and Smales, C.M. (2014). Measurement and control of host cell proteins (HCPs) in CHO cell bioprocesses. Curr. Opin. Biotechnol. 30: 153–160. Swaney, D.L., Wenger, C.D., and Coon, J.J. (2010). Value of using multiple proteases for large-scale mass spectrometry-based proteomics. J. Proteome Res. 9: 1323–1329. Schenauer, M.R., Flynn, G.C., and Goetze, A.M. (2012). Identification and quantification of host cell protein impurities in biotherapeutics using mass spectrometry. Anal. Biochem. 428: 150–157. Zhu, G., Sun, L., Linkous, T. et al. (2014). Absolute quantitation of host cell proteins in recombinant human monoclonal antibodies with an automated CZE-ESI-MS/MS system. Electrophoresis 35: 1448–1452. Zhang, Q., Goetze, A.M., Cui, H. et al. (2014). Comprehensive tracking of host cell proteins during monoclonal antibody purifications using mass spectrometry. mAbs 6: 659–670. Yuk, I.H., Nishihara, J., Walker, D. Jr., et al. (2015). More similar than different: host cell protein production using three null CHO cell lines. Biotechnol. Bioeng. 112: 2068–2083. Potty, A.S., Xenopoulos, A., Patel, S. et al. (2014). The effect of antiapoptosis genes on clarification performance. Biotechnol. Prog. 30: 100–107. Tait, A.S., Tarrant, R.D., Velez-Suberbie, M.L. et al. (2013). Differential response in downstream processing of CHO cells grown under mild hypothermic conditions. Biotechnol. Prog. 29: 688–696. Rossi, A., Kontarakis, Z., Gerri, C. et al. (2015). Genetic compensation induced by deleterious mutations but not gene knockdowns. Nature 524: 230–233. Li, Y. (2017). Effective strategies for host cell protein clearance in downstream processing of monoclonal antibodies and Fc-fusion proteins. Protein Expr. Purif. 134: 96–103. Chollangi, S., Parker, R., Singh, N. et al. (2015). Development of robust antibody purification by optimizing protein-A chromatography in combination with precipitation methodologies. Biotechnol. Bioeng. 112: 2292–2304. Shukla, A.A. and Hinckley, P. (2008). Host cell protein clearance during protein A chromatography: development of an improved column wash step. Biotechnol. Prog. 24: 1115–1121. Ng, P.K., He, J., and Snyder, M.A. (2009). Separation of protein mixtures using pH-gradient cation-exchange chromatography. J. Chromatogr. A 1216: 1372–1376. Chen, J., Tetrault, J., Zhang, Y. et al. (2010). The distinctive separation attributes of mixed-mode resins and their application in monoclonal antibody downstream purification process. J. Chromatogr. A 1217: 216–224. Eriksson, K., Ljungolf, A., Rodrigo, G., and Brekkan, E. (2009). MAb contaminant removal with a multimodal anion exchanger. Bioprocess Int. 7: 52–56.

References

73 Lewis, N.E., Liu, X., Li, Y. et al. (2013). Genomic landscapes of Chinese

74 75

76

77

hamster ovary cell lines as revealed by the Cricetulus griseus draft genome. Nat. Biotechnol. 31: 759–765. Godovac-Zimmermann, J. and Brown, L.R. (2001). Perspectives for mass spectrometry and functional proteomics. Mass Spectrom. Rev. 20: 1–57. Koren, E., De Groot, A.S., Jawa, V. et al. (2007). Clinical validation of the “in silico” prediction of immunogenicity of a human recombinant therapeutic protein. Clin. Immunol. 124: 26–32. Croughan, M.S., Konstantinov, K.B., and Cooney, C. (2015). The future of industrial bioprocessing: batch or continuous? Biotechnol. Bioeng. 112: 648–651. Levison, P. (2016). Moving one unit operation at a time toward continuous biomanufacturing: an overview of Pall solutions for integrated continuous biopharmaceutical production. Bioprocess Int. 14: s6–s9.

311

313

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals William C. Yang United States Food and Drug Administration, Center for Drug Evaluation and Research, 10903 New Hampshire Avenue, Silver Spring, MD 20993, USA

13.1 Introduction Recombinant proteins, such as enzymes, monoclonal antibodies, and fusion proteins, comprise an increasing number of drug and vaccine therapies in the market and coming to market. Proteins, with their primary, secondary, tertiary, and quaternary structures and posttranslational modifications, are structurally more complex than small-molecule drugs and are thus capable of more complex mechanisms of action, such as modulating the immune response or sequestering certain disease-causing molecules. Because of their complexity, therapeutic proteins cannot yet be made using chemical synthesis. These biological macromolecules intended for therapeutic use in humans and animals are manufactured using bacterial, yeast, or mammalian cells whose cellular machinery has been genetically engineered to overproduce the humanized proteins of interest using recombinant DNA technology. Although some of the biopharmaceuticals in the market are produced using Escherichia coli and Saccharomyces cerevisiae, the majority (40–56%) of today’s biopharmaceuticals are produced using mammalian cells (Chinese hamster ovary [CHO], Sp 2/0, NS0) because of the ability of mammalian cells to glycosylate proteins in a manner similar to that of human cells [1, 2]. The preceding chapters of this book have covered the genetic manipulations that enable mammalian cells to produce heterologous proteins, e.g. protein engineering, DNA delivery and integration, host cell engineering, selection, gene amplification, and cell line screening. This chapter focuses on one of the key roles an industrial cell culture process scientist plays in bringing biopharmaceutical proteins to the market: designing a mammalian cell culture process that maximizes productivity while also balancing product quality.

Disclaimer: This book chapter reflects the views of the author and should not be construed to represent FDA’s views or policies. Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

314

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

13.2 Objectives of Cell Culture Process Development 13.2.1

Yield and Product Quality

One of the main goals of commercial cell culture process development is to produce large quantities of the recombinant protein of interest to supply clinical trials and the commercial market. As biologics form an increasing proportion of the pipelines of pharmaceutical companies, there is increasing pressure for the manufacturers to increase protein production yield. This push to increase yield is further compelled by the high failure rate of new compounds. It is not cost-effective to build, staff, and maintain additional manufacturing facilities to accommodate demand that could potentially vaporize based on clinical trial results. Therefore, it is advantageous for cell culture scientists to design processes that can maximize the amount of protein produced per unit time using their existing manufacturing infrastructure. However, cell culture process development cannot be singularly focused on yield because a biological macromolecule is more complex than a chemical small molecule. Not only does the primary amino acid sequence have to be correct, but the product quality attributes also have to be correct and consistent as well. The cell line, media, chemical additives, and process variables can all greatly affect product quality attributes such as glycosylation, charge heterogeneity, and protein aggregation. These product quality attributes are important drug release specification parameters because they often govern the stability, function, and efficacy of the protein drug product. Therefore, not only is it important to maximize the yield of the protein but it is also important to maximize the manufacture of the safe and effective species of the protein of interest. This chapter will not perform in-depth analysis of protein product quality attributes as other publications have reviewed this topic in great detail [3–5]. Rather, this chapter will introduce the basic background and comment on the effects that particular fed-batch process design choices can have on these product quality attributes. Often times, these effects are highly dependent on the cell line, media, protein of interest, or a combination of the three. Therefore, the reported findings in the literature can only be taken as a reference and are not predictive of your process or protein of interest. 13.2.2

Glycosylation

Depending on the protein of interest and its intended use, the glycosylation pattern can be almost as important as the primary sequence itself. Mammalian cells attach sugar species to specific sites on proteins. In recombinant proteins for therapeutic uses, there are two types of glycosylation sites: N-linked and O-linked. N-linked glycosylation has sugar structures attached to asparagine, whereas O-linked glycosylation has sugar structures attached to serine and threonine. The most common type of glycosylation found in biopharmaceutical proteins is N-linked glycosylation. This glycosylation occurs at a conserved asparagine at site 297 in the Fc constant region of monoclonal antibodies and Fc fusion proteins. Some antibodies have additional N-linked glycosylation sites in the variable

13.2 Objectives of Cell Culture Process Development

region. These oligosaccharide structures have complex branching patterns and are composed of many different types of sugars. The branches are composed of N-acetylglucosamine and mannose species. These branches are either capped by galactose, mannose, or sialic acid. These oligosaccharide caps affect the biological function of the antibody. High mannose has high antibody-dependent cell-mediated cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) effects and high serum clearance [6]. High galactose enhances ADCC and CDC, which can be the main mechanisms of action for certain recombinant protein drugs [7]. High sialylation in the form of N-acetylneuraminic acid (NANA) promotes longer drug half-life and tempers the immunogenic response. The presence of the competing sialic acid, N-glycolylneuraminic acid (NGNA), results in higher levels of immunogenicity [6]. Human cells cannot make NGNA because of a mutated enzyme [8], but the CHO and NS0 cells commonly used in biomanufacturing are capable of producing it and do so at low but still immunogenic levels. Thus, recombinant proteins that contain NGNA are immunogenic and are an undesired glycoform that must be minimized. The greatest determinant of ADCC and CDC levels for an antibody is the absence or presence of a fucose sugar at one of the core N-acetylglucosamines near the asparagine. Lack of fucosylation elicits a stronger immune response [6]. Recombinant proteins for therapeutic purposes can also have O-linked glycosylation. This is typically found in nonantibody proteins such as erythropoietin or fusion proteins where a portion of a biologically active protein (e.g. ligand or antigen) is combined with the Fc domain of an antibody (to obtain stability or pharmacologic benefits) using recombinant DNA technology. These O-linked glycosylation sites also have similar functional ends as N-linked glycosylation, e.g. sialylation, to provide biological and pharmacological effects [9, 10]. 13.2.3

Charge Heterogeneity

As recombinant proteins reside in the bioreactor from synthesis to harvest, the chemical, physical, and enzymatic environment of the bioreactor process can modify certain moieties on the protein. These chemical modifications are heterogeneous and affect the charge of the protein; charge variants with a lower isoelectric point (pI) are considered acidic, whereas charge variants with a higher pI are considered basic. Chemical modifications such as deamidation, glycation, C-terminal lysine cleavage, or adduct formation (addition of COOH or loss of NH2 ) result in acidic species. Conversely, chemical modifications such as C-terminal lysine or glycine amidation, succinimide formation, amino acid oxidation, or removal of sialic acid by cell-released sialidases result in basic species [11]. These charge variants can affect antibody structure, stability, and biological functions [12]. It has been hypothesized that a number of degradative and oxidative processes can increase the acidic species, which can affect protein stability [12, 13]. However, a study on the effect of global charge heterogeneity on protein stability did not appear to suggest a linkage between acidic species and stability [14]. Basic species have been shown to have higher binding to Fc receptors, potentially increasing serum half-life [15]. However, the literature also speaks to the protein-specific nature of these types of conclusions as other

315

316

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

researchers have fractionated the acidic, basic, and main species of an antibody, administered it in animals, and showed that all species had similar Fc receptor binding and pharmacokinetics [11]. 13.2.4

Aggregation

Protein folding is a very complex process that involves the primary sequence as well as the chemical environment. Often, there is a proportion of the product that does not fold correctly into the soluble, active, monomeric form. An incorrectly folded subpopulation aggregates with itself or other incorrectly folded species in soluble or insoluble form and thus results in a loss of yield. A few of the most common causes are the thermodynamics of the amino acid primary sequence, premature truncation, and the redox environment of the bioreactor [16]. There are process levers that can potentially modulate the folding of the protein, such as temperature and pH.

13.3 Cells and Cell Culture Formats 13.3.1

Adherent Cells

The earliest biopharmaceutical cell culture processes originated from laboratory tissue culture with adherent cells. These processes were an extension of primary cell cultures on petri dishes, where cells were plated on tissue cell culture plastic, covered in complex, serum-containing media, in an oxygen-rich environment, and incubated at a physiological temperature [17]. Scale-up of these processes was challenging – they were scaled out rather than up. Amgen’s original Epogen (erythropoietin) process used rooms of cells seeded in roller bottles [18]; vaccine manufacturers seeded their cells on microcarriers and cultivated the microcarriers in stirred tank bioreactors [17]. Productivity was also a challenge because of contact inhibition and the low surface area to volume ratio; the original Epogen yield was approximately 50–200 mg/l [18]. 13.3.2

Suspended Cells

Over the years, cell culture scientists weaned adherent cells from their anchorage-dependent nature and created suspension-adapted cell lines to take advantage of the industrial scalability offered by stirred tank bioreactors [17, 19]. These cells can be further adapted in serial bioreactor cultures to thrive in the more hydrodynamically stressful environment of a stirred-tank, sparged bioreactor [20]. By culturing the cells free in suspension, the cultures could be easily and rapidly expanded in volume through subsequent dilutions of the batch cultures with fresh cell culture media in a series of progressively larger bioreactors, typically called a seed train. The starting concentration of the suspended cells (viable cells/ml) gradually increases along each stage of the seed train until it finally reaches the last and largest bioreactor, the production bioreactor (Figure 13.1). The number of stages and seeding densities depend on the size of the final production bioreactor and the growth rate of the cell.

13.4 Fed-batch Cultures

Rocking bag bioreactor

Cell bank

Shaker flasks

Shaker flasks

N-X Bioreactor

N-1 Bioreactor

Production bioreactor

Feed tank

Figure 13.1 Representative mammalian fed-batch cell culture process steps in manufacturing. A cell bank containing frozen, preserved starter culture is thawed into either shaker flasks or a rocking bag bioreactor containing cell culture media. Cultures are serially propagated in a series of vessels increasing in volume until it reaches the production bioreactor. The production bioreactor process is optimized for yield and productivity. Product will be harvested at the end of the culture duration and sent for further downstream processing, e.g. purification and formulation.

13.3.3

Batch Cultures

In batch culture, the inoculum and media are combined on day 0 and the culture is cultivated until the harvest day. No additional nutrients (e.g. amino acids, glucose, and vitamins) are provided to the cells, but temperature, pH, dissolved oxygen, and foam levels can be controlled [21]. If the batch culture is used as part of the seed train where the cells are the product, the culture is diluted down to a lower cell density with fresh media at the end of the culture stage to seed the subsequent stage/bioreactor of the seed train. The process then repeats. If the batch culture is used as the production vessel where the product is the protein of interest, the bioreactor will be harvested at the end of the culture duration. The cells and spent media would be separated, and the spent media would be sent for downstream purification of the protein of interest. Alternatively, the batch production bioreactor could be partially harvested and the remaining cells diluted with fresh media to restart a subsequent batch production stage [22]. The advantage of batch cultures is in its simplicity. There are no input modifications post-inoculation and the culture simply needs to be monitored. However, without additional nutrients, the cell densities and culture durations are relatively low. Correspondingly, the productivity is relatively low, as volumetric productivity is dependent on the specific productivity of the cells, number of cells, and culture volume.

13.4 Fed-batch Cultures As biopharmaceutical products increasingly commanded a larger market share of all pharmaceuticals, the industry needed to find ways to increase productivity

317

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

(a)

Nutrients

Viable cell count

in order to meet the protein demand in a cost-effective manner. Generally, the approach to increase productivity has been two-pronged: the first is to increase the specific productivity of the host cell through cell line engineering and the second is to increase the volumetric productivity through process development, e.g. through higher cell densities, higher product titers, higher culture volumes, and higher throughput. One of the main reasons batch cultures fail to reach high cell densities is due to nutrient depletion. By supplementing the batch culture with additional nutrients such as glucose, amino acids, vitamins, and trace metals, we can meet the cell nutrient needs and prolong the cultures to reach higher cell densities, thereby increasing volumetric productivity (Figure 13.2). However, these gains come with a cost. The higher cell densities and higher volumetric productivities of fed-batch cell culture have undesirable trade-offs that complicate process design, e.g. mass transfer limitations, waste accumulation, changes in product quality, and more complex operational workflows with additional feeds and mid-process perturbations providing more opportunities for human error. As fed-batch cultures reach higher cell densities, mass transfer of nutrients to the cell and waste from the cell become a limiting factor. The push for larger culture volumes to increase volumetric productivity further complicates mass transfer by introducing bioreactor scale-up complexities. The higher cell densities also result in higher levels of cell waste. As fed-batch culture has only inflows and no outflows, cell waste such as dead cell debris,

Culture days

(c)

(b)

Culture days

Titer

Waste

318

Culture days

(d)

Culture days

Figure 13.2 Representative schematics comparing process parameters of batch (dashed) and fed-batch cell culture (solid). (a) In general, fed-batch cell cultures have longer durations and have higher cell mass. (b) Feeding results in higher nutrient levels in fed-batch cell culture, (c) while the longer duration results in higher levels of waste accumulation. (d) Nutrient feeding, higher cell mass, and longer culture duration results in higher fed-batch titers.

13.5 Cell Culture Media

ammonia, lactate, and other metabolic by-products accumulate and poison the cell. The higher levels of cell waste not only affect cell viability and productivity but also affect product quality, such as charge heterogeneity, glycosylation, and other posttranslational modifications. Therefore, developing a robust cell culture process for a given protein of interest cannot be singularly focused on volumetric productivity as there are typically trade-offs between productivity and product quality. This chapter discusses key considerations necessary to develop a fed-batch process: media design, process variable (temperature, pH, oxygen, and culture duration) design, optimizing the process, bioreactor scale-up, product quality, and process throughput. We will also discuss new and emerging technologies to improve fed-batch cell culture development and throughput.

13.5 Cell Culture Media 13.5.1

Basal Media

Basal media is the initial growth media for the batch cultures. It contains the basic components necessary for the propagation of cells, for example, glucose, amino acids, vitamins, inorganic salts, and growth factors [23]. Most of the mammalian cell culture media used in the industry is based on a classical 50/50 volume per volume mixture of Dulbecco’s modified Eagle’s medium (DMEM) and Ham’s F-12 (F-12) Media. DMEM contains nutrients and buffering agents for pH while F-12 further enriches the media by providing additional trace elements, nonessential amino acids, and other nutrients [24]. Jayme et al. describes the composition of DMEM/F12 as well as other commonly used versions of classical media. However, mammalian cells need other components such as lipids, hormones, growth factors, and carrier proteins in order to achieve robust growth and productivity. Early cell culture media obtained these additional factors from fetal, calf, or bovine sera [25]. As the dominance and reach of biopharmaceuticals grew, the concern of virus and prion contamination from the animal sera medium components grew as well [26, 27]. In response, biopharmaceutical manufacturers replaced sera with yeast, soy, wheat, pea, cotton, and other plant-derived hydrolysates to obtain similar nutrients and factors, but without the animal-derived risk [28]. The hydrolysates lacked hormones such as insulin and carrier protein-like albumin or transferrin, so cell culture scientists supplemented these components exogenously [29]. Even though the hydrolysates offered a greater level of control over the quality and components of the complex serum-containing media and avoided the animal-derived risk, there were still issues of raw material variability because of the difficulty to control and undefined nature of the hydrolysis input components [30, 31]. Although legacy processes still utilize media containing sera or hydrolysates, current state-of-the-art cell culture media now do not contain any sera or hydrolysates. All components such as amino acids, inorganic salts, buffers, trace elements, vitamins, hormones, and lipids are added separately in known

319

320

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

quantities and the resulting media are referred to as chemically defined media (CDM). Although hydrolysate had quality control issues because of its nature as a complex mixture of digested proteins, it nevertheless provided a high level of shear protection for the cells because of the high protein content [32]. The advancements made in cell culture media enabled growth to high viable cell densities and resulted in greater volumes of aeration mass transfer. This further accentuated the shear protection benefit of hydrolysate because the bubbling within the bioreactor intensified as the sparging increased to keep up with the cell growth. In the absence of hydrolysate, key chemical surfactants such as Pluronic F68 were added to the CDM to provide the missing shear stress and bubble protection for cultured cells in the high-intensity sparging environment of the bioreactor [33]. The absence of rich, undefined media components such as sera or hydrolysates also highlights the importance of trace metal and vitamin levels in CDM. Classical media [34] often suggest a baseline level of the micronutrients, but cell growth and titer production can often be improved by increasing and modulating the levels [35, 36]. Individual trace metals also play a role in determining the aforementioned product quality of the protein products because of their role as cofactors in many glycosylation enzymes. Manganese is a glycosyltransferase cofactor and has been shown to increase galactosylation and sialylation [37, 38]. Copper plays a role in changing the charge variants – high levels have been shown to promote higher basic species through increased C-terminal proline amidation [39] and C-terminal lysine processing [30, 40]. Higher copper levels have also been shown to not only increase basic species but also promote a lactate consumption phenotype leading to healthier cultures and higher titers [41]. Although iron is not a trace metal, elevated levels of iron have been shown to increase acidic species and darken the color of the resulting protein product [42]. The same study also demonstrated that increased B vitamin levels of B2 (riboflavin), B6 (pyridoxine and pyridoxal), B9 (folic acid), and B12 (cyanocobalamin) also have a similar effect. The major commercial cell culture media vendors have created extensive menus of CDMs, and many cell culture scientists have used them to develop, improve, and optimize their processes [43, 44]. Many of the large biopharmaceutical manufacturers have also developed their own proprietary CDM formulations to reduce costs and maintain a greater level of control over their processes [36, 45, 46]. These large biopharmaceutical companies have even made portions of their CDM libraries available for the public to pursue through various published patents and patent applications. 13.5.2

Feed Media

The basal media typically provides sufficient nutrients for batch cultures because the viable cell densities do not grow to high levels and the culture duration is not long. Some batch cultures may supplement or front-load higher levels of selected, highly consumed nutrients such as glucose or glutamine. However, fed-batch culture requires additional, more complex nutrient supplementation because of its higher cell densities and longer culture duration. These additional, more sizable nutrient supplementations are referred to as feeds. Depending on the cell line and

13.6 Feeding Strategies

process, the feeds can occur sporadically on preset days [47, 48] or daily [49, 50], either as a bolus feed [47–50] or as a continuous feed [51]. The earliest forms of feed media were simply more concentrated forms of the basal media with the salts removed [52]. However, because of solubility issues and potentially uneven consumption based on the cell line of interest, this would lead to undesired depletion and accumulation of certain nutrients. A more turnkey solution would be to use commercially available feed media from media vendor companies. Vendors typically suggest pairings of their proprietary basal and feed media and recommend a simple feeding schedule [53]. Scientists have also mixed and matched basal and feed media in order to quickly arrive at a semioptimized solution [43]. However, there is little control or insight into these types of processes as the formulation is not public. In order to optimize the cell culture process and gain more control over the process inputs, scientists may choose to rationally design the feed media based on the consumption needs of the cell line of interest using metabolic analysis.

13.6 Feeding Strategies Before we optimize the cell culture basal and feed media, we must first utilize a cell culture process with a baseline chemically defined process in order to generate the samples for metabolic analysis to obtain starting point metabolite data for the subsequent design of the custom feed media. A popular baseline or starting medium and feed combination is the aforementioned DMEM/Ham’s F12 (1 : 1) basal medium with a 10× concentrated version of the basal medium without salts as feed [52]. 13.6.1

Metabolite Based

During the initial experiment to generate baseline data using a never-beforetested combination of cells, media, and feed, cell culture scientists must strictly monitor the metabolite levels in order to make appropriate feed amount and timing decisions that enable the culture to grow robustly and generate high-quality data for subsequent analysis. Today’s rapid metabolite analyzers can utilize spent media to quickly provide a measure of commonly monitored nutrients and process indicators such as glucose, glutamine, glutamate, ammonium, lactate, pH, osmolality, and salt levels [54]. A common strategy for feed determination is to regularly sample before feeding and monitor the levels of glucose, glutamine, ammonium, lactate, and osmolality. Mammalian cells need a baseline level of glucose and glutamine (unless cells containing a glutamine synthase gene are used) in order to survive, so these levels provide a good indicator for determining the feed amount. The glucose uptake rate can be calculated and used to approximate the amount of feeding necessary to maintain a glucose set point. Too much glucose can deleteriously affect the health of the culture. Exposure to high levels of glucose can shift the metabolism of the cell from a more efficient state where the cell

321

322

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

primarily channels glucose and other nutrients through the tricarboxylic acid (TCA) cycle to a less efficient state where the cell channels glucose through lactic acid fermentation [55]. Lactate accumulation is a key indicator of this metabolic shift. Elevated levels of glucose have also been shown to adversely affect product quality by increasing the acidic species through glycation [56]. Glycation is the spontaneous, concentration-driven covalent attachment of glucose to lysine moieties on the antibody. A pH-dependent feeding strategy has also been developed to combat lactate accumulation. When glucose levels are low, cells consume lactate as an alternative carbon source. This results in an increase in culture pH. Cell culture scientists have used the uptick in culture pH as a trigger to feed glucose as necessary [57]. The cycle of lactate consumption and glucose feeding results in lower lactate consumption and allows the cells to reach a higher density. This in turn leads to higher titers. This feeding strategy illustrates the complex interaction between lactate and glucose metabolic states of mammalian cells in bioreactors. There have been in-depth studies into what controls the balance between the high glycolysis, high lactate production state, and the low glycolysis, low lactate production state [58, 59]. The transition between these two states has been characterized as bistable, with bioreactor glucose concentrations governing the transition between the two states. Higher glucose concentrations push the metabolism toward the high glycolysis state. Therefore, feeding strategies need to consider the delicate balance between glucose and lactate metabolism to maintain the culture in a low lactate, high productivity regime. Ammonium and osmolality are additional indicators of overfeeding. If there are too many amino acids for the cells to utilize, the amino groups on the excess amino acids deaminate and result in an accumulation of ammonium, which is toxic to the cell [60]. Even if the elevated levels of ammonia have not reached the point of cytotoxicity, they can still inhibit glycosylation through increasing the trans-Golgi compartment pH, which inhibits glycosyltransferase activity [61] and expression [62]. In addition to reducing overall feeding, specifically reducing the amounts of glutamine and asparagine or replacing them with glutamate or pyruvate in the feed can reduce the culture ammonia because the side chains of glutamine and asparagine contribute additional ammonium molecules [63]. Thus, controlling asparagine at lower levels in cell culture can increase antibody galactosylation, likely from reducing the glycosylation inhibitory ammonium levels in the cultures [64]. Osmolality also increases if the cells were fed an excess amount of nutrients that they were not able to consume. High osmolality and rapid increases in osmolality are growth inhibitory and can cause osmotic stress to the cell [65]. Specific levels that trigger inhibition depend on the cell line as some cell lines are more or less tolerant than others. However, moderate and gradual increases in osmolality shift the cell cycle from growth to production and can increase the specific productivity of the cell, resulting in higher titers [66–68]. High osmolality has also been observed to affect glycosylation and can generate higher levels of mannose [38] and lower levels of fucose [69] glycoforms.

13.7 Feed Media Design

13.6.2

Respiration Based

Another popular feeding strategy is based on the oxygen uptake rate (OUR) of the cells [55]. Oxygen consumption is a direct and real-time measurement of the cell metabolism and can provide a more adaptive way of feeding if a stoichiometric relationship between OUR and nutrient consumption can be determined. We can calculate OUR in a variety of ways using the following equation: dC (13.1) = kL a × (C ∗ − C) − OUR dt where dC/dt is the concentration of oxygen over time, k L a is the oxygen transfer coefficient empirically determined for the vessel and conditions of interest, C* is the oxygen saturation in the bulk liquid, C is the dissolved oxygen concentration in the bulk liquid, and OUR is the oxygen uptake rate. The first method for calculating the OUR requires changing the dissolved oxygen content of the culture such that the dissolved oxygen is in equilibrium with the liquid phase. This renders the (C* − C) term = 0. By measuring the change in oxygen concentration over time, we can determine the OUR. This may be more easily and practically performed by taking a cell culture sample from the bioreactor and performing the oxygen perturbation and measurement offline so that the main culture is not disturbed. An alternative method maintains the dissolved oxygen content of the bioreactor but requires foreknowledge and predetermination of the k L a. This renders the dC/dt term = 0 and the OUR can be calculated by the product of the k L a and (C* − C). However, the k L a term depends on the bioreactor configuration, baffles, impellers, cell culture volume, and agitation rate. In any of these scenarios, it is important that certain key pieces of data such as the cell culture conditions, feed timing, and feed volume are collected for the media optimization calculation exercise. The most important aspect of the baseline study is the timed and proper collection of the spent media, taken before adding the feed media. The cells must be quickly separated from the spent media and the samples must be quickly frozen in order to preserve the state of the metabolic indicators as much as possible.

