Yeast Surface Display 9781071622858, 1071622854

Citation preview

Methods in Molecular Biology 2491

Michael W. Traxlmayr Editor

Yeast Surface Display

METHODS

IN

MOLECULAR BIOLOGY

Series Editor John M. Walker School of Life and Medical Sciences University of Hertfordshire Hatfield, Hertfordshire, UK

For further volumes: http://www.springer.com/series/7651

For over 35 years, biological scientists have come to rely on the research protocols and methodologies in the critically acclaimed Methods in Molecular Biology series. The series was the first to introduce the step-by-step protocols approach that has become the standard in all biomedical protocol publishing. Each protocol is provided in readily-reproducible step-bystep fashion, opening with an introductory overview, a list of the materials and reagents needed to complete the experiment, and followed by a detailed procedure that is supported with a helpful notes section offering tips and tricks of the trade as well as troubleshooting advice. These hallmark features were introduced by series editor Dr. John Walker and constitute the key ingredient in each and every volume of the Methods in Molecular Biology series. Tested and trusted, comprehensive and reliable, all protocols from the series are indexed in PubMed.

Yeast Surface Display Methods and Protocols

Edited by

Michael W. Traxlmayr Department of Chemistry, Institute of Biochemistry, University of Natural Resources and Life Sciences, Vienna, Austria

Editor Michael W. Traxlmayr Department of Chemistry, Institute of Biochemistry University of Natural Resources and Life Sciences Vienna, Austria

ISSN 1064-3745 ISSN 1940-6029 (electronic) Methods in Molecular Biology ISBN 978-1-0716-2284-1 ISBN 978-1-0716-2285-8 (eBook) https://doi.org/10.1007/978-1-0716-2285-8 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022, Corrected Publication 2022 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Humana imprint is published by the registered company Springer Science+Business Media, LLC, part of Springer Nature. The registered company address is: 1 New York Plaza, New York, NY 10004, U.S.A.

Preface Yeast surface display is based on covalent or noncovalent anchoring of heterologous target proteins to the yeast cell surface. Since its first description in the 1990s, yeast surface display has rapidly evolved into an extensively used protein engineering technology. Notable advantages of this display platform include the availability of a eukaryotic protein expression machinery, as well as the versatility of selection and screening strategies based on magnetic beads, mammalian cell panning, or quantitative flow cytometric screening. Moreover, once mutants have been enriched, their basic biochemical properties such as affinity and stability can be analyzed directly on the surface of yeast. While the major application of yeast surface display is the generation or affinity maturation of antigen binding sites, it has also been used for a range of other applications, including enzyme engineering, epitope mapping, identification of peptide-MHC ligands, engineering of fluorescence-activating proteins, and protein stability engineering, among others. This volume starts with a review on the development, characteristics, and applications of the yeast surface display technology (Part I). Part II contains detailed protocols for the construction and efficient selection/screening of yeast surface display libraries, as well as for the analysis of individual yeast-displayed protein variants. Part III comprises a collection of protocols describing the selection of yeast surface display libraries for binding to mammalian cells or to extracellular matrix. Finally, in Part IV, protocols for a broad spectrum of specialized yeast surface display applications are presented, demonstrating the versatility of this display platform. We are confident that this volume with its detailed protocols will be a comprehensive resource, enabling the implementation of this powerful and versatile technique in virtually any molecular biology laboratory, even in the absence of any prior yeast surface display experience. In addition to more general protocols, this volume also contains specialized and improved yeast surface display strategies, thus also providing valuable information for the advanced yeast surface display expert. I would like to thank all authors for their contributions to this volume. In addition, I would like to express my thanks to John M. Walker for initiating this volume on yeast surface display and for his help and advice, and to Patrick Marton and Anna Rakovsky from Springer for their support during the publishing process. Vienna, Austria

Michael W. Traxlmayr

v

Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

PART I

INTRODUCTION TO YEAST SURFACE DISPLAY AND ITS APPLICATIONS

1 Yeast Surface Display: New Opportunities for a Time-Tested Protein Engineering System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

PART II

v xi

3

CONSTRUCTION AND SELECTION OF YEAST SURFACE DISPLAY LIBRARIES AND ANALYSIS OF ISOLATED VARIANTS

2 Yeast Surface Display for Protein Engineering: Library Generation, Screening, and Affinity Maturation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Byong H. Kang, Brianna M. Lax, and K. Dane Wittrup 3 Site-wise Diversification of Combinatorial Libraries Using Insights from Structure-guided Stability Calculations. . . . . . . . . . . . . . . . . . . . . . . . Benedikt Dolgikh and Daniel Woldring 4 Ancestral Sequence Reconstruction and Alternate Amino Acid States Guide Protein Library Design for Directed Evolution . . . . . . . . . . . . . James VanAntwerp, Patrick Finneran, Benedikt Dolgikh, and Daniel Woldring 5 Machine Learning-driven Protein Library Design: A Path Toward Smarter Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mehrsa Mardikoraem and Daniel Woldring 6 Kinetic Competition Screening of Yeast-Displayed Libraries for Isolating High Affinity Binders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nicole J. Yang 7 Engineering Proteins by Combining Deep Mutational Scanning and Yeast Display. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Preeti Sharma, Erik Procko, and David M. Kranz 8 Engineering Binders with Exceptional Selectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . Kai Wen Teng, Akiko Koide, and Shohei Koide 9 Affinity and Stability Analysis of Yeast Displayed Proteins . . . . . . . . . . . . . . . . . . . . Charlotte U. Zajc, Magdalena Teufl, and Michael W. Traxlmayr

vii

29

63

75

87

105

117 143 155

viii

Contents

PART III 10

11

12

13

Antibody Library Screening Using Yeast Biopanning and Fluorescence-Activated Cell Sorting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stefania C. Carrara, Jan P. Bogen, Julius Grzeschik, Bjo¨rn Hock, and Harald Kolmar A Hybrid Adherent/Suspension Cell-Based Selection Strategy for Discovery of Antibodies Targeting Membrane Proteins. . . . . . . . . . . Patrick J. Krohl and Jamie B. Spangler Ligand Selection by Combination of Recombinant and Cell Panning Selection Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rojhae A. Panton and Lawrence A. Stern Identification of Brain ECM Binding Variable Lymphocyte Receptors Using Yeast Surface Display. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Benjamin J. Umlauf, John S. Kuo, and Eric V. Shusta

PART IV 14

15

16

17

18

19

20

SELECTION OF YEAST SURFACE DISPLAY LIBRARIES FOR BINDING TO MAMMALIAN CELLS OR EXTRACELLULAR MATRIX 177

195

217

235

SPECIALIZED YEAST SURFACE DISPLAY APPLICATIONS

Guidelines, Strategies, and Principles for the Directed Evolution of Cross-Reactive Antibodies Using Yeast Surface Display Technology . . . . . . . . Sara Linciano, Ee Lin Wong, Ylenia Mazzocato, Monica Chinellato, Tiziano Scaravetti, Alberto Caregnato, Veronica Cacco, Zhanna Romanyuk, and Alessandro Angelini Yeast Display for the Identification of Peptide-MHC Ligands of Immune Receptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Brooke D. Huisman, Beth E. Grace, Patrick V. Holec, and Michael E. Birnbaum Yeast Display Guided Selection of pH-Dependent Binders . . . . . . . . . . . . . . . . . . . Jenna N. Meanor, Albert J. Keung, Balaji M. Rao, and Nimish Gera Yeast Mating as a Tool for Highly Effective Discovery and Engineering of Antibodies via Display Methodologies . . . . . . . . . . . . . . . . . . . Du-San Baek, Seong-Wook Park, Cynthia Adams, Dimiter S. Dimitrov, and Yong-Sung Kim Humanization of Chicken-Derived Antibodies by Yeast Surface Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jan P. Bogen, Adrian Elter, Julius Grzeschik, Bjo¨rn Hock, and Harald Kolmar Engineering Tissue Inhibitors of Metalloproteinases Using Yeast Surface Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh Discovery of Cyclic Peptide Binders from Chemically Constrained Yeast Display Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kaitlyn Bacon, Stefano Menegatti, and Balaji M. Rao

251

263

293

313

335

361

387

Contents

Generation of Thermally Stable Affinity Pairs for Sensitive, Specific Immunoassays. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Elliot Corless, Yining Hao, Huan Jia, Patthara Kongsuphol, Dousabel M. Y. Tay, Say Yong Ng, and Hadley D. Sikes 22 Isolating Anti-amyloid Antibodies from Yeast-Displayed Libraries . . . . . . . . . . . . Alec A. Desai, Jennifer M. Zupancic, Matthew D. Smith, and Peter M. Tessier 23 Engineering Proteins Containing Noncanonical Amino Acids on the Yeast Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rebecca L. Hershman, Arlinda Rezhdo, Jessica T. Stieglitz, and James A. Van Deventer 24 Construction of Yeast Display Libraries for Selection of Antigen-Binding Variants of Large Extracellular Loop of CD81, a Major Surface Marker Protein of Extracellular Vesicles . . . . . . . . . . . . Stefan Vogt, Gerhard Stadlmayr, Katharina Stadlbauer, Florian Stracke, Madhusudhan Reddy Bobbili, Johannes Grillari, ¨ ker, and Gordana Wozniak-Knopp Florian Ru 25 Isolating and Engineering Fluorescence-Activating Proteins Using Yeast Surface Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lina El Hajji, Hela Benaissa, and Arnaud Gautier 26 Simultaneous Display of Multiple Kinds of Enzymes on the Yeast Cell Surface for Multistep Reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kouichi Kuroda and Mitsuyoshi Ueda Correction to: Yeast Surface Display for Protein Engineering: Library Generation, Screening, and Affinity Maturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

21

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

417

471

491

561

593

627

C1 643

Contributors CYNTHIA ADAMS • Center for Antibody Therapeutics, Division of Infectious Diseases, Department of Medicine, University of Pittsburgh Medical School, Pittsburgh, PA, USA ALESSANDRO ANGELINI • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy; European Centre for Living Technology (ECLT), Ca’ Bottacin, Venice, Italy KAITLYN BACON • Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, NC, USA DU-SAN BAEK • Center for Antibody Therapeutics, Division of Infectious Diseases, Department of Medicine, University of Pittsburgh Medical School, Pittsburgh, PA, USA HELA BENAISSA • Sorbonne Universite´, E´cole Normale Supe´rieure, Universite´ PSL, CNRS, Laboratoire des biomole´cules, LBM, Paris, France MICHAEL E. BIRNBAUM • Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA, USA; Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA MADHUSUDHAN REDDY BOBBILI • Institute of Molecular Biotechnology, Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna (BOKU), Vienna, Austria; Ludwig Boltzmann Institute for Experimental and Clinical Traumatology in the AUVA Research Center, Vienna, Austria ERIC T. BODER • Department of Chemical and Biomolecular Engineering, University of Tennessee, Knoxville, TN, USA JAN P. BOGEN • Institute for Organic Chemistry and Biochemistry, Technische Universit€ at Darmstadt, Darmstadt, Germany; Ferring Darmstadt Laboratory, Biologics Technology and Development, Darmstadt, Germany VERONICA CACCO • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy ALBERTO CAREGNATO • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy STEFANIA C. CARRARA • Institute for Organic Chemistry and Biochemistry, Technische Universit€ at Darmstadt, Darmstadt, Germany; Ferring Darmstadt Laboratory, Biologics Technology and Development, Darmstadt, Germany MONICA CHINELLATO • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy; Department of Medicine (DIMED), University of Padua, Padua, Italy ELLIOT CORLESS • Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA ALEC A. DESAI • Department of Chemical Engineering, University of Michigan, Ann Arbor, MI, USA; Biointerfaces Institute, University of Michigan, Ann Arbor, MI, USA DIMITER S. DIMITROV • Center for Antibody Therapeutics, Division of Infectious Diseases, Department of Medicine, University of Pittsburgh Medical School, Pittsburgh, PA, USA BENEDIKT DOLGIKH • Department of Chemical Engineering and Materials Science, Michigan State University, East Lansing, MI, USA; Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, MI, USA

xi

xii

Contributors

LINA EL HAJJI • Sorbonne Universite´, E´cole Normale Supe´rieure, Universite´ PSL, CNRS, Laboratoire des biomole´cules, LBM, Paris, France ADRIAN ELTER • Institute for Organic Chemistry and Biochemistry, Technische Universit€ at Darmstadt, Darmstadt, Germany; Merck Lab @ Technical University of Darmstadt, Darmstadt, Germany PATRICK FINNERAN • Menten AI, Palo Alto, CA, USA ARNAUD GAUTIER • Sorbonne Universite´, E´cole Normale Supe´rieure, Universite´ PSL, CNRS, Laboratoire des biomole´cules, LBM, Paris, France; Institut Universitaire de, Paris, France NIMISH GERA • Mythic Therapeutics, Waltham, MA, USA BETH E. GRACE • Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA, USA; Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA JOHANNES GRILLARI • Institute of Molecular Biotechnology, Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna (BOKU), Vienna, Austria; Ludwig Boltzmann Institute for Experimental and Clinical Traumatology in the AUVA Research Center, Vienna, Austria; Austrian Cluster for Tissue Regeneration, Vienna, Austria JULIUS GRZESCHIK • Ferring Darmstadt Laboratory, Biologics Technology and Development, Darmstadt, Germany YINING HAO • Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA REBECCA L. HERSHMAN • Chemical and Biological Engineering Department, Tufts University, Medford, MA, USA BJO¨RN HOCK • Ferring International Center S.A., Saint-Prex, Switzerland PATRICK V. HOLEC • Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA, USA; Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA BROOKE D. HUISMAN • Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA, USA; Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA HUAN JIA • Antimicrobial Resistance Interdisciplinary Research Group (AMR-IRG), Singapore-MIT Alliance in Research and Technology (SMART), Singapore, Singapore BYONG H. KANG • Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA; Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA, USA ALBERT J. KEUNG • Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, NC, USA YONG-SUNG KIM • Department of Molecular Science and Technology, Ajou University, Suwon, Republic of Korea AKIKO KOIDE • Perlmutter Cancer Center, New York University Langone Medical Center, New York, NY, USA; Department of Medicine, New York University School of Medicine, New York, NY, USA SHOHEI KOIDE • Perlmutter Cancer Center, New York University Langone Medical Center, New York, NY, USA; Department of Biochemistry and Molecular Pharmacology, New York University School of Medicine, New York, NY, USA HARALD KOLMAR • Institute for Organic Chemistry and Biochemistry, Technische Universit€ at Darmstadt, Darmstadt, Germany

Contributors

xiii

PATTHARA KONGSUPHOL • Antimicrobial Resistance Interdisciplinary Research Group (AMR-IRG), Singapore-MIT Alliance in Research and Technology (SMART), Singapore, Singapore DAVID M. KRANZ • Department of Biochemistry, University of Illinois, Urbana, IL, USA; Cancer Center at Illinois, University of Illinois, Urbana, IL, USA PATRICK J. KROHL • Department of Chemical and Biomolecular Engineering, Johns Hopkins University, Baltimore, MD, USA JOHN S. KUO • Department of Neurosurgery, Dell Medical School, The University of Texas at Austin, Austin, TX, USA; Mulva Clinic for the Neurosciences, The University of Texas at Austin, Austin, TX, USA KOUICHI KURODA • Division of Applied Life Sciences, Graduate School of Agriculture, Kyoto University, Kyoto, Japan BRIANNA M. LAX • Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA; Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA, USA SARA LINCIANO • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy MEHRSA MARDIKORAEM • Department of Chemical Engineering and Materials Science, Michigan State University, East Lansing, MI, USA; Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, MI, USA YLENIA MAZZOCATO • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy JENNA N. MEANOR • Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, NC, USA STEFANO MENEGATTI • Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, NC, USA; Golden LEAF Biomanufacturing Training and Education Center (BTEC), North Carolina State University, Raleigh, NC, USA SAY YONG NG • Antimicrobial Resistance Interdisciplinary Research Group (AMR-IRG), Singapore-MIT Alliance in Research and Technology (SMART), Singapore, Singapore ROJHAE A. PANTON • Department of Chemical, Biological, and Materials Engineering, University of South Florida, Tampa, FL, USA SEONG-WOOK PARK • Department of Molecular Science and Technology, Ajou University, Suwon, Republic of Korea ERIK PROCKO • Department of Biochemistry, University of Illinois, Urbana, IL, USA; Cancer Center at Illinois, University of Illinois, Urbana, IL, USA MARYAM RAEESZADEH-SARMAZDEH • Department of Chemical and Materials Engineering, University of Nevada, Reno, NV, USA BALAJI M. RAO • Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, NC, USA; Golden LEAF Biomanufacturing Training and Education Center (BTEC), North Carolina State University, Raleigh, NC, USA ARLINDA REZHDO • Chemical and Biological Engineering Department, Tufts University, Medford, MA, USA ZHANNA ROMANYUK • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy FLORIAN RU¨KER • Institute of Molecular Biotechnology, Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna (BOKU), Vienna, Austria

xiv

Contributors

TIZIANO SCARAVETTI • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy PREETI SHARMA • Department of Biochemistry, University of Illinois, Urbana, IL, USA; Cancer Center at Illinois, University of Illinois, Urbana, IL, USA ERIC V. SHUSTA • Departments of Chemical and Biological Engineering and Neurological Surgery, University of Wisconsin-Madison, Madison, WI, USA HADLEY D. SIKES • Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA; Antimicrobial Resistance Interdisciplinary Research Group (AMR-IRG), Singapore-MIT Alliance in Research and Technology (SMART), Singapore, Singapore MATTHEW D. SMITH • Department of Chemical Engineering, University of Michigan, Ann Arbor, MI, USA; Biointerfaces Institute, University of Michigan, Ann Arbor, MI, USA JAMIE B. SPANGLER • Department of Chemical and Biomolecular Engineering, Johns Hopkins University, Baltimore, MD, USA; Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, USA; Translational Tissue Engineering Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA; Department of Oncology, Johns Hopkins University School of Medicine, Baltimore, MD, USA; Bloomberg~Kimmel Institute for Cancer Immunotherapy, Johns Hopkins University School of Medicine, Baltimore, MD, USA; Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA; Department of Ophthalmology, Johns Hopkins University School of Medicine, Baltimore, MD, USA KATHARINA STADLBAUER • Christian Doppler Laboratory for Innovative Immunotherapeutics, Institute of Molecular Biotechnology, Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna (BOKU), Vienna, Austria GERHARD STADLMAYR • Christian Doppler Laboratory for Innovative Immunotherapeutics, Institute of Molecular Biotechnology, Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna (BOKU), Vienna, Austria LAWRENCE A. STERN • Department of Chemical, Biological, and Materials Engineering, University of South Florida, Tampa, FL, USA JESSICA T. STIEGLITZ • Chemical and Biological Engineering Department, Tufts University, Medford, MA, USA FLORIAN STRACKE • Christian Doppler Laboratory for Innovative Immunotherapeutics, Institute of Molecular Biotechnology, Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna (BOKU), Vienna, Austria DOUSABEL M. Y. TAY • Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA KAI WEN TENG • Perlmutter Cancer Center, New York University Langone Medical Center, New York, NY, USA; Discovery Biologics, Merck & Co., Inc., Boston, MA, USA PETER M. TESSIER • Department of Chemical Engineering, University of Michigan, Ann Arbor, MI, USA; Biointerfaces Institute, University of Michigan, Ann Arbor, MI, USA; Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI, USA; Department of Pharmaceutical Sciences, University of Michigan, Ann Arbor, MI, USA MAGDALENA TEUFL • Department of Chemistry, Institute of Biochemistry, University of Natural Resources and Life Sciences, Vienna, Austria; CD Laboratory for Next Generation CAR T Cells, Vienna, Austria MARI R. TOUMAIAN • Chemical and Materials Engineering, University of Nevada, Reno, NV, USA; Cell and Molecular Biology, University of Nevada, Reno, NV, USA

Contributors

xv

MICHAEL W. TRAXLMAYR • Department of Chemistry, Institute of Biochemistry, University of Natural Resources and Life Sciences, Vienna, Austria MITSUYOSHI UEDA • Division of Applied Life Sciences, Graduate School of Agriculture, Kyoto University, Kyoto, Japan BENJAMIN J. UMLAUF • Department of Neurosurgery, Dell Medical School, The University of Texas at Austin, Austin, TX, USA; Mulva Clinic for the Neurosciences, The University of Texas at Austin, Austin, TX, USA JAMES A. VAN DEVENTER • Chemical and Biological Engineering Department, Tufts University, Medford, MA, USA; Biomedical Engineering Department, Tufts University, Medford, MA, USA JAMES VANANTWERP • Department of Chemical Engineering and Materials Science, Michigan State University, East Lansing, MI, USA; Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, MI, USA STEFAN VOGT • acib GmbH (Austrian Centre of Industrial Biotechnology), Graz, Austria; Institute of Molecular Biotechnology, Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna (BOKU), Vienna, Austria K. DANE WITTRUP • Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA; Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA; Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA, USA DANIEL WOLDRING • Department of Chemical Engineering and Materials Science, Michigan State University, East Lansing, MI, USA; Institute for Quantitative Health Science and Engineering, Michigan State University, East Lansing, MI, USA EE LIN WONG • Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Mestre, Italy GORDANA WOZNIAK-KNOPP • Christian Doppler Laboratory for Innovative Immunotherapeutics, Institute of Molecular Biotechnology, Department of Biotechnology, University of Natural Resources and Life Sciences, Vienna (BOKU), Vienna, Austria NICOLE J. YANG • Department of Immunology, Harvard Medical School, Boston, MA, USA CHARLOTTE U. ZAJC • Department of Chemistry, Institute of Biochemistry, University of Natural Resources and Life Sciences, Vienna, Austria; CD Laboratory for Next Generation CAR T Cells, Vienna, Austria JENNIFER M. ZUPANCIC • Department of Chemical Engineering, University of Michigan, Ann Arbor, MI, USA; Biointerfaces Institute, University of Michigan, Ann Arbor, MI, USA

Part I Introduction to Yeast Surface Display and Its Applications

Chapter 1 Yeast Surface Display: New Opportunities for a Time-Tested Protein Engineering System Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder Abstract Yeast surface display has proven to be a powerful tool for the discovery of antibodies and other novel binding proteins and for engineering the affinity and selectivity of existing proteins for their targets. In the decades since the first demonstrations of the approach, the range of yeast display applications has greatly expanded to include many different protein targets and has grown to encompass methods for rapid protein characterization. Here, we briefly summarize the development of yeast display methodologies and highlight several selected examples of recent applications to timely and challenging protein engineering and characterization problems. Key words Yeast display, Library screening, Directed evolution, Protein engineering, Antibody engineering, In vitro antibody selections, Deep mutational scanning, Protein therapeutics

1

Overview of the Yeast Display Approach Directed evolution of proteins requires a relation between genotype and phenotype, a requirement frequently met by the use of display techniques, especially for engineering binding activities. Cell surface display provides a robust platform for the directed evolution of protein variants using high-throughput methods such as fluorescence-activated cell sorting (FACS) to screen highly diverse mutant libraries that can contain millions of variants, allowing rapid selection of protein variants with desired functions, such as improved binding affinity, selectivity, stability, or enzyme activity (Fig. 1a). Yeast surface display (YSD) combines the power of mutagenic library screening and directed evolution with a robust and easily manipulated eukaryotic expression host and, over the past 24 years, has become one of the dominant tools used in the practice of protein engineering. YSD offers the ability to target complex heterologous proteins that require post-translational modifications

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_1, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

3

4

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

A

B Binding and activity assays

Directed Evolution

Sequencing

FACS

Fig. 1 Schematic overview of yeast surface display for library screening and directed evolution. (a) Directed protein evolution cycle using YSD is involved in generating (mutagenesis), expressing, displaying on the yeast surface (yeast display), and screening libraries (screening), and sequencing and analysis of isolated variants. (b) The mechanism for directing a protein-of-interest (POI) to the yeast surface is sketched for a popular yeast display technology, the Aga2p fusion system. Part of the figure was created using BioRender.com

such as disulfide bond formation and glycosylation, while maintaining the genotype–phenotype linkage required for performing library screening and directed evolution [1]. Engineering of proteins using YSD benefits from eukaryotic post-translational machinery for functional expression of glycoproteins. The size of yeast cells also makes them amenable to cell sorting using FACS, enabling quantitative library screening and opening the door to simultaneous screening for multiple properties via multiparameter FACS [2]. Among the limitations of YSD are slower replication and reduced transformation efficiency compared to bacteria, which limits the maximum experimentally-accessible mutant library size compared to methods such as phage display. A wide range of anchors are available to provide both N- and C-terminal display on yeast, allowing flexibility in controlling protein orientation [3, 4]. Peptides or proteins are displayed on the surface of yeast by fusion to endogenous yeast proteins that are expressed in the mannoprotein layer of the cell wall. Several such proteins are transported to the plasma membrane as GPI-anchored species prior to being transferred to a covalent linkage between a remnant of the GPI tail and cell wall β-1,6 glucan [4–6], although other non-GPI-based anchoring partners are also possible. A number of fusion partners capable of directing the protein to be displayed to the cell wall have been demonstrated and previously reviewed [4], and a recent study used electron microscopy to demonstrate that displayed proteins can be directed to different

Continuing Acceleration of Yeast Display Applications

5

parts of the cell wall by different fusion partners [7]; however, most applications of yeast display for protein engineering to date have made use of the a-agglutinin mating adhesion receptor composed of the Aga1p and Aga2p domains (Fig. 1b) [8–14]. Aga2p assembles covalently in the secretory pathway with the GPI-anchored Aga1p domain via two disulfide bonds and is thus exported to the cell surface. Unlike single-domain GPI-based fusion partners, which require C-terminal processing to create the GPI tail, Aga2p can be successfully fused to the protein of interest via either its Nor C-terminus [8, 15–17], a feature which can be important for optimal function of the displayed protein [16, 18, 19]. The incorporation of features allowing surface expression density to be determined simultaneous to antigen or ligand binding levels represents an important feature of yeast display that enhances the quantitative nature of FACS-based library screening and avoids biasing selections toward high expressing clones [1]; this can be accomplished by immunofluorescent labeling of epitope tags [14], estimated via intracellular accumulation of truncated green fluorescent protein resulting from viral 2A peptide-induced ribosomal skipping [20, 21], or determined by dual display of a fluorescent protein at the opposite terminus of the Aga2p fusion protein [22].

2

Methodological Extensions and Developments YSD systems based on intracellular association of interacting species have been developed; initially, these demonstrations were limited to multimeric protein display, such as Fab or class II major histocompatibility complex (MHC-II), where one subunit was anchored to a yeast cell wall protein with another subunit expressed as a soluble partner [17, 23, 24]; Fab display efficiency was recently systematically optimized by using an ER retention signal on the soluble partner and simultaneous chaperone or foldase overexpression [25]. Methods using noncovalent intracellular interactions between anchored and soluble proteins in the secretory pathway as the basis for detecting target binding, referred to as co-display or the yeast surface two-hybrid (YS2H) system, were subsequently developed. Detection of interacting pairs of proteins was demonstrated by immunofluorescent labeling to identify the solubly expressed partner anchored to the surface by way of its Aga2p-fused binding partner [26] or by the use of split GFP complementation, wherein GFP fluorescence of the yeast cell was demonstrated to correlate with the strength of the interaction between the binding partners [27]. This general approach was applied to a model MHC-II that was quantitatively assessed with respect to its antigenic peptide binding preferences at a single anchor position and enabling isolation of mutant MHC-II binding to noncognate peptide [28], to

6

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

the screening of fibronectin type III domain libraries [27], and to characterizing the strength of alpha helical coiled-coil interactions and antibody–antigen interactions [27]. Several other enhancements to the detailed methods used in library generation and screening by yeast display have been developed since its inception. One noteworthy adaptation was the replacement of ligation-based cloning with the eukaryotic gap repair homologous recombination mechanism [29] for inserting gene libraries into yeast display vectors, as first demonstrated by David Kranz’s laboratory [30]; this approach simplified library DNA handling by allowing efficient, spontaneous insertion of amplified gene fragments into linearized acceptor plasmid upon co-transformation into yeast, creating circular plasmid libraries in vivo. Electroporation conditions were optimized to enhance the transformation of such DNA libraries into the yeast host [31], somewhat reducing the diversity bottleneck. While the majority of yeast display applications to isolate novel binders from libraries have used soluble fluorescent targets and FACS-based screening in at least the final rounds of enrichment, panning methods have also been developed to select binders against difficult targets such as membrane proteins or to make isolation of binders more economical and accessible by, for example, obviating the need for purified target ligand and in some cases FACS accessibility. Initially demonstrated for endothelial cells modified with a model antigen [32], panning against whole cells has been applied and optimized to isolate antibodies, T cell receptors, and other binding scaffolds such as affibodies and fibronectin domains [33– 38], in at least two cases using FACS to sort yeast-target cell complexes [35, 39]. Solubilized cell membranes, rather than whole cells, have also been used in library screening to isolate antibodies to membrane proteins and complexes [40–42] and to identify a Lyme disease-associated bacterial pathogen binder from a tick cDNA library expressed on yeast [43]. Recently, antigen targets co-expressed on yeast along with an iron oxide-binding protein (itself engineered via yeast display [44]) were used to magnetically isolate Sso7d-based and nanobody binders to a target antigen by selecting the yeast library against the magnetized antigendisplaying yeast [45], obviating the need for purified antigen. YSD has also been used to perform selections for mechanical force-bearing interactions by panning yeast subjected to controlled shearing flow over a ligand-coated surface [46]. In general, panning methods have and should continue to enhance the applicability of YSD to difficult targets. Systems allowing simple switching between surface display and soluble secretion of proteins based on amber stop codon suppression [47], biotin-avidin modification of the yeast surface and secretion of biotinylated (via intracellular biotin ligase) protein [48], or decoration of the surface with Fc-binding ZZ domains to capture secreted IgG [49, 50] have been demonstrated. In addition, a

Continuing Acceleration of Yeast Display Applications

7

system for simultaneous display and secretion of proteins based on inefficient ribosomal skipping truncation has been deployed [51]. Methods of this type can accelerate secondary screenings and initial characterization of novel proteins by eliminating the requirement for subcloning into dedicated soluble expression systems for generating purified protein. Recently, the building blocks for proteins engineered by YSD have been expanded beyond the 20 directly-encoded amino acids. Using amber stop codon suppression methodologies, non-canonical amino acids can be incorporated into yeast displayed proteins, including those bearing azide side chains enabling postexpression, biorthogonal modification reactions or side chains enabling protein photo-crosslinking [52–54]. These new methods could potentiate novel YSD applications beyond the engineering of binding functions and allow yeast display-based engineering of therapeutic proteins modified post-expression with synthetic chemistries. High-throughput screening of yeast displayed protein libraries for directed protein evolution usually require several rounds of screening of yeast cells followed by the tedious and time, cost and labor intensive process of screening colonies, and sequencing analysis [55]. An innovative method to facilitate high-throughput cell screening based on a microcapillary array platform with a laser detection strategy, called microcapillary single-cell analysis and laser extraction (μSCALE), was recently introduced and used to engineer various protein types including binding affinity of an antibody, intensity of fluorescent proteins, and an enzyme activity [56]. Recruiting the high precision laser-based technology, μSCALE offers robust spatial segregation of single cells and realtime measurement, and recovering of the microcapillary content, with five order of magnitude greater high-throughput screening capacity compared to conventional microwell plate or petri dish screening methods. Furthermore, μSCALE allows isolation and growth of each individual cell after analysis, eliminating the need for growth and screening single colonies after each highthroughput screening round in a conventional cell display based directed evolution [56]. In the μSCALE method, a library of protein variants mixed with opaque microbeads is pipetted into the array at a concentration that results in single-cell occupancy in each microcapillary well. Each well is quantified based on the fluorescence signal detection, and clones with the desired property are extracted using a laser-based extraction method and cultured. The plasmid DNA is extracted from the selected variants with improved protein function such as binding affinity or enzymatic activity and analyzed via DNA sequencing. μSCALE technique has its limitations such as requirement of complex machinery (laserbased detection), however, it provides a robust method for highthroughput screening which is flexible for adoption to various assays [57].

8

3

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

Application of Yeast Display to Antibody Engineering A few years following the initial demonstration of heterologous protein display on the yeast surface [58], YSD for protein engineering was demonstrated via affinity maturation of a single-chain Fv antibody fragment against fluorescein [1, 59]; kinetic selections were employed in this work, leading to isolation of improved scFv with measured binding half-life of days and highlighting the robustness of YSD with respect to long-term stability of both the cell and the display mechanism. In the years since, many examples of antibody affinity maturation by yeast display have been accomplished [8, 13]. Feldhaus, et al., achieved a milestone in the development of antibody engineering by YSD by demonstrating display of the first large human V gene library and recovery of primary antibodies against several diverse target antigens [60]. This study further established the effectiveness of magnetic-assisted cell sorting to pre-select libraries prior to fluorescence-activated cell sorting and demonstrated the ability to propagate and expand yeast libraries without expression bias or loss of diversity. Numerous examples of affinity maturation or primary isolation of scFv fragments, Fab fragments, and full-length IgG were demonstrated over the subsequent decade and have been previously reviewed [8, 10, 61], and the proliferation of successful yeast display antibody engineering studies addressing increasingly challenging targets has continued to date. For example, yeast display was recently applied to the challenging problem of isolating glycoform-specific antibodies [62, 63]; a novel method for cloning natively paired VH and VL genes from single B cells for screening by YSD was applied to identify Fab antibody fragments against HIV-1, influenza hemagglutinin, and Ebola virus envelope glycoprotein [64] and against whole Zika virus [65] from vaccinated or convalescent patient samples; conformationally selective nanobodies to two distinct human GPCRs were engineered by yeast display [66]; finally, highly cross-reactive antibodies against multiple inflammation-driving chemokines were isolated by YSD screening methods [67], and scFv libraries with reduced propensity for nonspecificity were generated based on screening by yeast display [68].

4

Engineering Thermostability and Secretion Efficiency Yeast offers the benefit of a eukaryotic quality control system wherein unstable and unfolded or misfolded proteins are retrotranslocated from the endoplasmic reticulum and degraded while folded proteins are secreted and display on the yeast surface [69, 70]. YSD has therefore been applied to improve the thermostability and expression of proteins using directed evolution

Continuing Acceleration of Yeast Display Applications

9

[13, 71–74]. In the initial example, single-chain T cell receptor (scTCR) mutants with improved thermal stability that correlated with their expression and display level were isolated using YSD [69]. Subsequent adaptations to this method applying increased temperatures to promote surface denaturation and/or in vivo misfolding during expression were developed [13, 71, 73, 74]. A partial list of examples of proteins that have been engineered for improved stability and/or expression include the loop-engineered Fc domain of IgG1 [74], loop-inserted green fluorescent protein [71], the epidermal growth factor receptor extracellular domain [75], and brain-derived neurotrophic factor [76]. This method has excellent and still largely untapped potential for enabling the isolation of stabilized and expressible mutants of refractory proteins for in vitro studies and structural determination, as very recently demonstrated for hematopoietic progenitor kinase [77].

5

YSD to Engineer Non-antibody Proteins Identifying highly stable scaffolds with unique structural features establishes a reliable backbone for further protein engineering to develop novel therapeutics, and yeast surface display has been utilized to engineer a number of non-antibody protein scaffolds, as has been previously reviewed [9, 13, 78]. A small selection of recent examples is summarized here. Knottin: Among these scaffolds is a cysteine knot miniprotein (knottin), which consists of three interwoven disulfide bridges—forming a pseudoknot which provides very high levels of thermal and proteolytic stability and tolerance to harsh pH [79, 80]. Knottin-based agents are particularly of interest in bioimaging [80], and multiple FDA-approved drugs have been derived from altering the surface-exposed loops of knottin [81]. Using YSD, the 28 amino acid Ecballium elaterium trypsin inhibitor was engineered to target αvβ3, αvβ5, and α5β1 integrins with high affinities, without altering cysteines responsible for the stability of knottin variants [82]. High affinity inhibitor of matriptase was also developed by engineering knottin miniprotein libraries using YSD [83]. Additionally, 28 novel CXC-tick evasin knottin scaffolds with altered chemokine-binding specificities have been created through the manipulation of the knottin scaffold [84]. Evasin contains a single antiparallel β-sheet with six conserved cysteine residues forming a disulfide-bonded knottin scaffold that creates a contiguous solvent-accessible surface [84]. Sso7d: Other non-antibody scaffold proteins that have been engineered to withstand wide ranges of pH, temperature, and chemical reagents include hyperthermophilic Sso7d scaffolds.

10

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

Sso7d scaffolds were screened using YSD against various targets from a library of mutants randomizing 10 residues of Sso7d [85]. The evolved Sso7d scaffold had melting temperatures over 89  C and showed resistance to guanidine hydrochloride [85]. Sso7d variants with affinity for silica surfaces were obtained though mutagenesis of 10 surface-exposed residues on three beta-strands and screened via YSD to isolate proteins that bind to iron oxide nanopowder [44]. Furthermore, Notch1 specific binders have been identified after screening ~108 Sso7d variants, with the most promising mutant binding to Notch1 with an equilibrium dissociation constant in the ~100 nM range [86], and nanomolar binding Sso7dbased proteins specific for tumor-associated K-Ras mutants were recently engineered by yeast display [87]. Non-antibody protein scaffolds have a plethora of uses, which can be identified and engineered all via YSD. Natural inhibitors of enzymes: Other non-antibody scaffolds engineered using YSD include natural enzyme inhibitors. Metalloproteinases (MPs) are known for their central role in remodeling of extracellular matrix (ECM). Dysregulation of MPs is related to several diseases such as cancer, neurological disorders, and cardiovascular diseases [88]. Thus, engineering their endogenous inhibitors, tissue inhibitor of metalloproteinases (TIMP), offer potential therapeutics for such diseases. TIMPs have two domains with an N-terminal region and five flexible interactive loops that resembles CDRs in heavy and light chain of antibodies [89]. They both interact with their targets through multiple discontinuous loops that pack together to form a binding surface [89]. The N-terminus motif “CXC” of TIMP has a great importance in MMP binding and inhibition, thus, TIMP was genetically fused to the N-terminus of Aga2p for displaying libraries of TIMPs for directed evolution [19]. Selective MMP-14 binders were engineered by screening a library of the N-terminal domain of TIMP-2 (N-TIMP-2) displayed on the yeast surface using FACS for high affinity and selectivity toward MMP-14, resulting in more than a 500-fold increase of selective inhibition of MMP-14 [90]. TIMPs bind to MMPs with a broad range of binding affinities. The MMP binding spectrum of N-TIMP2 were narrowed to MMP-9 and MMP-14, MMPs with significant role in breast cancer, by labeling of yeast displayed N-TIMP-2 library of mutants by soluble catalytic domains of MMP-9 and MMP-14 and screening these two labeled populations via FACS [91]. Like other protein engineering approaches, there is a tradeoff between using rational design or directed evolution. Directed evolution using yeast surface display offers high-throughput screening

Continuing Acceleration of Yeast Display Applications

11

of millions of TIMP binders without any required structural knowledge for each pair of TIMP/MP complex. However, directed evolution is limited by the size and quality (e.g., frequency and combination of mutations) of the library of TIMP variants. A rationally designed combinatorial library of TIMP-1 variants focused on the interacting residues at the interface of TIMP-1/ MMP-3 complex was generated, yeast displayed, and screened toward MMP-3 binding. The isolated TIMP-1 mutants were improved in binding and inhibition of MMP-3 catalytic domain (MMP-3cd) with low picomolar inhibition constant (Ki) [19]. Although most of the focus for engineering TIMPs were focused on the N-terminal/inhibitory domain of TIMPs, it was shown that cooperation between N- and C-terminal domains of TIMP-1 improved binding to the target MMP-3 [19]. Matriptase is another protease that activates growth factors such as hepatocyte growth factor (HGF) and thus plays a role in cancer progression when dysregulated. The hepatocyte growth factor activator inhibitor type-1 (HAI-1), a natural matriptase inhibitor, was engineered using YSD and “rationally designed” directed evolution against matriptase. A library of KD (Kunitz domain) were displayed on the yeast surface and screened for binding to matriptase using FACS. Surprisingly, it was found that a recombinant chimeric first and second Kunitz domain (KD1/KD2) showed the best binding to matriptase, and thus used for further tests. The chimeric KD1/KD2 followed by KD1 (active inhibitory domain) was fused to an antibody Fc domain (to increase valency and circulating serum half-life) and showed the best Ki of 70  5 pm for matriptase, with 120-fold improvement compared with the natural HAI-1 inhibitor [92].

6

Enzyme Engineering by YSD There is a high demand for engineering of biocatalysts, and directed evolution offers to reorganize the enzyme stability, substrate recognition, and catalytic efficiency [93]. However, there are unique challenges in engineering and designing new enzymes. In directed evolution of enzymes using YSD, display of both the enzyme library and the substrate on the cell surface is often required. The challenge in designing such system that links enzyme activity and/or selectivity to a detectable phenotype limits the use of cell surface display for enzyme evolution, similar to other cell display methods, and despite the extensive use of YSD for improving affinity and selectivity of protein binders [8], the use of YSD for engineering enzymes has remained limited. However, modified YSD strategies that link the product of enzymatic reaction to a detectable flow cytometry signal have provided robust quantitative methods to evolve enzyme activity and/or selectivity in several cases. Activity of horseradish

12

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

peroxidase (HRP) was measured and engineered using YSD via addition of a fluorescent-labeled tyrosine on the yeast surface using flow cytometry. HRP variants with higher selectivity for attaching either L- or D-tyrosinol (Alexa fluor 488-conjugated HRP substrates) were developed using yeast display and library screening by FACS [94]. YSD was also used to engineer enzymes where the enzyme reaction takes place in the yeast ER, highlighting the potential of YSD as an eukaryotic cell display platform for secreting complex proteins, and providing ER environment for engineering enzyme activity and substrate specificity without interference of other cytosolic proteases [95]. This yeast endoplasmic sequestration screening (YESS) technique was used for engineering the Tobacco Etch Virus protease (TEV-P) and similar proteases [95], where the conventional C-terminal fusion to Aga2p was used to display a protease substrate, followed by epitope tags, for detection of peptide expression and display, and a C-terminal ER retention sequence to prevent secretion of Aga2p fused proteins before protease cleavage. The engineered protease mutants with high activity and substrate selectivity were able to cleave the ER retention tag and allow the secretion and yeast display of the Aga2p fusion protein for a quantitative detection of the TEV-P activity using flow cytometry [95]. The substrate specificity of the engineered TEV was altered from Glu to Gln or His at P1 pocket with over 1000-fold improvement in selectivity. Unfortunately, the YESS strategy fails to completely control or turn off protease expression, and also contains false positives since the signal is observed through a loss of fluorescence, and thus, mutations or stop codons located within the C-terminal epitope tag or the substrate itself can be enriched via FACS. This technique was further improved, upgraded to YESS 2.0, by manipulating enzyme and substrate transcription levels and spatial sequestration. In YESS 2.0 system, a titratable protease expression was used by orthogonal β-estradiol-inducible promoter and engineered EBY100 (EBY100-Tune) with a chromosomally integrated synthetic transcription factor that enables β-estradiol induction that provides a switch off option for protease activity independently of substrate expression to be used in negative screening rounds [96]. The YSD platform was further modified to enable dual protein display on the yeast surface to engineer a transpeptidase such as S. aureus sortase A (SrtA) with two substrates (LPETG, multiglycine). A library of SrtA variants was cloned to the C-terminus of Aga2p and one of the SrtA substrates (LPETG) was enzymatically linked to Aga1p using a reactive peptide handle. Active SrtA mutants displayed on the yeast surface via Aga2p were able to ligate the soluble biotinyalted SrtA substrate (triglycine) to the LPETG peptide anchored to Aga1p. SrtA activity yielding biotin-tagged triglycine binding was then measured using streptavidin-

Continuing Acceleration of Yeast Display Applications

13

conjugated fluorophores and flow cytometry. SrtA variants with improved activity, kcat and/or Km, were screened using FACS with up to 140-fold in improvement in enzyme activity [97]. Another dual display of enzyme and substrate was developed by fusion of enzyme and GFP-LPETG to the two termini of the yeast surface anchor (Aga2p). Simultaneous fusion of a fluorescent protein to Aga2p also eliminates the need for immunolabeling to measure protein expression level using flow cytometry [22].

7

Applications in T Cell Antigen Presentation and Yeast-Based Vaccines In addition to its long-term use in T cell receptor engineering [13], yeast display has been applied to address questions related to T cell antigen presentation. MHC-II proteins were displayed on yeast [17], and yeast displaying stimulatory ligands were demonstrated to activate T cells in vitro [98]. These observations were extended to screen MHC-complexed peptide libraries from known antigenic proteins or whole organisms to identify peptides stimulating specific T cells [99, 100]. Recently, peptide-MHC and pMHC/ ICAM-1 (intercellular adhesion molecule-1) fusions to dockerin were applied to yeast displaying Aga2p-cohesin fusions, affording control of T cell ligand surface density and assessment of spatial effects on T cell stimulation by yeast acting as artificial antigen presenting cells [101]. Random peptide libraries displayed on yeast in the context of MHC-II were also applied to screening based on soluble T cell receptor (TCR) binding; in combination with bioinformatic analysis, this approach enabled rapid elucidation of TCR specificity [102, 103] and identification of antigens for T cells of unknown specificity [104]. Additional yeast display methods have been employed to assess the antigenic peptide binding specificity of MHC-II. In one case, conditional display of MHC-II based on its intracellular binding to peptide was used to determine specificity preferences at a key position (the P1 anchor position) for a model peptide/MHC-II interaction and to isolate MHC-II mutants capable of binding to noncognate peptide [28]; this simple system yielded strikingly quantitative results comparable to biochemical assays conducted with purified species. This concept was recently extended by the addition of HLA-DM-mediated peptide exchange catalysis and acidic conditions in an effort to better mimic antigen loading in antigen presenting cells, and deep sequencing analysis of very large peptide libraries yielded previously unmatched, high quality datasets suggesting the ability to significantly improve peptide binding prediction algorithms [105]. Yeast are natural adjuvants and have long been of interest as vehicles for vaccine delivery, and YSD has been employed in recent years in ongoing pursuit of this goal [106]. Yeast displaying relevant

14

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

pathogen antigens have shown promise recently in inducing immune responses against Toxoplasma gondii [107], Candida albicans [108], Porcine circovirus type 2 [109], and influenza A virus [110, 111] in animal models, including when delivered as oral vaccines in these examples. Use of yeast displaying immunogen proteins for vaccine design is considered an area poised for significant future growth [106].

8

Screening cDNA or Natural Protein Libraries YSD has potentiated several lines of research exploring ligand or whole cell interactions with unknown protein partners. Eukaryotic yeast expression affords the opportunity to express high quality libraries of post-translationally-modified eukaryotic proteins for screening against potential binding partners. This approach has been used to screen proteins or peptides expressed from random libraries created from cDNA cloned into surface display vectors [43, 112, 113], as well as smaller libraries of designated proteins such as putative tick evasins identified using bioinformatics [114, 115] or a defined set of secreted human proteins numbering in the thousands [116]. Screening such libraries by FACS, panning against target cells, or a combination of these enabled identification of tick [43] and human [116] proteins mediating interactions with Lyme-disease-causing Borrelia burgdorferi, inflammationmodulating tick evasins binding to specific chemokines [114, 115], post-translationally-modified peptides [113], and phospholipids [113].

9

Combining YSD and Deep Sequencing A chance for a number of potential mutants to get eliminated in several rounds of screening in directed evolution of YSD protein variants always exists. Deep mutational scanning or next generation sequencing (NGS) techniques provide a more stringent survey through amino acid mutations and allows detection of individual mutations responsible for subtle changes in desired property of protein mutants, which broadens the fitness landscape and decreases the chances of mutant clones being missed in several rounds of high-throughput screening due to the low expression of protein variants or other technical limits that overlooks small changes in protein properties such as binding or activity [117– 119]. Using deep mutational scanning in combination with YSD can bypass the need for several rounds of screening that might eliminate binders with higher affinity due to their low expression or display level.

Continuing Acceleration of Yeast Display Applications

15

A combination of NGS, yeast surface display, and FACS were recently used to improve binding affinity of a meditope cyclic peptide for binding to the specific antigen-binding pocket of the therapeutic antibody cetuximab, which in turn improved cetuximab-mediated targeting of EGFR-overexpressing tumor cells [120], guiding mutagenesis to engineer TCR/peptide-MHC interaction affinity [121], rapidly mapping antibody conformational epitopes [122], studying sequence-function relationships for protein-protein interactions [123], and generating quantitative binding landscapes [124]. Conformational epitopes on protein binders can be determined using these high-throughput screening methods with exquisite detail [123, 125]. Recently, CRISPR/ Cas9-based mutagenesis was combined with yeast display library screening and deep sequencing to map mutational effects on scFv antibody affinity [126]. Bioinformatic tools and analysis typically play an important role in all of these methods [117]. Considering dramatic growth in DNA sequencing and analysis technology, it is expected that NGS will be used more widely with yeast displayed libraries of mutants going forward.

10

Yeast Display in the Rapid Response to the SARS-CoV-2 Pandemic The emergence of COVID-19 in 2020 induced many labs across the world to quickly pivot and apply their expertise and tools to develop better understanding of the SARS-CoV-2 virus and to identify and develop solutions to its control. Yeast display is among those tools that have been rapidly brought to bear in this context. In one straightforward use of YSD, a library of rationally designed human scFv antibody fragments was screened for binding to the SARS-CoV-2 spike (S) protein receptor binding domain (RBD), including a final screening step where clones were isolated based on successful competitive binding to RBD in the presence of its native receptor (ACE2). Full-size IgG1 antibody constructed using one of these variable domains demonstrated both inhibition of viral infection in a cell culture model and improvement in some clinical manifestations of disease in a Syrian hamster model [127]. Similarly, a large yeast displayed nanobody library (singledomain antibodies based on those found in camelid species) was screened to identify ACE2-competitive binders to an S protein extracellular domain mutagenically-stabilized in its pre-activation conformation. Some of these nanobodies retained binding to S in the presence of isolated RBD (indicating they targeted regions outside the RBD) yet inhibited SARS-CoV-2 infection in cell culture model, putatively by stabilizing S against activation. One such antibody was further affinity matured by yeast display and potently blocked infection when formatted as a trimer [128]. Another recent study used yeast display to affinity mature three antibodies

16

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

isolated from a recovered SARS patient (infected by SARS-CoV-1 in 2003) that demonstrated some neutralizing activity against similar coronaviruses. The most potent mutant demonstrated SARSCoV-2 neutralization in a cell culture model that was equal to or better than that of antibodies currently in clinical use, while also displaying superior resistance to loss of S protein RBD binding due to RBD mutations found in natural SARS-CoV-2 variants, significantly broader reactivity with RBDs from a wide range of the sarbecovirus subgenus, and efficacy as both a prophylactic or therapeutic agent in animal models of SARS and COVID-19 [129]. Taking another approach, a combination of deep mutational scanning, computational design, and yeast display affinity maturation was applied to rapidly generate mutant ACE2 extracellular domains with substantially increased affinity for the SARS-CoV2 RBD. Several of these engineered ACE2 receptors were effective blockers of viral infection in cell culture models [130]. A trio of recent studies applied yeast display and deep mutational scanning to comprehensively assess mutational effects on proteins critical to SARS-CoV-2 infection. One study mapped sites on the RBD at which mutations perturb folding/expression and/or ACE2 binding and therefore represent potential targets for therapeutics resistant to escape mutants [131]. Another study identified mutations conferring reduced binding to three monoclonal antibodies in clinical use for severe COVID-19 cases [132]; the third study similarly identified mutations yielding escape from antibodies in plasma from convalescent COVID-19 patients [133]. In each of these examples, YSD represented a critical experimental tool for rapidly generating large datasets that enabled generation of powerful insight and potentially potent reagents.

11

Future Prospects As yeast display approaches three decades of age, what remains clear is that the method has enabled an impressive breadth of characterization and engineering of proteins—their properties and especially their interactions with other species. This chapter touches on only some of the many developments to date. Going forward, we anticipate seeing new methods developed to promote YSD’s use in some of the exciting areas with plenty of room for growth, such as enzyme engineering, development of optimized receptors for cellbased therapies such as CAR T cells, applications of affinity maturation methods to stabilize multimeric complexes for structural characterization (as used for interleukin-10 and its receptor [134, 135]), and investigations of yeast as vaccine vehicles, among others. In the near term, we expect to see a proliferation of YSD and deep mutational screening applications to

Continuing Acceleration of Yeast Display Applications

17

comprehensively characterize mutational impacts on functional protein properties, and shed light on protein sequence–structure– function relation addressing a diversity of applications, no doubt in addition to plenty of clever and novel adaptations and improvements.