13.7 Feed Media Design After obtaining the daily (or perhaps more frequent) spent media retain samples from the baseline study, we then measure the level of metabolites to determine the degree of consumption and therefore needs for supplementation. Currently, spent media analysis is typically performed for each clone selected to be the cell line for each cell culture process. This is because different clones behave differently when subjected to the same culture conditions and media [70]. The most commonly used form of cell line generation is still nonhomologous random integration of the gene of interest [71]; this stochastic nature leads to clonal variation. The rapid metabolite analyzers described above provide data at a level sufficient for daily monitoring and feeding decisions, but not at a level sufficient for

323

324

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

media design and development. To fully capture all of the nutrients that need to be considered in a media development project, more granular information is needed in the form of total amino acids [72], vitamins [73], and metals [74]. The current state of the art has been to obtain the complete panel of metabolic information in a piecewise manner through multiple mass spectrometry and chromatography-based methods. In order to simplify analysis, there has been promising research into using nuclear magnetic resonance (NMR) in order to obtain all the information using one assay to save time and resources [75]. Using the empirically determined cell culture, feed timing, feed volume, and metabolite data from the baseline study, we can calculate the amount of a given nutrient consumed by the cells through the mass balance equation of: Accumulation = Input − Output + Generation − Consumption

(13.2)

where accumulation is the change in the nutrient concentration from time point 1 to time point 2 and input is the amount of the nutrient in the feed. As this is fed-batch culture, there are no outputs, and we assume that the cell cannot generate its own nutrients for the sake of simplicity. If there is indeed nutrient generation, e.g. nonessential amino acids, then the calculated consumption rate would be negative. Bioreactor volume would need to be tracked and accounted for over time as the volume contribution effect of the bolus feed additions are inherently captured by the metabolite analyzer concentration results. Depending on the sample volume, cell growth, and culture duration, the feed volume contributions can be significant. The nutrient needs for each particular cell line are then converted to a concentrated feed solution by accounting for the desired bioreactor harvest volume, desired cumulative feed volume, feed timing, culture duration, and the desired minimum concentration. There are often many challenges and constraints in formulating the calculated feed solution. Solubility is a key issue as the bioreactor volume constraints and the cell uptake amount can result in the feed component exceeding its solubility limit in the feed solution. This has been addressed by using surfactants such as polysorbate 20 or 80 to solubilize the problematic components [76]. In other instances, the presence of certain feed components in high concentrations along with the sodium bicarbonate buffer resulted in precipitation. Methods to address this issue include adding pyruvate to the formulation to improve stability [77]. Tyrosine and cysteine are two key amino acids that are not very soluble in water. As extreme pH can help solubilize recalcitrant feed components, cell culture scientists have formulated these two amino acids and other hard-to-dissolve components separately from the other feed components in a second more concentrated but highly basic feed solution [78]. Another approach used di-peptides and tri-peptides to increase the solubility of tyrosine and cysteine [79], but this approach is temporary as the polymers eventually degrade into monomers and fall out of solution. A more elegant solution chemically synthesized more water-soluble versions of the amino acids that the cells could still utilize after some enzymatic processing. For example, feeding a chemically-synthesized soluble cysteine derivative that uses an intracellular enzyme to convert to the metabolically active l-cysteine [80] and a similarly solubilized tyrosine

13.8 Process Variable Design

derivative that uses extracellular phosphatases to convert to the metabolically active l-tyrosine [81], both improved the growth and yield of fed-batch cultures.

13.8 Process Variable Design 13.8.1

Temperature

A key consideration of any cell culture process is the balance between growth and production. A cell has finite resources; it either devotes energy and building blocks toward producing more of itself or more of the protein of interest. The main objective is to enable sufficient growth of cells to get to a critical mass and then focus the cell energy and resources toward producing protein while sustaining themselves and total cell mass. One of the most effective ways to trigger the transition from growth to production is shifting the culture temperature downward. Typically, cells grow the fastest at physiological temperature, e.g. 35–37 ∘ C. Shifting to a condition of mild hypothermia, e.g. 28–33 ∘ C, can arrest the cell cycle and shift the cells from S phase to G1 phase [82]. This increases the productivity of the cells and can result in higher harvest titer [83, 84]. However, like many of the approaches discussed in this chapter, the results may vary from cell line to cell line and process to process. Culture temperature also plays a role in determining the protein product quality. Although we have reports that reducing the temperature can increase productivity without changing the glycosylation pattern [85, 86], we also have reports that reduced temperatures decrease the sialylation [87]. Reducing the culture temperature increases the yield of the properly folded protein and reduces product aggregation by slowing down the molecular interactions and allowing more time for protein folding to occur [88]. Lower culture temperatures have also been shown to reduce the percentage of acidic charge variants [84]. The timing (culture day) of the temperature shift also plays a role in determining the extent of the productivity and product quality change. 13.8.2

pH and pCO2

Another key process variable that globally affects cell growth and protein production is pH. Again, mammalian cells prefer physiological conditions, so the majority of cell culture processes control bioreactor pH at approximately 7.0. Most mammalian cell culture processes start off at a pH >7.0 as cells grow faster at higher pH [89]. However, as cells grow, die, and lyse, they release acidic compounds such as lactic acid that acidify the bioreactor environment. Thus, processes either shift the pH set point lower after the initial growth to reduce lactate production or maintain the pH set point using a weak base in the form of sodium carbonate. However, there are instances where an alternative base such as sodium hydroxide is preferred [90] because pH and CO2 are related by chemical equilibrium and the Henderson–Hasselbalch equation: [HCO−3 ] pH = 6.1 × log (13.3) [0.03 × pCO2 ]

325

326

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

The additional CO3 from sodium carbonate can increase the pCO2 of the process if the cell mass is high and the mass transfer cannot efficiently strip the dissolved CO2 from the bioreactor. High pCO2 acidifies the culture and leads to a vicious feed-forward pH control cycle of base addition that increases the sodium concentration and upsets the osmotic balance of the bioreactor. However, there are instances where high pCO2 can be desired. High levels of pCO2 been shown to be a process lever in reducing the amount of NGNA [91]. Because of the Henderson–Hasselbalch relationship, it can be difficult to decouple pH control and pCO2 . However, this can be done by feeding lactate instead of sparging CO2 to lower the culture pH [48]. In this specific study, feeding lactate led to healthier cultures, which had the indirect benefit of reducing ammonia in the culture. The drawback from tight pH control in a mammalian cell culture process is the additional osmolality that result from the different perturbations to maintain pH. Additions of sodium carbonate, hydroxide, or lactate all contribute toward increasing the culture osmolality and sodium concentrations. Therefore, many current cell culture processes employ a wide pH dead band around a neutral, physiological pH and allow the culture to naturally drift within an acceptable range unless there was a need for tighter or specific level of pH control required to achieve a particular product quality attribute [70]. However, the differences in the surface aeration contribution, mixing, and other scale-dependent effects make commercial implementation of the wide pH dead band in the large production bioreactor difficult without a properly qualified scale-down model that mimics the fluid dynamics of the production scale at lab scale [92–95]. There are instances where product quality dictates control at a specific pH or within a pH range. For example, shifting to a lower pH can reduce protein aggregation [96], while a higher pH can reduce glycosylation through disrupting Golgi apparatus activities as previously mentioned in this chapter. 13.8.3

Dissolved Oxygen

Mammalian cells can tolerate a wide range of dissolved oxygen (DO) concentrations ranging from 10% to 100%; DO variations have a cell line- and protein-specific effect on product quality [3]. Generally, high DO set points (>50%) can result in oxidative stress [97], whereas low DO set points (100 × 106 vc/ml and result in a packed cell volume of up to 40%, both of which are typically an order of magnitude higher than the current fed-batch processes [153]. The higher cell masses thus enable higher titers, ranging from 12 to 27 g/l [153–156]. However, such a high cell mass makes mass transfer challenging. It can become difficult to deliver oxygen as well as remove CO2 . Furthermore, the higher packed cell volumes not only need to be factored into volume-adjusted productivity and capacity calculation but the higher packed cell volumes also create cell separation difficulties for purification process development because typical centrifuges are not equipped to handle packed cell volumes greater than 15%. New separation techniques must be developed to accommodate higher cell masses, such as flocculation [157, 158].

13.11 Future Directions Mammalian cell culture has come a long way from the early days of entire manufacturing facilities filled with rooms of roller bottle batch cultures. It is a constantly evolving field that makes full use of the available technology to increase productivity and capacity while reducing costs. For example, the commercial roller bottle process that originally produced Epogen was recently converted to one using state-of-the-art stirred-tank bioreactors [159]. Process development scientists have made the most of the available media, process, and bioreactor technologies to maximize the productivity of CHO cell hosts (Figure 13.3). Most of the recent gains in productivity were incrementally obtained through increased cell mass. In order to disruptively move biopharmaceutical productivity and capacity while maintaining control over product quality, a greater proportion of the next wave of technological advances in biopharmaceutical production may need to come from new cell lines with higher specific productivities and customized glycosylation machinery. Even then, the biopharmaceutical industry may be limited by the relatively slow growth and lower specific productivities of mammalian cell lines. Therefore, to meet the increased demand and expectations, the next generation of biopharmaceutical

References

14 12 Titer (g/l)

10 8 6 4 2 0 1980

1990

2000 Year

2010

2020

Figure 13.3 Improvements in fed-batch cell culture titers over time. Advances in cell line technology, cell culture media development, cell culture process techniques, and bioreactor equipment have enabled mammalian cell culture fed-batch yields to increase from 10 g/l in less than 20 years. The majority of current cell culture processes consistently yield approximately 5 g/l titers. These data points were compiled from examples of fed-batch titers in the literature [18, 35, 36, 48, 49, 51, 57, 70, 81, 84, 105, 120, 133, 154, 160–176].

cell lines could come from fungi, yeast, or bacteria engineered using the latest advances in synthetic biology to produce full-length humanized, glycosylated, antibodies, and fusion proteins at a high rate [71, 177, 178]. When that day arrives, cell culture process scientists of that era will again have the exciting opportunity to explore the capabilities of their new hosts and rewrite the chapter on fed-batch cell culture.

References 1 Nielsen, J. (2013). Production of biopharmaceutical proteins by yeast:

advances through metabolic engineering. Bioengineered 4 (4): 207–211. 2 Baeshen, N.A., Baeshen, M.N., Sheikh, A. et al. (2014). Cell factories for

insulin production. Microb. Cell Fact. 13: 141. 3 Hossler, P., Khattak, S.F., and Li, Z.J. (2009). Optimal and consistent protein

glycosylation in mammalian cell culture. Glycobiology 19 (9): 936–949. 4 Gramer, M.J. (2014). Product quality considerations for mammalian cell cul-

5 6 7

8

ture process development and manufacturing. In: Mammalian Cell Cultures for Biologics Manufacturing (eds. W. Zhou and A. Kantardjieff), 123–166. Berlin, Heidelberg: Springer-Verlag. Liu, H., Gaza-Bulseco, G., Faldu, D. et al. (2008). Heterogeneity of monoclonal antibodies. J. Pharm. Sci. 97 (7): 2426–2447. Reusch, D. and Tejada, M.L. (2015). Fc glycans of therapeutic antibodies as critical quality attributes. Glycobiology 25 (12): 1325–1334. Thomann, M., Reckermann, K., Reusch, D. et al. (2016). Fc-galactosylation modulates antibody-dependent cellular cytotoxicity of therapeutic antibodies. Mol. Immunol. 73: 69–75. Varki, A. (2010). Colloquium paper: uniquely human evolution of sialic acid genetics and biology. Proc. Natl. Acad. Sci. U.S.A. 107 (Suppl. 2): 8939–8946.

333

334

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

9 Spahr, C., Shi, S.D., and Lu, H.S. (2014). O-glycosylation of glycine-serine

10 11

12

13

14

15

16 17 18 19

20

21

22

23 24

25

linkers in recombinant Fc-fusion proteins: attachment of glycosaminoglycans and other intermediates with phosphorylation at the xylose sugar subunit. MAbs 6 (4): 904–914. Sola, R.J. and Griebenow, K. (2010). Glycosylation of therapeutic proteins: an effective strategy to optimize efficacy. BioDrugs 24 (1): 9–21. Khawli, L.A., Goswami, S., Hutchinson, R. et al. (2010). Charge variants in IgG1: isolation, characterization, in vitro binding properties and pharmacokinetics in rats. MAbs 2 (6): 613–624. Du, Y., Walsh, A., Ehrick, R. et al. (2012). Chromatographic analysis of the acidic and basic species of recombinant monoclonal antibodies. MAbs 4 (5): 578–585. Ponniah, G., Kita, A., Nowak, C. et al. (2015). Characterization of the acidic species of a monoclonal antibody using weak cation exchange chromatography and LC-MS. Anal. Chem. 87 (17): 9084–9092. Sule, S.V., Fernandez, J.E., Mecozzi, V.J. et al. (2017). Assessing the impact of charge variants on stability and viscosity of a high concentration antibody formulation. J. Pharm. Sci. 106 (12): 3507–3514. Hintersteiner, B., Lingg, N., Zhang, P. et al. (2016). Charge heterogeneity: basic antibody charge variants with increased binding to Fc receptors. MAbs 8 (8): 1548–1560. Cromwell, M.E., Hilario, E., and Jacobson, F. (2006). Protein aggregation and bioprocessing. AAPS J. 8 (3): E572–E579. Chu, L. and Robinson, D.K. (2001). Industrial choices for protein production by large-scale cell culture. Curr. Opin. Biotechnol. 12 (2): 180–187. Wurm, F.M. (2004). Production of recombinant protein therapeutics in cultivated mammalian cells. Nat. Biotechnol. 22 (11): 1393–1398. Astley, K., Naciri, M., Racher, A., and Al-Rubeai, M. (2007). The role of p21cip1 in adaptation of CHO cells to suspension and protein-free culture. J. Biotechnol. 130 (3): 282–290. Prentice, H.L., Ehrenfels, B.N., and Sisk, W.P. (2007). Improving performance of mammalian cells in fed-batch processes through "bioreactor evolution". Biotechnol. Progr. 23 (2): 458–464. Lim, H.C. and Shin, H.S. (2013). Fed-Batch Cultures: Principles and Applications of Semi-Batch Bioreactors, Cambridge Series in Chemical Engineering, 1–457. Cambridge University Press. Drapeau, D., Luan, Y.T., Popoloski, J.A. et al. (1994). Extracellular insulin degrading activity creates instability in a cho-based batch-refeed continuous process. Cytotechnology 15 (1–3): 103–109. Landauer, K. (2014). Designing media for animal cell culture: CHO cells, the industrial standard. Methods Mol. Biol. 1104: 89–103. Jayme, D., Watanabe, T., and Shimada, T. (1997). Basal medium development for serum-free culture: a historical perspective. Cytotechnology 23 (1–3): 95–101. von Seefried, A. and MacMorine, H.G. (1976). The use of foetal, calf and adult bovine sera for the growth of serially subcultivated diploid cells. Dev. Biol. Stand. 37: 83–89.

References

26 Merten, O.W. (2002). Virus contaminations of cell cultures - a biotechnologi-

cal view. Cytotechnology 39 (2): 91–116. 27 Francis, G.L. (2010). Albumin and mammalian cell culture: implications for

biotechnology applications. Cytotechnology 62 (1): 1–16. 28 Heidemann, R., Zhang, C., Qi, H. et al. (2000). The use of peptones as

29 30

31

32

33

34 35

36

37

38

39

40

41

medium additives for the production of a recombinant therapeutic protein in high density perfusion cultures of mammalian cells. Cytotechnology 32 (2): 157–167. Barnes, D. and Sato, G. (1980). Serum-free cell culture: a unifying approach. Cell 22 (3): 649–655. Mitchelson, F.G., Mondia, J.P., and Hughes, E.H. (2016). Effect of copper variation in yeast hydrolysate on C-terminal lysine levels of a monoclonal antibody. Biotechnol. Progr. 33 (2): 463–468. Luo, Y. and Pierce, K.M. (2012). Development toward rapid and efficient screening for high performance hydrolysate lots in a recombinant monoclonal antibody manufacturing process. Biotechnol. Progr. 28 (4): 1061–1068. Siemensma, A., Babcock, J., Wilcox, C., and Huttinga, H. (2010). Towards an understanding of how protein hydrolysates stimulate more efficient biosynthesis in cultured cells. In: Protein Hydrolysates in Biotechnology (eds. V.K. Pasupuleti and A.L. Demain), 33–54. Netherlands, Dordrecht: Springer. Murhammer, D.W. and Goochee, C.F. (1990). Structural features of nonionic polyglycol polymer molecules responsible for the protective effect in sparged animal cell bioreactors. Biotechnol. Progr. 6 (2): 142–148. Gregory Hamilton, W. and Ham, R.G. (1977). Clonal growth of Chinese hamster cell lines in protein-free media. In Vitro 13 (9): 537–547. Xie, L. and Wang, D.I. (1996). High cell density and high monoclonal antibody production through medium design and rational control in a bioreactor. Biotechnol. Bioeng. 51 (6): 725–729. Huang, Y.M., Hu, W., Rustandi, E. et al. (2010). Maximizing productivity of CHO cell-based fed-batch culture using chemically defined media conditions and typical manufacturing equipment. Biotechnol. Progr. 26 (5): 1400–1410. Crowell, C.K., Grampp, G.E., Rogers, G.N. et al. (2007). Amino acid and manganese supplementation modulates the glycosylation state of erythropoietin in a CHO culture system. Biotechnol. Bioeng. 96 (3): 538–549. Pacis, E., Yu, M., Autsen, J. et al. (2011). Effects of cell culture conditions on antibody N-linked glycosylation—what affects high mannose 5 glycoform. Biotechnol. Bioeng. 108 (10): 2348–2358. Kaschak, T., Boyd, D., Lu, F. et al. (2011). Characterization of the basic charge variants of a human IgG1: effect of copper concentration in cell culture media. MAbs 3 (6): 577–583. Luo, J., Zhang, J., Ren, D. et al. (2012). Probing of C-terminal lysine variation in a recombinant monoclonal antibody production using Chinese hamster ovary cells with chemically defined media. Biotechnol. Bioeng. 109 (9): 2306–2315. Yuk, I.H., Russell, S., Tang, Y. et al. (2015). Effects of copper on CHO cells: cellular requirements and product quality considerations. Biotechnol. Progr. 31 (1): 226–238.

335

336

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

42 Vijayasankaran, N., Varma, S., Yang, Y. et al. (2013). Effect of cell culture

43

44

45

46

47

48

49

50

51

52

53

54

55 56

medium components on color of formulated monoclonal antibody drug substance. Biotechnol. Progr. 29 (5): 1270–1277. Reinhart, D., Damjanovic, L., Kaisermayer, C., and Kunert, R. (2015). Benchmarking of commercially available CHO cell culture media for antibody production. Appl. Microbiol. Biotechnol. 99 (11): 4645–4657. Pan, X., Streefland, M., Dalm, C. et al. (2017). Selection of chemically defined media for CHO cell fed-batch culture processes. Cytotechnology 69 (1): 39–56. Landauer, K., Woischnig, H., Hepp, N. et al. (2011). Development of a chemically defined CHO medium by engineering based on a feed solution. BMC Proceedings 5 (Suppl. 8): P41. Zhang, J. and Robinson, D. (2005). Development of animal-free, protein-free and chemically-defined media for NS0 cell culture. Cytotechnology 48 (1–3): 59–74. Templeton, N., Dean, J., Reddy, P., and Young, J.D. (2013). Peak antibody production is associated with increased oxidative metabolism in an industrially relevant fed-batch CHO cell culture. Biotechnol. Bioeng. 110 (7): 2013–2024. Li, J., Wong, C.L., Vijayasankaran, N. et al. (2012). Feeding lactate for CHO cell culture processes: impact on culture metabolism and performance. Biotechnol. Bioeng. 109 (5): 1173–1186. Yu, M., Hu, Z., Pacis, E. et al. (2011). Understanding the intracellular effect of enhanced nutrient feeding toward high titer antibody production process. Biotechnol. Bioeng. 108 (5): 1078–1088. Khattak, S.F., Xing, Z., Kenty, B. et al. (2010). Feed development for fed-batch CHO production process by semisteady state analysis. Biotechnol. Progr. 26 (3): 797–804. Lu, F., Toh, P.C., Burnett, I. et al. (2013). Automated dynamic fed-batch process and media optimization for high productivity cell culture process development. Biotechnol. Bioeng. 110 (1): 191–205. Frahm, B., Lane, P., Markl, H., and Portner, R. (2003). Improvement of a mammalian cell culture process by adaptive, model-based dialysis fed-batch cultivation and suppression of apoptosis. Bioprocess. Biosyst. Eng. 26 (1): 1–10. Kochanowski, N., Siriez, G., Roosens, S., and Malphettes, L. (2011). Medium and feed optimization for fed-batch production of a monoclonal antibody in CHO cells. BMC Proceedings 5 (Suppl. 8): P75. Kurtz, C. and Smelko, J.P. (2014). Which factors to consider when selecting an analytical method for cell culture fermentation: a comparison of four different metabolic analyzer instruments. Bioprocess. J. 13 (3): 12–31. Wlaschin, K.F. and Hu, W.S. (2006). Fedbatch culture and dynamic nutrient feeding. Adv. Biochem. Eng./Biotechnol. 101: 43–74. Yuk, I.H., Zhang, B., Yang, Y. et al. (2011). Controlling glycation of recombinant antibody in fed-batch cell cultures. Biotechnol. Bioeng. 108 (11): 2600–2610.

References

57 Gagnon, M., Hiller, G., Luan, Y.T. et al. (2011). High-end pH-controlled

58

59

60 61

62 63

64

65

66

67

68

69

70 71

72

73

delivery of glucose effectively suppresses lactate accumulation in CHO fed-batch cultures. Biotechnol. Bioeng. 108 (6): 1328–1337. Mulukutla, B.C., Yongky, A., Grimm, S. et al. (2015). Multiplicity of steady states in glycolysis and shift of metabolic state in cultured mammalian cells. PLoS One 10 (3): e0121561. Yongky, A., Lee, J., Le, T. et al. (2015). Mechanism for multiplicity of steady states with distinct cell concentration in continuous culture of mammalian cells. Biotechnol. Bioeng. 112 (7): 1437–1445. Schneider, M., Marison, I.W., and von Stockar, U. (1996). The importance of ammonia in mammalian cell culture. J. Biotechnol. 46 (3): 161–185. Gawlitzek, M., Ryll, T., Lofgren, J., and Sliwkowski, M.B. (2000). Ammonium alters N-glycan structures of recombinant TNFR-IgG: degradative versus biosynthetic mechanisms. Biotechnol. Bioeng. 68 (6): 637–646. Chen, P. and Harcum, S.W. (2006). Effects of elevated ammonium on glycosylation gene expression in CHO cells. Metab. Eng. 8 (2): 123–132. Kim, D.Y., Chaudhry, M.A., Kennard, M.L. et al. (2013). Fed-batch CHO cell t-PA production and feed glutamine replacement to reduce ammonia production. Biotechnol. Progr. 29 (1): 165–175. McCracken, N.A., Kowle, R., and Ouyang, A. (2014). Control of galactosylated glycoforms distribution in cell culture system. Biotechnol. Progr. 30 (3): 547–553. Fan, L., Zhao, L., Sun, Y. et al. (2009). A high-yielding, generic fed-batch process for recombinant antibody production of GS-engineered cell lines. J. Microbiol. Biotechnol. 19 (12): 1695–1702. Nasseri, S.S., Ghaffari, N., Braasch, K. et al. (2014). Increased CHO cell fed-batch monoclonal antibody production using the autophagy inhibitor 3-MA or gradually increasing osmolality. Biochem. Eng. J. 91: 37–45. Han, Y.K., Koo, T.Y., and Lee, G.M. (2009). Enhanced interferon-β production by CHO cells through elevated osmolality and reduced culture temperature. Biotechnol. Progr. 25 (5): 1440–1447. Kim, N.S. and Lee, G.M. (2002). Response of recombinant Chinese hamster ovary cells to hyperosmotic pressure: effect of Bcl-2 overexpression. J. Biotechnol. 95 (3): 237–248. Konno, Y., Kobayashi, Y., Takahashi, K. et al. (2012). Fucose content of monoclonal antibodies can be controlled by culture medium osmolality for high antibody-dependent cellular cytotoxicity. Cytotechnology 64 (3): 249–265. Li, F., Vijayasankaran, N., Shen, A.Y. et al. (2010). Cell culture processes for monoclonal antibody production. MAbs 2 (5): 466–479. Matthews, C.B., Wright, C., Kuo, A. et al. (2017). Reexamining opportunities for therapeutic protein production in eukaryotic microorganisms. Biotechnol. Bioeng. 114 (11): 2432–2444. Carrillo-Cocom, L.M., Genel-Rey, T., Araiz-Hernandez, D. et al. (2015). Amino acid consumption in naive and recombinant CHO cell cultures: producers of a monoclonal antibody. Cytotechnology 67 (5): 809–820. Blake, C.J. (2007). Analytical procedures for water-soluble vitamins in foods and dietary supplements: a review. Anal. Bioanal. Chem. 389 (1): 63–76.

337

338

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

74 Cerchiaro, G., Manieri, T.M., and Bertuchi, F.R. (2013). Analytical methods

75

76

77

78

for copper, zinc and iron quantification in mammalian cells. Metallomics 5 (10): 1336–1345. Bradley, S.A., Ouyang, A., Purdie, J. et al. (2010). Fermentanomics: monitoring mammalian cell cultures with NMR spectroscopy. J. Am. Chem. Soc. 132 (28): 9531–9533. Hossler, P., McDermott, S., Racicot, C., and Fann, J.C. (2013). Improvement of mammalian cell culture performance through surfactant enabled concentrated feed media. Biotechnol. Progr. 29 (4): 1023–1033. McCoy, R.E., Costa, N.A., and Morris, A.E. (2015). Factors that determine stability of highly concentrated chemically defined production media. Biotechnol. Progr. 31 (2): 493–502. Paul, W., Davis, R., Campbell, A. et al. (2013). Optimized platform medium and feed for rCHO cell lines using the CHEF1 expression system. BMC Proceedings 7 (6): P98. Kishishita, S., Katayama, S., Kodaira, K. et al. (2015). Optimization of chemically defined feed media for monoclonal antibody production in Chinese hamster ovary cells. J. Biosci. Bioeng. 120 (1): 78–84. Hecklau, C., Pering, S., Seibel, R. et al. (2016). S-Sulfocysteine simplifies fed-batch processes and increases the CHO specific productivity via anti-oxidant activity. J. Biotechnol. 218: 53–63. Zimmer, A., Mueller, R., Wehsling, M. et al. (2014). Improvement and simplification of fed-batch bioprocesses with a highly soluble phosphotyrosine sodium salt. J. Biotechnol. 186: 110–118. Kumar, N., Gammell, P., and Clynes, M. (2007). Proliferation control strategies to improve productivity and survival during CHO based production culture: a summary of recent methods employed and the effects of proliferation control in product secreting CHO cell lines. Cytotechnology 53 (1–3): 33–46. Oguchi, S., Saito, H., Tsukahara, M., and Tsumura, H. (2003). Control of temperature and pH enhances human monoclonal antibody production in CHO cell culture. In: Animal Cell Technology: Basic & Applied Aspects: Proceedings of the 15th Annual Meeting of the Japanese Association for Animal Cell Technology (JAACT), Fuchu, Japan (11–15 November 2002) (eds. K. Yagasaki, Y. Miura, M. Hatori and Y. Nomura), 169–172. Netherlands, Dordrecht: Springer. Kishishita, S., Nishikawa, T., Shinoda, Y. et al. (2015). Effect of temperature shift on levels of acidic charge variants in IgG monoclonal antibodies in Chinese hamster ovary cell culture. J. Biosci. Bioeng. 119 (6): 700–705. Yoon, S.K., Choi, S.L., Song, J.Y., and Lee, G.M. (2005). Effect of culture pH on erythropoietin production by Chinese hamster ovary cells grown in suspension at 32.5 and 37.0 degrees C. Biotechnol. Bioeng. 89 (3): 345–356. Clark, K.J., Chaplin, F.W., and Harcum, S.W. (2004). Temperature effects on product-quality-related enzymes in batch CHO cell cultures producing recombinant tPA. Biotechnol. Progr. 20 (6): 1888–1892. Trummer, E., Fauland, K., Seidinger, S. et al. (2006). Process parameter shifting: Part II. Biphasic cultivation-A tool for enhancing the volumetric

®

79

80

81

82

83

84

85

86

87

References

88

89

90

91 92

93

94

95

96

97

98

99

100

101

productivity of batch processes using Epo-Fc expressing CHO cells. Biotechnol. Bioeng. 94 (6): 1045–1052. Jing, Y., Borys, M., Nayak, S. et al. (2012). Identification of cell culture conditions to control protein aggregation of IgG fusion proteins expressed in Chinese hamster ovary cells. Process Biochem. 47 (1): 69–75. Brunner, M., Fricke, J., Kroll, P., and Herwig, C. (2017). Investigation of the interactions of critical scale-up parameters (pH, pO2 and pCO2 ) on CHO batch performance and critical quality attributes. Bioprocess. Biosyst. Eng. 40 (2): 251–263. Borys, M.C., Dalal, N.G., Abu-Absi, N.R. et al. (2010). Effects of culture conditions on N-glycolylneuraminic acid (Neu5Gc) content of a recombinant fusion protein produced in CHO cells. Biotechnol. Bioeng. 105 (6): 1048–1057. Kimura, R. and Miller, W.M. (1997). Glycosylation of CHO-derived recombinant tPA produced under elevated pCO2 . Biotechnol. Progr. 13 (3): 311–317. Tsang, V.L., Wang, A.X., Yusuf-Makagiansar, H., and Ryll, T. (2014). Development of a scale down cell culture model using multivariate analysis as a qualification tool. Biotechnol. Progr. 30 (1): 152–160. Tescione, L., Lambropoulos, J., Paranandi, M.R. et al. (2015). Application of bioreactor design principles and multivariate analysis for development of cell culture scale down models. Biotechnol. Bioeng. 112 (1): 84–97. Li, F., Hashimura, Y., Pendleton, R. et al. (2006). A systematic approach for scale-down model development and characterization of commercial cell culture processes. Biotechnol. Progr. 22 (3): 696–703. Ahuja, S., Jain, S., and Ram, K. (2015). Application of multivariate analysis and mass transfer principles for refinement of a 3-L bioreactor scale-down model--when shake flasks mimic 15,000-L bioreactors better. Biotechnol. Progr. 31 (5): 1370–1380. Kaisermayer, C., Reinhart, D., Gili, A. et al. (2016). Biphasic cultivation strategy to avoid Epo-Fc aggregation and optimize protein expression. J. Biotechnol. 227: 3–9. Lewis, A.M., Croughan, W.D., Aranibar, N. et al. (2016). Understanding and Controlling Sialylation in a CHO Fc-Fusion Process. PLoS One 11 (6): e0157111. Miller, W.M., Wilke, C.R., and Blanch, H.W. (1987). Effects of dissolved oxygen concentration on hybridoma growth and metabolism in continuous culture. J. Cell. Physiol. 132 (3): 524–530. Baltz, R.H., Demain, A.L., Davies, J.E. et al. (eds.) (2010). Manual of Industrial Microbiology and Biotechnology, 3e. Washington, DC: ASM Press http:// www.asmscience.org/content/book/10.1128/9781555816827. Hu, W., Berdugo, C., and Chalmers, J.J. (2011). The potential of hydrodynamic damage to animal cells of industrial relevance: current understanding. Cytotechnology 63 (5): 445–460. Gramer, M.J., Goochee, C.F., Chock, V.Y. et al. (1995). Removal of sialic acid from a glycoprotein in CHO cell culture supernatant by action of an extracellular CHO cell sialidase. Biotechnology (NY) 13 (7): 692–698.

339

340

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

102 Koterba, K.L., Borgschulte, T., and Laird, M.W. (2012). Thioredoxin 1 is

103

104

105

106

107

108

109

110

111

112

113

114

115

responsible for antibody disulfide reduction in CHO cell culture. J. Biotechnol. 157 (1): 261–267. Robert, F., Bierau, H., Rossi, M. et al. (2009). Degradation of an Fc-fusion recombinant protein by host cell proteases: Identification of a CHO cathepsin D protease. Biotechnol. Bioeng. 104 (6): 1132–1141. Sung, Y.H., Song, Y.J., Lim, S.W. et al. (2004). Effect of sodium butyrate on the production, heterogeneity and biological activity of human thrombopoietin by recombinant Chinese hamster ovary cells. J. Biotechnol. 112 (3): 323–335. Yang, W.C., Lu, J., Nguyen, N.B. et al. (2014). Addition of valproic acid to CHO cell fed-batch cultures improves monoclonal antibody titers. Mol. Biotechnol. 56 (5): 421–428. Park, J.H., Noh, S.M., Woo, J.R. et al. (2016). Valeric acid induces cell cycle arrest at G1 phase in CHO cell cultures and improves recombinant antibody productivity. Biotechnol. J. 11 (4): 487–496. Liu, C., Chu, I., and Hwang, S. (2001). Pentanoic acid, a novel protein synthesis stimulant for Chinese hamster ovary (CHO) cells. J. Biosci. Bioeng. 91 (1): 71–75. Wong, N.S., Wati, L., Nissom, P.M. et al. (2010). An investigation of intracellular glycosylation activities in CHO cells: effects of nucleotide sugar precursor feeding. Biotechnol. Bioeng. 107 (2): 321–336. Baker, K.N., Rendall, M.H., Hills, A.E. et al. (2001). Metabolic control of recombinant protein N-glycan processing in NS0 and CHO cells. Biotechnol. Bioeng. 73 (3): 188–202. Gawlitzek, M., Valley, U., and Wagner, R. (1998). Ammonium ion and glucosamine dependent increases of oligosaccharide complexity in recombinant glycoproteins secreted from cultivated BHK-21 cells. Biotechnol. Bioeng. 57 (5): 518–528. Gramer, M.J., Eckblad, J.J., Donahue, R. et al. (2011). Modulation of antibody galactosylation through feeding of uridine, manganese chloride, and galactose. Biotechnol. Bioeng. 108 (7): 1591–1602. Gu, X. and Wang, D.I. (1998). Improvement of interferon-gamma sialylation in Chinese hamster ovary cell culture by feeding of N-acetylmannosamine. Biotechnol. Bioeng. 58 (6): 642–648. Zhang, A., Tsang, V.L., Markely, L.R. et al. (2016). Identifying the differences in mechanisms of mycophenolic acid controlling fucose content of glycoproteins expressed in different CHO cell lines. Biotechnol. Bioeng. 113 (11): 2367–2376. Allen, J.G., Mujacic, M., Frohn, M.J. et al. (2016). Facile modulation of antibody fucosylation with small molecule fucostatin inhibitors and cocrystal structure with GDP-mannose 4,6-dehydratase. ACS Chem. Biol. 11 (10): 2734–2743. Brühlmann, D., Muhr, A., Parker, R. et al. (2017). Cell culture media supplemented with raffinose reproducibly enhances high mannose glycan formation. J. Biotechnol. 252: 32–42.

References

116 Hossler, P., McDermott, S., Racicot, C. et al. (2014). Cell culture media

117

118

119

120

121

122

123

124

125

126

127

128

129

supplementation of uncommonly used sugars sucrose and tagatose for the targeted shifting of protein glycosylation profiles of recombinant protein therapeutics. Biotechnol. Progr. 30 (6): 1419–1431. Hossler, P., Racicot, C., Chumsae, C. et al. (2017). Cell culture media supplementation of infrequently used sugars for the targeted shifting of protein glycosylation profiles. Biotechnol. Progr. 33 (2): 511–522. Huang, C.J., Lin, H., and Yang, J.X. (2015). A robust method for increasing Fc glycan high mannose level of recombinant antibodies. Biotechnol. Bioeng. 112 (6): 1200–1209. Hossler, P., Chumsae, C., Racicot, C. et al. (2017). Arabinosylation of recombinant human immunoglobulin-based protein therapeutics. mAbs 9 (4): 715–734. Mulukutla, B.C., Kale, J., Kalomeris, T. et al. (2017). Identification and control of novel growth inhibitors in fed-batch cultures of Chinese hamster ovary cells. Biotechnol. Bioeng. 114 (8): 1779–1790. Abu-Absi, N.R., Kenty, B.M., Cuellar, M.E. et al. (2011). Real time monitoring of multiple parameters in mammalian cell culture bioreactors using an in-line Raman spectroscopy probe. Biotechnol. Bioeng. 108 (5): 1215–1221. Berry, B., Moretto, J., Matthews, T. et al. (2015). Cross-scale predictive modeling of CHO cell culture growth and metabolites using Raman spectroscopy and multivariate analysis. Biotechnol. Progr. 31 (2): 566–577. Li, B., Ray, B.H., Leister, K.J., and Ryder, A.G. (2013). Performance monitoring of a mammalian cell based bioprocess using Raman spectroscopy. Anal. Chim. Acta 796: 84–91. Hakemeyer, C., Strauss, U., Werz, S. et al. (2012). At-line NIR spectroscopy as effective PAT monitoring technique in Mab cultivations during process development and manufacturing. Talanta 90: 12–21. Clavaud, M., Roggo, Y., Von Daeniken, R. et al. (2013). Chemometrics and in-line near infrared spectroscopic monitoring of a biopharmaceutical Chinese hamster ovary cell culture: prediction of multiple cultivation variables. Talanta 111: 28–38. Matthews, T.E., Berry, B.N., Smelko, J. et al. (2016). Closed loop control of lactate concentration in mammalian cell culture by Raman spectroscopy leads to improved cell density, viability, and biopharmaceutical protein production. Biotechnol. Bioeng. 113 (11): 2416–2424. Berry, B.N., Dobrowsky, T.M., Timson, R.C. et al. (2016). Quick generation of Raman spectroscopy based in-process glucose control to influence biopharmaceutical protein product quality during mammalian cell culture. Biotechnol. Progr. 32 (1): 224–234. Kirdar, A.O., Chen, G., Weidner, J., and Rathore, A.S. (2010). Application of near-infrared (NIR) spectroscopy for screening of raw materials used in the cell culture medium for the production of a recombinant therapeutic protein. Biotechnol. Progr. 26 (2): 527–531. Trunfio, N., Lee, H., Starkey, J. et al. (2017). Characterization of mammalian cell culture raw materials by combining spectroscopy and chemometrics. Biotechnol. Progr. 33 (4): 1127–1138.

341

342

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

130 Opel, C.F., Li, J., and Amanullah, A. (2010). Quantitative modeling of viable

131

132

133

134

135

cell density, cell size, intracellular conductivity, and membrane capacitance in batch and fed-batch CHO processes using dielectric spectroscopy. Biotechnol. Progr. 26 (4): 1187–1199. Cannizzaro, C., Gugerli, R., Marison, I., and von Stockar, U. (2003). On-line biomass monitoring of CHO perfusion culture with scanning dielectric spectroscopy. Biotechnol. Bioeng. 84 (5): 597–610. Dowd, J.E., Jubb, A., Kwok, K.E., and Piret, J.M. (2003). Optimization and control of perfusion cultures using a viable cell probe and cell specific perfusion rates. Cytotechnology 42 (1): 35–45. Zhang, A., Tsang, V.L., Moore, B. et al. (2015). Advanced process monitoring and feedback control to enhance cell culture process production and robustness. Biotechnol. Bioeng. 112 (12): 2495–2504. Bareither, R. and Pollard, D. (2011). A review of advanced small-scale parallel bioreactor technology for accelerated process development: current state and future need. Biotechnol. Progr. 27 (1): 2–14. Moses, S., Manahan, M., Ambrogelly, A., and Ling, W.L.W. (2012). Assessment of AMBR as a model for high-throughput cell culture process development strategy. Adv. Biosci. Biotechnol. 3 (07): 918–927. Betts, J.P.J., SRC, W., Finka, G.B. et al. (2014). Impact of aeration strategies on fed-batch cell culture kinetics in a single-use 24-well miniature bioreactor. Biochem. Eng. J. 82: 105–116. Hsu, W.-T., Aulakh, R.P.S., Traul, D.L., and Yuk, I.H. (2012). Advanced microscale bioreactor system: a representative scale-down model for bench-top bioreactors. Cytotechnology 64 (6): 667–678. Rameez, S., Mostafa, S.S., Miller, C., and Shukla, A.A. (2014). Highthroughput miniaturized bioreactors for cell culture process development: reproducibility, scalability, and control. Biotechnol. Progr. 30 (3): 718–727. Chen, A., Chitta, R., Chang, D., and Amanullah, A. (2009). Twenty-four well plate miniature bioreactor system as a scale-down model for cell culture process development. Biotechnol. Bioeng. 102 (1): 148–160. Xu, P., Clark, C., Ryder, T. et al. (2017). Characterization of TAP Ambr 250 disposable bioreactors, as a reliable scale-down model for biologics process development. Biotechnol. Progr. 33 (2): 478–489. Janakiraman, V., Kwiatkowski, C., Kshirsagar, R. et al. (2015). Application of high-throughput mini-bioreactor system for systematic scale-down modeling, process characterization, and control strategy development. Biotechnol. Progr. 31 (6): 1623–1632. Siva, S., Zhang, A., Koepf, E. et al. (2016). Leveraging high-throughput technology to accelerate the time to clinic: a case study of a mAb. Eng. Life Sci. 16 (2): 143–151. Paul Smelko, J., Rae Wiltberger, K., Francis Hickman, E. et al. (2011). Performance of high intensity fed-batch mammalian cell cultures in disposable bioreactor systems. Biotechnol. Progr. 27 (5): 1358–1364. Jacquemart, R., Vandersluis, M., Zhao, M. et al. (2016). A single-use strategy to enable manufacturing of affordable biologics. Comput. Struct. Biotechnol. J. 14: 309–318. TM

136

137

138

139

140

141

142

143

144

References

145 Lopes, A.G. (2015). Single-use in the biopharmaceutical industry: a review of

146

147

148

149

150

151

152

153

154

155

156

157

158

159

current technology impact, challenges and limitations. Food Bioprod. Process. 93: 98–114. Hammond, M., Marghitoiu, L., Lee, H. et al. (2014). A cytotoxic leachable compound from single-use bioprocess equipment that causes poor cell growth performance. Biotechnol. Progr. 30 (2): 332–337. Wood, J., Mahajan, E., and Shiratori, M. (2013). Strategy for selecting disposable bags for cell culture media applications based on a root-cause investigation. Biotechnol. Progr. 29 (6): 1535–1549. Klutz, S., Magnus, J., Lobedann, M. et al. (2015). Developing the biofacility of the future based on continuous processing and single-use technology. J. Biotechnol. 213: 120–130. Walther, J., Godawat, R., Hwang, C. et al. (2015). The business impact of an integrated continuous biomanufacturing platform for recombinant protein production. J. Biotechnol. 213: 3–12. Yang, W.C., Lu, J., Kwiatkowski, C. et al. (2014). Perfusion seed cultures improve biopharmaceutical fed-batch production capacity and product quality. Biotechnol. Progr. 30 (3): 616–625. Pohlscheidt, M., Jacobs, M., Wolf, S. et al. (2013). Optimizing capacity utilization by large scale 3000 L perfusion in seed train bioreactors. Biotechnol. Progr. 29 (1): 222–229. Padawer, I., Ling, W.L.W., and Bai, Y. (2013). Case study: an accelerated 8-day monoclonal antibody production process based on high seeding densities. Biotechnol. Progr. 29 (3): 829–832. Kuczewski, M., Schirmer, E., Lain, B., and Zarbis-Papastoitsis, G. (2011). A single-use purification process for the production of a monoclonal antibody produced in a PER.C6 human cell line. Biotechnol. J. 6 (1): 56–65. Yang, W.C., Minkler, D.F., Kshirsagar, R. et al. (2016). Concentrated fed-batch cell culture increases manufacturing capacity without additional volumetric capacity. J. Biotechnol. 217: 1–11. Xu, S., Hoshan, L., and Chen, H. (2016). Improving lactate metabolism in an intensified CHO culture process: productivity and product quality considerations. Bioprocess. Biosyst. Eng. 39 (11): 1689–1702. Hiller, G.W., Ovalle, A.M., Gagnon, M.P. et al. (2017). Cell-controlled hybrid perfusion fed-batch CHO cell process provides significant productivity improvement over conventional fed-batch cultures. Biotechnol. Bioeng. 114 (7): 1438–1447. Kang, Y., Hamzik, J., Felo, M. et al. (2013). Development of a novel and efficient cell culture flocculation process using a stimulus responsive polymer to streamline antibody purification processes. Biotechnol. Bioeng. 110 (11): 2928–2937. McNerney, T., Thomas, A., Senczuk, A. et al. (2015). PDADMAC flocculation of Chinese hamster ovary cells: enabling a centrifuge-less harvest process for monoclonal antibodies. MAbs 7 (2): 413–428. Wurm, F.M. and de Jesus, M. (2016). Manufacture of recombinant therapeutic proteins using Chinese hamster ovary cells in large-scale bioreactors. In: Biosimilars of Monoclonal Antibodies, 327–353. Hoboken, NJ: Wiley.

343

344

13 Mammalian Fed-batch Cell Culture for Biopharmaceuticals

160 Sauer, P.W., Burky, J.E., Wesson, M.C. et al. (2000). A high-yielding, generic

161

162 163

164 165 166

167

168

169

170

171 172 173

174

175

fed-batch cell culture process for production of recombinant antibodies. Biotechnol. Bioeng. 67 (5): 585–597. Brown, M.E., Renner, G., Field, R.P., and Hassell, T. (1992). Process development for the production of recombinant antibodies using the glutamine synthetase (GS) system. Cytotechnology 9 (1): 231–236. Broad, D., Boraston, R., and Rhodes, M. (1991). Production of recombinant proteins in serum-free media. Cytotechnology 5 (1): 47–55. Oh, S.K.W., Vig, P., Chua, F. et al. (1993). Substantial overproduction of antibodies by applying osmotic pressure and sodium butyrate. Biotechnol. Bioeng. 42 (5): 601–610. Bibila, T.A. and Robinson, D.K. (1995). In pursuit of the optimal fed-batch process for monoclonal antibody production. Biotechnol. Progr. 11 (1): 1–13. Birch, J.R. and Froud, S.J. (1994). Mammalian cell culture systems for recombinant protein production. Biologicals 22 (2): 127–133. Zhou, W., Chen, C.C., Buckland, B., and Aunins, J. (1997). Fed-batch culture of recombinant NS0 myeloma cells with high monoclonal antibody production. Biotechnol. Bioeng. 55 (5): 783–792. deZengotita, V.M., Miller, W.M., Aunins, J.G., and Zhou, W. (2000). Phosphate feeding improves high-cell-concentration NS0 myeloma culture performance for monoclonal antibody production. Biotechnol. Bioeng. 69 (5): 566–576. Butler, M. (2005). Animal cell cultures: recent achievements and perspectives in the production of biopharmaceuticals. Appl. Microbiol. Biotechnol. 68 (3): 283–291. Zhang, J., Robinson, D., and Salmon, P. (2006). A novel function for selenium in biological system: Selenite as a highly effective iron carrier for Chinese hamster ovary cell growth and monoclonal antibody production. Biotechnol. Bioeng. 95 (6): 1188–1197. Farid, S.S. (2006). Established bioprocesses for producing antibodies as a basis for future planning. In: Cell Culture Engineering (ed. W.-S. Hu), 1–42. Berlin, Heidelberg: Springer-Verlag. Tchoudakova, A., Hensel, F., Murillo, A. et al. (2009). High level expression of functional human IgMs in human PER.C6 cells. MAbs 1 (2): 163–171. Kelley, B. (2009). Industrialization of mAb production technology The bioprocessing industry at a crossroads. mAbs 1 (5): 443–452. Zhou, M., Crawford, Y., Ng, D. et al. (2011). Decreasing lactate level and increasing antibody production in Chinese hamster ovary cells (CHO) by reducing the expression of lactate dehydrogenase and pyruvate dehydrogenase kinases. J. Biotechnol. 153 (1): 27–34. Kshirsagar, R., McElearney, K., Gilbert, A. et al. (2012). Controlling trisulfide modification in recombinant monoclonal antibody produced in fed-batch cell culture. Biotechnol. Bioeng. 109 (10): 2523–2532. Kunert, R. and Reinhart, D. (2016). Advances in recombinant antibody manufacturing. Appl. Microbiol. Biotechnol. 100 (8): 3451–3461.

References

176 Takagi, Y., Kikuchi, T., Wada, R., and Omasa, T. (2017). The enhancement of

antibody concentration and achievement of high cell density CHO cell cultivation by adding nucleoside. Cytotechnology 69 (3): 511–521. 177 Spadiut, O., Capone, S., Krainer, F. et al. (2014). Microbials for the production of monoclonal antibodies and antibody fragments. Trends Biotechnol. 32 (1): 54–60. 178 Vogl, T., Hartner, F.S., and Glieder, A. (2013). New opportunities by synthetic biology for biopharmaceutical production in Pichia pastoris. Curr. Opin. Biotechnol. 24 (6): 1094–1101.

345

347

14 Continuous Biomanufacturing Sadettin S. Ozturk MassBiologics, Process and Analytical Development, 460 Walk Hill Street, Mattapan, MA 02126, USA

14.1 Introduction Over the past four decades, the biomanufacturing industry has been monumental in generating drugs for the treatment of life-threatening diseases, including cancer, autoimmune diseases, and metabolic and blood disorders. Therapeutic proteins, such as interferons, blood clotting factors, enzymes, vaccines, and monoclonal antibodies are produced by genetically engineered cells (microbial, yeast, and mammalian) cultivated in large-scale bioreactors [1]. These proteins are then isolated, purified, formulated, and filled as biopharmaceutical drugs under current good manufacturing practice (cGMP) conditions using validated biomanufacturing processes [2]. Each unit operation in biomanufacturing is operated either as batch or continuous, although a hybrid mode is also possible. The choice between batch and continuous biomanufacturing depends on the specific product, organization, scale, process economics, and many other factors. Although a batch process is well-established and easier to operate and control, it is not very efficient, requires frequent downtime, and poses product quality issues, especially for labile and unstable molecules [3]. Increasing pressure on process economics, less cost of goods (COGs), and higher demands on the drugs necessitate process intensification and higher process yields [4, 5]. Although it is more complicated to design and operate, continuous biomanufacturing is preferred over batch biomanufacturing under these conditions. Continuous bioprocessing can be applied to all unit operations or can be applied to only certain upstream or downstream processes [6]. In this chapter, we will focus on continuous upstream processes, particularly on continuous perfusion.

14.2 Continuous Upstream (Cell Culture) Processes A continuous cell culture process involves running a bioreactor where the product is continuously removed in the harvest stream while a fresh medium is added to feed the cells [6]. The feed and harvest flow rates are kept the same (if there are no other additives) so that the volume in the bioreactor is maintained. Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

348

14 Continuous Biomanufacturing

Continuous Chemostat

Continuous with cell retention Perfusion

Waste products

Harvest refeed Nutrients

Waste products Nutrients

Nutrients Time

Waste products

Time

Time

Medium

Medium Cell-containing harvest

Harvest continuously

Medium Harvest w/ or w/o cells

“Cell-free” harvest

Harvest continuously

Harvest at intervals

Figure 14.1 Continuous cell culture processes used for the production of biopharmaceuticals.

Continuous cell culture processes can run for a long time (typically >2 months) and are operated under “steady-state” conditions. The harvest (or medium) flow rate, when normalized to the bioreactor volume, is called the “medium exchange rate.” Continuous processes can run up to 10 volume exchange rates per day, although low exchange rates are preferred to minimize medium costs and increase product titer [6, 7]. There are three types of continuous cell culture processes: (i) continuous culture without cell retention, (ii) continuous culture with cell retention (perfusion), and (iii) semicontinuous culture that can be operated with or without cell retention [8]. Figure 14.1 provides some examples of these systems and their operations are discussed in the following sections. 14.2.1

Continuous Culture without Cell Retention (Chemostat)

These systems are commonly referred to as chemostat cultures and are run in a true steady-state mode. In the absence of a cell retention device, cells are removed continually from the bioreactor at a cell density equal to cell concentration in the bioreactor. As cell growth in the bioreactor must keep up with cell removal in the harvest, chemostat cultures cannot be operated at high flow rates and at high cell densities because of the limitations in how fast mammalian cells grow (doubling time around 24 hours). Despite these limitations, chemostat cultures enable continuous biomanufacturing using conventional bioreactors and are easy to operate and control. 14.2.2

Continuous Culture with Cell Retention (Perfusion)

Retention of cells in the bioreactor by physical means allows for medium exchange at faster rates, thereby achieving higher cell densities and productivities [7, 8]. The degree of cell retention in the bioreactor depends on the method of cell retention and varies between 10% and 100%. Continuous culture with

14.2 Continuous Upstream (Cell Culture) Processes

(a)

(b)

Figure 14.2 Examples of perfusion bioreactors with cell retention: (a) iCellis bioreactor with cell entrapment and (b) stirred-tank bioreactor with external ATF (alternating tangential flow) cell retention device.

cell retention is commonly referred to as perfusion, where the cells are perfused by exchanging the medium, and it is the most preferred format of continuous culture employed by the industry [9]. Cell retention in the bioreactor can be achieved by two ways: (i) separating cells by immobilization or entrapment and (ii) using a cell retention device. Figure 14.2 illustrates two types of perfusion bioreactors, one with cell entrapment and one with a cell retention device. 14.2.2.1

Cell Retention by Immobilization or Entrapment

Several methods have been employed to retain cells without incorporating a special cell retention device. Cells can be immobilized or encapsulated in alginate beads, entrapped in a hollow fiber bioreactor between the fibers, or trapped or attached to a solid support (woven disks, fibers, and porous or nonporous microcarriers) [10]. Cells in or on beads, microcarriers, and fibers receive nutrients from the medium that is exchanged (perfused) continually. In most cases, the product secreted from the cells is recovered from the harvest stream, although in the case of the hollow fiber bioreactor, it can be collected from the cell layer. Packed bed, fluidized bed, and hollow fiber bioreactors achieve cell retention by immobilization and entrapment of the cells, thus preventing them from leaving the bioreactor in the harvest. Special bioreactor configurations must be employed for heat and mass transfer, for monitoring and control of process parameters (temperature, pH, and dissolved oxygen), and for aeration of cultures. In addition to a feed and harvest flow, a circulation loop is often incorporated to the design for placing probes and heat and gas exchange units. In the gas exchanger unit, the circulating medium is saturated by air or oxygen, and CO2 can be added

349

350

14 Continuous Biomanufacturing

to or removed from the culture for pH control. Beads or microcarriers can be used in stirred-tank bioreactors as well. Under these conditions, the bioreactor can be sparged to supply oxygen and remove CO2 . The immobilization or entrapment systems are heterogeneous in nature as they offer no mixing in the cell layer. In the absence of mixing, cells are not uniformly distributed in the bioreactor or in the medium and the cellular environment cannot be controlled. The nutrient delivery and product removal from the cells are accomplished through diffusion only. Diffusion is not an efficient form of mass transfer and therefore creates gradients in cell density, nutrient and product concentrations, dissolved oxygen, and pH. Because of nutrient starvation or oxygen limitations, a necrotic core or zone is established where cells are either not healthy or dead. This situation impacts the overall yield, process scalability, and the product quality from these systems. Even though these systems are not homogeneous or easily scalable, they can be very effective in the production of biopharmaceuticals at a small (a few grams) to medium (100 g) scale. Tissue-like cell densities (about 500 million cells/ml) can be reached locally and the systems can be operated for a long time (up totwo to three months). 14.2.2.2

Cell Retention by Cell Retention Device

A cell retention device can be employed into a conventional stirred-tank bioreactor to allow continuous perfusion for suspension cells [6]. Although a cell retention device adds complexities for the operation, it allows continuous cultures to be run in conventional bioreactors (stirred tank) where mixing and aeration are achieved very effectively. In these bioreactors, a homogeneous environment is provided to the cells and the cellular environment is well controlled. Not only are scalability and process consistency well established for stirred-tank bioreactors but these bioreactors are also easy to operate and control. Cell retention devices use several mechanisms to concentrate the cells and separate them from the harvest steam (see Section 14.3.1). Depending on the mechanism and the design of the device, the harvest stream can be cell free or can have only a fraction of the cells in the bioreactor. Cell retention devices can be used as an external or internal device [7]. The internal cell retention device enables simpler bioreactor configuration and maintains the optimal conditions (pH, temperature, dissolved oxygen, etc.) for the cells during cell separation and concentration. As they are part of the bioreactor, however, they cannot be replaced if they fail. External cell retention devices allow replacement of the unit in case of failure, but they require taking the cells outside of the bioreactor where cellular environment cannot be maintained at optimal levels. The external cell retention devices often require a circulation loop where the cells are removed from one port on the bioreactor and concentrated in the device continuously. Concentrated cells that return to the bioreactor from another port and harvest with reduced cell concentration are collected. The circulation loop requires a pump to move cells around, which can be used either to pump cells to the cell retention device or to pump the concentrated cells back to the bioreactor. Some cell retention devices require only one port, which simplifies the operation.

14.3 Advantages of Continuous Perfusion

Cell retention devices will be discussed in detail in Section 14.4 where the focus will be devoted to perfusion bioreactors. 14.2.3

Semicontinuous Culture

Semicontinuous cultures can run for months continuously; however, the operation consists of repeated batch cycles that can last three to four days (Figure 14.1). At the end of each cycle, a fraction of culture, typically about 80%, is removed as harvest, and fresh medium is added so that a new cycle can be initiated. Semicontinuous culture can be used with (cells attached/trapped to beads or microcarriers) or without (suspension cells) cell retention and often employ stirred-tank bioreactors.

14.3 Advantages of Continuous Perfusion Although any of the continuous cell culture processes outlined above can be used for the production of biopharmaceuticals, continuous culture with cell retention, or perfusion process, is the most popular method in industry. Continuous perfusion yields higher volumetric productivities, allows better utilization of biomanufacturing facilities, and results in better product quality and consistency. In addition, the continuous processes enable scaling out (increase scale by number of units) as opposed to scaling up by volume, and meet capacity requirements from a more compact facility. 14.3.1

Higher Volumetric Productivities

One of the main advantages of continuous perfusion is achieving higher cell densities and volumetric productivities as compared to fed-batch and other types of continuous cultures (chemostat and semicontinuous) [6, 9]. High media exchange rate allows reaching high cell densities and high viabilities while retaining the cells in the bioreactor. Figure 14.3 presents data from a stirred-tank bioreactor coupled with an alternating tangential flow (ATF) cell retention device (Repligen, MA). After a cell build-up phase of 12 days, the bioreactor cell density reached about 60 million cells/ml and maintained at this level by a cell density control algorithm. The total cell density increased to about 85 million cells/ml during the course of the run that lasted 45 days. The titers in the harvest varied between 1 and 1.2 g/l. Although this titer is lower than what could be obtained in a fed-batch bioreactor (6 g/l in this case), it should be noted that the material from the continuous perfusion is collected continuously without waiting for 14 days (typical fed-batch run time). In terms of volumetric productivities, perfusion culture provides 1.8 g/l-day as opposed to 0.43 g/l-day from fed-batch. Accordingly, the same size of bioreactor can produce about four times more product when it is operated in continuous perfusion, or for the same amount of material to be produced, continuous perfusion will require a bioreactor that is four times smaller [6].

351

Million cells/ml

14 Continuous Biomanufacturing

100 90 80 70 60 50 40 30 20 10 0

Viable cells Total cells 0

5

10

15

20

(a)

25 Days

30

35

40

45

50

45

50

1600 1400 1200 Titer (mg/l)

352

1000 800 600

Titer, mg/l

400 200 0 0 (b)

5

10

15

20

25 Days

30

35

40

Figure 14.3 Performance of a continuous perfusion bioreactor operated at 1.5 volume per day exchange rate with ATF as the perfusion device. (a) Cell densities and (b) titers.