Acknowledgments Partial support for this work was provided by the National Institute of General Medical Sciences of the National Institutes of Health under award number P20GM103650 (M.R.S.) and R15GM122326 (E.T.B.). Content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. References 1. Boder ET, Wittrup KD (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol 15(6):553–557. https://doi.org/10.1038/nbt0697-553 2. Schro¨ter C, Beck J, Krah S, Zielonka S, Doerner A, Rhiel L, Gu¨nther R, Toleikis L, Kolmar H, Hock B, Becker S (2018) Selection of antibodies with tailored properties by application of high-throughput multiparameter fluorescence-activated cell sorting of yeastdisplayed immune libraries. Mol Biotechnol 60(10):727–735. https://doi.org/10.1007/ s12033-018-0109-0 3. Smith MR, Khera E, Wen F (2015) Engineering novel and improved biocatalysts by cell surface display. Ind Eng Chem Res 54(16): 4021–4032. https://doi.org/10.1021/ ie504071f 4. Andreu C, Del Olmo ML (2018) Yeast arming systems: pros and cons of different protein anchors and other elements required for display. Appl Microbiol Biotechnol 102(6): 2543–2561. https://doi.org/10.1007/ s00253-018-8827-6 5. Klis FM, Boorsma A, De Groot PW (2006) Cell wall construction in Saccharomyces cerevisiae. Yeast 23(3):185–202. https://doi. org/10.1002/yea.1349 6. Orlean P (2012) Architecture and biosynthesis of the Saccharomyces cerevisiae cell wall. Genetics 192(3):775–818. https://doi.org/ 10.1534/genetics.112.144485 7. Inokuma K, Kurono H, den Haan R, van Zyl WH, Hasunuma T, Kondo A (2020) Novel strategy for anchorage position control of GPI-attached proteins in the yeast cell wall

using different GPI-anchoring domains. Metab Eng 57:110–117. https://doi.org/ 10.1016/j.ymben.2019.11.004 8. Boder ET, Raeeszadeh-Sarmazdeh M, Price JV (2012) Engineering antibodies by yeast display. Arch Biochem Biophys 526(2): 99–106. https://doi.org/10.1016/j.abb. 2012.03.009 9. Gera N, Hussain M, Rao BM (2013) Protein selection using yeast surface display. Methods 60(1):15–26. https://doi.org/10.1016/j. ymeth.2012.03.014 10. Pepper LR, Cho YK, Boder ET, Shusta EV (2008) A decade of yeast surface display technology: where are we now? Comb Chem High Throughput Screen 11(2):127–134. h t t p s : // d o i . o r g / 1 0 . 2 1 7 4 / 138620708783744516 11. Smith SN, Harris DT, Kranz DM (2015) T cell receptor engineering and analysis using the yeast display platform. Methods Mol Biol 1319:95–141. https://doi.org/10.1007/ 978-1-4939-2748-7_6 12. Stone JD, Chervin AS, Aggen DH, Kranz DM (2012) T cell receptor engineering. Methods Enzymol 503:189–222. https:// doi.org/10.1016/b978-0-12-396962-0. 00008-2 13. Cherf GM, Cochran JR (2015) Applications of yeast surface display for protein engineering. Methods Mol Biol 1319:155–175. https://doi.org/10.1007/978-1-49392748-7_8 14. Chao G, Lau WL, Hackel BJ, Sazinsky SL, Lippow SM, Wittrup KD (2006) Isolating

18

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

and engineering human antibodies using yeast surface display. Nat Protoc 1(2):755–768. https://doi.org/10.1038/nprot.2006.94 15. Brophy SE, Holler PD, Kranz DM (2003) A yeast display system for engineering functional peptide-MHC complexes. J Immunol Methods 272(1–2):235–246. https://doi. org/10.1016/s0022-1759(02)00439-8 16. Wang Z, Mathias A, Stavrou S, Neville DM Jr (2005) A new yeast display vector permitting free scFv amino termini can augment ligand binding affinities. Protein Eng Des Sel 18(7): 337–343. https://doi.org/10.1093/pro tein/gzi036 17. Boder ET, Bill JR, Nields AW, Marrack PC, Kappler JW (2005) Yeast surface display of a noncovalent MHC class II heterodimer complexed with antigenic peptide. Biotechnol Bioeng 92(4):485–491. https://doi.org/10. 1002/bit.20616 18. Uchan´ski T, Zo¨gg T, Yin J, Yuan D, Wohlko¨nig A, Fischer B, Rosenbaum DM, Kobilka BK, Pardon E, Steyaert J (2019) An improved yeast surface display platform for the screening of nanobody immune libraries. Sci Rep 9(1):382. https://doi.org/10.1038/ s41598-018-37212-3 19. Raeeszadeh-Sarmazdeh M, Greene KA, Sankaran B, Downey GP, Radisky DC, Radisky ES (2019) Directed evolution of the metalloproteinase inhibitor TIMP-1 reveals that its N- and C-terminal domains cooperate in matrix metalloproteinase recognition. J Biol Chem 294(24):9476–9488. https:// doi.org/10.1074/jbc.RA119.008321 20. Grzeschik J, Hinz SC, Ko¨nning D, Pirzer T, Becker S, Zielonka S, Kolmar H (2017) A simplified procedure for antibody engineering by yeast surface display: coupling display levels and target binding by ribosomal skipping. Biotechnol J 12(2):1600454. https://doi. org/10.1002/biot.201600454 21. Hinz SC, Elter A, Grzeschik J, Habermann J, Becker B, Kolmar H (2020) Simplifying the detection of surface presentation levels in yeast surface display by intracellular tGFP expression. Methods Mol Biol 2070: 211–222. https://doi.org/10.1007/978-14939-9853-1_12 22. Lim S, Glasgow JE, Filsinger Interrante M, Storm EM, Cochran JR (2017) Dual display of proteins on the yeast cell surface simplifies quantification of binding interactions and enzymatic bioconjugation reactions. Biotechnol J 12(5). https://doi.org/10.1002/biot. 201600696 23. Lin Y, Tsumuraya T, Wakabayashi T, Shiraga S, Fujii I, Kondo A, Ueda M (2003)

Display of a functional hetero-oligomeric catalytic antibody on the yeast cell surface. Appl Microbiol Biotechnol 62(2–3):226–232. https://doi.org/10.1007/s00253-0031283-x 24. van den Beucken T, Pieters H, Steukers M, van der Vaart M, Ladner RC, Hoogenboom HR, Hufton SE (2003) Affinity maturation of Fab antibody fragments by fluorescentactivated cell sorting of yeast-displayed libraries. FEBS Lett 546(2–3):288–294. https://doi.org/10.1016/s0014-5793(03) 00602-1 25. Mei M, Li J, Wang S, Lee KB, Iverson BL, Zhang G, Ge X, Yi L (2019) Prompting Fab yeast surface display efficiency by ER retention and molecular chaperon co-expression. Front Bioeng Biotechnol 7:362. https://doi.org/ 10.3389/fbioe.2019.00362 26. Dutta S, Koide A, Koide S (2008) Highthroughput analysis of the protein sequencestability landscape using a quantitative yeast surface two-hybrid system and fragment reconstitution. J Mol Biol 382(3):721–733. https://doi.org/10.1016/j.jmb.2008. 07.036 27. Hu X, Kang S, Chen X, Shoemaker CB, Jin MM (2009) Yeast surface two-hybrid for quantitative in vivo detection of proteinprotein interactions via the secretory pathway. J Biol Chem 284(24):16369–16376. https:// doi.org/10.1074/jbc.M109.001743 28. Jiang W, Boder ET (2010) High-throughput engineering and analysis of peptide binding to class II MHC. Proc Natl Acad Sci U S A 107(30):13258–13263. https://doi.org/10. 1073/pnas.1006344107 29. Raymond CK, Pownder TA, Sexson SL (1999) General method for plasmid construction using homologous recombination. BioTechniques 26(1):134–138, 140-131. https://doi.org/10.2144/99261rr02 30. Starwalt SE, Masteller EL, Bluestone JA, Kranz DM (2003) Directed evolution of a single-chain class II MHC product by yeast display. Protein Eng 16(2):147–156. https:// doi.org/10.1093/proeng/gzg018 31. Benatuil L, Perez JM, Belk J, Hsieh CM (2010) An improved yeast transformation method for the generation of very large human antibody libraries. Protein Eng Des Sel 23(4):155–159. https://doi.org/10. 1093/protein/gzq002 32. Wang XX, Shusta EV (2005) The use of scFvdisplaying yeast in mammalian cell surface selections. J Immunol Methods 304(1): 30–42. https://doi.org/10.1016/j.jim. 2005.05.006

Continuing Acceleration of Yeast Display Applications 33. Wang XX, Cho YK, Shusta EV (2007) Mining a yeast library for brain endothelial cellbinding antibodies. Nat Methods 4(2): 143–145. https://doi.org/10.1038/ nmeth993 34. Stern LA, Lown PS, Kobe AC, AbouElkacem L, Willmann JK, Hackel BJ (2019) Cellular-based selections aid yeast-display discovery of genuine cell-binding ligands: targeting oncology vascular biomarker CD276. ACS Comb Sci 21(3):207–222. https://doi. org/10.1021/acscombsci.8b00156 35. Yang Z, Wan Y, Tao P, Qiang M, Dong X, Lin CW, Yang G, Zheng T, Lerner RA (2019) A cell-cell interaction format for selection of high-affinity antibodies to membrane proteins. Proc Natl Acad Sci U S A 116(30): 14971–14978. https://doi.org/10.1073/ pnas.1908571116 36. Zorniak M, Clark PA, Umlauf BJ, Cho Y, Shusta EV, Kuo JS (2017) Yeast display biopanning identifies human antibodies targeting glioblastoma stem-like cells. Sci Rep 7(1): 15840. https://doi.org/10.1038/s41598017-16066-1 37. Richman SA, Healan SJ, Weber KS, Donermeyer DL, Dossett ML, Greenberg PD, Allen PM, Kranz DM (2006) Development of a novel strategy for engineering high-affinity proteins by yeast display. Protein Eng Des Sel 19(6):255–264. https://doi.org/10. 1093/protein/gzl008 38. Stern LA, Lown PS, Hackel BJ (2020) Ligand engineering via yeast surface display and adherent cell panning. Methods Mol Biol 2070:303–320. https://doi.org/10.1007/ 978-1-4939-9853-1_17 39. Bogen JP, Storka J, Yanakieva D, Fiebig D, Grzeschik J, Hock B, Kolmar H (2020) Isolation of common light chain antibodies from immunized chickens using yeast biopanning and fluorescence-activated cell sorting. Biotechnol J 16:e2000240. https://doi.org/10. 1002/biot.202000240 40. Lajoie JM, Cho YK, Frost D, Bremner S, Li L, Shusta EV (2019) A yeast display immunoprecipitation screen for targeted discovery of antibodies against membrane protein complexes. Protein Eng Des Sel 32(5):219–230. https://doi.org/10.1093/protein/gzz035 41. Tillotson BJ, Cho YK, Shusta EV (2013) Cells and cell lysates: a direct approach for engineering antibodies against membrane proteins using yeast surface display. Methods 60(1):27–37. https://doi.org/10.1016/j. ymeth.2012.03.010 42. Tillotson BJ, Lajoie JM, Shusta EV (2015) Yeast display-based antibody affinity

19

maturation using detergent-solubilized cell lysates. Methods Mol Biol 1319:65–78. https://doi.org/10.1007/978-1-49392748-7_4 43. Narasimhan S, Coumou J, Schuijt TJ, Boder E, Hovius JW, Fikrig E (2014) A tick gut protein with fibronectin III domains aids Borrelia burgdorferi congregation to the gut during transmission. PLoS Pathog 10(8): e1004278. https://doi.org/10.1371/jour nal.ppat.1004278 44. Cruz-Teran CA, Bacon K, McArthur N, Rao BM (2018) An engineered Sso7d variant enables efficient magnetization of yeast cells. ACS Comb Sci 20(10):579–584. https://doi. org/10.1021/acscombsci.8b00084 45. Bacon K, Burroughs M, Blain A, Menegatti S, Rao BM (2019) Screening yeast display libraries against magnetized yeast cell targets enables efficient isolation of membrane protein binders. ACS Comb Sci 21(12): 817–832. https://doi.org/10.1021/ acscombsci.9b00147 46. Pepper LR, Parthasarathy R, Robbins GP, Dang NN, Hammer DA, Boder ET (2013) Isolation of αL I domain mutants mediating firm cell adhesion using a novel flow-based sorting method. Protein Eng Des Sel 26(8): 515–521. https://doi.org/10.1093/pro tein/gzt028 47. Van Deventer JA, Kelly RL, Rajan S, Wittrup KD, Sidhu SS (2015) A switchable yeast display/secretion system. Protein Eng Des Sel 28(10):317–325. https://doi.org/10.1093/ protein/gzv043 48. Rakestraw JA, Aird D, Aha PM, Baynes BM, Lipovsek D (2011) Secretion-and-capture cell-surface display for selection of targetbinding proteins. Protein Eng Des Sel 24(6): 525–530. https://doi.org/10.1093/pro tein/gzr008 49. Krah S, Gu¨nther R, Becker S, Zielonka S, Rhiel L (2020) Chemical modification of the yeast cell surface allows the switch between display and soluble secretion of full-length antibodies. Methods Mol Biol 2070: 335–349. https://doi.org/10.1007/978-14939-9853-1_19 50. Rhiel L, Krah S, Gunther R, Becker S, Kolmar H, Hock B (2014) REAL-select: full-length antibody display and library screening by surface capture on yeast cells. PLoS One 9(12):e114887. https://doi.org/ 10.1371/journal.pone.0114887 51. Cruz-Teran CA, Tiruthani K, Mischler A, Rao BM (2017) Inefficient ribosomal skipping enables simultaneous secretion and display of proteins in Saccharomyces cerevisiae. ACS

20

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

Synth Biol 6(11):2096–2107. https://doi. org/10.1021/acssynbio.7b00144 52. Islam M, Kehoe HP, Lissoos JB, Huang M, Ghadban CE, Berumen Sa´nchez G, Lane HZ, Van Deventer JA (2021) Chemical diversification of simple synthetic antibodies. ACS Chem Biol 16(2):344–359. https://doi.org/ 10.1021/acschembio.0c00865 53. Stieglitz JT, Kehoe HP, Lei M, Van Deventer JA (2018) A robust and quantitative reporter system to evaluate noncanonical amino acid incorporation in yeast. ACS Synth Biol 7(9): 2256–2269. https://doi.org/10.1021/ acssynbio.8b00260 54. Van Deventer JA, Le DN, Zhao J, Kehoe HP, Kelly RL (2016) A platform for constructing, evaluating, and screening bioconjugates on the yeast surface. Protein Eng Des Sel 29(11):485–494. https://doi.org/10.1093/ protein/gzw029 55. Packer MS, Liu DR (2015) Methods for the directed evolution of proteins. Nat Rev Genet 16(7):379–394. https://doi.org/10.1038/ nrg3927 56. Lim S, Chen B, Kariolis MS, Dimov IK, Baer TM, Cochran JR (2017) Engineering high affinity protein-protein interactions using a high-throughput microcapillary array platform. ACS Chem Biol 12(2):336–341. https://doi.org/10.1021/acschembio. 6b00794 57. Chen B, Lim S, Kannan A, Alford SC, Sunden F, Herschlag D, Dimov IK, Baer TM, Cochran JR (2016) High-throughput analysis and protein engineering using microcapillary arrays. Nat Chem Biol 12(2):76–81. https://doi.org/10.1038/nchembio.1978 58. Schreuder MP, Brekelmans S, Van Den Ende H, Klis FM (1993) Targeting of a heterologous protein to the cell wall of Saccharomyces cerevisiae. Yeast 9(4):399–409. https://doi.org/10.1002/yea.320090410 59. Boder ET, Midelfort KS, Wittrup KD (2000) Directed evolution of antibody fragments with monovalent femtomolar antigenbinding affinity. Proc Natl Acad Sci U S A 97(20):10701–10705. https://doi.org/10. 1073/pnas.170297297 60. Feldhaus MJ, Siegel RW, Opresko LK, Coleman JR, Feldhaus JM, Yeung YA, Cochran JR, Heinzelman P, Colby D, Swers J, Graff C, Wiley HS, Wittrup KD (2003) Flowcytometric isolation of human antibodies from a nonimmune Saccharomyces cerevisiae surface display library. Nat Biotechnol 21(2): 163–170. https://doi.org/10.1038/nbt785

61. Boder ET, Jiang W (2011) Engineering antibodies for cancer therapy. Annu Rev Chem Biomol Eng 2:53–75. https://doi.org/10. 1146/annurev-chembioeng061010-114142 62. Lu Z, Kamat K, Johnson BP, Yin CC, Scholler N, Abbott KL (2019) Generation of a fully human scFv that binds tumor-specific glycoforms. Sci Rep 9(1):5101. https://doi. org/10.1038/s41598-019-41567-6 63. Sharma P, Marada V, Cai Q, Kizerwetter M, He Y, Wolf SP, Schreiber K, Clausen H, Schreiber H, Kranz DM (2020) Structureguided engineering of the affinity and specificity of CARs against Tn-glycopeptides. Proc Natl Acad Sci U S A 117(26):15148–15159. h t t p s : // d o i . o r g / 1 0 . 1 0 7 3 / p n a s . 1920662117 64. Wang B, DeKosky BJ, Timm MR, Lee J, Normandin E, Misasi J, Kong R, McDaniel JR, Delidakis G, Leigh KE, Niezold T, Choi CW, Viox EG, Fahad A, Cagigi A, Ploquin A, Leung K, Yang ES, Kong WP, Voss WN, Schmidt AG, Moody MA, Ambrozak DR, Henry AR, Laboune F, Ledgerwood JE, Graham BS, Connors M, Douek DC, Sullivan NJ, Ellington AD, Mascola JR, Georgiou G (2018) Functional interrogation and mining of natively paired human V(H):V(L) antibody repertoires. Nat Biotechnol 36(2):152–155. https://doi.org/10.1038/nbt.4052 65. Fahad AS, Timm MR, Madan B, Burgomaster KE, Dowd KA, Normandin E, Gutie´rrezGonza´lez MF, Pennington JM, De Souza MO, Henry AR, Laboune F, Wang L, Ambrozak DR, Gordon IJ, Douek DC, Ledgerwood JE, Graham BS, Castilho LR, Pierson TC, Mascola JR, DeKosky BJ (2021) Functional profiling of antibody immune repertoires in convalescent zika virus disease patients. Front Immunol 12:615102. https://doi.org/10. 3389/fimmu.2021.615102 66. McMahon C, Baier AS, Pascolutti R, Wegrecki M, Zheng S, Ong JX, Erlandson SC, Hilger D, Rasmussen SGF, Ring AM, Manglik A, Kruse AC (2018) Yeast surface display platform for rapid discovery of conformationally selective nanobodies. Nat Struct Mol Biol 25(3):289–296. https://doi.org/ 10.1038/s41594-018-0028-6 67. Angelini A, Miyabe Y, Newsted D, Kwan BH, Miyabe C, Kelly RL, Jamy MN, Luster AD, Wittrup KD (2018) Directed evolution of broadly crossreactive chemokine-blocking antibodies efficacious in arthritis. Nat Commun 9(1):1461–1461. https://doi.org/10. 1038/s41467-018-03687-x

Continuing Acceleration of Yeast Display Applications 68. Kelly RL, Le D, Zhao J, Wittrup KD (2018) Reduction of nonspecificity motifs in synthetic antibody libraries. J Mol Biol 430(1): 119–130. https://doi.org/10.1016/j.jmb. 2017.11.008 69. Shusta EV, Kieke MC, Parke E, Kranz DM, Wittrup KD (1999) Yeast polypeptide fusion surface display levels predict thermal stability and soluble secretion efficiency. J Mol Biol 292(5):949–956. https://doi.org/10.1006/ jmbi.1999.3130 70. Raeeszadeh-Sarmazdeh M, Patel N, Cruise S, Owen L, O’Neill H, Boder ET (2018) Identifying stable fragments of arabidopsis thaliana cellulose synthase subunit 3 by yeast display. Biotechnol J 14:e1800353. https://doi.org/ 10.1002/biot.201800353 71. Pavoor TV, Wheasler JA, Kamat V, Shusta EV (2012) An enhanced approach for engineering thermally stable proteins using yeast display. Protein Eng Des Sel 25(10):625–630. https://doi.org/10.1093/protein/gzs041 72. Shusta EV, Holler PD, Kieke MC, Kranz DM, Wittrup KD (2000) Directed evolution of a stable scaffold for T-cell receptor engineering. Nat Biotechnol 18(7):754–759. https://doi. org/10.1038/77325 73. Traxlmayr MW, Shusta EV (2017) Directed evolution of protein thermal stability using yeast surface display. Methods Mol Biol 1575:45–65. https://doi.org/10.1007/ 978-1-4939-6857-2_4 74. Traxlmayr MW, Faissner M, Stadlmayr G, Hasenhindl C, Antes B, Ru¨ker F, Obinger C (2012) Directed evolution of stabilized IgG1Fc scaffolds by application of strong heat shock to libraries displayed on yeast. Biochim Biophys Acta 1824(4):542–549. https://doi. org/10.1016/j.bbapap.2012.01.006 75. Kim YS, Bhandari R, Cochran JR, Kuriyan J, Wittrup KD (2006) Directed evolution of the epidermal growth factor receptor extracellular domain for expression in yeast. Proteins 62(4):1026–1035. https://doi.org/10. 1002/prot.20618 76. Burns ML, Malott TM, Metcalf KJ, Hackel BJ, Chan JR, Shusta EV (2014) Directed evolution of brain-derived neurotrophic factor for improved folding and expression in Saccharomyces cerevisiae. Appl Environ Microbiol 80(18):5732–5742. https://doi.org/ 10.1128/aem.01466-14 77. Lau WL, Pearce B, Malakian H, Rodrigo I, Xie D, Gao M, Marsilio F, Chang C, Ruzanov M, Muckelbauer JK, Newitt JA, Lipovsˇek D, Sheriff S (2021) Using yeast surface display to engineer a soluble and crystallizable construct of hematopoietic progenitor

21

kinase 1 (HPK1). Acta Crystallogr F Struct Biol Commun 77(Pt 1):22–28. https://doi. org/10.1107/s2053230x20016015 78. Ko¨nning D, Kolmar H (2018) Beyond antibody engineering: directed evolution of alternative binding scaffolds and enzymes using yeast surface display. Microb Cell Factories 17(1):32. https://doi.org/10.1186/ s12934-018-0881-3 79. Postic G, Gracy J, Perin C, Chiche L, Gelly JC (2018) KNOTTIN: the database of inhibitor cystine knot scaffold after 10 years, toward a systematic structure modeling. Nucleic Acids Res 46(D1):D454–D458. https://doi.org/ 10.1093/nar/gkx1084 80. Moore SJ, Cochran JR (2012) Engineering knottins as novel binding agents. Methods Enzymol 503:223–251. https://doi.org/10. 1016/B978-0-12-396962-0.00009-4 81. Simeon R, Chen Z (2018) In vitroengineered non-antibody protein therapeutics. Protein Cell 9(1):3–14. https://doi. org/10.1007/s13238-017-0386-6 82. Kimura RH, Cheng Z, Gambhir SS, Cochran JR (2009) Engineered knottin peptides: a new class of agents for imaging integrin expression in living subjects. Cancer Res 69(6): 2435–2442. https://doi.org/10.1158/ 0008-5472.CAN-08-2495 83. Glotzbach B, Reinwarth M, Weber N, Fabritz S, Tomaszowski M, Fittler H, Christmann A, Avrutina O, Kolmar H (2013) Combinatorial optimization of cystine-knot peptides towards high-affinity inhibitors of human matriptase-1. PLoS One 8(10):e76956. https://doi.org/10.1371/ journal.pone.0076956 84. Lee AW, Deruaz M, Lynch C, Davies G, Singh K, Alenazi Y, Eaton JRO, Kawamura A, Shaw J, Proudfoot AEI, Dias JM, Bhattacharya S (2019) A knottin scaffold directs the CXC-chemokine-binding specificity of tick evasins. J Biol Chem 294(29): 11199–11212. https://doi.org/10.1074/ jbc.RA119.008817 85. Gera N, Hussain M, Wright RC, Rao BM (2011) Highly stable binding proteins derived from the hyperthermophilic Sso7d scaffold. J Mol Biol 409(4):601–616. https://doi.org/ 10.1016/j.jmb.2011.04.020 86. Gocha T, Rao BM, DasGupta R (2017) Identification and characterization of a novel Sso7d scaffold-based binder against Notch1. Sci Rep 7:12021. https://doi.org/10.1038/ s41598-017-12246-1; ARTN 12021 87. Kauke MJ, Traxlmayr MW, Parker JA, Kiefer JD, Knihtila R, McGee J, Verdine G,

22

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

Mattos C, Wittrup KD (2017) An engineered protein antagonist of K-Ras/B-Raf interaction. Sci Rep 7(1):5831. https://doi.org/ 10.1038/s41598-017-05889-7 88. Raeeszadeh-Sarmazdeh M, Do LD, Hritz BG (2020) Metalloproteinases and their inhibitors: potential for the development of new therapeutics. Cell 9(5):1313. https://doi. org/10.3390/cells9051313 89. Radisky ES, Raeeszadeh-Sarmazdeh M, Radisky DC (2017) Therapeutic potential of matrix metalloproteinase inhibition in breast cancer. J Cell Biochem 118(11):3531–3548. https://doi.org/10.1002/jcb.26185 90. Arkadash V, Yosef G, Shirian J, Cohen I, Horev Y, Grossman M, Sagi I, Radisky ES, Shifman JM, Papo N (2017) Development of high affinity and high specificity inhibitors of matrix metalloproteinase 14 through computational design and directed evolution. J Biol Chem 292(8):3481–3495. https://doi. org/10.1074/jbc.M116.756718 91. Shirian J, Arkadash V, Cohen I, Sapir T, Radisky ES, Papo N, Shifman JM (2018) Converting a broad matrix metalloproteinase family inhibitor into a specific inhibitor of MMP-9 and MMP-14. FEBS Lett 592(7): 1122–1134. https://doi.org/10.1002/ 1873-3468.13016 92. Mitchell AC, Kannan D, Hunter SA, Parra Sperberg RA, Chang CH, Cochran JR (2018) Engineering a potent inhibitor of matriptase from the natural hepatocyte growth factor activator inhibitor type-1 (HAI-1) protein. J Biol Chem 293(14): 4969–4980. https://doi.org/10.1074/jbc. M117.815142 93. Chen K, Arnold FH (2020) Engineering new catalytic activities in enzymes. Nat Catal 3(3): 203–213. https://doi.org/10.1038/ s41929-019-0385-5 94. Lipovsek D, Antipov E, Armstrong KA, Olsen MJ, Klibanov AM, Tidor B, Wittrup KD (2007) Selection of horseradish peroxidase variants with enhanced enantioselectivity by yeast surface display. Chem Biol 14(10): 1176–1185. https://doi.org/10.1016/j. chembiol.2007.09.008 95. Yi L, Gebhard MC, Li Q, Taft JM, Georgiou G, Iverson BL (2013) Engineering of TEV protease variants by yeast ER sequestration screening (YESS) of combinatorial libraries. Proc Natl Acad Sci U S A 110(18): 7229–7234. https://doi.org/10.1073/pnas. 1215994110 96. Denard CA, Paresi C, Yaghi R, McGinnis N, Bennett Z, Yi L, Georgiou G, Iverson BL (2021) YESS 2.0, a Tunable Platform for

Enzyme Evolution, Yields Highly Active TEV Protease Variants. ACS Synth Biol 10(1):63–71. https://doi.org/10.1021/ acssynbio.0c00452 97. Chen I, Dorr BM, Liu DR (2011) A general strategy for the evolution of bond-forming enzymes using yeast display. Proc Natl Acad Sci U S A 108(28):11399–11404. https:// doi.org/10.1073/pnas.1101046108 98. Cho BK, Kieke MC, Boder ET, Wittrup KD, Kranz DM (1998) A yeast surface display system for the discovery of ligands that trigger cell activation. J Immunol Methods 220(1–2):179–188. https://doi.org/10. 1016/s0022-1759(98)00158-6 99. Wen F, Esteban O, Zhao H (2008) Rapid identification of CD4+ T-cell epitopes using yeast displaying pathogen-derived peptide library. J Immunol Methods 336(1):37–44. https://doi.org/10.1016/j.jim.2008. 03.008 100. Wen F, Smith MR, Zhao H (2019) Construction and screening of an antigen-derived peptide library displayed on yeast cell surface for CD4+ T cell epitope identification. Methods Mol Biol 2024:213–234. https://doi.org/ 10.1007/978-1-4939-9597-4_13 101. Smith MR, Tolbert SV, Wen F (2018) Protein-scaffold directed nanoscale assembly of T cell ligands: artificial antigen presentation with defined valency, density, and ratio. ACS Synth Biol 7(6):1629–1639. https://doi. org/10.1021/acssynbio.8b00119 102. Birnbaum ME, Mendoza JL, Sethi DK, Dong S, Glanville J, Dobbins J, Ozkan E, Davis MM, Wucherpfennig KW, Garcia KC (2014) Deconstructing the peptide-MHC specificity of T cell recognition. Cell 157(5): 1073–1087. https://doi.org/10.1016/j.cell. 2014.03.047 103. Mendoza JL, Fischer S, Gee MH, Lam LH, Brackenridge S, Powrie FM, Birnbaum M, McMichael AJ, Garcia KC, Gillespie GM (2020) Interrogating the recognition landscape of a conserved HIV-specific TCR reveals distinct bacterial peptide crossreactivity. elife 9:e58128. https://doi.org/ 10.7554/eLife.58128 104. Gee MH, Han A, Lofgren SM, Beausang JF, Mendoza JL, Birnbaum ME, Bethune MT, Fischer S, Yang X, Gomez-Eerland R, Bingham DB, Sibener LV, Fernandes RA, Velasco A, Baltimore D, Schumacher TN, Khatri P, Quake SR, Davis MM, Garcia KC (2018) Antigen identification for orphan T cell receptors expressed on tumor-infiltrating lymphocytes. Cell 172(3):549–563.e516.

Continuing Acceleration of Yeast Display Applications https://doi.org/10.1016/j.cell.2017. 11.043 105. Rappazzo CG, Huisman BD, Birnbaum ME (2020) Repertoire-scale determination of class II MHC peptide binding via yeast display improves antigen prediction. Nat Commun 11(1):4414. https://doi.org/10.1038/ s41467-020-18204-2 106. Kumar R, Kumar P (2019) Yeast-based vaccines: New perspective in vaccine development and application. FEMS Yeast Res 19(2):foz007. https://doi.org/10.1093/ femsyr/foz007 107. Wang LJ, Xiao T, Xu C, Li J, Liu GZ, Yin K, Cui Y, Wei QK, Huang BC, Sun H (2018) Protective immune response against Toxoplasma gondii elicited by a novel yeast-based vaccine with microneme protein 16. Vaccine 36(27):3943–3948. https://doi.org/10. 1016/j.vaccine.2018.05.072 108. Shibasaki S, Aoki W, Nomura T, Miyoshi A, Tafuku S, Sewaki T, Ueda M (2013) An oral vaccine against candidiasis generated by a yeast molecular display system. Pathog Dis 69(3):262–268. https://doi.org/10.1111/ 2049-632x.12068 109. Patterson R, Eley T, Browne C, Martineau HM, Werling D (2015) Oral application of freeze-dried yeast particles expressing the PCV2b Cap protein on their surface induce protection to subsequent PCV2b challenge in vivo. Vaccine 33(46):6199–6205. https:// doi.org/10.1016/j.vaccine.2015.10.003 110. Lei H, Xie B, Gao T, Cen Q, Ren Y (2020) Yeast display platform technology to prepare oral vaccine against lethal H7N9 virus challenge in mice. Microb Cell Factories 19(1): 53. https://doi.org/10.1186/s12934-02001316-1 111. Wasilenko JL, Sarmento L, Spatz S, PantinJackwood M (2010) Cell surface display of highly pathogenic avian influenza virus hemagglutinin on the surface of Pichia pastoris cells using alpha-agglutinin for production of oral vaccines. Biotechnol Prog 26(2): 542–547. https://doi.org/10.1002/ btpr.343 112. Bidlingmaier S, Liu B (2011) Construction of yeast surface-displayed cDNA libraries. Methods Mol Biol 729:199–210. https://doi.org/ 10.1007/978-1-61779-065-2_13 113. Bidlingmaier S, Liu B (2015) Identification of novel protein-ligand interactions by exon microarray analysis of yeast surface displayed cDNA library selection outputs. Methods Mol Biol 1319:179–192. https://doi.org/ 10.1007/978-1-4939-2748-7_9

23

114. Singh K, Davies G, Alenazi Y, Eaton JRO, Kawamura A, Bhattacharya S (2017) Yeast surface display identifies a family of evasins from ticks with novel polyvalent CC chemokine-binding activities. Sci Rep 7(1): 4267. https://doi.org/10.1038/s41598017-04378-1 115. Eaton JRO, Alenazi Y, Singh K, Davies G, Geis-Asteggiante L, Kessler B, Robinson CV, Kawamura A, Bhattacharya S (2018) The N-terminal domain of a tick evasin is critical for chemokine binding and neutralization and confers specific binding activity to other evasins. J Biol Chem 293(16):6134–6146. https://doi.org/10.1074/jbc.RA117. 000487 116. Gupta A, Arora G, Rosen CE, Kloos Z, Cao Y, Cerny J, Sajid A, Hoornstra D, Golovchenko M, Rudenko N, Munderloh U, Hovius JW, Booth CJ, Jacobs-Wagner C, Palm NW, Ring AM, Fikrig E (2020) A human secretome library screen reveals a role for Peptidoglycan Recognition Protein 1 in Lyme borreliosis. PLoS Pathog 16(11):e1009030. https://doi.org/10. 1371/journal.ppat.1009030 117. Rouet R, Jackson KJL, Langley DB, Christ D (2018) Next-generation sequencing of antibody display repertoires. Front Immunol 9: 118. https://doi.org/10.3389/fimmu. 2018.00118 118. Whitehead TA, Chevalier A, Song Y, Dreyfus C, Fleishman SJ, De Mattos C, Myers CA, Kamisetty H, Blair P, Wilson IA, Baker D (2012) Optimization of affinity, specificity and function of designed influenza inhibitors using deep sequencing. Nat Biotechnol 30(6):543–548. https://doi.org/10. 1038/nbt.2214 119. Wrenbeck EE, Faber MS, Whitehead TA (2017) Deep sequencing methods for protein engineering and design. Curr Opin Struct Biol 45:36–44. https://doi.org/10.1016/j. sbi.2016.11.001 120. van Rosmalen M, Janssen BM, Hendrikse NM, van der Linden AJ, Pieters PA, Wanders D, de Greef TF, Merkx M (2017) Affinity maturation of a cyclic peptide handle for therapeutic antibodies using deep mutational scanning. J Biol Chem 292(4): 1477–1489. https://doi.org/10.1074/jbc. M116.764225 121. Harris DT, Wang N, Riley TP, Anderson SD, Singh NK, Procko E, Baker BM, Kranz DM (2016) Deep mutational scans as a guide to engineering high affinity T cell receptor interactions with peptide-bound major

24

Maryam Raeeszadeh-Sarmazdeh and Eric T. Boder

histocompatibility complex. J Biol Chem 291(47):24566–24578. https://doi.org/10. 1074/jbc.M116.748681 122. Kowalsky CA, Faber MS, Nath A, Dann HE, Kelly VW, Liu L, Shanker P, Wagner EK, Maynard JA, Chan C, Whitehead TA (2015) Rapid fine conformational epitope mapping using comprehensive mutagenesis and deep sequencing. J Biol Chem 290(44): 26457–26470. https://doi.org/10.1074/ jbc.M115.676635 123. Medina-Cucurella AV, Whitehead TA (2018) Characterizing protein-protein interactions using deep sequencing coupled to yeast surface display. Methods Mol Biol 1764: 101–121. https://doi.org/10.1007/978-14939-7759-8_7 124. Heyne M, Papo N, Shifman JM (2020) Generating quantitative binding landscapes through fractional binding selections combined with deep sequencing and data normalization. Nat Commun 11(1):297. https:// doi.org/10.1038/s41467-019-13895-8 125. Naftaly S, Cohen I, Shahar A, Hockla A, Radisky ES, Papo N (2018) Mapping protein selectivity landscapes using multi-target selective screening and next-generation sequencing of combinatorial libraries. Nat Commun 9(1):3935. https://doi.org/10.1038/ s41467-018-06403-x 126. Oh EJ, Liu R, Liang L, Freed EF, Eckert CA, Gill RT (2020) Multiplex evolution of antibody fragments utilizing a yeast surface display platform. ACS Synth Biol 9(8): 2197–2202. https://doi.org/10.1021/ acssynbio.0c00159 127. Fagre AC, Manhard J, Adams R, Eckley M, Zhan S, Lewis J, Rocha SM, Woods C, Kuo K, Liao W, Li L, Corper A, Challa D, Mount E, Tumanut C, Tjalkens RB, Aboellail T, Fan X, Schountz T (2020) A potent SARS-CoV2 neutralizing human monoclonal antibody that reduces viral burden and disease severity in syrian hamsters. Front Immunol 11: 614256. https://doi.org/10.3389/fimmu. 2020.614256 128. Schoof M, Faust B, Saunders RA, Sangwan S, Rezelj V, Hoppe N, Boone M, Billesbølle CB, Puchades C, Azumaya CM, Kratochvil HT, Zimanyi M, Deshpande I, Liang J, Dickinson S, Nguyen HC, Chio CM, Merz GE, Thompson MC, Diwanji D, Schaefer K, Anand AA, Dobzinski N, Zha BS, Simoneau CR, Leon K, White KM, Chio US, Gupta M, Jin M, Li F, Liu Y, Zhang K, Bulkley D, Sun M, Smith AM, Rizo AN, Moss F, Brilot AF, Pourmal S, Trenker R, Pospiech T,

Gupta S, Barsi-Rhyne B, Belyy V, Barile-Hill AW, Nock S, Liu Y, Krogan NJ, Ralston CY, Swaney DL, Garcı´a-Sastre A, Ott M, Vignuzzi M, Walter P, Manglik A (2020) An ultrapotent synthetic nanobody neutralizes SARS-CoV-2 by stabilizing inactive Spike. Science 370(6523):1473–1479. https://doi. org/10.1126/science.abe3255 129. Rappazzo CG, Tse LV, Kaku CI, Wrapp D, Sakharkar M, Huang D, Deveau LM, Yockachonis TJ, Herbert AS, Battles MB, O’Brien CM, Brown ME, Geoghegan JC, Belk J, Peng L, Yang L, Hou Y, Scobey TD, Burton DR, Nemazee D, Dye JM, Voss JE, Gunn BM, McLellan JS, Baric RS, Gralinski LE, Walker LM (2021) Broad and potent activity against SARS-like viruses by an engineered human monoclonal antibody. Science 371(6531):823–829. https://doi.org/10. 1126/science.abf4830 130. Glasgow A, Glasgow J, Limonta D, Solomon P, Lui I, Zhang Y, Nix MA, Rettko NJ, Zha S, Yamin R, Kao K, Rosenberg OS, Ravetch JV, Wiita AP, Leung KK, Lim SA, Zhou XX, Hobman TC, Kortemme T, Wells JA (2020) Engineered ACE2 receptor traps potently neutralize SARS-CoV-2. Proc Natl Acad Sci U S A 117(45):28046–28055. h t t p s : // d o i . o r g / 1 0 . 1 0 7 3 / p n a s . 2016093117 131. Starr TN, Greaney AJ, Hilton SK, Ellis D, Crawford KHD, Dingens AS, Navarro MJ, Bowen JE, Tortorici MA, Walls AC, King NP, Veesler D, Bloom JD (2020) Deep mutational scanning of SARS-CoV-2 receptor binding domain reveals constraints on folding and ACE2 binding. Cell 182(5):1295–1310. e1220. https://doi.org/10.1016/j.cell. 2020.08.012 132. Starr TN, Greaney AJ, Addetia A, Hannon WW, Choudhary MC, Dingens AS, Li JZ, Bloom JD (2021) Prospective mapping of viral mutations that escape antibodies used to treat COVID-19. Science 371(6531): 850–854. https://doi.org/10.1126/science. abf9302 133. Greaney AJ, Loes AN, Crawford KHD, Starr TN, Malone KD, Chu HY, Bloom JD (2021) Comprehensive mapping of mutations in the SARS-CoV-2 receptor-binding domain that affect recognition by polyclonal human plasma antibodies. Cell Host Microbe 29(3): 463–476.e466. https://doi.org/10.1016/j. chom.2021.02.003 134. Mendoza JL, Schneider WM, Hoffmann HH, Vercauteren K, Jude KM, Xiong A, Moraga I, Horton TM, Glenn JS, de Jong YP, Rice CM,

Continuing Acceleration of Yeast Display Applications Garcia KC (2017) The IFN-λ-IFN-λR1-IL10Rβ complex reveals structural features underlying type III IFN functional plasticity. Immunity 46(3):379–392. https://doi.org/ 10.1016/j.immuni.2017.02.017 135. Saxton RA, Tsutsumi N, Su LL, Abhiraman GC, Mohan K, Henneberg LT, Aduri NG,

25

Gati C, Garcia KC (2021) Structure-based decoupling of the pro- and anti-inflammatory functions of interleukin-10. Science 371(6535):eabc8433. https://doi.org/10. 1126/science.abc8433

Part II Construction and Selection of Yeast Surface Display Libraries and Analysis of Isolated Variants

Chapter 2 Yeast Surface Display for Protein Engineering: Library Generation, Screening, and Affinity Maturation Byong H. Kang, Brianna M. Lax, and K. Dane Wittrup Abstract Yeast surface display is a powerful directed evolution method for developing and engineering protein molecules to attain desired properties. Here, updated protocols are presented for purposes of identification of lead binders and their affinity maturation. Large libraries are screened by magnetic bead selections followed by flow cytometric selections. Upon identification and characterization of single clones, their affinities are improved by an iterative process of mutagenesis and fluorescence-activated cell sorting. Key words Yeast surface display, Directed evolution, Protein engineering, Library screening, Affinity maturation

1

Introduction Directed evolution by display technologies has been a powerful in vitro method in developing and engineering proteins for research, diagnostic, and therapeutic applications. This is made possible by the direct linkage between genotype and phenotype, which allows for translating genetic diversity into protein diversity and tracing back to the gene responsible once the desired phenotype is identified. Conceptually, directed evolution mimics natural selection in that protein variants in a large combinatorial library or a library generated by random or site-directed mutagenesis are screened in a high-throughput manner to select variants with desired properties, such as specificity, affinity, stability, and function. By creating new libraries with the lead candidates, this process can be iterated to further optimize these properties.

The original version of this chapter was revised. The correction to this chapter is available at https://doi.org/ 10.1007/978-1-0716-2285-8_27 Byong H. Kang and Brianna M. Lax contributed equally to this work. Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_2, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022, Corrected Publication 2022

29

30

Byong H. Kang et al.

Since the introduction of phage display in 1985 [1], many other display technologies have been developed, including, in chronological order, bacterial surface display [2, 3], ribosome display [4], yeast surface display (YSD) [5], mRNA display [6, 7], DNA display [8], and mammalian cell surface display [9]. However, compared to other display technologies, YSD has notable advantages such as its eukaryotic expression system that allows for complex post-translational modifications and quality control of protein folding [10]. YSD also allows for quantitative library screening by fluorescence-activated cell sorting (FACS) [11], which enables enrichment of the desired yeast population in equilibrium conditions or in kinetic competition with the antigen, with fine discrimination between binders with varying affinities [12]. In addition, affinity and stability of lead candidates can be assessed in real-time by flow cytometry [13]. While these advantages are present in mammalian cell surface display as well, YSD has been widely used and preferred to mammalian cell surface display due to its lower cost, faster library passaging, and enhanced ease of handling [14]. Initially, YSD was shown to be capable of generating singlechain variable fragments (scFvs) that can bind to fluorescein with a high affinity [5]. Since then, various immunoglobulin scaffolds have been incorporated into YSD to develop high-affinity immunoglobulin molecules, including immune [15], nonimmune [16], and synthetic [17, 18], scFv, full IgG [19, 20], Fab [21, 22], Fc [23–25], and camelid [26, 27] or shark [28–31] single-domain antibody. YSD has been also applied to non-immunoglobulin scaffold proteins, including but not limited to, T cell receptor (TCR) [32–34], lamprey variable lymphocyte receptor [35], the tenth type III domain of human fibronectin (Fn3) [36–38], green fluorescent protein [39], Sso7d [40, 41], knottin [42], Kringle domain [43], affibody [44, 45], designed ankyrin repeat proteins (DARPins) [46], and Gp2 [47]. These immunoglobulin and other protein scaffolds have been used to develop high-affinity binders to various targets [48–50]. Over the years, YSD has been adapted in various applications. For membrane-bound proteins that are difficult to prepare solubly, cell-based selection methods were developed to identify binders that are able to form complexes with mammalian cells expressing the antigen on their membrane [51, 52]. Binders have been developed to have cross-reactivity to botulinum neurotoxins [53], chemokines [54], and hemagglutinins (HA) [55, 56]. In addition to protein libraries, cDNA libraries of primary human tumors have been used to identify tumor antigens [57, 58]. By directed evolution of existing proteins, variants with improved affinity to their cognate partners have been identified for epidermal growth factor (EGF) [59], interleukin-2 (IL-2) [60, 61], leptin [62], Axl receptor [63], signal-regulatory protein α (SIRPα) [64], interleukin-17A receptor [65], and programmed cell death protein-1 (PD-1) [66]. Enzymes such as horseradish peroxidase (HRP) [67, 68],

Yeast Surface Display for Protein Engineering

31

sortase A [69], firefly luciferase [70], TEV protease [71], and glucose oxidase (GOx) [72] have been engineered to have higher catalytic activity. Libraries of protein variants have been displayed on the surface of yeast to map the epitope of binders to EGFR [73, 74], H5 HA [75], HIV-1 gp120 [76], alpha toxin from Staphylococcus aureus [77], tumor necrosis factor, and pertussis toxin [78]. YSD has also been applied to identify peptide specificity of major histochemistry complex (MHC) molecules [79–81] and specificity of peptide-MHC complexes for a given TCR [82]. While YSD has been successful in developing proteins for various applications, there are potential limitations. Compared to phage (~1011), bacterial (~1011), and mRNA (~1014) display, the apparent clonal size of a YSD library (~109) is nominally smaller—although true functional displayed protein diversity may be superior in a eucaryotic host. For example, it has been shown that YSD was more effective than phage display in identifying binders to HIV-1 gp120 and the binders isolated by YSD were not displayed well on phages [83]. Another potential issue is that because the glycosylation machinery of yeast differs from that of mammals, proteins are often hyperglycosylated and may improperly present their epitopes. YSD utilizes two heterologous a-agglutinin proteins, Aga1p and Aga2p, that are naturally used by yeast to mediate cell–cell contacts during mating. Aga1p is a glycosylphosphatidylinositol (GPI)anchored membrane protein linked to the cell wall, and Aga2p is an adhesion protein that attaches to Aga1p by two disulfide bonds [84]. By expressing a scaffold fused to Aga2p, the scaffold is displayed as a covalent complex on the yeast cell surface. Aga1p is stably integrated into the chromosome in a galactose-inducible expression cassette, whereas Aga2p-scaffold fusion is expressed under a galactose-inducible promoter on the yeast display plasmid, which is maintained episomally with a nutritional marker. The protein scaffold of interest can be expressed as N- or C-terminal fusion to Aga2p with N-terminal HA and C-terminal c-Myc affinity tags (Fig. 1). YSD was first developed and is most commonly used in the yeast species Saccharomyces cerevisiae; however, it has been applied to Pichia pastoris as well [85, 86]. Other anchoring proteins have been explored, such as α-agglutinin and Flo1p, but the Aga1pAga2p system remains the most widely used in YSD [87]. Here, we combine, consolidate, and improve upon several YSD protocols that have been published previously [88–92] to streamline the overall workflow and update with methods to perform affinity maturation by kinetic competition (Fig. 2). We focus on isolating de novo binders to the antigen of interest from a naı¨ve yeast library and developing lead candidates into high-affinity binders. However, this protocol can be easily applied to improving the affinity of an existing binder, developing a cross-reactive version of an existing binder, or the mutagenesis of a displayed protein for epitope mapping. The overall workflow consists of enrichment for

32

Byong H. Kang et al.

Fig. 1 Schematic representation of yeast surface display and corresponding simplified plasmid maps. (a) By pCTcon2 vector, the protein of interest (POI) with N-terminal hemagglutinin (HA) and C-terminal c-myc affinity tags is displayed as a C-terminal fusion to Aga2p. (b) By pCHA vector, POI with the same affinity tags is displayed as an N-terminal fusion to Aga2p

binders by magnetic bead selections followed by FACS selections. After isolation and characterization of single clones, a new library can be generated by mutagenesis and enriched by FACS under equilibrium or kinetic competition to isolate variants with higher affinity. This process can be iterated until the desired affinity is achieved.

2

Materials

2.1 Yeast Library Growth and Induction

1. Naı¨ve yeast display libraries generated with Saccharomyces cerevisiae strain EBY100 or RJY100 (see Subheading 3.5 for generating a library; the libraries listed below are available from the authors). (a) Nonimmune human scFv library [16]. (b) Synthetic switchable display/secretion human scFv library G [17]. (c) Human fibronectin type III (Fn3) library [38].

Yeast Surface Display for Protein Engineering

33

Fig. 2 Overall workflow of yeast surface display for developing high-affinity binders. Naive yeast library is grown, induced, and subjected to magnetic bead selections to deplete undesired binders and to enrich for binders to the biotinylated antigen of interest. Resulting populations are further enriched by FACS in equilibrium conditions to select for higher-affinity binders and monitored by flow analysis. By extracting plasmids from the enriched populations, their sequences are analyzed and prevalent sequences are transformed into yeast to characterize single clones. Once clones of interest are identified, a new library is generated by performing error-prone PCR and transforming yeast via homologous recombination. For nanomolar affinity, higher-affinity binders are enriched by iterating FACS selection in equilibrium conditions and for sub-nanomolar affinity, they are enriched by iterating FACS selection in kinetic competition until desired affinity is achieved

(d) Charge-neutralized Sso7d library [41]. (e) Synthetic nonspecificity-reduced human scFv library [18].

34

Byong H. Kang et al.

2. Yeast extract Peptone Dextrose (YPD) medium: Dissolve 20 g peptone and 10 g yeast extract in 950 mL ddH2O and autoclave. Separately, dissolve 20 g dextrose in 50 mL ddH2O and sterilize by filtration. Combine and allow to cool before use. 3. YPD plates: Dissolve 20 g peptone, 10 g yeast extract, and 15 g agar in 950 mL ddH2O and autoclave. Separately, dissolve 20 g dextrose in 50 mL ddH2O and sterilize by filtration. Combine and pour plates while warm. 4. Synthetic Dextrose medium with CasAmino Acids (SDCAA), pH 4.5: Dissolve 20 g dextrose, 6.7 g yeast nitrogen base, 5 g casamino acids, 10.4 g sodium citrate, 7.4 g citric acid monohydrate in 1 L ddH2O and sterilize by filtration (see Note 1). 5. SDCAA plates, pH 6.0: Dissolve 182 g sorbitol, 15 g agar, 8.56 g sodium phosphate monobasic monohydrate, 5.4 g sodium phosphate dibasic anhydrous in 900 mL ddH2O and autoclave. Dissolve 20 g dextrose, 6.7 g yeast nitrogen base, 5 g casamino acids in 100 mL ddH2O. When the autoclaved solution has cooled to 60
C, sterilize the non-autoclaved solution by filtering it into the autoclaved solution, and pour plates. 6. Synthetic Galactose medium with CasAmino Acids (SGCAA): Dissolve 20 g galactose, 2 g dextrose, 6.7 g yeast nitrogen base, 5 g casamino acids, 8.56 g sodium phosphate monobasic monohydrate, 5.4 g sodium phosphate dibasic anhydrous in 1 L ddH2O and sterilize by filtration. 7. Sterile glass culture tubes (For 16-mm diameter tubes, do not exceed 5 mL; see Note 2). 8. Sterile baffled glass flasks (For flasks smaller than 1 L, do not exceed 1/3 of the volume; see Note 2). 9. Stationary incubator at 30
C. 10. Shaking incubator at 30
C, 250 rpm. 11. Shaking incubator at 20
C, 250 rpm. 12. Spectrophotometer capable of measuring OD600. 2.2 Library Enrichment by Magnetic Bead Selections

1. Refrigerated microcentrifuge. 2. Refrigerated benchtop centrifuge. 3. 2.0-mL microcentrifuge tubes. 4. 1 PBS. 5. PBSA: Sterile PBS + 0.1 w/v% bovine serum albumin (Millipore Sigma). 6. Target antigen and control antigen. 7. EZ-Link Sulfo-NHS-LC-Biotin (Thermo Fisher). 8. Dynabeads biotin binder (Thermo Fisher). 9. Dynamag-2 magnet (Thermo Fisher).

Yeast Surface Display for Protein Engineering

35

10. Rotary wheel at 4
C. 11. 30 v/v% glycerol in SDCAA. 2.3 FACS Selection (Equilibrium)

1. Primary antibodies: (a) Chicken anti-c-myc (Exalpha). (b) Mouse anti-HA (clone 16B12) (Biolegend). 2. Secondary antibodies (Invitrogen): (a) Goat anti-chicken Alexa Fluor 488 (AF488). (b) Goat anti-mouse Alexa Fluor 647 (AF647). (c) Streptavidin AF647. 3. Cell-strainer capped round-bottom tubes (compatible with flow cytometer).

2.4 Identification and Characterization of Single Clones

1. Zymoprep™ Yeast Plasmid Miniprep II Kit (Zymo Research). 2. GenCatch™ Plasmid DNA Miniprep Kit (or comparable miniprep kit) (Epoch). 3. Stellar™ cells (or comparable competent Escherichia coli) (Takara). 4. Water bath at 42
C. 5. SOC media. 6. Carbenicillin (50 mg/mL) (1000): Dissolve in ddH2O and sterile filter. 7. Luria broth (LB) + carbenicillin medium: Dissolve 25 g LB powder (Miller, BD™) in 1 L ddH2O. Sterilize by filtration or autoclaving. Add 1 mL 1000 carbenicillin (after cooled if autoclaved). 8. LB + carbenicillin plates: Dissolve 25 g LB powder and 15 g agar in 1 L ddH2O and autoclave. Once cooled to 60
C, add 1 mL 1000 carbenicillin and pour plates while warm. 9. Sequencing primers: (a) pCTcon2 forward primer: 50 - CAATAGCTCGACGATT GAAGGTAGA-30 . (b) pCTcon2 reverse primer: 50 - ACACTGTTGTTATCA GATCTCG-30 . (c) pCHA forward primer: TAACGTCAAGG-30 .

50 - ATACCTCTATACTT

(d) pCHA reverse primer: 50 -CGCAATTACTGACAAACGT TACTGA-30 . 10. Saccharomyces cerevisiae strain RJY100 (or EBY100). 11. Frozen-EZ Yeast Transformation II Kit (Zymo Research).

36

Byong H. Kang et al.

12. 96-well v-bottom plates (compatible with HTS attachment on flow cytometer). 13. Stationary incubator at 37
C. 14. Shaking incubator at 37
C, 250 rpm. 2.5 Library Generation

1. Yeast surface display vector pCTcon2 or pCHA (Addgene). 2. Restriction enzymes (with CutSmart buffer) (New England Biolabs): (a) BamHI-HF. (b) NheI-HF. (c) SalI-HF. 3. Recombination primers: (a) pCTcon2 forward primer: 50 CGACGATTGAAGGTAGATACCCATAC GACGTTCCAGACTACGCTCTGCAG-30 . (b) pCTcon2 reverse primer: 50 - CAGATCTCGAGCTATTACAAGTCCTCTTCA GAAATAAGCTTTTGTTC-30 . (c) pCHA forward primer: 50 - GCTCTTTGGACAAGAGA GAAGCTTACCCATACGACGTTCCAGACTACGCT 30 . (d) pCHA reverse primer: 50 TCCTGCAAGTCTTCTTCGGAGA TAAGCTTTTGTTCTGCACGCGTGGATCC-30 , 4. 20 -deoxynucleoside-50 -triphosphates dCTP, dGTP, dTTP).

(dNTP

mix:

dATP,

5. 8-oxo-20 -deoxyguanosine-50 -triphosphate (8-oxo-dGTP) (TriLink Biotechnologies). 6. 20 -deoxy-p-nucleoside-50 -triphosphate Biotechnologies).

(dPTP)

(TriLink

7. Taq DNA polymerase (with ThermoPol buffer) (New England Biolabs). 8. Thin-walled PCR tubes. 9. Thermocycler Thermal cycler. 10. Low-melt agarose. 11. GelGreen nucleic acid stain (10,000) (Biotium Inc.). 12. Tris Acetate-EDTA (TAE) buffer. 13. Gel loading dye (6). 14. Fast DNA ladder (New England Biolabs). 15. Gel electrophoresis equipment. 16. Gel extraction + PCR clean up kit (Takara).

Yeast Surface Display for Protein Engineering

37

17. Pellet Paint NF co-precipitant (Millipore Sigma). 18. 100 mM lithium acetate (sterile filtered). 19. 1 M dithiothreitol (DTT) (sterile filtered). 20. 2-mm electroporation cuvettes. 21. Bio-Rad Gene Pulser XCell Total System (Bio-Rad).

3

Methods

3.1 Yeast Library Growth and Induction

1. Thaw an aliquot of the yeast library of interest at room temperature (see Note 3). 2. Inoculate the library into 1 L SDCAA (~OD600 ¼ 1) and grow the culture overnight in the 30
C shaking incubator until OD600 reaches 6–8 (see Notes 2, 4, and 5). 3. Pellet at least 10 the maximum library diversity at 2000  g for 3 min. Aspirate the supernatant and resuspend the cells in fresh SDCAA to OD600 of 1. Grow the culture in the 30
C shaking incubator until OD600 reaches 2–5. 4. Pellet at least 10 the maximum library diversity at 2000  g for 3 min. Aspirate the supernatant and resuspend the cells in SGCAA to OD600 of 1. Induce the culture in the 20
C shaking incubator overnight. OD600 should be around 2–3. 5. Induced cells can be used for selections or stored at 4
C for up to 1 month.

3.2 Library Enrichment by Magnetic Bead Selections

To identify binders to the antigen of interest, the naı¨ve library is screened by magnetic bead selections and FACS. A FACS sorter can typically sort 5  107 cells/h, and so a naı¨ve library with a diversity of 107 or smaller could be directly screened by FACS (see Subheading 3.3.1 and Note 6). However, for a library with diversity of 108 or greater, magnetic bead selections should be performed to reduce the library to a manageable size for FACS selections while depleting undesired binders and enriching for binders to the antigen. In magnetic bead selections, the biotinylated antigen is immobilized on streptavidin-coated magnetic beads, which increase the avidity of the antigen and allow for capture of low-affinity binders. Initially, a negative selection against streptavidin-coated beads is performed to deplete streptavidin binders. If the antigen has an affinity tag or fusion partner, the unbound population is negatively selected against the biotinylated affinity tag or fusion partner. Lastly, positive selection is performed on the unbound population against the biotinylated antigen.

3.2.1 Antigen Biotinylation

If the antigen has an affinity tag or a fusion partner, it is recommended that biotinylated control antigen (just affinity tag/fusion partner or an irrelevant protein with the same affinity tag/fusion

38

Byong H. Kang et al.

partner) is used for negative bead selection and that non-biotinylated control antigen is included in positive bead selection (see Note 7). For most antigens, chemical biotinylation of primary amines with NHS-reactive biotin allows for random biotinylation around the antigen such that all of the epitopes are accessible on average. This can be done by using EZ-Link Sulfo-NHSLC-Biotin and following the manufacturer’s protocol. However, depending on the antigen, site-specific biotinylation can be achieved by adding an AviTag to the antigen and enzymatically biotinylating by co-expression or in vitro incubation with biotin ligase. 3.2.2 Negative Selection Against Magnetic Beads

1. Pellet at least 10 the maximum library diversity at 2000  g for 3 min in 50-mL conical tubes. (This protocol assumes the library diversity of 1  109 and use of 1  1010 cells and 50 μL of magnetic beads; scale accordingly). 2. Aspirate the supernatant and resuspend the pellet to a final concentration of 2  109 cells/mL. 3. Transfer 2  109 cells into five 2-mL microcentrifuge tubes (for a total of 1010 cells) and pellet at 13,000  g for 1 min. 4. Aspirate the supernatant and resuspend the cells in each tube with 900 μL of PBSA. 5. Vortex magnetic beads in the vial for 30 s and transfer 50 μL of beads into a 2-mL microcentrifuge tube. 6. Wash the beads with 1 mL of PBSA and place the tube on a magnet for 2 min. 7. Aspirate the supernatant and resuspend the beads with 500 μL of PBSA. 8. Add 100 μL of beads into each tube with cells and incubate on a rotary wheel at 4
C for 2 h. (Antigen-coated beads can be prepared at the same time by following Subheading 3.2.3) (see Note 8). 9. Place the tubes on a magnet for 5 min. Carefully transfer the liquid in the cap into the tube during this process (see Note 9). 10. Collect the supernatant from each tube containing unbound yeast cells into fresh 2-mL centrifuge tubes (see Note 10). 11. Repeat negative selection against magnetic beads. If there is a control antigen, prepare control antigen-coated beads by following Subheading 3.2.3 and perform negative selection by following steps 8–10 (see Note 11).

3.2.3 Preparation of Antigen-Coated Beads

1. Prepare beads as in Subheading 3.2.2 steps 5–7. 2. Add 165 pmole of biotinylated antigen and incubate on a rotary wheel at 4
C for 2 h. 3. Place the tube on a magnet for 2 min.

Yeast Surface Display for Protein Engineering

39

4. Aspirate the supernatant and wash the beads with 1 mL of PBSA by resuspending. 5. Repeat steps 3 and 4 two more times. 6. Resuspend the beads with 500 μL of PBSA. 3.2.4 Positive Selection Against Target Antigen

1. Add 100 μL of antigen-coated beads into each tube containing negatively selected yeast cells from Subheading 3.2.2 (see Note 12). 2. Incubate on a rotary wheel at 4
C for 2 h. 3. Place the tubes on a magnet for 5 min. 4. Aspirate the supernatant and gently wash the beads in each tube with 1 mL of PBSA. 5. Place the tubes on a magnet for 2 min. 6. Aspirate the supernatant and resuspend the beads in each tube with 1 mL SDCAA. 7. Transfer the cells from each tube into a baffled culture flask containing 20 mL SDCAA (total volume will be 25 mL). 8. Plate serially diluted culture on SDCAA plates to determine the library diversity. 9. Mix the flask well, take 25 μL of the culture, and dilute into 175 μL of SDCAA (8 dilution). Make 80, 800, and 8000 dilutions by serially diluting 10 μL into 90 μL of SDCAA, using a new tip for each transfer. 10. Draw quadrants on a SDCAA plate and spread 20 μL of each dilution onto each quadrant. 11. Grow the liquid culture to saturation in the 30
C shaking incubator. This will take 1–2 days. 12. Pellet the culture at 2000  g for 3 min and aspirate the supernatant. 13. Resuspend the pellet in 1 mL SDCAA and transfer into a 2-mL microcentrifuge tube. 14. Place the tube on a magnet for 5 min. 15. Collect the supernatant and passage the cells into two cultures: one for cryopreservation and the other for induction. 16. Determine the diversity of the library by counting the number of colonies on the plate. Each colony in 8, 80, 800, and 8000 dilutions corresponds to 1  104, 1  105, 1  106, and 1  107 cells recovered, respectively. 17. Cryopreserve at least tenfold the library diversity by pelleting, resuspending in 500 μL SDCAA, adding 500 μL 30% sterile glycerol in SDCAA, and cooling in 80
C at approximately 1
C/min cooling rate (use Mr. Frosty to achieve this cooling rate; see Note 13).

40

Byong H. Kang et al.

18. Induce the library as in Subheading 3.1. The cells are now ready for flow analysis or FACS selection. Based on the flow analysis, perform additional magnetic bead selections or proceed to FACS selection. 3.3 FACS Selection (Equilibrium)

FACS enrichment of binders to a particular antigen of interest is a high-throughput and selective method once the library is of a manageable size (107). It is advantageous to first enrich for fulllength binders and then for binders to the antigen of interest. Sequential equilibrium sorts are performed to continuously enrich for higher-affinity binders. Kinetic sorting (see Subheading 3.6) must be used when the concentrations become so low that equilibrium sorting becomes impractical.

3.3.1 FACS Selection for Enrichment of Full-Length Binders

As shown in Fig. 1, the YSD platform is designed to have both an N-terminal HA tag and a C-terminal c-myc tag to allow for the detection of full display and proper folding of the protein scaffold. Following magnetic bead selections and prior to FACS enrichment for target antigen binders, it is beneficial to sort for binders that are double positive for both affinity tags. This will eliminate clones that have a nonsense or frameshift mutation that results in misfolding of the protein scaffold. These clones will appear as positive for only one of the two tags, likely the N-terminal HA tag. 1. Pellet at least tenfold the maximum library diversity of induced yeast at 13,000  g for 1 min in a 1.7-mL microcentrifuge tube. Aspirate the supernatant and wash with 1 mL of PBSA. 2. Pellet, aspirate, and wash one more time. 3. Resuspend cells in PBSA to a final concentration of 1  108 cells/mL. Add primary antibodies (mouse antiHA + chicken anti-c-myc) at a final concentration of 1 μg/ mL each. 4. Incubate cells with primary antibodies for 30 min to 1 h at room temperature on a rotary wheel. 5. Pellet cells at 13,000  g for 1 min. Aspirate the supernatant and wash with 1 mL of PBSA. 6. Pellet, aspirate, and wash two more times. 7. Resuspend cells in PBSA to a final concentration of 1  108 cells/mL. Add secondary antibodies (goat anti-mouse AF647 + goat anti-chicken AF488) at a final concentration of 2 μg/mL each. 8. Incubate cells with secondary antibodies for 15–30 min at 4
C on a rotary wheel. 9. Pellet cells at 13,000  g for 1 min. Aspirate the supernatant and wash with 1 mL of PBSA. 10. Pellet, aspirate, and wash two more times.

Yeast Surface Display for Protein Engineering

41

11. Keep the cells pelleted in the dark on ice until immediately prior to sorting double positive cells. 12. Resuspend cells in at least 500 μL for sorting immediately before loading tube on the sorter. They should be at a concentration of no more than 1  108 cells/mL. Filter through a cell-strainer capped round-bottom tube compatible with the flow cytometer to remove any cell clumps. 13. Gate single cells based on FSC-A vs. SSC-A, FSC-H vs. FSC-W, and SSC-H vs. SSC-W. The cells in the last gate should appear in a diagonal line on a plot of FSC-A vs. FSC-H. 14. Sort double positive cells on a plot of AF488 (c-myc) vs. AF647 (HA) using a diagonal gate, as shown in Fig. 3. Collect cells into a 15-mL glass culture tube containing 2 mL of SDCAA (see Note 14).