14.3.2

Better Utilization of Biomanufacturing Facilities

Biopharmaceutical medicines are produced in cGMP facilities that are very expensive to be built, validated, and maintained. Facility cost is a major factor determining COGs, and the facility utilization needs to be maximized for cost-effective production [4, 11, 12]. The fed-batch process is run in stages, and each stage needs to wait for the completion of the previous stage. These stages or holding times reduce the facility utilization significantly. Continuous perfusion, on the other hand, allows maximum facility utilization. 14.3.3

Better Product Quality and Consistency

Although most biopharmaceuticals can be produced using batch1 and continuous processes, the mode of operation impacts the product quality and product profile. The mode of operation creates differences in (i) the residence time of the product in the bioreactor; (ii) concentration profiles for nutrients, metabolic 1 We will use batch as a general term to refer to both straight batch (no feed) and fed batch operation.

14.3 Advantages of Continuous Perfusion

by-products, and product; (iii) levels of product and process-related impurities, and (iv) cell physiology, growth, metabolic, and production rates. When the product is not stable, the residence time in the bioreactor impacts product quality significantly. Therapeutic enzymes or blood coagulation factors such as rFVIII degrade quickly in the culture because of proteases and their product profile can also change because of sialidases and other enzymes accumulated in the culture [13]. Continuous perfusion minimizes these issues because of the shorter residence times it offers, making it the method of choice over batch culture. The mode of operation also influences the concentration profiles for nutrients, metabolic by-products, and product. In batch operation, these concentrations change all the time, while they are maintained at constant levels in continuous perfusion. Changing cellular environment impacts not only cellular function but also the quality attribute profile of the secreted product. These issues are minimized or eliminated for continuous perfusion culture leading to more stable and consistent production [6]. The impurity levels for a biopharmaceutical product need to meet stringent limits imposed by regulatory agencies. Both process-related (host cell protein, DNA, selective agents, media and feed additives) and product-related (aggregates, nonglycosylated heavy chain, etc.) impurities are reduced significantly by purification processes [1, 5]. The efficiency of purification processes for the removal of these impurities is impacted by the starting levels in the bioreactor. Because of its advantages for lower residence time, higher viability, and continuous removal of culture, lower impurity levels are expected in continuous culture. As the culture progresses in batch mode, the cellular environment changes and cells undergo different phases (growth, stationary, and decline). The cell physiology (growth, metabolic, and production rates) cannot be controlled, so obtaining consistent product profile under these conditions is more difficult for batch cultures. On the other hand, continuous cultures are typically operated at “steady-state” (after an initial build up phase), so a uniform environment is provided to the cells and stable production and consistent product profiles are easily achieved [8].

14.3.4

Scale-up and Commercial Production

Compared to batch mode, continuous perfusion can achieve the same amount of production capacity using smaller and more compact bioreactors because of higher productivities achieved. Using multiple (3–5×) 1000 l bioreactors, for instance, can provide the same output as a 20000 l bioreactor. This feature of continuous perfusion allows scaling up the process by increasing the number of bioreactors instead of increasing the volume of the bioreactor. Scaling up by numbers, or scaling out, allows for the use of the same size bioreactor for process development, clinical production, and commercial manufacturing. This eliminates the potential issues that can arise by increasing the bioreactor size and scale-up in general.

353

354

14 Continuous Biomanufacturing

The use of small-scale equipment also impacts the size of the facility and its capital cost. A typical large-scale biomanufacturing facility can cost $500 million to $2 billion to build and millions of dollars to operate [6]. A continuous perfusion-based facility on the other hand can be built at a lower cost and operated more efficiently. Compact bioreactors also allow flexibility and ease of expansion of biomanufacturing by adding new bioreactors to the facility.

14.4 Cell Retention Systems for Continuous Perfusion As mentioned in Section 14.2, there are several methods used for cell retention in the bioreactor for perfusion operation. The use of a solid support for immobilization, encapsulation, attachment, and entrapment eliminates the need for a special cell retention device but requires a special bioreactor design in most cases in an attempt to control the cellular environment without adequate mixing. The issues related to bioreactor design and operation, culture homogeneity, and process scalability limit the utilization of these systems. The use of a cell retention device, on the other hand, solves these issues. These devices, which are preferred by the industry, will be covered in detail in this section. 14.4.1

Cell Retention Devices

There are several mechanisms for separating cells from the culture medium and there have been many devices developed to utilize them. Filtration, centrifugation, gravitational settling, acoustic sound separation, and fluid dynamics are used successfully for cell retention during perfusion [6–8]. Table 14.1 shows these cell retentions systems and indicate which mechanisms they use for cell retention. 14.4.1.1

Filtration-Based Devices

The relatively large size of mammalian cells (15–20 μm in diameter) allows for effective use of microfiltration units (with 0.2 μm pore size) for cell separation. Although a straight filtration device provides 100% cell retention, it cannot be Table 14.1 Cell retention devices and the mechanisms for their operation. Cell retention system

Principle for cell retention

ATF or TFF

Tangential filtration

Spin filters

Rotating mesh-based filtration

Continuous centrifugation

Centrifugal separation

Settler

Gravitational settling

BioSep

Acoustic sound

Hydrocyclones

Fluid dynamics

Source: Ozturk and Kompala 2006 [7]. Reproduced with permission of Springer Nature.

14.4 Cell Retention Systems for Continuous Perfusion

operated for a long time because of fouling or clogging issues that develop over time. This fouling of the filters can be minimized when tangential flow filtration (TFF) mode is applied as opposed to the dead-end mode. Tangential filtration mode utilizes a cross flow on the surface of the filter in an effort to push the particles away from the surface while a permeate across the filter material is removed at a slower flow rate [6, 9]. TFF can be achieved by a circulation pump (conventional TFF) or for the case of ATF, by a diaphragm actuated by an alternating pressure/vacuum cycle [9, 14]. Both TFF and ATF create a cell-free harvest (100% cell retention). This is a great advantage as it eliminates the cell clarification step from the manufacturing process. TFF operation requires two connections to the bioreactor: one to remove the culture from the bioreactor and the other to return it back to the bioreactor. The ATF, on the other hand, needs only one connection that functions as inlet and as well as an outlet to the device. Another difference between conventional TFF and ATF operation is that the TFF requires a circulation pump to move cells from the bioreactor while ATF accomplishes this movement by the action of a diaphragm. Both TFF and ATF are used as external devices and they can be replaced if one unit fails because of clogging or fouling. Although this creates flexibility, the use of an external unit requires cells to spend time in the device and the circulation line with no control of cellular environment (nutrient levels, temperature, pH, dissolved oxygen, etc.). In addition, shear is created through the use of pump or the action of a diaphragm. The residence time outside of the bioreactor, shear forces due to pumping or diaphragm actuation, filtration area per volume, and fouling rates need to be considered for the operation and scalability of these devices. 14.4.1.2

Spin Filters

Spin filters use a fast rotating, or spinning, screen to separate cells and retain them in the bioreactor [7, 15]. These devices can be used as an internal or external device. The spinning motion creates a fluid flow on the surface of the screen, which pushes the cells away and separates them from the harvest stream. Spin filters achieve cell separation efficiencies around 90%, but their performance depends on the rotation speed, cell density, mesh type and size of the screen, surface area, and harvest rate. Even though the opening on the screen is larger than the cells, the spinning motion prevents cells from moving through the screen. Some cells make it through the screen to the harvest and subsequently need to be removed in the cell clarification step [16]. The spin filters are typically placed inside the bioreactor as a cage-centered around the impeller. Even though they have very big openings, they can get clogged over time as cell debris accumulates on the screen material. Spin filters can typically run for one to two months and are operated at relatively lower cell densities to minimize clogging. Using a spin filter unit as an external device allows switching of the units in the case of failure and enables a longer operation. The capacity of a spin filter is proportional to the surface area of the screen, although many other variables play a role as well. The surface area to volume ratio for the spin filters decreases as the volume of the bioreactor increases. This phenomenon impacts the scalability and consistency of continuous perfusion

355

356

14 Continuous Biomanufacturing

between different scales. Spin filter systems also occupy a significant volume when used at large scale, making it more difficult to control the cellular environment. 14.4.1.3

Continuous Centrifugation

Centrifugation is a very effective separation method for cells and can be used for continuous perfusion [7, 17]. Although centrifugation is typically performed in batch mode, continuous centrifugation is preferred for this application. Continuous centrifugation devices are used as external units and require a circulation loop driven by a pump. Cells that are removed from the bioreactor enter to centrifugation unit and concentrated cells return to the bioreactor while harvest with a lower cell concentration are continuously collected. Modifications can be introduced to the centrifugation unit to increase the efficiency of cell separation. A multistack or disk stack centrifuge uses multiple conical disks spaced out around the bowl, which are rotated at high speed. The cells from the bioreactor are fed from the top, distributed on the disks, and separated between the spacers, or in channels. The concentrated cells from each channel are collected continuously and are combined at the exit. The multistack centrifuges can be used as continuous, but an intermittent operation can provide more efficient separation. The separation of cells creates a cell-clarified stream or concentrate with a lower cell concentration, and this stream is used as harvest. The multistack centrifuges are difficult to be cleaned after use and they can have operational issues in long-term use. Even though the centrifuge is an open perfusion system, build-up of cell debris and dried solids in the channels can cause clogging of the centrifuge resulting in termination of the run. Another popular centrifugation device is the CentriTech unit, which uses a plastic bag or pouch instead of a stainless steel centrifugal chamber. Cells are taken into the bag from the bioreactor by pumping and are then separated by centrifugation. The concentrated cells are collected at the bottom of the bag and are removed intermittently by applied air pressure. As the bags are of single-use and discarded after use, cleaning is no longer an issue for this type of device. The efficiency of centrifugal separation is impacted by the size of the cells, degree of aggregation in the culture, centrifugation speed, temperature, flow rates, cell density in the bioreactor, and the geometry and dimensions of the centrifugal chamber. 14.4.1.4

Settler

Gravity can be used to separate and concentrate cells as long as the settling rate of the cells is higher than the linear flow rate of the harvest stream [7, 18]. As cells are small and the density difference between cells and the medium is not significant, cell settling rate is typically low (about 1 cm/h. for single cells), so gravity-based cell retention can only operate at low flow rates. Proportionally larger settling areas are required to increase the volumetric capacity (l/h) of a settler in order to maintain a constant linear rate during scale-up. When the settling area is increased, the volume of the settling device and the residence time outside of the bioreactor also increase. This situation limits the scalability of the settler device, thereby requiring an optimized design for scalability.

14.4 Cell Retention Systems for Continuous Perfusion

An inclined settler with a multilayer settling area allows large-scale continuous perfusion culture up to 2000 l/day capacity [7]. These settlers are typically operated with an angle (typically 60∘ ) and settling takes place in the channels between settling plates. The cell suspension from the bioreactor is first introduced to the settler’s bottom chamber and then moves along the channels upward by the harvest flow. As the cell suspension moves up, the cells gradually settle on the plates and then slide down toward the bottom chamber. After most of the cells are separated in the settler, the clarified harvest is removed from the top port of the settler [7, 18]. The settled cells at the bottom chamber are returned to the bioreactor by a pump. In order to increase the efficiency of settling, minimize cell attachment and build-up on the plates, and facilitate the return of the cells to the bioreactor, the settlers often utilize a vibrator placed at the top to vibrate the plates. Separation efficiency of the settler depends on the cell density, harvest and return rate, settling area, aggregation level of the cells, temperature, and overall design. The number of plates, their length and width, distance between plates, and the geometry of the bottom chamber can also impact performance. Removing the cells from the bioreactor to be separated in the settler can cause both mechanical and physiological stress to the cells and complicates the bioreactor operation. The pump used for the return of concentrated cells to the bioreactor can create significant shear damage if it is not properly designed and operated. The cellular environment (nutrient levels, dissolved oxygen, pH, temperature, etc.) cannot be controlled during the circulation of the cells to and from the settler. In order to alleviate these problems, the cells can be cooled using an in-line heat exchanger during circulation. Settlers can be operated for a long time (>6 months) and they do not suffer issues such as clogging. However, a settler cannot have 100% cell retention, so active cell growth is required to compensate the losses of cells in the harvest. Because of the size and density difference between viable and dead cells, a settler can selectively remove more dead cells than viable cells, resulting in a higher culture viability [19]. In long-term operation, the use of settlers can result in an increased level of aggregation in the bioreactor, as they selectively retain aggregated cells because of their higher settling velocity. 14.4.1.5

BioSep Device

Acoustic separation utilizes ultrasonic waves to align cells together so that they can be separated from the harvest flow [7, 20, 21]. The BioSep device (Applikon, Foster City, CA) uses a cell separating chamber (resonator) that is typically mounted to the head plate of the bioreactor. Two parallel glass plates on opposite side of the chamber are used to generate and reflect ultrasonic waves. At certain frequencies, a highly frequent acoustic standing field is generated in the chamber. This standing wave aligns the cells, horizontally creates layers of cells and “cell free” streams, and acts as a filter. The cells separated in the cell layer settle down back to the bioreactor and the “cell-free” layer is removed from the top of the chamber as harvest. BioSep is an open cell retention device that can be operated for a long time without clogging and is compact and effective, especially at small scale [21]. BioSep does not require a circulating loop and has no moving parts.

357

358

14 Continuous Biomanufacturing

The scalability of BioSep suffers from a fundamental problem; the distance between the transducer and reflector plates cannot be too big for the system to be operated without generating too much heat. Although some cooling can be provided to the chamber by blowing cold air, the chambers cannot have more than a 2.54 cm path. To increase the capacity for perfusion, multiple units in parallel are employed, a 50 l/day unit, for instance, is made of five 10 l/day units. The largest size of acoustic separation device could operate at 1000 l/day. 14.4.1.6

Hydrocyclones

The geometry of a hydrocyclone and its fluid dynamics in the chamber create a centrifugal force and a vortexing action [22]. The heavier cells are forced to move toward the center and bottom while the lighter particles and the harvest are allowed to escape the chamber. The cell suspension enters into the cylindrical section of the hydrocyclone tangentially and it is then forced to follow the centrifugal path in the vortex. The cells are concentrated in the middle, move to the conical base, and are pumped back to the bioreactor. Hydrocyclones are simple and compact devices with no moving parts. One important issue that prevents hydrocyclones from being widely used is the requirement of very high flow rates. Pumping cells at high flow rates can damage the cells irreversibly.

14.5 Operation and Control of Continuous Perfusion Bioreactors Although continuous perfusion with cell retention uses conventional stirred-tank bioreactors with temperature, pH, agitation rate, and dissolved oxygen control, it requires additional elements as presented in Figure 14.4. 14.5.1

Feed and Harvest Flow and Volume Control

Control of the bioreactor volume requires coordination between feed and harvest rates. As material such as base and antifoam can be added and culture can be removed as cell bleed, simply running the same feed and harvest rates do not guarantee constant bioreactor volume. Most of the continuous perfusion Feed

Bioreactor

Harvest

Cell retention device

Bleed

Return Weight

Figure 14.4 A typical continuous perfusion bioreactor system with an external cell retention device.

14.5 Operation and Control of Continuous Perfusion Bioreactors

bioreactors use harvest as the main pump and operate it continuously at the rate determined by perfusion and/or cell density control strategy. The feed pump, on the other hand, is actuated by a weight or level control and is often operated as on/off mode; when the weight of the level is below the set point, the feed is added, and when the weight or the level reaches the set point, it is turned off (Figure 14.4). 14.5.2

Circulation or Return Pump

Most of the external cell retention devices require a pump to move the cell suspension to the device and/or to return the concentrated cells back to the bioreactor. This pump is typically used in the return line and operated at a constant flow rate. Having the harvest and return pump in place, the flow to the device is accomplished without an additional pump. 14.5.3

Control of Perfusion Rate and Cell Density

Continuous perfusion bioreactor operation is typically divided into two phases: (i) the cell build-up phase in the first 7–10 days and (ii) a production phase for the rest of the run. It is customary to treat these phases separately for perfusion rate and cell density control. 14.5.3.1

Cell Build-up Phase

A continuous perfusion bioreactor starts in batch mode first, and when the cell density reaches to a certain level, the perfusion is initiated. The perfusion (harvest) rate starts low and is adjusted according to the perfusion control strategy adapted. These adjustments can be made based on cell density, metabolic rate, and metabolite concentrations. The increase to perfusion rate can be made continuously or in stepwise increments [6, 10]. Maintaining a constant cellular environment for the cells during a perfusion run is critical for stable and consistent production. Although controlling environmental variables such as pH, temperature, and dissolved oxygen use conventional controllers, controlling the levels of nutrients, metabolic by-products, and product is not an easy task. One popular method utilizes the cell-specific perfusion rate (CSPR), which increases the perfusion rate proportional to cell density. The concept of CSPR was first introduced in 1996 by Ozturk and is now widely applied to control the cellular environment, characterize and optimize perfusion cultures, and develop media and processes for perfusion [10]. The CSPR is defined as the ratio of medium exchange rate (D, harvest flow rate/volume of the bioreactor) to viable cell density (X v ): CSPR = D∕Xv Operating a bioreactor at a given CSPR establishes a “steady state” and results in an environment where the concentration of nutrient (S) and products (P) will be constant: S = So − qs ∕CSPR (nutrients) P = Po + qp ∕CSPR (products)

359

360

14 Continuous Biomanufacturing

With known levels in the feed (So and Po ) and information on specific consumption (qs ) and specific production (qp ), the targeted concentration of substrate and product can be achieved adjusting the CSPR using the calculations above. Once the CSPR target is set based on target nutrient and product levels, it is used to adjust the perfusion (harvest) rate proportional to cell density. Target CSPR can also be set based on process optimization and characterization studies. Adjustments to the perfusion rate can be made incrementally, based on off-line cell density measurements (typically performed daily), or continuously, based on on-line measurements or estimates of cell densities [3, 6, 10]. Cell density probes based on turbidity and capacitance and metabolic rates such as oxygen glucose consumption can be used for cell density estimations [10]. Some companies use a ramping strategy in the perfusion rate with a predetermined schedule in the cell build-up phase. 14.5.3.2

Production Phase

Once a certain cell density or a perfusion rate target is reached, the perfusion culture enters the production phase. The desired CSPR in production phase can be achieved by targeting the perfusion rate and the cell density [3, 6]. Typically, perfusion rate is capped at 1–2 volume/day medium exchange for monoclonal antibodies to minimize the media use and to increase the titer. The cell density is allowed to increase until a target is reached, and a cell density control is initiated by cell bleed or purge. 14.5.3.3

Cell Bleed or Purge

For a given medium exchange rate, each medium formulation results in a maximum cell density in the bioreactor. Although the bioreactor could be operated at this density, stability of the run, operational consistency, and robustness, other factors necessitate controlling the cell density at a level lower than the maximum. This is accomplished by bleed or purge stream as indicated in Figure 14.4. The bleed rate can be fixed (typically at 10% of the bioreactor volume) or it can vary based on a cell density control algorithm [10, 23]. The bleed results in positive growth, flushes cell debris, enhances viability, and allows the bioreactor to run stably and consistently.

14.6 Current Status of Continuous Perfusion Although most companies are using fed-batch as their platform of choice, continuous cultures have been adopted for many products because it offers advantages in productivity, product quality, consistency, and COGs [6]. In 2017, about 24 commercial products have been produced using continuous cultures that include not only labile proteins such as enzymes and blood coagulation factors but also stable molecules such as monoclonal antibodies (Table 14.2). Although continuous perfusion had always been popular, in the past three decades, its wide spread adoption by the pharmaceutical industry was rather slow. Fed-batch mode had been preferred because of its operational simplicity,

14.6 Current Status of Continuous Perfusion

Table 14.2 Commercial biopharmaceuticals produced using continuous culture with or without cell retention (based on Ref. [7], updated with latest product approvals). Product

Company

Year

Molecule

Centoxin*

Centocor*

1991

IgG1

Recombinate

Baxter

1992

rFVIII

Cerezyme

Genzyme

1994

β-glucocerebrosidase

ReoPro

Centocor/J&J

1994

Fab fragment

Gonal-F

Serono

1997

Human follicle stimulating hormone (FSH)

Benefix

Wyeth

1997

rFIX

®

Centocor/J&J

1998

IgG1

Merck (Serono)

1998

IFN β-1a

Novoseven

Novo Nordisk

1999

rFVIIa

ReFacto/Xyntha

Wyeth

2000

BDD-rFVIII

Kogenate-FS

Bayer

2000

rFVIII

Xigris

Eli Lilly

2001

Activated protein C

Replagal

Shire

2001

Human α-galactosidase A

Advate

Baxter

2003

rFVIII

Aldurazyme

Biomarin

2003

Iduronidase

Fabrazyme

Sanofi (Genzyme)

2003

Agalsidase β

Naglazyme

Biomarin

2005

Galsulfase

Myozyme

Sanofi (Genzyme)

2006

Alglucosidase alfa

Campath

Sanofi (Genzyme)

2007

IgG1

Stelara

Centocor/J&J

2009

IgG1

Simphoni

Centocor/J&J

2009

IgG1

VPRIV

Shire

2012

Velaglucerase alfa

SYLVANTTM

Centocor/J&J

2014

IgG1

Nuwiq

Octapharma

2015

BDD-rFVIII

Remicade

®

Rebif

® ® ® ®

* Not produced commercially.

robustness, and consistency. Because of advancement in feed development and medium enrichment as well as in bioreactor control, fed-batch can reach very high cell densities without medium exchange and yields very high titers [2, 11]. By comparison, continuous perfusion is complex, requires a high level of automation and control, employs complicated perfusion devices, uses more medium, and requires more work for process characterization and validation [6]. Therefore, unless fed-batch does not work, as is the case for labile proteins, companies had been reluctant to use continuous perfusion. Especially for the production of monoclonal antibodies and other relatively robust proteins, fed-batch is chosen over continuous perfusion [4]. The field of continuous perfusion advanced significantly in the past three decades, and most of the concerns about its technology were addressed. New cell retention devices were introduced, the operating conditions for the existing

361

362

14 Continuous Biomanufacturing

Table 14.3 Advances in continuous perfusion performance in the past three decades based on author’s own work and experience. Parameter

Unit

1990s

Now (2017)

Cell density

Million cells/ml

10

140

Titer

g/l

0.4

2

Bioreactor volume

l

1000

2000

Medium exchange rate

Volume/day

1

2

CSPR

nl/Million cells -day

0.1

0.014

Volumetric productivity

g/l-day

0.4

4

devices were optimized, new control strategies were implemented, and specially developed perfusion media were developed [6, 9, 24]. In addition, continuous perfusion now incorporates single-use (disposable) systems for better flexibility and modularity [6]. Table 14.3 presents the state-of-the-art results for continuous perfusion and compares process yields from today’s processes to those from 1990s. As shown in Table 14.3, significant improvements were achieved for cell density, titer, and volumetric productivity. Cell densities of 140 million/ml are achievable at two volumes per day medium exchange rates, whereas only 10 million/ml could be obtained three decades ago. The titers increased significantly as well, reaching about 2 g/l. New media formulations for perfusion allowed running the bioreactor at extremely low CSPRs, drastically minimizing medium use. Volumetric productivity levels increased an order of magnitude higher as well. It should be noted that the values in Table 14.3 are those obtained under cell density control, at steady state. Higher titers (2.5–3 g/l) and higher volumetric productivities (>5 g/l-day) were reported when the cell densities are allowed to go higher. In terms of bioreactor volumes and volumetric throughput, running a 2000 l bioreactor with 4000 l/day media exchange rate is possible and a continuous perfusion at 5000 l scale was demonstrated (J. Bonham-Carter, private communication).

14.7 Conclusions In this chapter, we focused on continuous perfusion cultures and discussed their advantages, the systems they use, their operations, and implementations. Perfusion with and without cell retention devices was covered and examples of retention devices were provided. Finally, current state-of-the-art perfusion technology, products from continuous culture, and the advancements in the field were outlined” with what has been learned and the perspectives of where the technology is heading.

Acknowledgment I would like to thank Lindsay Hock for the review of the chapter and for her valuable edits.

References

References 1 Ozturk, S.S. and Hu, W.-S. (eds.) (2006). Cell Culture Technology for Phar-

2 3

4

5

6

7

8

9

10 11 12 13

14

15

16

maceutical and Cell-Based Therapies. Boca Raton, FL: CRC Press, Taylor & Francis. Li, F., Vijayasankaran, N., Shen, A.Y. et al. (2010). Cell culture processes for monoclonal antibody production. mAbs 2: 466–479. Ozturk, S.S. (2014). Equipment for large scale mammalian cell culture. In: Mammalian Cell Cultures for Biologics Manufacturing, Advances in Biochemical Engineering/Biotechnology (eds. W. Zhou and A. Kantardjieff). New York, NY: Springer. Pollock, J., Ho, S.V., and Farid, S.S. (2013). Fed-batch and perfusion culture processes: economic, environmental, and operational feasibility under uncertainty. Biotechnol. Bioeng. 110: 206–219. Shukla, A.A. and Thommes, J. (2010). Recent advances in large-scale production of monoclonal antibodies and related proteins. Trends Biotechnol. 28: 253–261. Ozturk, S.S. (2014). Continuous manufacturing opportunities and challenges for the implementation of continuous processes to biomanufacturing. In: Continuous Manufacturing in Biopharmaceutical Manufacturing (ed. G. Subramanian), 457–478. Wiley. Ozturk, S.S. and Kompala, D. (2006). Optimization of high cell density perfusion cultures. In: Cell Culture Technology for Pharmaceutical and Cell-Based Therapies (eds. S.S. Ozturk and W.-S. Hu), 387–416. Boca Raton, FL: CRC Press, Taylor & Francis Group. Fenge, C. and Lüllau, E. (2006). Cell culture bioreactors. In: Cell Culture Technology for Pharmaceutical and Cell-Based Therapies (eds. S.S. Ozturk and W.-S. Hu), 155–224. Boca Raton, FL: CRC Press, Taylor & Francis Group. Bonham-Carter, J. and Shevitz, J. (2011). A brief history of perfusion biomanufacturing: how high-concentration cultures will characterize the factory of the future. BioProcess Int. 9: 24–31. Ozturk, S.S. (1996). Engineering challenges in high density cell culture systems. Cytotechnology 22: 3–16. Kelley, B. (2009). Industrialization of mAb production technology: the bioprocessing industry at a crossroads. mAbs 1: 443–452. Ecker, D.M., Jones, S.D., and Levine, H.L. (2015). The therapeutic monoclonal antibody market. mAbs 7: 9–14. Vogel, J.H., Nguyen, H., Giovannini, R. et al. (2012). A new large-scale manufacturing platform for complex biopharmaceuticals. Biotechnol. Bioeng. 109: 3049–3058. Clincke, M.F., Molleryd, C., Zhang, Y. et al. (2013). Very high density of CHO cells in perfusion by ATF or TFF in WAVE bioreactor. Part I. Effect of the cell density on the process. Biotechnol. Progr. 29: 754–767. Yabannavar, V.M., Singh, V., and Connelly, N.V. (1994). Scaleup of spin filter perfusion bioreactor for mammalian cell retention. Biotechnol. Bioeng. 43: 159–164. Fenge, C., Buzsaky, F., Fraune, E., and Lindner-Olsson, E. (1992). Evaluation of spin filter during perfusion culture of recombinant CHO cells. In: Animal

363

364

14 Continuous Biomanufacturing

17

18

19

20 21

22

23

24

Cell Technology: Developments, Process and Products (eds. R.E. Spier, J.B. Griffiths and C. MacDonald), 429–433. Butterworth-Heinemann. Takamatsu, H., Hamamoto, K., Ishimaru, K. et al. (1996). Large-scale perfusion culture process for suspended mammalian cells that uses a centrifuge with multiple settling zones. Appl. Microbiol. Biotechnol. 45: 454–457. Lamotte, D., Syraczek, J., and Marc, A. (1998). Cell-settler perfusion system for the production and glycosylation of human interferon-g by clumped cells. In: New Developments and New Applications in Animal Cell Technology (eds. O.-W. Merten, B. Perrin and B. Griffiths), 395–397. Kluwer Academic Publishers. Searles, J.A., Todd, P., and Kompala, D.S. (1994). Perfusion culture of suspended CHO cells employing inclined sedimentation. In: Animal Cell Technology: Products of Today, Prospects for Tomorrow (eds. R.E. Spier, J.B. Griffiths and W. Berthold), 240–242. Butterworth-Heinemann. Zhang, J., Collins, A., Chen, M. et al. (1998). High-density perfusion culture of insect cells with a biosep ultrasonic filter. Biotechnol. Bioeng. 59: 351–359. Ryll, T., Dutina, G., Reyes, A. et al. (2000). Performance of small-scale CHO perfusion cultures using an acoustic cell filtration device for cell retention: Characterization of separation efficiency and impact of perfusion on product quality. Biotechnol. Bioeng. 69: 440–449. Jockwer, A., Medronho, R.A., Wagner, R. et al. (2001). The use of hydrocyclones for mammalian cell retention in perfusion bioreactors. In: Animal Cell Technology: From Target to Market. ESACT Proceedings, vol. 1 (eds. E. Lindner-Olsson, N. Chatzissavidou and E. Lüllau). Dordrecht: Kluwer Academic Publishers. Deschênes, J.S., Desbiens, A., Perrier, M., and Kamen, A. (2006). Use of cell bleed in a high cell density perfusion culture and multivariable control of biomass and metabolite concentrations. Asia-Pac. J. Chem. Eng. 1: 82–91. Ray, N.G., Ozturk, S.S., Tung, A.S. et al. (1993). Medium optimization for perfused culture to provide high product titer. In: Animal Cell Technology: Basic & Applied Aspects Animal Cell Technology (eds. S. Kaminogawa, A. Ametani and S. Hachimura), 383–390. Dordrecht: Springer.