Antigen Binding (AF647)

105

Gate 1 0.081%

Gate 2 4.99%

104

103

0

–103 –103

0

103

104

–105

Display (c-myc) (AF488)

Fig. 3 Representative FACS plot with gating strategy for a library. A library of yeast displaying scFv binders previously subjected to two rounds of magnetic bead selections and three rounds of FACS enrichment were labeled with chicken anti-c-myc and 1 μM biotinylated antigen in a primary incubation until equilibrium binding was reached. A secondary incubation was performed with goat anti-chicken AF488 (display) and streptavidin AF647 (antigen binding). Gate 1 is an example of a stringent gate which will collect only the highest-affinity clones but may miss rare clones. Gate 2 is an example of a more generous gate that will still enrich for binders but at a slower rate and potentially include unique, lower affinity binders. The important distinction between the gates is the percentage of cells collected without changing the slope of the diagonal in the gate

42

Byong H. Kang et al.

15. After sorting is completed, rinse the inside of the culture tube with the SDCAA to rescue any cells stuck to the side. The new maximum diversity of the library is the number of double positive cells collected. 16. Grow the cells in the 30
C shaking incubator to saturation. 17. Passage and induce cells as described in Subheading 3.1. 3.3.2 FACS Selection for Enrichment of HigherAffinity Binders to Target Antigen

Following isolation of full-length protein scaffold displaying cells, flow cytometry is also used to enrich for binders to a target antigen. As compared to magnetic bead selections, which non-discriminately isolate all binders to a target antigen, FACS can be used to selectively enrich for only the highest affinity clones. As in Subheading 3.3.1, the c-myc affinity tag is used as a marker for full-length display. The cells are also incubated with biotinylated antigen and a fluorophore-conjugated streptavidin for detection of antigen binding. Sorting for cells that are double positive will eliminate full-length displaying clones that have no affinity for the target antigen. These clones will appear positive for the c-myc fluorophore only (see Notes 15 and 16). 1. Pellet at least tenfold the maximum library diversity of induced yeast at 13,000  g for 1 min in a 1.7-mL microcentrifuge tube. Aspirate the supernatant and wash with 1 mL of PBSA. 2. Pellet, aspirate, and wash one more time. 3. Resuspend cells in PBSA with chicken anti-c-myc antibody at a final concentration of 1 μg/mL and the desired amount of soluble, biotinylated antigen (see Notes 12, 17, and 18). 4. Incubate cells at room temperature (or 4
C if the antigen is unstable) for time sufficient to reach equilibrium (see Note 19). 5. Pellet cells at 13,000  g for 1 min. Aspirate the supernatant and wash with 1 mL of PBSA. 6. Pellet, aspirate, and wash two more times. 7. Resuspend cells in PBSA to a final concentration of 1  108 cells/mL. Add secondaries (streptavidin Alexa Fluor 647 + goat anti-chicken Alexa Fluor 488), each at a final concentration of 2 μg/mL. 8. Incubate cells with secondaries for 15–30 min at 4
C on a rotary wheel. 9. Pellet cells at 13,000  g for 1 min. Aspirate the supernatant and wash with 1 mL of PBSA. 10. Pellet, aspirate, and wash two more times. 11. Keep the cells pelleted in the dark on ice until immediately prior to sorting double positive cells.

Yeast Surface Display for Protein Engineering

43

12. Resuspend cells in at least 500 μL for sorting immediately before loading tube on the sorter. They should be at a concentration of no more than 1  108 cells/mL. Filter through a cell-strainer capped round-bottom tube compatible with the flow cytometer to remove any cell clumps. 13. Gate single cells as described previously in Subheading 3.3.1. 14. Sort double positive cells on a plot of AF488 (c-myc) vs. AF647 (antigen binding) using a diagonal gate, as shown in Fig. 3 (see Note 20). 15. After sorting is completed, rinse the inside of the culture tube with the SDCAA to rescue any cells stuck to the side. The new maximum diversity of the library is the number of double positive cells collected. 16. Grow the cells in the 30
C shaking incubator to saturation. 17. Passage and induce cells as described in Subheading 3.1. 18. Repeat sorting until sufficient enrichment of binders is achieved (see Note 21). 3.4 Identification and Characterization of Single Clones

Once phenotypic convergence for antigen binding of the library is achieved, the DNA can be extracted for sequencing. The plasmid copy number in yeast cells is too low for sequencing, so the isolated DNA must be transformed into competent E. coli first. Sequencing of these colonies allows for identification of prevalent sequences, which can then be transformed back into competent yeast. These single clone yeast can be used for characterization of binding affinity directly on the surface of yeast, eliminating the need for protein expression and purification. The binding affinity, estimated via the equilibrium dissociation constant KD, of the interaction between the protein binders on the surface of yeast and soluble antigen can be determined using yeast surface titrations. In general, the KDs obtained from yeast surface titrations correlate well with those of soluble proteins [93].

3.4.1 Identification of Single Clones via DNA Extraction, Transformation, and Sequencing

The protocol for DNA extraction from yeast is adapted from the Zymoprep™ Yeast Plasmid Miniprep II Kit and GenCatch™ Plasmid DNA Miniprep Kit protocols. This process allows for the identification of clones that are highly enriched and correspond to binders of the target antigen. 1. Passage yeast to be sequenced to an OD600 of 0.1 in 2 mL of SDCAA in a 15-mL glass culture tube. Grow the cells in the 30
C shaking incubator for 4 h. 2. Pellet 1 mL of the culture at 13,000  g for 1 min in a 1.7-mL microcentrifuge tube. Check to ensure that there is a pellet before aspirating the media.

44

Byong H. Kang et al.

3. Resuspend the yeast in 200 μL of the digestion buffer from the Zymoprep™ Kit. Add 5 μL of Zymolyase and mix well. 4. Incubate in a stationary 37
C incubator for at least 1 h. 5. Add 200 μL of lysis buffer from the Zymoprep™ Kit and mix by inverting the tube 5 times. Incubate for 2 min, add 400 μL of neutralization buffer from the Zymoprep™ Kit, and mix by inverting the tube 5 times. 6. Incubate for 2 min and pellet cell debris at maximum speed in microcentrifuge for 3 min. 7. Transfer supernatant to a DNA binding column from the GenCatch™ miniprep kit (or comparable miniprep column) while not disturbing the pelleted debris. It can help to leave 20–30 μL at the bottom of the tube or transfer the supernatant into a new 1.7-mL microcentrifuge tube and pellet again at maximum speed for 3 min before transferring. 8. Follow the washing and elution steps from the GenCatch™ miniprep kit (or comparable miniprep kit). Allow the elution buffer to sit on the column for 5 min before eluting. 9. Thaw Stellar™ cells (or comparable competent E. coli cells) on ice. 10. Add 5 μL of the plasmid from yeast mini-prep to 50 μL of Stellar™ cells in a 1.7-mL microcentrifuge tube, mix by tapping the tube, and incubate on ice for 20–30 min. 11. Transform by heat shocking the cells at 42
C for 45 s. Allow cells to recover on ice for 2 min. 12. Add 200 μL of SOC media and incubate in the 37
C shaking incubator for 1 h. 13. Plate the entire transformation on an LB carbenicillin plate. 14. Pick colonies and submit for sequencing. 3.4.2 Transformation and Characterization of Single Clones Displayed on Yeast

Once prevalent sequences have been identified, the plasmids can be transformed back into competent yeast follows that of the Zymo Research Frozen-EZ Yeast Transformation II Kit protocol. Flow cytometry is used to determine the affinity of these binders once they are transformed as single clones. 1. Inoculate 5 mL of LB + carbenicillin with single colonies with sequences of interest and incubate in the 37
C shaking incubator for 6 h or up to overnight 12–16 h. 2. Miniprep the E. coli culture to obtain isolated plasmid DNA following the GenCatch™ protocol. 3. Prepare fresh competent RJY100 yeast following the Zymo Research Frozen-EZ Yeast Transformation II Kit protocol (or thaw vial on ice from previously prepared frozen stock).

Yeast Surface Display for Protein Engineering

45

4. Transform competent RJY100 yeast following the Zymo Research Frozen-EZ Yeast Transformation II Kit protocol (see Note 22). 5. Plate the entire transformation on an SDCAA plate. 6. Incubate in the 30
C shaking incubator for 2–3 days. 7. Pick colonies and inoculate in 2 mL of SDCAA in a 15-mL glass culture tube (see Note 23). 8. Grow the cells in the 30
C shaking incubator to saturation. 9. Passage and induce yeast as described in Subheading 3.1 and continue to perform titrations to quantify affinity of single clones. 10. Pellet 105 induced cells for each concentration point on the titration curve while making sure to have enough volume (i.e., for a 12-point titration, pellet 1.3  106 cells) at 13,000  g for 1 min. Aspirate the supernatant and wash with 1 mL of PBSA. 11. Pellet, aspirate, and wash one more time. 12. Resuspend cells in at 106 cells/mL in PBSA for each concentration point (i.e., 1300 μL for a 12-point titration). Aliquot 100 μL of the yeast into separate 1.7-mL microcentrifuge tubes or a 96-well v-bottom plate. Pellet cells and aspirate the supernatant. 13. Resuspend the cells in PBSA with chicken anti-c-myc antibody at a final concentration of 1 μg/mL and the desired concentration of soluble, biotinylated antigen (see Note 24). 14. Perform primary and secondary incubations, with required washing steps, as described in Subheading 3.3.1 (see Note 25). 15. Resuspend cells in at least 500 μL if using tubes, or 100 μL/ well if using plates, immediately before analyzing on the flow cytometer. 16. Analyze the cells on a flow cytometer and record data for 10,000 events. 17. Calculate the equilibrium dissociation constant, KD, as a metric for quantifying affinity. Representative plots and titrations for two clones with different affinities are shown in Fig. 4. 3.4.3 Further Characterization of Single Clones

Following yeast surface titrations of single clones, the best binders can be subcloned into expression vectors for protein expression and purification. Alternatively, if the yeast surface titrations reveal that the binders have affinities that are lower than desirable, affinity maturation can be performed by generating a new library and selecting for higher-affinity binders (see Subheadings 3.5 and 3.6).

3.5 Library Generation

To generate a new YSD library from an existing binder, random mutations are introduced into the gene of interest, combined with linearized backbone, and introduced into electrocompetent yeast

46

Byong H. Kang et al.

A

B Clone 1 (KD = 94.0 ±16.0)

1.0

MFI, Normalized

Antigen Binding (AF647)

Clone 2 (KD = 1.10 ±0.25) 105

104

103

0.5

0

–103

0.0 –103

0

103

104

105

0.01

Display (c-myc) (AF488)

0.1

1

10

100

1000 10000

Antigen Concentration (nM)

Fig. 4 Representative FACS data for two different clones. Yeast displaying scFv binders were labeled with chicken anti-c-myc and biotinylated antigen in a primary incubation until equilibrium binding was reached. A secondary incubation was performed with goat anti-chicken AF488 (display) and streptavidin AF647 (antigen binding). (a) Flow cytometry plot of wild-type binder (Clone 1, purple) and an affinity-matured binder (Clone 2, teal). Both clones are labeled with 100 nM biotinylated target antigen. (b) Titration curves for Clone 1 and affinity-matured Clone 2. Twelve-point titrations were performed in triplicate. Curve fitting resulted in a KD of 93.0  16.0 nM for Clone 1 (wild-type) and 1.10  0.25 nM for Clone 2 (affinity matured). The mean fluorescence intensity (MFI) was normalized by subtraction of the background signal and division by Bmax

via electroporation. The new library is created in vivo via homologous recombination, a DNA repair mechanism naturally occurring in yeast with ~30 bp overlap on both ends. 3.5.1 Library Construction via ErrorProne PCR

Combinatorial libraries require both a linearized yeast display vector and mutagenized inserts. The diversity in the inserts can be introduced in a number of ways, with error-prone PCR as the most preferred due to the level of control over the mutation rate [88], as well as the ability to randomize pooled clones without knowing their individual sequence identity. The mutagenesis rate is tunable by changing either the concentration of the nucleotide analogs, the number of PCR cycles, or both. It is also advantageous because both transitional and transversional mutations can occur this way (see Note 26). 1. Triple digest pCTcon2 yeast display vector using the following reaction components (see Note 27): Component

Concentration

Volume

Vector DNA

X μg/μL

20/X μL

NEB CutSmart buffer

10

10 μL

SalI-HF

20 U/μL

5 μL

NheI-HF

20 U/μL

5 μL

BamHI-HF

20 U/μL

5 μL (continued)

Yeast Surface Display for Protein Engineering

Component

Concentration

47

Volume

ddH2O

75–20/X μL

Total volume

100 μL

2. Mix well and incubate overnight at 37
C. 3. Store at 20
C until needed. 4. Mutagenize insert via error-prone PCR using the following reaction components (see Note 28): Component

Stock conc.

Final conc./amt.

Volume

DNA template

1 ng/μL

1 ng

1 μL

Forward primer

10 μM

500 nM

2.5 μL

Reverse primer

10 μM

500 nM

2.5 μL

dNTPs

10 mM each

200 μM each

1 μL

8-oxo-dGTP

20 μM

2 μM

5 μL

dPTP

20 μM

2 μM

5 μL

ThermoPol reaction buffer

10

1

5 μL

Taq DNA polymerase

5 U/μL

2.5 U

0.5 μL

ddH2O

27.5 μL

Total volume

50 μL

The PCR reaction should be run with the following conditions (see Note 29): Cycle

Step

Time

Temp. (
C)

1 (1)

Initial denaturation

3 min

95

2 (15)

Denaturation

45 s

95

Annealing

30 s

55

Elongation

90 s

72

Final elongation

10 min

72

3 (1)

5. Perform gel electrophoresis to separate the mutagenized insert from the template DNA. Run the entire PCR reaction, mixed with loading dye, on a 1 w/v% low-melt agarose gel in TAE buffer at 100 V for 30 min. Run a ladder alongside the product to ensure the product is of the proper size (see Note 30). 6. Cut out the band of the proper size and purify the product with 30 μL of elution buffer.

48

Byong H. Kang et al.

7. PCR amplify mutagenized insert using the following reaction components: Component

Stock conc. Final conc./amt. Volume

Mutagenized DNA

X ng/μL

80 ng

80/X μL

Forward primer

10 μM

500 nM

5 μL

Reverse primer

10 μM

500 nM

5 μL

dNTPs

10 mM each 200 μM each

2 μL

ThermoPol reaction buffer 10

1

10 μL

Taq DNA polymerase

5U

1 μL

5 U/μL

ddH2O

77–80/X μL

Total volume

100 μL

The PCR reaction should be run with the following conditions: Cycle

Step

Time

Temp. (
C)

1 (1)

Initial denaturation

3 min

95

2 (30)

Denaturation

45 s

95

Annealing

30 s

55

Elongation

90 s

72

Final elongation

10 min

72

3 (1)

8. Purify the product using the PCR purification kit protocol (see Notes 31 and 32). 9. Mix 1 μg of digested backbone and 4 μg of amplified, mutagenized insert DNA in a 1.7-mL microcentrifuge tube. Also, prepare a backbone only control. 10. Precipitate the DNA following the Pellet Paint protocol. In the final step, let the pellet air dry and do not use a heat block. 11. Once the DNA pellet is dried, resuspend in 10 μL of ddH2O. Store at 20
C until needed. 3.5.2 Preparation and Transformation of Electrocompetent Yeast

The new yeast library will be created in vivo via homologous recombination, a DNA repair mechanism naturally occurring in yeast. Electrocompetent yeast cells are generated by treating yeast cells with lithium acetate and DTT. The linearized backbone and mutagenized inserts are then transformed into the electrocompetent yeast via electroporation [94].

Yeast Surface Display for Protein Engineering

49

1. Streak RJY100 on a YPD plate and let grow 1–2 days at 30
C. 2. Pick a colony and inoculate in a 5 mL YPD culture in a 15-mL glass culture tube. Grow overnight in the 30
C shaking incubator. 3. Passage the yeast cells by transferring 100 μL of the culture into a fresh 5 mL YPD the day before electroporation. 4. Dilute the culture to an OD600 of 0.2 in YPD. Grow in the 30
C shaking incubator until an OD600 of 1.5 is reached. 50 mL of the culture at an OD600 of 1.5 will provide enough cells for two transformations (see Note 33). 5. Transfer cells to 50-mL conical tubes. Pellet the cells at 2000  g for 3 min. 6. Decant the supernatant and resuspend the cells in 25 mL of sterile 100 mM lithium acetate (half the original culture volume). Do not put more than 25 mL per tube to allow for adequate aeration during incubation. Resuspend the cells by vigorously shaking the tube. 7. Add 0.25 mL of sterile, freshly prepared 1 M DTT to each tube. To allow for adequate aeration, loosen and then tape on the caps. Incubate cells in the 30
C shaking incubator for 10 min. Following this incubation, cells should remain on ice until electroporation. 8. Tighten the lids and pellet cells at 2000  g for 3 min. 9. Decant the supernatant and add 25 mL of chilled ddH2O to each tube. Resuspend with vigorous shaking. Once resuspended, pellet cells at 2000  g for 3 min. 10. Decant the supernatant and add 0.25 mL of chilled ddH2O to each tube. Resuspend the cells via repeated pipetting and check to ensure that all cells are resuspended. The cells are now electrocompetent. 11. Thoroughly mix 250 μL of the electrocompetent yeast with the 10 μL of resuspended DNA with gentle pipetting. Leave cells on ice until immediately before electroporation. 12. One transformation at a time, transfer the cells to the prechilled 2-mm electroporation cuvette, dry the outside of the cuvette, place in the shock pad of the Bio-Rad Gene Pulser XCell, and shock the cells using a single square wave at 500 V for 15 ms. Make sure to input specifications for one pulse only and for a 2-mm cuvette. 13. After the shock, remove the cuvette from the shock pad and add 1 mL of pre-warmed YPD. Pipette out as much as possible into a 15-mL glass culture tube. 14. Add another 1 mL of YPD to cuvette to capture any remaining yeast. Add this to the same culture tube.

50

Byong H. Kang et al.

15. Incubate the newly transformed yeast at 30
C, with no shaking, for 1 h. Warm up one SDCAA plate per transformation. 16. After 1 h, remove glass culture tubes from the incubator. Vortex each tube very gently, just enough to resuspend all yeast cells. 17. Remove 10 μL from each tube and pipette into 990 μL SDCAA media in a 1.7-mL microcentrifuge tube (100 dilution). Set aside. 18. Pellet the remainder of the transformed yeast in the glass tubes at 900  g for 5 min. Carefully aspirate the supernatant. 19. Resuspend the cells in each tube with 5 mL of SDCAA. Add these 5 mL to 95 mL of SDCAA for a total of 100 mL in a baffled glass culture flask. Grow these cultures in a shaking incubator at 30
C. The transformed cells should be saturated and ready for passaging the following day. 20. Prepare three additional 1.7-mL microcentrifuge tubes for each transformation with 90 μL of SDCAA each for additional dilutions. 21. Vortex the original 100 dilution and transfer 10 μL into 90 μL SDCAA (1000 dilution). Vortex and transfer another 10 μL to the next tube with a new tip for each transfer. 22. Draw quadrants on a SDCAA plate and spread 20 μL of each dilution onto each quadrant. 23. Incubate the plates at 30
C for 2–3 days until colonies are visible. Each colony in the 100, 1000, 10,000, and 100,000 quadrants corresponds to 1  104, 1  105, 1  106, 1  107 transformants, respectively. 24. Cryopreserve newly transformed libraries using methods described in Subheading 3.2.4. Each stock should contain tenfold the estimated number of transformants. 25. Using methods described in Subheading 3.4.1, isolate DNA from yeast in the newly generated libraries. Sequencing of 10 clones should give an adequate representation of the number of mutations per clone. 26. Using methods described in Subheading 3.3.1, prepare a portion of the library to check for full-length display and proper folding using flow cytometry with detection of both the HA and c-myc tags (see Note 34). 3.6 FACS Enrichment of Binders for Affinity Maturation

Newly created yeast libraries can generally skip bead selections and immediately be subjected to FACS enrichment. It is advantageous to perform at least one equilibrium sort before moving to kinetic sorting with a newly generated library. However, kinetic sorting will be preferential to equilibrium sorts when the concentration

Yeast Surface Display for Protein Engineering

51

required to enrich for binders is below 10 nM [95]. At this concentration, the incubation volumes for equilibrium conditions become impractical. Tighter binders are instead enriched for via kinetic sorts, which enrich for binders with slower dissociation rates, which correlate to improved KDs. 3.6.1 Perform Equilibrium Sort(s) to Enrich for Full-Length Clones that Bind to Antigen of Interest

It is advantageous to perform at least one equilibrium sort will eliminate clones from the library that have mutations that abrogated binding to the target antigen, reducing the overall number of yeast cells needed for kinetic sorting. 1. As described in Subheading 3.4.2, induce, prepare, and titrate the library with the biotinylated antigen of interest. Based on the curve, select a concentration to perform the equilibrium sort that will provide the greatest potential for enrichment of clones with increased affinity (see Note 35). 2. Using methods described in Subheading 3.3.2, perform an equilibrium sort with the library at the identified concentration. Converging on the highest affinity binders is obtained with stringent diagonal gating as shown in Fig. 3. 3. Subsequent equilibrium sorts can be performed on the library if the majority of the clones have KDs greater than 10 nM for the desired antigen, based on the previously performed titration of the library. Performing equilibrium sorts at low concentrations will require larger incubation volumes (see Note 17). If the incubation volume is above 50 mL, consider moving to kinetic sorting (see Note 36).

3.6.2 Measure koff of the Library to Determine the Proper Time Scale for Kinetic Sorting

The off rate (koff) is the rate at which the binder dissociates from the target antigen. This is measured by saturating the binders with biotinylated antigen, then allowing the biotinylated antigen to dissociate and be replaced by non-biotinylated antigen for different amounts of time. 1. Incubate cells with a saturating concentration of biotinylated antigen (at least 10 the estimated KD) and allow cells to reach equilibrium (see Note 37). 2. Pellet, aspirate, and wash cells with PBSA three times. 3. Resuspend cells in PBSA with non-biotinylated antigen at a concentration at least 10–100 times the effective antibody concentration. Allow cells to incubate at room temperature for various amounts of time (ranging from 20 min to 24 h should provide an adequate curve for estimating koff). 4. Pellet, aspirate, and wash cells with PBSA three times. Resuspend in PBSA with anti-c-myc and allow to incubate for 30 min to 1 h.

52

Byong H. Kang et al.

5. Pellet, aspirate, and wash cells with PBSA three times. Resuspend in PBSA with secondaries. Allow to incubate for 15–30 min. 6. Pellet, aspirate, and wash with PBSA three times. Analyze all samples (with different dissociation times) on a flow cytometer. 7. Calculate koff (see Note 38). 3.6.3 Perform Kinetic Sorts to Enrich for HigherAffinity Binders Based on koff

Kinetic sorting allows for continued improvement of koff. Once the yeast cells have been saturated with biotinylated antigen, an excess of non-biotinylated antigen is introduced to compete off the biotinylated antigen. After sufficient incubation, the tightest binders will have the highest signal in the binding channel and are sorted by FACS. 1. Perform a kinetic sort using the same protocol as described in Subheading 3.6.2, but allow cells to dissociate for a length of time equal to 5/koff to get approximately a threefold enrichment of binders with an increased off rate (see Notes 39–42). 2. Repeat sorting until sufficient enrichment of binders is achieved. 3. Identify and characterize single clones from the library using methods described in Subheading 3.4. Multiple rounds of affinity maturation may need to be performed to identify clones with the desired affinities.

4

Notes 1. The use of SDCAA at pH 4.5 prevents bacterial contamination but does not affect the growth of yeast cells. 2. Yeast cultures need to be aerated well for optimal growth. For 16 mm diameter tubes, do not exceed 5 mL, and for culture flasks smaller than 1 L, do not exceed more than 1/3 of the volume. However, for a 2-L flask, 1 L culture can grow well with sufficient aeration. 3. To ensure that the full library diversity is screened, it is recommended that at least tenfold the library diversity is used for the enrichment steps. Thaw additional vials as needed and replenish the stocks by cryopreservation (see Subheading 3.2.4 step 16). 4. OD600 measurement is linear up to 1, which is approximately 1  107 cells/mL. For saturated or induced cultures, dilute 10–20 fold and measure OD600 to calculate the yeast cell concentration.

Yeast Surface Display for Protein Engineering

53

5. RJY100 or EBY100 doubles in SDCAA every 3–4 h and saturates at around OD600 of 8. It doubles in YPD every 1.5–2 h and saturates at around OD600 of 10. 6. If the library is expected to have members with diverse specificities, it is important to deplete binders to streptavidin and affinity tag or fusion partner if present. This can be achieved by performing magnetic bead selections, or by including non-fluorescent streptavidin, affinity tag, or fusion partner in excess (tenfold or higher) while incubating yeast cells with biotinylated antigen for initial FACS enrichment. 7. It is recommended to reserve non-biotinylated control antigen to be included during positive selection and for FACS enrichment if needed. If non-antigen specific binders are not depleted during negative selections, the control antigen can serve as a competitor to deplete them further. 8. To save time, antigen-coated beads can be prepared simultaneously by following Subheading 3.2.3. 9. Carefully transfer the liquid in the cap into the tube during this process. Whenever a tube with yeast cells is placed on a magnet, wait the full 5 min to make sure the magnetic beads end up on the side of the tube. 10. To determine the diversity of streptavidin binders, wash the beads with 1 mL of PBSA, place the tube on a magnet for 2 min, discard the supernatant, resuspend in 1 mL of SDCAA, pool, and plate serially diluted culture on SDCAA plates. 11. If there is a control antigen, prepare control antigen-coated beads by following Subheading 3.2.3 and perform negative selection by following Subheading 3.2.2 steps 8–10. 12. If there is a control antigen, adding non-biotinylated control antigen at 1 μM during this step will help prevent positive selection of binders to the affinity tag or fusion partner. 13. The recovery of cryopreserved yeast cells is greatly improved by cryopreserving at a controlled rate. This can be achieved by using a controlled rate freezer or a freezing container like Mr. Frosty (Thermo Fisher). 14. In early stages of enrichment by FACS, it is especially important to oversample to capture low-frequency binders. Sort at least 10 the diversity to recover rare binders. 15. Only the c-myc affinity tag is used as a marker for display in this section to avoid the need for three color gating and potential compensation requirements. Additional samples (105 to 106 cells each) should be prepared as controls in parallel with the sample to be sorted. These controls should include an unstained sample, a secondary only sample, and a sample for dual display (HA and c-myc).

54

Byong H. Kang et al.

16. For smaller protein scaffolds, such as the Sso7d, binding of primary and secondary antibodies to the c-myc tag can sterically hinder binding to the target antigen. For these cases, FACS should be performed using the HA tag as a marker for display instead of c-myc. 17. The concentration of biotinylated antigen used for the sorting is a function of the KD of the binders. With a naı¨ve library, it is not feasible to know the KD of the binders that will be enriched for. In this case, a concentration of 1 μM is generally good for the first sort, and then lower antigen concentrations can be used for subsequent sorts to discriminate between different affinity binders. The concentration of the antigen will determine the volume of the primary incubation. The antigen should be in tenfold molar excess to the effective antibody concentration (assume each yeast displays 105 scFvs or other binders). The minimum required volume can be determined using the following formula, where Nyeast is the number of yeast cells in solution, CAg is the molar concentration of antigen, and NA is Avogadro’s number

ð10Þ 105 N yeast V min ¼ : C Ag N A 18. Negative sorts can also be performed to remove binders to a specific protein or epitope. Negative sorts are commonly performed against affinity tags or protein fusion partners to the target antigen to ensure that binding is to an epitope on the antigen. 19. The time to equilibrium is dependent on both the antigen concentration and the apparent affinity of the binders. The equilibrium time constant (τ) can be estimated using the following formula, where CAg is the molar concentration of antigen, and kon and koff are the association and dissociation rate constants of the binder–antigen binding reaction, respectively,
1 τ ¼ kon C Ag þ koff The binding reaction will asymptotically approach equilibrium, and reaches 95% at 3τ and 99% at 4.6τ. The association and dissociation rates of the binder–antigen binding interaction are related via the equilibrium dissociation constant, KD, using the following formula: KD ¼

koff : kon

For clones from a naı¨ve library, it is reasonable to assume a kon of 105/M/s and a KD of no less than 1 nM (usually 10 nM is good for calculating τ). For most naı¨ve libraries, 3 h at room

Yeast Surface Display for Protein Engineering

55

temperature should be sufficient for the binding reaction to approach equilibrium. For an antigen concentration below 10 nM or for clones with a KD below 1 nM, a 16- to 20-h incubation at room temperature should be sufficient. 20. The stringency of the sort can be adjusted via the percentage of cells collected from the library. To favor only the highest affinity clones (at the cost of diversity), extremely stringent sorts (0.1%) can be advantageous for library convergence with fewer rounds of sorting. However, less stringent sorts (5–10%) can allow for additional, unique clones to be sorted (that may be less abundant, lower affinity, etc.), but convergence will take several additional rounds of sorting. Regardless of the stringency, the sorting gate should only include cells with high levels of expression. 21. In between sorts, cells can be run on an analyzer to check for binding to the target antigen (at 1 nM, 10 nM, 100 nM, and 1 μM) and to ensure there is no enrichment for streptavidin or other undesired binders. 22. The amount of competent yeast used for a single transformation can be reduced to 15 μL (and the amount of DNA and Solution 3 should also be scaled appropriately). The incubation time can also be increased from 45 min to 2–3 h for increased transformation efficiency. 23. Transformations of single clones can be inoculated directly into SDCAA instead of plating on SDCAA plates. At least 3 mL of media should be used for growth to sufficiently dilute the solution used for the transformation. The culture should be saturated in 2–3 days if the plating step is skipped. 24. Incubation volumes for the lowest concentrations on the titration curve can be impractical when ensuring tenfold 10 molar excess antigen as compared to the number of binders displayed on the recommended 105 cells. To reduce incubation volumes while still having sufficient yeast to pellet, add 105 carrier yeast to 2000 induced yeast cells. Yeast that have not been induced or transformed can both serve as carrier yeast. 25. If performing titrations in a 96-well v-bottom plate, pellet at 2000  g for 3 min. Wash with 200 μL of PBSA for each well. Flick supernatant off into a biological waste container to be autoclaved instead of aspirating. 26. Pre-circularized plasmids can also be used for transformation; however, the use of the native yeast machinery for homologous recombination allows for additional recombination events to occur (allowing for greater diversity) and also reduces the amount of DNA preparation steps for transformation [94].

56

Byong H. Kang et al.

27. These restriction enzymes are for the pCTcon2 plasmid available through Addgene. This plasmid contains CD20 as the insert and the backbone can be linearized by triple digestion. The pCHA plasmid on Addgene contains VRC01 scFv and should be linearized with only NheI-HF and BamHI-HF (not SalI-HF). If the digestion went to completion, smaller fragments do not have overlapping sequences with the backbone and will not form a plasmid by homologous recombination. 28. Error-prone PCR can also be performed on a library or a pool of selected clones if enrichment for one specific clone is not seen. 29. 15 cycles of error-prone PCR under these conditions is expected to yield one or more amino acid mutations per 500 bp. This can be adjusted depending on the size of the gene of interest or the number of desired amino acid mutations. 30. For cycle numbers of error-prone PCR lower than 15, it may not be possible to see the band on an agarose gel. A sizematched control (PCR reaction run for 30 cycles without the nucleotide analogs) can be run on a lane next to the errorprone PCR product (it may be beneficial to skip a well to prevent contamination during loading and excising). The portion of the gel in the lane with the error-prone PCR product that is in alignment with the control can then be excised and purified. 31. If the error-prone PCR product was not visible on the gel during Subheading 3.5.1, it is recommended to run a small amount of the amplified, mutagenized product on a gel to confirm the size and purity of the reaction. 32. Even if 4 μg of mutagenized insert DNA is not obtained from the amplification PCR reaction, electroporation with as little as 1 μg of insert DNA can be performed and will likely result in a library of at least 106 transformants. 33. When measuring the OD600 of the culture, be sure to blank the cuvette with YPD. After diluting the culture to an OD600 of 0.2, double check to make sure it is actually 0.2 and make adjustments as needed. The culture should reach an OD600 of 1.5 after around 6 h. 34. Error-prone PCR and homologous recombination can introduce mutations in affinity tags, particularly in c-myc, which can result in reduced binding by anti-c-myc antibody. This will result in a leftward shift on the display vs. antigen binding plot, which may be misinterpreted as an upward shift and improved binding. Perform quality check on the library to make sure that HA and c-myc expression levels form a diagonal and sort these populations before enriching for higher-affinity

Yeast Surface Display for Protein Engineering

57

binders. Alternatively, HA can be labeled with a different fluorophore (e.g., PE) to sort higher-affinity populations that are properly expressing both of their affinity tags. 35. If the library was created from a single clone, these titrations can also be performed with the single clone rather than the library. The KD of the binder can be calculated by fitting the titration data to a monovalent binding isotherm and solving via global nonlinear least-square regression for KD, MFImin, and MFIrange. The monovalent binding isotherm is described by the following formula, where MFItot is the total mean fluorescence intensity in the AF647 channel (antigen binding) and CAg is the molar concentration of antigen: MFItot ¼ MFImin þ

MFIrange þ C Ag K D þ C Ag

36. If the incubation volume is above 1 mL, do not incubate with biotinylated antigen and c-myc at the same time. Perform the equilibrium sort with the low concentration of biotinylated antigen, and after sufficient incubation time, pellet and resuspend cells in 100 μL PBSA (per 107 cells) with anti-c-myc antibody at a final concentration of 1 μg/mL. Allow cells to incubate with anti-c-myc antibody for 30 min to 1 h before washing and labeling with secondaries, as described in Subheading 3.4.2. 37. As in Note 35, if the library was created from a single clone, koff can be measured for the single clone instead of the library. 38. The dissociation rate constant, koff, can be calculated by fitting the dissociation data with one phase exponential decay and solving via global nonlinear least-square regression [22]. 39. If the dissociation time of a kinetic sort is longer than 24 h, it can be beneficial to pellet the cells and resuspend in fresh PBSA with non-biotinylated antigen every 12 or 24 h. This will remove any dissociated biotinylated antigen and prevent it from rebinding. 40. If enrichment for cysteine residues is seen after long kinetic sorts, N-ethylmaleimide or iodoacetamide can be used to cap cysteines on the antigen used for sorting. Incubate the biotinylated antigen with 20 molar ratio of N-ethylmaleimide or iodoacetamide (to the number of free cysteines on the antigen) for 2 h at room temperature in the dark. Excess N-ethylmaleimide or iodoacetamide can be removed by running the reaction mixture through a desalting column. Perform kinetic sorts normally. The non-biotinylated antigen does not need to be capped since the binders will be selected based on their ability to bind the biotinylated antigen.

58

Byong H. Kang et al.

41. If the incubation time for the kinetic competition is greater than 3 days, it can be advantageous to incubate at 30
C instead of room temperature to speed up the overall incubation time. 42. If the target antigen is multivalent, avid interactions may interfere with the ability to select for binders that have a true enhanced affinity as opposed to avidity, especially in kinetic sorts. The effect of avidity can be reduced by inducing cells for only 4 h instead overnight, lowering the number of binders expressed on the surface of yeast by at least an order of magnitude. References 1. Smith GP (1985) Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface. Science 228: 1315–1317. https://doi.org/10.1126/sci ence.4001944 2. Brown S (1992) Engineered iron oxideadhesion mutants of the Escherichia coli phage lambda receptor. Proc Natl Acad Sci U S A 89:8651–8655. https://doi.org/10. 1073/pnas.89.18.8651 3. Georgiou G, Stathopoulos C, Daugherty PS et al (1997) Display of heterologous proteins on the surface of microorganisms: from the screening of combinatorial libraries to live recombinant vaccines. Nat Biotechnol 15: 2 9 – 3 4 . h t t p s : // d o i . o r g / 1 0 . 1 0 3 8 / nbt0197-29 4. Hanes J, Plu¨ckthun A (1997) In vitro selection and evolution of functional proteins by using ribosome display. Proc Natl Acad Sci U S A 94: 4937–4942. https://doi.org/10.1073/pnas. 94.10.4937 5. Boder ET, Wittrup KD (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol 15:553–557. https://doi.org/10.1038/nbt0697-553 6. Nemoto N, Miyamoto-Sato E, Husimi Y, Yanagawa H (1997) In vitro virus: bonding of mRNA bearing puromycin at the 30 -terminal end to the C-terminal end of its encoded protein on the ribosome in vitro. FEBS Lett 414: 405–408. https://doi.org/10.1016/s00145793(97)01026-0 7. Roberts RW, Szostak JW (1997) RNA-peptide fusions for the in vitro selection of peptides and proteins. Proc Natl Acad Sci U S A 94: 12297–12302. https://doi.org/10.1073/ pnas.94.23.12297 8. Doi N, Yanagawa H (1999) STABLE: proteinDNA fusion system for screening of combinatorial protein libraries in vitro. FEBS Lett 457:

227–230. https://doi.org/10.1016/s00145793(99)01041-8 9. Ho M, Nagata S, Pastan I (2006) Isolation of anti-CD22 Fv with high affinity by Fv display on human cells. Proc Natl Acad Sci U S A 103: 9637–9642. https://doi.org/10.1073/pnas. 0603653103 10. Brodsky JL, Skach WR (2011) Protein folding and quality control in the endoplasmic reticulum: recent lessons from yeast and mammalian cell systems. Curr Opin Cell Biol 23:464–475. https://doi.org/10.1016/j.ceb.2011.05.004 11. Georgiou G (2001) Analysis of large libraries of protein mutants using flow cytometry. In: Advances in protein chemistry. Academic, pp 293–315 12. VanAntwerp JJ, Wittrup KD (2000) Fine affinity discrimination by yeast surface display and flow cytometry. Biotechnol Prog 16:31–37. https://doi.org/10.1021/bp990133s 13. Orr BA, Carr LM, Wittrup KD et al (2003) Rapid method for measuring ScFv thermal stability by yeast surface display. Biotechnol Prog 19:631–638. https://doi.org/10.1021/ bp0200797 14. Mei M, Li J, Wang S et al (2019) Prompting Fab yeast surface display efficiency by ER retention and molecular chaperon co-expression. Front Bioeng Biotechnol 7:362. https://doi. org/10.3389/fbioe.2019.00362 15. Miller KD, Pefaur NB, Baird CL (2008) Construction and screening of antigen targeted immune yeast surface display antibody libraries. Curr Protoc Cytom Chapter 4:Unit 4.7. https://doi.org/10.1002/0471142956. cy0407s45 16. Feldhaus MJ, Siegel RW, Opresko LK et al (2003) Flow-cytometric isolation of human antibodies from a nonimmune Saccharomyces cerevisiae surface display library. Nat

Yeast Surface Display for Protein Engineering Biotechnol 21:163–170. https://doi.org/10. 1038/nbt785 17. Van Deventer JA, Kelly RL, Rajan S et al (2015) A switchable yeast display/secretion system. Protein Eng Des Sel 28:317–325. https://doi.org/10.1093/protein/gzv043 18. Kelly RL, Le D, Zhao J, Wittrup KD (2018) Reduction of nonspecificity motifs in synthetic antibody libraries. J Mol Biol 430:119–130. https://doi.org/10.1016/j.jmb.2017.11.008 19. Rakestraw JA, Aird D, Aha PM et al (2011) Secretion-and-capture cell-surface display for selection of target-binding proteins. Protein Eng Des Sel 24:525–530. https://doi.org/ 10.1093/protein/gzr008 20. Rhiel L, Krah S, Gu¨nther R et al (2014) REALSelect: full-length antibody display and library screening by surface capture on yeast cells. PLoS One 9:e114887. https://doi.org/10. 1371/journal.pone.0114887 21. van den Beucken T, Pieters H, Steukers M et al (2003) Affinity maturation of Fab antibody fragments by fluorescent-activated cell sorting of yeast-displayed libraries. FEBS Lett 546: 288–294. https://doi.org/10.1016/s00145793(03)00602-1 22. Rosowski S, Becker S, Toleikis L et al (2018) A novel one-step approach for the construction of yeast surface display Fab antibody libraries. Microb Cell Factories 17:3. https://doi.org/ 10.1186/s12934-017-0853-z 23. Wozniak-Knopp G, Bartl S, Bauer A et al (2010) Introducing antigen-binding sites in structural loops of immunoglobulin constant domains: Fc fragments with engineered HER2/neu-binding sites and antibody properties. Protein Eng Des Sel 23:289–297. https:// doi.org/10.1093/protein/gzq005 24. Traxlmayr MW, Wozniak-Knopp G, Antes B et al (2011) Integrin binding human antibody constant domains—probing the C-terminal structural loops for grafting the RGD motif. J Biotechnol 155:193–202. https://doi.org/10. 1016/j.jbiotec.2011.06.042 25. Traxlmayr MW, Hasenhindl C, Hackl M et al (2012) Construction of a stability landscape of the CH3 domain of human IgG1 by combining directed evolution with high throughput sequencing. J Mol Biol 423:397–412. https://doi.org/10.1016/j.jmb.2012.07.017 26. McMahon C, Baier AS, Pascolutti R et al (2018) Yeast surface display platform for rapid discovery of conformationally selective nanobodies. Nat Struct Mol Biol 25:289–296. https://doi.org/10.1038/s41594-0180028-6

59

27. Uchan´ski T, Zo¨gg T, Yin J et al (2019) An improved yeast surface display platform for the screening of nanobody immune libraries. Sci Rep 9:382. https://doi.org/10.1038/ s41598-018-37212-3 28. Zielonka S, Weber N, Becker S et al (2014) Shark attack: high affinity binding proteins derived from shark vNAR domains by stepwise in vitro affinity maturation. J Biotechnol 191: 236–245. https://doi.org/10.1016/j.jbiotec. 2014.04.023 29. Ko¨nning D, Rhiel L, Empting M et al (2017) Semi-synthetic vNAR libraries screened against therapeutic antibodies primarily deliver antiidiotypic binders. Sci Rep 7:9676. https:// doi.org/10.1038/s41598-017-10513-9 30. Ko¨nning D, Hinz S, Grzeschik J et al (2018) Construction of histidine-enriched shark IgNAR variable domain antibody libraries for the isolation of pH-sensitive vNAR fragments. In: Nevoltris D, Chames P (eds) Antibody engineering: methods and protocols. Springer, New York, NY, pp 109–127 ˜ as S, Ayala-Avila M 31. Cabanillas-Bernal O, Duen et al (2019) Synthetic libraries of shark vNAR domains with different cysteine numbers within the CDR3. PLoS One 14:e0213394. https://doi.org/10.1371/journal.pone. 0213394 32. Kieke MC, Shusta EV, Boder ET et al (1999) Selection of functional T cell receptor mutants from a yeast surface-display library. Proc Natl Acad Sci U S A 96:5651–5656. https://doi. org/10.1073/pnas.96.10.5651 33. Holler PD, Holman PO, Shusta EV et al (2000) In vitro evolution of a T cell receptor with high affinity for peptide/MHC. Proc Natl Acad Sci U S A 97:5387–5392. https://doi. org/10.1073/pnas.080078297 34. Chlewicki LK, Holler PD, Monti BC et al (2005) High-affinity, peptide-specific T cell receptors can be generated by mutations in CDR1, CDR2 or CDR3. J Mol Biol 346: 223–239. https://doi.org/10.1016/j.jmb. 2004.11.057 35. Xu G, Tasumi S, Pancer Z (2011) Yeast surface display of lamprey variable lymphocyte receptors. Methods Mol Biol 748:21–33. https:// doi.org/10.1007/978-1-61779-139-0_2 36. Koide A, Bailey CW, Huang X, Koide S (1998) The fibronectin type III domain as a scaffold for novel binding proteins. J Mol Biol 284: 1141–1151. https://doi.org/10.1006/jmbi. 1998.2238 37. Hackel BJ, Kapila A, Wittrup KD (2008) Picomolar affinity fibronectin domains engineered

60

Byong H. Kang et al.

utilizing loop length diversity, recursive mutagenesis, and loop shuffling. J Mol Biol 381: 1238–1252. https://doi.org/10.1016/j.jmb. 2008.06.051 38. Chen TF, de Picciotto S, Hackel BJ, Wittrup KD (2013) Chapter fourteen—engineering fibronectin-based binding proteins by yeast surface display. In: Keating AE (ed) Methods in enzymology. Academic, pp 303–326 39. Pavoor TV, Cho YK, Shusta EV (2009) Development of GFP-based biosensors possessing the binding properties of antibodies. Proc Natl Acad Sci U S A 106:11895–11900. https://doi.org/10.1073/pnas.0902828106 40. Gera N, Hussain M, Wright RC, Rao BM (2011) Highly stable binding proteins derived from the hyperthermophilic Sso7d scaffold. J Mol Biol 409:601–616. https://doi.org/10. 1016/j.jmb.2011.04.020 41. Traxlmayr MW, Kiefer JD, Srinivas RR et al (2016) Strong enrichment of aromatic residues in binding sites from a charge-neutralized hyperthermostable Sso7d scaffold library. J Biol Chem 291:22496–22508. https://doi. org/10.1074/jbc.M116.741314 42. Silverman AP, Levin AM, Lahti JL, Cochran JR (2009) Engineered cystine-knot peptides that bind alpha(v)beta(3) integrin with antibodylike affinities. J Mol Biol 385:1064–1075. https://doi.org/10.1016/j.jmb.2008.11.004 43. Lee C-H, Park K-J, Sung E-S et al (2010) Engineering of a human kringle domain into agonistic and antagonistic binding proteins functioning in vitro and in vivo. Proc Natl Acad Sci U S A 107:9567–9571. https://doi. org/10.1073/pnas.1001541107 44. Stern LA, Lown PS, Kobe AC et al (2019) Cellular-based selections aid yeast-display discovery of genuine cell-binding ligands: targeting oncology vascular biomarker CD276. ACS Comb Sci 21:207–222. https://doi.org/10. 1021/acscombsci.8b00156 45. Jing L, Liu J, Cui D et al (2020) Screening and production of an affibody inhibiting the interaction of the PD-1/PD-L1 immune checkpoint. Protein Expr Purif 166:105520. https://doi.org/10.1016/j.pep.2019.105520 46. Binz HK, Amstutz P, Kohl A et al (2004) High-affinity binders selected from designed ankyrin repeat protein libraries. Nat Biotechnol 22:575–582. https://doi.org/10.1038/ nbt962 47. Kruziki MA, Bhatnagar S, Woldring DR et al (2015) A 45-amino-acid scaffold mined from the PDB for high-affinity ligand engineering. Chem Biol 22:946–956. https://doi.org/10. 1016/j.chembiol.2015.06.012

48. Boder ET, Raeeszadeh-Sarmazdeh M, Price JV (2012) Engineering antibodies by yeast display. Arch Biochem Biophys 526:99–106. https:// doi.org/10.1016/j.abb.2012.03.009 49. Cherf GM, Cochran JR (2015) Applications of yeast surface display for protein engineering. Methods Mol Biol 1319:155–175. https:// doi.org/10.1007/978-1-4939-2748-7_8 50. Mei M, Zhou Y, Peng W et al (2017) Application of modified yeast surface display technologies for non-antibody protein engineering. Microbiol Res 196:118–128. https://doi. org/10.1016/j.micres.2016.12.002 51. Richman SA, Healan SJ, Weber KS et al (2006) Development of a novel strategy for engineering high-affinity proteins by yeast display. Protein Eng Des Sel 19:255–264. https://doi. org/10.1093/protein/gzl008 52. Yang Z, Wan Y, Tao P et al (2019) A cell-cell interaction format for selection of high-affinity antibodies to membrane proteins. Proc Natl Acad Sci U S A 116:14971–14978. https:// doi.org/10.1073/pnas.1908571116 53. Garcia-Rodriguez C, Levy R, Arndt JW et al (2007) Molecular evolution of antibody crossreactivity for two subtypes of type A botulinum neurotoxin. Nat Biotechnol 25:107–116. https://doi.org/10.1038/nbt1269 54. Angelini A, Miyabe Y, Newsted D et al (2018) Directed evolution of broadly crossreactive chemokine-blocking antibodies efficacious in arthritis. Nat Commun 9:1461. https://doi. org/10.1038/s41467-018-03687-x 55. Gaiotto T, Hufton SE (2016) Crossneutralising nanobodies bind to a conserved pocket in the hemagglutinin stem region identified using yeast display and deep mutational scanning. PLoS One 11:e0164296. https:// doi.org/10.1371/journal.pone.0164296 56. Gaiotto T, Ramage W, Ball C et al (2021) Nanobodies mapped to cross-reactive and divergent epitopes on A(H7N9) influenza hemagglutinin using yeast display. Sci Rep 11: 3126. https://doi.org/10.1038/s41598021-82356-4 57. Wadle A, Mischo A, Imig J et al (2005) Serological identification of breast cancer-related antigens from a Saccharomyces cerevisiae surface display library. Int J Cancer 117:104–113. https://doi.org/10.1002/ijc.21147 58. Wadle A, Kubuschok B, Imig J et al (2006) Serological immune response to cancer testis antigens in patients with pancreatic cancer. Int J Cancer 119:117–125. https://doi.org/10. 1002/ijc.21744

Yeast Surface Display for Protein Engineering 59. Cochran JR, Kim Y-S, Lippow SM et al (2006) Improved mutants from directed evolution are biased to orthologous substitutions. Protein Eng Des Sel 19:245–253. https://doi.org/ 10.1093/protein/gzl006 60. Rao BM, Girvin AT, Ciardelli T et al (2003) Interleukin-2 mutants with enhanced α-receptor subunit binding affinity. Protein Eng Des Sel 16:1081–1087. https://doi.org/ 10.1093/protein/gzg111 61. Rao BM, Driver I, Lauffenburger DA, Wittrup KD (2005) High-affinity CD25-binding IL-2 mutants potently stimulate persistent T cell growth. Biochemistry 44:10696–10701. https://doi.org/10.1021/bi050436x 62. Shpilman M, Niv-Spector L, Katz M et al (2011) Development and characterization of high affinity leptins and leptin antagonists. J Biol Chem 286:4429–4442. https://doi.org/ 10.1074/jbc.M110.196402 63. Kariolis MS, Miao YR, Jones DS 2nd et al (2014) An engineered Axl “decoy receptor” effectively silences the Gas6-Axl signaling axis. Nat Chem Biol 10:977–983. https://doi.org/ 10.1038/nchembio.1636 64. Weiskopf K, Ring AM, Ho CCM et al (2013) Engineered SIRPα variants as immunotherapeutic adjuvants to anticancer antibodies. Science 341:88–91. https://doi.org/10.1126/ science.1238856 65. Zaretsky M, Etzyoni R, Kaye J et al (2013) Directed evolution of a soluble human IL-17A receptor for the inhibition of psoriasis plaque formation in a mouse model. Chem Biol 20:202–211. https://doi.org/10.1016/ j.chembiol.2012.11.012 66. Maute RL, Gordon SR, Mayer AT et al (2015) Engineering high-affinity PD-1 variants for optimized immunotherapy and immuno-PET imaging. Proc Natl Acad Sci U S A 112: E6506–E6514. https://doi.org/10.1073/ pnas.1519623112 67. Lipovsek D, Antipov E, Armstrong KA et al (2007) Selection of horseradish peroxidase variants with enhanced enantioselectivity by yeast surface display. Chem Biol 14:1176–1185. https://doi.org/10.1016/j.chembiol.2007. 09.008 68. Antipov E, Cho AE, Wittrup KD, Klibanov AM (2008) Highly L and D enantioselective variants of horseradish peroxidase discovered by an ultrahigh-throughput selection method. Proc Natl Acad Sci U S A 105:17694–17699. https://doi.org/10.1073/pnas.0809851105 69. Chen I, Dorr BM, Liu DR (2011) A general strategy for the evolution of bond-forming enzymes using yeast display. Proc Natl Acad

61

Sci U S A 108:11399–11404. https://doi. org/10.1073/pnas.1101046108 70. Fushimi T, Miura N, Shintani H et al (2013) Mutant firefly luciferases with improved specific activity and dATP discrimination constructed by yeast cell surface engineering. Appl Microbiol Biotechnol 97:4003–4011. https://doi.org/10.1007/s00253-0124467-4 71. Yi L, Gebhard MC, Li Q et al (2013) Engineering of TEV protease variants by yeast ER sequestration screening (YESS) of combinatorial libraries. Proc Natl Acad Sci U S A 110: 7229–7234. https://doi.org/10.1073/pnas. 1215994110 72. Ostafe R, Prodanovic R, Nazor J, Fischer R (2014) Ultra-high-throughput screening method for the directed evolution of glucose oxidase. Chem Biol 21:414–421. https://doi. org/10.1016/j.chembiol.2014.01.010 73. Cochran JR, Kim Y-S, Olsen MJ et al (2004) Domain-level antibody epitope mapping through yeast surface display of epidermal growth factor receptor fragments. J Immunol Methods 287:147–158. https://doi.org/10. 1016/j.jim.2004.01.024 74. Chao G, Cochran JR, Wittrup KD (2004) Fine epitope mapping of anti-epidermal growth factor receptor antibodies through random mutagenesis and yeast surface display. J Mol Biol 342:539–550. https://doi.org/10.1016/j. jmb.2004.07.053 75. Han T, Sui J, Bennett AS et al (2011) Fine epitope mapping of monoclonal antibodies against hemagglutinin of a highly pathogenic H5N1 influenza virus using yeast surface display. Biochem Biophys Res Commun 409: 253–259. https://doi.org/10.1016/j.bbrc. 2011.04.139 76. Mata-Fink J, Kriegsman B, Yu HX et al (2013) Rapid conformational epitope mapping of antigp120 antibodies with a designed mutant panel displayed on yeast. J Mol Biol 425:444–456. https://doi.org/10.1016/j.jmb.2012.11.010 77. Van Blarcom T, Rossi A, Foletti D et al (2018) Epitope mapping using yeast display and next generation sequencing. Methods Mol Biol 1785:89–118. https://doi.org/10.1007/ 978-1-4939-7841-0_7 78. Kowalsky CA, Faber MS, Nath A et al (2015) Rapid fine conformational epitope mapping using comprehensive mutagenesis and deep sequencing *. J Biol Chem 290: 26457–26470. https://doi.org/10.1074/jbc. M115.676635 79. Wen F, Esteban O, Zhao H (2008) Rapid identification of CD4+ T-cell epitopes using yeast

62

Byong H. Kang et al.

displaying pathogen-derived peptide library. J Immunol Methods 336:37–44. https://doi. org/10.1016/j.jim.2008.03.008 80. Jiang W, Boder ET (2010) High-throughput engineering and analysis of peptide binding to class II MHC. Proc Natl Acad Sci U S A 107: 13258–13263. https://doi.org/10.1073/ pnas.1006344107 81. Rappazzo CG, Huisman BD, Birnbaum ME (2020) Repertoire-scale determination of class II MHC peptide binding via yeast display improves antigen prediction. Nat Commun 11:4414. https://doi.org/10.1038/s41467020-18204-2 82. Birnbaum ME, Mendoza JL, Sethi DK et al (2014) Deconstructing the peptide-MHC specificity of T cell recognition. Cell 157: 1073–1087. https://doi.org/10.1016/j.cell. 2014.03.047 83. Bowley DR, Labrijn AF, Zwick MB, Burton DR (2007) Antigen selection from an HIV-1 immune antibody library displayed on yeast yields many novel antibodies compared to selection from the same library displayed on phage. Protein Eng Des Sel 20:81–90. https://doi.org/10.1093/protein/gzl057 84. Cappellaro C, Baldermann C, Rachel R, Tanner W (1994) Mating type-specific cell-cell recognition of Saccharomyces cerevisiae: cell wall attachment and active sites of a- and alphaagglutinin. EMBO J 13:4737–4744. https:// doi.org/10.1002/j.1460-2075.1994. tb06799.x 85. Jacobs PP, Ryckaert S, Geysens S et al (2008) Pichia surface display: display of proteins on the surface of glycoengineered Pichia pastoris strains. Biotechnol Lett 30:2173–2181. https://doi.org/10.1007/s10529-0089807-1 86. Wasilenko JL, Sarmento L, Spatz S, PantinJackwood M (2010) Cell surface display of highly pathogenic avian influenza virus hemagglutinin on the surface of Pichia pastoris cells using alpha-agglutinin for production of oral vaccines. Biotechnol Prog 26:542–547. https://doi.org/10.1002/btpr.343

87. Yang X, Tang H, Song M et al (2019) Development of novel surface display platforms for anchoring heterologous proteins in Saccharomyces cerevisiae. Microb Cell Factories 18:85. https://doi.org/10.1186/s12934-0191133-x 88. Colby DW, Kellogg BA, Graff CP et al (2004) Engineering antibody affinity by yeast surface display. In: Methods in enzymology. Academic, pp 348–358 89. Chao G, Lau WL, Hackel BJ et al (2006) Isolating and engineering human antibodies using yeast surface display. Nat Protoc 1:755–768. https://doi.org/10.1038/nprot.2006.94 90. Orcutt KD, Wittrup KD (2010) Yeast display and selections. In: Kontermann R, Du¨bel S (eds) Antibody engineering. Springer, Berlin, pp 207–233 91. Van Deventer JA, Wittrup KD (2014) Yeast surface display for antibody isolation: library construction, library screening, and affinity maturation. In: Ossipow V, Fischer N (eds) Monoclonal antibodies: methods and protocols. Humana Press, Totowa, NJ, pp 151–181 92. Angelini A, Chen TF, de Picciotto S et al (2015) Protein engineering and selection using yeast surface display. In: Liu B (ed) Yeast surface display: methods, protocols, and applications. Springer, New York, NY, pp 3–36 93. Lipovsek D, Lippow SM, Hackel BJ et al (2007) Evolution of an interloop disulfide bond in high-affinity antibody mimics based on fibronectin type III domain and selected by yeast surface display: molecular convergence with single-domain camelid and shark antibodies. J Mol Biol 368:1024–1041. https://doi. org/10.1016/j.jmb.2007.02.029 94. Swers JS, Kellogg BA, Wittrup KD (2004) Shuffled antibody libraries created by in vivo homologous recombination and yeast surface display. Nucleic Acids Res 32:e36. https://doi. org/10.1093/nar/gnh030 95. Boder ET, Wittrup KD (1998) Optimal screening of surface-displayed polypeptide libraries. Biotechnol Prog 14:55–62

Chapter 3 Site-wise Diversification of Combinatorial Libraries Using Insights from Structure-guided Stability Calculations Benedikt Dolgikh and Daniel Woldring Abstract Many auspicious clinical and industrial accomplishments have improved the human condition by means of protein engineering. Despite these achievements, our incomplete understanding of the sequence–structure–function relationship prevents rapid innovation. To tackle this problem, we must develop and integrate new and existing technologies. To date, directed evolution and rational design have dominated as protein engineering principles. Even so, prior to screening for novel or improved functions, a large collection of variants, within a protein library, exist along an ambiguous mutational terrain. Complicating things further, the choice of where to initialize investigation along a vast sequence space becomes even more difficult given that the majority of any sequence lacks function entirely. Unfortunately, even when considering functionally relevant positions, random substitutions can prove to be destabilizing, causing a hindrance to an otherwise function-inducing, stability-reliant folding process. To enhance productivity in the field, we seek to address this issue of destabilization, and subsequent disfunction, at protein–protein and protein–ligand interacting regions. Herein, the process of choosing amenable positions – and amino acids at those positions – allows for a refined, knowledge-based approach to combinatorial library design. Using structural data, we perform computational stability prediction with FoldX’s PositionScan and Rosetta’s ddG_monomer in tandem, allowing for the refinement of our thermodynamic stability data through the comparison of results. In turn, we provide a process for selecting in silico predicted mutually stabilizing positions and avoiding overly destabilizing ones that guides the site-wise diversification of combinatorial libraries. Key words Computational, Protein engineering, Stability, Library design, Site-wise diversification

1

Introduction Engineering novel or improved protein function often requires making several mutations at various positions. Given that random mutations tend to be destabilizing [1] and that most natural proteins are only marginally stable to begin with, designing functional protein variants necessitates the exclusion of overly destabilizing mutations. Identifying the most mutationally amenable positions is assisted by computational modeling tools such as Rosetta’s

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_3, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

63

64

Benedikt Dolgikh and Daniel Woldring

ddg_monomer [2] and ddg_cartesian [3] applications or FoldX PositionScan and BuildModel commands [4] (see Note 1). To screen proteins with computational tools for engineered design and functionality, a computational approach is limited by the type of data, that is a structural model or protein sequence, and the number or caliber of extractable properties, such as high-resolution structural data or sufficiently large deep sequencing datasets. In locating amenable candidates for library design, a stability prediction tool greatly depends on the quality of input structure [5]. Fortunately, the physicochemical terms that influence tertiary structure are present in high-resolution structural information and necessary for energy function calculations [6, 7]. By modeling the mechanics of functional regions to explore their potential for thermodynamic variation, structurally guided stability prediction detects positions that are most and least amenable for site-wise diversification of combinatorial libraries [8]. Stability prediction tools iteratively sample an ensemble of conformations from high-resolution structural information, applying an energy function or force field to acquire the most stable fold of a given protein [2, 9]. Energy functions for stability prediction can be varied in their accuracy (see Note 2) and computational resource usage [10, 11] (see Note 3) in a multitude of use cases [12–18]. In our case, computationally predicted stability difference (ΔΔG) between wildtype and mutant folds helps to understand the mode for how variation in stability influences protein folding and unfolding. Going one step further in our methodology, a recent comparison of experimental stability data to the combined prediction of FoldX and Rosetta has rationalized the efficacy of combinatorial approaches [19]. Recently, the discussion of integrating multiple computational tools for protein engineering has gained momentum [20–22]. By maximizing the utility between two stability prediction tools, we present the opportunity to seek out additional computational approaches (Chapters 4 and 5 of this volume); the drawbacks of separate datatypes still converge at their source system, biomolecule, or protein. To achieve more accurate predictions we combine the data from FoldX, a user friendly tool [23] for researchers with little computational experience, and Rosetta’s ddg_monomer, implementing backbone flexibility in structural minimization that parallels more realistic structural changes upon mutagenesis [2]. Overall, the informative, albeit imperfect [12, 19, 24, 25], stability prediction capabilities of Rosetta and FoldX force fields enable site-wise diversification, creating a platform for combinatorial libraries [26–29]. Here, we locate positions of interest around a known active site location, followed by our general protocols for performing stability prediction within Rosetta and FoldX (Fig. 1).