365

15 Process Analytical Technology and Quality by Design for Animal Cell Culture Hae-Woo Lee 1 , Hemlata Bhatia 2 , Seo-Young Park 2 , Mark-Henry Kamga 2 , Thomas Reimonn 2,3 , Sha Sha 2 , Zhuangrong Huang 2 , Shaun Galbraith 2 , Huolong Liu 2 , and Seongkyu Yoon 2 1 Daegu-Gyeongbuk Medical Innovation Foundation, Drug Manufacturing Center, 88 Dongnae-ro, Daegu 41061, South Korea 2 University of Massachusetts Lowell, Department of Chemical Engineering, 1 University Ave, Lowell, MA 01854, USA 3 University of Massachusetts Medical School, Program in Bioinformatics and Integrative Biology, 55 Lake Avenue North, Worchester, MA 01605, USA

15.1 PAT and QbD – US FDA’s Regulatory Initiatives In September 2004, the US Food and Drug Administration (FDA) issued a guidance document “PAT-A framework for innovative pharmaceutical development, manufacturing and quality assurance” [1, 2]. This guidance defines PAT as “a system for designing, analyzing, and controlling manufacturing through timely measurements (i.e. during processing) or critical quality and performance attributes of raw and in-process materials and processes, with the goal of ensuring final product quality.” After the initial release of the PAT guidance, this strategy was further extended into the initiative of quality by design (QbD) [1], including the concepts of creating a manufacturing knowledge base, risk management principles, process design space, and PAT [3]. Along with these recent efforts made by the pharmaceutical industry as well as the US FDA to introduce science-based and innovative manufacturing practice, conventional regulative paradigm based on predefined fixed recipes and product testing is shifting into more flexible and risk-based control schemes with the goal of more efficiently achieving the desired product quality, safety, and efficacy [4].

15.2 PAT and QbD – Challenges The PAT and QbD approach was firstly adopted in the manufacturing processes of pharmaceuticals with low molecular weights, which are usually produced by a series of unit operations, such as powder blending, granulation, tablet compression, and coating. Application of modern analytical techniques, such as near infrared (NIR), acoustic sensor, nuclear magnetic resonance (NMR), Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

366

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

Raman, Fourier transform infrared (FTIR), gas chromatography (GC), and high-performance liquid chromatography (HPLC), to these processes allowed real-time (on- or in-line) or near real-time (i.e. at-line) monitoring of critical process parameters (CPPs) and product quality attributes [5]. Among those technologies, NIR spectroscopy has been widely used because of its fast, cheap, noninvasive, and nondestructive nature. Over the last few years, numerous publications of NIR spectroscopy have appeared in the field of pharmaceutical technology [6, 7], covering identification of raw materials, characterization of process performance, and quality control of intermediate/final product attributes in a real-time manner. The examples include a diverse range of applications such as estimation of active content [8], powder-flow characterization [9], raw material analysis [10], estimation of dissolution rate [11], and many others. These innovations in the process analytics were followed by technical advances in the manipulations of large amounts of data for identifying/reducing variation, managing process risks, relating process information to critical quality attributes (CQAs), and determining process improvement opportunities. Thus, an adoption of PAT and QbD is being further in pharmaceutical industries [12]. Accordingly, these trends are being adopted by another newly emerging sector of biopharmaceuticals, which have complex physicochemical structures and high molecular weight, to adopt these modern bioprocess monitoring tools in all stages of the manufacturing processes [13]. However, for the biopharmaceuticals, the related work is still at the initial phase and various techniques are being tested at present time for evaluating their feasibility in terms of critical requirements posed by intrinsic complexity of bioprocess. A biggest challenge in applying PAT and QbD framework lies in the use of several complex raw materials, variations that originated from raw material, media and inoculums, uncertainty in the biological processes, limited knowledge about basic mechanism, and their unknown effects on the CQAs of final products [4, 14]. In addition, regulatory issues and concerns in the use of biopharmaceuticals make the situation more complicated and require extra activities such as method validation and post-approval change control. However, similar applications of advanced process analytics have been practiced for several decades in other biological industries [15], providing a good starting point for applying PAT and QbD to biopharmaceuticals.

15.3 PAT and QbD Implementations 15.3.1

NIR Spectroscopy

The monitoring of mammalian cell cultures in manufacturing sites is still largely based on routine physical sensors and electrodes for measuring basic properties such as temperature, pH, dissolved oxygen, pressure, airflow, liquid flow, foam level, and stirring rate [13]. However, to develop a scheme for realizing advanced bioprocess control, monitoring, and optimization, a number of measurements have to be made for raw material, inoculums, cell culture media, broth, metabolites, and products. Under these circumstances, of particular interest is the

15.3 PAT and QbD Implementations

rapidly evolving field of spectroscopic techniques, such as NIR, fluorescence, IR, Raman, etc. At the same time, other various instrumental analysis techniques have also been exploited as well mostly in academia for the purpose of providing a rich source of information about the process conditions [16]. Among them, NIR spectroscopy has been most extensively studied for the determination of individual component concentration in cell culture broth [17–23]. In Harthun et al., NIR was used to predict the concentration of the recombinant human antithrombin III, as well as glucose, lactate, glutamine, glutamate, and ammonia in mammalian cell culture process. Although their initial results were based on off-line measurements, in other studies, a similar approach could be easily extended into in situ on-line monitoring of protein product, ammonium and phosphate during the cell culture [23]. Another strength of NIR spectroscopy in biopharmaceutical manufacturing processes lies in the raw materials identification and culture medium qualification as demonstrated by many research groups [24, 25], where NIR was utilized to distinguish between good and poor performing media lots in mammalian cell culture producing a recombinant therapeutic protein. In Riley et al. [26], totally 19 components in serum-containing animal cell culture media were simultaneously quantified by the use of Fourier transform NIR spectroscopy. Furthermore, recent advances in NIR spectroscopy hardware enabled the use of fiber optics probes, multiplexer [27], and 2D imaging technique [28], which accelerated the wide acceptance of NIR as a fast, cost-effective, and noninvasive monitoring tool in biopharmaceutical industries. 15.3.2

Mid-Infrared (MIR) Spectroscopy

Another useful application of vibrational spectroscopy is the use of the mid-infrared (MIR) region, where a better degree of spectral resolution can be achieved [13]. MIR spectroscopy has been used for identification of foreign materials in biopharmaceutical manufacturing [8], monitoring of key components in cell culture broth [29–32], and quality assay of final products [33]. However, MIR has been received less attention compared to NIR because of higher water absorbance as well as difficulties of multivariate calibration [31, 34]. 15.3.3

Raman Spectroscopy

Raman spectroscopy has different mechanism with those of NIR and IR absorption and often provides complementary information about chemical composition and molecular structure of biopharmaceutical products [35]. The use of Raman spectrometry in cell culture processes has been demonstrated for the analysis of broth components profile [36, 37] and complex cell culture media solutions [38]. Recently, it has also been confirmed that Raman spectra can provide information about cell-line differences including phenotypes of cells and growth states [39]. However, a more attractive feature of Raman spectroscopy in the area of biopharmaceuticals is its ability to monitor structural and chemical changes of protein products [40–42]. This usefulness of Raman spectroscopy has been verified by many studies and examples include determination of protein conformation [43, 44], estimation of protein aggregation/fibrillation [45, 46],

367

368

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

analysis of side chain structure [47, 48], detection of site-specific mutant [49], and inspection of protein–protein interactions [50, 51]. Although most of them were based on off-line measurements, a recent study by Mungikar and Kamat [35] has revealed in-line measurement was also possible to measure therapeutic protein’s aggregation induced by manipulation of process conditions and formulations. A few years ago, a major limiting factor in applying Raman spectroscopy to biopharmaceuticals was the fluorescence interference from biomolecules, but recent progresses in optical technology and spectroscopic instrumentation overcome this handicap, gaining more and more attention in the biopharmaceutical field [52]. 15.3.4

Fluorescence Spectroscopy

Fluorescence spectroscopy is another useful tool for bioprocess monitoring, particularly because of the fact that many biologically relevant components, including amino acids, enzymes, cofactors, and vitamins, exhibit fluorescence properties [53]. Especially, 2D fluorescence spectroscopy, where several excitation and emission wavelengths were measured simultaneously, has been widely utilized to monitor biomass as well as cell culture profiles in the bioprocess [53–55]. In Ryan et al. [56], novel fluorescence spectroscopic techniques were employed for routine analysis of cell culture media in Chinese hamster ovary (CHO) cell culture, and it was demonstrated that the fluorescence spectra of media can predict the end product yield reliably when it was combined with chemometric modeling technique. 15.3.5

Chromatographic Techniques

Many instrumental analysis techniques, such as liquid chromatography (LC) and HPLC, are routinely performed in an off-line manner to assess CQAs of intermediate and final products. Significant improvements in recent on-line HPLC techniques now allow its direct connection to cell culture system, and this was employed to monitor various substrates, such as glucose, glutamine, lactate, etc., in bioreactors [57, 58]. Among many studies, Larson et al. [59] implemented on-line HPLC in mammalian cell culture, and they were capable of measuring 17 amino acids and glucose from their experiments. Characterization of therapeutic proteins produced from cell cultivation system is also a common application area of HPLC, and it serves as a routine analysis method, having high sensitivity and resolution [60, 61]. Mass spectrometry can also provide valuable information about the identity of analytes on the molecular level [62] and has become an important analytical tool in the field of biopharmaceutical research. Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) has been used extensively for the qualitative analysis of peptides, proteins, and amino acids in cell culture media [63–65], as demonstrated by Dally et al. [64]. In their study, 12 underivatized free amino acids, including arginine, aspartic acid, glutamic acid, glycine, histidine, methionine, phenylalanine, proline, serine, threonine, tyrosine, and valine in mammalian cell culture medium, were successfully

15.3 PAT and QbD Implementations

quantified for rapid process optimization. Shen et al. [66] also quantified lipids, determined as fatty acid methyl esters (FAMEs), in complex serum-free cell culture media using GC/MS. For the characterization of therapeutic protein structure, such as glycosylation, phosphorylation, and primary/secondary conformation, various mass spectrometric techniques have been studied, and these examples include the uses of electrospray mass (ES-MS), MALDI-TOF-MS, fast atom bombardment mass (FAB-MS), GC/MS, LC/MS, tandem MS/MS, and many others [67–69]. Currently, these techniques are routinely being used as a validated assay to allow characterization of different biochemical, which are present in the biopharmaceutical manufacturing processes, covering from small synthetic peptides to conjugated antibody molecules [70]. 15.3.6

Other Useful Techniques

Apart from the above mentioned techniques, there are other advanced analytical methods that support PAT and QbD in biopharmaceutical field. They generally include diverse ranges of techniques, such as dielectric spectroscopy [71–73], electric nose [74–76], cytometry [77, 78], enzyme-linked immunosorbent assay (ELISA) [67], NMR [79, 80], etc. For example, in Cannizzaro et al. [71] and Lee et al. [81], in situ dielectric spectroscopy was applied for measuring CHO cell density, and they could quantify viable cell density in bioreactors using these in situ measurements. In another application of Bachinger et al. [74], gaseous emissions from mammalian cell cultures were measured using electronic nose noninvasively and they could recognize the microbial contamination of bioreactors much earlier than conventional methods. Intracellular metabolic fluxes in CHO cell cultures have also been estimated using cytometry [78] and NMR [79], providing valuable information about current physiological status of cells in the bioprocess. Rudt et al. [82] has also reviewed spectroscopy application in downstream processing of biologics under the PAT framework. 15.3.7

Data Analysis and Modeling Tools

As illustrated in the above examples, there exist a diverse range of advanced process analytics for measuring several crucial process parameters and CQAs in real-time or near real-time manner. In order to explore abundant data sets and take full advantages of those process analytics, efficient data analysis and knowledge extraction tools are generally needed [83]. From the perspectives of efficient database analysis, chemometric techniques are normally the method of choice to handle large sets of multidimensional data and to link physicochemical information of process analytics to the desired product quality attributes [13, 53]. Principal component analysis (PCA), partial least squares (PLS), and soft independent modeling of the class analogy (SIMCA) are some examples of chemometric approaches, which can provide high level of process and product understanding from PAT-based process measurements. They have been applied to multivariate batch modeling/monitoring, design space monitoring, and post-mortem analyses of batch performance, real-time bioprocess monitoring, and soft sensor development in biopharmaceutical manufacturing [12].

369

370

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

Alternately, hybrid modeling approaches can also be used by combining mechanistic and chemometric approaches when a priori mechanistic knowledge is available [84, 85]. Mercier et al. [86] has reviewed various multivariate data analytics and chemometric techniques in context of PAT and biopharmaceutical applications. In addition, FDA officials and industry consortium have published two case studies of A-Mab and A-VAX to demonstrate QbD and PAT in the biological processes [87, 88]. These two case studies provide comprehensive demonstration on how PAT and QbD can be implemented under FDA guideline.

15.4 Case Studies In this section, a few case studies of PAT and QbD are summarized to demonstrate how the PAT and QbD concepts can be utilized in animal cell culture development and production. These case studies are based on the research and collaborations with a few biopharmaceutical companies and US FDA. It is advised to refer to references provided for full contents of the case studies. 15.4.1 Estimation of Raw Material Performance in Mammalian Cell Culture Using Near-Infrared Spectra Combined with Chemometrics Approaches Understanding variability in raw materials and their impacts on product quality is the key component to realize the benefits of QbD in the biopharmaceutical manufacturing processes [25]. Several spectroscopic techniques, such as NIR and Raman, have been studied for raw material characterization, providing fast, cheap, and nondestructive ways to measure raw material variability. However, investigations of link between spectra of raw materials and cell culture performance have been scarce because of their complexity and uncertainty. In this case study, NIR spectra and cell culture performance of multiple soy hydrolysate lots manufactured by different vendors were analyzed using chemometrics approaches in order to address variability of raw materials as well as correlation between raw material properties and corresponding CQA of cell culture process. PCA of NIR spectra was conducted, and it revealed that the NIR spectra of different soy lots contain enough physicochemical information about soy hydrolysates to allow identification of lot-to-lot variability as well as vendor-to-vendor differences. Figure 15.1 shows clustering pattern of multiple soy hydrolysate lots inspected by first two principal components (PCs) in PCA model. The identified compositional variability was further analyzed in order to link them into CQA of cell growth and protein production in two different mammalian cell lines. Under the condition of varying soy dosages, PLS regression combined with optimal variable selection was utilized to regress NIR spectra to cell growth and protein production. The performance of the resulting models demonstrates the potential of NIR spectroscopy as a robust lot selection tool for raw materials while providing a biological link between chemical composition of raw materials and cell culture performance. By using a chemometrics approach, this case study demonstrates that NIR spectra can be used to reveal lot-to-lot variability, as well as vendor-to-vendor differences of soy hydrolysate, which are commonly encountered in chemically undefined medium.

15.4 Case Studies

Scores on second PC (10.89%)

0.004

Lot #9

0.002

Lot #2

Lot #3

Lot #8

Lot #1

Lot #14 Lot #10

0.000

Lot #6

Lot #15

Lot #7

Lot #4 Lot #5

Lot #13 Lot #12

–0.002

Lot #11

–0.004

–0.005 0.000 0.005 Scores on first PC (72.14%)

–0.010 (a)

0.010

Scores on first PC (52.31%)

10 Lot #3

5

Lot #1

Lot #2

Lot #9 Lot #5

Lot #8 Lot #6

0

Lot #12

–5

Lot #15

Lot #7 Lot #10

Lot #14 Lot #11

–10

8 (b)

Lot #4

Lot #13

6

4

2

0

–2

–4

–6

–8

Scores on second PC (16.82%)

Figure 15.1 Clustering pattern of multiple soy hydrolysate lots inspected by first two PCs in PCA model. (a) Score plot of near-infrared spectra. (b) Score plot of bioassay data. Lots with different vendors are represented by different color for clear comparisons (blue – vendor A; pink – vendor B; yellow – vendor C; green – vendor D). Dashed line represents 95% confidence limit in the corresponding score dimension. Source: Lee et al. 2012 [25]. Reproduced with permission of John Wiley and Sons.

371

372

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

15.4.2 Design Space Exploration for Control of Critical Quality Attributes of mAb In this case study, a design space (DS) exploration strategy defined as a function of four key scenarios was successfully integrated and validated to enhance the DS building exercise, by increasing the accuracy of analyses and interpretation of processed data [89]. The four key scenarios, defining the strategy, were based on cumulative analyses of individual models developed for the CQAs (23 glycan profiles) considered for the study. The analyses of the CQA estimates and model performances were interpreted as (i) inside specification/significant model, (ii) inside specification/nonsignificant model, (iii) outside specification/significant model, and (iv) outside specification/nonsignificant model. Each scenario was defined and illustrated through individual models of CQA aligning the description. The R2 , Q2 , model validity, and model reproducibility estimates of G2, G2FaGbGN, G0, and G2FaG2, respectively, signified the four scenarios stated. Figure 15.2 demonstrates a step-by-step process of building design space used in the case study. Through further optimizations, including the estimation of edge of failure and set point analysis, wider and accurate DS were created for each scenario, establishing a critical functional relationship between CPPs and CQAs. A DS provides the optimal region for systematic evaluation, mechanistic understanding and refining of a QbD approach. DS exploration strategy will aid the critical process of consistently and reproducibly achieving predefined quality of a product throughout its lifecycle. 15.4.3 Quantification of Protein Mixture in Chromatographic Separation Using Multiwavelength UV Spectra In therapeutic protein production, the protein purification with chromatographic processes is of high importance in separating the qualified proteins from the impurities for consistent product quality [90]. Therefore, to aid real-time monitoring of the protein purification processes, various kinds of methodologies have been proposed until now. However, the majority of them still rely on the use of a single ultraviolet (UV) absorbance or the utilization of expensive and time-consuming instruments, thus requiring a simple, fast, and cost-effective methodology for protein quantification. In this study, the feasibility of using multiwavelength UV spectroscopy was investigated as an alternative tool for the real-time monitoring of the protein mixtures in protein purification. Three different proteins were selected as a model system for the protein mixture, and the multivariate UV spectra were analyzed to construct the reliable quantification models for different proteins of interest. By using various chemometrics tools, such as PLS, the validity of estimating the protein concentration from the UV spectra of the mixture samples was rigorously analyzed with their prediction performance. Figure 15.3 illustrates the actual vs. predicted values of the protein concentration obtained from the PLS model for the calibration and external test sets. The results indicated that the multiwavelength UV spectra had sufficient sensitivity and accuracy to estimate the protein concentrations in mixture,

3. Classification

1. Experimental setup and data ∙ Temperature ∙ Mixing ∙ Aeration ∙ pH ∙ Nutrients ∙ Harvest

Yes

No Is Model Significant

Is Model Significant

Yes Batch

Are the Results inside specification

Input VariablesCPP

No

Yes

No

Results-CQA Case A

CPP CPP1 CQA CQA CQA1 CPP2 1 1 1 2 1 1

Low Low

Low 0.22 0.24

0.35

2

High Low

Low 0.25 0.34

0.32

3

Low High Low 1.12 0.54

0.34

4

High High Low 0.78 0.64

0.82

5

Low Low High 0.64 0.72

0.24

6

High High High 0.34 1.23

0.78

Case B

Case A

Case B

2. Mathematical Analysis 1

R2 Q2 Model validity Reproducibility

0.8

0.6

4. Design space exploration 0.4

%

37 10

36.8

0.2

50 10

36.6 Temp

0

–0.2 M3

M4

G0

M5

G16 G1F6 G2

M7 G2FaG G2FSa G2FSG

5

36.4

2 36.2

5

1

36

0.5

35.8 35.6 20

22

24

26

28

30 DO

Figure 15.2 A guide to design space implementation. Source: Bhatia et al. 2016 [89]. Reproduced with permission of Elsevier.

32

34

36

38

Sparge = 0.5 Agit = 105.4 Inoc = 74 PHshift = 0.3 Tempshift = 0.011 Feed = 0 40 NE-AA = 0 FA = 0.33 hCortisone = 0.005

15 Process Analytical Technology and Quality by Design for Animal Cell Culture 1.6

1.6

1.4

1.4

Predicted BLG (mg/ml)

Predicted BSA (mg/ml)

1.2 1.0 0.8 0.6 0.4 0.2

Calibration Test

0.0

True BSA (mg/ml)

1.2 1.0 0.8 0.6 0.4 0.2

Calibration Test

0.0

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

(a)

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

True BLG (mg/ml)

(b)

1.6

Predicted IgG (mg/ml)

374

1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0

Calibration Test 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

(c)

True IgG (mg/ml)

Figure 15.3 Actual vs. predicted values of the protein concentrations obtained from the best PLS models for the calibration and external test sets. (a) Bovine serum albumin (BSA), (b) β-lactoglobulin (BLG), and (c) IgG. Source: Kamga et al. 2013 [90]. Reproduced with permission of John Wiley and Sons.

demonstrating its usefulness for the rapid quantification of the protein mixtures in protein purification. 15.4.4 Characterization of Mammalian Cell Culture Raw Materials by Combining Spectroscopy and Chemometrics Two of the primary issues with characterizing the variability of raw materials used in mammalian cell culture, such as wheat hydrolysate, is that the analyses of these materials can be time-consuming, and the results of the analyses are not straightforward to interpret [91, 92]. To solve these issues, spectroscopy can be combined with chemometrics to provide a quick, robust, and easy-to-understand methodology for the characterization of raw materials, which will improve cell culture performance by providing an assessment of the impact that a given raw material will have on final product quality. In this study, four spectroscopic technologies, NIR spectroscopy, MIR spectroscopy, Raman spectroscopy, and fluorescence spectroscopy, were used in conjunction with PCA to characterize the

15.4 Case Studies

variability of wheat hydrolysates and to provide evidence that the classification of good and bad lots of raw material is possible. Then, the same spectroscopic platforms are combined with PLS regressions to quantitatively predict two cell culture CQAs: integrated viable cell density and IgG titer. The results showed that NIR spectroscopy and fluorescence spectroscopy are capable of characterizing the wheat hydrolysate’s chemical structure, with NIR performing slightly better, and that they can be used to estimate the raw materials’ impact on the CQAs. Figure 15.4 demonstrates raw spectra of wheat hydrolysate from four spectral technologies. These results were justified by demonstrating that all the components present in the wheat hydrolysates, six amino acids, arginine, glycine, phenylalanine, tyrosine, isoleucine, and threonine, and five trace elements, copper, phosphorus, molybdenum, arsenic, and aluminum, had a large, statistically significant effect on the CQAs and that NIR and fluorescence spectroscopy performed the best for characterizing the important amino acids. It was also found that the trace elements of interest were not characterized well by any of the spectral technologies used; however, the trace elements were also shown to have a less significant effect on the CQAs than the amino acids. 15.4.5 Effect of Amino Acid Supplementation on Titer and Glycosylation Distribution in Hybridoma Cell Cultures Genome-scale flux balance analysis (FBA) is a powerful systems biology tool to characterize intracellular reaction fluxes during cell cultures [93]. FBA estimates intracellular reaction rates by optimizing an objective function, subject to the constraints of a metabolic model and media uptake/excretion rates. A dynamic extension to FBA, dynamic flux balance analysis (DFBA), can calculate intracellular reaction fluxes as they change during cell cultures. In a previous study by Read et al. [94], a series of informed amino acid supplementation experiments were performed on 12 parallel murine hybridoma cell cultures, and these data were leveraged for further analysis. In order to understand the effects of media changes on the model murine hybridoma cell line, a systems biology approach is applied in the current study. DFBA was performed using a genome-scale mouse metabolic model, and multivariate data analysis was used for interpretation. The calculated reaction fluxes were examined using PLS and PLS discriminant analysis. Figure 15.5 illustrates the flux values calculated during the exponential growth period. The results indicate that media supplementation increases product yield because it raises nutrient levels extending the growth phase, and the increased viable cell density allows for greater culture performance. At the same time, the directed supplementation does not change the overall metabolism of the cells. This supports the conclusion that product quality, as measured by glycoform assays, remains unchanged because the metabolism remains in a similar state. Additionally, the DFBA shows that metabolic state varies more at the beginning of the culture but less by the middle of the growth phase, possibly because of stress on the cells during inoculation.

375

(a)

(b)

100

Transmittance (%)

Intensity

1000

500

700

600

500

400

300

Emmission wavelength (nm)

350

300

400

80 70 60

450 3500

Excitation wavelength (nm)

(c)

3000

2500 2000 Wavenumber (cm–1)

1500

1000

500

(d) 60

450 k 400 k

50

350 k 40

Intensity

Reflectance (%)

90

30 20

300 k 250 k 200 k 150 k

10

100 k 12000

11000

10000

9000 8000 7000 Wavenumber (cm–1)

6000

5000

4000

1000

1500 2000 Raman shift (cm–1)

2500

3000

Figure 15.4 Raw spectra of wheat hydrolysate from all four technologies. (a) Raw fluorescence spectrum, (b) raw middle-infrared spectra, (c) raw near-infrared spectra, and (d) raw Raman spectra. Source: Trunfio et al. 2017 [92]. Reproduced with permission of John Wiley and Sons.

15.4 Case Studies Glucose uptake 0 -2E-10

5E-25 0 -5E-25 -1E-24 -1.5E-24 -2E-24

Glucose-6P 6P-Gluconate

-4E-10 -6E-10

1E-11 Ax1d Ax2d Ax3d Ax4d Ay1d Ay2d Ay3d Ay4d Bc2a Bc2b Bc2c Bc2d

Glycogen G1P

-3E-11

Ax1d Ax2d Ax3d Ax4d Ay1d Ay2d Ay3d Ay4d Bc2a Bc2b Bc2c Bc2d

Glucose

Glycogen

G1P

Ax1d Ax2d Ax3d Ax4d Ay1d Ay2d Ay3d Ay4d Bc2a Bc2b Bc2c Bc2d

0 -1E-11 -2E-11

Glucose-6P

6P-Gluconate

Glucose-6P Fructose-6P 0 -1E-11

Pyruvate Malate

-2E-11 1E-24 Ax1d Ax2d Ax3d Ax4d Ay1d Ay2d Ay3d Ay4d Bc2a Bc2b Bc2c Bc2d

-3E-11

Fructose-6P

-3E-24 Ax1d Ax2d Ax3d Ax4d Ay1d Ay2d Ay3d Ay4d Bc2a Bc2b Bc2c Bc2d

Ribulose-5P

0 -1E-24 -2E-24

Cytosol Lactate

Pyruvate

Malate

Pyruvate Lactate Acetyl-CoA

2E-11

CoA

1.5E-11

0

Ax1d Ax2d Ax3d Ax4d Ay1d Ay2d Ay3d Ay4d Bc2a Bc2b Bc2c Bc2d

1E-11 5E-12

Glu Oxoglutarate Glutamate 2E-11 1.5E-11

0

Mitochondria

Ax1d Ax2d Ax3d Ax4d Ay1d Ay2d Ay3d Ay4d Bc2a Bc2b Bc2c Bc2d

1E-11 5E-12

Figure 15.5 Flux values calculated during the middle of the exponential growth phase for key reactions in the central carbon metabolism. Reaction fluxes are shown in (mol/[cell × hour]). Bars represent one standard deviation of the calculated fluxes when the input uptake/excretion rates were randomly varied between 25% and 5% of their measured value. Source: Reimonn et al. 2016 [93]. Reproduced with permission of John Wiley and Sons.

15.4.6 Metabolic Responses and Pathway Changes of Mammalian Cells Under Different Culture Conditions with Media Supplementations Amino acids and glucose consumption, cell growth, and antibody production in mammalian cell culture are key considerations during upstream process and particularly media optimization [95]. Understanding the interrelations and the relevant cellular physiology will provide insight for setting strategy of robust and effective antibody production. The aim of this study was to further understand

377

378

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

the nutrient consumption metabolism, as this could have a significant impact on enhancing productivity, cell proliferation, designing feeding strategies, and development of feed media. The nutrient consumption pattern, monoclonal antibody (mAb) concentration, and cell growth were analyzed in hybridoma cell cultures under three different cell cultures with media supplementation of glucose, methionine, threonine, tryptophan, and tyrosine. The amino acid metabolism and its impact on cell growth and mAb production during the batch and fed-batch culture were closely analyzed. Figure 15.6 shows calculated metabolic flux in central metabolism and amino acid metabolism during the exponential growth phase. It was shown that the phenylalanine, tyrosine, and tryptophan biosynthesis pathways were significantly altered under different culture processes and media composition. These changes were more apparent in the fed-batch process in which higher mAb productivity was observed because of the metabolic changes than productivity in the batch process. The pathway analysis approach was well utilized for evaluating the impact on the relevant pathways involved under different cell culture conditions to enhance mAb production and cell growth. 15.4.7 Estimation and Control of N-Linked Glycoform Profiles of Monoclonal Antibody with Extracellular Metabolites and Two-Step Intracellular Models Real-time estimation of mAb glycosylation is highly desirable in manufacturing processes, but related progress has not been sufficiently reported [96]. Altering hybridoma cell culture conditions were found to change model monoclonal antibodies’ (mAbs) level of galactosylation. In this work, a two-component modeling framework integrating FBA and glycosylation kinetic model was demonstrated. The method was based on estimating nucleotide sugars as using those values as the input for generating predictions of glycosylation profiles. With the data tested from multiple altered cultures, it was shown that the model could successfully estimate the variation of glycosylation profiles from extracellular metabolite dynamic information. Through computational FBA using extracellular metabolite profiles, the substrate uridine diphosphate galactose (UDP-Gal) was found to be the key limiting factor of galactosylation. The FBA-estimated nucleotide sugar synthesis was followed by a mathematical model describing the kinetics of oligosaccharide attachment to antibodies and finally generated predictions of glycoform distribution of mAbs. Figure 15.7 shows a simplified concept of metabolic pathway in the FBA. The model outputs matched well with experimental outcomes across batches. The integrated mathematical approach combining FBA and kinetic models in which measured extracellular metabolites were the only inputs provided the robust predictive capability, thus building a basis for glycosylation control by manipulating these culture conditions in the media. Thus, the model can be potentially conducted in real time and used to monitor glycosylation against culture perturbations. Based on this, timely control strategies can even be further implemented to maintain the consistency of ongoing processes. As an extension of this work, different cell lines could be tested with the model to further explore the feasibility of this framework.