Stability Prediction for Site-Wise Combinatorial Libraries

A

B

65

C FoldX PositionScan

Rosetta HighRes (or LowRes)

F

D

E Clone Library

=

G H2 H1 I K L

Amino Acids

M N P Q R S T V W

Amino Acids

ΔΔG

Y

Mutually Stabilizing Posions (i) and Amino Acids (x)

Stabilizing

F

Positions Rosea HighRes (or LowRes)

FoldX PosionScan

Destabilizing

Fig. 1 Computational stability prediction workflow. (a) The active site (Green) within an enzyme is evaluated for mutational tolerance. (b) Residues proximal to the active site are chosen. (c) Multiple stability prediction tools independently calculate all possible point mutations surrounding the active site resulting in a range of stabilizing (Red) and destabilizing (Blue) effects. (d) Graphical analysis of ΔΔG data provides insights for rational library design. (e) The two datasets inform library size by the product of the number of mutually stabilizing amino acids (x) offered at each position of interest (i). (f) Library construction with overlap extension PCR uses degenerate codon oligonucleotides to introduce diversity. Protein and residue images were created using PyMOL

2

Materials 1. A UNIX-based operating system available on your local desktop or through a computing cluster (see Note 2) for running command line and python scripts (see Note 4). 2. Download and unpack FoldX 5.0 (http://foldxsuite.crg.eu/). 3. Download, unpack, and compile Rosetta 3.12 source + binaries (https://www.rosettacommons.org/) (see Notes 4 and 5). 4. PDB structure file(s) for a protein of interest (see Note 6). 5. Download and unpack PyMOL (https://pymol.org/2/).

3

Methods

3.1 Choosing Residues of Interest

1. Open PyMOL and load PDB structures for your protein of interest. 2. Locate the active site, then highlight positions within 5 Å of the selection using:

66

Benedikt Dolgikh and Daniel Woldring show sticks, byres all within 5 of

(see Notes 7 and 8). 3. Record and use these positions of interest in all downstream stability prediction. 3.2 FoldX Stability Prediction

1. Using a text editor (see Note 9), create a file called RepairPDB_.cfg.

2. Use .cfg file templates from GitHub to optimize your configuration file (see Note 10). 3. Run the FoldX RepairPDB command (see Note 11) for each PDB structure separately using: ./foldx -f ./RepairPDB_.cfg

4. Repeat steps 1 and 2 but name the .cfg file me>_Repair.cfg.

PS_

2. Run the minimize_with_cst.linuxgccrelease executable (see Note 16) in the directory with all necessary files: /PATH/TO/ROSETTA/main/source/bin/minimize_with_cst.linuxgccrelease \ -in:file:l ./ \ -in:file:fullatom \ -ignore_unrecognized_res \ -fa_max_dis 9.0 \

68

Benedikt Dolgikh and Daniel Woldring -database /PATH/TO/ROSETTA/main/database/ \ -ddg::harmonic_ca_tether 0.5 \ -score:weights pre_talaris_2013_standard \ -restore_pre_talaris_2013_behavior \ -ddg::constraint_weight 1.0 \ -ddg::out_pdb_prefix min_cst_0.5 \ -ddg::sc_min_only false \ -score:patch /PATH/TO/ROSETTA/main/database/ scoring/weights/score12.wts_patch > mincst.log

3. After pre-minimization, convert the mincst.log file to a distance-restraint file (see Notes 10 and 18): tcsh

./convert_to_cst_file.sh

./mincst.log

>

.cst

4. Run the ddg_monomer.linuxgccrelease executable in the directory with all necessary files: /PATH/TO/ROSETTA/main/source/bin/ddg_monomer.linuxgccrelease \ -in:file:s ./min_cst_0.5._001. pdb \ -ddg::weight_file soft_rep_design \ -ddg:minimization_scorefunction pre_talaris_2013_standard \ -restore_pre_talaris_2013_behavior \ -ddg::minimization_patch /PATH/TO/ROSETTA/main/ database/scoring/weights/score12.wts_patch \ -ddg::iterations 50 \ -ddg::dump_pdbs true \ -database /PATH/TO/ROSETTA/main/database/ \ -ignore_unrecognized_res \ -ddg::local_opt_only false \ -ddg::min_cst true \ -ddg::mean false \ -ddg::min true \ -ddg::sc_min_only false \ -in:file:fullatom \ -resfile ./.resfile \ -ignore_zero_occupancy false \ -ddg::ramp_repulsive true \ -constraints::cst_file ./.cst \ -ddg:suppress_checkpointing true

5. Use the ΔΔGfolding (REU) values from the ddg_predictions.out (first value in row, “total”) file for downstream

Stability Prediction for Site-Wise Combinatorial Libraries

69

graphical analyses and degenerate codon design (Subheading 3.4). Examples are provided on GitHub (see Note 10). 3.4 Library Design for Overlap Extension PCR and Electroporation into Yeast

1. For library design, at each analyzed position, enumerate the number of residues that balance the stringency of stabilizing mutations (e.g. between ΔΔG <
1.5–0 kcal/mol) with the estimated library size (e.g. 109 for Yeast Surface Display) to cater to the library design goals. Create a list of individual positions and the corresponding amino acids that are most amenable to mutation (i.e., sites that have the greatest number of stabilizing mutations). Aggregate these data with the results from both stability prediction tools. 2. Based on the created list, identify a degenerate codon that encodes for an identical or similar distribution of amino acids for each site (see Note 19). Examples are provided on GitHub (see Note 10). 3. Order oligonucleotide sequences that incorporate degenerate codons at the position(s) determined in step 1. 4. Construct full length gene using overlap extension PCR [27, 30–32]. Electroporate the newly constructed insert library with an appropriate, linearized yeast surface display vector (e.g., pCTcon2 [33]) [26–29, 34].

4

Notes 1. Web-servers such as FireProt [35], PoPMuSiC [36], ELASPIC [37], DUET [38], and HotSpotWizard [39] reduce computational demand from a user’s home desktop. There also exist combinatorial tools such as FRESCO [40], which has been optimized for integration of FoldX and Rosetta stability prediction in combination with the Dynamic Disulfide Discovery algorithm. 2. Limitations of accuracy: To further elaborate on this considerable limitation in stability prediction tools it is important to understand the breadth of variable training data. Energy functions have been trained using experimental data where many variants contain alanine substitution mutations, such as in FoldX, and many training datasets largely consist of destabilizing mutations. 3. Computing cluster resources are highly recommended for performing these commands. Rosetta’s ddg_monomer utilizes significant resources compared to FoldX, but both tools increase computational demand with the number of mutated positions, the mutational approach (e.g., all 20 amino acids, only charged amino acids, etc.), and iterations of structural energy minimization and optimization.

70

Benedikt Dolgikh and Daniel Woldring

4. Version information for other necessary modules: Python – 2.7.16 – https://www.python.org/; Biopython – 1.75 (Use pip install. For pdb_renumber.py); GCC – 8.3.0 https://gcc.gnu.org/; OpenMPI – 3.1.4 - https://www. open-mpi.org/ 5. Rosetta is parallelizable; calculation speed mainly depends on a Rosetta application’s utilization of computational resources (https://www.rosettacommons.org/docs/latest/build_docu mentation/Build-Documentation). 6. The number of structures analyzed is dependent on structures available through databases such as https://www.rcsb.org/, but should ideally be as many as are available. 7. To make a selection, simply click on a residue(s) and change the selection name by moving the mouse over to (sele), click A for Action, click rename selection, rename, and click ENTER on your keyboard (refer to Note 10). After running the command, consider starting with 20 positions to limit computational demand; adjust as needed. 8. Any code is in this FONT. Any code in these brackets should be replaced with the name of your choice (example: .pdb ¼ 1A2B.pdb). The. / before a command indicates that the file or executable is in the current directory. 9. A frequently used text editor in UNIX is called vim. It will allow you to read or create .txt, .cfg, .pdb, .out, .log, .resfile, and other files containing text. To use vim, type: vi .txt (example:

vi .

pdb)

To edit a vim file, press A. Exit editing by ESC. Save and exit your file by typing: wq. 10. Templates and examples for input and output files/data can be found here: https://GitHub.com/WoldringLabMSU/ Computational-Stability 11. More information about RepairPDB can be found here: http://foldxsuite.crg.eu/command/RepairPDB 12. More information about PositionScan can be found here: http://foldxsuite.crg.eu/command/PositionScan 13. Long commands are separated using a \ back slash to indicate that the line continues. 14. The authors of ddg_monomer indicate many approaches for stability prediction with their tool [2], two of which are used here. The major difference between both is the addition of distance-restraints for backbone flexibility during high-

Stability Prediction for Site-Wise Combinatorial Libraries

71

resolution minimization. As a result, the high-resolution method requires considerable computational time compared to the low-resolution method. Regardless of approach, using FoldX with either of them is still useful for improving predictive accuracy [19]. 15. Rosetta executable nomenclature varies depending on the Rosetta package that you downloaded and the coinciding method for compiling. While ddg_monomer.linuxgccrelease is most referenced, variations exist such as ddg_monoor ddg_monomer. mer.static.linuxgccrelease default.linuxgccrelease. For more information refer to Note 9. 16. More information about ddg_monomer can be found here: https://www.rosettacommons.org/docs/latest/application_ documentation/analysis/ddg-monomer 17. Rosetta uses arbitrary units that replace kcal/mol. The energy function and corresponding units we have chosen to use for our protocol is optimized for this Rosetta application. Newer energy functions and those that correlate more closely to kcal/ mol can be explored in recent literature on the Rosetta scorefunction [41]. 18. The tcsh command may not work on a computing cluster without administrator permissions, so you will need to transfer the convert_to_cst_file.sh file and the mincst.log file to your local desktop. To copy files from a server to a local desktop or otherwise, the scp -r command arguments will vary depending on your user ID (refer to Navigating Rosetta to run executables from Note 10). 19. When designing degenerate codons for a particular position, while the amino acid distribution encoded by the degenerate bases may not perfectly match the intended distribution of residues predicted to be most stabilizing at that site, care should be taken to avoid including residues that were predicted to be overly destabilizing (e.g. ΔΔG > 1.5 kcal/mol) [27]. Online tools such as SwiftLib are available to guide the degenerate codon design process [42]. References 1. Baase WA, Liu L, Tronrud DE et al (2010) Lessons from the lysozyme of phage T4. Protein Sci 19:631–641 2. Kellogg EH, Leaver-Fay A, Baker D (2011) Role of conformational sampling in computing mutation-induced changes in protein structure and stability. Proteins 79:830–838 3. Park H, Bradley P, Greisen P et al (2016) Simultaneous optimization of biomolecular energy functions on features from small

molecules and macromolecules. J Chem Theory Comput 12:6201–6212 4. Delgado J, Radusky LG, Cianferoni D et al (2019) FoldX 5.0: working with RNA, small molecules and a new graphical interface. Bioinformatics 35:4168–4169 5. Davey JA, Chica RA (2015) Optimization of rotamers prior to template minimization improves stability predictions made by

72

Benedikt Dolgikh and Daniel Woldring

computational protein design. Protein Sci 24: 545–560 6. Buß O, Rudat J, Ochsenreither K (2018) FoldX as protein engineering tool: better than random based approaches? Comput Struct Biotechnol J 16:25–33 7. Hou T, Wang J, Li Y et al (2011) Assessing the performance of the MM/PBSA and MM/GBSA methods. 1. The accuracy of binding free energy calculations based on molecular dynamics simulations. J Chem Inf Model 51: 69–82 8. Tokuriki N, Stricher F, Serrano L et al (2008) How protein stability and new functions trade off. PLoS Comput Biol 4:e1000002 9. Naganathan AN (2019) Modulation of allosteric coupling by mutations: from protein dynamics and packing to altered native ensembles and function. Curr Opin Struct Biol 54: 1–9 10. Pohorille A, Jarzynski C, Chipot C (2010) Good practices in free-energy calculations. J Phys Chem B 114:10235–10253 11. Klimovich PV, Shirts MR, Mobley DL (2015) Guidelines for the analysis of free energy calculations. J Comput Aided Mol Des 29:397–411 12. Strokach A, Corbi-Verge C, Kim PM (2019) Predicting changes in protein stability caused by mutation using sequence-and structurebased methods in a CAGI5 blind challenge. Hum Mutat 40:1414–1423 13. Mazurenko S (2020) Predicting protein stability and solubility changes upon mutations: data perspective. ChemCatChem 12:5590–5598 14. Beauchamp KA, Lin YS, Das R et al (2012) Are protein force fields getting better? A systematic benchmark on 524 diverse NMR measurements. J Chem Theory Comput 8:1409–1414 15. Pucci F, Bernaerts KV, Kwasigroch JM et al (2018) Quantification of biases in predictions of protein stability changes upon mutations. Bioinformatics 34:3659–3665 16. Thiltgen G, Goldstein RA (2012) Assessing predictors of changes in protein stability upon mutation using self-consistency. PLoS One 7: 46084 17. Huang P, Chu SKS, Frizzo HN et al (2020) Evaluating protein engineering thermostability prediction tools using an independently generated dataset. ACS Omega 5:6487–6493 18. Kumar V, Rahman S, Choudhry H et al (2017) Computing disease-linked SOD1 mutations: deciphering protein stability and patientphenotype relations article. Sci Rep 7:1–13 19. Nisthal A, Wang CY, Ary ML et al (2019) Protein stability engineering insights revealed

by domain-wide comprehensive mutagenesis. Proc Natl Acad Sci U S A 116:16367–16377 20. Adolf-Bryfogle J, Teets FD, Bahl CD (2021) Toward complete rational control over protein structure and function through computational design. Curr Opin Struct Biol 66:170–177 21. Sun J, Cui Y, Wu B (2021) GRAPE, a greedy accumulated strategy for computational protein engineering. In: Methods in enzymology. Academic, pp 207–230 22. Soni S (2021) Trends in lipase engineering for enhanced biocatalysis. Biotechnol Appl Biochem 59:13204–13231 23. Van DJ, Delgado J, Stricher F et al (2011) A graphical interface for the FoldX forcefield. Bioinformatics 27:1711–1712 24. Khan S, Vihinen M (2010) Performance of protein stability predictors. Hum Mutat 31: 675–684 25. Potapov V, Cohen M, Schreiber G (2009) Assessing computational methods for predicting protein stability upon mutation: good on average but not in the details. Protein Eng Des Sel 22:553–560 26. Woldring DR, Holec PV, Zhou H et al (2015) High-throughput ligand discovery reveals a sitewise gradient of diversity in broadly evolved hydrophilic fibronectin domains. PLoS One 10:e0138956 27. Woldring DR, Holec PV, Stern LA et al (2017) A gradient of sitewise diversity promotes evolutionary fitness for binder discovery in a threehelix bundle protein scaffold. Biochemistry 56: 1656–1671 28. Kruziki MA, Bhatnagar S, Woldring DR et al (2015) A 45-amino-acid scaffold mined from the PDB for high-affinity ligand engineering. Chem Biol 22:946–956 29. Kruziki MA, Sarma V, Hackel BJ (2018) Constrained combinatorial libraries of Gp2 proteins enhance discovery of PD-L1 binders. ACS Comb Sci 20:423–435 30. Bryksin AV, Matsumura I (2010) Overlap extension PCR cloning: a simple and reliable way to create recombinant plasmids. BioTechniques 48:463–465 31. Schimming O, Fleischhacker F, Nollmann FI et al (2014) Yeast homologous recombination cloning leading to the novel peptides ambactin and xenolindicin. Chembiochem 15: 1290–1294 32. An Y, Ji J, Wu W et al (2005) A rapid and efficient method for multiple-site mutagenesis with a modified overlap extension PCR. Appl Microbiol Biotechnol 68:774–778

Stability Prediction for Site-Wise Combinatorial Libraries 33. Chao G, Lau WL, Hackel BJ et al (2006) Isolating and engineering human antibodies using yeast surface display. Nat Protoc 1:755–768 34. Benatuil L, Perez JM, Belk J et al (2010) An improved yeast transformation method for the generation of very large human antibody libraries. Protein Eng Des Sel 23:155–159 35. Bednar D, Beerens K, Sebestova E et al (2015) FireProt: energy- and evolution-based computational design of thermostable multiple-point mutants. PLoS Comput Biol 11:e1004556 36. Dehouck Y, Kwasigroch JM, Gilis D et al (2011) PoPMuSiC 2.1: a web server for the estimation of protein stability changes upon mutation and sequence optimality. BMC Bioinformatics 12:151 37. Witvliet DK, Strokach A, Giraldo-Forero AF et al (2016) ELASPIC web-server: proteomewide structure-based prediction of mutation effects on protein stability and binding affinity. Bioinformatics 32:1589–1591 38. Pires DEV, Ascher DB, Blundell TL (2014) DUET: a server for predicting effects

73

of mutations on protein stability using an integrated computational approach. Nucleic Acids Res 42:W314 39. Sumbalova L, Stourac J, Martinek T et al (2018) HotSpot Wizard 3.0: web server for automated design of mutations and smart libraries based on sequence input information. Nucleic Acids Res 46:W356–W362 40. Wijma HJ, Fu¨rst MJLJ, Janssen DB (2018) A computational library design protocol for rapid improvement of protein stability: FRESCO. In: Methods in molecular biology. Humana Press, pp 69–85 41. Alford RF, Leaver-Fay A, Jeliazkov JR et al (2017) The rosetta all-atom energy function for macromolecular modeling and design. J Chem Theory Comput 13:3031–3048 42. Jacobs TM, Yumerefendi H, Kuhlman B et al (2015) SwiftLib: rapid degenerate-codonlibrary optimization through dynamic programming. Nucleic Acids Res 43:e34

Chapter 4 Ancestral Sequence Reconstruction and Alternate Amino Acid States Guide Protein Library Design for Directed Evolution James VanAntwerp, Patrick Finneran, Benedikt Dolgikh, and Daniel Woldring Abstract Engineered proteins possess nearly limitless possibilities in medical and industrial applications but finding a precise amino acid sequence for these applications is challenging. A robust approach for discovering protein sequences with a desired functionality uses a library design method in which combinations of mutations are applied to a robust starting point. Determining useful mutations can be tortuous, yet rewarding; in this chapter, we present a novel library design method that uses information provided by ancestral sequence reconstruction (ASR) to create a library likely to have stable proteins with diverse function. ASR computational tools use a multi-sequence alignment of homologous proteins and an evolutionary model to estimate the protein sequences of the numerous common ancestors. For all ancestors, these tools calculate the probability of every amino acid occurring at each position within the sequence alignment. The alternate amino acid states at individual positions corelate to a region of stability in sequence space around the ancestral sequence which can inform site-wise diversification within a combinatorial library. The method presented in this chapter balances the quality of results, the computational resources needed, and ease of use. Key words Library design, Ancestral sequence reconstruction, Phylogenetic analysis, Directed evolution, Rational design

1

Introduction Proteins are incredibly useful chemical machines which are capable of doing nearly any chemical process we could want [1–3]. Importantly, this depends on our ability to discover the correct amino acid sequence to code for that functionality. With a vast, rugged sequence landscape to choose from, it is challenging to find the single sequence that provides the functionality that we desire. This problem is often addressed by a library design method—choosing a region of sequence space to explore through combinatorial

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_4, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

75

76

James VanAntwerp et al.

mutations on a template protein. There is a need for a methodology for selecting mutations that will yield a desired functionality and stability. Ancestral sequence reconstruction (ASR) is a method for inferring the amino acid sequences of ancient proteins based on a multisequence alignment of modern proteins and an evolutionary model. The inference of common ancestral sequences provides a glimpse at combinations of mutations that nature has previously determined to be functional and stable through natural selection. In this way, ASR provides a roadmap for navigating complex mutational landscapes to yield stable and functional protein variants [4]. Many of the ancestral sequences calculated by this method are often found to have higher stability [5–7], improved tolerance to deleterious mutations [8], and may also demonstrate increased promiscuity with respect to their substrate specificity [9, 10]. In the context of protein library design, ASR is a versatile tool for identifying a highly stable starting point to impose mutations onto [11], acting as a powerful platform for directed evolution. ASR can also be leveraged to identify a region of stable sequence space proximal from a calculated common ancestor to rationally guide site-specific mutations [12, 13]. By gaining a perspective of a protein family that not only includes the breadth of modern sequences but also its evolutionary history, users can create large, diverse libraries filled with sequences that natural selection has already judged as fit. Here we describe a collection of best practices for building phylogenetic trees, inferring the protein sequences of common ancestors, and using the posterior probabilities of individual residues to design combinatorial protein libraries. This workflow balances quality of results against computational intensity and ease of use (see Note 1).

2

Materials 1. Python scripts and datasets associated with the workflow presented here are located on GitHub: https://github.com/ WoldringLabMSU/ASR_LibraryDesign_DirectedEvolution. 2. NCBI BlastP, protein database search tool: https://blast.ncbi. nlm.nih.gov/Blast.cgi?PAGE¼Proteins. 3. CD-Hit, sequence clustering tool: https://github.com/ weizhongli/cdhit [14].This software can be accessed via webserver (http://weizhongli-lab.org/cdhit_suite/cgi-bin/index. cgi) or can be compiled locally (on Mac OS, it requires gcc to be first downloaded). See Note 2. User guide: https://github. com/weizhongli/cdhit/blob/master/doc/cdhit-user-guide. pdf.

ASR Guides Library Design for Directed Evolution

77

4. MUSCLE, multiple sequence alignment tool: https://www. drive5.com/muscle/downloads.html [15]. User guide: https://www.drive5.com/muscle/manual/index.html. 5. IQ-Tree, phylogeny and ancestral sequence reconstruction tool: https://www.iqtree.org [16, 17]. Webserver: http://iqtree.cibiv.univie.ac.at/. User guide: http://www. iqtree.org/doc/iqtree-doc.pdf. 6. AliView, multi-sequence alignment visualization editing tool: https://github.com/AliView/AliView.

and

7. FigTree, phylogenetic tree visualization: https://github.com/ rambaut/figtree/releases. 8. SwiftLib, a web server for degenerate codon design found at http://rosettadesign.med.unc.edu/SwiftLib/ [18].

3

Methods 1. Select a protein sequence of interest. This may represent a family for which a stable and functional ancestor is desired. 2. Submit the sequence to NCBI’s BlastP. See Note 3 for BlastP settings. Download the results as a multi-FASTA file. See Note 4 for how to determine the number of BlastP results needed for ASR. To avoid issues in later steps related to invalid characters and length restrictions, it is helpful to convert sequence names using barcodes or reduced identifiers, keeping reference to the original names. 3. Use the program CD-Hit to remove highly similar sequences from the unaligned FASTA file downloaded from NCBI (see Note 5 for details on the parameters of CD-Hit): cd-hit -i BlastP_file.fasta -o CDHit_BlastP_file.fasta -c 0.95 -n 5

4. Use MUSCLE to align the output file of CD-Hit (see Note 6 for more information on common MUSCLE errors): muscle -in CDHit_BlastP_file.fasta -out MUSCLE_CDHit_BlastP_file.fasta

5. Manually curate the sequences within the resulting alignment. Sequences containing multiple domains, or large insertions at either the N- and C-termini, should be trimmed away so that only the domain of interest remains. Sequences that contain large insertions or deletions within the domain of interest should be removed from the alignment entirely. Remove

78

James VanAntwerp et al.

entries containing an “X” anywhere within the amino acid sequence of the domain of interest. Rerun the MUSCLE alignment after sequence curation. 6. Using the output of the MUSCLE alignment, submit a preliminary evaluation to IQ-Tree to inspect the phylogeny: iqtree -s MUSCLE_CDHit_BlastP_file.fasta -alrt 10000 -bb 10000

Option -s provides the alignment that will be made into a phylogeny. Option -bb specifies the number of ultrafast bootstraps (UFBoot) to be performed for testing branch supports [19, 20]. Option -alrt uses the SH-aLRT test, which is a method of testing branch support [21]. 7. After IQ-Tree has finished running, examine the maximumlikelihood tree. Identify areas of the tree with poor confidence and add sequences to improve that confidence. See Note 7 for methods that can be useful here, and Fig. 1 for how to evaluate the trees. The output file, MUSCLE_CDHit_BlastP_file. fasta.treefile can be viewed using FigTree software. The branch support approximations from each test will be written at each node as SH-aLRT / UFBoot. 8. It may be necessary to repeat steps 3–7 as needed until the phylogenetic tree produced by IQ-Tree no longer shows significant improvement in the branch supports upstream of individual clades with each iteration (see Note 7). It is recommended to begin relying on individual clades when SH-aLRT >80% and ultrafast bootstrapping >95%.

A

B

C

D

Fig. 1 (a) Phylogenetic tree created from an alignment of modern sequences with diverse functions (step 7). (b) Confidence of tree topology at individual nodes is quantified using branch support values. A range of small and red to large and green nodes indicate the gradient of high to low support values, respectively (step 8). (c) The branches downstream of the low confidence nodes are shown as dashed lines (step 9). To improve the branch support values (smaller red and yellow circles), alternate topologies will be explored with additional sequences proximal to the low confidence node (step 9; Note 7). (d) With the inclusion of additional sequences (hash filled circles), new topologies with improved branch support values at internal nodes are formed, allowing inference of reconstructed ancestor sequences at each internal node (step 10)

ASR Guides Library Design for Directed Evolution

79

9. Once the final set of sequences, alignment, and phylogeny have been established, submit a run in IQ-Tree using the additional arguments -asr (ancestral sequences will be inferred by the empirical Bayesian method, written to a .state file), te (load and fix tree topology from previous run), s (alignment file), m (substitution model), and -redo (rerun the analysis): iqtree -s MUSCLE_CDHit_BlastP_file.fasta -te MUSCLE_CDHit_BlastP_file.fasta.treefile -asr -m {substitution model} -redo

10. The posterior probabilities of each amino acid occurring at every position throughout individual ancestral nodes are listed in the .state file (output generated using the -asr argument). These data show the likelihood of an amino acid (the consensus residue and alternate amino acid states) being present at a specific position within each ancestral sequence. Example filename: MUSCLE_CDHit_BlastP_file.fasta. state.

11. Additional consideration is needed for handling gaps within the inferred ancestral sequences (see Note 8). IQ-Tree treats gaps within an aligned sequence as unknown or missing information. For this reason, the .state file generated in step 9 contains posterior probabilities for the 20 amino acids, but fails to include the likelihood of a position being a gap. To accurately predict which positions are gaps within the ancestral sequences, conduct an IQ-Tree run using a modified alignment file (converting all amino acids to 1 and gaps to 0) while fixing the phylogenetic tree as in step 9. Using the Fasta2Binary.py script to convert the amino acid sequence alignment (e.g. MUSCLE_CDHit_BlastP_file.fasta) to a binary alignment with identical sequences names and alignment length. This script produces a new FASTA file (e.g. MUSCLE_CDHit_BlastP_BINARY.fasta). Script located here: https://github.com/WoldringLabMSU/ASR_ LibraryDesign_DirectedEvolution. Run IQ-Tree using the newly created binary alignment file. Assign and fix the tree (-te and -blfix), specify the binary compatible substitution model (-m GTR2 + FO), and rerun the Baysian ASR analysis (-asr -redo): iqtree -s MUSCLE_CDHit_BlastP_BINARY.fasta -te MUSCLE_CDHit_BlastP_file.fasta.treefile -blfix -asr -m GTR2+FO –redo

12. The binary format ASR run in step 11 generates a .state file listing the posterior probabilities of each position being either a 0 (gap) or 1 (amino acid). For each ancestral sequence node within the binary .state file, positions identified as being

80

James VanAntwerp et al.

probable gaps (i.e. p(0) > 0.5) will dictate which positions in the amino acid representation of the ancestral sequence . state file to replace with gaps. Merge the .state files from the binary (step 11) and the initial (step 9) run to accurately place gaps into the inferred ancestral sequences. Generate a FASTA file of the aligned, gapped sequences using the AA_Binary_State_Gap_FASTA.py script (https://github.com/ WoldringLabMSU/ASR_LibraryDesign_DirectedEvolution). 13. Assess the modified .state file to identify non-gap positions within specific ancestral nodes where the posterior probability of the consensus residue is below a cut-off value 0.95. In the context of library design, positions with a consensus residue having a likelihood greater than 0.95 will be conserved. The amino acids (alternate states) at the remaining positions (those having a consensus residue probability of less than 0.95) provide an ensemble of amino acids to choose from in the combinatorial library design. Priority is given to amino acids with the highest posterior probabilities (Fig. 2) and setting a cut-off value for the cumulative posterior probability of multiple, probable amino acids at a site (Fig. 3; Note 9). 14. For each position with an ensemble of alternate amino acids, design a degenerate codon that codes for all of those amino acids while avoiding stop codons or other detrimental mutations. Design of degenerate codons can be done via SwiftLib web server: http://rosettadesign.med.unc.edu/SwiftLib/ [18].

Posterior Probability, Cumulative

100% 80%

N S

G

S

Q

H

E

K

L

A

S

60%

Y Y

A

40%

M

N

R

1

2

3

4

5

V

Posterior Probabilies of Amino Acid Alternate States at Each Site

R

W

20% 0%

F

6

7 … Hn

Ancestral Sequence Posions

Fig. 2 Posterior probabilities of amino acids predicted at individual low confidence sites within an ancestral node of interest. Posterior probabilities from the .state files indicate the likelihood of an amino acid existing at particular positions of the inferred ancestral sequences. Often, over 80% of the positions will have a consensus residue with a posterior probability >0.95; however, at the remainder of the positions (shown above), a collection of multiple residues are predicted to be likely. These alternate amino acid states have been found to reflect a region of stable sequence space proximal to a calculated common ancestor [12]. A library is designed using diverse combinations among these alternate amino acids

ASR Guides Library Design for Directed Evolution

81

12 cell surface display

Library Size, log10

10 8

mammalian expression

6 4 plate reader

2 0 0%

20%

40%

60%

80%

100%

Cumulative Posterior Probability Cut-off Fig. 3 A representation of the relationship between choosing a cumulative posterior probability cut-off value and its impact on the theoretical size of the resulting combinatorial library. Theoretical library size is calculated by taking the product of the number alternate sites at every diversified position (e.g., if 10 positions each had 4 probable alternate amino acid states, theoretical combinatorial library size: 410 ¼ 1e6)

15. To implement and validate the newly designed libraries, oligonucleotides containing degenerate codons can be assembled using overlap extension PCR (Fig. 4), then combined with linearized yeast surface display vector (pCT-CON2), and electroporated into yeast (EBY-100) for expression, screening, and characterization [22–29].

4

Notes 1. There are a growing number of available phylogeny and ancestral sequence reconstruction programs, some even offering automated web servers such as the recently developed FireProtASR [30]. Another notable example of useful ASR tools is that of BAli-Phy [31], found at bali-phy.org. BAli-Phy is a program for simultaneous Bayesian inference of alignment and phylogeny. See the user guide here: http://www.bali-phy. org/docs.php#usersguide. BAli-Phy uses a Markov-Chain Monte Carlo algorithm, so two parallel analyses can have the same effect on final confidence as one analysis run for twice as long. For this reason, it is recommended to run several BAli-Phy analyses on the same set of sequence data at once. The Bayesian methods used by BAli-Phy have been shown to be more precise than maximum-likelihood methods [32]. In

82

James VanAntwerp et al.

Fig. 4 Construction of combinatorial library. (a) Overlapping DNA primers are designed to include degenerate codons (colored) at each diversified library position. (b) Regions between diversified sites (gray) are amplified via overlapping primers or using a full ancestral gene as the template. (c) PCR products are pooled and amplified to generate final combinatorial gene library

addition, BAli-Phy can simultaneously modify sequence alignment, phylogenetic tree, and ancestral sequences to find the most likely combination of all three. BAli-Phy is simple to run, yet is computationally demanding and generates outputs that are laborious to parse. 2. The executables for the command line software described here must be added to the user’s path following their installation. All commands in the methods section assume this setup of the path has been done. 3. A Blast search uses sequence similarity as an indicator of common evolutionary history. Similarity is measured in NCBI’s database using an “E-value,” a number from near-zero to ten which represents the number of random matches of the same quality that would be expected. E-values above 1 should be excluded, as their similarity could be random and not due to evolutionary history. To expand the collection of modern sequences, re-submit other known members of the family under study to a BlastP search. Using annotated or non-redundant databases are preferred. These will avoid retrieving sequences that share no evolutionary history, such as cloning vectors, engineered proteins, partial domains, and

ASR Guides Library Design for Directed Evolution

83

other erroneous matches. Alternatively, Position-Specific Iterative BLAST (PSI-BLAST) performs a sequence search for more distantly related sequences. PSI-BLAST begins exactly as BlastP for the first iteration, using an input sequence to locate similar sequences within the database of choice at a specified e-value threshold. Based on these findings, a calculated position-specific scoring matrix (PSSM) represents a profile of the “multiple alignment of the highest scoring pairs,” indicating regions of conservation between the sequences that can be used in further iterations to uncover distantly related sequences. This is iterated as necessary or until new sequences above a desired threshold no longer appear. Refer to https:// www.ncbi.nlm.nih.gov/books/NBK2590/ for more information. 4. The number of BlastP results can vary wildly depending on the protein, yet the total number of sequences in the final alignment heavily impacts the quality and computational resources needed for ASR. When using IQ-Tree, it is best to include no more than 100,000 total characters, as measured by the product of the number of sequences and sequence length. Ideally, an even distribution of sequences should be drawn from across the scope of the family under study. Achieving this can be made much easier with CD-Hit, a uniform method of determining similarity and removing sequences that are too closely similar. 5. The number of sequences removed from a large dataset can be controlled by the cut-off of CD-Hit. When using CD-HIT option -c specifies the similarity cut-off threshold for a cluster, between 0 and 1. A cut-off of 1 will only remove duplicates, while a cut-off of 0.7 is typically too aggressive for ASR. Option -n specifies the word size for comparisons, and it must be set depending on the cut-off threshold and if the sequences are amino acid or DNA sequences. See the user guide for a full description of this parameter, but for comparisons of protein sequence with cut-off between 0.7 and 1, use a word size of 5. 6. MUSCLE software has limitations in character type within input files. A common error is. *** ERROR *** Cannot open ‘input.fasta’ errno=2

which indicates a character in the name of the input file is not allowed, typically a space. MUSCLE can run in just a few seconds if the input has been properly cleaned (e.g., removal of excessively long sequences). 7. High confidence in the tree topology is critical to ancestral sequence reconstruction and low confidence areas of the tree should be rectified when possible. This can be accomplished

84

James VanAntwerp et al.

through the introduction of additional homologous sequences, removal of irrelevant sequences, as well as the inclusion of outgroups (modern proteins which are closely related to the node of interest, yet reside in a separate clade). The primary means of evaluating the quality of a phylogenetic tree is to examine the confidence placed in each internal node, measured by branch support, commonly found through bootstrapping, ultrafast bootstrapping, or SH-aLRT. Bootstrapping involves subtle manipulations of the underlying data and evaluating the effect on tree topology [33]. The resulting bootstrap values for a node are most often reported as the percentage of bootstrapped trees which contain the given node. To find more sequences that fit under a given node, submit a sequence descended from that node to BlastP. The highest-similarity sequences returned likely belong to the same clade and may not have been returned by the original BlastP search or may have been removed by CD-Hit. In this process, there are two potential pitfalls. First, constantly attempting to improve sequence coverage at the far ends of the phylogeny causes an ever-expanding tree with ever decreasing relevance to the sequence of interest. The first solution is to limit the number of sequences and identifying “good enough” coverage that may have low confidence in low-level branches not of interest, but which retains high confidence in older nodes. Second, all available sequence data is found, and there is still not enough for ancestral sequence reconstruction. This problem is not easily solved (e.g. sequencing novel wild type proteins or finding an alternate family to study). This decision is somewhat subjective and requires experience to make well. 8. A challenging aspect of ancestral sequence reconstruction is determining the relevance and positioning of indels (e.g. gaps) within a multiple sequence alignment. At a position where some descendants contain a gap and some do not, should the ancestral sequence contain a gap, no gap, or a certain likelihood of a gap? These are questions that IQ-Tree is not equipped to handle because of the problems it causes for sequence alignment and maximum-likelihood applications of a mutational matrix. Instead, IQ-Tree simply ignores gaps in an alignment, and will make predictions for ancestral positions based on the amino acids that are present – effectively treating a gap as though it did not contain information. This is not an optimal approach for positions for which the majority of descendants contain a gap, and this is why the user is advised to aggressively remove those positions from the input alignment in step 5 of this workflow.

ASR Guides Library Design for Directed Evolution

85

9. In order to design a library from ancestral sequence uncertainty, the investigator uses a cut-off value, typically between 70% and 95%. For example, if a site has a 60% chance of valine, a 25% chance of isoleucine but a 15% chance of leucine, an 80% cut-off library would include only valine and isoleucine at that position. A 90% cut-off library made from the same ancestral uncertainty would include all three amino acids at that position. By changing the cut-off value, library size can be adjusted to accommodate for high- or low-throughput screening. Libraries made with this method are likely to produce a very high proportion of functional proteins, as the ancestral sequences are highly robust to uncertainty [11–13, 34] with this method. References 1. Siegel JB, Zanghellini A, Lovick HM et al (2010) Computational design of an enzyme catalyst for a stereoselective bimolecular dielsalder reaction. Science 329:309–313 2. Hyster TK, Kno¨rr L, Ward TR et al (2012) Biotinylated Rh(III) complexes in engineered streptavidin for accelerated asymmetric C-H activation. Science 338:500–503 3. Coelho PS, Brustad EM, Kannan A et al (2013) Olefin cyclopropanation via carbene transfer catalyzed by engineered cytochrome P450 enzymes. Science 339:307–310 4. Chen F, Gaucher EA, Leal NA et al. (2010) Reconstructed evolutionary adaptive paths give polymerases accepting reversible terminators for sequencing and SNP detection. Proceedings of the National Academy of Sciences of the United States of America, 107(5), 1948–1953. https://doi.org/10.1073/pnas. 0908463107 5. Gaucher EA, Govindarajan S, Ganesh OK (2008) Palaeotemperature trend for Precambrian life inferred from resurrected proteins. Nature 451:704–707 6. Risso VA, Gavira JA, Sanchez-Ruiz JM (2014) Thermostable and promiscuous Precambrian proteins. Environ Microbiol 16:1485–1489 7. Nguyen V, Wilson C, Hoemberger M et al (2017) Evolutionary drivers of thermoadaptation in enzyme catalysis. Science 355:289–294 8. Bloom JD, Labthavikul ST, Otey CR et al (2006) Protein stability promotes evolvability. Proc Natl Acad Sci U S A 103:5869–5874 9. Risso VA, Gavira JA, Mejia-Carmona DF et al (2013) Hyperstability and substrate promiscuity in laboratory resurrections of precambrian β-lactamases. J Am Chem Soc 135:2899–2902

10. Thornton JW, Need E, Crews D (2003) Resurrecting the ancestral steroid receptor: ancient origin of estrogen signaling. Science 301: 1714–1717 11. Bar-Rogovsky H, Stern A, Penn O et al (2015) Assessing the prediction fidelity of ancestral reconstruction by a library approach. Protein Eng Des Sel 28:507–518 12. Eick GN, Bridgham JT, Anderson DP et al (2017) Robustness of reconstructed ancestral protein functions to statistical uncertainty. Mol Biol Evol 34:247–261 13. Wheeler LC, Harms MJ (2021) Were ancestral proteins less specific? Mol Biol Evol 38(6): 1–35 14. Li W, Godzik A (2006) Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences. Bioinformatics 22:1658–1659 15. Edgar RC (2004) MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res 32:1792– 1797 16. Nguyen L-T, Schmidt HA, von Haeseler A et al (2015) IQ-TREE: a fast and effective stochastic algorithm for estimating maximumlikelihood phylogenies. Mol Biol Evol 32: 268–274 17. Minh BQ, Schmidt HA, Chernomor O et al (2020) IQ-TREE 2: new models and efficient methods for phylogenetic inference in the genomic era. Mol Biol Evol 37:1530–1534 18. Jacobs TM, Yumerefendi H, Kuhlman B et al (2015) SwiftLib: rapid degenerate-codonlibrary optimization through dynamic programming. Nucleic Acids Res 43:1–9

86

James VanAntwerp et al.

19. Minh BQ, Nguyen MAT, von Haeseler A (2013) Ultrafast approximation for phylogenetic bootstrap. Mol Biol Evol 30:1188–1195 20. Hoang DT, Chernomor O, von Haeseler A et al (2018) UFBoot2: improving the ultrafast bootstrap approximation. Mol Biol Evol 35: 518–522 21. Guindon S, Dufayard JF, Lefort V et al (2010) New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0. Syst Biol 59: 307–321 22. Woldring DR, Holec PV, Zhou H et al (2015) High-throughput ligand discovery reveals a sitewise gradient of diversity in broadly evolved hydrophilic fibronectin domains. PLoS One 10:e0138956 23. Woldring DR, Holec PV, Stern LA et al (2017) A gradient of sitewise diversity promotes evolutionary fitness for binder discovery in a threehelix bundle protein scaffold. Biochemistry 56: 1656–1671 24. Kruziki MA, Bhatnagar S, Woldring DR et al (2015) A 45-amino-acid scaffold mined from the PDB for high-affinity ligand engineering. Chem Biol 22:946–956 25. Kruziki MA, Sarma V, Hackel BJ (2018) Constrained combinatorial libraries of Gp2 proteins enhance discovery of PD-L1 binders. ACS Comb Sci 20:423–435 26. Bryksin AV, Matsumura I (2010) Overlap extension PCR cloning: a simple and reliable

way to create recombinant plasmids. BioTechniques 48:463–465 27. Schimming O, Fleischhacker F, Nollmann FI et al (2014) Yeast homologous recombination cloning leading to the novel peptides ambactin and xenolindicin. Chembiochem 15:1290– 1294 28. An Y, Ji J, Wu W et al (2005) A rapid and efficient method for multiple-site mutagenesis with a modified overlap extension PCR. Appl Microbiol Biotechnol 68:774–778 29. Benatuil L, Perez JM, Belk J et al (2010) An improved yeast transformation method for the generation of very large human antibody libraries. Protein Eng Des Sel 23:155–159 30. Khan RT, Musil M, Stourac J et al (2021) Fully automated ancestral sequence reconstruction using FireProtASR. Curr Protoc 1:1–13 31. Suchard MA, Redelings BD (2006) BAli-Phy: simultaneous Bayesian inference of alignment and phylogeny. Bioinformatics 22:2047–2048 32. Williams PD, Pollock DD, Blackburne BP et al (2006) Assessing the accuracy of ancestral protein reconstruction methods. PLoS Comput Biol 2:0598–0605 33. Felsenstein J (1985) Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39:783–791 34. Thomas A, Cutlan R, Finnigan W et al (2019) Highly thermostable carboxylic acid reductases generated by ancestral sequence reconstruction. Commun Biol 2:429

Chapter 5 Machine Learning-driven Protein Library Design: A Path Toward Smarter Libraries Mehrsa Mardikoraem and Daniel Woldring Abstract Proteins are small yet valuable biomolecules that play a versatile role in therapeutics and diagnostics. The intricate sequence–structure–function paradigm in the realm of proteins opens the possibility for directly mapping amino acid sequence to function. However, the rugged nature of the protein fitness landscape and an astronomical number of possible mutations even for small proteins make navigating this system a daunting task. Moreover, the scarcity of functional proteins and the ease with which deleterious mutations are introduced, due to complex epistatic relationships, compound the existing challenges. This highlights the need for auxiliary tools in current techniques such as rational design and directed evolution. To that end, the state-of-the-art machine learning can offer time and cost efficiency in finding high fitness proteins, circumventing unnecessary wet-lab experiments. In the context of improving library design, machine learning provides valuable insights via its unique features such as high adaptation to complex systems, multi-tasking, and parallelism, and the ability to capture hidden trends in input data. Finally, both the advancements in computational resources and the rapidly increasing number of sequences in protein databases will allow more promising and detailed insights delivered from machine learning to protein library design. In this chapter, fundamental concepts and a method for machine learning-driven library design leveraging deep sequencing datasets will be discussed. We elaborate on (1) basic knowledge about machine learning algorithms, (2) the benefit of machine learning in library design, and (3) methodology for implementing machine learning in library design. Key words Library design, Directed evolution, Machine learning, Deep learning

1

Introduction Proteins are molecules with a wide variety of applications in biological processes. They have fundamental functions in living organisms such as being catalysts, receptors, structural elements, transporters, and regulators [1–5]. Accordingly, increased potential for mapping the protein sequence to its function will result in comprehension and regulation of biological processes associated with functional disorders. As a result, techniques to efficiently navigate the protein fitness landscape are at the forefront of protein

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_5, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

87

88

Mehrsa Mardikoraem and Daniel Woldring

engineering. Nevertheless, the complexity and ruggedness of the protein fitness landscape, and the high potential of failure in searching through the fitness landscape (mutations resulting in unstable and non-functional variants), make this task more demanding. A growing number of computational and experimental approaches (e.g. high-resolution stability calculations [Chapter 3 of this volume], virtual screening [6], deep sequencing [7], cytometry-based selections [8], ancestral sequence reconstruction [Chapter 4 of this volume]) seek to address these gaps in knowledge. Machine learning algorithms offer a platform for harnessing large, diverse datasets for the purpose of understanding natural protein features and guiding protein engineering efforts. This provides the opportunity to map protein sequence to function without requiring explicit biophysical knowledge of individual sequences. An additional advantage of such algorithms relates to expanding the utility of experimentally derived sequences beyond the often small subsets of lead variants, i.e. while directed evolution discards low fitness protein variants, machine learning can learn the characteristics of these sub-optimal variants and, in turn, increase the model’s predictive performance [9]. Machine learning can be particularly advantageous for protein engineering campaigns that involve low-throughput or laborious selections. Therefore, its usefulness depends on factors such as library size, screening difficulty, fitness landscape ruggedness, and the accuracy of the predictive model. As a result, the ability to capture complicated trends among protein datasets, aided by non-linear functionality to reveal important features, makes machine learning a powerful tool for guiding protein library design. Machine learning has played an important role in protein engineering for more than 30 years, yielding improved prediction of tertiary structure [10] and protein–protein interactions [11] using amino acid sequence information alone [10, 11]. Machine learning algorithms such as support vector machines (SVMs), random forest (RF), and gradient boosting machines (GBMs) have provided platforms for protein function and property prediction including stability, catalytic activity, and secondary structure [12–15]. Machine learning models can also be used in predicting the developability and evolvability of protein sequences [16]. Deep learning (DL) [17], as a sub-field of machine learning, imitates human brain functionality in decision making and learning experiences. Utilizing non-linear functions, the algorithm can learn and extract desired features from the provided input data, well suited for dealing with rich datasets with high dimensionality. This makes deep learning methods particularly promising for evaluating sequence– function trends among a rapidly growing number of protein sequences. Although practical and promising, developing a finetuned strategy to employ machine learning in the field demands an awareness of the existing challenges and capabilities. Here, we present a procedure for establishing deep learning models that

Machine Learning-Driven Protein Library Design

89

guide protein library design. Common challenges and best practices in this burgeoning field will be highlighted. We consider multiple practical applications of machine learning within the context of protein library design: (1) combinatorial library design based on deep sequencing data following highthroughput directed evolution, (2) process parallelization and parameterization of features within far-reaching parts of the fitness landscape, and (3) the ability to sample the diversity applied by specific degenerate codon techniques and oligonucleotide combinations prior to experimental implementation. 1.1 Providing a Better Starting Point for Directed Evolution

Directed evolution campaigns are initialized using a parent sequence to implement mutations onto. Therefore, the pathdependency of directed evolution benefits from a high-fitness or highly stable sequence as a starting point to increase the probability of finding optimal regions within a fitness landscape [18]. Machine learning can guide directed evolution by identifying a large collection of promising sequences based on a curated input data. In a recent example, a machine learning model was trained based on the initial library of fluorescent proteins to build a second small yet enriched protein library [19].

1.2 Investigating Unexplored Parts of the Fitness Landscape

Machine learning algorithms provide some unique advantages that enable finding unexplored variants which may possess high fitness. Large datasets which are processed with high-complexity algorithms are not only able to predict functionality but can learn hidden characteristics and rules that exist in provided data (see Note 1). As an example, UniRep [20, 21] has been trained on 24 million protein sequences from Uniref50 [22] to obtain general trends in protein sequences and statistical representations of amino acids. The authors claim strong generalizations and high performance in downstream tasks taking advantage of their embedding methods. Another factor which is effective in finding the unexplored high fitness sequences is the application of parallelism and multi-tasking within machine learning. Multi-task learning is a subset of machine learning algorithms that trains multiple tasks simultaneously in one unique model [23]. This feature enables capturing the epistatic relationships in protein sequences by providing a path-independent search in the fitness landscape [24]. In this way, applying machine learning to protein engineering enables an increased likelihood of accessing undetected regions of the fitness landscape.

1.3 Estimating Degenerate Codon Performance via Fitness Distribution Analysis

Implementing a well-trained machine learning algorithm enables the evaluation of multiple design strategies and reduces experimental effort. Various experimental techniques have been developed to improve the efficiency of directed evolution such as gene shuffling [25] and neutral drift library screening [26] to manage the library size and increase the likelihood of finding the desired property.

90

Mehrsa Mardikoraem and Daniel Woldring

Another highly used method is to generate libraires based on degenerate codons in order to introduce tailored diversity at individual positions. As an example, while NNK is used to generate a library coding for all amino acids (with 3% stop codons), other combinations such as NYC (coding for hydrophobic residues), KST (coding for small residues), and NDT (coding for a balanced set of all amino acid properties) are available. In addition, several impressive computational attempts have been used to even go beyond these techniques and optimize the oligonucleotide combinations [27, 28]. Among these potential degenerate codon techniques for the library design, the user can pre-analyze their performance for their desired protein. Figure 1 proposes a comparison of candidate degenerate codons. As a base platform, an initial deep learning model is trained on the objective protein. Subsequently, the algorithm can generate sequences that incorporate candidate degenerate codons and assign a predicted fitness score to each based on the previously trained model. Finally, the distribution of fitness scores is calculated for all variants of each degenerate codon strategy. Statistical tools such as Jensen–Shannon Divergence [29] and survival function [30] provide a direct comparison between individual distributions in terms of diversity and fitness (see Note 2). The criterion for choosing the best performance depends on the particular application the platform is used for.

2

Materials There are a plethora of software libraries and packages provided for implementing model optimization, machine learning, and deep learning algorithms. Choosing one depends on the specific application that the user has in mind. TensorFlow [31], Keras [32], PyTorch [33], and Scikit-learn [34] are among the most popular libraries with each having some trade-offs (see Note 3 for analysis). The discussion found here focuses on deep learning in Keras. Please refer to our GitHub (https://github.com/WoldringLabMSU/ DeepLearning.git) for a collection of relevant Python scripts to guide model implementation as well. 1. The latest version of python (https://www.python.org) installed (Installing Anaconda (https://www.anaconda.com) is highly suggested for beginners as it is straightforward and easy to use). 2. Spyder or Jupyter Notebook environments in Anaconda are both popular and practical in doing machine learning and deep learning projects. 3. pandas [35] (For dealing with data frames). 4. numpy [36] (For efficient mathematical operations).

Machine Learning-Driven Protein Library Design

91

Fig. 1 Evaluation of degenerate codon performance within multiple design strategies using the trained model based on experimental data on the protein of interest. (a) Elaboration of different hypothetical degenerate codon strategies in three different sites of the protein (the previously trained model needs to be trained on high-quality experimental data in order to predict the fitness of generated sequences in each technique.) The first utilizes the NNK technique in all sites, while the second uses different degenerate codons at each site. The third strategy uses custom mix codons (versatile ratios of base pairs). Sequences compatible with the rubrics of each strategy can be generated and a fitness score for the generated sequences will be assigned to each sequence based on the previously trained model in the protein of interest. (b) The score function (transformed and standardized predicted scores) distribution will be obtained for each candidate library. One method to comprehend the predicted data is by clustering the generated data. The exemplary plot shown in here is based on both the sequences and scores of the three strategies when each dot represents one sequence produced by one of the strategies. The radius of the dots will represent their individual fitness scores, and the distance between any two dots will represent how distant those amino acid sequences are from each other in sequence space. (c) The Jenson–Shannon Divergence is useful for quantifying differences that exist between individual distributions and a reference distribution or the extent of similarity between candidates’ distributions. (d) The survival function provides the probability of improved score functions compared to the current score function, providing more insight for analysis of distributions. (e) Box plots can be produced based on the generated fitness scores and are informative for recognizing the quantiles and making comparison between them in the three different strategies

5. scikit_learn [29] (This is mostly used in machine learning but its preprocessing section has a myriad of useful functions for analysis and fine tuning the data). 6. tensorflow [37] (For using keras, its backend should be installed (CNTK and Theano are other options, as well). 7. keras [32]. X)

In order to use these libraries, use pip command (pip install in the command prompt.

92

3

Mehrsa Mardikoraem and Daniel Woldring

Methods The following is a general workflow representing the required actions in building a deep learning algorithm from deep sequencing data (Fig. 2). Here, the goal is to produce a supervised learning algorithm for predicting protein function based on amino acid sequence.

3.1 Data Processing as an Initial Yet Pivotal Step in Any DL Algorithm

The main purpose of this step is to prepare the data to be fed into the deep learning algorithm. Importantly, what can be learned by the model strongly depends on what is provided to the algorithm as an input. If the aim is to map the sequence to function, protein sequence should be as an input labeled with the desired functionality (output). Three important steps in data processing include input data refinement, input data representation, and output data representation (if dealing with supervised learning). See Note 3.

3.1.1 Input Data Refinement

Scaling and refining the data aims to remove the “importance” of the raw magnitude of one factor relative to another and as a result reduces estimation errors and calculation times. One package which

Fig. 2 This figure provides the overall workflow of what the user may encounter when using deep sequencing data and wants to apply some state-of-the-art machine learning algorithm. The main steps for using this path are data processing, choosing an appropriate learning algorithm and deciding based on the algorithms’ findings. The detailed explanation of steps is mentioned in the content

Machine Learning-Driven Protein Library Design

93

can be used for standardization is StandardScaler or RobustScaler (advantageous when dealing with outliers) from sklearn preprocessing package: from sklearn.preprocessing import StandardScaler scaler1 = StandardScaler() scaler1.fit(Feature) Feture_standardized1 = scaler1.transform(Feature) ### from sklearn.preprocessing import RobustScaler scaler2 = RobustScaler() scaler2.fit(Feature) Feture_standardized2 = scaler2.transform(Feature)

3.1.2 Input Data Representation

One-hot encoding (see Note 4), integer encoding (see Note 5), physiochemical property-based encoding (see Note 6), and sequence embedding (e.g. UniRep [20] (https://github.com/ churchlab/UniRep), and TAPE [38] (https://github.com/ songlab-cal/tape)) (see Note 7) are notable options for representing amino acid sequence data. The user can manually one-hot encode the input sequences via defining dictionaries or using packages in py thon (Refer to ht tps://github.com/ WoldringLabMSU/DeepLearning.git for more information on encoding). from sklearn.preprocessing import LabelEncoder from sklearn.preprocessing import OneHotEncoder Amino_Acids = ["A","C", "D", "E","F", "G", "H", "I", "K", "L", "M", "N", "P","Q","R", "S", "T", "V","W", "Y"] label_encoder = LabelEncoder() onehot_encoder = OneHotEncoder(sparse = False) integer_encode = label_encoder.fit_transform (Amino_Acids) integer_encoded = integer_encoded.reshape(len (integer_encode), 1) Amino_Acids_onehot = onehot_encoder.fit_transform (integer_encoded)

3.1.3 Output Data Representation

Following high-throughput selection and deep sequencing of an initial combinatorial library, amino acid sequences can be labeled based on the observed enrichment ratios as a metric for relative fitness [39]. Depending on the experimental conditions for selection and depth of sequencing, the distribution of enrichment ratio data may take on various forms and require further refinement (see Note 9).

94

Mehrsa Mardikoraem and Daniel Woldring

3.2 Deep Learning Algorithm Selection Requires an Understanding of Each Algorithm Structure 3.2.1 Overview

Artificial neural network (ANN) architecture is inspired by human brain function which can learn from various input data. The building blocks for this network (neurons) receive information from adjacent neurons, then process this information with the aid of an activation function (see Note 9) before being sent to other downstream neurons. The effect and significance of each connection is proportional to its assigned weight. There are many deep learning methods utilized in the field from feed-forward neural network (FNN) to the convolutional neural network (CNN) [40] and recurrent neural network (RNN) [41]. FNNs are primary neural network algorithms which are rigorous and powerful in capturing high dimensional features. It consists of three distinct layers (each composed of neurons): input layer, hidden layer, and output layer. The input layer receives the data features, the hidden layer transforms the input layer to the output by updating the weights iteratively to minimize the loss function. For each of these neural network algorithms, backpropagation is used to update model weights to minimize the loss function via gradient decent. Finally, the output layer captures the results. Figure 3 illustrates one example structure of a feed-forward neural network. Convolutional neural networks are an additional deep learning algorithm inspired by the visual cortex in animals. CNNs capture hidden relationships and spatial dependencies in provided input via various filters, pooling, and convolutional kernels. Therefore, local properties will be obtained by using sliding filters and non-linear functions [42]. In recurrent neural network, there is a cyclic connection in the algorithm structure, whereby the current state of the algorithm will be updated based upon the past state and the current input data [43]. This characteristic makes RNN useful in time series data and capturing temporal relationships in sequences. In addition, variations of RNN such as LSTM [44] and GRU [45] have been developed to resolve the potential problems in its architecture such as capturing long distance relationships and resolving vanishing gradient problems. The best performance in the mentioned algorithms depends on the type of data and purpose for using deep learning. However, more advanced architectures like RNN/CNN are able to learn properties more efficiently than a traditional FNN. Transformers are another promising deep learning model utilizing attention-based mechanisms [46]. They enable learning the context and extract meaning from unlabeled data. This characteristic makes them very attractive to be used in protein engineering platform as the number of unlabeled sequences in protein databases outweighs the labeled data [20, 38, 47]. Descriptions of commonly used terms and techniques used in deep learning and optimization are shown in Table 1.

Machine Learning-Driven Protein Library Design

95

Fig. 3 This figure represents one simple example for a feed forward neural network. It consists of three main layers: the input layer, hidden layer(s), and output layer. It is a symbolic representation of the path from sequence to function. The magnifying glass focuses on one neuron (a21), illustrating how information passes through each neuron. Moreover, the information from six neurons in the previous hidden layer are multiplied by their weights (different colors are representative of activation level in each corresponding neuron). Afterwards, the result will be summed with bias and passes through the activation function to determine the output of that neuron 3.2.2 Guidance for Building a Deep Learning Structure

Here we build a simple feed-forward neural network in Keras to show one possible and simple format for building a neural network structure (Refer to https://github.com/WoldringLabMSU/ DeepLearning.git for more advanced structures). 1. Importing the required libraries from Keras. from tensorflow import keras import keras from keras.models import Sequential from keras.layers.core import Dense from keras.optimizers import Adam from keras.activations import relu, from keras.activations import sigmoid from sklearn.model_selection import train_test_split

2. Splitting the data set into train and test, X_train, X_test, Y_train, Y_test = train_test_split(features, labels, test_size=0.2, random_state=42)

96

Mehrsa Mardikoraem and Daniel Woldring

Table 1 Deep learning and optimization terminology and usage Concepts and terms

Definition

Train set

Part of data used for fitting the parameters

Test set

Part of data used for evaluation of the trained model

Evaluation set

Part of data (it can also be considered as segment of training data) used for tunning the hyperparameters

Classification

Prediction when output values are discrete classes

Regression

Prediction when output values are continues (mostly called as quantities than discrete labels)

Overfitting

Lack of model ability to generalize from trained data to unseen data

Loss function

A function for evaluating the algorithm calculating the difference between predicted and actual value

Application and suggestion

Avoid overfitting

Common for classification: accuracy, cross-entropy Common for regression: mse,mape

Backpropagation Algorithm used for updating model weights based upon the gradient of the loss function Epoch

Each epoch represents the entire training dataset has been passed to the system entirely

Batch size

Number of samples in each break of training set (called as Needs optimization batch) where backpropagation will be carried out

Optimizer

Strategy for minimizing the loss function

Needs optimization ADAM and SGD (two potential candidates)

Learning rate

The magnitude of steps in each iteration during backpropagation (implemented via gradient*learning rate)

Needs optimization (Adaptive learning rate is suggested)

Shuffling

Randomly change the order of existed data

Dropout

Randomly dropping some connections between neurons Avoid overfitting

Dense layer

Linear operation which maps every input to every output by weights

Convolutional layer

Consists of filters convolving through the provided matrix

Pooling layer

Used after convolutional layer for reducing the spatial size

Max pooling Average pooling

Data Increasing the number of samples by adding augmentation implementing modifications to the original samples

Avoid overfitting

Regularization

Needs optimization

Modification in the original cost function to reduce bias Avoid overfitting and increase penalty pertinent to magnitude of the weights

Machine Learning-Driven Protein Library Design

97

3. Defining the model. My_Model = Sequential()

4. Defining the input layer, number of neurons in that layer, its shape and its activation function. My_Model.add(Dense(20,

activation=’relu’,

input_shape=

(100,)))

5. Defining the hidden layers (the # of hidden layers is another hyperparameter), number of neurons (hyperparameter) in each hidden layer and corresponding activation function. My_Model.add(Dense(10, activation=’relu’)) My_Model.add(Dense(5, activation=’relu’))

6. Defining the output layer [the last activation function will depend on the task (see Note 8)]: My_Model.add(Dense(1, activation=’linear’))

7. Compiling the model and determining the optimizer, learning rate, loss function, and evaluation metrics (numbers should be optimized based on the problem): My_Model.compile(Adam(lr=0.01, decay=0.003),loss=’mean_squared_error’,metrics=[’mse’])

8. Fitting the model is accomplished by defining the batch size, epochs, and verbose. Validation split specifies the fraction of data to be used for validation: My_Model.fit(X_train,Y_train,batch_size=100,epochs=40,verbose=2, validation_split=0.2,

shuffle= True)

9. Evaluating the fitting performance: Metric = My_Model.evaluate(X_test, Y_test, verbose =2)

10. Predicting the labels for the test data set: y_test_predicted= My_Model.predict(X_test, verbose = 2)

98

Mehrsa Mardikoraem and Daniel Woldring

3.3 Visualization Guidance

Evaluation metrics provide useful insights to algorithm performance. Visualizations of these results help to further interpret and communicate the findings. For example, the seaborn library allows for showing the correlation between predicted output values versus actual values in regression: import seaborn as sns sns.jointplot(Prediciton,Actual_value, kind=’scatter’)

For classification, one simple method is using a confusion matrix to show model performance on each class prediction consisting of: true positives, true negatives, false positives, and false negatives values. from sklearn.metrics import confusion_matrix import matplotlib.pyplot as plt import seaborn as sns cm = confusion_matrix(Y_test, Prediction) tn, fp, fn, tp= confusion_matrix(Y_test,Prediction).ravel () ax= plt.subplot() sns.heatmap(cm, annot=True, fmt=’g’, ax=ax) ax.set_xlabel(’Predicted labels’) ax.set_ylabel(’True labels’) ax.set_title(’Confusion Matrix’)

Understanding and building such a structure is an important first step to take for utilizing deep learning in different projects. However, even with a powerful algorithm and appropriate data processing step, the prediction might be drastically off which leads us to decision making and further fine-tuning steps. 3.4 Decision Making and Evaluating Parameters

After processing the data and choosing the appropriate algorithm, one should be able to accurately evaluate the performance of that algorithm. Evaluation metrics depend on whether the problem is regression or classification. In classification, metrics such as accuracy, confusion matrix, AUC, and Recall are useful. While in regression, metrics such as RMSE, MSE, MAPE are better suited (see Note 10). Obtaining poor metrics for a model may arise from various stages in the algorithm such as inadequate input data, deficient preprocessing step, overfitting, and untuned hyperparameters. Choosing the right hyperparameters and preventing the system from overfitting are two necessary tasks in training any deep learning algorithm. Two generally used methods for resolving some of these issues are elaborated in the following.

Machine Learning-Driven Protein Library Design

99

3.4.1 Hyperparameter Optimization [48]

Hyperparameters are set of variables that dictate various characteristics of the algorithm’s structure and influence the process used for training models. It can be considered as a meta-optimization technique whereby parameter value fitness is monitored via the loss function during the training process [49]. Multiple methods can be employed for searching through hyperparameter space (e.g., manual search, grid search, random search, and Bayesian optimization). The manual search involves tuning the hyperparameters by a user based on guess and check. In grid search, for each parameter of interest, the user defines a list of values to implement. The algorithm then calculates the loss for all possible combinations of parameter values. In random search, some but not all combinations of parameters will be selected randomly, and the best loss will be chosen based on the selected parameters. Bayesian optimization offers high efficiency by using information from the past as an experience to choose the next set of hyperparameters. HYPEROPT (http://hyperopt.github.io/hyperopt/) and OPTUNA (https:// optuna.org/) are among the python libraries designed for hyperparameter optimization based on the objective function. (Refer to https://github.com/WoldringLabMSU/ DeepLearning.git for more information).

3.4.2 K-Fold CrossValidation

This resampling approach allows for more accurate prediction in model performance using even a limited data set. It splits the data into k complementary groups and uses k 1 groups for training and one group for evaluating the performance. The performance of the cross-validation is calculated by taking the mean and variance over all k performances. This enables a less biased estimate of model p e r f o r m a n c e [ 4 9 ] . ( R e f e r t o h t t p s : // g i t h u b . c o m / WoldringLabMSU/DeepLearning.git, for example, K-fold crossvalidation).

3.5 Protein Library Construction

Machine learning techniques enable the design of smart libraries by identifying the protein positions that are highly amenable to mutation and determining the most suitable degenerate codon candidate(s) for the protein of interest. Based on these designs, full length genes can then be constructed using overlap extension PCR of degenerate oligos to incorporate the intended site-wise amino acid diversity. Finally, the newly constructed full length gene library, combined with a linearized yeast surface display vector (e.g. pCT-CON2), can be electroporated into yeast (EBY100) [50] and evaluated by high-throughput techniques [16, 51–56].

100

4

Mehrsa Mardikoraem and Daniel Woldring

Notes 1. Informed models: It should be clarified that it not always large datasets and high-complexity algorithms that improve the prediction performance (a simple feed-forward neural network including the epistasis relationship in feature representation may perform better than a convolutional neural network with one-hot encoding as its feature representation method). 2. Learning paradigms in ANN: The learning paradigms in ANN are supervised learning, unsupervised learning, semisupervised learning, and reinforcement learning. In supervised learning, the data are provided with assigned labels that then guide the algorithm to learn the input–output mapping through a backpropagation process. Unsupervised learning algorithms (e.g. K-nearest neighbors and K-means) are used when there are no values assigned to the input features. This allows for the detection of patterns even when additional information is not provided with the input data. Semi-supervised learning is a method well suited where only partially labeled data exists, but the algorithm aims to fully benefit from provided information either labeled or unlabeled [57]. Reinforcement learning is agent-based learning interacting with the environment. Therefore, the algorithm learns based on rewards from correct prediction and penalties from incorrect guesses (i.e., trial and error methodology) [58]. 3. Deep learning packages analysis: TensorFlow is one of the most popular and fastest evolving open source deep learning tools [37]. It is compatible with both GPU/CPU computation and is well suited for working with multi-dimensional arrays. One downside could be the low-level API which makes it to be not the ideal choice for the direct creation of deep learning algorithms. Keras (open source) can support backends such as CNTK, TensorFlow, and Theano and its simple API makes it straightforward to implement. PyTorch is among the more flexible programming packages in python and supports tensor computation and GPU-acceleration. Its dynamic graph and easy debugging make it a strong option to choose. At its core, it uses CPU and GPU tensor and NN backends. Therefore, PyTorch brings speed and flexibility to deep learning models despite needing third-party visualization [59]. Regardless of the choice in packages, the magnitude and quality of the input data will have a strong impact on the predictive power of the resulting model. 4. One-hot encoding: The simplest method is to use one-hot encoding by constructing a matrix of one and zeros where one represents the existence of the element at a specific position

Machine Learning-Driven Protein Library Design

101

of the sequence. One-hot encoding is easy to implement and has been proven to be effective in many cases, yet it is highly memory intensive and struggles to capture the relationship between amino acids in protein sequences. This large and sparse encoding often leads to complications in training as a result of the inherently high dimensionality. 5. Integer encoding: Integer encoding is implemented by representing each amino acid by an integer. For integer encoding one observed drawback is the tendency of the system to assume linear relationships among the provided labels. For example, if the labels are 1, 2, and 3, the system assumes a relationship between the amino acids (such that 1 is closer to 2 than 3) that are assigned to these labels. As a result, orthogonality between the labels matters. However, integer encoding is often used with a linear embedding layer (see tf.keras.layers.embedding) whereby integer encoding calls an embedding column like a “lookup” table. 6. Property-based encoding: Some practical encodings are obtained based on the physiochemical properties (e.g. charge, hydrophobicity and size) of the sequences [60]. One example is principal components score Vectors of Hydrophobic, Steric, and Electronic properties (VHSE8) [61]. 7. Sequence embedding via self-supervised learning: Utilizing language-based models and taking advantage of techniques such as next token prediction and masked token prediction, fixed vector representations of protein sequences will be reached. Two benchmark studies in this area are UniRep [20] and TAPE [38] embeddings. 8. Desired data distribution for DL algorithms: Gaussian-like distributions tend to have better performance in deep learning. Therefore, dealing with data having a skewed distribution may benefit from applying power transform functions to make the data more gaussian. For example, simple power transform functions may take the nth root or the nth order logarithm of the variable. More advanced power-transformers include the Box-Cox and Yeo-Johnson methods. In order to obtain new distributions with the mentioned methods, Scipy [62] library or scikit-learn preprocessing package power transformer can be used: import scipy import scipy.stats Yeo-Johnson = scipy.stats.yeojohnson(label_List) Box_Cox = scipy.stats.boxcox(label_List)

102

Mehrsa Mardikoraem and Daniel Woldring

9. Activation function: One suggestive method is to use ReLU in all hidden layers and choose an appropriate activation function for the output layer to match the distribution of data with the nature of the task. Generally, for the output layer, sigmoid (for binary-class), softmax, and tanh (for multi-class) are used for classification tasks, and linear activation function is used for regression. 10. Evaluation metrics: Evaluation metrics are representative of algorithm performance and should be considered within the context of the nature of the problem and origin of the input data. As an example, in biased data when 90% of the population are in category 1 and the remainder are in category 2, it is highly probable that the algorithm predicts all the data to be in the first category. In this case, if one wants to rely on the metrics, the accuracy will be 90% which is not addressing the performance of the algorithm (i.e., not representing the poor performance in prediction of other class). In this case, the confusion matrix provides useful information about the number of false positives, false negatives, true positives, and true negatives. References 1. Hogan BL (1996) Bone morphogenetic proteins: multifunctional regulators of vertebrate development. Genes Dev 10:1580–1594 2. Schlessinger J (2000) Cell signaling by receptor tyrosine kinases. Cell 103:211–225 3. Syrovatkina V, Alegre KO, Dey R et al (2016) Regulation, signaling, and physiological functions of G-proteins. J Mol Biol 428: 3850–3868 4. Hellinga HW, Marvin JS (1998) Protein engineering and the development of generic biosensors. Trends Biotechnol 16:183–189 5. Mishra NK, Chang J, Zhao PX (2014) Prediction of membrane transport proteins and their substrate specificities using primary sequence information. PLoS One 9:e100278 6. Yang T, Wu JC, Yan C et al (2011) Virtual screening using molecular simulations. Proteins 79:1940–1951 7. Wrenbeck EE, Faber MS, Whitehead TA (2017) Deep sequencing methods for protein engineering and design. Curr Opin Struct Biol 45:36–44 8. Kronqvist N, Lo¨fblom J, Jonsson A et al (2008) A novel affinity protein selection system based on staphylococcal cell surface display and flow cytometry. Protein Eng Des Sel 21: 247–255

9. Yang KK, Wu Z, Arnold FH (2019) Machinelearning-guided directed evolution for protein engineering. Nat Methods 16:687–694 10. Bohr H, Bohr J, Brunak S et al (1990) A novel approach to prediction of the 3-dimensional structures of protein backbones by neural networks. FEBS Lett 261:43–46 11. Ofran Y, Rost B (2003) Predicted proteinprotein interaction sites from local sequence information. FEBS Lett 544:236–239 12. Ward JJ, McGuffin LJ, Buxton BF et al (2003) Secondary structure prediction with support vector machines. Bioinformatics 19: 1650–1655 13. Petrova NV, Wu CH (2006) Prediction of catalytic residues using Support Vector Machine with selected protein sequence and structural properties. BMC Bioinformatics 7:1–12 14. Li BQ, Feng KY, Chen L et al (2012) Prediction of protein-protein interaction sites by random forest algorithm with mRMR and IFS. PLoS One 7:1–10 15. Quan L, Lv Q, Zhang Y (2016) STRUM: structure-based prediction of protein stability changes upon single-point mutation. Bioinformatics 32:2936–2946 16. Golinski AW, Mischler KM, Laxminarayan S et al (2021) High-throughput developability

Machine Learning-Driven Protein Library Design assays enable library-scale identification of producible protein scaffold variants. Proc Natl Acad Sci U S A 118:1–11 17. Tahir M, Tayara H, Chong KT (2019) iRNAPseKNC(2methyl): identify RNA 2’-O-methylation sites by convolution neural network and Chou’s pseudo components. J Theor Biol 465: 1–6 18. Bloom JD, Labthavikul ST, Otey CR et al (2006) Protein stability promotes evolvability. Proc Natl Acad Sci U S A 103:5869–5874 19. Saito Y, Oikawa M, Nakazawa H et al (2018) Machine-learning-guided mutagenesis for directed evolution of fluorescent proteins. ACS Synth Biol 7:2014–2022 20. Alley EC, Khimulya G, Biswas S et al (2019) Unified rational protein engineering with sequence-based deep representation learning. Nat Methods 16:1315–1322 21. Biswas S, Khimulya G, Alley EC, Esvelt, KM, Church GM (2021) Low-N protein engineering with dataefficient deep learning. Nat Methods 18(4):389–396 https://doi.org/10. 1038/s41592-021-01100-y 22. Suzek BE, Wang Y, Huang H et al (2015) UniRef clusters: a comprehensive and scalable alternative for improving sequence similarity searches. Bioinformatics 31:926–932 23. Crawshaw M (2020) Multi-Task Learning with Deep Neural Networks: A Survey. arXiv:2009.09796 24. Im J, Park B, Han K (2019) A generative model for constructing nucleic acid sequences binding to a protein. BMC Genomics 20:1–13 25. Ness JE, Kim S, Gottman A et al (2002) Synthetic shuffling expands functional protein diversity by allowing amino acids to recombine independently. Nat Biotechnol 20:1251–1255 26. Gupta RD, Tawfik DS (2008) Directed enzyme evolution via small and effective neutral drift libraries. Nat Methods 5:939–942 27. Engqvist MKM, Nielsen J (2015) ANT: software for generating and evaluating degenerate codons for natural and expanded genetic codes. ACS Synth Biol 4:935–938 28. Jacobs TM, Yumerefendi H, Kuhlman B et al (2015) SwiftLib: rapid degenerate-codonlibrary optimization through dynamic programming. Nucleic Acids Res 43:e34 29. Mene´ndez ML, Pardo JA, Pardo L et al (1997) The Jensen-Shannon divergence. J Frankl Inst 334:307–318 30. Bewick V, Cheek L, Ball J (2004) Statistics review 12: survival analysis. Crit Care 8: 389–394

103

3 1 . Te n s o r fl o w ( 2 0 1 7 ) I n d e x @ W w w. Tensorflow.Org 32. Chollet F, & others (2015) Keras. GitHub. Retrieved from https://github.com/fchollet/ keras 33. Mazza D, Pagani M (2021) Automatic differentiation in PCF. Proc ACM Program Lang 5: 1–4 34. Pedregosa F, Varoquaux G, Gramfort A et al (2012) Scikit-learn: machine learning in Python. J Mach Learn Res 12 35. McKinney W (2010) Data structures for statistical computing in Python. In: Proc 9th Python Sci Conf 1, pp 56–61 36. Harris CR, Millman KJ, van der Walt SJ et al (2020) Array programming with NumPy. Nature 585:357–362 37. Abadi M, Barham P, Chen J et al (2016) TensorFlow: a system for large-scale machine learning. In: Proc 12th USENIX Symp Oper Syst Des implementation, vol 2016. OSDI, pp 265–283 38. Rao R, Bhattacharya N, Thomas N et al (2019) Evaluating protein transfer learning with tape. Adv Neural Inf Process Syst 32:9689 39. Whitehead TA, Chevalier A, Song Y et al (2012) Optimization of affinity, specificity and function of designed influenza inhibitors using deep sequencing. Nat Biotechnol 30:543–548 40. Shroff R, Cole AW, Diaz DJ et al (2020) Discovery of novel gain-of-function mutations guided by structure-based deep learning. ACS Synth Biol 9:2927–2935 41. Zhao Z, Gong X (2019) Protein-protein interaction interface residue pair prediction based on deep learning architecture. IEEE/ACM Trans Comput Biol Bioinformatics 16: 1753–1759 42. Zhang Q, Zhang M, Chen T et al (2019) Recent advances in convolutional neural network acceleration. Neurocomputing 323: 37–51 43. Yu Y, Si X, Hu C et al (2019) A review of recurrent neural networks: LSTM cells and network architectures. Neural Comput 31: 1235–1270 44. Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9: 1735–1780 45. Cho K, Merrie¨nboer B Van, Gulcehre C, et al (2014) Learning phrase representations using RNN encoder-decoder for statistical machine translation. EMNLP 2014 - 2014 Conf Empir Methods Nat Lang Process Proc Conf 1724–1734

104

Mehrsa Mardikoraem and Daniel Woldring

46. Vaswani A, Shazeer N, Parmar N, et al (2017) Attention is all you need, In: Advances in neural information processing systems, pp. 5998–6008 47. Brandes N, Ofer D, Peleg Y et al (2021) ProteinBERT: a universal deep-learning model of protein sequence and function. Comput Biol Chem 95:107596 48. Bergstra J, Yamins D, Cox DD (2013) Making a science of model search: Hyperparameter optimizationin hundreds of dimensions for vision architectures. Presented at the 30th International Conference on Machine Learning (ICML 2013), Atlanta, Gerorgia, June 16–21, 2013. In JMLR Workshop and Conference Proceedings 28(1):115–123 49. Raschka S (2018) Model Evaluation, Model Selection, and Algorithm Selection in Machine Learning. arXiv:1811.12808 50. Chao G, Lau WL, Hackel BJ et al (2006) Isolating and engineering human antibodies using yeast surface display. Nat Protoc 1:755–768 51. Woldring DR, Holec PV, Zhou H et al (2015) High-throughput ligand discovery reveals a sitewise gradient of diversity in broadly evolved hydrophilic fibronectin domains. PLoS One 10:e0138956 52. Woldring DR, Holec PV, Stern LA et al (2017) A gradient of sitewise diversity promotes evolutionary fitness for binder discovery in a threehelix bundle protein scaffold. Biochemistry 56: 1656–1671 53. Kruziki MA, Bhatnagar S, Woldring DR et al (2015) A 45-amino-acid scaffold mined from the pdb for high-affinity ligand engineering. Chem Biol 22:946–956

54. Kruziki MA, Sarma V, Hackel BJ (2018) Constrained combinatorial libraries of Gp2 proteins enhance discovery of PD-L1 binders. ACS Comb Sci 20:423–435 55. Stern LALA, Csizmar CMCM, Woldring DRDR et al (2017) Titratable avidity reduction enhances affinity discrimination in mammalian cellular selections of yeast-displayed ligands. ACS Comb Sci 19:315–323 56. Hasenhindl C, Traxlmayr MW, WozniakKnopp G et al (2013) Stability assessment on a library scale: a rapid method for the evaluation of the commutability and insertion of residues in C-terminal loops of the CH3 domains of IgG1-Fc. Protein Eng Des Sel 26:675–682 57. Zhu X, Goldberg AB (2009) Introduction to semi-supervised learning. Synth Lect Artif Intell Mach Learn 3:1–130 58. Sutton RS, Barto AG (1998) Reinforcement learning: an introduction. In: Sutton RS, Barto AG (eds) Bradford book. The MIT Press 59. Nguyen G, Dlugolinsky S, Boba´k M et al (2020) Machine learning and deep learning frameworks and libraries for large-scale data mining : a survey. Artif Intell Rev 52:77–124 60. Yang KK, Wu Z, Bedbrook CN et al (2018) Learned protein embeddings for machine learning. Bioinformatics 34:2642–2648 61. Mei HU, Liao ZH, Zhou Y et al (2005) A new set of amino acid descriptors and its application in peptide QSARs. Pept Sci Orig Res Biomol 80:775–786 62. Virtanen P, Gommers R, Oliphant TE et al (2020) SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat Methods 17:261–272

Chapter 6 Kinetic Competition Screening of Yeast-Displayed Libraries for Isolating High Affinity Binders Nicole J. Yang Abstract Yeast surface display is a robust platform for obtaining binders with high affinity. Kinetic competition screening is an effective method for maturing the affinity of binders with strong starting affinities, or when dissociation kinetics are a key consideration for the protein of interest. In this chapter, we describe detailed protocols for setting up and performing a kinetic competition screen. The duration of competition is determined based on the dissociation rate constant of the parental clone measured on the yeast surface. This methodology was successfully used to improve the affinity of a viral double-stranded RNA binding protein with starting affinity in the sub-nanomolar range. Key words Yeast surface display, Kinetic screening, Kinetic competition screening, Affinity maturation

1

Introduction Yeast surface display (YSD) is a leading platform for engineering various protein characteristics, including expression and stability [1, 2], binding specificity [3, 4], catalytic activity [5, 6], and perhaps most representatively, binding affinity [7]. YSD has been successfully utilized to improve the affinity of existing binding interactions, as well as isolate de novo binders from naı¨ve combinatorial libraries based on antibody or alternative scaffolds [8]. Generally, the library is created using random mutagenesis to yield a typical diversity of 107–109 variants, which are screened using avid magnetic beads [9] or FACS. Compatibility with flow cytometry is a major advantage of YSD, which enables quantitative screening approaches that finely discriminate between clones [10]. Equilibrium screening is a major screening strategy which selects improved mutants based on their equilibrium dissociation constants (KD). Here, the library is allowed to reach equilibrium with a fixed concentration of ligand, typically below the KD of the wild-type interaction. To allow equilibrium to be reached without

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_6, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

105

106

Nicole J. Yang

Biotinylated ligand

Yeast library

Saturate with labeled ligand

Unlabeled ligand

Dissociation with excess unlabeled competitor

Secondary detection

FACS to isolate improved mutants

Fig. 1 General strategy for kinetic competition screening. Following generation of the library, sorting is performed by saturating the yeast surface with labeled ligand (biotinylated or fluorescently labeled), washing away any unbound material, and initiating competition in the presence of excess unlabeled ligand. After allowing competition to proceed for a set duration which is typically determined by the dissociation kinetics of the parental clone, secondary staining is performed and improved mutants are isolated by FACS

depleting free ligand in solution, it is important to provide an excess of ligand over the number of displayed proteins that are present. Thus at low ligand concentrations, the incubation volume must be increased to provide sufficient molar excess of the ligand. When screening for binders with higher affinities (KD in the low nanomolar range), incubation volumes may then become impractically large to work with. An alternative screening strategy is kinetic competition screening, where clones are selected based on their dissociation rate constants (koff) (Fig. 1). Cells are first labeled to saturation with a labeled ligand and washed. Surface-bound, labeled ligand is then allowed to dissociate in the presence of competing unlabeled ligand, which is provided in excess to prevent re-association of labeled ligand. Here, the duration of competition is a key parameter determining the stringency and success of the sort: a competition time that is too short cannot provide adequate discrimination between improved mutants and the wild-type clone, whereas a competition time that is too long will result in the loss of improved mutants, as eventually the surface labeling will decay to background. Properly designed kinetic competition screens can be an effective method for isolating binders with high affinity [11] or engineering proteins for which slow dissociation kinetics is an important characteristic. In this chapter, we introduce methods that were used to mature the affinity of the double-stranded RNA (dsRNA)-binding protein p19 of the Carnatian Italian Ringspot Virus (CIRV), which has a starting affinity reported in the sub-nM range [12]. p19 was engineered as an siRNA carrier for delivery applications, to be loaded with siRNA prior to administration to cells or animals, after which

Kinetic Competition Screening of Yeast-Displayed Libraries

107

the ligand would start to dissociate from the carrier. The library was created by diversifying the parental clone via random mutagenesis following previously reported methods [13, 14], which we do not describe here. Kinetic competition screening successfully isolated mutants with six- to tenfold improvements in binding affinity without further rounds of mutagenesis [15].

2

Materials

2.1

Yeast Library

2.2

Media and Plates

1. Yeast library based on the Saccharomyces cerevisiae strain EBY100 and the display vector pCTCON2. Detailed methods for performing random mutagenesis via error-prone PCR, transformation of yeast and library characterization are available elsewhere [13, 14] and in Chapter 2 of this volume. For affinity maturation by kinetic screening, the parental clone will typically have a starting KD at or below 1–10 nM. The affinity of clones with KDs higher than this range can be improved effectively using equilibrium screening strategies. Media and plates may be stored at 4
C for up to 6 months. 1. SD-CAA media, pH 4.5 citrate buffer (see Note 1): 20 g/L dextrose (D-glucose), 6.7 g/L yeast nitrogen base, 5 g/L casamino acids (ade, ura, trp), 7.4 g/L citric acid monohydrate, 10.4 g/L sodium citrate. Filter-sterilize. 2. SD-CAA plates, pH 6.0: 20 g/L dextrose (D-glucose), 6.7 g/L yeast nitrogen base, 5 g/L casamino acids, 5.4 g/L Na2HPO4, 8.6 g/L NaH2PO4  H2O, 16 g/L agar, 182 g/L sorbitol (see Note 2). 3. SG-CAA media, pH 6.0 phosphate buffer: 18 g/L galactose, 2 g/L dextrose (D-glucose), 6.7 g/L yeast nitrogen base, 5 g/ L casamino acids, 5.4 g/L Na2HPO4, 8.6 g/L NaH2PO4  H2O. Filter-sterilize.

2.3 Reagents and Buffers

1. PBSA, pH 7.4: 10 mM Na2HPO4, 2 mM KH2PO4, 137 mM NaCl, 2.7 mM KCl, 1 g/L BSA. 2. Labeled ligand, typically biotinylated (see Note 3). Fluorescently labeled ligands are also suitable. 3. Unlabeled ligand. 4. Detection reagents: Chicken anti-c-myc IgY (1 mg/mL) (Gallus Immunotech). Goat anti-chicken IgG-Alexa Fluor 488 (2 mg/mL) (Thermo Fisher Scientific). Goat anti-chicken IgG-Alexa Fluor 647 (2 mg/mL) (Thermo Fisher Scientific).

108

Nicole J. Yang

Streptavidin-Alexa Fluor 488 (2 mg/mL) (Thermo Fisher Scientific). Streptavidin-Alexa Fluor 647 (2 mg/mL) (Thermo Fisher Scientific). Mouse anti-biotin IgG-PE (BK-1/39) (0.2 mg/mL) (eBioscience). 2.4 Other Consumables and Equipment

1. Microcentrifuge tubes and appropriate flow cytometry tubes with filter caps. 2. V-bottom 96 well plates and multi-channel pipet. 3. Orbital shaker or rotating platforms stationed at the desired temperature (see Note 4).

3

Methods

3.1 Growth and Induction of Yeast

1. Inoculate yeast into fresh SD-CAA media and grow overnight (16–20 h) at 30
C with shaking (250 rpm). For a yeast library, include 10–100-fold the size of the library to prevent the loss of rare clones. Aim for a starting density of 1  106 to 1  107 cells/mL. For clonal yeast, pick 1–2 colonies into 5 mL of SD-CAA media in a 15 mL glass culture tube. Overnight growth will typically yield a culture at OD600 6–10. 2. Passage into fresh SD-CAA. Measure the OD600 of the overnight culture. An OD600 of 1 corresponds to approximately 1  107 cells/mL. For a yeast library, pellet at least tenfold the library diversity by centrifuging at 3000  g for 5 min and resuspend in fresh SD-CAA media at an OD600 of 0.5–1. For clonal yeast, dilute the overnight culture into fresh SD-CAA media to an OD600 of 0.5–1. Grow at 30
C with shaking (250 rpm) until cells reach the mid-log growth phase (OD600 of 2–5) (see Note 5), during which induction is optimal. 3. Induce cells. For a yeast library, pellet at least tenfold the library diversity (3000  g for 5 min), discard the supernatant and resuspend in SG-CAA media to an OD600 of 1. For clonal yeast, also resuspend in SG-CAA media at an OD600 of 1. Induce overnight (16–20 h) at 20
C with shaking (250 rpm) (see Note 6). 4. Cells can be used directly for selection and analysis or stored at 4
C for several weeks.

3.2 Measurement of Dissociation Rate Constants on the Yeast Surface

A critical parameter for kinetic screening is the duration of competition, which can be estimated based on the dissociation rate constant (koff) of the parental clone. Yeast display allows convenient measurement of binding kinetics on the yeast surface using flow cytometry, without having to express and purify soluble protein.

Kinetic Competition Screening of Yeast-Displayed Libraries

109

Below we describe a protocol for measuring the dissociation rate constant using induced yeast. Analysis is performed using 96 well plates for convenience, but the protocol can be readily adapted for use with microcentrifuge tubes. Dissociation is performed in a reverse time course and samples are stained for flow cytometry together at the end. 1. Grow and induce clonal yeast as described in Subheading 3.1. 2. Wash yeast. Centrifuge the appropriate number of yeast (see Note 7) in a 1.5 mL microcentrifuge tube at 14,000  g for 1 min. Discard the supernatant and resuspend in 1 mL of cold PBSA. Centrifuge again at 14,000  g for 1 min and discard the supernatant. 3. Complex with labeled antigen. Resuspend yeast in 100 μL of PBSA containing labeled antigen (see Note 8). Include a primary antibody for detecting full length expression (such as the Chicken anti-c-myc antibody (Gallus Immunotech) at a 1:200– 1:1000 dilution or 1–5 μg/mL) (see Note 9). To drive saturation, the labeled antigen should be present at a concentration higher than the expected KD by tenfold or higher, and in stoichiometric excess of displayed protein present in solution by tenfold or higher (see Note 10). Incubate with rotation for 15–30 min at room temperature, after which saturation can be assumed (see Note 11). Then transfer to 4
C with continued rotation for the duration of the experiment. 4. Prepare dissociation buffer containing unlabeled ligand in PBSA. Also include the primary antibody for detecting full length expression at a 1:1000 dilution (1 μg/mL) to maintain binding. The unlabeled ligand should be present at an approximately 100-fold excess of the expected concentration of the labeled antigen, assuming saturation (see Note 10). The unlabeled ligand competes for binding with the labeled ligand, and discourages re-association of any labeled ligand that has dissociated. 5. Initiate competition. Samples are prepared in the reverse order, starting with the last timepoint first. For each timepoint, perform a partial wash by transferring 1  105 cells (previously incubating with labeled antigen) per well to a V-bottom 96 well plate containing 250 μL of cold PBSA. Centrifuge for 3000  g for 5 min at 4
C and discard the supernatant. Resuspend the pellet in 250 μL of dissociation buffer and transfer to a separate V-bottom 96 well plate. Incubate at room temperature or 37
C as appropriate (see Note 12) on an orbital shaker. Samples can be continued to be processed in the remaining wells of these two 96 well plates. 6. After kinetic competition is completed for all samples, all subsequent steps should be performed at 4
C with ice-cold

110

Nicole J. Yang

reagents to minimize dissociation of bound ligand. Wash all samples by spinning the plate at 3000  g for 5 min at 4
C, discarding the supernatant, and resuspending in 250 μL of cold PBSA. Spin and discard the supernatant. 7. Perform secondary labeling for flow cytometry. Incubate with the appropriate secondaries for 5–10 min at 4
C with shaking in a 50 μL volume. Typical reagents used are Goat anti-chicken IgG-Alexa Fluor 488 or 647 at a 1:1000 dilution (2 μg/mL) and Streptavidin-Alexa Fluor 647 or 488 at a 1:1000 dilution (2 μg/mL) (see Notes 13 and 14). The secondary antibodies should also be present in stoichiometric excess over displayed protein in solution. 8. Wash cells by adding 250 μL of cold PBSA, centrifuging for 3000  g for 5 min at 4
C, and discarding the supernatant. Repeat the wash and keep the cells pelleted on ice until they are to be run on the flow cytometer to prevent premature dissociation. 9. Analyze on a flow cytometer and plot the median fluorescence intensities of yeast displaying full length protein. Fit to exponential decay to determine the dissociation rate constant (see Note 15). 10. Determine the duration of kinetic competition to be used for screening the library. Mathematical models allow straightforward estimation of the optimal competition time which maximizes the difference in binding between the parental clone and improved mutants [16]. Generally, aim for when fluorescent signal from the parental clone has mostly decayed. 3.3 Kinetic Competition Sorting by FluorescenceActivated Cell Sorting (FACS)

1. Grow and induce the yeast library as described in Subheading 3.1. Also prepare yeast displaying the parental clone to guide gating. The following groups should be prepared in parallel with 5  106 cells per group: (a) unlabeled control; (b) singlecolor controls; and (c) the parental clone subject to the same procedures below as the library. 2. Wash cells. Pellet at least tenfold the library diversity by centrifuging for 5 min at 3000  g in 50 mL conical tubes. Remove the supernatant and resuspend in 50 mL of cold PBSA to wash. Centrifuge again for 5 min at 3000  g and discard the supernatant. 3. Repeat wash. Resuspend the library in 1 mL PBSA and transfer to a 1.5 mL microcentrifuge tube. Centrifuge at 14,000  g for 1 min and discard the supernatant. 4. Complex with labeled ligand (see Note 16). Resuspend the yeast library at 1  108 cells/mL in labeled antigen prepared in PBSA. The concentration of the labeled antigen should be sufficiently high to saturate the yeast surface (see Note 10).

Kinetic Competition Screening of Yeast-Displayed Libraries

111

Also include the primary antibody for detecting full length expression (e.g., Chicken anti-cmyc antibody at a 1:1000 dilution or 1 μg/mL). Incubate at room temperature for 15–30 min with rotation (see Note 11). 5. Wash cells. Centrifuge at 14,000  g for 1 min, discard the supernatant, and resuspend in 1 mL cold PBSA. Repeat and discard the supernatant. 6. Initiate competition. Resuspend the yeast library in PBSA containing excess unlabeled antigen. The unlabeled antigen is included at a concentration that is approximately 100-fold of the estimated concentration of labeled antigen. Allow competition to occur at room temperature or 37
C as appropriate with rotation for the duration determined in Subheading 3.2 (see Note 17). 7. Wash cells. Centrifuge for 14,000  g for 1 min, discard the supernatant, and resuspend in 1 mL cold PBSA. Repeat and discard the supernatant. 8. Perform secondary staining. Resuspend the yeast library at 1  108 cells/mL in PBSA containing secondary detection reagents (e.g. Goat anti-chicken IgG-Alexa Fluor 647, 1:100 (20 μg/mL) combined with Streptavidin-Alexa Fluor 488, 1: 100 (20 μg/mL) or anti-biotin IgG-PE, 1:100) (2 μg/mL) (see Note 18). Incubate at 4
C for 10 min with rotation. 9. Wash cells. Centrifuge for 14,000  g for 1 min, discard the supernatant, and resuspend in 1 mL cold PBSA. 10. Repeat the wash. After resuspending in 1 mL of cold PBSA, aliquot the library into 1.5 mL microcentrifuge tubes each containing 2  107 cells. Centrifuge at 14,000  g for 1 min and discard the supernatant. The aliquots should be stored undiluted on ice until immediately prior to sorting. 11. Fluorescence-activated cell sorting (FACS). First run the unlabeled and single-color controls to establish settings on the flow cytometer. Next, collect approximately 105 events for the parental clone and a small fraction of the library to set the sorting gate. The gate should mostly exclude the parental clone while collecting all cells in the library with discernable binding (Fig. 2) (see Note 19). 12. Resuspend each aliquot of the yeast library in 0.4 mL of cold PBSA and filter into the appropriate FACS tube. Collect cells into a 10 mL glass culture tube containing 2 mL of SD-CAA with antibiotics (see Note 20). After the sort, rinse the side of the walls with 3 mL of SD-CAA to collect all cells (see Note 21). 13. Incubate the isolated cells at 30
C with shaking (250 rpm) overnight or until the culture reaches saturation OD600.

112

Nicole J. Yang

Fig. 2 Library enrichment from a kinetic competition screen. Representative FACS plots and sorting gates from three rounds of kinetic screens during affinity maturation of p19. The duration of competition was held constant in the presented campaign. The wild-type parental clone processed in parallel with the 1.0 generation library is shown on the left for comparison

Passage and induce the library as described in Subheading 3.1 for subsequent screening. 14. Repeat additional rounds of kinetic screening and FACS until the desired enrichment is achieved. If necessary, perform additional rounds of mutagenesis and screening for further improvement (see Note 22).

4

Notes 1. The pH is lowered to discourage bacterial growth. When collecting yeast during FACS, further supplement the media with 50 μg/mL kanamycin, 100 kU/L penicillin, and 0.1 g/L streptomycin (or with commercially available pen-strep solutions) to prevent bacterial contamination. Lower pH is not recommended during induction in SG-CAA media to facilitate proper folding of displayed protein. 2. Filter-sterilize the dextrose, yeast nitrogen base and amino acids in 10% of the final volume. Autoclave the buffer salts, agar, and sorbitol in 90% of the final volume. Once the autoclaved mixture has sufficiently cooled, combine the two solutions and pour. 3. Biotinylation kits are available from Thermo Fisher Scientific which non-specifically label primary amines that are surfaceexposed (the side chains of lysines and the N terminus). Alternatively, biotin ligase (BirA) can be used to specifically label proteins containing a 15 amino acid tag (Avi tag). BirA may also be co-expressed with Avi-tagged constructs for labeling during recombinant expression in E. coli. 4. All incubations with yeast should be performed with rotation or shaking to avoid cells from settling.

Kinetic Competition Screening of Yeast-Displayed Libraries

113

5. The doubling time of the EBY100 strain in SD-CAA is approximately 3–4 h. 6. Induction at lower temperature aids the surface expression of less thermally stable scaffolds. If desired, the induction temperature can be raised to 30–37
C to select for stable clones [2]. The p19 library was induced at 37
C overnight. 7. Prepare enough cells to analyze 1  105 cells per timepoint, with multiple timepoints spanning 0 min of dissociation to approximately 10 times the expected dissociation half time. Triplicates are recommended for each timepoint. 8. Typical labeling volumes are 50–100 μL for up to 1  107 cells and 0.5–1 mL for 1  108 cells. 9. If desired, yeast may be incubated with the anti-c-myc primary antibody at the end after competition is complete. In this case, incubation is performed at 4
C with ice-cold reagents to minimize dissociation of cell-bound ligand. Here, we have included the antibody earlier to eliminate this future step and further prevent any unwanted dissociation. 10. Typical levels of expression are 105 molecules per yeast, which can be used to estimate the number and concentration of displayed protein present in solution. For instance, 1  105 cells in a volume of 100 μL will have approximately 105 cells  (1  105 protein/cell)  (1 mol/6.02  1023 protein)/100 μL ¼ 0.17 nM of displayed protein. 11. The half-life to approach equilibrium (τ1/2) can be calculated using the following formula: τ1=2 ¼

ln 2 kon ½L 0  þ koff

kon is the association rate in units of M1 s1 (typically 1  105 M1 s1 for protein–protein interactions), koff is the dissociation rate in units of s1, and [L]0 is the ligand concentration in units of M. Five half-lives are sufficient to reach 97% of equilibrium. If only the KD is known, the koff can be estimated using the following formula: KD ¼

koff kon

For example, for an expected KD of 1 nM, labeling at a ligand concentration of 100 nM can be estimated to reach 97% of equilibrium in 5τ1=2 ¼ 5  ln 2 ¼ 6 min . ð1105 M1 s1 Þð100 nMÞþð104 s1 Þ 12. Yeast can be incubated in PBSA at room temperature or 37
C for several hours without affecting viability. However, for longer incubation times, cell viability following incubation should be tested in a pilot experiment.

114

Nicole J. Yang

13. Secondary detection reagents for biotin are omitted if fluorescently labeled ligands were used. 14. The combination of Alexa Fluor 488 or PE and Alexa Fluor 647 fluorophores avoids the need for compensation. 15. Background (MFI(1)) is subtracted from the median fluorescence intensities at each timepoint (MFI(t)), and the resulting values are normalized and fitted to the following formula to calculate the dissociation rate koff: Fraction bound ¼

MFIðt Þ  MFIð1Þ ¼ e ðkoff t Þ MFIðt ¼ 0Þ  MFIð1Þ

where t is the duration of competition at each timepoint, MFI(t ¼ 0) is the fluorescence intensity at 0 min of dissociation, and MFI(1) is the background fluorescence signal of cells once dissociation is complete. 16. For screening of the p19 library, siRNAs directly labeled with fluorophores were used for selection. Three separate sequences of siRNA each labeled with a different fluorophore were used sequentially for subsequent rounds of selection to avoid promoting binding to specific siRNA sequences or fluorophores. 17. For screening of the p19 library, kinetic competition was performed under conditions mimicking the intended in vivo use of the protein, i.e. PBSA containing 55% mouse serum at 37
C. The presence of serum slightly accelerated the decay of cellbound fluorescence signal, potentially due to RNase-mediated degradation of siRNA. 18. Although unlikely at this stage of affinity maturation, alternating between streptavidin and anti-biotin antibodies can be a strategy to prevent the isolation of streptavidin binders. 19. For initial rounds of sorting, lower stringency (collecting approximately 1–5% of displaying cells) is recommended to avoid losing unique clones. Stringency is increased in later rounds to collect 0.1–0.5% of displaying cells with strong display and binding. 20. The upper limit of the diversity of the new library can be estimated as the total number of cells collected, which will increasingly overestimate the actual size of the library as additional rounds of screening are performed and improved mutants are enriched. 21. If excess volume is collected after sorting, rinse the side of the tubes with SD-CAA, centrifuge at 3000  g for 5 min, remove the supernatant and resuspend in 5 mL of fresh SD-CAA. Sheath fluid in excess of 1 mL may cause flocculation. 22. For kinetic screening of the p19 library, three rounds of screening were performed each using different siRNA ligands with

Kinetic Competition Screening of Yeast-Displayed Libraries

115

15 min of competition, followed by three rounds of screening with 1 h of competition. This campaign resulted in the enrichment of two dominant clones, the affinities of which were improved by 6-fold and 40-fold, respectively, over the wildtype clone.

Acknowledgments The author would like to thank Dane Wittrup and Wittrup lab alumni for training in yeast surface display. References 1. Traxlmayr MW, Faissner M, Stadlmayr G et al (2012) Directed evolution of stabilized IgG1Fc scaffolds by application of strong heat shock to libraries displayed on yeast. Biochim Biophys Acta Prot Proteom 1824:542–549. https://doi.org/10.1016/j.bbapap.2012. 01.006 2. Shusta EV, Holler PD, Kieke MC et al (2000) Directed evolution of a stable scaffold for T-cell receptor engineering. Nat Biotechnol 18:754– 759. https://doi.org/10.1038/77325 3. Chen TF, Li KK, Zhu EF et al (2018) Artificial anti-tumor opsonizing proteins with fibronectin scaffolds engineered for specificity to each of the murine FcγR types. J Mol Biol 430: 1786–1798. https://doi.org/10.1016/j.jmb. 2018.04.021 4. Angelini A, Miyabe Y, Newsted D et al (2018) Directed evolution of broadly crossreactive chemokine-blocking antibodies efficacious in arthritis. Nat Commun 9:1461. https://doi. org/10.1038/s41467-018-03687-x 5. Chen I, Dorr BM, Liu DR (2011) A general strategy for the evolution of bond-forming enzymes using yeast display. PNAS 108: 11399–11404. https://doi.org/10.1073/ pnas.1101046108 6. Parthasarathy R, Bajaj J, Boder ET (2005) An immobilized biotin ligase: surface display of Escherichia coli BirA on Saccharomyces cerevisiae. Biotechnol Prog 21:1627–1631. https:// doi.org/10.1021/bp050279t 7. Boder ET, Wittrup KD (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol 15:553–557. https://doi.org/10.1038/nbt0697-553 8. Ko¨nning D, Kolmar H (2018) Beyond antibody engineering: directed evolution of alternative binding scaffolds and enzymes using yeast surface display. Microb Cell Factories 17:32. https://doi.org/10.1186/s12934018-0881-3

9. Ackerman M, Levary D, Tobon G et al (2009) Highly avid magnetic bead capture: an efficient selection method for de novo protein engineering utilizing yeast surface display. Biotechnol Prog 25:774–783. https://doi.org/10.1002/ btpr.174 10. VanAntwerp JJ, Wittrup KD (2000) Fine affinity discrimination by yeast surface display and flow cytometry. Biotechnol Prog 16:31–37. https://doi.org/10.1021/bp990133s 11. Boder ET, Midelfort KS, Wittrup KD (2000) Directed evolution of antibody fragments with monovalent femtomolar antigen-binding affinity. PNAS 97:10701–10705. https://doi.org/ 10.1073/pnas.170297297 12. Vargason JM, Szittya G, Burgyán J, Hall TMT (2003) Size selective recognition of siRNA by an RNA silencing suppressor. Cell 115:799– 811. https://doi.org/10.1016/S0092-8674 (03)00984-X 13. Angelini A, Chen TF, de Picciotto S et al (2015) Protein engineering and selection using yeast surface display. Methods Mol Biol 1319:3–36. https://doi.org/10.1007/978-14939-2748-7_1 14. Van Deventer JA, Wittrup KD (2014) Yeast surface display for antibody isolation: library construction, library screening, and affinity maturation. In: Ossipow V, Fischer N (eds) Monoclonal antibodies: methods and protocols. Humana Press, Totowa, NJ, pp 151–181 15. Yang NJ, Kauke MJ, Sun F et al (2017) Cytosolic delivery of siRNA by ultra-high affinity dsRNA binding proteins. Nucleic Acids Res 45:7602–7614. https://doi.org/10.1093/ nar/gkx546 16. Boder ET, Wittrup KD (1998) Optimal screening of surface-displayed polypeptide libraries. Biotechnol Prog 14:55–62. https:// doi.org/10.1021/bp970144q

Chapter 7 Engineering Proteins by Combining Deep Mutational Scanning and Yeast Display Preeti Sharma, Erik Procko, and David M. Kranz Abstract Protein engineering using display platforms such as yeast display and phage display has allowed discovery of proteins with therapeutic and industrial applications. Antibodies and T cell receptors developed for therapeutic applications are often engineered by constructing libraries of mutations in loops of five to ten residues called complementarity determining regions that are in proximity to the antigen. In the past decade, deep mutational scanning has become a powerful tool in a protein engineer’s toolbox, as it allows one to compare the impact of all 20 amino acids at each position, across the length of the protein. Thus, a single experiment can provide a sequence-activity landscape with information about hotspots or suboptimal binding sites in the original proteins. These residues or regions may be overlooked by engineering methods that are driven solely by structures or directed evolution of error-prone PCR libraries. Here, we describe experimental methods to engineer proteins by combining yeast display and deep mutational scanning mutagenesis, using T cell receptors as an example. Key words Protein engineering, Yeast display, Deep mutational scanning, Heat maps, Enrichment ratio, T cell receptors, Sequence-activity landscape

1

Introduction Protein engineering in order to generate enhanced stability, affinity, or even altered specificity has become a valuable approach for both basic and clinical applications. Over several decades, approaches have included the use of directed evolution with libraries of mutated proteins followed by selection schemes that involve methods such as phage display [1, 2] or yeast display [3, 4]. The choice of where in the protein to generate mutations (typically by site directed, PCR-based methods) is often guided by structural information about the protein and its ligand. Alternatively, it is possible to generate random mutations throughout the protein using errorprone PCR. Each of these approaches is limited by the amount of

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_7, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

117

118

Preeti Sharma et al.

sequence space that can be sampled. While error-prone PCR conditions can be chosen that exhaustively generate all possible amino acid substitutions at each residue, the required error rate would be so high as to riddle the protein with mutations. More recently, a new approach that can be used to engineer proteins has been added, that of deep mutational scans [5–10]. In this approach, each of the 20 amino acids can be substituted at every position of a protein by the introduction of degenerate codons, and these single codon libraries can be subjected to selection for stability or affinity, using a display method. Deep sequencing of the input and selected output yields information about the substitutions that reduce or improve stability or affinity. As opposed to random mutations generated by error-prone PCR that are primarily confined to single nucleotide changes within a codon, this approach samples all sequence space at each position. In addition, further enhancement of protein stability or affinity can be achieved by combination of individual mutations from a deep mutational scan, which if appropriately chosen can act additively. Here we describe methods used in this approach, focusing on T cell receptors displayed on the yeast surface as an example [11–13].

2

Materials

2.1 Yeast Display Strain and Plasmid

1. EBY100: Saccharomyces cerevisiae yeast display strain (GAL1AGA1::URA3 ura3-52 trp1 leu2Δ1 his3Δ200 pep4::HIS2 prb1Δ1.6R can1 GAL). 2. pCT302 plasmid (Drug resistance: Ampicillin; deposited with Addgene) (Fig. 1a).

2.2 Bacterial Strain for Propagating Plasmid DNA 2.3

DNA Purification

E. coli competent strain DH5α (subcloning efficiency).

1. QIAprep Spin miniprep kit (QIAGEN) for isolating plasmid DNA from E. coli strain DH5α. 2. Zymoprep Yeast plasmid miniprep II (Zymo Research) for isolating plasmid DNA from EBY100. 3. QIAquick Gel Extraction Kit. 4. QIAquick PCR Purification Kit. 5. Pellet Paint Co-Precipitant (Novagen).

2.4 Restriction Enzymes

NheI, XhoI, DpnI, Calf Intestinal Phosphatase (CIP).

Deep Mutational Scanning of T Cell Receptors

119

Fig. 1 Yeast surface display of a protein of interest. (a) A schematic of yeast display vector (pCT302). A gene of interest is cloned in-frame with yeast mating protein Aga2p for cell surface display, in addition to N-ter HA and C-ter c-myc tags. Ampicillin drug resistance gene and auxotrophic TRP1 marker allow propagation of the plasmid in E. coli and yeast, respectively. Galactose-inducible promoter induces protein expression. Splice4L, YRS, and T7 are primers that can be used for sequencing and PCR amplification of the gene. (b) Protein expression on the surface of pCT302-transformed, EBY100 yeast cells, can be monitored by antibodies that bind to the expression tags or to the protein of interest. In addition, a fluorescent ligand for the protein of interest can also be used to assess protein folding 2.5 Polymerase Chain Reaction

1. Forward and reverse primers: For sequencing or amplifying the gene cloned in pCT302 vector, the following primers can be used: (a) Forward primer: Splice4L (50 -GGC AGC CCC ATA AAC ACA CAG TAT-30 ). (b) Reverse primer: YRS (50 -CGA GCT AAA AGT ACA GTG GG-30 ) or T7 (50 -TAA TAC GAC TCA CTA TAG GG-30 ). 2. Deoxynucleotide (dNTP) solution mix (10 mM each dNTP) (NEB). 3. Phusion high-fidelity DNA polymerase and reaction buffer (ThermoFisher). 4. Molecular biology grade water. 5. Thermocycler.

120

2.6

Preeti Sharma et al.

Culture Media

1. LB (Luria-Bertani) liquid medium for E. coli: Dissolve 5 g yeast extract, 10 g bacto-tryptone, and 10 g NaCl in ddH2O. Adjust final volume of the solution to 1 L. Sterilize by autoclaving. For solid medium plates, add 15 g bacto agar to 1 L liquid medium and sterilize by autoclaving. Allow the medium to cool to 50–60  C, add antibiotic (ampicillin to a working concentration of 100 μg/mL), and mix by gentle swirling to avoid bubbles. Pour ~25 mL in 100 mm  15 mm plates in a laminar flow hood. Cool and store at 4  C. 2. YPD media for propagating EBY100: Dissolve 10 g yeast extract, 20 g bacto-peptone, and 20 g dextrose in ddH2O. Adjust final volume of the solution to 1 L. Sterilize by autoclaving. For solid medium plates, add 15 g bacto agar to 1 L liquid medium and sterilize by autoclaving. Allow the medium to cool to 50–60  C, and pour ~25 mL in 100 mm  15 mm plates in a laminar flow hood. Cool and store at 4  C. 3. SD-CAA media for propagating EBY100 transformed with pCT302: Dissolve 14.8 g sodium citrate dihydrate, 4.2 g citric acid monohydrate, 5 g casamino acids, 6.7 g yeast nitrogen base without amino acids, and 20 g dextrose in ddH2O. Add 10 mL penicillin-streptomycin (10,000 U/mL). Adjust final volume of the solution to 1 L. Sterile filter through a 0.22 μm membrane, and store at 4  C. 4. SG-CAA media for inducing galactose-inducible pCT302 construct: Replace dextrose with equal amount of galactose in the SD-CAA recipe to make SG-CAA. Sterile filter through a 0.22 μm membrane, and store at 4  C. 5. For SD-CAA plates: Dissolve 91.1 g sorbitol, 7.4 g sodium citrate dihydrate, 2.1 g citric acid monohydrate in 400 mL ddH2O. Add 7.5 g bacto-agar and sterilize by autoclaving. Allow the medium to cool to 50–60  C. In a separate container, dissolve 2.5 g casamino acids, 10 g dextrose, and 3.35 g yeast nitrogen base without amino acids in ddH2O. Adjust final volume of the solution to 100 mL, sterile filter through a 0.22 μm membrane, and add to cooled agar-containing medium. Mix by gentle swirling to avoid bubbles, and pour ~25 mL in 100 mm  15 mm plates in a laminar flow hood. Cool and store at 4  C.

2.7 Reagents for Transforming Yeast with Plasmid DNA by Heat Shock

1. 50% w/v PEG 3350 (sterile filter through a 0.22 μm membrane) (see Note 1). 2. 1 M Lithium acetate dihydrate (sterile filter through a 0.22 μm membrane). 3. Single stranded carrier DNA: 2 mg/mL salmon sperm DNA in 1 TE buffer.

Deep Mutational Scanning of T Cell Receptors

1. Sterile ddH2O.

2.8 Reagents for Preparation of Electrocompetent Yeast for Making Libraries

3. 0.1 M Lithium acetate with 10 mM DTT.

2.9

1. 1:1 mixture of 1 M sorbitol and YPD medium.

Electroporation

121

2. 1 M sorbitol with 1 mM CaCl2. 4. 1 M sorbitol.

2. 0.2 cm electroporation cuvettes. 3. Electroporator (Biorad Gene Pulser II or equivalent). 4. Temperature controlled orbital shaker. 2.10 Flow Cytometry and Sorting

1. Anti-HA antibody (for example, clone 16B12 anti-HA antibody, mouse IgG1, Covance). 2. Anti-c-myc antibody (for example, clone 2G8D5 biotinylated anti-c-myc antibody, mouse IgG2A, GenScript or FITC conjugated, chicken anti-c-myc antibody, ICL Inc., or Alexa Fluor® 647 conjugated mouse anti-myc-tag, clone 9B11, Cell Signaling Technology). 3. Fluorophore-linked secondary antibodies for unlabeled primary antibodies indicated above (for example, F(ab0 )2-Goat anti-Mouse IgG (H + L) Secondary Antibody, Alexa Fluor 647, ThermoFisher). 4. Ligand for the protein of interest (for example, peptide-MHC complex for TCRs as biotinylated monomers or fluorescent tetramers). 5. Fluorophore-linked streptavidin (for example, PE Streptavidin, BD Biosciences). 6. Conformation-specific, monoclonal antibodies for the protein of interest and appropriate fluorophore-linked secondary antibody. 7. PBS containing 1% BSA (sterile filter through a 0.22 μm membrane). 8. Flow cytometer and sorter.

2.11 Illumina Deep Sequencing

1. Zymoprep Yeast plasmid miniprep II (Zymo Research). 2. QIAprep Spin miniprep kit (QIAGEN). 3. Exonuclease I (E. coli) (NEB). 4. Lambda exonuclease (NEB). 5. Illumina Miseq sequencing primers (Table 2). 6. Primers to anneal end sequences that anneal to Illumina Miseq sequencing primers (Table 2). 7. Primers to add adaptors for annealing to flowcell (Table 2). 8. Primers to add indexes or barcodes for unique identification (Table 2).

122

Preeti Sharma et al.

Table 2 Primers for Illumina deep sequencinga

a

In red and green are the sequences that anneal to the Illumina flow cell cluster oligonucleotides. Various unique index or barcode sequences are colored in blue.

9. Qubit 2.0 fluorimeter to quantify PCR-amplified DNA. 10. Qubit dsDNA HS Assay Kit. 11. MiSeq Sequencer.

Deep Mutational Scanning of T Cell Receptors

2.12 Software for Deep Sequencing Data Analysis

3

123

Enrich software [14, 15].