Figure 15.6 Overview of calculated metabolic flux in central metabolism and amino acid metabolism during the middle of exponential growth phase under 12 different culture conditions, represented in units of (mol/cell/h). The highest negative value (smallest value) in each reaction maps to the first (bright) color in the color map that indicates the consumption reaction that activated the most. The highest positive value (largest value) maps to the last (dark) color that indicates the production reaction that activated the most due to different culture conditions. Source: Park et al. 2018 [95]. Reproduced with permission of John Wiley and Sons.

Glc Glycogen

17

G6P

51

R5P

G1P Man

2 3

DHAP

17

G6P 81

GDPMan

R5P

52

4

82

F6P

UDP-GlcNAc 83

GDP-Fuc 5

Pyr

UDP-Gal

85

GlcNAc 86

UDP-GalNAc 87

90

92

GalNAc

Fuc Neu5Ac

TCA

6

15

CMPNeu5Ac

Glycolysis

91

3PG

Lac

80

UDP-Gluc

84

G3P

78

89

1

88

Nucleotide/ Nulcleotide Sugar

F6P 3

Glyc3P

Gal

Glc

1 53

CMPNeu5Gc 93

7

Neu5Gc

(b)

AcCoA 8

Oxal

Cit 9

14

aKG

Mal

Glu

10

(a)

12

SucCoA Succ

UDP-GlcNAc

G1F GDP GDP-Fuc

UDP UDP-Gal

CMP CMPNeu5Ac

82

13

Fum

UDP

29, 31, 34, 35, 39, 43

Gln

limiting

11

(c)

97

G0F G2F

Figure 15.7 Simplified view of metabolic pathways included in the flux balance analysis (FBA). (a) Central metabolism; (b) nucleotide sugar synthetic pathways; and (c) oligosaccharide modification process. The numbers labeled with reactions correspond with the ones initially used in the reference [97]. Source: Sou et al. 2015 [97]. Reproduced with permission of John Wiley and Sons.

15.4 Case Studies

15.4.8 Quantitative Intracellular Flux Modeling and Applications in Biotherapeutic Development and Production Using CHO Cell Cultures CHO cells have been widely used for producing many recombinant therapeutic proteins [98]. Constraint-based modeling, such as FBA and metabolic flux analysis (MFA), has been developing rapidly for the quantification of intracellular metabolic flux distribution at a systematic level. Such methods would produce detailed maps of flows through metabolic networks, which contribute significantly to better understanding of metabolism in cells (Figure 15.8). Although these approaches have been extensively established in microbial systems, their application to mammalian cells is sparse. This study brings together the recent development of constraint-based models and their applications in CHO cells. The further development of constraint-based modeling approaches driven by multi-omics data sets is discussed, and a framework of potential modeling application in cell culture engineering is proposed. Improved cell culture system understanding will enable robust developments in cell line and bioprocess engineering, thus accelerating consistent process quality control in biopharmaceutical manufacturing.

Figure 15.8 Recent development of constraint-based metabolic models and applications of models to CHO cell culture including media formation, feed strategy development, bioprocess control optimization, and cell line engineering. Source: Huang et al. 2017 [98]. Reproduced with permission of John Wiley and Sons.

381

382

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

The predominant concerns in the biopharmaceutical industry are centering on the cost-effective recombinant antibody production while satisfying regulatory requirements in safety and efficacy. Despite great improvement in therapeutic protein production from labor-intensive empirical methods in CHO cells, knowledge-driven approaches are not widely applied. As a result, it still struggles to find the answer for unpredictable behavior during cell culture bioprocess transfer and scale-up in industrial cell culture systems without a fundamental understanding of how or why these improvements are generated [99]. It is clear that future progress will be increasingly dependent on rational and efficient modifications through knowledge-based approaches, in which global, high-throughput approaches for data collection and analysis are necessary to understand the complicated relationship among genotype, bioprocess, and phenotype [100]. The final purpose of constraint-based metabolic models is to perform a systematic in silico evaluation of cellular behaviors under targeted genetic modifications or environmental perturbations. In addition, more importantly, such models can be used as platforms to generate and identify hypothesis and allow many improvement strategies to be simulated before experimental practice. The researchers now have access to perform in silico metabolic flux simulation and analysis at a system level based on genome-scale CHO cell metabolic network [101, 102]. Furthermore, increasing volumes of omics data in CHO cells are becoming publicly available with the help of advanced analytical techniques [99, 103], and it is believed that genome-scale constraint-based metabolic models are a reliable framework to fully utilize these multi-omics data. However, constraint-based approaches have several limitations. First, the models can only characterize the intracellular fluxes at steady state. Another shortage is the exclusion of metabolite concentrations, enzyme levels, and metabolic regulation. Thus, those models cannot predict the complex dynamic cellular responses to environmental and genetic perturbations [104]. Mechanistic-based kinetic models, alternatively, incorporate enzyme kinetics and metabolic regulation and enable to quantify optimal enzyme levels and metabolite concentrations required to enhance the productive capabilities. Some recent efforts have been made to build large-scale or even genome-scale kinetic models in E. coli [105, 106]. The kinetic models in CHO cells are still poorly developed especially because of the lack of a large number of kinetic parameters and the high complexity of CHO cell metabolic networks. Because of the characteristic features of stoichiometric models and kinetic models, the combined usage of these two methods is likely to be complementary and effective for systematic CHO cell engineering. As biosimilars in the biopharmaceutical industry become more popular, the need to achieve predefined CQAs to innovators’ products and fundamental understanding of cellular behaviors in response to bioprocess conditions and medium alterations are increasingly important [107]. The efficiency of research would be dramatically accelerated if constraint-based models could be employed to scan and define optimal culturing conditions and genetic modifications to produce desired glycosylation profiles. Genome-scale constraint-based metabolic models linking the information from entire cell culture system including cell culture conditions, cell metabolism, and protein glycosylation will offer great benefit to the rapid and economic development of biopharmaceutical manufacturing.

References

15.5 Conclusion The current status of PAT and QbD in the real world of biopharmaceutical manufacturing sites is still largely at the development stage, although significant advances have been accomplished in academia for past decades with regard to our ability to analyze and monitor key process and quality attributes. Furthermore, emphasis of the research had been mainly given toward the development of efficient innovative online monitoring tools until now and less attention has been paid toward the development of systematic methods and tools for efficient utilization of the existing measurement methods and accompanying control system [5, 83]. It means that some of the methods mentioned in the above have not been fully validated through intensive applications and the comprehensive integration of these advanced analytical tools to bring consistent product quality in biopharmaceuticals has also not been appeared so far. Nevertheless, there is a general agreement in biotech community that PAT and QbD approach through efficient monitoring system will provide assurance of the continued capability of processes and controls to meet product quality and to identify areas for continual improvement [2, 3, 108]. Implementation of advanced monitoring systems and their utilization toward further knowledge discovery is likely to result in gains in consistency of product quality as well as efficiency in manufacturing of biopharmaceutical products and bring us closer to full implementation of PAT and QbD from raw material assessment to final product assay, realizing its full benefits. Eventually, PAT and QbD frameworks can help to increase the safety and efficacy of medicines while reducing the time-to-market for new products and the operational costs of manufacturing. To realize these activities, much more rigorous analysis needs to be done with regard to utilizing the collected data for finding fundamental relationship between CPPs and product quality attributes for subsequent control of the process and it can be done by comprehensively combining all sources of information coming from these analytical bioprocess monitoring tools. Some efforts to bring these integrated approaches have been coming out recently [83]. Once the potential of applying such techniques for process control is clearly demonstrated, they will progressively become routine at the industrial environment of biopharmaceutical companies [13, 53].

References 1 ICH (2008). Q8(R1): Pharmaceutical development. International Confer-

ence on Harmonisation of Technical Requirements for the Registration of Pharmaceuticals for Human Use. Geneva, Switzerland. 2 FDA (2004). PAT-A Framework for Innovative Pharmaceutical Development, Manufacturing and Quality Assurance. Rockville, MD: U.S. Department of Health and Human Services Food and Drug Administration Center for Biologics Evaluation and Research.

383

384

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

3 Read, E.K., Park, J.T., Shah, R.B. et al. (2010). Process analytical technology

4

5

6 7

8

9

10

11

12

13

14

15

16 17

18

(PAT) for biopharmaceutical products: Part II. Concepts and Applications. Biotechnol. Bioeng. 105: 285–295. Read, E.K., Park, J.T., Shah, R.B. et al. (2010). Process analytical technology (PAT) for biopharmaceutical products: Part I. Concepts and Applications. Biotechnol. Bioeng. 105: 276–284. Rathore, A.S., Bhambure, R., and Ghare, V. (2010). Process analytical technology (PAT) for biopharmaceutical products. Anal. Bioanal. Chem. 398: 137–154. Luypaert, J., Massart, D.L., and Heyden, Y.V. (2007). Near-infrared spectroscopy applications in pharmaceutical analysis. Talanta 72: 865–883. Roggo, Y., Chalus, P., Maurer, L. et al. (2007). A review of near infrared spectroscopy and chemometrics in pharmaceutical technologies. J. Pharm. Biomed. Anal. 44: 683–700. Li, W., Bagnol, L., Berman, M. et al. (2009). Applications of NIR in early stage formulation development. Part II. Content uniformity evaluation of low dose tablets by principal component analysis. Int. J. Pharm. 380: 49–54. Benedetti, C., Abatzoglou, N., Simard, J.S. et al. (2007). Cohesive, multicomponent, dense powder flow characterization by NIR. Int. J. Pharm. 336: 292–301. Paris, I., Janoly-Dumenil, A., Paci, A. et al. (2006). Near infrared spectroscopy and process analytical technology to master the process of busulfan paediatric capsules in a university hospital. J. Pharm. Biomed. Anal. 41: 1171–1178. Tabasi, S.H., Moolchandani, V., Fahmy, R., and Hoag, S.W. (2009). Sustained release dosage forms dissolution behavior prediction: a study of matrix tablets using NIR spectroscopy. Int. J. Pharm. 382: 1–6. Undey, C., Ertunc, S., Mistretta, T., and Looze, B. (2010). Applied advanced process analytics in biopharmaceutical manufacturing: challenges and prospects in real-time monitoring and control. J. Process Control 20: 1009–1018. Teixeira, A.P., Oliveira, R., Alves, P.M., and Carrondo, M.J.T. (2009). Advances in on-line monitoring and control of mammalian cell cultures: supporting the PAT initiative. Biotechnol. Adv. 27: 726–732. Boudreau, M.A. and McMillan, G.K. (2007). New Direction in Bioprocess Modeling and Control: Maximizing Process Analytical Technology Benefits. ISA - The Instrumentation, System and Automation Society. Kourti, T. (2006). The process analytical technology initiative and multivariate process analysis, monitoring and control. Anal. Bioanal. Chem. 384: 1043–1048. Whitford, W. and Julien, C. (2007). Analytical technology and PAT. Bioproc. Int. 5: S32–S41. Arnold, S.A., Gaensakoo, R., Harvey, L.M., and McNeil, B. (2002). Use of at-line and in-situ near-infrared spectroscopy to monitor biomass in an industrial fed-batch Escherichia coli process. Biotechnol. Bioeng. 80: 405–413. Arnold, S.A., Crowley, J., Woods, N. et al. (2003). In-situ infrared spectroscopy to monitor analytes in mammalian cell cultivation. Biotechnol. Bioeng. 84: 13–19.

References

19 Card, C., Hunsaker, B., Smith, T., and Hirsch, J. (2008). Near-infrared spec-

20

21

22

23

24

25

26

27

28 29

30

31

32

33 34

troscopy for rapid, simultaneous monitoring of multiple components in mammalian cell culture. Bioprocess Int. 6: 58–67. Harthun, S., Matischak, K., and Friedl, P. (1998). Simultaneous prediction of human antithrombin III and main metabolites in animal cell culture processes by near-infrared spectroscopy. Biotechnol. Tech. 12: 393–398. Holm-Nielsen, J.B., Lomborg, C.J., Oleskowicz-Popiel, P., and Esbensen, K.H. (2008). On-line near infrared monitoring of glycerol-boosted anaerobic digestion processes: evaluation of process analytical technologies. Biotechnol. Bioeng. 99: 302–313. Navratil, M., Norberg, A., Lembren, L., and Mandenius, C.F. (2005). On-line multi-analyzer monitoring of biomass, glucose and acetate for growth rate control of a Vibrio cholerae fed-batch cultivation. J. Biotechnol. 115: 67–79. Triadaphillou, S., Martin, E., Montague, G. et al. (2007). Fermentation process tracking through enhanced spectral calibration modeling. Biotechnol. Bioeng. 97: 554–567. Kirdar, A.I., Chen, G., Weidner, J., and Rathore, A.S. (2010). Application of near-infrared (NIR) spectroscopy for screening of raw materials used in the cell culture medium for the production of a recombinant therapeutic protein. Biotechnol. Progr. 26: 527–531. Lee, H., Christie, A., Liu, J., and Yoon, S. (2012). Estimation of raw material performance in mammalian cell culture using near infrared spectra combined with chemometrics approaches. Biotechnol. Progr. 28 (3): 824–832. Riley, M.R., Crider, H.M., Nite, M.E. et al. (2001). Simultaneous measurement of 19 components in serum-containing animal cell culture media by Fourier transform near-infrared spectroscopy. Biotechnol. Progr. 17: 376–378. Roychoudhury, P., O’Kennedy, R., McNeil, B., and Harvey, L.M. (2007). Multiplexing fibre optic near infrared (NIR) spectroscopy as an emerging technology to monitor industrial bioprocesses. Anal. Chim. Acta 590: 110–117. Reich, G. (2005). Near-infrared spectroscopy and imaging: basic principles and pharmaceutical applications. Adv. Drug Delivery Rev. 57: 1109–1143. Armenta, S., Garrugues, S., Guardia, M., and Rondeau, P. (2005). Attenuated total reflectance-Fourier transform infrared analysis of the fermentation process of pineapple. Anal. Chim. Acta 545: 99–106. Gordon, S.H., Whatel, S.C., Wheeler, B.C., and James, C. (1993). Multivariate FTIR analysis of substrates for protein, polysaccharide, lipid and microbe content: potential for solid-state fermentations. Biotechnol. Adv. 11: 665–675. Kornmann, H., Valentinotti, S., Duboc, P. et al. (2004). Monitoring and control of Gluconacetobacter xylinus fed-batch cultures using in situ mid-IR spectroscopy. J. Biotechnol. 113: 231–245. Rhiel, M., Ducommun, P., Bolzonella, I. et al. (2002). Real-time in situ monitoring of freely suspended and immobilized cell cultures based on mid-infrared spectroscopic measurements. Biotechnol. Bioeng. 77: 174–185. Grob, P.C. and Zeppezauer, M. (2010). Infrared spectroscopy for biopharmaceutical protein analysis. J. Pharm. Biomed. Anal. 53: 29–36. Roychoudhury, P., Harvey, L.M., and McNeil, B. (2006). The potential of mid infrared spectroscopy (MIRS) for real time bioprocess monitoring. Anal. Chim. Acta 571: 159–166.

385

386

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

35 Mungikar, A. and Kamat, M. (2010). Use of in-line Raman spectroscopy as a

36

37

38

39

40

41

42

43

44

45 46 47 48

49

50

non-destructive and rapid analytical technique to monitor aggregation of a therapeutic protein. Am. Pharm. Rev. 13 (7): 78–83. Shaw, A.D., Kaderbhai, N., Jones, A. et al. (1999). Noninvasive, on-line monitoring of biotransformation by yeast of glucose to ethanol using dispersive Raman spectroscopy and chemometrics. Appl. Spectrosc. 53: 1419–1428. Sivakesava, S., Irudayaraj, J., and Demirci, A. (2001). Monitoring a bioprocess for ethanol production using FT-MIR and FT-Raman spectroscopy. J. Ind. Microbiol. Biotechnol. 26: 185–190. Li, B., Ryan, P.W., Ray, B.H. et al. (2010). Rapid characterization and quality control of complex cell culture media solutions using Raman spectroscopy and chemometrics. Biotechnol. Bioeng. 107: 290–301. Short, K.W., Carpenter, S., Freyer, J.P., and Mourant, J.R. (2005). Raman spectroscopy detects biochemical changes due to proliferation in mammalian cell cultures. Biophys. J. 88: 4274–4288. Gryniewicz, C.M. and Kauffman, J.K. (2008). Multivariate calibration of covalent aggregate fraction to the Raman spectrum of regular human insulin. J. Pharm. Sci. 97: 3727–3734. Hedoux, A. and Ionov, R. (2006). Evidence of a two-stage thermal denaturation process in lysozyme: a Raman scattering and differential scanning calorimetric investigation. J. Chem. Phys. 124: 14703–14709. Sanes, S., Wong, R., and Hsu, C. (2004). Raman spectroscopic characterization of drying-induced structural changes in a therapeutic antibody: correlating structural changes with long-term stability. J. Pharm. Sci. 93: 1005–1018. Asher, S.A., Ianoul, A., Mix, G. et al. (2001). Dihedral angle dependence of Amide III vibration: a unique sensitive UV resonance Raman secondary structure probe. J. Am. Chem. Soc. 123: 11775–11781. Chi, Z.H. and Asher, S.A. (1999). Ultraviolet resonance Raman examination of horse apomyoglobin acid unfolding intermediates. Biochemistry 38: 8196–8203. Huang, K., Dong, J., Phillips, N.B. et al. (2005). Pro-insulin is refractory to protein fibrillation. J. Biol. Chem. 280: 42345–42355. Zheng, R., Zheng, X.J., Dong, J., and Carey, P.R. (2004). Proteins can covert to β-sheet in single crystals. Protein Sci. 13: 1288–1294. Hannum, C.H., Wilcox, C.J., Arend, W.P. et al. (1990). Interleukin-1 receptor antagonist activity of a human interleukin-1 inhibitor. Nature 343: 336–340. Wen, Z.Q., Cao, X.L., and Vance, A. (2008). Conformation and side chains environments of recombinant human interleukin-1 receptor antagonist (rh-IL-1ra) probed by Raman, Raman optical activity, and UV-resonance Raman spectroscopy. J. Pharm. Sci. 97: 2228–2241. Raso, S.W., Clark, P.C., Haase-Pettingell, C. et al. (2001). Distinct cystein sulfhydryl environment detected by analysis of Raman S-H markers of Cys→Ser mutant proteins. J. Mol. Biol. 307: 899–911. Fagnano, C., Fini, G., and Torreggiani, A. (1995). Raman spectroscopic study of the avidin-biotin complex. J. Raman Spectrosc. 26: 991–995.

References

51 Li, T., Talvenheimo, L., Zeni, L. et al. (2002). Changes in protein conforma-

52 53

54

55

56

57

58

59

60

61

62

63

64

tion and dynamics upon formation of brain-derived neutrophic factor and its receptor. Biopolymers 67: 10–19. Wen, Z.Q. (2007). Raman spectroscopy of protein pharmaceuticals. J. Pharm. Sci. 96: 2861–2878. Teixeira, A.P., Portugal, C., Carinhas, N. et al. (2009). In situ 2D fluorometry and chemometric monitoring of mammalian cell cultures. Biotechnol. Bioeng. 102: 1098–1106. Boehl, D., Solle, D., Hitzmann, B., and Scheper, T. (2003). Chemometric modelling with two-dimensional fluorescence data for Claviceps purpurea bioprocess characterization. J. Biotechnol. 105: 179–188. Hisiger, S. and Jolicoeur, M. (2005). A multiwavelength fluorescence probe: is one probe capable for on-line monitoring of recombinant protein production and biomass activity? J. Biotechnol. 117: 325–336. Ryan, P.W., Li, B., Shanahan, M. et al. (2010). Prediction of cell culture media performance using fluorescence spectroscopy. Anal. Chem. 82: 1311–1317. Kurokawa, H., Park, Y.S., Iijima, S., and Kobayashi, T. (1994). Growth characteristics in fed-batch culture of hybridoma cells with control of glucose and glutamine concentrations. Biotechnol. Bioeng. 44: 95–103. Ozturk, S.S., Thrift, J.C., Blackie, J.D., and Naveh, D. (1995). Real-time monitoring of protein secretion in mammalian cell fermentation: Measurement of monoclonal antibodies using a computer-controlled HPLC system (BioCad/RPM). Biotechnol. Bioeng. 48: 201–206. Larson, T.M., Gawlitzek, M., Evans, H. et al. (2002). Chemometric evaluation of on-line high-pressure liquid chromatography in mammalian cell cultures: analysis of amino acids and glucose. Biotechnol. Bioeng. 77: 553–563. Anumula, K.R. (2006). Advances in fluorescence derivatization methods for high-performance liquid chromatography analysis of glycoprotein carbohydrates. Anal. Biochem. 350: 1–23. Hoffmann, H. (2000). Analytical methods and stability testing of biopharmaceuticals. In: Drugs and the Pharmaceutical Science: Protein Formulation and Delivery (ed. E.J. McNally), 71–110. New York, NY: Marcel Dekker, Inc. Vuleti´c, M., Cindri´c, M., and Koružnjak, J.D. (2005). Identification of unknown impurities in simvastatin substance and tablets by liquid chromatography/tandem mass spectrometry. J Pharm Biomed Anal. 37 (4): 715–721. Bernardo, K., Pakulat, N., Macht, M. et al. (2002). Identification and discrimination of Staphylococcus aureus strains using matrix-assisted laser desorption/ionization-time of flight mass spectrometry. Protemics 2: 747–754. Dally, J.E., Gorniak, J., Bowie, R., and Bentzley, C.M. (2003). Quantitation of underivatized free amino acids in mammalian cell culture media using matrix assisted laser desorption ionization time-of-flight mass spectrometry. Anal. Chem. 75: 5046–5053.

387

388

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

65 Villanueva, J., Canals, F., Querol, E., and Aviles, F.X. (2001). Monitoring the

66

67

68

69

70 71

72 73

74

75

76

77

78

79

expression and purification of recombinant proteins by MALDI-TOF mass spectrometry. Enzyme Microb. Technol. 29: 99–103. Shen, C.F., Hawari, J., and Kamen, A. (2004). Micro-quantitation of lipids in serum-free cell culture media: a critical aspect is the minimization of interference from medium components and chemical reagents. J. Chromatogr. B 810: 119–127. Cox, K.M., Sterling, J.D., Regan, J.T. et al. (2006). Glycan optimization of a human monoclonal antibody in the aquatic plant Lemma minor. Nat. Biotechnol. 24: 1591–1597. Greer, F. and Reason, A. (2002). Post-translational modification of biopharmaceuticals - a challenge for analytical characterization. European BioPharmaceutical Review Summer, 106–111. Wu, S.-L., Jardine, I., Hancock, W.S., and Karger, B.L. (2004). A new and sensitive on-line liquid chromatography/mass spectrometric approach for top-down protein analysis: the comprehensive analysis of human growth hormone in an E. coli lysate using hybrid linear ion trap/Fourier transform ion cyclotron resonance mass spectrometer. Rapid Commun. Mass Spectrom. 18: 2201–2207. Greer, F. (2003). Biopharmaceutical characterisation - considering key questions. BioPharmaceutical Review Summer, 64–67. Cannizzaro, C., Gugerli, R., Marison, I., and von Stockar, U. (2003). On-line biomass monitoring of CHI perfusion culture with scanning dielectric spectroscopy. Biotechnol. Bioeng. 84: 597–610. Noll, T. and Biselli, M. (1998). Dielectric spectroscopy in the cultivation of suspended and immobilized hybridoma cells. J. Biotechnol. 63: 187–198. Zeiser, A., Bedard, C., Voyer, R. et al. (1999). On-line monitoring of the progress of infection in St-9 insect cell cultures using relative permittivity measurements. Biotechnol. Bioeng. 63: 122–126. Bachinger, T., Riese, U., Eriksson, R., and Mandenius, C.F. (2000). Monitoring cellular state transitions in a production-scale CHO-cell process using an electric nose. J. Biotechnol. 76: 61–71. Bachinger, T., Riese, U., Eriksson, R.K., and Mandenius, C.F. (2002). Gas sensor arrays for early detection of infection in mammalian cell culture. Biosens. Bioelectron. 17: 395–403. Kreij, K., Mandenius, C.F., Clemente, J.J. et al. (2005). On-line detection of microbial contaminations in animal cell reactor cultures using an electric nose devise. Cytotechnology 48: 41–58. Abu-Absi, N.R., Zamamiri, A., Kacmar, J. et al. (2003). Automated flow cytometry for acquisition of time-dependent population data. Cytometry, Part A 51A: 87–96. Henry, O., Kamen, A., and Perrier, M. (2007). Monitoring of physiological state of mammalian cell perfusion process by on-line estimation of intracellular fluxes. J. Process Control 17: 241–251. Goudar, C., Biener, R., Boisart, C. et al. (2010). Metabolic flux analysis of CHO cells in perfusion culture by metabolite balancing and 2D [13 C, 1 H] COSY NMR spectroscopy. Metab. Eng. 12: 138–149.

References

80 Luo, Y. and Chen, G. (2007). Combined approach of NMR and chemo-

81

82

83

84 85 86

87

88

89

90

91

92

93

94

metrics for screening peptones used in the cell culture medium for the production of a recombinant therapeutic protein. Biotechnol. Bioeng. 97: 1654–1659. Lee, H., Christie, A., Starkey, J. et al. (2015). Intracellular metabolic flux analysis of CHO cells supplemented with wheat hydrolysates for improved mAb production and cell-growth. J. Chem. Technol. Biotechnol. https://doi .org/10.1002/jctb.4523. Rudt, M., Briskot, T., and Hubbuch, J. (2017). Advances in downstream processing of biologics – spectroscopy: an emerging process analytical technology. J. Chromatogr. A 1490: 2–9. Singh, R., Gernaey, K.V., and Gani, R. (2009). ICAS-PAT: a software for design, analysis and validation of PAT systems. Comput. Chem. Eng. 34: 1108–1136. Oliveira, R. (2004). Combining first principles modelling and artificial neural networks: a general framework. Comput. Chem. Eng. 28: 755–766. Psichogios, D.C. and Ungar, L.H. (1992). A hybrid neural network- first principle approach to process modeling. AlChE J. 38: 1499–1511. Mercier, S.M., Diepenbroek, B., Wijffels, R.H., and Streefland, M. (2014). Multivariate PAT solutions for biopharmaceutical cultivation: current progress and limitations. Trends Biotechnol. 32 (6): 329–336. CMC Biotech Working Group (2012). A-Mab: a case study in bioprocess development. http://c.ymcdn.com/sites/www.casss.org/resource/resmgr/ imported/A-Mab_Case_Study_Version_2-1.pdf (accessed 30 October 2009). CMC Vaccines Working Group (2012). A-VAX: applying quality by design to vaccines. http://qbdworks.com/wp-content/uploads/2014/06/a-vax-applyingqbd-to-vaccines.pdf (accessed May 2012). Bhatia, H., Read, E., Agarabi, C. et al. (2016). A design space exploration for control of critical quality attributes of mAb. Int. J. Pharm. 512 (1): 242–52. PMID: 27575657. Kamga, M., Lee, H., Liu, J., and Yoon, S. (2013). Quantification of protein mixture in chromatographic separation using multi-wavelength UV spectra. Biotechnol. Progr. 29 (3): 664–671. Bhatia, H., Mehdizadeh, H., Drapeau, D. et al. (2017). In-line monitoring of amino acids in mammalian cell cultures using raman spectroscopy and multivariate chemometrics models. Eng. Life Sci. 18: 55–61. Trunfio, N., Lee, H., Starkey, J. et al. (2017). Characterization of mammalian cell culture raw materials by combining spectroscopy and chemometrics. Biotechnol. Progr. https://doi.org/10.1002/btpr.2480. Reimonn, T.M., Park, S.Y., Agarabi, C.D. et al. (2016). Effect of amino acid supplementation on titer and glycosylation distribution in hybridoma cell cultures-systems biology-based interpretation using genome-scale metabolic flux balance model and multivariate data analysis. Biotechnol. Progr. 32 (5): 1163–1173. PMID: 27452371. Read, E.K., Bradley, S.A., Smitka, T.A. et al. (2013). Fermentanomics informed amino acid supplementation of an antibody producing mammalian cell culture. Biotechnol Prog. 29: 745–753.