Methods

3.1 Designing the Construct for Yeast Surface Display of a Protein of Interest

As a first step to engineer any protein by yeast surface display, the cDNA encoding the protein of interest is cloned in the NheI and XhoI restriction sites in the yeast display vector, pCT302, to allow protein expression on the cell surface (Fig. 1) [3, 4, 16–19]. This cloning results in an in-frame fusion of the protein of interest with the yeast mating agglutinin surface protein, Aga2p, at the N-terminus. The Aga2p fusion partner serves to link the protein of interest to the yeast cell wall, where it is surface-exposed for detection. The construct also contains a HA tag (amino acid sequence: YPYDVPDYA) downstream of Aga2p, followed by a Gly-Ser linker ((G4S)3) and NheI-XhoI restriction sites for cloning the gene of interest. The gene encoding the protein can be commercially synthesized (e.g., from GenScript or IDT) and codon-optimized for expression in yeast, although we have not found codonoptimization to be critical. Alternatively, the gene can be subcloned from a plasmid carrying the cDNA. During gene synthesis (or subcloning using primers with overhangs), a C-terminal c-myc tag (amino acid sequence: EQKLISEEDL) followed by two stop codons should be included in frame. This ensures that mutations in future libraries causing truncations can be eliminated by selecting for the presence of the C-terminal c-myc tag. Finally, flanking DNA sequences containing an in-frame 50 NheI site and a 30 XhoI site should be added to allow subcloning into the yeast display vector pCT302. To illustrate with the example of a single-chain TCR, the gene inserted between the NheI and XhoI sites of pCT302 encodes, from the 50 to 30 end: the variable domain of the TCR alpha chain, a flexible connecting linker, the variable domain of the TCR beta chain, a c-myc epitope tag, and two stop codons. This leads to the translation of a single-chain polypeptide that is displayed via the Aga2p fusion on the yeast surface.

3.2 Transformation of EBY100 with Yeast Display Plasmid Containing the Gene of Interest

Once the gene encoding the protein of interest is cloned in pCT302, the plasmid can be transformed in the yeast strain EBY100 by lithium acetate mediated heat shock method or by electroporation (as described in Subheading 3.4.6). The yeast display plasmid contains the TRP1 auxotrophic marker, which allows EBY100 cells (TrpLeu) containing pCT302 to grow on selective medium (SD-CAA and SG-CAA) missing tryptophan. The procedure to transform EBY100 with pCT302 plasmid by heat shock method is described below (adapted from [20–22]).

124

Preeti Sharma et al.

1. Streak a YPD plate with a frozen glycerol stock of EBY100 in a laminar flow hood. Incubate the plate in a 30  C incubator for 36–48 h until colonies are visible (see Notes 2 and 3). 2. Inoculate 2–3 mL YPD medium in a sterile test tube with a single colony of EBY100. Incubate the tube in a 30  C orbital shaker for 48 h at 220 rpm until a thick culture is obtained. 3. For a single transformation, transfer 1 mL yeast culture in a sterile 1.7 mL microfuge tube, and pellet cells using a tabletop centrifuge by centrifuging at top speed (18,000  g) for 1 min. Aspirate the supernatant. 4. Resuspend the pelleted cells in the following reagents in the order listed below: (a) 240 μL 50% w/v PEG 3350. (b) 36 μL 1 M Lithium acetate dihydrate. (c) 50 μL boiled single stranded carrier DNA (2 mg/mL) (boil for 2–3 min, chill on ice, vortex before use) (see Note 4). (d) 0.1–1 μg plasmid DNA plus sterile water in a final volume of 34 μL. 5. Incubate the 1.7 mL microfuge tube in a 42  C water bath for 2–3 h. 6. Centrifuge the tube at 18,000  g for 1 min. Aspirate the supernatant. 7. Resuspend the cells gently in 1 mL sterile water, and plate 100 μL on SD-CAA plates (TRP selection). Incubate the plate in a 30  C incubator for 36–48 h until colonies are visible (see Notes 3 and 5). 3.3 Analysis of Yeast Cells Transformed with the Gene of Interest 3.3.1 Sequencing Analysis

1. Inoculate a single colony from SD-CAA plate in 3 mL SD-CAA medium in a sterile test tube. Incubate the tube in a 30  C orbital shaker for ~36–48 h at 220 rpm (until a dense culture is obtained). 2. To confirm the sequence of the transformed construct, isolate plasmid DNA from yeast cells using the Zymoprep plasmid miniprep II kit following manufacturer’s instructions (see Note 6). 3. Transform 25–50 μL DH5α cells with 2.5–5 μL zymoprep DNA following manufacturer’s instructions (see Note 7). 4. Inoculate a single DH5α colony in 3 mL LB medium containing 100 μg/mL Ampicillin in a sterile test tube. Incubate the tube in a 37  C orbital shaker overnight (16 h) at 220 rpm. 5. Isolate plasmid DNA using QIAprep Spin miniprep kit following manufacturer’s instructions. Elute DNA in molecular biology grade water and sequence by Sanger sequencing using the Splice 4L and YRS primers (Fig. 1a).

Deep Mutational Scanning of T Cell Receptors

125

6. After confirming sequences, the SD-CAA cultures can be frozen in SD-CAA + 10% DMSO at 80  C and used for future subculturing. 3.3.2 Flow Cytometric Analysis

Once the sequence of plasmid DNA transformed in EBY100 is confirmed, the yeast cells can be assessed for surface expression of the displayed protein of interest, in addition to the N- and C-terminal HA and c-myc tags, respectively, by flow cytometry (Fig. 1b). A general procedure for inducing surface expression and flow cytometry is described in the following. 1. Measure the density of the SD-CAA culture by measuring absorbance at 600 nm (OD600). An OD600 ¼ 1 is equivalent to a concentration of 107 cells/mL. Yeast cells are typically induced at a concentration of 107 cells/mL. For a test induction, induce 3  107 cells. 2. Transfer SD-CAA culture containing 3  107 cells in a sterile microfuge tube. Centrifuge at 1800  g for 3 min at 4  C, aspirate supernatant. 3. To wash cells, resuspend cells in 1 mL cold SG-CAA medium and centrifuge at 1800  g for 3 min at 4  C. Aspirate supernatant. 4. Wash cells once more as described in step 3 above. 5. Resuspend cells in 1 mL SG-CAA, and transfer to a sterile test tube. Add 2 mL SG-CAA to achieve a final concentration of 107 cells/mL. 6. Incubate cultures in a 20  C orbital shaker for 48 h at 220 rpm. 7. For flow cytometry, aliquot 50 μL cells in flow tubes. 8. To wash cells, add 1 mL PBS + 1% BSA, centrifuge at 1800  g for 5 min at 4  C, and aspirate supernatant. 9. Resuspend cells in 50 μL primary antibody solution (for example, PBS + 1% BSA containing 1–10 μg/mL of anti-HA or antic-myc or conformation-specific antibody or ligand for your protein of interest). Incubate on ice for 30–60 min with occasional mixing. Alternatively, tubes can be incubated on a rocker, placed in cold room (see Note 8). Also, set up an unstained control tube and secondary antibody only control tube to aid with setting gates during data acquisition. A sample of uninduced yeast may also help as an additional control for monitoring non-specific binding of antibodies. 10. Wash cells twice as described in step 8 above. 11. Resuspend cells in 50 μL secondary antibody solution (for example, fluorophore-linked anti-mouse antibody for a mouse primary antibody, or streptavidin-PE for a biotinylated primary reagent). Incubate on ice for 30–60 min.

126

Preeti Sharma et al.

Fig. 2 Yeast surface display in combination with deep mutational scanning to design protein variants. (a) A schematic of SOE PCR for making single codon libraries. To introduce variability in a single codon, a forward primer with a single degenerate codon (NNK, shown as *) is used in combination with T7 reverse primer in a Pre-SOE PCR. In a second Pre-SOE PCR, Splice4L is used in combination with a reverse primer that has ~20–25 bases overlap with the forward degenerate primer. The products of the two Pre-SOE PCRs can hybridize via their overlapping regions, and the entire gene containing single codon libraries at various locations can be amplified by using Splice4L and T7 reverse primers. (b) Single codon library PCR from (a) is mixed with cut vector (pCT302) and transformed into yeast cells to generate libraries in yeast. Yeast cells induced with galactose in culture medium can be selected for a desired property (for example, high binding affinity to ligand). Sorted and unsorted cells are expanded, DNA is isolated and prepared for deep sequencing. Deep sequencing data can then be used to guide further mutagenesis strategies to isolate lead mutants. As an example, isolation of a higher affinity TCR for binding to the human cancer antigen MART-1/HLA-A2 is shown. (Representative data from Sharma et al., JBC, 2018 are shown with permission from The Journal of Biological Chemistry (JBC), Elsevier)

12. Wash cells twice as described in step 8 above. 13. Resuspend cells in 200–500 μL PBS + 1% BSA and acquire fluorescence data on a flow cytometer (see Note 9). 3.4 Construction of Single Codon Libraries

Single codon libraries in the protein of interest can be created by splice overlap extension (SOE) PCR, randomizing one residue at a time to generate diversity (Fig. 2a). All codons within the gene sequence can be sequentially replaced with degenerate codons in a comprehensive mutational scan, or alternatively SOE PCR-based mutagenesis can be targeted to residues of interest. For example, when engineering TCRs, it may be desirable to focus mutagenesis to the antigen-binding loops [11–13]. SOE PCR-based mutagenesis is a two-step PCR method that is used to prepare two “pre-

Deep Mutational Scanning of T Cell Receptors

127

SOE” PCR products with overlapping overhangs in the first PCR, which hybridize to create the complete “SOE” PCR product in the second PCR. A mixture of the purified PCR products containing mutations in a single codon (or, insert), and linearized pCT302 vector (digested with NheI and XhoI), are transformed into the yeast cells allowing the vector and insert sequences to recombine by homologous recombination, thus creating a yeast single codon library (Fig. 2b). Various steps of this process are described below. 3.4.1 Primer Design

To construct a single codon library at every residue in the protein, a set of forward primers is designed that replaces one codon at a time with a degenerate codon (i.e., NNK codon, where N is any nucleotide (A/C/T/G) and K is G/T) thus allowing 19 other residues to be encoded, in addition to the wild-type residue at the target location (see Note 10) [8]. The forward primer contains a 20–30 nucleotide overlap, corresponding to melting temperatures of 52–54  C, with the template DNA (pCT302 containing the gene of interest) on both sides (i.e., upstream and downstream) of the degenerate codon. In addition, a set of reverse primers is designed that anneals to 20–30 nucleotides (again corresponding to melting temperatures of 52–54  C) of the template DNA immediately upstream of the degenerate codon (Table 1). Obtain synthesized primers in 96-well plates from the vendor and dilute to 10 μM with molecular biology grade water in a new 96-well deep well plate.

Table 1 An example of primer design to make single codon libraries in 15 contiguous residues (NNK ¼ degerate codon, nucleotide positions in red)

128

Preeti Sharma et al.

3.4.2 Pre-SOE PCR

1. Set up the following Pre-SOE PCR reactions in two 96-well plates: Pre-SOE 1 (generates 30 fragment for the SOE PCR): (a) 10 μL Phusion high-fidelity buffer (5). (b) 50 ng pCT302 DNA containing the gene of interest. (c) 5 μL Forward degenerate primer (10 μM). (d) 3 μL T7 reverse primer (10 μM). (e) 1.25 μL dNTP mix (10 mM each). (f) 0.5 μL Phusion high-fidelity DNA polymerase (2 U/μL). (g) Molecular biology grade water (to 50 μL). Pre-SOE 2 (generates 50 fragment for the SOE PCR): (a) 10 μL Phusion high-fidelity buffer (5). (b) 50 ng pCT302 DNA containing the gene of interest. (c) 5 μL Reverse primer (10 μM). (d) 3 μL Splice4L forward primer (10 μM). (e) 1.25 μL dNTP mix (10 mM each). (f) 0.5 μL Phusion high-fidelity DNA polymerase (2 U/μL). (g) Molecular biology grade water (to 50 μL). 2. Place the pre-SOE reactions in a thermocycler and run the following program: (a) 98  C for 20 s. (b) 20 cycles of: l

98  C for 10 s.

l

56  C for 10 s.

l

72  C for 40 s.

(c) 72  C for 5 min. (d) 4  C forever. 3. Load 5 μL PCR products on a 1% agarose gel to confirm their sizes (see Note 11). 3.4.3 SOE PCR

1. Set up the following SOE PCRs in 96-well plates: (a) 10 μL Phusion high-fidelity buffer (5). (b) 5 μL 1:10 diluted Pre-SOE PCR 1 product. (c) 5 μL 1:10 diluted Pre-SOE PCR 2 product. (d) 3 μL Splice4L forward primer (10 μM). (e) 3 μL T7 reverse primer (10 μM). (f) 1.25 μL dNTP mix (10 mM each).

Deep Mutational Scanning of T Cell Receptors

129

(g) 0.5 μL Phusion high-fidelity DNA polymerase (2 U/μL). (h) Molecular biology grade water (to 50 μL). 2. Place the SOE reactions in a thermocycler and run the following program: (a) 98  C for 20 s. (b) 20 cycles of: l

98  C for 10 s.

l

56  C for 10 s.

l

72  C for 40 s.

(c) 72  C for 5 min. (d) 4  C forever. 3. Load 5 μL PCR products on a 1% agarose gel to confirm sizes. Each SOE PCR will generate a major product amplified from the Splice 4L forward and T7 reverse primers corresponding to the length of the insert between NheI and XhoI sites plus an additional ~500 bp of vector sequence. 4. Pool approximately equal amount of each SOE product containing individual single codon library, and purify the full, amplified gene by QIAquick Gel Extraction Kit following manufacturer’s instructions. Elute DNA in molecular biology grade water (see Note 12). 3.4.4 Preparation of Cut Vector

Digest pCT302 vector with NheI and XhoI enzymes following manufacturer’s instructions. After NheI, XhoI digestions, dephosphorylate the linearized vector using calf intestinal phosphatase (CIP). Purify cut vector by QIAquick Gel Extraction Kit following manufacturer’s instructions. Elute DNA in molecular biology grade water (see Note 13).

3.4.5 Precipitate Vector and Insert

Mix 4 μg pooled, purified SOE PCR product with 1 μg digested vector DNA in eight to ten 1.7 mL microfuge tubes. Precipitate DNA as colored pellets using Pellet Paint Co-Precipitant (Novagen) following manufacturer’s instructions. In addition, in two separate tubes, precipitate 4 μg SOE PCR product or 1 μg digested vector DNA to use as “Insert only” and “Vector only” controls. Store pelleted DNA at 20  C until ready to transform into yeast.

3.4.6 Preparation of Electrocompetent Yeast and Transformation of Library DNA

1. Two days before preparing the yeast library, inoculate 3 mL YPD with single colony of EBY100 in a sterile test tube. Incubate in 30  C orbital shaker overnight at 220 rpm. 2. Subculture 3 mL overnight culture into 50 mL YPD in a 250 mL Erlenmeyer flask. Incubate in 30  C orbital shaker overnight at 220 rpm.

130

Preeti Sharma et al.

3. The next morning, inoculate prewarmed YPD (400 mL in 2 L flask) to an OD600 of 0.3 from overnight cultures. Incubate in 30  C orbital shaker at 220 rpm until OD600 of 1.3–1.6 is achieved. 4. Transfer 50 mL culture into 50 mL conical tubes in laminar flow hood. Open and close the tubes inside the hood to maintain sterility. Each tube will be used for one electroporation. Pellet cells by centrifuging cultures at 2000  g for 5 min at 4  C. Discard supernatant. Ensure that all reagents and cells remain on ice during the procedure, except when indicated otherwise. 5. Resuspend cells in each conical tube with 25 mL cold, sterile ddH2O. Pellet cells by centrifuging cultures at 2000  g for 5 min at 4  C. Discard supernatant. 6. Resuspend cells in each conical tube with 25 mL cold, sterile 1 M sorbitol with 1 mM CaCl2. Pellet cells by centrifuging cultures at 2000  g for 5 min at 4  C. Discard supernatant. 7. Resuspend cells in each conical tube with 25 mL 0.1 M lithium acetate with 10 mM DTT. Since the solution is labile, prepare lithium acetate-DTT right before use. Pool resuspended cells from various tubes in a single autoclaved 1 L Erlenmeyer flask. Incubate in 30  C orbital shaker at 220 rpm for 30 min. 8. After 30 min, pellet 25 mL culture by centrifuging in 50 mL conical tubes at 2000  g for 5 min at 4  C. Discard supernatant. 9. Resuspend cells in each conical tube with 25 mL cold, sterile 1 M sorbitol with 1 mM CaCl2. Pellet cells by centrifuging cultures at 2000  g for 5 min at 4  C. Discard supernatant. 10. Wash cells in each conical tube with 25 mL cold, sterile 1 M sorbitol (no CaCl2). Pellet cells by centrifuging cultures at 2000  g for 5 min at 4  C. Discard supernatant. 11. Add ~100 μL 1 M sorbitol (no CaCl2) to resuspend cells in each tube. Pool cells in a single tube. Make up the volume such that 100–200 μL is available for each electroporation (see Note 14). Aliquot 100–200 μL cell suspension (per electroporation) to a prechilled 0.2 cm electroporation cuvette on ice. Typically, we conduct 8–10 electroporations for the library (“vector + insert”), and single electroporation each for “vector only,” “insert only,” and “cells only” controls. 12. Resuspended precipitated DNA (“vector + insert” or “vector only” or “insert only” from Subheading 3.4.5) in 10 μL water. In addition, also set up a tube for “cells only” control. 13. Add DNA to yeast cells in the prechilled cuvette. Incubate on ice for 5 min. Using Biorad Gene Pulser II, electroporate cells

Deep Mutational Scanning of T Cell Receptors

131

with DNA at 2.5 kV, 25 μF capacitance. Typical time constants range between 3 and 4.5 ms. 14. Following the electric pulse, recover the electroporated cells by adding 1 mL 1:1 mixture of 1 M sorbitol and YPD medium to the cuvette. Transfer the cells to a sterile test tube. Add an additional 1 mL 1:1 mixture of 1 M sorbitol and YPD medium to the cuvette to collect any residual cells and transfer to the sterile test tube. Incubate tubes in 30  C orbital shaker without shaking for 1 h. 15. Pool eight to ten electroporations representing the library (“vector + insert”) in a single conical tube. Pellet cells by centrifuging at 900  g for 5 min. Resuspend cells in 10 mL SD-CAA medium. Prepare small volumes (~50–100 μL) at various dilutions (1:10, 1:102, 1: 103, 1: 104, 1: 105, 1: 106) and plate 10 μL of each dilution on SD-CAA plates to determine the library size. Incubate the plates in a 30  C incubator for 36–48 h until colonies are visible (see Note 15). 16. Similarly, transfer control electroporations to conical tubes, pellet cells, and resuspend in 10 mL SD-CAA. Plate 10 μL of undiluted culture on SD-CAA plate to determine any background (i.e., colonies that resulted from uncut vector DNA). Incubate the plates in a 30  C incubator for 36–48 h until colonies are visible. Remaining control cultures can be discarded. 17. Transfer the remaining library culture into 0.5–1 L SD-CAA media containing penicillin-streptomycin, and expand for 36–48 h in 30  C orbital shaker at 220 rpm until saturation is achieved (OD600 > 8). Yeast libraries in SD-CAA can be stored at 4  C for several months. Expand sufficient number of cells in fresh medium to oversample the library before sorting. Alternatively, make several aliquots oversampling library diversity, freeze in SD-CAA containing 10% DMSO, and store at 80  C. 18. Diversity in the libraries can be confirmed by expanding colonies from step 15 in 3 mL SD-CAA medium, isolating plasmid DNA and sequencing as described in Subheading 3.3.1 (see Note 16). 19. In addition, the phenotype of the libraries can be measured by assessing surface expression by staining with conformationspecific antibodies for your protein of interest or anti-c-myc antibody as described in Subheading 3.3.2. Alternatively, if a fluorescent ligand is available, the binding profile of the library can be determined by flow cytometry compared to the wildtype protein.

132

3.5

Preeti Sharma et al.

Sorting Libraries

Based on the characteristic of interest (for example, high stability or protein expression, high binding affinity to the ligand, high thermostability, etc.) [4, 11–13, 23–27], selection pressures can be applied to the library to select mutations that would impart such characteristics (Fig. 2b). To do this, it is important to oversample the diversity contained in the libraries to reduce the chances of losing mutants due to under sampling, i.e., if the theoretical size of your library is 106 unique mutations, subject at least tenfold more yeast cells (i.e., 107 cells) to selection. A general procedure to sort libraries is listed below: 1. Expand an aliquot of your yeast library (for example, 108 cells) in fresh SD-CAA medium at 107 cells/mL. Incubate the culture for 36–48 h in 30  C orbital shaker at 220 rpm. 2. Induce cultures in SG-CAA medium for 36–48 h in 20  C orbital shaker at 220 rpm, as described in Subheading 3.3.2. 3. Wash cells in PBS + 1% BSA and stain (for example, 108 cells) with at least 1 mL primary staining solution to select for a certain property by flow cytometry, as described in Subheading 3.3.2 [17–19]. For example, if approximate dissociation constant of the wild-type protein for binding to its ligand is known, and one is interested in isolating higher affinity mutants, one can use a concentration of a biotinylated ligand that is five- to tenfold lower than the dissociation constant to isolate higher affinity mutants (see Note 17). If the ligand is monomeric, it is most important to ensure that reagent excess is maintained compared to the number of protein molecules expressed on the yeast cell surface, otherwise low affinity binders can be lost at sub-saturating ligand concentrations. When using ligands with two or more binding sites (for example, bivalent antibodies or tetramers of peptide-MHC), avidity of the interaction facilitates the selection of low affinity binders. As another example, if one is interested in isolating thermostable variants, then the yeast library can be subjected to a thermal stress (for example, incubating at 45  C for 30 min), followed by staining with a conformation-specific primary reagent. 4. After the primary stain, wash cells with excess PBS + 1% BSA (20-fold volume of primary reagent) to remove unbound reagent. 5. Bound ligand can be detected by staining cells with a fluorescent, secondary reagent specific for the primary reagent. 6. After the completion of secondary staining, wash cells by adding large excess of PBS + 1% BSA solution, and resuspend in an appropriate volume for sorting. Be sure to include unstained control and secondary antibody only control samples to set gates during data acquisition. When multiple fluorescent

Deep Mutational Scanning of T Cell Receptors

133

probes are used, controls should include positive samples stained with each fluorescent probe individually to set compensation on the FACS instrument. 7. Run samples on a FACS instrument to first set up a gate on live, healthy cells followed by singlets within that population. Finally, observe the staining of your stained library with the fluorescent reagent to collect cells that exhibit a desired property (see Note 18). Collect at least 50,000 cells and expand in SD-CAA medium in a 30  C orbital shaker at 220 rpm to desired density. Store cultures at 4  C or freeze in SD-CAA containing 10% DMSO and store at 80  C. 3.6

Deep Sequencing

3.6.1 Preparation of DNA for Deep Sequencing

This procedure uses both Zymoprep Yeast plasmid miniprep II and QIAprep Spin miniprep kits [8]. 1. Pellet 1  107 to 4  107 cells from unsorted and sorted libraries in a 1.7 mL microfuge tube. 2. Resuspend cells in 200 μL Solution 1 (from Zymoprep kit). Add 5 μL Zymolyase (5 U/μL). Incubate at 37  C for 4 h, mixing the cells every 1 h. 3. Perform 3–4 freeze-thaw cycles in dry ice-ethanol bath or liquid nitrogen, and 42  C water bath, to facilitate lysis. 4. Add 200 μL Solution 2 (from Zymoprep kit), mix the contents, and incubate at room temperature for 3–5 min. 5. Add 400 μL Solution 3 (from Zymoprep kit), mix the contents and centrifuge at top speed (18,000  g) on a tabletop microfuge for 5 min. 6. Transfer supernatant from previous step to Qiagen miniprep spin column and centrifuge at 18,000  g for 1 min. 7. Wash column by adding 700 μL PB buffer (from Qiagen miniprep kit), and centrifuging at 18,000  g for 30 s. 8. Wash column by adding 700 μL PE buffer (from Qiagen miniprep kit), and centrifuging at 18,000  g for 30 s. Wash the column a second time with the PE buffer. 9. Dry the column, by emptying out the flow through and spinning the column at 18,000  g for 30 s. 10. Transfer the column to a clean microfuge tube. Add 30–35 μL molecular biology grade H2O to the column, and let it stand at room temperature for 1 min. Elute DNA by centrifuging at 18,000  g for 1 min. Reload the column with the eluate and elute a second time by centrifugation. At this step, the DNA can be stored at 80  C. 11. Optional: Genomic DNA fragments may interfere with sample preparation for Illumina sequencing and can be removed with exonucleases. To remove genomic DNA fragments from the DNA prep, set up the following reaction:

134

Preeti Sharma et al.

(a) 15 μL DNA from step 10. (b) 2 μL Exonuclease I. (c) 1 μL Lambda Exonuclease. (d) 2 μL 10 Lambda Exonuclease Reaction Buffer. Incubate the reaction in a thermocycler for 90 min at 30  C. Purify the DNA using QIAquick PCR Purification Kit following manufacturer’s instructions. Elute DNA in molecular biology grade water. 3.6.2 Design Primers to PCR Amplify Mutated Regions of the Gene for Paired End Illumina Sequencing

There are multiple strategies for deep sequencing the libraries and determining the relative enrichment or depletion of variants. Popular sequencing methods on the Illumina NovaSeq 6000 or MiSeq instruments generally do not go beyond 2  250 nt read lengths using a paired end protocol. For libraries where the mutated codons are confined within a stretch of not more than 200 bp, a 2  250 nt paired end protocol will fully cover the relevant region of the gene. Otherwise, the diversified region/s will need to be sequenced as a set of fragments that together provide full coverage. Note that when separate fragments are sequenced independently to provide full coverage, there is no connecting information that would allow the full-length sequence to be reconstructed. If a mutation is observed in a sequenced fragment, it can be assumed that the rest of the sequence is likely wild-type due to mutagenesis being targeted to single codons. If no mutation is observed in a sequenced fragment, then it is unknown whether a mutation might have resided elsewhere in the gene. Accordingly, the frequency of the wild-type sequence is not accurately measured when the library diversity covers a region longer than the Illumina read length. In practice, this has minimal effect on protein engineering problems when one is searching for single mutations with high activity, but it may be an important consideration in other kinds of experiments. In such cases, readers are referred elsewhere for alternative strategies [28, 29]. Preparation of the DNA for Illumina sequencing requires amplifying the gene as one or more fragments of ~250 bp length. The amplification primers add 50 and 30 ends to the PCR product for compatibility with Illumina sequencing kits. This is accomplished in two steps. First, gene-specific primers amplify the relevant region and add annealing sites for the Illumina sequencing primers. Second, a universal set of primers add annealing sites for the Illumina flow cell and barcodes for cataloging each DNA sample (e.g., naive library prior to FACS, sort condition 1, sort condition 2, etc.).

Deep Mutational Scanning of T Cell Receptors

135

1. For the first round of PCR, design primers to add end sequences to the library DNA for annealing to the Illumina sequencing primers (Table 2). (a) To design the gene-specific forward primer, add a 20–25 nt long sequence (melting temperature 51–53  C) that anneals to the target gene upstream of the first mutated codon, to the 30 end of the Read 1 sequencing primer i.e., 50 TCTTTCCCTACACGACGCTCTTCCGATCT_gene-specific sequence 30 . (b) Similarly, to design the gene-specific reverse primer, add the reverse complement of a 20–25 nt long sequence downstream of the last mutated codon in the target gene to the 30 end of the Read 2 sequencing primer i.e., 50 GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT_gene-specific sequence 30 . Again, the unique sequence that anneals to the target DNA should have a melting temperature of 51–53  C. (c) For a 2  250 nt paired end read protocol on an Illumina sequencer, the gene-specific forward and reverse primer binding sites should be separated by 250 bp. 2. For the second round of PCR, order the following primers: (a) Synthesize Illumina start adaptamer primer and a reverse primer (Table 2). These add adaptor sequences to the PCR product from the first PCR. The adaptor sequences are homologous to the capture oligonucleotides on the flow cell surface during the Illumina sequencing reaction. The adaptor sequences hence allow annealing of the PCR amplicons to the flow cell cluster oligonucleotides, which then undergo bridge amplification to generate clusters on the flow cell for Illumina sequencing. (b) The reverse adaptamer primers also provide a choice of unique indexes or barcode sequences that allow experiment identification. Different indexes can be used for unsorted and sorted libraries. This allows for many samples to be pooled together on a single Illumina sequencing run. 3.6.3 PCR to Add Flanking Sequences to DNA for Illumina Deep Sequencing

1. For PCR 1, set up the following reactions to add sequences that are complementary to Illumina sequencing primers: (a) 4 μL Phusion high-fidelity buffer (5). (b) 0.4 μL dNTP mix (10 mM each). (c) 1 μL gene-specific forward primer (10 μM) (from Subheading 3.6.2, step 1). (d) 1 μL gene-specific reverse primer (10 μM) (from Subheading 3.6.2, step 1).

136

Preeti Sharma et al.

(e) 0.2 μL Phusion high-fidelity DNA polymerase (2 U/μL). (f) 13.4 μL DNA isolated from unsorted or sorted libraries (from Subheading 3.6.1). Place the reactions in a thermocycler and run the following program (see Note 19): (a) 98  C for 20 s. (b) 12 cycles of: l

98  C for 10 s.

l

60  C for 10 s.

l

72  C for 40 s.

(c) 72  C for 5 min. (d) 4  C hold. 2. For PCR 2, set up the following reactions to add adaptors and index sequences: (a) 10 μL Phusion high-fidelity buffer (5). (b) 1.25 μL dNTP mix (10 mM each). (c) 1 μL Illumina_Start_Adaptamer primer (10 μM) (from Subheading 3.6.2, step 2). (d) 1 μL Illumina_Index_1/2/3. . .11/12_Adaptamer primer (10 μM) (from Subheading 3.6.2, step 2) (use unique index primers for unsorted and sorted libraries). (e) 0.5 μL Phusion high-fidelity DNA polymerase (2 U/μL). (f) 1 μL PCR product from PCR 1. (g) Molecular biology grade water (to 50 μL). Place the reactions in a thermocycler and run the following program: (a) 98  C for 20 s. (b) 12 cycles of: l

98  C for 10 s.

l

60  C for 10 s.

l

72  C for 40 s.

(c) 72  C for 5 min. (d) 4  C forever. 3. Load 5 μL PCR 2 product on a 1–1.5% agarose gel to confirm successful amplification and PCR product sizes. Purify remaining PCR product by QIAquick Gel Extraction Kit following manufacturer’s instructions. Elute DNA in molecular biology grade water.

Deep Mutational Scanning of T Cell Receptors

137

4. Quantify DNA using Qubit 2.0 fluorimeter following manufacturer’s instructions. 5. Pool PCR-amplified, purified unsorted and sorted library DNA in an equimolar ratio, to a final concentration of 10 nM (or the preferred concentration of the sequencing vendor). Submit DNA sample for Illumina sequencing, e.g. paired end 2  250 nt deep sequencing, on a MiSeq, HiSeq, or NovaSeq instrument. 3.7 Deep Sequencing Data Analysis with Enrich

Deep sequencing runs on an Illumina sequencer can yield 20 million (MiSeq V3) to over 800 million (NovaSeq 6000) paired reads, or even higher if the read length is less than 250 nt. Sequences are sorted by index using instrument software and provided as FASTQ files, in which each read is recorded with a unique identifier and with quality scores to match the nucleotide sequence. To parse through this data, we use the open-source software Enrich [5, 14, 15], developed by Dr. Doug Fowler at University of Washington. The software package runs on a Python environment and is available for installation from http://depts.washington.edu/sfields/ software/enrich/. On the website, the authors of Enrich give detailed documentation on the use of their software, and we therefore only provide a conceptual outline of the processing steps. Enrich has been updated to Enrich2 [15] which is available on GitHub at https://github.com/FowlerLab/Enrich2/. The main updates in Enrich2 are algorithms for processing multiple deep sequencing data sets at different levels of selection stringency. For FACS based selections of yeast displayed libraries, these data sets may correspond to different collection gates (e.g. yeast can be sorted into different bins based on low to high fluorescence signal) or to separate sorts in which yeast are incubated with increasing ligand concentrations, similar to a titration experiment. By analyzing data from multiple bins or selections of varying stringency, the effects of a mutation on protein expression/activity can be quantitatively determined with some level of statistical confidence. However, for many engineering purposes, data from a small number of FACS experiments is sufficient to qualitatively determine the mutations of highest activity in the yeast displayed library. Enrich software processes the FASTQ files in a series of steps as follows: 1. For paired reads, the read 1 (R1) and read 2 (R2) FASTQ files need to be “fused” using the Fuser module, whereby the region of overlap between the R1 and R2 sequences is aligned and compared. When R1 and the reverse complement of R2 are mismatched, the nucleotide of higher quality score is chosen. This reduces the rate of sequencing errors. The final output is a list of sequences that correspond to the fused R1 and R2 reads.

138

Preeti Sharma et al.

2. The Aligner module uses the output file from the Fuser module to align each sequence and its translation with the reference (i.e. wild-type) nucleotide and amino acid sequences, respectively. The output of the Aligner module lists the mutations present in each sequence. 3. Next, using the Map Counts module, Enrich identifies and enumerates unique sequences. That is, all sequences with the same mutation (same substitution at the same position) are summed to find their total frequency in the library. 4. FASTQ files from the unsorted/naive and sorted libraries are processed independently using steps 1–3 above. Now, using the Map Ratios module, the Map Counts output files are compared. Enrich calculates the enrichment ratio of each unique sequence by comparing the frequencies with which they occur in the unsorted/naive and sorted libraries (i.e., freqsorted/frequnsorted). 5. Following this, the Map Ratios output file can be parsed in different ways to visualize the data (refer to online documentation for Enrich and Enrich2 for different visualization tools). In our group, python scripts are used to exclude sequences with more than one mutation (these should be a small fraction of the library as a result of PCR errors and sequencing errors) and to arrange the list of enrichment ratios as a table based on amino acid substitution at each sequential position. The data can then be easily inspected and manipulated in Microsoft Excel, and tables can be plotted as color-coded heat maps using gnuplot (http://www.gnuplot.info/). For instance, in the example data shown (Fig. 2b), blue color represents enrichment of a mutation in the sorted library compared to its sequence frequency in the unsorted library (i.e. a positive enrichment ratio on a log2 scale). Similarly, orange color represents depletion of a mutation in the sorted library compared to the unsorted library (i.e. a negative log2 enrichment ratio). White color indicates that the mutation occurs at the same frequency in the sorted and unsorted libraries. The intensity of the orange and blue colors represents the magnitude of the mutation’s depletion or enrichment on a log2 scale. 3.8 Targeted Mutagenesis Based on Deep Mutagenesis Data

The large amount of mutational data generated by deep mutational scanning provides a comprehensive snapshot of sequence–function relationships for single amino acid substitutions. Mutations with negative log2 enrichment ratios provide insights about residues that are most important for overall stability and folding of the protein, and also those that upon being mutated negatively impact the desired property of interest. The substitutions with high, positive log2 enrichment ratios can be used to guide targeted mutagenesis to yield candidates with improved properties. For example, this

Deep Mutational Scanning of T Cell Receptors

139

strategy can be used to engineer T cell receptors with higher affinity and/or stability by introducing single mutations with high enrichment (Fig. 2b) [11, 13], that can then be combined to yield additional improvements in affinity [12]. This can be easily conducted by Quikchange site directed mutagenesis kit (Agilent) following manufacturer’s instructions or by SOE PCR as outlined here. Additionally, single codon library deep mutational scanning can also be used to guide further rounds of library design by focusing combinatorial libraries to sets of mutations predicted to have the greatest positive effect on activity. Directed evolutionbased selection strategies can then be used to discover candidates with multiple mutations working in concert.

4

Notes 1. It is important to ensure that the lid of the 50% PEG container is secured tightly to prevent evaporation. Change in the concentration of PEG can impact transformation efficiency. 2. EBY100 on YPD forms butyrous, cream colored, smooth surface colonies. 3. To avoid overgrowth, check plates after 36 h and remove accordingly. 4. This mixture can be dispensed into smaller volumes if multiple plasmids are to be transformed. 5. Save remaining liquid culture after transformation at 4  C. This can be used to plate at various dilutions if required. 6. The purity of zymoprep DNA isolated from yeast cells is not suitable for sequencing, hence we recommend propagating the plasmid in E. coli DH5α cells for sequencing, maintaining plasmid DNA stocks, and for future transformation purposes. 7. When plating on LB-Ampicillin plates, centrifuge the recovered cells in SOC medium and aspirate supernatant leaving 100–200 μL in the tube. Resuspend the pelleted cells and plate the entire volume on LB-Ampicillin plates. 8. A titration of various dilutions of primary antibodies may be necessary to determine optimal concentration for staining. 9. For certain proteins (for example, single-chain T cell receptors), it is possible that the protein may not fold properly and hence cannot be detected by conformation-specific antibodies or fluorescent ligand proteins on the yeast cell surface. To identify a stabilized version of the protein, random mutagenesis by error-prone PCR can be used to generate a library which can then be sorted by fluorescent, conformation-specific antibodies

140

Preeti Sharma et al.

or ligand protein. Methods describing generation and selection of error-prone libraries have been previously described by our laboratory [17–19]. 10. Theoretical diversity of a single NNK library at the nucleotide level is 4  4  2 ¼ 32 codons, resulting in 20 amino acids plus 1 stop codon. Accordingly, the total diversity of combined single codon libraries at the protein level can be calculated as the number of residues in the protein replaced by NNK codons multiplied by 20. For example, single codon libraries that fully span a 250 residue protein will result in 20  250 ¼ 5000 variants. 11. It is important that all PCR products at this stage are verified on an electrophoretic gel. The products from the pre-SOE PCR are used directly as templates, without purification, in the next round to assemble full-length insert. This means that some of the original pCT302 plasmid is also transferred to the final SOE PCR mixture, such that full-length products of wildtype sequence templated from pCT302 can be amplified even when the pre-SOE PCR fails. This issue is resolved by validating that all pre-SOE PCR products are correct before advancing to the SOE PCR. 12. Certain PCR reactions may require optimization of annealing temperature and buffer conditions to ensure undesired amplifications are reduced and sufficient quantity of expected SOE product is formed. 13. Load undigested pCT302, pCT302 digested with one enzyme (NheI or XhoI) and double-digested DNA on the same agarose gel to distinguish between undigested vector DNA, linearized vector and double-digested vector. 14. At this step, electrocompetent cells can be frozen and stored at 80  C. The transformation efficiency of freeze-thawed cells can decrease greater than tenfold compared to freshly prepared competent cells, but sufficient enough to transform single construct miniprep DNA. 15. Library size can be calculated by multiplying the number of colonies on the plate by the dilution factor and the total volume of the resuspended library in μL, divided by the volume plated in μL. Since the single codon libraries have lower diversity (104 variants for a 500 amino acids containing protein) compared to when multiple codons are diversified in combination (~107 variants for a 6-codon library), these libraries are not limited by the transformation efficiency of the yeast cells (107 to 108 independent transformants). 16. We recommend sequencing 10 individual plasmid clones at this stage to check for possible errors. Among the sequenced plasmids, there should be few wild-type clones with most plasmids

Deep Mutational Scanning of T Cell Receptors

141

having a single amino acid substitution at sites targeted during SOE PCR. Codon changes should be a mixture of 1, 2, and 3 nt substitutions. 17. If the dissociation constant is unknown and expected to be 1.2 mL/well). 3. SD-CAA media: 20.0 g/L dextrose, 6.7 g/L yeast nitrogen base, 5.0 g/L casamino acids, 10.19 g/L Na2HPO4·7 H2O, 8.56 g/L NaH2HPO4·H2O, in water. 4. SG-CAA induction media: 20.0 g/L galactose, 6.7 g/L yeast nitrogen base, 5.0 g/L casamino acids, 10.19 g/L Na2HPO4·7 H2O, 8.56 g/L NaH2HPO4·H2O, in water. 5. 50 mM HEPES buffer, pH ¼ 7.2. 6. 50 mM MESNA in 50 mM HEPES buffer, pH ¼ 7.2 (see Note 3). 7. DMSO. 8. PBS. 9. Tween-20. 10. Goat anti-Myc (9E10) HRP antibody. 11. 3,30 ,5,50 -tetramethylbenzidine (TMB). 12. 1 M HCl.

2.4 Verifying VLR Binding to Brain ECM Using Murine Brain Sections

1. Snap frozen, non-fixed murine brain sections on microscope slides. 2. SD-CAA media: 20.0 g/L dextrose, 6.7 g/L yeast nitrogen base, 5.0 g/L casamino acids, 10.19 g/L Na2HPO4·7 H2O, 8.56 g/L NaH2HPO4·H2O, in water. 3. SG-CAA induction media: 20.0 g/L galactose, 6.7 g/L yeast nitrogen base, 5.0 g/L casamino acids, 10.19 g/L Na2HPO4·7 H2O, 8.56 g/L NaH2HPO4·H2O, in water. 4. 15-mL round bottom culture tube. 5. 2-Mercaptoethanesulfonic acid (MESNA). 6. PBS. 7. Bovine serum albumin (BSA). 8. Goat serum. 9. Rabbit anti-c-myc antibody. 10. Goat anti-rabbit AF555. 11. Isolectin GS-IB4 AF488. 12. Hoechst 33342.

240

Benjamin J. Umlauf et al.

13. 4% paraformaldehyde. 14. Prolong gold mounting media. 15. Glass coverslip. 16. Nail polish.

3

Methods

3.1 Generating Mammalian ECM

The next sections describe the methods used to generate ECM used for both biopanning and the ELISA-based screen protocols (Fig. 1b, c). Different combinations of cell lines could be substituted into the method described below to isolate VLRs that preferentially accumulate in different tissues.

3.1.1 Mammalian Cell Culture

This method describes the basic steps for culturing bEnd.3 and 3T3 murine cell lines to generate ECM used for biopanning and screening throughout this protocol. For applications requiring ECM from other cell culture sources, substitute culture procedure here. 1. Make DMEM +10% heat inactivated FBS as media for 3T3 cells and DMEM:F12 + 10% heat inactivated FBS as media for bEnd.3 cells. 2. Thaw or seed bEnd.3 or 3T3 cells into a T75 tissue culture flask using appropriate media. 3. Grow cells till approximately 80% confluent. 4. Wash cells two times with PBS. 5. Add 3 mL of Accutase and incubate at 37
C for 3–10 min (see Note 4). 6. Tap the flask gently to dislodge cells. 7. Wash two times with complete media. 8. Count cells. 9. Plate 30,000 cells/well in six-well plate or 1000 cells/well in 96-well plate. 10. Incubate plates at 37
C, 5% CO2 for 48–72 h until wells are completely confluent. These plates are then used in the decellularization step (Subheading 3.1.2) described below.

3.1.2 Decellularizing Mammalian Culture Substrates to Expose ECM

This section describes the gentle removal of mammalian cells from tissue culture plates using a detergent and enzyme-free method to expose “native” ECM used for biopanning. 1. After bEnd.3 or 3T3 cell culture plates are 100% confluent, wash cells two times with PBS. 2. Incubate cells with 1000 μL of versene/well for six-well plate or 100 μL of versene/well for a 96-well plate for 10 min at 37
C.

ECM Biopanning with a VLR Yeast Surface Display Library

241

3. Gently tap the plates to dislodge cells. 4. Flick plate to aspirate cells+versene into a waste dish. 5. Repeat steps 2–4 three to four times to completely remove cells from plate, leaving the ECM. 6. Validate mammalian cell removal using light microscopy. 7. Add 1000 μL of PBS + 1% BSA + 1% goat serum/well for six-well plate or 100 μL/well for 96-well plate to serve as storage and blocking buffer. 8. Seal plates with mylar adhesive plate sealer and store at 4
C (see Note 5). 3.2 ECM Biopannning with a VLR YSD Library

The following sections describe using a VLR YSD library to identify clones that bind to ECM (Fig. 1). These sections describe enriching the library ECM-binding clones that can subsequently be screened for selective binding to tissue-specific ECM.

3.2.1 Yeast Culture

This method describes preparing a VLR yeast surface display library for biopanning. Here we use an existing yeast surface display library [11] where the VLR is fused to Aga2p for yeast surface expression. See Refs [19–23] or Chaps. 2, 7, or 18 of this volume for information on creating yeast surface display libraries. 1. Thaw aliquot of frozen VLR yeast surface display library (pCT surface display plasmid backbone in EBY100 yeast display strain) rapidly at 37
C. 2. Add 10–100 μL of yeast library to 3 mL of SD-CAA media in 15-mL round bottom culture tube, being sure to thaw enough to oversample library diversity. 3. Culture O.N. at 30
C with 260 rpm shaking. 4. The following morning use the O.N. culture to inoculate a fresh SD-CAA culture to OD600 ¼ 0.3. Ensure five times the headspace in the size of the culture flask compared to volume of media. We typically use 3–50 mL cultures depending on downstream experimental requirements. 5. Incubate at 30
C for ~4–6 h with 260 rpm shaking till OD600 ¼ 0.8. 6. Pellet yeast by centrifugation at 2000  g for 5 min. 7. Remove media and resuspend yeast pellet with the same volume of SG-CAA induction media that was removed. 8. Incubate at 20
C for 24 h with 260 rpm shaking to induce surface display of VLR-Aga2p fusions. 9. Harvest yeast pellet by centrifuging cultures at 2000  g for 5–10 min. 11. Yeast pellet is retained for biopanning as described below (Subheading 3.2.2).

242

Benjamin J. Umlauf et al.

3.2.2 bEnd.3 ECM Biopanning

This section describes the biopanning steps using ECM and VLR displaying yeast prepared in the previous steps of this protocol. 1. Resuspend yeast pellet from Subheading 3.2.1, step 11 in 1000 μL PBS + 1% BSA. 2. Centrifuge cultures at 2000  g for 5–10 min. 3. Repeat steps 1 and 2. 4. Aspirate blocking buffer from 6-well bEnd.3 ECM plate generated in Subheading 3.1.2. 5. Resuspend yeast pellet in 1000 μL of PBS + 1% BSA/well. Add yeast at a density of ~0.85  106 yeast/cm2 to a 6-well bEnd.3 ECM plate (see Note 6). 6. Incubate plate for 1 h at room temperature, agitating the plate gently every 15–20 min. 7. Wash well three times with 3 mL of PBS/wash discarding supernatant after each wash (see Note 7). 8. Add 500 μL of 0.1 mM Glycine pH ¼ 2 to the well. Steps 8 and 9 are used to recover the ECM-binding yeast for amplification and analysis. 9. Incubate for 3–5 min. 10. Harvest supernatant. 11. Repeat steps 8–10 two times pooling acid wash after each elution. 12. Neutralize acid elution by adding 150 μL (10) of 1 M Trisbase, pH ¼ 9. 13. Add elution to 1.5 mL of SD-CAA media in a 15-mL round bottom culture tube. 14. Culture O.N. at 30
C with 260 rpm shaking. 15. The following morning use the O.N. culture to inoculate a fresh SD-CAA culture to OD600 ¼ 0.3. Ensure five times the headspace in the size of the culture flask compared to volume of media. We typically use 3–50 mL cultures depending on downstream experimental requirements. 16. Incubate at 30
C for ~4–6 h with 260 rpm shaking till OD600 ¼ 0.8. 17. Pellet yeast by centrifugation at 2000  g for 5 min. 18. Remove media and resuspend yeast pellet with the same volume of SG-CAA induction media that was removed. 19. Incubate at 20
C for 24–72 h with 260 rpm shaking to induce surface display of VLR-Aga2p fusions. 20. Harvest yeast pellet by centrifuging cultures at 2000  g for 5–10 min.

ECM Biopanning with a VLR Yeast Surface Display Library

243

21. Repeat steps 1–20 to complete a minimum of two rounds of biopanning, or until you have seen substantial recovery of ECM-binding yeast clones (examine binding yeast coverage between steps 7 and 8 above). After the second round of biopanning, save yeast pellet for brain ECM screening protocol described below (Subheading 3.3). 3.3 ELISA-Based Screen for Identifying Clonal VLRs That Demonstrate Preferential Accumulation in bEnd.3 ECM

This section describes the method for identifying VLR ECM binding clones that demonstrate preference for binding brain ECM compared to fibroblast ECM (Fig. 1). Single YSD VLR clones that have undergone two rounds of biopanning are screened to determine the relative binding to bEnd.3 ECM (target) and 3T3 ECM (negative control) using an ELISA-based method with VLR directly released from displaying yeast. 1. Yeast pellet from Subheading 3.2.2, step 20 is resuspended in 100 μL of PBS. 2. Streak yeast on SD-CAA agar plates. 3. Pick single yeast colonies into individual wells of a 96-well polypropylene deep well block (total volume of each well should be 1.2–2 mL) containing 200 μL of SD-CAA media/ well. 4. Incubate O.N. at 30
C, 260 rpm. 5. Harvest 180 μL of yeast culture and add 20 μL of DMSO (final concentration is 10%). Freeze yeast in controlled rate freezing container for archiving. 6. Add 180 μL of SD-CAA media to each well. 7. Incubate at 30
C for ~4–6 h with 260 rpm. The expected OD600 range should be between 0.8 and 1. 8. Pellet yeast by centrifuging cultures at 2000  g for 5–10 min. 9. Remove media and resuspend yeast with 200 μL of SG-CAA induction media. 10. Incubate at 20
C for 24 h with 260 rpm shaking to induce surface display of VLR-Aga2p fusions. 11. Pellet yeast by centrifuging cultures at 2000  g for 5–10 min. 12. Wash with 200 μL of 50 mM HEPES pH ¼ 7.2. 13. Pellet yeast by centrifuging cultures at 2000  g for 5–10 min. 14. Resuspend with 20 μL of 50 mM MESNA in 50 mM HEPES, pH ¼ 7.2 to reduce VLR-Aga2p fusions off of the yeast surface (see Note 3). 15. Incubate at 20
C, 260 rpm shaking for 45 min to dissociate VLR-Aga2p complexes from yeast surface. 16. Pellet yeast by centrifuging cultures at 2000  g for 5–10 min.

244

Benjamin J. Umlauf et al.

17. Remove the 20 μL of MESNA solution containing released VLR-Aga2p fusions into 180 μL of PBS pH ¼ 7.4 (see Note 8). 18. Mix wells well by pipetting up and down to ensure VLR is uniformly distributed in the solution. 19. Remove blocking buffer from 96-well bEnd.3 and 3T3 ECM plates. 20. Add 100 μL of VLR-Aga2p solution from step 18 to bEnd.3 ECM plate. 21. Add remaining 100 μL of VLR solution from step 18 to 3T3 ECM plate. 22. Incubate plates at room temperature for 1 h with intermittent agitation every 15–20 min. 23. Wash wells five times with PBS + 0.05% tween 20. 24. Add 100 μL of goat anti-Myc HRP antibody (1 mg/mL stock solution diluted 1:1000 in PBS + 1% BSA + 1% goat serum) to each well. 25. Incubate for 45 min at room temperature. 26. Wash well seven times with PBS + 0.05% tween 20, 1 min/ wash. 27. Add 100 μL of TMB solution to each well. 28. Incubate at room temperature for 15–30 min for colorimetric reaction to proceed. 29. Add 50 μL of 1 M HCl to each well to stop peroxidase reaction before signal saturation. 30. Quantify absorbance at 450 nm. 31. Compare absorbance signal from bEnd.3 and 3T3 ECM to quantify differences in individual VLR accumulation in neural ECM compared to fibroblast ECM. 3.4 Verification of VLR Brain ECM Binding Using Murine Brain Sections

The protocol below describes the validation of VLR clones for their binding to brain ECM using murine brain sections (Figs. 1d and 2). This assay is key to demonstrate that the VLR clones bind to ECM in an ex vivo brain environment and provide potential value beyond ECM generated by cultured cells in vitro. 1. Thaw a single brain ECM binding, candidate VLR yeast surface display clone from plate generated in Subheading 3.3, step 5 by placing a finger on the bottom of the well. 2. Add 10 μL of thawed yeast stock culture to 3 mL of SD-CAA media in 15-mL round bottom culture tube. 3. Culture O.N. at 30
C with 260 rpm shaking. 4. The following morning use the O.N. culture to inoculate a fresh SD-CAA culture to OD600 ¼ 0.3. Ensure five times the

ECM Biopanning with a VLR Yeast Surface Display Library

245

headspace in the size of the culture flask compared to volume of media. We typically use 3 mL cultures. 5. Incubate at 30
C for ~4–6 h with 260 rpm shaking till OD600 ¼ 0.8. 6. Pellet yeast by centrifugation at 2000  g for 5 min. 7. Remove media and resuspend yeast pellet with the same volume of SG-CAA induction media that was removed. 8. Incubate at 20
C for 24 h with 260 rpm shaking to induce surface expression of VLR-Aga2p fusions. 9. Harvest yeast pellet by centrifuging cultures at 2000  g for 5 min. 10. Wash with 1000 μL of 50 mM HEPES, pH ¼ 7.2. 11. Pellet yeast by centrifuging cultures at 2000  g for 5 min. 12. Resuspend with 50 μL of 50 mM MESNA in 50 mM HEPES, pH ¼ 7.2. 13. Incubate at 20
C, 260 rpm shaking for 45 min to release VLR-Aga2p complexes from yeast surface. 14. Pellet yeast by centrifugation at 2000  g for 5 min. 15. Harvest 50 μL of MESNA solution into 450 μL of PBS, pH ¼ 7.4. 16. Mix well. 17. Thaw snap frozen murine brain sections that have been cut and placed on microscope slides. 18. Remove excess optimal cutting temperature (O.C.T.) compound with razor blade. 19. Outline the murine brain section on the microscope slide using a PAP pen which creates a hydrophobic barrier around the brain section to contain washing and staining fluids. 20. Wash brain section two times with PBS. 21. Block sections for 30–60 min with PBS + 1%BSA + 1% goat serum. 22. Aspirate blocking buffer and add 100–200 μL of VLR solution. 23. Incubate at room temperature for 60 min. 24. Wash brain sections three times, 5 min/wash with PBS + 1% BSA + 1% goat serum. 25. Add 100 μL of a master mix containing: rabbit anti c-myc antibody (1:500 of 1 mg/mL stock solution for labeling VLR), goat anti-rabbit AF555 (1:1000 of 1 mg/mL stock solution for labeling anti-c-myc antibody), and Isolectin GS-IB4 AF488 (1:400 of 1 mg/mL stock solution for labeling murine blood vessels) in PBS + 1%BSA + 1% goat serum to each brain section (see Note 9).

246

Benjamin J. Umlauf et al.

26. Incubate for 45 min at room temperature. 27. Without aspirating the master mix, add Hoechst 33342 (1:800 of 10 mg/mL solution in water for labelling nuclei) to each brain section. 28. Incubate for 15 min at room temperature. 29. Wash brain sections two times with PBS + 1%BSA + 1% goat serum and 1 time with PBS, 5 mins/wash. 30. Add 100 μL of 4% paraformaldehyde to fix each section. 31. Incubate at room temperature for 10 min. 32. Wash brain sections three times, 1 min/wash with PBS. 33. Add one drop of mounting media to each section for fluorescent signal preservation. 34. Drop glass coverslip onto brain section ensuring air bubbles are removed. 35. Use nail polish to seal coverslip to microscope slide. 36. Image slide on a fluorescent microscope. Nuclei are visible in the blue channel, blood vessels visible in the green channel, and VLRs visible in the red channel.

4

Notes 1. Other mammalian cell types, including cells of human origin, may be substituted into this protocol. 2. Other types of acid wash including citrate buffer pH ¼ 3 are suitable for this protocol. 3. 5 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP) in 50 mM HEPES, pH 7.2, can also be used as the reducing agent. TCEP can be beneficial compared to MESNA because phosphate in PBS will inactivate TCEP, thus prevent the reducing agent from interacting with ECM. Additionally, the phosphine mechanism of TCEP reduction will not allow for VLR multimers to form which could potentially occur with MESNA as it uses the typical thiol substitution mechanism to reduce Aga2p disulfide bonds. 4. 0.25% trypsin + EDTA can be substituted for Accutase. 5. We have successfully stored sealed ECM plates at 4
C for up to 6 months when sodium azide is added to blocking/storage buffer. 6. Number of yeast cells is estimated by measuring the OD600. An OD600 of 1.0 corresponds to approximately 1.5  107 cells/ mL. This degree of accuracy is usually sufficient for biopanning;

ECM Biopanning with a VLR Yeast Surface Display Library

247

however, the exact value can be calculated for each individual spectrophotometer using a hemocytometer and bright field microscope. 7. Wash well vigorously enough to remove all unbound yeast. This is generally achieved by evacuating pipet rapidly against the wall of the six-well plate. It is important not to touch the ECM layer with the pipet tip. 8. We anticipate having nM levels of VLR released from the yeast surface using the 200 μL scale. 9. Since we observe lot-to-lot variation, the amount of Isolectin GS-IB4 AF488 may need to be optimized.

Acknowledgments This work was funded by National Science Foundation grant CBET1403350, a Catalyst Award from the Falk Medical Research Trust, and the Clayton Foundation for Research. References 1. Frantz C, Stewart KM, Weaver VM (2010) The extracellular matrix at a glance. J Cell Sci 123 (Pt 24):4195–4200. https://doi.org/10. 1242/jcs.023820 2. Yue B (2014) Biology of the extracellular matrix: an overview. J Glaucoma 23(8 Suppl 1):S20–S23. https://doi.org/10.1097/IJG. 0000000000000108 3. Bonnans C, Chou J, Werb Z (2014) Remodelling the extracellular matrix in development and disease. Nat Rev Mol Cell Biol 15(12): 7 8 6 – 8 0 1 . h t t p s : // d o i . o r g / 1 0 . 1 0 3 8 / nrm3904 4. Colley KJ, Varki A, Kinoshita T (2015) Cellular organization of glycosylation. In: Varki A, Cummings RD et al (eds) Essentials of glycobiology. Cold Spring Harbor Press, Cold Spring Harbor, pp 41–49. https://doi.org/ 10.1101/glycobiology.3e.004 5. Abbott NJ, Patabendige AAK, Dolman DEM, Yusof SR, Begley DJ (2010) Structure and function of the blood-brain barrier. Neurobiol Dis 37(1):13–25. https://doi.org/10.1016/j. nbd.2009.07.030 6. Davies DC (2002) Blood-brain barrier breakdown in septic encephalopathy and brain tumours. J Anat 200(6):639–646 7. Obermeier B, Verma A, Ransohoff RM (2016) The blood-brain barrier. Handb Clin Neurol 133:39–59. https://doi.org/10.1016/B9780-444-63432-0.00003-7

8. Sarkaria JN, Hu LS, Parney IF, Pafundi DH, Brinkmann DH, Laack NN, Giannini C, Burns TC, Kizilbash SH, Laramy JK, Swanson KR, Kaufmann TJ, Brown PD, Agar NYR, Galanis E, Buckner JC, Elmquist WF (2018) Is the blood-brain barrier really disrupted in all glioblastomas? A critical assessment of existing clinical data. Neuro Oncol 20(2):184–191. https://doi.org/10.1093/neuonc/nox175 9. Umlauf BJ, Shusta EV (2019) Exploiting BBB disruption for the delivery of nanocarriers to the diseased CNS. Curr Opin Biotechnol 60: 146–152. https://doi.org/10.1016/j.copbio. 2019.01.013 10. Yang Y, Rosenberg GA (2011) Blood-brain barrier breakdown in acute and chronic cerebrovascular disease. Stroke 42(11): 3323–3328. https://doi.org/10.1161/ STROKEAHA.110.608257 11. Umlauf BJ, Clark PA, Lajoie JM, Georgieva JV, Bremner S, Herrin BR, Kuo JS, Shusta EV (2019) Identification of variable lymphocyte receptors that can target therapeutics to pathologically exposed brain extracellular matrix. Sci Adv 5(5):eaau4245. https://doi.org/10. 1126/sciadv.aau4245 12. Han BW, Herrin BR, Cooper MD, Wilson IA (2008) Antigen recognition by variable lymphocyte receptors. Science 321(5897): 1834–1837. https://doi.org/10.1126/sci ence.1162484

248

Benjamin J. Umlauf et al.