389

390

15 Process Analytical Technology and Quality by Design for Animal Cell Culture

95 Park, S., Reimonn, T.M., Agarabi, C.D. et al. (2018). Metabolic responses and

96

97

98

99

100

101

102

103

104 105

106

107

108

pathway changes of mammalian cells under different culture conditions with media supplementations. Biotechnol. Prog. 34: 793–805. Sha, S., Agarabi, C., Brorson, K., and Yoon, S. (2016). N-glycosylation design and control of therapeutic monoclonal antibodies. Trends Biotechnol. https:// doi.org/10.1016/j.tibtech.2016.02.013. Sou, S.N., Sellick, C., Lee, K. et al. (2015). How does mild hypothermia affect monoclonal antibody glycosylation. Biotechnol. Bioeng. 112: 1165–1176. Huang, Z., Lee, D.Y., and Yoon, S. (2017). Quantitative intracellular flux modeling and applications in biotherapeutic development and production using CHO cell cultures. Biotechnol. Bioeng. 114: 2717–2728. Lewis, A.M., Abu-Absi, N.R., Borys, M.C., and Li, Z.J. (2016). The use of ’Omics technology to rationally improve industrial mammalian cell line performance. Biotechnol. Bioeng. 113 (1): 26–38. Schaub, J., Clemens, C., Schorn, P. et al. (2010). CHO gene expression profiling in biopharmaceutical process analysis and design. Biotechnol. Bioeng. 105 (2): 431–438. Hefzi, H., Ang, K.S., Hanscho, M. et al. (2016). A consensus genome-scale reconstruction of Chinese hamster ovary cell metabolism. Cell Syst. 3 (5): 434–443. Yusufi, F.N.K., Lakshmanan, M., Ho, Y.S. et al. (2017). Mammalian systems biotechnology reveals global cellular adaptations in a recombinant CHO cell line. Cell Syst. 4 (5): 530–542.e6. Gao, Y., Ray, S., Dai, S. et al. (2016). Combined metabolomics and proteomics reveals hypoxia as a cause of lower productivity on scale-up to a 5000-liter CHO bioprocess. Biotechnol. J. 11 (9): 1190–1200. Costa, R.S., Hartmann, A., and Vinga, S. (2016). Kinetic modeling of cell metabolism for microbial production. J. Biotechnol. 219: 126–141. Khodayari, A. and Maranas, C.D. (2016). A genome-scale Escherichia coli kinetic metabolic model k-ecoli457 satisfying flux data for multiple mutant strains. Nat. Commun. 7: 13806. Khodayari, A., Zomorrodi, A.R., Liao, J.C., and Maranas, C.D. (2014). A kinetic model of Escherichia coli core metabolism satisfying multiple sets of mutant flux data. Metab. Eng. 25: 50–62. Le, H., Vishwanathan, N., Jacob, N.M. et al. (2015). Cell line development for biomanufacturing processes: recent advances and an outlook. Biotechnol. Lett. 37 (8): 1553–1564. Konold, P., Woolfenden, R., Undey, C., and Rathore, A.S. (2009). Monitoring of biopharmaceutical processes: present and future approaches. Biopharm. Int. 22: 26–39.

391

16 Development and Qualification of a Cell Culture Scale-Down Model Sarwat Khattak and Valerie Pferdeort Biogen, Cell Culture Development, 5000 Davis Drive, RTP, Durham, NC 27516, USA

16.1 Purpose of the Scale-Down Model According to International Council for Harmonisation (ICH) of technical requirements for pharmaceuticals for human use Q11 Step 4, developing a scale-down model (SDM) supports process development studies and incorporate scale effects while sufficiently representing and predicting performance at the final commercial scale [1]. Understanding and quantifying how well the SDM represents the commercial-scale process is critical in determining what conclusions can be drawn from small-scale experiments [2]. The SDM should address raw materials, components, cell source, equipment, and engineering operational parameters [2]. 16.1.1

Development Challenges

Cell line engineering and cell culture process development are done at bench/ microscale to conserve on costs and to remain efficient [3–6]. Thousands of clones may be produced and evaluated before choosing the ideal/best clone [6, 7]. Once candidate clones are chosen, they are subjected to an environment that should predict or be like the final manufacturing scale [4, 8–12]. To achieve an appropriate SDM environment, a variety of methods and techniques have been applied and evaluated [13–25]. Iterative methods may be used to demonstrate comparability and the final SDM should be validated for the chosen process [20, 24, 26, 27]. The challenges in creating an appropriate SDM are multiple: mass transfer and mixing within the bioreactor [8, 12, 28–36], the sampling techniques and online/offline monitoring capabilities [37–41], control strategies, analytical methods, and purification. The SDM is critical for process development studies and ensuring the benchscale process is predictive of the final commercial scale process. The model should address scale effects and be proven to sufficiently represent the commercial process through univariate, multivariate, and equivalency analysis. A robust and scientifically rigorous model should result in accurate product quality predictions and produce operating conditions across scales [1]. Scale-down limitations should be recognized and incorporated into the risk assessment. Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

392

16 Development and Qualification of a Cell Culture Scale-Down Model

16.2 Types of Scale-Down Models Essentially, an SDM addresses agitation and aeration. One of the simplest methods is geometric scaling using equal power per volume (P/V ) [42]. Agitation can be determined by P/V and further refinement can be achieved by understanding the oxygen mass transfer rate and oxygen mass transfer coefficient [43, 44]. Aeration in the bioreactor is determined by the sparger design and gas flow rate alignment between scales [28, 45, 46]. 16.2.1

Power/Volume (P/V) and Air velocity

The P/V ratio can be measured and maintained equivalent at different scales of a stirred tank bioreactor. For a SDM, a constant P/V between scales will determine the agitation speed given a working volume and known impeller power number. The equation for P/V is P0 ⋅ 𝜌 ⋅ N 3 ⋅ D5i P = (16.1) V V where P0 is impeller power number, 𝜌 is density (kg/m3 ), N is impeller rotation (s− 1), Di is impeller diameter (m), and V is culture volume (m3 ). This P/V calculation assumes an ungassed environment. P/V is a function of agitation, working volume, impeller size, and impeller type. The volumetric gas flow rate expressed as the volumetric flow rate at standard conditions per volume in the bioreactor (VVM) is determined by the following equation: Qg VVM = (16.2) V where Qg is the gas flow rate (m3 /min) and V is the volume of cell culture (m3 ) in the bioreactor. The air sparge and maximum oxygen sparge can be calculated using a constant gas velocity (VVM) to match across scales. When P/V and VVM are used together, a good starting point for a SDM can be achieved. 16.2.2

Oxygen Transfer Coefficient (kL a)

Oxygen plays a critical role in ensuring cell viability and growth and is dependent on the bioreactor parameters and cell culture media. The oxygen transfer coefficient, k L a, is a function of the media viscosity, media density, bioreactor geometry, and agitation/sparge. The k L a in a bioreactor can be calculated experimentally in a bioreactor outfitted with a dissolved oxygen (DO) probe and the appropriate media. Typically, a salt buffer solution is used at large scale instead of cell culture media to minimize costs and test for k L a [31, 32, 35, 47]. By varying agitation with a fixed sparge rate, k L a can be determined under different conditions and then applied to a SDM to represent the k L a seen at large scale. The impeller pumping direction can also impact k L a and should be evaluated as part of the k L a determination. Empirically, k L a can be determined by the Van’t Riet equation [48]: ( )𝛼 P kL a = 0.21 ⋅ ⋅ Gg𝛽 (16.3) V

16.2 Types of Scale-Down Models

P/V 187 W/m3

45

15 μm frit

kLa (1 h−1) 37 °C/1 atm

40 35 30

50 μm frit

25 20

100 μm frit

15 10

Drilled hole

5 0 0.04 0.02 Gas velocity (VVM)

0

0.06

Figure 16.1 kL a as a function of sparger design and gas velocity (VVM).

where k L a (h−1 ), P/V (W/m3 ), and Gg (air sparge rate, ml/min), and 𝛼 = 0.4, and 𝛽 = 0.5. Both 𝛼 and 𝛽 can also be determined iteratively to achieve more accurate results for a specified system. The type of sparger used will also impact the k L a. At a fixed P/V with varying gas flow rates, the k L a will vary according to sparger design. The sparge hole size, hole number, weep holes (drain holes), and angle all play a critical role in reducing bubble damage, increasing bubble homogeneity, and improving or aligning k L a across scales [23, 35]. See Figure 16.1 for an example of how k L a is modulated with different spargers over a range of gas velocities and fixed P/V . 16.2.3

Gas Entrance Velocity (GEV)

The gas entrance velocity (GEV) (m/s) is a function of gas velocity (V g , standard liter per minute [slpm]), cross-sectional area of each hole in the sparger (Ah , m2), and the number of holes (n) in the sparger. 0.001 (16.4) 60 × n Cell damage can be caused by high GEV [8, 49]. The GEV does not need to be equivalent in the SDM but it should be evaluated to determine the upper limit. High sparge rates in the SDM may correlate with unacceptable GEVs at commercial scale. GEV = Vg × Ah

16.2.4

Oxygen Transfer Rate (OTR)

If P/V and k L a are matched between scales but result in cell damage or is not representative of the commercial process, the next strategy should be to used matching oxygen transfer rate (OTR). The assumption is that matching OTRs should result in lower gas flow rates and less impact to shear-sensitive cells. The OTR is a

393

20 000 l 1:1.55

1000 l 1:1.08 50 l 1:1.20

Figure 16.2 Aspect ratio (H/D) and impeller placement in a bioreactor.

16.3 Evaluation of a Scale-Down Model

function of the k L a and the oxygen driving force ΔC L (the difference in saturated level of oxygen and the measured level of oxygen) in the media [31, 32, 45]: OTR = kL a ⋅ ΔCL

(16.5)

OTR can be modeled at small scale by varying the agitation rate and/or the ratio of oxygen to air in the sparged gas. Typically, this requires oxygen enrichment because of cell sensitivity to the sparger design and/or agitation speed. Using an OTR model could minimize the sparge rate, which is beneficial to shear-sensitive cells but usually requires a higher P/V at small scale and could also lead to higher pCO2 levels. 16.2.5

Model Refinement Workflow

Further refinement of the SDM should include an assessment of the bioreactor geometry, online pH profile, online DO profile, pCO2 profile, and base utilization. These profiles may still differ between scales but can help discern if the OTR or k L a model needs further modifications of the sparger, gas flow rates, and/or impeller/bioreactor design. The SDM should have geometric similarity to the final manufacturing scale including aspect ratio (height to diameter) and impeller physical characteristics (Figure 16.2). If the final scale has dual impellers, the pilot SDM should have dual impellers to mimic the agitation and coverage of each impeller during a production run.

16.3 Evaluation of a Scale-Down Model The purpose of evaluation of a SDM is to verify that the performance of the SDM for a given process step is representative of the commercial-scale process and therefore applicable to the intended use. The degree of verification desired will determine the exact methodology of evaluation employed. The evaluation is focused on output parameters for the process to confirm that the cell culture and resulting product quality match between scales, as opposed to controlled input parameters of how the process is operated. 16.3.1

Univariate Analysis

Univariate evaluation involves evaluation of single-variable parameters, one-at-a-time, often over the time course of the process. This approach does not allow for an analysis of potential interactions of performance outputs but does provide a visual inspection of the time-based progression of the variable. An example is shown in Figure 16.3. Univariate analysis is often a straightforward evaluation of means and standard deviations of the data sets from the two scales. As a result, the strength of the conclusion is related to the ability of the data sets to accurately reflect the true population mean and variability of the process at each scale. Therefore, data sets of at least five representative runs at each scale should be used, preferably with 10–20 runs at each scale if possible.

395

16 Development and Qualification of a Cell Culture Scale-Down Model

Figure 16.3 Example of a univariate time course analysis chart.

Large scale Scale down model

2.0

1.5 Parameter

396

1.0

0.5

0.0 0

4 2 Culture duration

6

As this type of analysis is most useful in consideration of time course parameters, it is most often applied to the evaluation of cell culture parameters such as viable cell density, viability, pCO2 profiles, pH profiles, glucose, lactate, and other key metabolites or nutrients concentrations. Gross offsets observed at this preliminary phase may be used to drive control changes or even changes to the scaling parameters used in refining the SDM. 16.3.2

Multivariate Analysis

16.3.2.1

Statistical Background

Multivariate analysis (MVA) allows the comparison of large, complex data sets across scales and is often used in cell culture processes [13, 23, 24, 50]. By applying matrix algebra, MVA can reduce multidimensional data sets into a few uncorrelated variables. The resulting uncorrelated variables are optimized through least-squares regression so that as much of the variance of the original data sets are represented in the transformed variables [51]. Software packages such as Simca (Umetrics, Sartorius Stedim, Sweden) are available to execute MVA. 16.3.2.2

Qualification Data Set

The analysis to evaluate alignment between the SDM and large-scale performance requires a qualification data set. Generally, the manufacturing-scale process runs should be used as the training set, and the SDM runs are used as the projected model set. If fewer than eight manufacturing batches are available to develop the training set ellipse, then representative pilot scale batches can be used to augment the data set. If fewer than eight runs are still available from the large-scale process, then the SDM runs can be used as the training set and the manufacturing and pilot scale runs evaluated as the model set. The recommended parameters to include should define a cell culture process in terms of growth, metabolism, productivity, and product quality. Because of the complex interactions in biological systems, parameters may be related and correlate with each other. The parameters that have a strong linear correlation

16.3 Evaluation of a Scale-Down Model

(collinearity) should be eliminated to prevent overweighting the model. For example, base addition may or may not be collinear with osmolality. Strong collinearity between independent variables can make it impossible to determine which of the variables account for the variance in the dependent variable. 16.3.2.3

Observation Level Analysis

Using a partial least squares (PLSs) model built from the large-scale data set, a time-based progression of the scale-down batches can be evaluated in the batch control charts of the observation level analysis. This analysis determines average ±3 standard deviation limits of the large-scale process and applies the predictive scores of the SDM runs over the time course (see Figure 16.4). This high-level evaluation is most useful as an initial assessment of alignment and troubleshooting of basic consistency of behavior between the scales. 16.3.2.4

Batch-Level Analysis

Batch-level analysis is useful to assess scalability of the SDM and large-scale data sets. Using principal component analysis (PCA) of the training set, a 95% confidence ellipse established and the model set is then projected on to evaluation alignment (see Figure 16.5). In the score scatter plot, each data point represents a single scale-down batch in the model set. As shown in Figure 16.5, if all scale-down batches fall within the 95% confidence ellipse of the large-scale training set, then there is no statistically significant difference between the data sets. SDM data sets from different model scaling criteria will overlay the training set ellipse differently. If a model set falls outside of the ellipse for any or all batches, then further consideration should be given to the suitability of the selected scaling criteria. More information regarding associated parameters related to an observed offset is provided in the score contribution plots in the following section. 4 3 2

+3SD

t[1]

1 0 –1 –2

–3SD

–3 –4 –5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 Time (shifted and normalized)

Figure 16.4 Example observation level batch control chart for MVA.

397

16 Development and Qualification of a Cell Culture Scale-Down Model

8 6 CSS16-04 C1

4 tPS[2]

CSS16-07 C2

CSS16-07 B3 RSS16-01 V5 CSS16-07 C3 CSS16-05 B3 CSS16-07 C4 CSS16-02 B6 CSS16-04 B7 RSS16-01 V5 CSS16-04 B1 RSS16-01 V5 CSS16-02 C2 CSS16-05 B1 RSS16-01 V4 CSS16-05 B6 RSS16-02 V5 CSS16-02 C1 CSS16-02 B7 RSS16-01 V5 CSS16-05 B7 RSS16-04 V5 CSS16-02 B1 CSS16-01 B3 CSS16-07 C2 C1 CSS16-04

2 0 –2 –4

CSS16-01 B1 CSS16-01 C2

–6

RSS16-07 V4

–8 –10 –10

–8

–6

–2

–4

0 tPS[1]

R2x[1] = 0.218 Ellipse: Hotelling’s T2PS (95%)

2

4

6

8

R2x[2] = 0.151

Figure 16.5 Example of score scatter plot of batch level analysis plot for MVA.

16.3.2.5

Scores Contribution Plots

Score contribution plot is a useful tool that provides insight as to which parameters contribute the largest differences between the training and model data sets. An example score contribution plot is provided in Figure 16.6, where the various process parameters included in the data sets are along the x-axis (not listed explicitly) and the contribution score of each is shown on the y-axis. If differences in the data sets are observed at the batch level in the score scatter plot, then the

Score contribution (Batch N - model average)

398

4 3 2 1 0 –1 –2 –3 –4 –5 10

20

30

40

50

60 70 80 90 Model parameter

Figure 16.6 Example scores contribution plot for MVA.

100

110

120 130

16.3 Evaluation of a Scale-Down Model

contribution plot may help to illuminate which process parameters are causing the misalignment between scales. Process parameters with the largest scores contribution differences can subsequently be targeted for further evaluation and refinement in the SDM. For example, if the largest contributions are coming from the later days of offline pH and pCO2 readings, then this may indicate a difference in gassing strategies between scales that needs to be rectified. As a rule of thumb, typically scores contribution differences with absolute values greater than or equal to 10 are typically targeted for further evaluation and refinement between the scales.. 16.3.3

Equivalence Testing

16.3.3.1

Statistical Background

Equivalence testing [52] is an evaluation of the difference in means between the historical, large-scale process (𝜇H ) and the new, SDM (𝜇N ), with a: ∣ 𝜇H − 𝜇N ∣< EAC

(16.6)

where the EAC (equivalence acceptance criterion) is a difference in means between the two scales that is of practical importance. The evaluation is typically carried out by determining the 90% confidence interval on 𝜇H – 𝜇N , and if that complete confidence interval falls within −EAC to +EAC, then average ¯ H) equivalence can be claimed. Based on the sample means of the historical (U ¯ N ) process data sets, average equivalence can also be demonstrated and new (U as follows: U H − U N + ME < EAC

(16.7)

U H − U N − ME > −EAC

(16.8)

where ME is the 90% margin of error for the confidence interval on 𝜇H – 𝜇N . Reorganizing Eqs. (16.7) and (16.8) yields U N − ME > U H − EAC

(16.9)

U N + ME < U H + EAC

(16.10)

The right-hand side of Eqs. (16.9) and (16.10) can be defined as the lower equivalence limit (LEL) and upper equivalence limit (UEL), respectively. Similarly, the left side of Eqs. (16.9) and (16.10) can be defined as the lower test limit (LTL) and upper test limit (UTL), respectively. As with the requirement that the complete 90% confidence interval around 𝜇H – 𝜇N must fall within −EAC to +EAC to prove average equivalency, so too must the UTL and LTL both fall between the UEL and LEL to prove equivalency. Graphically, this is illustrated in Figure 16.7. 16.3.3.2

Considerations for Evaluation and Test Data Sets

The analysis can be completed with the historical data set as either the large-scale or the SDM batches but is typically down with the large-scale process defined as the “historical” data set as this is the standard to which the SDM (i.e. “new” data set) needs to conform. A minimum of eight data points in each set is typically

399

16 Development and Qualification of a Cell Culture Scale-Down Model

1.1 UEL

Parameter U

400

0.9 UTL

UH UN

LTL 0.7 LEL

0.5 H1 H2 H3 H4 H5 H6 H7 H8 H9 H10H11H12H13H14H15H16

N1 N2 N3 N4 N5 N6 N7 N8

Lot

Figure 16.7 Graphical representation of average equivalence evaluation between historical large scale (H) and new SDM (N) batches.

generated to ensure adequate statistical power. In addition, parameters chosen are most usually for discrete process parameters, such as final harvest product quality attributes, because of the cumbersome process of compiling and evaluating multiple analysis charts for time course parameters. 16.3.3.3

Types of Analysis Outcomes

The conclusion from the comparison illustrated in example Figure 16.7 is that the data sets are equivalent, as the UTL and LTL fall within the UEL and LEL. Further, this is a special circumstance where the data sets are not only equivalent but there is no statistically significant difference between them at all as the LTL and UTL also encompass the historical mean. By contrast, Figure 16.8 illustrates a scenario where the UTL and LTL do not encompass the historical mean and thus indicates that a statistically significant offset does exist between the scales. However, as the UTL and LTL still both fall between the UEL and LEL, average equivalence can still be claimed based on the allowable offset (EAC) defined for the evaluation. It is worthwhile to note that this is true, despite the fact that two of the new data set data points fall outside of the UEL. As stated above, this evaluation is relevant to the averages for each data set and should not be applied to the individual data points as a tolerance interval approach would be. Lastly, Figure 16.9 provides an example of a process evaluation that would fail equivalence testing as the UTL falls above the UEL. The expansive distance between the UTL and LTL are due to higher levels of variability in the data for the new data set. As a result, process or analytical improvements aimed at process robustness and decreased variability in the SDM process data set may allow a smaller estimate for the margin of error for the confidence interval on 𝜇H – 𝜇N and therefore provide tighter test limits that fall within the equivalence acceptance criterion.

16.4 Conclusions and Perspectives

35 UEL UTL

Parameter U

UN 30 LTL UH

25

20

LEL

10 H1 H2 H3 H4 H5 H6 H7 H8 H9

N1 N2 N3 N4 N5 N6 N7 N8 N9 N10

Lot

Figure 16.8 Graphical representation of average equivalence evaluation with statistically significant offset between historical large scale (H) and new SDM (N) batches. 40 UTL

35

Parameter U

UEL UN 30 LTL

UH 20

LEL 10 H1 H2 H3 H4 H5 H6 H7 H8 H9

N1 N2 N3 N4 N5 N6 N7 N8 N9 N10

Lot

Figure 16.9 Graphical representation of average equivalence evaluation with nonequivalent data sets between historical large scale (H) and new SDM (N) batches.

16.4 Conclusions and Perspectives The creation and demonstration of an appropriate SDM is a regulatory requirement to support the development and characterization of large-scale cell culture processes. The mass transfer and mixing strategy chosen as the basis of the SDM can range from a simple power/volume ratio (P/V ) model or increasingly more complex models based on oxygen transfer coefficient (k L a) or OTR. The statistical approach to validate or verify the SDM can also range from simple (univariate analysis) to complex (MVA and/or equivalency testing). For

401

402

16 Development and Qualification of a Cell Culture Scale-Down Model

a complete understanding of the alignment and gaps in the SDM, all three statistical approaches may be applied for appropriate subsets of parameters. The complexity of the design and evaluation methods applied may be tailored to fit the needs of the process control approach employed.

References 1 Federal Register (2012). International Conference on Harmonisation;

Guidance on Q11 Development and Manufacture of Drug Substances. 2 Federal Register (2011). Guidance for Industry Process Validation: General

Principles and Practices. 3 Alves, C.S., Gilbert, A., Dalvi, S. et al. (2015). Integration of cell line and pro-

4

5 6

7

8

9

10

11

12

13

cess development to overcome the challenge of a difficult to express protein. Biotechnol. Progr. 31: 1201–1211. DeMaria, C.T., Cairns, V., Schwarz, C. et al. (2007). Accelerated clone selection for recombinant CHO cells using a FACS-based high-throughput screen. Biotechnol. Progr. 23: 465–472. Li, F., Vijayasankaran, N., Shen, A. et al. (2010). Cell culture processes for monoclonal antibody production. mAbs 2: 466–477. Omasa, T., Onitsuka, M., and Kim, W.D. (2010). Cell engineering and cultivation of Chinese hamster ovary (CHO) cells. Curr. Pharm. Biotechnol. 11: 233–240. Wright, C., Alves, C., Kshirsagar, R. et al. (2017). Leveraging a CHO cell line toolkit to accelerate biotherapeutics into the clinic. Biotechnol. Progr. 33: 1468–1475. Hu, W., Berdugo, C., and Chalmers, J.J. (2011). The potential of hydrodynamic damage to animal cells of industrial relevance: current understanding. Cytotechnology 63: 445–460. Legmann, R., Benoit, B., Fedechko, R.W. et al. (2011). A strategy for clone selection under different production conditions. Biotechnol. Progr. 27: 757–765. Matsunaga, N., Kano, K., Maki, Y., and Dobashi, T. (2009). Culture scale-up studies as seen from the viewpoint of oxygen supply and dissolved carbon dioxide stripping. J. Biosci. Bioeng. 107: 412–418. Prentice, H.L., Ehrenfels, B.N., and Sisk, W.P. (2007). Improving performance of mammalian cells in fed-batch processes through “bioreactor evolution”. Biotechnol. Progr. 23: 458–464. Zhou, H., Purdie, J., Wang, T., and Ouyang, A. (2010). pH measurement and a rational and practical pH control strategy for high throughput cell culture system. Biotechnol. Progr. 26: 872–880. Ahuja, S., Jain, S., and Ram, K. (2015). Application of multivariate analysis and mass transfer principles for refinement of a 3-L bioreactor scale-down model--when shake flasks mimic 15,000-L bioreactors better. Biotechnol. Progr. 31: 1370–1380.

References

14 Bareither, R., Bargh, N., Oakeshott, R. et al. (2013). Automated disposable

15

16

17

18

19

20

21

22

23

24

25

26

27

small scale reactor for high throughput bioprocess development: a proof of concept study. Biotechnol. Bioeng. 110: 3126–3138. Chen, A., Chitta, R., Chang, D., and Amanullah, A. (2009). Twenty-four well plate miniature bioreactor system as a scale-down model for cell culture process development. Biotechnol. Bioeng. 102: 148–160. Gomez, N., Ambhaikar, M., Zhang, L. et al. (2017). Analysis of Tubespins as a suitable scale-down model of bioreactors for high cell density CHO cell culture. Biotechnol. Progr. 33: 490–499. Hsu, W.T., Aulakh, R.P., Traul, D.L., and Yuk, I.H. (2012). Advanced microscale bioreactor system: a representative scale-down model for benchtop bioreactors. Cytotechnology 64: 667–678. Janakiraman, V., Kwiatkowski, C., Kshirsagar, R. et al. (2015). Application of high-throughput mini-bioreactor system for systematic scale-down modeling, process characterization, and control strategy development. Biotechnol. Progr. 31: 1623–1632. Legmann, R., Schreyer, H.B., Combs, R.G. et al. (2009). A predictive highthroughput scale-down model of monoclonal antibody production in CHO cells. Biotechnol. Bioeng. 104: 1107–1120. Li, F., Hashimura, Y., Pendleton, R. et al. (2006). A systematic approach for scale-down model development and characterization of commercial cell culture processes. Biotechnol. Progr. 22: 696–703. Nagashima, H., Watari, A., Shinoda, Y. et al. (2013). Application of a quality by design approach to the cell culture process of monoclonal antibody production, resulting in the establishment of a design space. J. Pharm. Sci. 102: 4274–4283. Sieck, J.B., Cordes, T., Budach, W.E. et al. (2013). Development of a scaledown model of hydrodynamic stress to study the performance of an industrial CHO cell line under simulated production scale bioreactor conditions. J. Biotechnol. 164: 41–49. Tescione, L., Lambropoulos, J., Paranandi, M.R. et al. (2015). Application of bioreactor design principles and multivariate analysis for development of cell culture scale down models. Biotechnol. Bioeng. 112: 84–97. Tsang, V.L., Wang, A.X., Yusuf-Makagiansar, H., and Ryll, T. (2014). Development of a scale down cell culture model using multivariate analysis as a qualification tool. Biotechnol. Progr. 30: 152–160. Xu, P., Clark, C., Ryder, T. et al. (2017). Characterization of TAP Ambr 250 disposable bioreactors, as a reliable scale-down model for biologics process development. Biotechnol. Progr. 33: 478–489. Joosten, C.E. (2006). BIOT 238-Establishment of a cell culture bioreactor scale-down model for process qualification studies. Abstr. Pap. Am. Chem. Soc. 232. Yeung, P.L., Okediadi, C., Lanter, P., and Figueroa, B. (2015). Development and qualification of cell culture bioreactor scale-down models. Abstr. Pap. Am. Chem. Soc. 249.

403

404

16 Development and Qualification of a Cell Culture Scale-Down Model

28 Betts, J.P.J., Warr, S.R.C., Finka, G.B. et al. (2014). Impact of aeration strate-

29

30

31 32 33

34

35

36

37

38

39

40

41

42 43

gies on fed-batch cell culture kinetics in a single-use 24-well miniature bioreactor. Biochem. Eng. J. 82: 105–116. Brunner, M., Braun, P., Doppler, P. et al. (2017). The impact of pH inhomogeneities on CHO cell physiology and fed-batch process performance two-compartment scale-down modelling and intracellular pH excursion. Biotechnol. J. 12: 201600633. Keane, J.T., Ryan, D., and Gray, P.P. (2003). Effect of shear stress on expression of a recombinant protein by Chinese hamster ovary cells. Biotechnol. Bioeng. 81: 211–220. Lavery, M. and Nienow, A.W. (1987). Oxygen transfer in animal cell culture medium. Biotechnol. Bioeng. 30: 368–373. Mooyman, J.G. (1987). Scaling oxygen mass transfer in agitated fermentors. Biotechnol. Bioeng. 29: 180–186. Neunstoecklin, B., Stettler, M., Solacroup, T. et al. (2015). Determination of the maximum operating range of hydrodynamic stress in mammalian cell culture. J. Biotechnol. 194: 100–109. Neunstoecklin, B., Villiger, T.K., Lucas, E. et al. (2016). Pilot-scale verification of maximum tolerable hydrodynamic stress for mammalian cell culture. Appl. Microbiol. Biotechnol. 100: 3489–3498. Nienow, A.W. (2015). Mass transfer and mixing across the scales in animal cell culture. In: Animal Cell Culture (ed. M. Al-Rubeai), 137–167. Cham: Springer International Publishing. Qian, Y., Xing, Z., Lee, S. et al. (2014). Hypoxia influences protein transport and epigenetic repression of CHO cell cultures in shake flasks. Biotechnol. J. 9: 1413–1424. Derfus, G.E., Abramzon, D., Tung, M. et al. (2010). Cell culture monitoring via an auto-sampler and an integrated multi-functional off-line analyzer. Biotechnol. Progr. 26: 284–292. Deshpande, R.R. and Heinzle, E. (2004). On-line oxygen uptake rate and culture viability measurement of animal cell culture using microplates with integrated oxygen sensors. Biotechnol. Lett. 26: 763–767. Leon, Z., Garcia-Canaveras, J.C., Donato, M.T., and Lahoz, A. (2013). Mammalian cell metabolomics: experimental design and sample preparation. Electrophoresis 34: 2762–2775. Li, B., Ray, B.H., Leister, K.J., and Ryder, A.G. (2013). Performance monitoring of a mammalian cell based bioprocess using Raman spectroscopy. Anal. Chim. Acta 796: 84–91. Sandor, M., Rudinger, F., Bienert, R. et al. (2013). Comparative study of noninvasive monitoring via infrared spectroscopy for mammalian cell cultivations. J. Biotechnol. 168: 636–645. Perry, R.H., Green, D.W., and Maloney, J.O. (1997). Perry’s Chemical Engineers’ Handbook. New York: McGraw-Hill. Nienow, A.W., Widsom, D.J., and Middleton, J.C. (1977). Effect of scale and geometry on flooding, recirculation and power in stirred vessels. Proceedings of the 2nd European Conference on Mixing, 1–16.

References

44 Oosterhuis, N.M., Kossen, N.W., Olivier, A.P., and Schenk, E.S. (1985).

45 46

47

48

49

50

51

52

Scale-down and optimization studies of the gluconic acid fermentation by Gluconobacter oxydans. Biotechnol. Bioeng. 27: 711–720. Ognean, T. (1993). Dimensionless criteria for estimating oxygen transfer in aeration systems. Biotechnol. Bioeng. 41: 1014–1020. Perez, R.E., Lasa, A.M., Rodriguez, R.S. et al. (2006). Scale-up of recombinant Opc protein production in Escherichia coli for a meningococcal vaccine. J. Biotechnol. 127: 109–114. Xing, Z., Kenty, B.M., Li, Z.J., and Lee, S.S. (2009). Scale-up analysis for a CHO cell culture process in large-scale bioreactors. Biotechnol. Bioeng. 103: 733–746. Van’t Riet, K. (1979). Review of measuring methods and results in nonviscous gas-liquid mass transfer in stirred vessels. Ind. Eng. Chem. Process Des. Dev. 18: 357–364. Zhu, Y., Cuenca, J.V., Zhou, W., and Varma, A. (2008). NS0 cell damage by high gas velocity sparging in protein-free and cholesterol-free cultures. Biotechnol. Bioeng. 101: 751–760. Le, H., Kabbur, S., Pollastrini, L. et al. (2012). Multivariate analysis of cell culture bioprocess data--lactate consumption as process indicator. J. Biotechnol. 162: 210–223. Rathore, A.S., Mittal, S., Pathak, M., and Arora, A. (2014). Guidance for performing multivariate data analysis of bioprocessing data: pitfalls and recommendations. Biotechnol. Progr. 30: 967–973. Burdick, R.K., Pferdeort, V., Sidor, L., and Tholudur, A. (2011). A graphical representation for a statistical test of average equivalence and variance comparison with process data. Qual. Reliab. Eng. Int. 27: 771–780.

405

407

Index a absolute quantification (AQUA) 170 ad hoc software tools 261 affinity chromatography process 12–13, 297 α-1,6-fucosyltransferase (FUT8) modulation 198, 229–231 alpha-spectrum 82 alternating tangential flow (ATF) 11, 283, 349, 351 Amino acid-Polyamine-organocation (APC) superfamily 136 anion exchange chromatography (AEX) 296 anNET 141 5-azacytidine (5-Aza) 107 5-aza-2′ -deoxycytidine (DAC) 107, 108

b baby hamster kidney (BHK21) cell lines 27 batch culture 10, 35, 105, 261, 264, 317, 318, 353 advantage 317 batch-level analysis 397–398 B-cell lymphoma 2 (BCL2) gene 74, 221 biologically active HCPs 300 biopharmaceutical cell culture process adherent cells 316 batch culture 317 fed-batch cultures 317–319 suspended cells 316

biopharmaceuticals 1, 3, 5, 6, 8, 11–13, 15–17, 23, 24, 28, 36, 49, 55, 56, 62, 73, 97, 98, 127–153, 185, 198–200, 207, 208, 233, 234, 265, 285, 295, 297, 303, 305, 313–333, 347, 348, 350–353, 361, 366–370, 381–383 Biosep device 357–358 biotherapeutics 172, 227, 279, 280, 282, 283, 381–382 bisulfite sequencing method 168

c calcium phosphate-based transient transfection 50 carboxypeptidase D 304 Cas9 endonuclease 186 cation exchange chromatography (CEX) 296 cell culture process, for TGE 59 culture longevity-enhancing factors 59, 60 qp -enhancing factors 59 cell viability on HCP profiles 303 Cell XpressTM system 32 centrifugation 51, 62, 257, 295, 296, 354, 356 CentriTech unit 356 chain termination Sanger sequencing method 165 chemically defined (CD) media development, for CHO cells basal and feed medium formulations 283

Cell Culture Engineering: Recombinant Protein Production, First Edition. Edited by Gyun Min Lee and Helene Faustrup Kildegaard. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

408

Index

cell line development 280 cell line selection 279–280 considerations 284 design of experiment (DoE) approach 280 fed-batch production process 283, 284 medium component classification 281 perfusion process 283 seed train expansion 281 standard workflow for 281 systems and synthetic biology 284–288 chemically defined feeds 264 chemically defined media (CDM) 10, 163, 234, 261, 266, 320 Chinese hamster ovary (CHO) cell line engineering autophagy 225, 226 cell division rates and proliferation 226 CRISPR/Cas9 system 197, 209 CRISPR/Cpf1 system 198 CRISPR-mediated targeted integration 199 gene knockout 209 improving recombinant protein production cellular metabolism 211 cellular protein synthesis 218 continuous acidification of culture medium 219 cytoskeleton dynamics 219 miRNAs, potential of 220 secretory capacity 218 telomeric region of chromosome 8, 219 necrosis/necroptosis 225 noncoding RNA-mediated gene silencing 209–211 overexpression of antiapoptotic genes 221 overexpression of engineering genes 208–209 posttranslational modifications 227

C-terminal amidated species of mAbs 233 glycosylation 227, 231 N-glycosylation 231 O-glycosylation 232 sialylation 232, 233 proapoptotic genes, inhibition of 225 programmed cell death/apoptosis 221, 222 sodium butyrate treatment 225 transcriptomics studies 225 viral safety 233 Chinese hamster ovary (CHO) cells chemically defined media development 279 epigenomics of 168 genetic/genomic instability 97, 99–101 genome-scale model 174, 178 genome sequencing of 166–167 genomic resource development 75 glutamine synthetase deficient 73 glycomics 172 host cell protein risk assessment in 305 lipidomics 172 low-temperature culture 174 metabolomics 171–172 microRNAs (miRNA) 167 multi-omics profiling bioprocess optimization 174 cell line characterization 174–176 engineering target identification 176–177 proteome database 305 proteomics 75–76, 170–171 transcriptomics of 75–76, 167 CHO-K1 cell line 6, 26, 28, 33, 100, 106, 172 CHO-K1 genome 5, 75, 166, 167 CHO Protein Predicted Immunogenicity (CHOPPI) 305 CHO-S cell line 26, 33 chromatin immunoprecipitation and sequencing (ChIP-seq) 168

Index

chromatographic method 12, 172 13 C-fluxomics 257, 261, 266 class 2 type V CRISPR/Cpf1 188 clonality 9, 36, 73 ClonePixTM system 9, 32 co-elution 298 cofilin-1 219 computational systems biology 84 concentrated fed-batch 332 constraint-based metabolic models 381, 382 constraint-based modeling approaches 80, 127, 381 context-specific models (CSMs) 13 C-labeled flux data 149 description 146 FAST Consistent Reconstruction 148 gene inactivity mediated by metabolism and expression 146 integrative metabolic analysis tool 147 integrative network inference for tissues method 147–148 metabolic adjustment by differential expression 147 metabolic context specificity assessed by deterministic reaction evaluation 148 model building algorithm 147 performance 148–149 thresholding value 149 continuous biomanufacturing 305, 347–362 continuous cell culture process description 347–348 with cell retention (perfusion) advantages 351 biomanufacturing facility utilization 352 Biosep device 357–358 centrifugation 356 current status of 360 external cell retention device 350 filtration based devices 354–355 high volumetric productivities 351

hydrocyclones 358 immobilization/entrapment systems 349–350 internal cell retention device 350 product quality and consistency 352–353 scale up and commercial production 353–354 settlers 356–357 spin filters 355–356 stirred tank bioreactor operation and control 358 without cell retention (chemostat) 348 continuous stirred tank perfusion bioreactor system 358 cell bleed/purge 360 cell build-up phase 359–360 circulation/return pump 359 external cell retention device 358 feed and harvest flow and volume control 358–359 production phase 360 core fucose 198, 231, 232 Cre/loxP system 33 Cricetulus griseus GEM 129 CRISPR activation (CRISPRa) 195 CRISPR/Cas9 system biotechnological potential of 185 components 186 genome editing experimental design 189–191 gene knockout 192–194 mechanism of 186–187 site-specific gene integration 194–195 multiplex genome engineering 192 RNA-targeting 196 CRISPR/Cas13 system 193, 196 CRISPR/Cas systems engineered Cas9 variants 188 evolutionary diversity in bacteria and archaea 187 CRISPR interference (CRISPRi) 195 CRISPR RNA (crRNA) 186–188 critical quality attribute (CQA) 298, 366, 372

409

410

Index

culture media, for transient gene expression 58 CycleFreeFlux 143 CycleFreeFVA approach 139 cytogenetic methods 99, 101

d dCas9-DNMT3A-DNMT3L fusion protein 112 dCas9-p300 fusion construct 110 dCpf1-based transcriptional activators 196 dead Cas9 (dCas9) 110, 188 difficult-to-remove host cell protein (HCP) 298 dihydrofolate reductase (DHFR)-deficient CHO sublines 207 dimeric cartilage matrix proteinangiopoietin-1 35 direct injection-mass spectrometry (DI-MS) 253 DNA demethylation 112, 113 DNA methylation 102, 107 CpG dinucleotides 105 global manipulation of 107–108 reader proteins of 104 targeted DNA methylation 112–113 DNA methyltransferases (DNMTs) 54, 102, 112 DNA targeting CRISPR/Cas9 system 187 DNA transposons 34 downstream process development, in therapeutic protein production purification 12–13 quality by design (QbD) concept 13–14 Drotrecogin alfa 34 DXB11 cell line 26 dynamic flux balance analysis (DFBA) 375

e electroporation 7, 8, 30, 51–52, 59–61, 191, 208

electroporation-based transfection 51, 52 endonuclease-mediated targeted integration method 33 energy dissipation rate (EDR) 326 engineered Cas9 variants 188 epigenetics 101 DNA methylation 102 histone modifications 102–104 long noncoding RNAs 104 targeted epigenetic modification 109–113 EpiMatrix 305 equivalence acceptance criterion (EAC) 399, 400 erythropoietin (EPO) 23, 174, 232, 266, 297, 315, 316 euchromatin 101, 110 extensions 147

f FAST consistent reconstruction (FASTCORE) 148 Fast-SNP 143 fed-batch cell culture process development 318, 329 bioreactor technologies 331–332 near infrared and Raman spectroscopy probes 330 perfusion 331, 332 fingerprint analysis 253 FlexFlux 145 Flp/FRT system 15, 33, 199 fluorescence-activated cell sorting (FACS) 9, 32, 74, 189 flux balance analysis (FBA) 81, 150, 257, 375, 380 assumptions 128 genome-scale model 129 limitations 131 biological objective function 133 genetic regulation 131 kinetics and metabolite concentrations 133 limited intracellular space 132 thermodynamically infeasible cycles 131

Index

regulatory networks 145 single optimal reaction network state 133 flux balance analysis with molecular crowding (FBAwMC) 151 flux variability analysis (FVA) 82, 133 footprint analysis 253 Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer 169

g gas chromatography–mass spectrometry (GC–MS) 253–256 gas entrance velocity 393 gene inactivity mediated by metabolism and expression (GIMME) algorithm 146–147 gene–protein-reaction (GPR) 77, 130 genetic compensation 304 genetic/genomic instability, CHO cells 97, 99–101 genome editing CRISPR/Cas9-mediated epigenetic modification 196 experimental design 189 gRNA target site selection and design 189–191 plasmid-based delivery 191, 192 timeline of 190 viral delivery 192 gene knockout 192–194 mechanism of 187 RNA targeting 196 site-specific gene integration 194–195 transcriptional regulation 195–196 genome editing tools, for HCP removal 304 genome-scale constraint-based metabolic models 382 genome-scale CRISPR screening 197 genome-scale metabolic models 76, 82, 127 automatic reconstruction methods 77–79 draft reconstruction 77

external exchange reaction loops 134 gene regulation, advantages of 144 H+ -coupled symport transporters 137 knowledge-base construction 77 mathematical format 79 modeling techniques 85 Na+ -linked transporters 136–137 omics data 83–84 reversible passive antiporters 136 reversible passive transporters 135–136 validation and evaluation 79–80 GIMME by Proteome (GIMMEp) 147 glutamine synthetase (GS)-deficient CHO cell lines 207 glutathione S transferase-α (GST-α) 302 glycoengineering 198 glycosylation 1, 5, 8, 12, 23, 24, 34, 35, 38, 54, 76, 85, 102, 172, 174, 197, 198, 225, 227–229, 231, 232, 265, 266, 280, 282, 284, 305, 314–315, 319, 320, 322, 325–329, 332, 369, 375–378, 382 pathways 231 guide RNA (gRNA) 33, 110, 186 target site selction and design 189, 191

H H+ -coupled symport transporters 137 Henderson–Hasselbalch equation 325, 326 high throughput omics technologies mass spectrometry 168–170 sequencing-based 165–168 histone acetylation 59, 107–109 homology-directed repair (HDR) 15, 34, 187, 194 homology-independent targeted integration (HITI) method 195 host cell proteins (HCPs) detection and monitoring methods 300–302

411

412

Index

host cell proteins (HCPs) (contd.) enzyme-linked immunosorbent assay 300–302 immunoassays 300, 302, 306 impurity removal process antibody product 296–297 depth filtration step 296 difficult-to-remove HCPs 298 harvested cell culture fluid 296 hydrophobic interaction chromatography 296 ion exchange chromatography 296 non-antibody protein product 297 protein A chromatography 296 non-gel based proteomics approaches 302 on polysorbate degradation 233 removal/control of 303 downstream strategies 304 upstream process 303 residual 298–300 risk assessment in CHO cells 305 sodium dodecyl sulfate-polyacrylamide gel electrophoresis 302 2D-differential in-gel electrophoresis 302 human embryonic kidney (HEK293) cells 5, 24, 35, 49 human tissue plasminogen activator 1, 49, 73, 211 hybrid perfusion/fed-batch process 16 hydrocyclones 358 hydrophobic interaction chromatography (HIC) 296

I Illumina sequencing 76, 166, 167 immunoassays 300, 302, 306 immunoglobulin G (IgG) sepharose affinity chromatography 297 insect cells 23, 24 insulin 23, 60, 218, 297, 319 integrative metabolic analysis tool (iMAT) 147 integrative network inference for tissues (INIT) method 147–148

interleukins (ILs) 297 internal ribosome entry sites (IRES) 7, 74 International Council for Harmonisation (ICH) quality guideline 295, 391 ion exchange chromatography (IEX) 296 ionizers 168 isobaric tags for relative and absolute quantitation (iTRAQ) 170, 302

l large-scale TGE-based protein production 55, 60–62 lectin-based microarrays 172 linear ion trap (LIT) 169 lipofection 7, 8, 52, 112, 208 liposome-based transfection reagents 52, 53 liposome-based transient transfection 52–53 liquid chromatography–mass spectrometry (LC–MS) 253–256 liquid chromatography–tandem mass spectrometry (LC–MS/MS) 302 ll-COBRA 143 long noncoding RNAs (lncRNAs) 104 loop law 139

m macromolecular crowding 132, 150 major facilitator superfamily (MFS) 135–136 mammalian cell culture process development aggregation 316 ammonium accumulation 322 basal media 319–320 batch cultures 10, 317 cell culture supplements 327–329 charge heterogeneity 315–316 culture environment optimization 10 fed-batch cultures 11, 317–319 feed media 321 design 323–325

Index

glycosylation 314–315, 328–329 lactate accumulation 322 media development 10 metabolite based feeding strategy 321–322 objectives of 314–316 osmolality 322 perfusion culture 11 pH-dependent feeding strategy 322 process variable design 325–327 culture duration 327 dissolved oxygen 326–327 pH and pCO2 325–326 temperature 325 product quality 314 quality analysis 12 respiration based feeding strategy 323 scale up process 11, 12 single use bioreactor systems 11–12 yield 314, 328 mammalian cell cultures 1, 3, 10, 11, 13, 58, 60, 73, 84, 252, 257, 313, 319, 325, 326, 332, 333, 367–371, 374, 377 mammalian host cell lines 3, 5, 24–27, 54 mass analyzers 168, 169 mass conservation constraints 140 mass spectrometry (MS) 38, 76, 164, 168–170, 252, 255, 256, 259, 285, 302, 324, 368 mass spectrometry-based omics technologies CHO cells glycomics 172 lipidomics 172 metabolomics 171–172 proteomics 170–171 liquid chromatography–mass spectrometry 169 mass spectrometer, components of 168 tandem mass spectrometry 169 matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOFMS) 368, 369

matrix attachment regions (MARs) 7, 74, 107 mechanistic-based kinetic models 382 mechanistic mathematical models 84 metabolic adjustment by differential expression (MADE) 147 metabolic context-specificity assessed by deterministic reaction evaluation (mCADRE) 148 metabolic data, in mammalian bioprocessing 261 clonally derived recombinant cell lines 263 genetic engineering 265–266 glycosylation profiles 265 growth phase vs. metabolism 261–263 improving biomass and recombinant protein yield 263–265 lactate shift 263 metabolic indicators of high productivity 263 UDP-Gal precursor 265 metabolic engineering of cell lines 85 metabolic flux 80, 151, 257, 281, 287, 369, 375, 378 metabolic modeling with enzyme kinetics (MOMENT) 151 metabolite-based feeding strategy 321–322 metabolite profile generation 253 mass spectrometry 255–256 nuclear magnetic resonance 254–255 metabolite profiling 251–267 metabolites 74, 76–83, 86, 127–131, 133–138, 140, 141, 144–146, 148, 149, 151–153, 171, 172, 178, 221, 251–267, 280, 281, 283, 285–287, 295, 321–324, 330, 331, 359, 366, 378–380, 382, 396 detection and quantitation 252 metabolite sample preparation extracellular 257 extraction methodology 257 metabolic flux analysis 257

413

414

Index

metabolite sample preparation (contd.) quenching of intracellular samples 257 sampling 256 metabolome 128, 174, 176, 251–253, 258 metabolomics 251, 253 data analysis 257–260 data interpretation and integration 260–261 data mining techniques 258, 260 data processing 258 univariate and multivariate methods 258 methylation-induced transcriptional silencing 37 microbial systems 23, 24, 381 microhomology-mediated end joining (MMEJ) 34, 187 microRNAs (miRNAs) 104, 167, 210, 211, 227 potential of 220 miR-143 217, 221 miR-2861 216, 220 miR-30 family, in CHO cells 220 miR-34a 217, 220, 227 miR-466h 227 miR-483 217, 221 mixed-integer linear programming (MILP) algorithm 82 mixed-mode chromatography 12, 303, 304 model building algorithm (MBA) 147 model extraction methods (MEMs) 83 molecular crowding 132, 150–152 monoclonal antibodies (mAbs) 5, 23, 60, 61, 74, 198, 208, 212, 215, 225, 230, 231, 233, 234, 266, 279–281, 295, 313, 314, 347, 360, 361, 378–380 multilocus genomic fingerprinting 99 multiomic approaches 14 multi-omics data integration 177 network-based modeling approaches 178 statistical methods 177

multiple reaction monitoring (MRM) 302 murine NS0 cell 3, 27, 108 multicolor fluorescence in situ hybridization (mFISH) 100

N Na+ linked transporters 136–137 network-embedded thermodynamic analysis (NET) 140 NExT 140 next-generation RNA sequencing technology 211 next-generation sequencing (NGS) 100, 127, 189, 285 nickase Cas9 (nCas9) 188 N-linked glycosylation 24, 229, 231, 232, 314, 315 nonchromatographic separation 12, 13 noncoding RNA mediated gene silencing 209–211 non-homologous end joining (NHEJ) 15, 187 non-viral-based gene delivery 50–53 nonviral gene transfer 7, 30 nuclear magnetic resonance (NMR) 127, 171, 253–256, 285, 324, 365 nucleosomes 101–104, 144, 196 nucleotide sequencing techniques, historical development of 165–166

O O-linked glycosylation 314, 315 omics data 15, 83–84, 127, 146, 148, 164, 165, 172–174, 176–178, 200, 252, 261, 286, 382 Orbitrap 169 oxygen transfer coefficient 323, 392–393, 401 oxygen transfer rate (OTR) 393–395 oxygen uptake rate (OUR) 323

p parsimonious flux balance analysis (pFBA) 131, 143, 148, 149, 152

Index

PAT-A framework 365 perfusion culture advantages 351 Biosep device 357–358 centrifugation 356 external cell retention device 350 filtration based devices 354–355 hydrocyclones 358 immobilization/entrapment systems 349–350 internal cell retention device 350 settlers 356–357 spin filters 355–356 phase partitioning method 13 pH-dependent feeding strategy 322 PiggyBac transposon 34 piwi-interfering RNAs (piRNAs) 104 plant cells 23, 24 platform technology development 3, 14–16 polyclonal antibodies 300 polyclonal anti-HCP antibody 300 polyethylenimine (PEI)-based transient transfection 52 post-transcriptional gene silencing (PTGS) 104 power/volume (P/V) ratio 392, 401 process analytical technology (PAT) advanced analytical methods 369 amino acid supplementation 375–377 challenges 365–366 chemometric techniques 369 data analysis and modeling tools 369–370 definition 365 design space (DS) exploration strategy 372 fluorescence spectroscopy 368 high performance liquid chromatography 366 mAb glycosylation 378 mammalian cell culture raw material characterization 374–375 mid-infrared (MIR) spectroscopy 367

multiwavelength UV spectroscopy 372–374 NIR spectroscopy 367 nutrient consumption metabolism 378 protein mixture quantification 372–374 quantitative intracellular flux modeling 381–382 Raman spectroscopy 367–368 raw material performance, in mammalian cell culture 370–371 process-related impurities 12, 295, 353 product-related impurities 12, 295 protein-A method 13 protein-based drugs 85 protein folding 1, 5, 109, 200, 316, 325 protein therapeutics 85, 279–288 proteomics approach, for HCP detection 302 pyrosequencing-based method 165

q quadrupole analyzer 169 quadrupole ion trap (QIT) 169 quality by design (QbD) advanced analytical methods 369 amino acid supplementation 375–377 challenges 365–366 chemometric techniques 369 data analysis and modeling tools 369–370 design space exploration strategy 372 fluorescence spectroscopy 368 high-performance liquid chromatography 366 mAb glycosylation 378 mammalian cell culture raw material characterization 374–375 mid-infrared spectroscopy 367 multiwavelength UV spectroscopy 372–374 NIR spectroscopy 367

415

416

Index

quality by design (QbD) (contd.) nutrient consumption metabolism 378 protein mixture quantification 372–374 quantitative intracellular flux modeling 381–382 Raman spectroscopy 367–368 raw material performance, in mammalian cell culture 370–371 quantification concatamers (QconCAT) 170

r reader proteins, of DNA methylation 104 recombinant cell line development, for therapeutic protein production CHO cells clone selection 31–32 dihydrofolate reductase (DHFR) system 28 gene amplification 30–31 glutamine synthetase (GS) system 28 selection medium 30 site-specific integration 32 stable expression system 27 transfection methods 30 transient expression system 27 vector construction 29–30 clonality 36 clone selection 8–9 expression vector 7 host selection 6 necessity for 34–35 quality 37–38 stability 36–37 transfection/selection 7 recombinant protein expression 176, 211, 220, 251, 263 recombinase-mediated cassette exchange (RMCE) 15, 33, 194, 199 reconstructed mammalian GEMs 129

regulatory flux balance analysis (rFBA) 145 regulatory steady-state analysis (RSA) 145 residual host cell proteins (HCPs) biological activity 299–300 drug efficacy 298–299 immunogenicity 299 quality and shelf life 298–299 resource balance analysis (RBA) 151 respiration based feeding strategy 323 RNA-induced silencing complex (RISC) 210 RNA interference (RNAi) 15, 197, 209 RNA sequencing (RNA-seq) 144, 165, 166, 195, 199, 211

s Saccharomyces cerevisiae iIN800 model 131 scale down model (SDM) development challenges 391 equivalence testing 399–401 evaluation 395–401 batch-level analysis 397 multivariate analysis 396–399 observation level analysis 397 qualification data set 396–397 scores contribution plots 398–399 univariate analysis 395–396 gas entrance velocity 393 oxygen transfer coefficient 392–393 oxygen transfer rate 393–395 power/volume ratio and air velocity 392 refinement workflow 395 scores contribution plots 398–399 secretory capacity, of CHO cells 218 selected reaction monitoring (SRM) 171 semicontinuous cultures 348, 351 semisolid matrix-based systems 9 sequential quadratic programming (SQP) method 140

Index

sequential window acquisition of theoretical mass spectra (SWATH) MS 171, 176 single-cell cloning process 36, 74, 209 single molecule real-time sequencing technology (SMRT) 166 single-stranded oligodeoxynucleotides (ssODNs) 194 small-interfering RNAs (siRNAs) 104, 209, 210 sodium butyrate (NaBu) 59, 74, 108, 109, 225, 328 Sp2/0-Ag14 cell lines 24, 27 SpCas9-mediated DNA recognition 187 spin filters 355–356 stable human cell line development process 16, 35–36 stable isotope labeling with amino acids (SILAC) 170 stoichiometric models 128, 138, 140, 142, 382 SV40-based episomal replication system 54 synthetic biology 114, 284–288, 333 systems biology 15, 84, 86, 284–288, 375

T TALE-Sin3a fusion constructs 111 tangential flow filtration cell retention device 283, 355 targeted integration (TI)-based CLD 15 targeted integration method 32, 33 targeted metabolomics approach 253 therapeutic protein production, in mammalian cells 5 baby hamster kidney (BHK21) cell lines 27 CHO cell lines 25–26 downstream process development 12–14 HEK293 cell lines 26 HKB-11 cell lines 26 improving specific productivity 5 increasing cell mass 5

murine NS0 cell lines 27 PER.C6 cell lines 26 protein quality 5 purification 12–13 quality by design (QbD) 13–14 recombinant cell line development guidelines 6–9 Sp2/0-Ag14 cell lines 27 therapeutic proteins 347 description 1 global market value 1 mammalian manufacturing platform 1 quality of 37–38 thermodynamic variability analysis (TVA) 141 thermodynamically infeasible cycles 131, 134 algorithms developed 138–139 exclusion of minimization of total flux 142–144 thermodynamic constraints 140 manual curation 139–140 Monte Carlo method 139 transporter types 135 visualizing flux on network maps 138 Wright and Wagner (WW) algorithm 138 thermodynamics-based metabolic flux analysis (TMFA) 141 time-of-flight (TOF) mass analyzer 169, 255 titer 3, 8, 9, 11, 31, 35, 49, 50, 54, 56–61, 73, 74, 85, 107, 163, 211–216, 218, 220, 222–225, 265, 266, 280, 318, 320, 322, 325, 327–330, 332, 333, 348, 351, 352, 360–362, 375–377 trans-activating crRNA (tracrRNA) 186 transcription activator-like effector nuclease (TALEN) 33, 197, 209, 304 transcription activator-like effectors (TALEs) 33, 110

417

418

Index

transforming growth factor β1 (TGFβ1) 300 transient gene expression (TGE)-based protein production cell culture strategies 58–60 CHO cells 56–58 coexpression strategies 54 episomal replication 53–54 expression vector composition and preparation 53 HEK293 cells 55–56 HKB-11 cells 58 large-scale 60–62 non-viral-based gene delivery 50 viral-based gene delivery 50 transposon-mediated gene integration method 34 2D gel electrophoresis 76

untargeted metabolomics approach 253 uridine diphosphate N-acetylglucosamine (UDP-GlcNAc) 328

v valeric acid 109, 328 valproic acid 59, 109 Van’t Riet equation 392 viral-based gene delivery 50 viral promoters 53, 105 virus-mediated transfection method 30

W Warburg effect

150, 151

z u ubiquitous chromatin opening element (UCOE) 107

zinc-finger nuclease (ZFN) 33, 197, 209, 266 zinc finger proteins (ZFP) 110, 212