13. Herrin BR, Alder MN, Roux KH, Sina C, Ehrhardt GRA, Boydston JA, Turnbough CL, Cooper MD (2008) Structure and specificity of lamprey monoclonal antibodies. Proc Natl Acad Sci U S A 105(6):2040–2045. https:// doi.org/10.1073/pnas.0711619105 14. Herrin BR, Cooper MD (2010) Alternative adaptive immunity in jawless vertebrates. J Immunol 185(3):1367–1374. https://doi. org/10.4049/jimmunol.0903128 15. Hirano M, Guo P, McCurley N, Schorpp M, Das S, Boehm T, Cooper MD (2013) Evolutionary implications of a third lymphocyte lineage in lampreys. Nature 501(7467):435–438. https://doi.org/10.1038/nature12467 16. Collins BC, Gunn RJ, McKitrick TR, Cummings RD, Cooper MD, Herrin BR, Wilson IA (2017) Structural insights into VLR fine specificity for blood group carbohydrates. Structure 25(11):1667–1678.e1664. https:// doi.org/10.1016/j.str.2017.09.003 17. Hong X, Ma MZ, Gildersleeve JC, Chowdhury S, Barchi JJ Jr, Mariuzza RA, Murphy MB, Mao L, Pancer Z (2013) Sugarbinding proteins from fish: selection of high affinity “lambodies” that recognize biomedically relevant glycans. ACS Chem Biol 8(1): 1 5 2 – 1 6 0 . h t t p s : // d o i . o r g / 1 0 . 1 0 2 1 / cb300399s 18. Boder ET, Wittrup KD (1997) Yeast surface display for screening combinatorial polypeptide

libraries. Nat Biotechnol 15(6):553–557. https://doi.org/10.1038/nbt0697-553 19. Wang XX, Shusta EV (2005) The use of scFvdisplaying yeast in mammalian cell surface selections. J Immunol Methods 304(1–2): 30–42. https://doi.org/10.1016/j.jim.2005. 05.006 20. Boder ET, Midelfort KS, Wittrup KD (2000) Directed evolution of antibody fragments with monovalent femtomolar antigen-binding affinity. Proc Natl Acad Sci U S A 97(20): 10701–10705. https://doi.org/10.1073/ pnas.170297297 21. Burns ML, Malott TM, Metcalf KJ, Puguh A, Chan JR, Shusta EV (2016) Pro-region engineering for improved yeast display and secretion of brain derived neurotrophic factor. Biotechnol J 11(3):425–436. https://doi. org/10.1002/biot.201500360 22. Tillotson BJ, Cho YK, Shusta EV (2013) Cells and cell lysates: a direct approach for engineering antibodies against membrane proteins using yeast surface display. Methods 60(1): 27–37. https://doi.org/10.1016/j.ymeth. 2012.03.010 23. Tillotson BJ, Lajoie JM, Shusta EV (2015) Yeast display-based antibody affinity maturation using detergent-solubilized cell lysates. Methods Mol Biol 1319:65–78. https://doi. org/10.1007/978-1-4939-2748-7_4

Part IV Specialized Yeast Surface Display Applications

Chapter 14 Guidelines, Strategies, and Principles for the Directed Evolution of Cross-Reactive Antibodies Using Yeast Surface Display Technology Sara Linciano, Ee Lin Wong, Ylenia Mazzocato, Monica Chinellato, Tiziano Scaravetti, Alberto Caregnato, Veronica Cacco, Zhanna Romanyuk, and Alessandro Angelini Abstract The ability of cross-reactive antibodies to bind multiple related or unrelated targets derived from different species provides not only superior therapeutic efficacy but also a better assessment of treatment toxicity, thereby facilitating the transition from preclinical models to human clinical studies. This chapter provides some guidelines for the directed evolution of cross-reactive antibodies using yeast surface display technology. Cross-reactive antibodies are initially isolated from a naı¨ve library by combining highly avid magnetic bead separations followed by multiple cycles of flow cytometry sorting. Once initial cross-reactive clones are identified, sequential rounds of mutagenesis and two-pressure selection strategies are applied to engineer cross-reactive antibodies with improved affinity and yet retained or superior cross-reactivity. Key words Yeast surface display, Directed evolution, Antibody engineering, Protein recognition, Combinatorial library screening, Cross-reactivity, Multispecificity, Promiscuity

1

Introduction Monoclonal antibodies (Abs) have emerged as one of the fastest growing and most successful class of biotherapeutics, and they are nowadays used widely in the clinic to treat a variety of cancers, inflammatory, and infectious diseases [1]. Advantages of Abs include high specificity, high affinity, long half-life, and low toxicity. Though high specificity is usually perceived as a key feature for clinical and commercial success of a therapeutic Ab, it often poses difficulties in assessing its efficacy and toxicity in preclinical animal models. Diseases usually involve multiple ligands and receptors acting in concert. As a result, highly specific therapeutic Ab targeting a single pathological target are often insufficient to achieve

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_14, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

251

252

Sara Linciano et al.

desired clinical outcomes. Contrariwise, cross-reactive antibodies (crAbs) capable of recognizing multiple targets of interest (TOI) have the potential to exhibit superior therapeutic efficacy. Moreover, their ability of binding multiple TOIs derived from different species allows better assessment of therapeutic efficacy and toxicity in immunocompetent animal models ultimately enabling a safer and faster progression of the same molecule through the preclinical and clinical phases in humans. Unfortunately, crAbs are challenging to obtain using traditional methodologies involving animal immunization and hybridoma techniques. The immune system tends to remove self-reactive Abs, making it difficult to generate in vivo crAbs against sequence and structurally related antigens derived from the same or different species. In contrast, in vitro antibody libraries associated with display technologies are unaffected by immune tolerance [2]. The two most widely used technologies to generate crAbs in vitro are phage and yeast display [3, 4]. While phage display was largely used at the beginning and proved to be highly effective in the isolation of crAbs against either related [5–7] or unrelated TOIs [8], in the last few years we have witnessed an increasing number of crAbs that have been developed using yeast display [9–16]. Abs are complex mammalian proteins containing multiple domains and posttranslational modifications. In addition, the presence of multiple framework mutations, usually included in broadly CrAbs, often results in lower stability and production yields. The ability of yeast to rely on a eukaryotic posttranslational machinery (e.g., disulfide isomerization and glycosylation) facilitates the folding of such complex proteins otherwise not attainable using bacteria. Indeed, yeast-encoded crAbs have been engineered as (1) a single-chain variable fragment (scFv) [9–12, 14, 15], (2) an antibody fragment antigen-binding portion (Fab) [13], and (3) a whole immunoglobulin (IgG) [16]. CrAbs often occur at lower frequency within the total selected variant pool and have weaker binding affinities compared to highly specific Abs. The ability of yeast to display multiple copies of the same Ab on the cell surface (~105 proteins per cell) facilitates highly avid interactions with the TOI promoting isolation of weak yet broadly cross-reactive binders otherwise difficult to identify. If necessary, avidity effects can subsequently be overcome by applying kinetic selections [17]. Finally, yeast display allows high-throughput quantitative screening and biophysical characterization of combinatorial libraries via fluorescence-activated cell sorting (FACS). Indeed, flow cytometry permits simultaneous screening of combinatorial libraries for binding to multiple TOIs. By applying different target labeling approaches, including (1) equilibrium binding, (2) competition for limited TOI, and (3) kinetic selections, it is possible to precisely discriminate yeast cells displaying Ab clones with varying target affinities and specificities [18], otherwise difficult to accomplish using phage display. While detailed protocols for the selection and engineering

Engineering Antibody Cross-Reactivity by Yeast Display

253

of Abs from large combinatorial libraries displayed on yeast have been extensively reported elsewhere [19] and in various chapters of this volume on “Yeast Surface Display” (e.g., Chaps. 2 and 22), in this chapter, we will focus solely on providing some basic guidelines on molecular evolution of crAbs using yeast surface display technology. CrAbs are initially isolated from a naı¨ve library by combining highly avid magnetic bead separations followed by multiple cycles of flow cytometry sorting. Once initial cross-reactive clones are identified, successive rounds of mutagenesis and two-pressure selection strategies are applied to engineer and select crAbs with improved affinity, yet retained or superior cross-reactivity.

2

Isolation of Cross-Reactive Antibodies from a Naı¨ve Yeast Display Library The initial selection process focuses primarily on the isolation of cross-reactive molecules with little or no pressure on binding affinity. The screening usually involves the use of naı¨ve yeast display antibody library and highly avid magnetic bead separations followed by multiple cycles of flow cytometry sorting (Fig. 1).

2.1 Choosing the Targets, the Labeling Strategy, and Antibody Library to Be Used

To force cross-reactivity and thus enhance the chances of isolating promiscuous Abs, we advise screening for binding to all TOIs during the selection process. If TOIs are numerous, we recommend prioritizing those sharing low-sequence identity. For therapeutic purposes, precedence should be given to TOIs derived from species commonly used in preclinical and clinical studies (e.g., mouse-, rat-, monkey-, and human-derived targets). To preserve the

Fig. 1 Schematic representation of the isolation of cross-reactive antibodies (crAbs) from a naı¨ve yeast display library. A representative heat map displaying the sequence identity of multiple targets of interest (TOIs) is shown on the left. The color of each element in the heat map indicates the sequence identity percentage, ranging from low (white) to high (dark gray). For selection purposes, highly diverse TOIs are chosen, incubated with a naı¨ve yeast display Ab library and further submitted to iterative selection pathways of magnetic bead separation followed by cycles of flow cytometry sorting. The initial magnetic bead-based screening allows for the isolation of crAbs through highly avid interactions between yeast cells and the immobilized TOI on magnetic beads. Yeast cells isolated after the magnetic bead-based processes are further screened against all TOIs using fluorescence-activated cell sorting. To favor the selection of broadly crAbs, the cell output of each round of selection is exposed to a diverse array of TOIs in the next cycle. The biophysical properties of isolated single clones are subsequently characterized using yeast surface display titrations

254

Sara Linciano et al.

functional regions of the TOIs unaltered, avoid loss of epitope recognition, and prevent additional structural heterogeneity, we advise against performing a chemistry-based labeling of the TOIs (e.g., amine-reactive succinimidyl esters, thiol-reactive maleimides). Site-specific labeling of the TOI can instead be achieved by exploiting fluorescently conjugated reagents capable of recognizing specific peptide tags (e.g., Avi-tag, His-tag, Flag-tag) or fusion protein partner (e.g., GST, MBP, GFP, Sumo, Fc fragment, and albumin) sequences located at well-tolerated N- or C-terminus of the TOI. For instance, site-specific labeling of the TOI can be achieved by performing enzymatic biotinylation of Avi-tag sequence. The specific addition of a single biotin molecule on Avi-tag sequence enables the TOI to bind avidin, streptavidin, and neutravidin proteins with extremely high affinity, fast on-rate, high specificity, and precise orientation. Though there are examples of crAbs that have been isolated starting from large naı¨ve libraries generated using peripheral blood B lymphocytes derived from non-immunized healthy donors [12, 15], chances of identifying crAbs are usually higher when selection is performed using libraries originated from B-cells of humans or mouse immunized with the TOI [10] or synthetic Ab repertoires [9, 16]. The latter are currently the most used as several studies have shown that synthetic Ab libraries are truly naı¨ve since they have not been subjected to the restrictions imposed by self-tolerance of natural repertoires [20]. Moreover, synthetic Ab repertoires provide epitope coverage wider than those obtained from animal immunizations, thus yielding specificities not otherwise attainable [21]. 2.2 Isolation of Cross-Reactive Antibodies Using Highly Avid Magnetic Beads

The initial use of magnetic bead-based screening allows the isolation of weak affinity crAbs within a relative short period of time and in a high-throughput manner [22]. The multivalency of the yeast display system in combination with the use of site-specifically labeled TOIs (e.g., enzymatically biotinylated) bound to micronsized magnetic beads (e.g., streptavidin-coated magnetic beads) results in a high avidity system with multiple copies of the Ab displayed on the yeast cell surface that can interact with multiple copies of the TOI immobilized on the bead. The highly avid interaction between yeast cells and magnetic beads allows for the isolation of weak binders that would be otherwise difficult to detect by flow cytometry [19]. Prior to “positive selection” against the TOI, we advise performing “negative selection” processes in order to deplete the naı¨ve library of crAb-binding streptavidin-coated magnetic beads. The negative selection is usually repeated multiple times using unbound collected cells and newly prepared magnetic beads. Additional negative selections might also be needed in between successive positive selection steps to prevent enrichment of previously undepleted streptavidin-coated magnetic bead binders [19]. If the TOIs share a common peptide tag or fusion

Engineering Antibody Cross-Reactivity by Yeast Display

255

protein partner, we recommend performing negative selections with either the tag or the fusion protein partner only in addition to the negative selections against streptavidin-coated magnetic beads [19]. Positive selection is further performed by exposing the previously depleted yeast display library against the TOIs, typically captured on magnetic beads. Two or more iterative cycles of magnetic bead selections are usually applied. To drive crossreactivity and thus enhance the likelihoods of isolating promiscuous Abs, we advise (1) incubating each TOI with tenfold cell output collected from previous selection round performed using a diverse TOI and (2) to maintain a constant and relatively high (0.1–1 μM) concentration of each TOI in early flow cytometry-based selection rounds. Each magnetic bead screening will comprise growth of yeast cells, expression of the Abs on the surface, binding to the immobilized biotinylated TOI, washing, and expansion of the isolated bound yeast cells [19]. 2.3 Selection CrossReactive Antibodies Using FluorescenceActivated Cell Sorting

Selected yeast cells isolated after the magnetic bead-based processes are further screened using FACS. The use of multi-color labeling schemes based on fluorescently conjugated detection reagents enables (1) quantitative screening of combinatorial libraries for binding to multiple TOIs in real time, (2) normalization of antigen-binding signal to cell surface expression, and (3) simultaneous selection for both protein stability and binding affinity, as the extent of protein surface expression has been shown to correlate with protein stability [19, 23, 24]. To increase the opportunities of isolating crAbs during initial FACS-based selection cycles, we recommend (1) using highly avid fluorescently labeled reagents (e.g., streptavidin and neutravidin) preloaded with diverse biotinylated TOIs, (2) performing consecutive flow cytometry sorts in which tenfold cell sorted output will be incubated with a different TOI at equivalent concentration, (3) using less-stringent gates and collecting >1% of the entire sorted yeast population in order to avoid loss of unique clones and to enrich for both low- and high-affinity clones. Importantly, secondary fluorescent-conjugated detection reagents for FACS should be constantly alternated to avoid enrichments of cross-reactive clones that could bind to them. Moreover, incubation volumes and the number of yeast cells stained should be chosen to ensure the number of target molecules to be tenfold in excess over the number of surface expressed Abs, assuming ~105 Abs displayed per yeast cell [19]. The number of FACS rounds to be performed is usually dictated by the number of biotinylated TOIs to be screened. The initial selection can be considered completed when the yeast population has been exposed to all TOIs, and clones with desired affinity and cross-reactivity have been enriched and collected. The identity of individual clones is further revealed by either Sanger or next-generation DNA sequencing and the binding affinities (KD) determined using yeast cell surface titrations [19].

256

Sara Linciano et al.

2.4 Single Clone Analysis and Characterization Using Yeast Surface Titrations

3

The equilibrium dissociation constant (KD) of each individual selected Ab clone toward each single TOI can be determined using yeast surface display titrations. Several studies have shown that KD values of a binding interaction determined using yeast cell surface titrations correlate well with KD values measured using soluble engineered binders [17, 19]. Moreover, the analysis of the binding affinity of selected Ab expressed on the surface of the yeast cell eliminates the need for additional sub-cloning, soluble protein expression, and purification steps [17–19]. At this stage, we recommend testing binding of isolated clones toward all desired TOIs to evaluate extent of both affinity and cross-reactivity. Binding should be tested also against (1) secondary fluorescent-conjugated detection reagents only and (2) some structurally unrelated antigens to assess unspecific binding. The TOI concentration range should ideally span two orders of magnitude both above and below of the expected KD value. The reaction incubation time and labeling reaction volume depend on the binding affinity (KD), and the concentration of TOI used should ensure a tenfold molar excess of TOI over the total number of displayed Abs present in solution [19].

Molecular Co-Evolution of Antibody Affinity and Cross-Reactivity The affinity of the isolated crAbs can be further increased by applying sequential rounds of mutagenesis and two-pressure selection strategies in which both affinity and cross-reactivity are further evolved concomitantly (Fig. 2a).

3.1 Generation of Genetic Diversity of Cross-Reactive Antibody Clones

The engineering of a CrAb with improved affinity toward multiple TOIs requires the introduction of genetic diversity into the Ab encoding gene. Combinatorial libraries of Ab variants can be generated using different random genetic diversification techniques. To maximize the likelihood of success, these mutagenesis strategies can be combined into a single diversification step or applied separately during successive rounds of evolution. Computational design methods can also be used to enable faster and effective exploration of larger sequence space. We suggest to initially apply approaches of random mutagenesis throughout the entire Ab gene. We have observed that key mutations do not only accumulate within the complementarity-determining regions (CDRs) but also within the framework regions (FWRs). The latter are generally difficult to predict. Random mutagenesis techniques commonly applied on yeast surface display to generate DNA inserts that encode for mutant Ab clones include error-prone PCR with nucleotide analogues [25] and DNA shuffling [26]. Error-prone PCR with nucleotide analogues is generally preferred to prepare mutagenic DNA inserts, since the mutation rate can be controlled by varying the

Engineering Antibody Cross-Reactivity by Yeast Display

257

Fig. 2 Schematic representation of the molecular co-evolution of antibody affinity and cross-reactivity. (a) Combinatorial libraries of Ab variants are generated using mutagenesis. Novel clones with improved affinity and cross-reactive toward multiple TOIs are further isolated by applying long series of two-pressure flow cytometry selections. The biophysical properties of isolated single clones are subsequently characterized, and, if necessary, Ab affinity and cross-reactivity are further enhanced through successive rounds of mutagenesis and flow cytometry selections; (b) Representative binding titrations of parental crAb (x) and its derived clones (y and z) isolated from sequential iterative processes of selection in which Ab variants are screened in order from lowest (dark gray) to highest (black) affinity targets. Binding of molecular evolved crAb clones against soluble TOIs is assessed by flow cytometry-based assay. The obtained fluorescence binding median values are normalized to the display median fluorescence intensities (y-axis) and plotted against varying concentrations of soluble TOIs (x-axis); (c) Plots of the binding affinities of crAb x, y, and z toward three different TOIs as determined from titration curves in panel b. Targets and the corresponding binding affinity values are reported as differently colored filled circles. Data are presented as inverted of equilibrium binding constants (1/KD). Iterative two-pressure co-evolutionary processes enable the development of promiscuous Abs whose affinity and cross-reactivity toward multiple TOIs is higher than that of their respective parental clones

number of PCR cycles, and both transition and transversion mutations can occur. Although crAbs often showed high mutation rates, the substitution of multiple amino acid at once may have deleterious effect on Ab stability and cross-reactivity. We therefore advise applying at each cycle a low mutagenesis rate, thus allowing, on average, one to two amino-acid mutations per Ab gene. If necessary, multiple rounds of mutagenesis followed by FACS can be employed iteratively until Ab clones with desired affinity and promiscuity are identified. The yeast library is finally generated using homologous recombination [19]. 3.2 Flow Cytometry Sorting to Enrich for Cross-Reactive Antibody Clones with Higher Affinity to Multiple Targets

Selection and enrichment of mutated Ab clones with increased binding affinity and yet preserved or improved cross-reactivity is further performed by allowing the mutants to evolve through sequential cycles of equilibrium-based selection using FACS screening. Higher affinity clones are isolated by incubating the mutant with decreasing concentrations of the TOIs. We recommend using TOI concentrations that are not more than tenfold below the measured KD values. Lower concentrations could favor the selection of high affinity and specificity clones at the expense of the lower

258

Sara Linciano et al.

affinity cross-reactive ones. The cross-reactivity should be forced concomitantly by exposing Ab variants isolated from each affinity selection cycle toward a different TOI in the following cycle. Importantly, the sequential order in which the TOIs are exposed to the Ab mutant libraries is key for the success of the selection process. Improvements in both affinity and cross-reactivity are usually observed only when the library is screened in order from lowest to highest affinity targets of the parental clone (Fig. 2b). Iterative two-pressure co-evolutionary processes, each comprising multiple and consecutive cycles of random mutagenesis and FACSbased selections, are often necessary to promote the isolation of Abs whose affinity and cross-reactivity toward multiple TOIs is higher than that of their respective parental clones. Each selection cycle should comprise growth of yeast cells, expression of the Ab on the surface, binding to the biotinylated TOI, washing, FACS sorting, and expansion of the isolated bound yeast cells [19]. Again, secondary fluorescent-conjugated detection reagents for FACS should be constantly alternated to avoid enrichments of cross-reactive clones that could bind to them. Moreover, incubation volumes and the number of yeast cells stained should be chosen to ensure the number of TOI molecules to be tenfold in excess over the number of surface expressed Abs. When selecting for very highaffinity binders, equilibrium screening might become impractical because the required reaction volumes are too large. Kinetic dissociation competition can be used to overcome this problem [17]. Finally, more stringent gates (collect 340 μL: • X μL vector DNA. • 430 – X μL water. • 50 μL 10 green FD buffer. • 10 μL FastDigest NheI. • 10 μL FastDigestD HindIII. 10. Incubate at 37C for 90 min. 11. During incubation, prepare a 100 mL 1% agarose gel with 1: 10,000 Gel Red in a wide gel box with a single comb. 12. Run the entire sample on gel at 110 V for 30–40 min or until adequate separation of vector and insert bands is achieved. 13. Extract the large vector band into a 15-mL conical tube. 14. Extract DNA from gel using 4–6 columns of a gel extraction kit. Elute DNA and pool. 10 μg of vector DNA is optimal for library creation. 3.2.5 Library Creation

Yeasts are transformed using the insert and vector DNA generated above, following a strategy similar to that described by Chao et al. [6]. 1. Isolation streak EBY100 yeast on a YPD plate. Grow 2 days at 30  C. 2. Inoculate 4–6  4 mL YPD with swipes of EBY100 colonies. Grow overnight at 30  C. 3. In the morning, inoculate 2  100 mL YPD to an optical density (OD600) of 0.2 in sterile 250-mL baffled flasks. 4. Grow at 30  C, shaking at 250 rpm, until OD600 ¼ 1.5–1.67. This usually takes 4–6 h. 5. While waiting, chill centrifuges and ~ 25–30 cuvettes. Prepare and filter the following: (a) 500 mL E buffer (chilled to 4  C). (b) 10 mL 1 M Lithium Acetate (LiAc).

Yeast-Displayed Peptide-MHC Libraries

279

(c) 1 L SDCAA. (d) In the last hour, 3 mL 1 M Tris-DTT--must be fresh. 6. When OD600 measures 1.5–1.67, add 1 mL Tris-DTT, 5 mL LiAc, and 200 μL 0.5 M EDTA to each flask. 7. Shake at 250 rpm, 30  C for 15 min. 8. While the flasks are shaking, combine 50 μg insert with 10 μg vector. Top up to 500 μL with cold E buffer, and sterile filter the DNA solution. 9. Divide the cultures into four 50-mL conical tubes. 10. Spin at 2500  g for 3 min at 4  C. Dispose of supernatants. 11. Keeping the cells on ice, resuspend each pellet in 50 mL cold E buffer. 12. Spin at 2500  g for 3 min at 4  C. Dispose of supernatants. 13. Resuspend two pellets with 25 mL cold E buffer each and combine into two conical tubes (ending with two tubes instead of four). 14. Spin at 2500  g for 3 min at 4  C. Dispose of supernatants. 15. Resuspend one pellet with 10 mL cold E buffer and use to resuspend both pellets. 16. Spin at 2500  g for 3 min at 4  C. Aspirate supernatant. 17. Resuspend pellet fully in filtered DNA solution from step 8. 18. Taking 50 μL at a time (see Note 10) and ensuring the yeasts are well mixed, electroporate in chilled cuvettes with settings 500 V (LV mode), no resistance, 25 μF capacitance. Rescue with 2  1 mL YPD per cuvette into a fresh 250 mL sterile baffled flask. Time constants should be ~40–60 ms. 19. Shake at 250 rpm, 30  C for 1 h after rescue. Allow four SDCAA plates to warm up. 20. Pour into one 50-mL tube and spin at 2500  g for 3 min at 4  C. Aspirate supernatant. 21. Resuspend in 10 mL SDCAA. 22. Titer onto four warm SDCAA plates. Change pipette tip for each dilution. (a) 104: 990 μL SDCAA +10 μL library (from 10 mL in step 21). Mix well. (b) 105: 900 μL SDCAA +100 μL 104 above, mix well. (c) 106: 900 μL SDCAA +100 μL 105 above, mix well. (d) 107: 900 μL SDCAA +100 μL 106 above, mix well. 23. Spread 100 μL of dilution onto each plate. Incubate for 2–3 days at 30  C until colonies are large enough to count.

280

Brooke D. Huisman et al.

24. Pour the remaining library into a sterile 2.8-L baffled flask with 490 mL SDCAA, using some SDCAA to wash the conical tube. Shake at 250 rpm, 30  C. 25. The following day, split the library flask 1:6 by pouring ~100 mL of library into the bottle containing ~500 mL SDCAA. Pour out the remaining library and pour the split library back into the flask. 26. 2–3 days after library creation, check the plates. Count colonies on the 106 and 107 plates. Use the smaller count as library size (e.g., if there are 12 colonies on the 107 plate and 150 colonies on the 106 plate, record the library size as 1.2  108). See Note 11 regarding library size. 27. 1–2 days after the library is split, remove it from the shaker and save 450-mL conical tubes of the library. Store at 4  C labeled with titer and OD for up to a month (see Note 12 for longterm storage tips). Remaining library can be induced immediately for use in selections or discarded. 3.3 Conducting Selections

Once a library of adequate size has been created, selections can be performed using the library. First, the library must be grown and induced to express the pMHC on the yeast surface.

3.3.1 Inducing the Library

1. If starting from a freshly made or freshly thawed and passaged library, skip to step 5. 2. If starting from yeast stored at 4  C, check the OD and calculate the volume needed to obtain 10–20 more yeasts than the size of the library to ensure no overall library diversity is lost via sampling (e.g., if the library is 1  108, start selections on 1–2  109 yeasts). 3. In a sterile baffled flask, dilute this amount of yeast down to OD 1 in SDCAA. 4. Grow for 1 day at 30  C, shaking at 250 rpm. 5. Check the OD. Beginning with at least 10–20 the number of yeasts as the titered library size, grow in SGCAA for 2–3 days at 20  C, shaking at 250 rpm.

3.3.2 Assessing Induction

1. Use 200 μL of induced yeast culture for an unstained control and 200 μL to stain with the relevant epitope tag or β2M antibody. 2. Spin at 5000  g for 1 min at 4  C. Remove supernatant and wash with 500 μL of FACS buffer. 3. Resuspend in 50 μL FACS buffer +1:50 dilution of antibody. Incubate for 20 min at 4  C. 4. Wash twice with FACS buffer.

Yeast-Displayed Peptide-MHC Libraries

281

Fig. 4 Sample selection data for TCR deorphanization. (a) Myc epitope tag staining shows enrichment over successive rounds of selection (gated on unstained yeast). (b) Selection scheme and data for five rounds of selection on a randomized 9mer library displayed by mouse MHCI H-2Db. In this example, “spot check” of a dozen colonies following round 3 indicated enrichment of three peptides

5. Run samples on a flow cytometer to assess percent of yeasts that are induced (displaying protein on surface). For an initial library, 15–50% induction is typical. If much lower, something may be wrong with expression. Much higher may suggest a contaminant. Next, design the selection strategy. If the goal is to select for peptide binders of a given TCR, a general strategy would be to do three rounds of selection with TCR on streptavidin beads and then increase the stringency by switching to tetramers of TCR for rounds 4+. Additional considerations might include tetramer concentration (lower to increase stringency further), negative selections with another TCR for which no binding is expected or desired (to select for specificity), or rounds of selections with multiple TCRs for which peptide cross-reactivity is desirable. Bead-based selections are conducted using streptavidin-coated magnetic microbeads which are flowed through a column attached to a magnet. Sample selection data is shown in Fig. 4. 3.3.3 Starting Selections: Removing Non-specific Binders

Especially in early rounds of selection, non-specific binders should be removed from the library by incubating yeast with streptavidin beads only. Any yeast that binds non-specifically to the beads will be retained in the column, and the TCR-specific selection can be performed on the cleared yeast. 1. To begin selections, check the OD600 of the induced library. Calculate the volume of culture needed to examine 10-20

282

Brooke D. Huisman et al.

yeasts as compared to the starting library size. Spin down this amount of yeast at 5000  g for 1 min at 4  C. 2. Wash with 10 mL of FACS buffer. Spin at 5000  g for 1 min at 4  C. 3. Resuspend in 5 mL and transfer to 15-mL conical tube. 4. Add 250 μL streptavidin beads. Place on a rotator at 4  C for 1 h. 5. Spin down (do not decant yet) and keep on ice. 6. In the cold room, place an LS column on the magnet stand. Place a 15-mL conical tube below. Add 5 mL FACS buffer to the column for equilibration. Dump this out from the 15 mL conical tube once it runs through the column. 7. Decant yeast tube and resuspend in 5 mL FACS buffer. Load solution onto the column. 8. Once all of the liquid has flowed through the column, add 3 mL FACS buffer. Repeat twice more for a total of 3  3 mL washes. 9. At this point, any yeast that is non-specifically bound to the streptavidin beads is magnetically trapped in the column. The rest of the yeasts are in the 15-mL conical tube (cleared library). Cap and save the tube of the cleared library—these are the yeasts to continue working with for TCR-specific selection. 10. Remove the LS column and place it into a new 15-mL conical tube. Add 5 mL FACS buffer and use the plunger to force the liquid through the column. Take a 500 μL sample and run it on the flow cytometer to count how many non-specific binders are present. Record this number for future reference to compare specific and non-specific enrichment. 3.3.4 TCR-Specific Selection: Round 1

1. With the 15-mL conical tube of cleared yeast, spin at 5000  g for 1 min at 4  C. Resuspend the yeast in 5 mL of FACS buffer. Keep on ice. 2. Incubate 250 μL streptavidin beads with protein of interest to bring the beads to a final biotinylated protein concentration of 400 nM. The degree of biotinylation can be assessed through various methods, such as a gel shift assay [29], and a low degree of biotinylation requires more total protein in selections. For example, if a TCR is only 50% biotinylated, calculate the amount of protein for 400 nM and then use double that amount to account for incomplete biotinylation. Mix well and incubate for 5 min on ice. 3. Add the beads to the resuspended yeast. Rotate at 4  C for 2–3 h.

Yeast-Displayed Peptide-MHC Libraries

283

4. Repeat steps 5–8 described for clearing non-specific binders above (Subheading 3.3.3). Now, the yeasts that flowed through the column are undesirable yeasts—they did not bind the TCR. Cap and remove the 15-mL conical tube. 5. The yeasts that have putatively bound to the protein of interest are magnetically bound to the column. Elute from the column as above (Subheading 3.3.3, step 10) and save the elution containing the selected yeast. Take a 250 μL sample and run on a flow cytometer to count selected yeast. 6. Spin down the selected yeast at 5000  g for 1 min at 4  C. Discard the supernatant and resuspend in 5 mL SDCAA to serve as a wash to remove residual FACS buffer. 7. Spin at 5000  g for 1 min at 4  C. Discard supernatant and resuspend in 3 mL SDCAA. Transfer to a 14-mL culture tube and grow at 30  C overnight, shaking at 250 rpm. 8. After growing overnight, if continuing to further rounds of selection immediately (see Note 13), check OD and induce 10-20 more yeasts than were selected in round 1. Induce in SGCAA at 20  C for 2–3 days, shaking at 250 rpm. 3.3.5 TCR-Specific Selection: Rounds 2–3

When checking induction, the percent of yeast that stain with an epitope tag antibody typically increases in successive rounds of selection (see Fig. 4). The number of non-specific binders should decrease in further rounds (see Note 14). 1. For rounds 2–3, use the protocol above (Subheadings 3.3.3 and 3.3.4) but with the yeast in 500 μL instead of 5 mL for incubation with beads, and use a 50 μL volume of streptavidin beads instead of 250 μL. 2. Continue to start with at least 10–20 more yeast than were selected in the previous round. If that calculated number is impractically small to manipulate, there is no detriment in using a larger number of yeast (such as 3  107, which can be cultured in 3 mL). After round 2, it may take more than 24 h for yeast to grow back up in SDCAA if the number selected was low.

3.3.6 TCR-Specific Selection: Rounds 4+

We find that for most selections, if a library contains pMHC binders, enrichment will be apparent after round 3 at latest. There are also often significant numbers of pMHCs that may specifically interact with highly avid TCR-coated microbeads yet not with lower-avidity multimers such as tetramers. Therefore, for round 4, consider increasing the stringency of selection by incubating the yeasts with tetramers of TCR rather than streptavidin beads to separate out pMHC sequences with higher affinity. Note that since the physiological affinity of the pMHC–TCR interaction is low (~1–50 μM), few if any library sequences would be expected to

284

Brooke D. Huisman et al.

bind if monovalent TCRs were used for selection. Tetramerpositive yeasts are enriched using magnetic beads decorated with antibodies to the tetramer fluorophore. Alternatively, for small libraries, these selections could be conducted using fluorescenceactivated cell sorting. 1. Create tetramers of TCR by making a 100 μL solution of 500 nM fluorescently labeled tetrameric streptavidin and 2.5 μM biotinylated TCR in FACS buffer (5:1 TCR:tetrameric streptavidin ratio, ensuring complete tetramer formation with margin for error). Incubate for 5–10 min at 4  C to allow tetramers to assemble. 2. Wash starting yeasts with FACS buffer; spin down and remove supernatant. Resuspend in 100 μL of tetramer solution. Place on a rotator at 4  C for 2–3 h in the dark. 3. Wash twice with 500 μL of FACS buffer. 4. Resuspend in 500 μL FACS buffer. Take a 50 μL sample to run on a flow cytometer and assess the percentage tetramer-positive yeast pre-column. 5. Add 50 μL anti-fluorophore beads to the remaining 450 μL of yeasts. Incubate for 20 min at 4  C. 6. Run selection on LS column as described above for other rounds of selection (Subheading 3.3.3, steps 5–8). Elute selected yeasts in 5 mL FACS buffer (similar to Subheading 3.3.3, step 10). Run a 250 μL sample on a flow cytometer to count selected yeasts and assess tetramer-positive percentage— there should be some enrichment of tetramer-positive yeasts compared to the pre-column sample. 7. Wash yeasts with 5 mL SDCAA. Resuspend in 3 mL SDCAA and grow overnight as above. 8. For further rounds, tetramer concentration can be lowered to increase selection stringency even further. It may be helpful to titrate tetramer on samples of yeasts that have grown up following the previous round of selection to determine an optimal tetramer concentration. 3.3.7 Growing and Sequencing Selected Yeasts: Individual Yeast Colonies

Once there is an enriched population of TCR-binding yeasts, it can be advantageous to sequence individual yeast clones to estimate the sequence diversity of the enriched pool and to ensure that the sequences are well behaved (in frame, predicted to bind to MHC, and not contaminants from previous experiments). To “spot check” hits before deep sequencing, individual yeast colonies can be grown to test tetramer staining or to sequence the peptides (see Note 15 for troubleshooting). An alternative is described in Subheading 3.3.8, which provides peptide sequences faster though does not allow for tetramer staining.

Yeast-Displayed Peptide-MHC Libraries

285

1. After the yeasts have grown up post-selection, plate 20 μL on an SDCAA plate and isolation streak. Grow 2–3 days at 30  C until colonies are visible. 2. Add 1 mL SDCAA into wells of a 2-mL 96-well block. Pick colonies from the plate and add to wells. Cover with breathable film. 3. Grow at 30  C for 1 day. 4. From here, either or both of these steps can be performed: (a) Option 1: Conduct a yeast miniprep on 100 μL of culture as described by manufacturer. Transform 10 μL elution in DH5α bacteria. Grow overnight at 37  C with 1:1000 dilution of carbenicillin. Miniprep bacteria and submit DNA for sequencing to see peptide sequences. (b) Option 2: spin plate, dump supernatants, and resuspend yeasts in SGCAA. Shake at 20  C at 250 rpm for 2–3 days. Test individual colonies for tetramer staining. 3.3.8 Growing and Sequencing Selected Yeasts: Bacteria Colonies

As an alternative to growing and sequencing individual yeast colonies (Subheading 3.3.7), individual bacteria colonies transformed with plasmid purified from the enriched pool of yeast binders can be sequenced. 1. After the yeasts have grown up post-selection, take a sample with 10 times the amount selected and perform a yeast mini prep as described by the manufacturer. 2. Transform into DH5α bacteria and plate. 3. Pick a dozen or so colonies and grow overnight in 3 mL LB with 1:1000 dilution of carbenicillin. 4. Mini prep and send for sequencing. 5. If desired, DNA can be transformed back into yeast for further tetramer staining or analysis. Transformation of DNA into yeast was described above in Subheading 3.1.2.

3.4 Preparing, Processing, and Analyzing NGS Data

After three rounds of TCR-based selections, the peptide library is likely highly converged such that sequencing individual colonies will provide a sense of what is in the library. However, depending on the diversity and size of the starting library, next-generation sequencing (NGS) may provide important insight into the content of the library, including identifying low-abundance sequences that may bind with lower affinity, covariation between residues, and relative enrichment of peptide sequences. Sequencing can be performed on the Illumina MiSeq (Illumina Incorporated; Irvine, CA), which can accommodate limited-diversity amplicon sequencing while providing sufficient coverage for an enriched yeast display library with ~ten million reads.

286

Brooke D. Huisman et al.

3.4.1 Preparing Sample for NGS

To prepare the sample for NGS, first perform a yeast miniprep to extract plasmid DNA from yeast, following manufacturer kit instructions (see Note 15 for troubleshooting). Sequencing each round of selection—not just the final round—is recommended to gain insight into the convergence of the library as selections progress. Ensure that the number of yeasts utilized in the miniprep exceeds the expected diversity in that round. The exception is with the unselected library, which is frequently intentionally not sequenced to full depth. For example, if the library starts at 100 million members, sequencing to full depth would require more reads than a standard MiSeq run. Additionally, because each individual member is present at low frequency, it will be challenging to distinguish between true members and PCR errors. Typically, including 10–50 million yeasts into the mini prep for each round is sufficient. In a standard NGS prep, PCR primers are designed to incorporate three components: sequence matching desired sequencing primers, i5 and i7 paired-end handles, and an index barcode unique to each sample being included in the multiplex (e.g., one index barcode for each round of selection). Depending on the sequencer, it may also be beneficial to include a short randomized sequence for cluster identification on Illumina sequencers [30]. These additions are usually made in two rounds of PCR as illustrated in Fig. 5. Typically, a 150 + 150 paired-end read will sufficiently cover the peptide sequence, although the primer positions can be moved to cover additional sequences of interest, such as signal peptides, flanking linkers, or epitope tags.

3.4.2 Processing NGS Data

Paired end reads can be assembled utilizing one of many available tools, such as FLASH [31] or PandaSeq [32], and can increase sequence fidelity by correcting for sequencing errors. Additional filtering steps include removing sequences with errors in regions flanking the peptide or in constant regions of a partially randomized peptide. Additionally, the most probable library contaminants are the constructs previously used for validation, which may need to be excluded.

i5 Anchor

F Primer

NNNNNN

Peptide

NNNNNN

R Primer

BC

i7 Anchor

Fig. 5 Amplicon sequencing formatting. Short randomized sequences for cluster identification are indicated as “NNNNNN”. F primer Forward NGS primer, R primer Reverse NGS primer, BC Index Barcode, “N” indicates any nucleotide

Yeast-Displayed Peptide-MHC Libraries

287

During each NGS processing step, low-frequency errors may arise, such as polymerase misincorporation during PCR amplification. Because these errors are typically stochastic and low frequency, they can be identified by processing each sample in replicate and comparing peptide presence and frequency across replicates. 3.4.3 Analyzing NGS Data

4

From the sequencing run, there are two pieces of information with which to understand the library: peptide sequences and the read count for each peptide. TCR selection data is frequently highly hierarchical, with ~10% of the peptides that are present after round 3 of selection accounting for >90% of the sequencing counts. As a result, read counts may be important to fully understanding the peptide motifs. Additionally, contaminating sequences and noise will account for some low-frequency peptides in the NGS data. Noise can be filtered by excluding peptides with fewer than a set number of reads. An arbitrary read count may prove helpful (e.g., greater than 10 reads), although read count cutoffs can be rationally set by considering factors like known noise such as stop-codon containing peptides. While it is not uncommon for the stop-codon template plasmid to be a contaminant at higher frequency, stopcodon sequences can be more frequently found at low read counts. We have found that the false-positive rate is low for pMHCTCR binders. For inferring negative binders, the opposite is true: a peptide that did not enrich may be absent due to stochastic dropout in early rounds of selection when each member is present at low frequency or due to poor processing in yeast such as incorrect signal peptide cleavage. When extending yeast display-enriched peptides to T cell phenotypic assays, it is important to note that a peptide that binds to a TCR of interest will not always result in T cell activation or phenotypic response [33]. Additionally, yeast display does not incorporate peptide processing pathways present in mammalian cells, meaning a peptide identified through yeast display may not be a natural ligand. The data resulting from sequencing these yeasts can be used for many exciting applications, such as identifying natural TCR ligands [2, 4], determining peptide mimotopes [12], characterizing TCR motifs [2, 13], and training prediction algorithms [2, 4, 13, 34].

Notes 1. Restriction sites used for cloning may differ depending on the construct design and/or epitope tags used (e.g., for a construct including a Myc epitope tag, there is a HindIII site within the Myc tag used for cloning).

288

Brooke D. Huisman et al.

2. As described by Chao et al. [6], SDCAA and SGCAA can also be prepared with sodium phosphate dibasic and sodium phosphate monobasic. However, use of sodium citrate and citric acid monohydrate instead are favorable for reducing the growth of bacterial contaminants. We have additionally noted that for some MHC constructs, foldedness is improved via low pH induction. 3. If there is persistent contamination in yeast cultures, ensure good aseptic technique. It may be helpful to add PenicillinStreptomycin at 1:100 to SDCAA and SGCAA media. Following the addition of Penicillin-Streptomycin, store media at 4  C. Filter tips may also reduce the frequency or severity of contamination. 4. When designing MHCI constructs, we typically use the following linker lengths: 3xGGGGS between peptide and β2M, 4xGGGGS between β2M and MHCI heavy chain, and an epitope tag followed by 3xGGGGS between MHCI heavy chain and Aga2. Similarly for MHCII: GSGS connecting MHCIIɑ to an epitope tag and P2A, GGSGGG, epitope tag, GG–optional protease site–GGSG between peptide and MHCIIβ (protease cut site can be used for peptide-MHC binding selections as done by Rappazzo et al. [13]), and GS, epitope tag, 3xGGGGS between MHCIIβ and Aga2. If primers may need to bind to part of a linker sequence during any cloning, we suggest varying the DNA sequences of the linkers to avoid multiple primer binding sites. 5. Signal peptide cleavage can impact peptide preferences, specifically by causing the absence of proline as peptide position 1, which may be improperly cleaved from an N-terminally adjacent signal peptide. 6. To prepare electrocompetent yeast, follow steps 1–17 in Subheading 3.2.5 with the following changes: in step 3, inoculate a 1 L YPD culture in a sterile 2.8-L baffled flask; in step 5, prepare 500 mL E buffer and 10 mL 2 M DTT in Tris; in step 6, add only 10 mL Tris-DTT per liter of cells; in step 9, use 500-mL centrifuge tubes; in step 17, resuspend in 15 mL E buffer, aliquot 50 μL/tube in sterile Eppendorf tubes, and slow freeze to 80  C. 7. Reference primer sequences for a randomized MHCI 9mer peptide library: (a) Forward—PCR 1 (peptide randomization): C A A T A T T T T C T G T T A T T G C TAGCGTTTTGGCTNNKNNKNNKNNKNNKNNKN NKNNKNNKGGTGGAGGAGGTTCTGGAGGTG (b) Forward—PCR 2 (scale up) (homology past NheI site): TTCAATTAAGATGCAGTTACTTCGCTGTTTTT CAATATTTTCTGTTATTGCTAGCGTTTTGGCT

Yeast-Displayed Peptide-MHC Libraries

289

8. For library preparation, if DNA yields are low, a library can be made with less DNA, although library size will decrease. For example, we have successfully made a 1  107 library using 5 μg of insert DNA. Proportionally scale the amount of vector DNA used. 9. Troubleshooting scale-up PCR: one common issue for the scale-up PCRs is a non-specific PCR product that appears on the quality control gel. If there is a non-specific band, optimization of PCR conditions such as altering PCR melting temperatures or DMSO concentration may remove the non-specific band; we recommend performing these optimizations at a small scale (50 μL reaction). If PCR optimizations fail, we recommend gel purifying all of the DNA, rather than PCR purifying, to remove the non-specific amplification product. If there are challenges scaling a 50 μL reaction to 5 mL, a medium scale (~1 mL) reaction may be successful instead. 10. When aliquoting yeast for electroporation, we have found aliquoting individual cuvettes for immediate electroporation or aliquoting 5–10 cuvettes at a time for subsequent electroporation works better than aliquoting all yeasts into cuvettes before beginning electroporation. Time constants are more consistent when the yeasts are well mixed and do not settle much before electroporation. 11. Regarding library size: for a fully randomized peptide MHCI library for TCR deorphanization, our goal is typically a library size of at least 1  108, which can be achieved with the protocol in Subheading 3.2. Other libraries may not need to be so large; for example, a mimotope library with four peptide residues randomized would include 204 ¼ 160,000 theoretical peptides, so even a library size of 1  107 would provide excess coverage. 12. Yeast grown in SDCAA may be stored at 4  C. For long-term storage of a yeast library, freeze an oversampling of the library. Protocols are available for long-term yeast storage and revival: see Supplementary Methods by Chao et al. [6] or Chap. 2 of this volume. 13. Yeasts are induced most robustly after growing for at least a day. For library selections, we suggest inducing yeasts in SGCAA directly after they have been growing in SDCAA. Thus, if pausing between rounds of selection, uninduced yeasts can be stored at 4  C, but reculture the library in SDCAA media for a day before inducing in SGCAA. 14. In selection rounds 3+, it may not be necessary to continue removing non-specific binders first, as there should be very few. However, to do so in tetramer rounds, yeast can be incubated first in 100 μL of 500 nM fluorescently labeled tetrameric

290

Brooke D. Huisman et al.

streptavidin, then with anti-fluorophore beads, and then run on an LS column. The cleared library can be incubated with TCR tetramers as described above for TCR-specific selections in tetramer rounds. 15. Troubleshooting yeast mini prep: one common challenge in performing colony sequencing of post-selection libraries is having no colonies appear after transforming the yeast mini prep into DH5α bacteria. Note that yeast genomic DNA will make the plasmid DNA from a yeast mini prep impure, so it is important to amplify in bacteria and to use most or all of the yeast mini prep output DNA. When eluting from the yeast mini prep column, allow the water to sit on the column for several minutes. If there continue to be no or few colonies, consider varying the number of yeasts utilized in the yeast mini prep.

Acknowledgments We would like to thank K. Christopher Garcia (Stanford University) for generous sharing of reagents and Christine Devlin for aiding in the creation of our yeast display protocols. This work was supported by National Science Foundation Graduate Research Fellowships to B.D.H, B.E.G., and P.V.H., and a Melanoma Research Alliance grant, the AACR-TESARO Career Development Award for Immuno-oncology Research (17-20-47-BIRN), Schmidt Futures, and the National Institutes of Health (P30CA14051 and 5U19AI110495) to M.E.B. References 1. Matsui K, Boniface J, Reay P et al (1991) Low affinity interaction of peptide-MHC complexes with T cell receptors. Science 254:1788–1791 2. Birnbaum ME, Mendoza JL, Sethi DK et al (2014) Deconstructing the peptide-MHC specificity of T cell recognition. Cell 157: 1073–1087 3. Wooldridge L, Ekeruche-Makinde J, van den Berg HA et al (2012) A single autoimmune T cell receptor recognizes more than a million different peptides*. J Biol Chem 287:1168– 1177 4. Gee MH, Han A, Lofgren SM et al (2018) Antigen identification for orphan T cell receptors expressed on tumor-infiltrating lymphocytes. Cell 172:549–563.e16 5. Adams JJ, Narayanan S, Liu B et al (2011) T cell receptor signaling is limited by docking geometry to peptide-major histocompatibility complex. Immunity 35:681–693

6. Chao G, Lau WL, Hackel BJ et al (2006) Isolating and engineering human antibodies using yeast surface display. Nat Protoc 1:755–768 7. Ramachandiran V, Grigoriev V, Lan L et al (2007) A robust method for production of MHC tetramers with small molecule fluorophores. J Immunol Methods 319:13–20 8. Boder ET, Dane Wittrup K (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol 15:553–557 9. Hansen T, Lawrence Yu YY, Fremont DH (2009) Preparation of stable single-chain trimers engineered with peptide, β2 microglobulin, and MHC heavy chain. Curr Protoc Immunol 87 10. Lybarger L, Lawrence Yu YY, Miley MJ et al (2003) Enhanced immune presentation of a single-chain major histocompatibility complex class I molecule engineered to optimize linkage

Yeast-Displayed Peptide-MHC Libraries of a C-terminally extended peptide. J Biol Chem 278:27105–27111 11. Pedersen LØ, Stryhn A, Holtet TL et al (1995) The interaction of beta 2-microglobulin (β2m) with mouse class I major histocompatibility antigens and its ability to support peptide binding. A comparison of human and mouse β2m. Eur J Immunol 25:1609–1616 12. Fernandes RA, Li C, Wang G et al (2020) Discovery of surrogate agonists for visceral fat Treg cells that modulate metabolic indices in vivo. Elife 9:e58463 13. Rappazzo CG, Huisman BD, Birnbaum ME (2020) Repertoire-scale determination of class II MHC peptide binding via yeast display improves antigen prediction. Nat Commun 11:4414 14. Almagro Armenteros JJ, Tsirigos KD, Sønderby CK et al (2019) SignalP 5.0 improves signal peptide predictions using deep neural networks. Nat Biotechnol 37:420–423 15. Kall L, Krogh A, Sonnhammer ELL (2007) Advantages of combined transmembrane topology and signal peptide prediction—the Phobius web server. Nucleic Acids Res 35: W429–W432 16. von Heijne G (1986) A new method for predicting signal sequence cleavage sites. Nucleic Acids Res 14:4683–4690 17. Sussman JL, Lin D, Jiang J et al (1998) Protein data Bank (PDB): database of threedimensional structural information of biological macromolecules. Acta Crystallogr D Biol Crystallogr 54:1078–1084 18. Kaas Q (2004) IMGT/3Dstructure-DB and IMGT/StructuralQuery, a database and a tool for immunoglobulin, T cell receptor and MHC structural data. Nucleic Acids Res 32:208D– 210D 19. O’Brien C, Flower DR, Feighery C (2008) Peptide length significantly influences in vitro affinity for MHC class II molecules. Immunome Res 4:6 20. Zavala-Ruiz Z, Strug I, Anderson MW et al (2004) A polymorphic pocket at the P10 position contributes to peptide binding specificity in class II MHC proteins. Chem Biol 11:1395– 1402 21. Lovitch SB, Pu Z, Unanue ER (2006) Aminoterminal flanking residues determine the conformation of a peptide-class II MHC complex. J Immunol 176:2958–2968

291

22. Rossjohn J, Gras S, Miles JJ et al (2015) T cell antigen receptor recognition of antigenpresenting molecules. Annu Rev Immunol 33: 169–200 23. Wieczorek M, Abualrous ET, Sticht J et al (2017) Major histocompatibility complex (MHC) class I and MHC class II proteins: conformational plasticity in antigen presentation. Front Immunol 8:292 24. Andreatta M, Nielsen M (2016) Gapped sequence alignment using artificial neural networks: application to the MHC class I system. Bioinformatics 32:511–517 25. Nielsen M, Lundegaard C, Worning P et al (2003) Reliable prediction of T cell epitopes using neural networks with novel sequence representations. Protein Sci 12:1007–1017 26. Vita R, Mahajan S, Overton JA et al (2019) The immune epitope database (IEDB): 2018 update. Nucleic Acids Res 47:D339–D343 27. Rammensee H, Bachmann J, Emmerich NP et al (1999) SYFPEITHI: database for MHC ligands and peptide motifs. Immunogenetics 50:213–219 28. Falk K, Ro¨tzschke O, Stevanovic´ S et al (1991) Allele-specific motifs revealed by sequencing of self-peptides eluted from MHC molecules. Nature 351:290–296 29. Fairhead M, Howarth M (2015) Site-specific biotinylation of purified proteins using BirA. Methods Mol Biol 1266:171–184 30. Mitra A, Skrzypczak M, Ginalski K, Rowicka M (2015) Strategies for achieving high sequencing accuracy for low diversity samples and avoiding sample bleeding using illumina platform. PLoS One 10:e0120520 31. Magocˇ T, Salzberg SL (2011) FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 27:2957– 2963 32. Masella AP, Bartram AK, Truszkowski JM et al (2012) PANDAseq: paired-end assembler for illumina sequences. BMC Bioinform 13:31 33. Sibener LV, Fernandes RA, Kolawole EM et al (2018) Isolation of a structural mechanism for uncoupling T cell receptor signaling from peptide-MHC binding. Cell 174:672–687.e27 34. Dai Z, Huisman BD, Zeng H et al (2021) Machine learning optimization of peptides for presentation by class II MHCs. Bioinformatics 37(19):3160–3167. https://doi.org/10. 1093/bioinformatics/btab131

Chapter 16 Yeast Display Guided Selection of pH-Dependent Binders Jenna N. Meanor, Albert J. Keung, Balaji M. Rao, and Nimish Gera Abstract pH-dependent antigen binding has proven useful in engineering next-generation therapeutics specifically via antibody recycling technology. This technology allows for half-life extension, thereby lowering the amount and frequency of dosing of therapeutics. Cell sorting, coupled with display techniques, has been used extensively for the selection of high-affinity binders. Herein, we describe a cell sorting methodology utilizing yeast surface display for selection of binding proteins with strong binding at physiological pH and weak to no binding at acidic pH. This methodology can be readily applied to engineer proteins and/or antibodies that do not have pH-dependent binding or for selection of de novo pH-dependent binders using library-based methods. Key words Yeast surface display, pH-dependent binding, Cell sorting, Histidine scanning, Antibody recycling, Half-life extension

1

Introduction Naturally occurring pH-dependent protein–protein interactions regulate biological functions by modulating the fate of receptors and their cargoes [1–3]. This characteristic has been engineered in proteins and antibodies for a variety of biomedical and biotechnological applications such as extending the half-life of therapeutics and generating pH-dependent affinity chromatography reagents [4–8]. High-affinity therapeutics, many of which are antibody based, are limited in their efficacy by the number of antigen molecules they can bind and get consumed in the process themselves [9]. Antibody recycling technology has been used for half-life extension in which antibody-antigen complexes enter the acidic endosomal environment and engineered pH-dependent antibodyantigen binding allows for antibody dissociation and recycling to the cell surface via interaction with neonatal Fc-receptor (FcRn) [9– 13]. This enables less frequent and/or lower dosing of these therapeutics.

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_16, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

293

294

Jenna N. Meanor et al.

Yeast surface display (YSD) is a well-established platform for protein and antibody engineering that utilizes eukaryotic expression, allowing proper folding of human-derived proteins [14]. Briefly, the protein of interest is expressed as a fusion to a yeast cell wall protein linking its phenotype to the corresponding genetic sequence. This platform can be used to generate and screen libraries of up to 1010 variants and surface expression can be easily controlled through the galactose-inducible GAL1 promoter [15– 17]. Screening for desired properties is mainly achieved through a combination of magnetic-activated cell sorting (MACS) and fluorescence-activated cell sorting (FACS) [18–22]. A variety of proteins, antibodies, and antibody fragments have been displayed on the yeast surface to generate libraries for selection of high-affinity variants; however, selection of pH-dependent binding proteins using YSD has only been demonstrated recently [4, 5, 7, 8, 23]. This can be accomplished by either (a) selection of de novo pH-dependent binding proteins from a naive/synthetic library or (b) generating a mutagenesis library using an existing “parent” binding protein to the antigen of interest. Library generation methodologies have been extensively described for naive/ synthetic libraries in the literature [15, 24–28]. However, libraries utilizing an existing binding protein to engineer pH-dependent binding have been described sparingly [5, 29]. One of the most common methods for introducing pH-dependent binding in an existing protein/antibody is through site-directed mutagenesis and the introduction of histidine residues along the binding interface (Fig. 1a). Histidine has a pKa of ~6.0 and becomes protonated upon endocytosis into the acidic environment (pH ~5.5–5.8) resulting in a structural transition due to altered electrostatic interactions [9]. In antibodies, histidine residues are commonly substituted into the complementaritydetermining regions (CDRs) [10, 23, 29] although it has been shown that the introduction of ionizable residues in the protein core can also lead to pH-dependent conformational changes [7, 8]. Libraries can also be generated through direct cloning from immunized animals (Fig. 1b); however, the feasibility of isolating a naturally occurring pH-dependent antibody is rare; generally, less than 5% of binders have a natural pH-dependency [9]. A more all-encompassing combinatorial library can be generated through random mutagenesis of either the binding interface or entire protein of interest through the introduction of triphosphate nucleoside analogues [7, 30] (Fig. 1c). This approach allows for the highest library diversity and introduction of other potentially pH-dependent amino acid substitutions. A comprehensive protocol for generation of yeast displayed combinatorial libraries is described in the literature [15]. The method of screening described in this chapter can be applied to either commercial libraries or libraries specifically

Yeast Display Guided Selection of pH-Dependent Binders

a

295

c

b

Original DNA Denaturation

CDR3 N

CDR1 D

B C‘

F A G

Primer Annealing

CDR2

Extension

E C

C‘’

8-oxo-dGTP dPTP

C

Histidine scanning

Direct cloning

Random mutagenesis

Fig. 1 Library construction technologies: (a) site-specific introduction of histidine residues, (b) direct generation of an antibody library via antigen immunization of a host species (created with BioRender.com) and (c) random mutagenesis of a protein of interest through error-prone PCR

generated for selection of pH-dependent binding proteins. An overview of this selection process is shown in Fig. 2. An initial round of MACS is performed to select for variants that bind the antigen and is followed by a round of FACS to further select for full-length and higher affinity binding proteins. A second round of FACS is performed at an acidic pH to select for variants that lose binding to the antigen at low pH. The final two rounds of FACS are end-point assays where, in each assay, the variants are incubated with the antigen at physiological pH and subsequently exposed to an acidic environment to select for variants that rapidly dissociate at lower pH. Once a clonal population is obtained, variants are evaluated for affinity and pH dependency via flow cytometry. Select binding proteins can be produced recombinantly and assessed for binding, stability, and pH dependency via established biophysical characterization methods.

2

Materials

2.1 Strains and Libraries

1. S. cerevisiae strain Collection, ATCC).

EBY100

(American

Type

Culture

2. E. coli strain DH5α. 3. Frozen yeast library aliquot (see Note 1). 2.2 Media and Buffers

1. SDCAA media: Dissolve 20 g Dextrose, 6.7 g Difco Yeast Nitrogen Base, 5 g Bacto Casamino Acids, 5.4 g Na2HPO4, and 8.6 g NaH2PO4 * H2O in 1000 mL deionized H2O. Filter sterilize using a 0.22 μm bottle top filter. 2. SGCAA media: Dissolve 2 g Dextrose, 20 g Galactose, 6.7 g Difco Yeast Nitrogen Base, 5 g Bacto Casamino Acids, 5.4 g

296

Jenna N. Meanor et al.

YSD Library

MACS

- magnetic beads

- non-target antigens

+ antigen

FACS

POI

HA

c-Myc

pH 7.4

+ for all binders

+ for binders that lose binding at low pH

pH 14.0

pH 1.0

pH 6.0 pH 14.0

pH 1.0

Agα2p pH 7.4 Agα1

pH 14.0

pH 1.0

+ end-point assay

pH 6.0 pH 1.0

Wash pH 14.0

further characterization

pH dependent binder

Fig. 2 General workflow for isolation of pH-dependent binding proteins. Briefly, a yeast surface displayed library based on a protein of interest (POI) undergoes multiple rounds of MACS screenings to remove nonspecific binders and to select for specific binders to the target antigen from the library population. Following MACS, initial FACS sorting is employed to isolate full-length binders to the antigen in an affinitybased manner. A second round of FACS at a low pH will isolate binders that lose binding in an acidic environment and a third round of FACS allows incubation of binders at physiological pH followed by acidic pH to select for binders that rapidly dissociate from the antigen at low pH. Any of these steps can be repeated, as necessary

Na2HPO4, and 8.6 g NaH2PO4 * H2O in 1000 mL deionized H2O. Filter sterilize using a 0.22 μm bottle top filter. 3. PBSA 0.1%: Dissolve 1 g BSA in 900 mL deionized H2O. Add 100 mL 10 PBS. Adjust pH to 7.4 using 1 M NaOH and 1 M HCl. Filter sterilize using a 0.22 μm bottle top filter. 4. PBSA 0.1% pH 6.0: Dissolve 1 g BSA in 900 mL deionized H2O. Add 100 mL 10 PBS. Adjust pH to 6.0 using 1 M HCl. Filter sterilize using a 0.22 μm bottle top filter.

Yeast Display Guided Selection of pH-Dependent Binders

297

5. PBSA 1%: Dissolve 10 g BSA in 900 mL deionized H2O. Add 100 mL 10 PBS. Adjust pH to 7.4 using 1 M NaOH and 1 M HCl. Filter sterilize using a 0.22 μm bottle top filter. 6. SDCAA agar plates: Dissolve 182 g Sorbitol, 12 g Agar, 5.4 g Na2HPO4, and 8.6 g NaH2PO4 * H2O in 900 mL deionized H2O. Sterilize by autoclaving. Dissolve 20 g Dextrose, 6.7 g Difco Yeast Nitrogen Base, and 5 g Bacto Casamino Acids in 100 mL deionized H2O. Filter sterilize using a 0.22 μm bottle top filter. Combine both solutions and pour plates once cooled. 2.3

Reagents

1. Biotinylated target antigen (see Note 2). 2. Biotinylated closely related proteins. 3. Dynal biotin binder beads (Thermo Fisher Scientific). 4. 1 mg/mL Anti-c-Myc, chicken IgY fraction (Thermo Fisher Scientific) (see Note 3). 5. 2 mg/mL Goat anti-chicken antibody conjugated to Alexa Fluor 488 (Thermo Fisher Scientific). 6. 1 mg/mL Streptavidin-phycoerythrin (SA-PE) (Thermo Fisher Scientific). 7. 0.2 mg/mL Biotin Monoclonal Antibody (BK-1/39) PE conjugate (Thermo Fisher Scientific). 8. Zymoprep Yeast plasmid Miniprep II (Genesee Scientific).

2.4

Equipment

1. 1.7-mL microcentrifuge tubes. 2. 15-mL conical tubes. 3. DynaMag™-2 Magnet (Thermo Fisher Scientific). 4. DynaMag™-15 Magnet (Thermo Fisher Scientific). 5. Flow cytometer with sorting capabilities. 6. Plastic petri plates. 7. Baffled flasks (see Note 4). 8. Shaking incubators (30  C and 20  C). 9. Multi-tube Rotator (can fit 1.7-mL and 15-mL tubes).

3

Methods

3.1 Yeast Library Preparation and Display

The following steps should be followed immediately before any screening to ensure healthy expressing cells. When handling yeast libraries, it is important to maintain the full library diversity through each passage; always start new yeast cultures at an optical density at 600 nm (OD) of 1.0 with enough cells to cover at least 10 the library diversity (i.e., if the library diversity is 1  108,

298

Jenna N. Meanor et al.

passage at least 1  109 cells into each new culture). An absorbance value of 1.0 at 600 nm, referring to an OD of 1.0, corresponds to a density of roughly 1  107 cells/mL (see Note 5). 1. Thaw an aliquot of the frozen yeast library at room temperature and resuspend cells in an appropriate volume of SDCAA media for a final OD of ~1.0. 2. Expand the culture overnight at 30  C with shaking at 225 rpm until the cells reach the stationary phase, OD ~4.0–6.0. This will take approximately 20–24 h. Cells can be stored at 4  C for 1–2 weeks. 3. Spin down enough cells (3000  g, 2 min) to maintain 10 the library diversity and resuspend in SGCAA media to a final OD of ~1.0. 4. Incubate cells at 20  C for 16–20 h with shaking at 225 rpm for protein expression. 3.2 Selection of Binder Pool Through Magnetic-Activated Cell Sorting

Magnetic-activated cell sorting (MACS) is a separation technique where antigen-coated magnetic beads can be incubated with the library to rapidly isolate many binding proteins. A round or two of MACS before FACS can help isolate all potential binders to an antigen while allowing for the removal of non-desirable binders to closely related proteins through the introduction of negative selections. The MACS in this section will assume the following selections and a library diversity of 1  108: Selection 1: Negative selection against non-coated magnetic beads—This selection will remove binders to the beads themselves and any streptavidin binders. Selection 2: Negative selection against closely related proteins—This selection will remove binders to closely related proteins with similar epitopes as the antigen of interest. Selection 3: Positive selection for antigen of interest—This selection will allow for the selection of binders to the antigen of interest.

3.2.1 Preparation of Beads

1. Aliquot 1  107 Dynal Biotin Binder beads per selection (4  108 beads/mL are present in the undiluted stock solution) into labeled microcentrifuge tubes (corresponding to each selection step above) and place them in the magnetic stand (see Note 6). 2. Once the beads localize to the side of the microcentrifuge tube (approximately 1–2 min), remove supernatant and wash the beads two times with 1 mL PBSA 0.1%. For all steps with the magnetic stand, let the tubes contact the magnet for at least 1 min to ensure all beads collect to the side of the tube before removing the supernatant.

Yeast Display Guided Selection of pH-Dependent Binders

299

3. Resuspend beads in 500 μL PBSA 0.1%. 4. Incubate tubes for selections 2 and 3 with appropriate biotinylated antigens (closely related proteins and antigen of interest, respectively) to saturate all streptavidin sites (1–1.5 μg biotinylated antigen/1  107 Dynal Biotin Binder beads) for 30 min to 1 h at 4  C under rotation. 5. Wash antigen-coated beads two times with 1 mL PBSA 0.1% by placing them on the magnetic stand and pipetting off the supernatant. 6. Resuspend all beads from tubes for selections 1, 2, and 3 in 1 mL PBSA 1% and block for 1 h rotating at 4  C. 7. Wash all beads from tubes for selections 1, 2, and 3 two times with 1 mL PBSA 0.1% and resuspend in PBSA 0.1% (100 μL/ selection). 3.2.2 Preparation of Cells

Prior to MACS, confirm library protein expression via detection of the c-Myc tag. Follow preparation of sample 3 in Subheading 3.3. 1. Collect enough cells from step 4 in Subheading 3.1 to ensure 10 diversity in the yeast displayed protein population. Assuming a library diversity of 1  108, 1  109 cells should be collected. 2. Centrifuge cells at 3000  g for 2 min and discard supernatant. 3. Wash cells with PBSA 0.1% and resuspend in appropriate volume of PBSA 0.1%. 1  109 cells should be resuspended in 10 mL or more (see Note 7).

3.2.3 Magnetic Selection

1. For selection 1, add non-coated magnetic beads from tube 1 from step 7 in Subheading 3.2.1 to the resuspended yeast in step 3 of Subheading 3.2.2. 2. Incubate at 4  C for an appropriate length of time to reach equilibrium (see Note 8). 3. Place the tube in the magnetic stand and let sit for 1 min. 4. Transfer the unbound yeast to a new tube and discard the used magnetic beads. 5. Repeat step 4 a total of three times to ensure all beads have been completely removed. 6. For selection 2, add magnetic beads coated with closely related proteins from tube 2 from Subheading 3.2.1 to the resuspended yeast. 7. Repeat steps 2–5 above. 8. For selection 3, add antigen-coated magnetic beads from tube 3 from Subheading 3.2.1 to the resuspend yeast (see Note 9). 9. Rotate at 4  C for an appropriate length of time.

300

Jenna N. Meanor et al.

10. Place the tube in the magnetic stand and let sit for 1 min. Discard the unbound yeast. Caution: do not discard the beads this time. 11. Resuspend antigen-coated magnetic beads in 1 mL PBSA 0.1% and wash five times with 1 mL PBSA 0.1%. 12. Resuspend washed beads in 5 mL SDCAA and place in 30  C shaking at 225 rpm for ~48 h for expansion of collected yeast. 13. Follow Subheading 3.1 for cell growth and expression. 3.3 FACS: Selection of Higher Affinity Binders to the Antigen of Interest

Fluorescent-activated cell sorting (FACS) allows for fine discrimination between high- and low-affinity binders. The samples to be run in this first round of FACS are: Sample 1: Unlabeled control—this sample will help adjust channel voltages to ensure all cells are being properly displayed and captured on the sorter. This sample is only needed if this is the first time this cell type is being run on the sorter. Sample 2: Secondary antibody control—this sample will help determine if cells are falsely displaying a positive binding signal due to nonspecific binding to assay reagents. Sample 3: Expression control—this sample will help determine how many cells in the population are displaying the protein on the yeast cell surface. Sample 4: Fully labeled sample to sort—this sample will be sorted to select for variants that bind to the antigen.

3.3.1 Cell Preparation

1. Aliquot enough cells collected from Subheading 3.2.3, step 13 to ensure 10 diversity into four separate tubes (corresponding to each sample above). The maximum library diversity corresponds to the number of beads used in selection 3 in Subheading 3.2. 2. Wash cells once with PBSA 0.1% and resuspend in an appropriate volume PBSA 0.1%.

3.3.2 Antigen and Primary Antibody Labeling

The samples that need antigen and/or primary antibody labeling are samples 3 and 4. 1. To the resuspended cells in samples 3 and 4, add chicken anti-cMyc antibody (1:100 dilution). 2. To the resuspended cells in sample 4, add biotinylated antigen (see Note 10). 3. Ensure the samples are mixed well, and let the tubes rotate at 4  C for an appropriate length of time (see Note 8). 4. Wash once with PBSA 0.1%, and resuspend in PBSA 0.1%.

Yeast Display Guided Selection of pH-Dependent Binders 3.3.3 Secondary Antibody Labeling (See Note 11)

301

The samples that need secondary labeling are samples 2, 3, and 4. 1. To the resuspended cells in sample 3, add goat-anti chicken Alexa Fluor 488 (1:250 dilution). 2. To the resuspended cells in samples 2 and 4, add goat-anti chicken Alexa Fluor 488 (1:250 dilution) and SA-PE (1:250 dilution). 3. Ensure the samples are well mixed and rotate at 4  C for 20 min shielded from light. 4. Wash samples once with PBSA 0.1%, and resuspend in PBSA 0.1%. 5. Keep samples on ice and shield from light until sorting. 1. Run sample 1 on the flow cytometer using an FSC-H vs. FSC-A plot as reference to ensure all cells are captured by the sorter. Adjust voltages of this laser as needed (Fig. 3a). Draw an oval gate (G1) to capture cells that have a 1:1 ratio of FSC-H: FSC-A. These are single cells, and the gate will exclude cell debris and doublets from the fluorescence analysis.

3.3.4 Sorting

2. Run sample 2 on the sorter and visualize cells from G1 using a two-parameter dot plot using the channels that will represent expression (i.e., Alexa Fluor-488) and antigen binding (i.e., PE). Create a quadrant style gate on this plot to delineate where negatively labeled cells appear as shown in Fig. 3b. 3. Run sample 3 and visualize using the gates created from sample 2. Draw a new gate (Fig. 3c, gate G2) in the upper right quadrant, representing double-positive cells; this new gate will be the sort gate. This gate can be as stringent as needed (see Note 12). a

b

G1

c G2

G1

G2 G3

Antigen binding

FSC-A

G1

FSC-H Yeast cell surface expression

Fig. 3 Representative data showing flow cytometry gating techniques used for isolation of binders to the antigen of interest in an affinity-based manner. (a) Sample 1: Unlabeled control, yeast cells are displayed on FSC-H vs FSC-A plot and gated (G1) to exclude doublets and debris. (b) Sample 3: Expression control, G1 gated yeast cells incubated with expression tag antibodies, gate G2 drawn in double-positive quadrant. (c) Sample 4: Fully labeled sample to sort, G1 gated yeast cells incubated with antigen and expression tag antibodies, cells in G2 are collected. A more stringent gating strategy is shown with gate G3

302

Jenna N. Meanor et al.

4. Run sample 4 and collect cells that fall in gate G2. 1 mL SDCAA media can be added to the collection tube before collecting sorted cells. 5. Transfer sorted yeast cells into SDCAA, and expand culture according to Subheading 3.1. 6. This sorting strategy can be repeated if needed to enrich the population further. 3.4 FACS: Isolation of Mutants That Lose Binding at Low pH

Once all the variants that bind to the antigen have been collected, the second round of FACS sorting will collect variants that lose binding in an acidic environment. The samples to be run in this round of FACS are: Sample 1: Unlabeled control—this sample will help adjust channel voltages to ensure all cells are being properly displayed and captured on the sorter. This sample is only needed if this is the first time this cell type is being run on the sorter. Sample 2: Secondary antibody control—this sample will help determine if cells are falsely displaying a positive binding signal due to nonspecific binding to assay reagents. Sample 3: Expression control—this sample will help determine how many cells in the population are displaying the protein on the yeast cell surface. Sample 4: Positive control for antigen binding at physiological pH— this sample will serve as a control to ensure variants still bind the antigen of interest at physiological pH. Sample 5: Sort sample for antigen binding at acidic pH—this sample will be sorted to select for variants that show little to no binding at an acidic pH.

3.4.1 Cell Preparation

1. Aliquot enough cells collected from Subheading 3.3.4, step 6 to ensure 10 diversity into five separate tubes (corresponding to each sample above). 2. Wash cells once with PBSA 0.1% and resuspend samples 1, 2, 3, and 4 in an appropriate volume PBSA 0.1% and sample 5 in an appropriate volume PBSA 0.1% pH 6.0.

3.4.2 Antigen and Primary Antibody Labeling

The samples that need primary labeling are samples 3, 4, and 5. 1. To the resuspended cells in samples 3, 4, and 5, add chicken anti-c-Myc antibody (1:100 dilution). 2. To the resuspended cells in samples 4 and 5, add biotinylated antigen. 3. Ensure the samples are mixed well and let rotate at 4  C for an appropriate length of time.

Yeast Display Guided Selection of pH-Dependent Binders

303

4. Wash samples 1, 2, 3, and 4 once with PBSA 0.1% and wash sample 5 once with PBSA 0.1% pH 6.0. 5. Resuspend all samples in PBSA 0.1% (see Note 13). 3.4.3 Secondary Antibody Labeling

The samples that need secondary labeling are samples 2, 3, 4, and 5. 1. To the resuspended cells in sample 3, add goat-anti chicken Alexa Fluor 488 (1:250 dilution). 2. To the resuspended cells in samples 2, 4, and 5, add goat-anti chicken Alexa Fluor 488 (1:250 dilution) and SA-PE (1:250 dilution). 3. Ensure the samples are well mixed and rotate at 4  C for 20 min shielded from light. 4. Wash samples once with PBSA 0.1% and resuspend in PBSA 0.1%. 5. Keep samples on ice and shield from light until sorting. 1. Run samples 1 and 2 following Subheading 3.3.4, steps 1 and 2. 2. Run sample 3 and visualize using the gates created from sample 2 to ensure cells are expressing properly. 3. Run sample 4 and draw a gate (Fig. 4a, gate G4) in the lower right quadrant, representing protein expressing cells that do not bind to the antigen at physiological pH; this new gate will be the sort gate. This gate can be as stringent as needed.

a

b

G1

G1

Antigen binding

3.4.4 Sorting

G4

G4

Yeast cell surface expression

Fig. 4 Representative data showing gating techniques used for isolation of binders that do not bind to the antigen of interest at low pH. (a) Sample 4: Positive control for antigen binding at physiological pH, G1 gated yeast cells incubated with antigen at physiological pH, gate G4 drawn below binding population. (b) Sample 5: Sort sample for antigen binding at acidic pH, G1 gated yeast cells incubated with antigen at acidic pH, cells in G4 are collected

304

Jenna N. Meanor et al.

4. Run sample 5 and collect cells that fall in gate G4. 1 mL SDCAA media can be added to the collection tube before collecting sorted cells (Fig. 4b). 5. Transfer sorted yeast cells into SDCAA and expand culture according to Subheading 3.1. 6. This sorting strategy can be repeated if needed to enrich the population. 3.5 FACS: Isolation of pH-Dependent Binders Through Endpoint Sorting

Following the previous two rounds, the variant population should effectively bind the antigen strongly at physiological pH and weakly in acidic conditions. This final round of sorting, employing an end-point sorting strategy, will expose the same variants to both physiological pH and low pH to collect variants that bind at physiological pH and rapidly dissociate upon exposure to low pH. The samples to be run in this round of FACS are: Sample 1: Unlabeled control—this sample will help adjust channel voltages to ensure all cells are being properly displayed and captured on the sorter. This sample is only needed if this is the first time this cell type is being run on the sorter. Sample 2: Secondary antibody control—this sample will help determine if cells are falsely displaying a positive binding signal due to nonspecific binding to assay reagents. Sample 3: Expression control—this sample will help determine how many cells in the population are displaying potential binders. Sample 4: Positive control for antigen binding at physiological pH— this sample will serve as a control to ensure variants still bind the antigen at physiological pH. Sample 5: End-point sort sample—this sample will be sorted to select for variants that bind at physiological pH and rapidly dissociate at an acidic pH.

3.5.1 Cell Preparation

1. Aliquot enough cells collected from Subheading 3.4.4, step 6 to ensure 10 diversity into five separate tubes (corresponding to each sample above). 2. Wash cells once with PBSA 0.1% and resuspend in an appropriate volume PBSA 0.1%.

3.5.2 Antigen and Primary Antibody Labeling

The samples that need primary labeling are samples 3, 4, and 5. 1. To the resuspended cells in samples 3, 4, and 5, add chicken anti-c-Myc antibody (1:100 dilution). 2. To the resuspended cells in samples 4 and 5, add biotinylated antigen.

Yeast Display Guided Selection of pH-Dependent Binders

305

3. Ensure all samples are well mixed and let rotate at 4  C for an appropriate length of time. 4. Wash sample 5 once in PBSA 0.1% and resuspend in PBSA 0.1% pH 6.0. 5. Ensure the sample is well mixed and rotate all samples at RT for an additional 5–10 min (see Note 14). 6. Wash sample 5 twice with PBSA 0.1% pH 6.0. 7. Wash all samples once with PBSA 0.1% and resuspend in PBSA 0.1%. 3.5.3 Secondary Antibody Labeling

The samples that need secondary labeling are samples 2, 3, 4, and 5. 1. To the resuspended cells in sample 3, add goat-anti chicken Alexa Fluor 488 (1:250 dilution). 2. To the resuspended cells in samples 2, 4, and 5, add goat-anti chicken Alexa Fluor 488 (1:250 dilution) and SA-PE (1:250 dilution). 3. Ensure the samples are well mixed and rotate at 4  C for 20 min shielded from light. 4. Wash samples once with PBSA 0.1% and resuspend in PBSA 0.1%. 5. Keep samples on ice and shield from light until sorting.

3.5.4 Sorting

1. Run samples 1 and 2 following Subheading 3.3.4, steps 1 and 2. 2. Run sample 3 and visualize using the gates created from sample 2 to ensure cells are expressing properly. 3. Run sample 4 and draw a gate (Fig. 5a, gate G5) in the lower right quadrant, representing expressing cells that do not bind to the antigen at physiological pH; this new gate will be the sort gate. This gate can be as stringent as needed. 4. Run sample 5 and collect cells that fall in gate G5. 1 mL SDCAA media can be added to the collection tube before collecting sorted cells (Fig. 5b). 5. Transfer sorted yeast cells into SDCAA and expand culture according to Subheading 3.1. 6. This sorting strategy can be repeated if needed to enrich the population. 7. After expanding the culture, plate serial dilutions onto SDCAA agar plates. Incubate at 30  C until colonies are visible and pick individual colonies into SDCAA media for further single clone analysis.

Jenna N. Meanor et al.

a

b

G1

G1

Antigen binding

306

G5

G5

Yeast cell surface expression

Fig. 5 Representative data showing gating techniques used for isolation of binders that rapidly dissociate from the antigen of interest at low pH. (a) Sample 4: Positive control for target binding at physiological pH, G1 gated yeast cells incubated with antigen at physiological pH, gate G5 drawn below binding population. (b) Sample 5: End-point sort sample, G1 gated yeast cells incubated initially with antigen at physiological pH and further incubated in acidic pH, cells in G5 are collected 3.6

Characterization

3.6.1 Sequencing

Resulting yeast clones should be sequenced from the final sorted population to ensure only unique clones are characterized further. Sequences may also provide insight into structural modifications required to bind to the target antigen. 1. Individually expand 25 single colonies from plates (see Subheading 3.5.4, step 8) in 5 mL SDCAA media for 24–36 h at 30  C shaking at 225 rpm. 2. Plasmids can be isolated from 2 mL of the above yeast culture using Zymoprep Yeast plasmid Miniprep II kit and following the manufacturer’s protocol. The rest of the culture can be stored at 4  C for single clone analysis via flow cytometry described in Subheading 3.6.2. 3. Isolated plasmids can be further transformed into E. coli DH5α cells using standard molecular biology methods, and colonies can be sent for sequencing.

3.6.2 Single Clone Analysis

The selected unique variants should be tested to ensure they still bind at physiological pH and do not undergo irreversible structure changes in an acidic environment. Single clone analysis will allow for each variant to be checked individually using flow cytometry (see Note 15). The samples to be run for each clone are: Sample 1: Secondary antibody control—this sample will help determine which cells are falsely displaying a positive binding signal due to nonspecific binding to assay reagents. Sample 2: Expression control—this sample will help determine how many cells in the population are displaying the selected variant.

Yeast Display Guided Selection of pH-Dependent Binders

307

Sample 3: Antigen binding at physiological pH—this sample will serve as a control to ensure variants still bind the antigen at physiological pH. Sample 4: Antigen binding at acidic pH—this sample will serve as a control to ensure variants have lower binding affinity to the antigen at acidic pH. Sample 5: Reversibility of structural change—this sample will ensure the variant can be exposed to an acidic environment and retain structural integrity upon returning to physiological pH. Run samples on a flow cytometer following cell preparation, antigen and primary antibody labeling, and secondary antibody labeling protocols described above in Subheading 3.5. Sample 3 will undergo antigen labeling in PBSA 0.1%, sample 4 will undergo antigen labeling in PBSA 0.1% pH 6.0, and sample 5 will be incubated in PBSA 0.1% pH 6.0 for 10 min prior to antigen labeling in PBSA 0.1%. All secondary antibody labeling will be done in PBSA 0.1%. 3.6.3 Biophysical Characterization of Selected Clones

4

Once promising clones and their sequences have been identified and confirmed for pH-dependent binding on the yeast cell surface, biophysical characterization of soluble proteins can be performed to confirm the binding and pH-dependent properties. The various characterization methods are beyond the scope of this chapter. Briefly, antibody/protein genes can be synthesized commercially and subcloned into an expression vector with purification tags (see Note 16). Proteins can then be expressed via transfection of plasmids into an expression host of choice (E. coli, yeast, or mammalian cells) and purified via standard purification methods. Purified antibodies/proteins can be tested for binding affinity, thermal stability, and monomeric content via size exclusion—HPLC. A more extensive developability assessment would be required if the molecule is being used as a therapeutic.

Notes 1. Yeast libraries can be stored at 80  C indefinitely in a freezing solution of SDCAA with 10–20% glycerol. Upon thawing, yeast should be passaged at least twice into fresh SDCAA to ensure a healthy population of cells. 2. Antigens and closely related proteins used in this protocol were purchased pre-biotinylated. However, they can be biotinylated with a variety of kits available commercially such as the EZ-Link Sulfo-NHS-Biotin from Thermo Fisher Scientific. 3. The library referred to in this protocol was constructed in the vector pCTcon2 [31] which includes a HA-tag at the N

308

Jenna N. Meanor et al.

terminus and a c-Myc tag on the C-terminus of the displayed protein. Specific labeling of the c-Myc tag can be used to check for full-length protein expression. If your plasmid does not contain a c-Myc tag, appropriate antibodies should be purchased for your specific expression tags. 4. When growing and inducing yeast cultures, ensure the baffled flask used can hold at least 5 the volume of the culture. This is to ensure proper aeration of the culture. 5. The initial library diversity can be calculated by plating serial dilutions on SDCAA plates. Library diversity will also change through different selection rounds. It can be assumed that the maximum library diversity after each MACS round will be the number of beads used, and the maximum library diversity after each FACS round will be the number of cells collected. 6. The number of magnetic beads used in the MACS step in this protocol maintains a minimum bead: yeast ratio of 1:20 and should be adjusted for the library diversity and desired stringency. A more stringent MACS would be performed on a library generated from an existing antigen binder while a lessstringent MACS would be desirable for a novel library. The text maintains an example of a library diversity of 1  108, volumes used should be scaled appropriately for differing library diversities. 7. The incubation steps described in this protocol are performed at a maximum OD of 10.0 to ensure for the constant suspension of yeast cells. In general, the volumes used for this protocol should be adjusted proportionally to how many cells are being used. As library diversity decreases, incubation OD should decrease as well. 8. All incubation steps in this protocol with target coated beads or biotinylated soluble antigen should be long enough to allow for each protein–protein interaction to reach equilibrium. One hour should be sufficient for a majority of interactions. 9. Along with varying the bead: yeast ratio as mentioned in Note 6, two other methods by which to increase the stringency of an MACS include incubating displaying yeast with DTT prior to a magnetic separation to decrease avidity effects (since a fraction of the disulfide-anchored surface constructs will dissociate) [32] as well as the addition of non-transformed EBY100 cells to decrease nonspecific selection. 10. The concentration of antigen added during FACS is dependent on the affinity of binders you wish to capture. This concentration should be no greater than 1 μM to minimize the selection of nonspecific binders. Throughout subsequent sorting rounds, this concentration can be decreased to increase the stringency of the sort.

Yeast Display Guided Selection of pH-Dependent Binders

309

11. Between successive screening rounds, labeling reagents should be alternated to avoid enrichment of binders to the labeling reagents, i.e., alternate between Streptavidin-PE and an antibiotin or a neutravidin fluorophore conjugate ensuring that fluorophores used for expression and binding emit in different channels based on the flow cytometer being used. 12. To create a more stringent gate, remove cells from the sort gate that have a higher surface expression to antigen binding by collecting cells above the diagonal 1:1 ratio of these fluorophores (Fig. 3c, gate G3). This gate can then be shifted to increase the ratio of binding to expression to further increase stringency. Ensure that this gating strategy is not too stringent in early sorting rounds; it is likely that many pH-dependent variants will not be among those with the highest affinity to the antigen. Too stringent of a screening strategy early in the sorting process might eliminate better pH-dependent variants. 13. Resuspension in the corresponding low pH buffer is unnecessary since excess antigen has been washed away and binders that did not bind at pH 6.0 will not bind the antigen at pH 7.4 now. All secondary antibody labeling should be performed at pH 7.4. 14. The length of time allowed for dissociation in this step can be optimized depending on desired binding traits. A shorter incubation time will select for pH-dependent binders while a longer incubation time might allow for selection of low affinity binders. 15. Following this same protocol (Subheading 3.6.2) but varying the antigen concentration allows for affinity analysis on the surface of yeast [15]. 16. Several commercial vectors are available for subcloning of proteins and antibodies with his-tag or human Fc. Once binders are expressed with these tags, affinity chromatography methods can be applied for protein purification.

Acknowledgments The work described here was supported by grants from the National Institute of Health (R21EB023377) and the National Science Foundation (EFMA 1830910). References 1. French AR, Tadaki DK, Niyogi SK, Lauffenburger DA (1995) Intracellular trafficking of epidermal growth factor family ligands is directly influenced by the pH sensitivity of the

receptor/ligand interaction. J Biol Chem 270: 4334–4340 2. Maeda K, Kato Y, Sugiyama Y (2002) pH-dependent receptor/ligand dissociation as

310

Jenna N. Meanor et al.

a determining factor for intracellular sorting of ligands for epidermal growth factor receptors in rat hepatocytes. J Control Release 82:71–82 3. Vaughn DE, Bjorkman PJ (1998) Structural basis of pH-dependent antibody binding by the neonatal fc receptor. Structure 6:63–73 4. Ko¨nning D, Zielonka S, Sellmann C et al (2016) Isolation of a pH-sensitive IgNAR variable domain from a yeast-displayed, histidinedoped master library. Marine Biotechnol 18: 161–167. https://doi.org/10.1007/s10126016-9690-z 5. Schro¨ter C, Gu¨nther R, Rhiel L et al (2015) A generic approach to engineer antibody pH-switches using combinatorial histidine scanning libraries and yeast display. MAbs 7: 1 3 8 – 1 5 1 . h t t p s : // d o i . o r g / 1 0 . 4 1 6 1 / 19420862.2014.985993 6. Chaparro-Riggers J, Liang H, DeVay RM et al (2012) Increasing serum half-life and extending cholesterol lowering in vivo by engineering antibody with pH-sensitive binding to PCSK9. J Biol Chem 287:11090–11097. https://doi. org/10.1074/jbc.M111.319764 7. Gera N, Hill AB, White DP et al (2012) Design of pH sensitive binding proteins from the Hyperthermophilic Sso7d scaffold. PLoS One 7:e48928. https://doi.org/10.1371/journal. pone.0048928 8. Traxlmayr MW, Lobner E, Hasenhindl C et al (2014) Construction of pH-sensitive Her2binding IgG1-fc by directed evolution. Biotechnol J 9:1013–1022. https://doi.org/10. 1002/biot.201300483 9. Igawa T, Mimoto F, Hattori K (2014) PH-dependent antigen-binding antibodies as a novel therapeutic modality. Biochim Biophys Acta Proteins Proteomics 1844:1943–1950 10. Schro¨ter C, Krah S, Beck J et al (2018) Isolation of pH-sensitive antibody fragments by fluorescence-activated cell sorting and yeast surface display. In: Methods in molecular biology. Humana Press, Totowa, NJ, pp 311–331 11. Fukuzawa T, Sampei Z, Haraya K et al (2017) Long lasting neutralization of C5 by SKY59, a novel recycling antibody, is a potential therapy for complement-mediated diseases. Sci Rep 7: 1080. https://doi.org/10.1038/s41598017-01087-7 12. Yang D, Giragossian C, Castellano S et al (2017) Maximizing in vivo target clearance by design of pH-dependent target binding antibodies with altered affinity to FcRn. MAbs 9: 1105–1117. https://doi.org/10.1080/ 19420862.2017.1359455 13. Sheridan D, Yu ZX, Zhang Y et al (2018) Design and preclinical characterization of

ALXN1210: a novel anti-C5 antibody with extended duration of action. PLoS One 13: e0195909. https://doi.org/10.1371/journal. pone.0195909 14. Boder ET, Wittrup KD (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol 15:553–557 15. Gera N, Hussain M, Rao BM (2013) Protein selection using yeast surface display. Methods 60:15–26. https://doi.org/10.1016/j.ymeth. 2012.03.014 16. Cherf GM, Cochran JR (2015) Applications of yeast surface display for protein engineering. Methods Mol Biol 1319:155–175. https:// doi.org/10.1007/978-1-4939-2748-7_8 17. Benatuil L, Perez JM, Belk J, Hsieh C-M (2010) An improved yeast transformation method for the generation of very large human antibody libraries. Protein Eng Des Sel 23:155–159. https://doi.org/10.1093/pro tein/gzq002 18. Ackerman M, Levary D, Tobon G et al (2009) Highly avid magnetic bead capture: an efficient selection method for de novo protein engineering utilizing yeast surface display. Biotechnol Prog 25:774–783. https://doi.org/10.1002/ btpr.174 19. Angelini A, Chen TF, De Picciotto S et al (2015) Protein engineering and selection using yeast surface display. Methods Mol Biol 1319:3–36. https://doi.org/10.1007/978-14939-2748-7_1 20. Ko¨nning D, Kolmar H (2018) Beyond antibody engineering: directed evolution of alternative binding scaffolds and enzymes using yeast surface display. Microb Cell Fact 17:32 21. Boder ET, Midelfort KS, Dane Wittrup K (2000) Directed evolution of antibody fragments with monovalent femtomolar antigenbinding affinity. PNAS 97:10701–10705 22. Gera N, Hussain M, Wright RC, Rao BM (2011) Highly stable binding proteins derived from the hyperthermophilic Sso7d scaffold. J Mol Biol 409:610–616. https://doi.org/10. 1016/j.jmb.2011.04.020 23. Bogen JP, Hinz SC, Grzeschik J et al (2019) Dual function pH responsive bispecific antibodies for tumor targeting and antigen depletion in plasma. Front Immunol 10:1892. https:// doi.org/10.3389/fimmu.2019.01892 24. Pepper LR, Cho YK, Boder ET, Shusta EV (2008) A decade of yeast surface display technology: where are we now? Comb Chem High Throughput Screen 11:127–134 25. Scholler N (2012) Selection of antibody fragments by yeast display. Humana Press, Totowa, NJ, pp 259–280

Yeast Display Guided Selection of pH-Dependent Binders 26. Rosowski S, Becker S, Toleikis L et al (2018) A novel one-step approach for the construction of yeast surface display fab antibody libraries. Microb Cell Fact 17:3. https://doi.org/10. 1186/s12934-017-0853-z 27. McMahon C, Baier AS, Pascolutti R et al (2018) Yeast surface display platform for rapid discovery of conformationally selective nanobodies. Nat Struct Mol Biol 25:289–296. https://doi.org/10.1038/s41594-0180028-6 28. Swers JS, Kellogg BA, Wittrup KD (2004) Shuffled antibody libraries created by in vivo homologous recombination and yeast surface display. Nucleic Acids Res 32:e36–e36. https://doi.org/10.1093/nar/gnh030 29. Bonvin P, Venet S, Fontaine G et al (2015) De novo isolation of antibodies with

311

pH-dependent binding properties. MAbs 7: 2 9 4 – 3 0 2 . h t t p s : // d o i . o r g / 1 0 . 1 0 8 0 / 19420862.2015.1006993 30. Zaccolo M, Williams DM, Brown DM, Gherardi E (1996) An approach to random mutagenesis of DNA using mixtures of triphosphate derivatives of nucleoside analogues. J Mol Biol 255:589–603 31. Chao G, Lau WL, Hackel BJ et al (2006) Isolating and engineering human antibodies using yeast surface display. Nat Protoc 1:755–768. https://doi.org/10.1038/nprot.2006.94 32. Stern LA, Csizmar CM, Woldring DR et al (2017) Titratable avidity reduction enhances affinity discrimination in mammalian cellular selections of yeast-displayed ligands. ACS Comb Sci 19:315–323. https://doi.org/10. 1021/acscombsci.6b00191

Chapter 17 Yeast Mating as a Tool for Highly Effective Discovery and Engineering of Antibodies via Display Methodologies Du-San Baek, Seong-Wook Park, Cynthia Adams, Dimiter S. Dimitrov, and Yong-Sung Kim Abstract Yeast surface display (YSD) is a powerful methodology for discovery and engineering of antibodies, and the yeast mating has been used to overcome low transformation efficiency of yeast in antibody library generation. We developed an optimized method of yeast mating for generating a large, combinatorial antibody fragment library and heterodimeric protein library by cellular fusion between two haploid cells carrying different library each other. This method allows for increased diversity in screening of target-specific fragment antigen-binding (Fab) antibodies as well as in the development of heterodimeric Fc variants for bi-specific antibody generation and T-cell receptor (TCR). Here we describe the efficient isolation of human antibodies against the activated GTP-bound form of the oncogenic Ras mutant (KRasG12D-GTP) by sequential isolation of their heavy chains (HCs) followed by combination with light chains (LCs) via the yeast mating process. This strategy facilitates guided selection of the antigen-specific HC with either a fixed functional LC, which has cytosol penetrating ability, or an LC library to generate the Fab. It also allows for deeper exploration of a sequence space with fixed diversity, leading to a higher probability of successful isolation of human antibodies with high specificity and affinity. Key words Yeast surface display, Yeast mating, Antibody screening, Antibody library, Guided selection, KRas

1

Introduction Great strides have been made in engineering of antibodies and other proteins with regard to yeast surface display technology (YSD) [1–7]. An increasing list of applications continues to be reported for enzymes, receptors, and alternative scaffolds [8– 11]. The majority of the research applying YSD to antibody engineering has been conducted with mating type a (MATa) haploid yeast strains. MATa yeast expresses a-agglutinin, composed of an anchorage subunit (Aga1p) that acts as a GPI attachment signal

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_17, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

313

314

Du-San Baek et al.

enabling anchoring to the cell wall of yeast, and an adhesion subunit (Aga2p) that is linked to the Aga1p through two disulfide bridges [12]. The subunit Aga2p is preferred for conjugation with the protein of interest (POI) to display POI on the yeast surface due to its smaller size (68 residues) compared to Aga1p (674 residues). It is also anchored away from the cell surface, minimizing potential interactions with other proteins on the yeast cell surface and reducing the effects of a fluorescent protein on structure and folding during labeling steps [13]. The other yeast used in antibody engineering is mating type alpha (MATα), which produces POI through its secretory pathway because MATα haploid yeast has α-agglutinin (607 residues), lacking the Aga2p for anchoring of POI [2, 14]. The mating of yeast only occurs with those two haploids, capable of mating with other haploid cells of the opposite mating type. The yeast mating is initiated by multiple interactions between sexual adhesion molecules of both haploid cells and eventually two haploid cells undergo cell fusion, requiring extensive remodeling of the cell wall to produce a continuous cytoplasm [15]. After mating, generated diploid yeast cells can also obtain the capability to display POI on its surface via a-agglutinin like MATa haploid. Initially, the yeast mating was exploited to overcome limited transformation efficiency of yeast leading to modest library size [1, 16], but it is now more actively being used as a versatile tool for generation and optimization of heterodimeric proteins including Fab, TCR, and heterodimeric Fc [1, 4, 10, 14]. Previously, we investigated optimal yeast mating conditions with regard to cell density, media pH, and the cell growth phase to generate a large synthetic human Fab antibody library having diversity larger than 109. We found that yeast mating with mid-exponential growth-phased cells at 3
107 cells/cm2 and use of pH 4.5 YPD plates kept at 30  C for 6 h achieved a mating efficiency rate of over 50% compared to the reported 29% of conventional approaches [1]. Here, we describe detailed procedures related to the construction of haploid libraries and the generation of a diploid library through yeast mating, which achieved greater than 50% mating efficiency in KRasG12D-GTP-specific antibody screening as an example. With optimized protocol for the yeast mating, we successfully isolated a potent bifunctional antibody, comprised of a heavy chain (HC) that selectively binds to the activated GTP-bound form of the oncogenic Ras mutant (KRasG12D-GTP) with a fixed light chain (LC) that has cytosol penetrating ability to deliver intact IgG to cytosolic space of living cells [5, 17]. The strategy described here can be applied for the isolation of monofunctional and conventional antibodies as well.

Yeast Mating for Antibody Discovery and Engineering

2 2.1

315

Materials Yeast Strains

1. JAR200 (MATa, GAL1-AGA1::KanM
4ura3D45, ura3–52 trp1 leu2D1 his3D200 pep4::HIS3 prb1D1.6R can1), kindly provided by Prof. Dane Wittrup (Massachusetts Institute of Technology, USA) [1]. 2. YVH10 (MATα, PDI::ADHII-PDI-Leu2 ura3-52 trp1 leu2D1 his3D200 pep4::HIS3 prb1D1.6p can1 GAL), kindly provided by Prof. Dane Wittrup (Massachusetts Institute of Technology, USA) [1].

2.2

Media

1. YPD media: 10 g/L yeast extract, 20 g/L Peptone, and 20 g/ L D-(+)-Glucose. 2. YPD media pH 4.5: 10 g/L yeast extract, 20 g/L Peptone, 20 g/L D-(+)-Glucose, 14.7 g/L sodium citrate (Na3C6H5O7) and 4.29 g/L citric acid monohydrate (C6H8O7  H2O), adjust the pH to 4.5. 3. SD-CAA (SD) media pH 6.4: 20 g/L glucose, 6.7 g/L yeast nitrogen base without amino acids, 5.4 g/L sodium phosphate dibasic (Na2HPO4), 8.6 g/L sodium phosphate monobasic monohydrate (NaH2PO4  H2O), and 5 g/L casamino acids, adjust the pH to 6.4. 4. SD-CAA (SD) media pH 4.5: 20 g/L glucose, 6.7 g/L yeast nitrogen base without amino acids, 14.7 g/L sodium citrate, 4.29 g/L citric acid monohydrate, and 5 g/L casamino acids, adjust the pH to 4.5. 5. SD-CAA + uracil (SDU) media: Add 10 mL 100
uracil stock (0.02 g/L) to 990 mL SD-CAA media. 6. SD-CAA + tryptophan (SDT) media: Add 10 mL 100
tryptophan stock (0.04 g/L) to 990 mL SD-CAA media. 7. SG-CAA (SG) media: Same as for SD-CAA pH 6.4, except for using 20 g D-(+)-Galactose instead of D-(+)-Glucose. 8. LB media: 10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl, adjust pH to 7.0 with NaOH. 9. LBA media: Add 1 mL 1000
ampicillin stock (1000 mg/mL) to 1L LB media. 10. LBCA media: Add 1 mL 1000
chloramphenicol stock (34 mg/mL) to 1L LBA media. 11. YPD plate pH 4.5: Add 15 g agar and magnetic stir bar to 1 L of YPD media pH 4.5 and then autoclave. 12. SD, SDT, or SDU plate pH 6.4: Add 15 g agar and magnetic stir bar to 1 L of each liquid media pH 6.4 and then autoclave.

316

Du-San Baek et al.

2.3 Reagents and Equipment

1. D-Sorbitol. 2. Calcium chloride dihydrate. 3. Lithium acetate dihydrate. 4. DL-Dithiothreitol (DTT). 5. Isopropyl β-D-1-thiogalactopyranoside (IPTG). 6. β-Mercaptoethanol. 7. Ammonium sulfate. 8. Zinc chloride. 9. Magnesium chloride. 10. Sodium chloride. 11. D-biotin. 12. Bicine. 13. EDTA. 14. Imidazole. 15. Bovine serum albumin powder. 16. Sephadex G-25 in PD-10 Desalting Columns. 17. Frozen-EZ Yeast Transformation II Kit. 18. Plasmid DNA mini-prep kit for S. cerevisiae (for yeast). 19. Plasmid DNA mini-prep kit for E. coli. 20. 0.2-cm electroporation cuvettes. 21. Ni-NTA resin. 22. Agar. 23. UltraPure™ Agarose. 24. Phosphatase alkaline, immobilized on Agarose. 25. 0.45-μm PES syringe filter. 26. 10,000 MWCO Amicon® centrifugal filter units. 27. Disposable 10 mL column. 28. Penicillin-streptomycin. 29. Ampicillin. 30. Chloramphenicol. 31. Q5® high-fidelity DNA polymerase. 32. Plasmid DNA mini-prep kit for E. coli. 33. Agarose gel-extraction kit. 34. pYDS-H VH3-23 cloned plasmid DNA for HC haploid yeast library. 35. pYDS-K Vκ1-16 cloned plasmid DNA for LC haploid yeast library. 36. pET23m His-Avi-KRasG12D plasmid DNA.

Yeast Mating for Antibody Discovery and Engineering

317

37. pBirAcm plasmid DNA. 38. Guanosine 50 -[β,γ-imido]triphosphate trisodium salt hydrate (GppNHp). 39. Guanosine 50 -diphosphate (GDP). 40. Alkaline phosphatase, immobilized on agarose beads. 41. MidiMACS Starting Kit with LS column. 42. Streptavidin microbeads. 43. Sonicator. 44. Flow cytometry apparatus equipped cell sorter unit. 45. Anti c-Myc antibody FITC conjugated (9E10 clone). 46. Streptavidin R-phycoerythrin conjugated (SA-PE). 47. Gene Pulser II Electroporator, Pulse Controller Plus, Capacitance Extender Plus, Gene Pulser II Shocking Chamber, 25-pin and 9-pin Interconnect Cables. 48. Yeast electroporation buffer: 1 M sorbitol and 1 mM CaCl2. 49. Yeast competent buffer: 100 mM lithium acetate (LiAc) and 10 mM DTT. 50. Yeast resuspension buffer: 1:1 mix of 1 M sorbitol in ddH2O: YPD media. 51. E. coli resuspension buffer: 50 mM Tris-HCl, 300 mM NaCl, and 2 mM β-mercaptoethanol, adjust the pH to 8.0. 52. KRas protein storage buffer: 50 mM Tris-HCl, 2 mM magnesium chloride (MgCl2), and 1 mM DTT, adjust the pH to 8.0. 53. GppNHp reaction buffer: 50 mM Tris-HCl, 200 mM ammonium sulfate ((NH4)2SO4), 10 μM zinc chloride (ZnCl2), and 5 mM DTT, adjust the pH to 7.5. 54. GDP reaction buffer: 50 mM Tris-HCl, 2 mM MgCl2, 1 mM DTT, and 20 mM EDTA, adjust the pH 8.0. 55. PBSM buffer: Add 5 g/L bovine serum albumin, and 744 mg/ L EDTA to PBS pH 7.4. 56. PBSF buffer: Add 1 g/L bovine serum albumin to PBS pH 7.4. 57. Yeast storage buffer: 25% glycerol containing YPD, SD, SDU, or SDT media (w/v); YPD for yeast strains, SD for diploid yeast after mating, SDU for HC haploid yeast library, and SDT for LC haploid yeast library. 2.4

Primers

See Table 1 for a list of primers and sequences used for library construction and sequencing.

318

Du-San Baek et al.

Table 1 List of primers and sequences used for library construction and sequencing Name

Length Nucleotide sequences (50 –30 )

H1-For

32

TAA AGA AGA AGG GGT ACA ATT GGA TAA AAG AG

H1-Rev

54

AGC CTG CCT GAC CCA GCT CAT CSM GTA AKH AGA GAA TGT GAA TCC AGA GGC TGC

H2-For

21

ATG AGC TGG GTC AGG CAG GCT

H2-Rev

75

GCC CTT GAC AGA ATC TGC GTA GTA TDT ANY ACY TCC AYB AKA CSH AAT KVY TGA CAC CCA TTC CAA GCC TTT TCC

H3-For

24

TAC TAC GCA GAT TCT GTC AAG GGC

H3-Rev

24

AGC ACA GTA ATA CAC AGC CGT GTC

H4-Rev (CDR-H3 length 11)

147

GGT GCT CTT GGA GGA GGG TGC CAG GGG GAA GAC CGA TGG GCC CTT GGT GGA GGC GGA AGA GAC GGT AAC CAA TGT TCC CTG TCC CCA GTA GTC MAW MNN MNN MNN MNN MNN MNN MNN MNN CYT AGC ACA GTA ATA CAC AGC CGT GTC

K1-For

22

TAA AGA AGA AGG GGT ACA ATT G

K1-Rev

47

AGG CTT CTG CTG ATA CCA TSC CAA ATM AKW GGA GAT GCC TTG AGA GG

K2-For

19

ATG GTA TCA GCA GAA GCC T

K2-Rev

62

AAA ATC TAG AAG GAA CTC CAG ATT SCA AAG WAG AAG CAK CAT AGA TCA ACA ATT TAG GAG CC

K3-For

22

TCT GGA GTT CCT TCT AGA TTT T

K3-Rev

22

TTG TTG GCA GTA ATA TGT AGC G

K4-Rev (CDR-L3 length 9)

116

CAG ATG GCG GGA AGA TGA AGA CAG ATG GTG CAG CCA CCG TAC GCT TAA TCT CCA CCT TCG TTC CCT GGC CGA ATG TAN NAG GAD WAG AAT HAD RTT GTT GGC AGT AAT ATG TAG CG

Sequencing primer (SP)_For

21

GAT TTC GAT GCT GCT GCT TTG

SP_Rev (VH)

18

CAC CGG TTC GGG GAA GTA

SP_Rev (VL)

22

CTT CCA CTG TAC TTT GGC CTC T

Lined sequences: Nucleotides for homologous recombination sites in pYDS-H and pYDS-K plasmid

3

Methods

3.1 Amplification of Insert DNA for HC and LC Libraries

Insert DNA preparation for the fully synthetic HC and LC libraries was performed by serial overlapping polymerase chain reaction (PCR) with degenerative primers (Fig. 1) [1]. The constructed HC and LC libraries are designed to have diversified all six CDRs (complementarity-determining regions) of the variable domains

Yeast Mating for Antibody Discovery and Engineering

319

Materials

Step 1: PCR reaction for fragment DNA preparation

HMS

CDR-H1

H1-For

CDR-H2 H2-For

Products

H1-Rev

CDR-H3 H3-For

HMS

H3-Rev

H2-Rev

HMS

CDR-H2

CDR-H1

(F1)

(F2)

(F3)

Product

Materials

Step 2: Overlap extension PCR reaction for insert DNA preparation HMS

CDR-H1 CDR-H2

H1-For

HMS

FR1F

CDR-H1

R2

CDR-H2

H4-Rev

FR3F

(Insert DNA for HC library)

CDR-H3

R4

HMS Homologous Recombination Site (HMS)

Fig. 1 Gene library preparation. Overall scheme of insert DNA preparation for haploid HC library construction. Initial step is amplification of each fragment DNAs with designed primers mimicking the human germline antibody repertoires by degenerate codons. Second step is assembling amplified fragments to generate full length of insert DNA with two extra sequences at both 50 and 30 termini for the homologous recombination. For haploid LC library construction, the overall process is identical except for use of different primer sets

with a specific aim of mimicking the human germline antibody repertoires to selected human antibody frameworks of VH3-23 and Vκ1-16 germline sequences (Table 2). 1. Prepare Q5® High-Fidelity DNA polymerase reaction mix for a typical 50–100 μL reaction per tube as following: (a) 25 ng template DNA (VH3-23 cloned pYDS-H or Vκ116 cloned pYDS-K plasmid). (b) 200 μM each deoxyribonucleotide triphosphate (dNTP). (c) 0.5 μM forward primer (Tm: ~58  C for HC library, Tm: ~50  C for LC library) (see Note 1). (d) 0.5 μM reverse primer (Tm: ~58  C for HC library, Tm: ~50  C for LC library). (e) 1
Q5® reaction buffer.

320

Du-San Baek et al.

Table 2 DNA sequences for VH3-23 and Vκ1-16 frameworks used for library construction DNA name

Nucleotide sequences (50 –30 )

VH3-23

GAA GTG CAA TTG GTG GAG TCT GGC GGC GGA TTG GTG CAA CCA GGA GGA TCT TTG AGA TTG TCT TGC GCA GCC TCT GGA TTC ACA TTC TCT TCT TAC GCT ATG AGC TGG GTC AGA CAG GCT CCA GGA AAA GGC TTG GAA TGG GTG TCA GCA ATT TCT GGA TCT GGA GGA TCT ACA TAC TAC GCA GAT TCT GTC AAG GGC AGG TTC ACC ATC TCC AGG GAC AAC TCC AAG AAC ACA TTG TAC TTG CAG ATG AAT TCC TTG AGA GCT GAA GAC ACA GCT GTG TAT TAC TGC GCT AAA TGG GGA GGG GAC GGC TTC TAT GCT ATG GAC TAC TGG GGA CAG GGA ACA TTG GTT ACC GTC TCT TCC

Vκ1-16

GAT ATT CAG ATG ACA CAG TCT CCT TCC TCT TTG TCT GCT TCT GTG GGC GAT AGG GTT ACA ATA ACA TGC AGG GCC TCT CAA GGC ATC TCC TCC TAC TTG GCT TGG TAT CAG CAG AAG CCT GGA AAG GCT CCT AAA TTG TTG ATC TAT GCT GCT TCT TCT TTG CAG TCT GGA GTT CCT TCT AGA TTT TCT GGA TCT GGA TCT GGA ACA GAC TTC ACA TTG ACC ATA TCC TCC TTG CAA CCT GAA GAT TTC GCT ACA TAT TAC TGC CAA CAA TAC AAC TCT TAC CCT GGC ACA TTC GGC CAG GGA ACG AAG GTG GAG ATT AAG CGT

(f) 0.02 U/μL of Q5® polymerase. (g) Bring volume to 50–100 μL with ddH2O. 2. Run PCR reaction on following program (see Note 2), (a) Initial denaturation: 98  C (1 min). (b) 30 cycles: 98  C (10 s), 60  C for HC library or 52  C for LC library (30 s), and 72  C (10 s). (c) Final Extension: 72  C (2 min). (d) Cooling: 16  C (5 min). 3. Run DNA fragments using gel electrophoresis on a 2.0% agarose gel. 4. Purify fragment DNA using gel-extraction kit. 5. Prepare overlapping PCR reaction mix with purified DNA for each fragment as following: (a) 10 nM fragment #1 (F1). (b) 10 nM fragment #2 (F2). (c) 10 nM fragment #3 (F3). (d) 200 μM each deoxyribonucleotide triphosphate (dNTP). (e) 0.5 μM forward primer (Tm: ~58  C). (f) 0.5 μM reverse primer (Tm: ~58  C). (g) 1
Q5® reaction buffer. (h) 0.02 U/μL of Q5® polymerase. (i) Bring volume to 50–100 μL with ddH2O.

Yeast Mating for Antibody Discovery and Engineering

321

6. Run PCR reaction on following program: (a) Initial denaturation: 98  C (1 min). (b) 20 cycles: 98  C (10 s), 62  C (30 s), and 72  C (20 s). (c) Final Extension: 72  C (2 min). (d) Cooling: 16  C (5 min). 7. Purify insert DNA after electrophoresis on 1.2% agarose gel. 8. Measure concentration and purity (A260/280) after finishing gel-extraction (see Note 3). 3.2 Enzymatic Digestion for Vector Preparation

1. Prepare double enzymatic digestion for pYDS-H as following: (a) 25 μg of pYDS-H plasmid. (b) 50 U of NheI-HF endonuclease. (c) 100 U of ApaI endonuclease. (d) 10 μL of 10
Cut-Smart buffer (NEB). (e) Bring volume to 50 μL with ddH2O. 2. Prepare double enzymatic digestion for pYDS-K as following: (a) 25 μg of pYDS-K plasmid. (b) 50 U of NheI-HF endonuclease. (c) 50 U of BsiWI-HF endonuclease. (d) 10 μL of 10
Cut-Smart buffer (NEB). (e) Bring volume to 50 μL with ddH2O. 3. Incubate the enzymes and DNA mixture at 37  C for 6 h. 4. Purify digested plasmid DNA by using gel-extraction kit after electrophoresis on 1.0% agarose gel. 5. Measure concentration and purity (A260/280) after finishing gel-extraction.

3.3 Electroporation for Haploid Library Construction

1. Grow S. cerevisiae yeast cells (JAR200 for HC library or YVH10 for LC library) overnight in YPD media on a shaking incubator at 220 rpm and temperature set to 30  C. 2. Next day, inoculate aliquot of the overnight culture into 50 mL YPD media (initial OD600 is 0.2) (see Note 4). 3. Grow inoculated cells on shaking incubator at 220 rpm and 30  C until OD600 reaches approximately 1.4, corresponding to mid-log phase in growth curve (see Note 5). 4. Spin down grown yeast cells at 2500
g for 5 min and discard supernatant. 5. Wash the cells with 20 mL ice-cold ddH2O. 6. Repeat steps 4 and 5. 7. Spin down yeast cells and wash yeast cells again with 20 mL ice-cold yeast electroporation buffer [18].

322

Du-San Baek et al.

8. Spin down yeast cells and resuspend cells in 10 mL yeast competent buffer. 9. Incubate conditioned yeast cells on shaking incubator at 220 rpm and 30  C for 30 min. 10. Spin down yeast cells at 4  C, 2500
g for 5 min. 11. Wash once with 20 mL ice-cold yeast electroporation buffer. 12. Spin down again then resuspend yeast cell pellet in 500 μL ice-cold yeast electroporation buffer. 13. Keep electrocompetent yeast cells on ice until transformation. 14. Add 4 μg of digested vector and 12 μg of insert DNA to 400 μL electrocompetent yeast cells (see Notes 4 and 6). 15. Mix the cells and added DNA gently and transfer mixture to pre-chilled cuvette (0.2-cm electrode gap). 16. Electroporate the cells at 2.5 kV and 25 μF with Bio-Rad Gene Pulser II Electroporator (see Notes 7 and 8). 17. Resuspend transformed cells in 2 mL pre-warmed yeast resuspension buffer. 18. Transfer resuspended cells to additional 2.6 mL yeast resuspension buffer. 19. Incubate collected cells (approximately 5 mL) on a shaking incubator at 220 rpm and 30  C for an hour for the recovery of cells. 20. Spin down transformed cells at 2500
g for 5 min and discard supernatant. 21. Resuspend cells in 1 mL SDU (for HC library) or SDT (for LC library) media and transfer to 200 mL SDU or SDT media (pH 6.4) for the auxotrophic selection. In parallel, take part of resuspension and use it for the titration on SDU or SDT plate to determine library size by serial dilution (see Note 9). 22. Incubate cells on shaking incubator at 220 rpm and 30  C. 23. To make stock vials for library, resuspend 1
109 cells (assuming library size is 1
108 diversity) in 25% glycerol containing SDU or SDT media (w/v); storage buffer. 24. Vortex tube vigorously to thoroughly disperse the cells. 25. Place to vials in an isopropanol bath and store at 80  C deep freezer. 3.4 Purification of KRasG12D

1. Co-transform pET23m 8
His-Avi-KRasG12D and pBirAcm plasmid to BL21(DE3) E. coli strain (see Note 10) [5]. 2. Spread cells on a LBCA plate. 3. Incubate plate at 37  C overnight.

Yeast Mating for Antibody Discovery and Engineering

323

4. Inoculate large single colony to 5 mL LBCA media and incubate cells on shaking incubator at 37  C, 180 rpm for overnight. 5. Next day, transfer overnight culture to 200 mL LBCA media in 1 L flask (initial OD600 is 0.02) and incubate cells on shaking incubator at 37  C, 180 rpm. 6. When an OD600 of 0.6 is reached, add 0.2 mM IPTG and 8 mL solution containing 5 mM D-biotin and 5 mM Bicine (pH 8.3) to growing cells. 7. Grow induced cells for 4 h on shaking incubator at 30  C, 150 rpm. 8. Spin down cells at 4  C, 6000
g for 10 min. 9. Resuspend pellet in 15 mL E. coli resuspension buffer. 10. Place cells on ice and sonicate cells to lysis at 35% amplitude (5-s pulse/10-s rest cycles) three times with a 1/400 (6 mm) scale probe. 11. Spin down the whole lysate at 12,000
g for 30 min at 4  C. 12. Collect the supernatant and filter through a 0.45-μm PES syringe filter. 13. Pack 2 mL Ni-NTA resin into a 10 mL disposable column and let resin storage buffer flow through. 14. Wash resin with 20 mL ddH2O and let water flow through. 15. Close the bottom cap of disposable column and equilibrate resin with 4 mL resuspension buffer. 16. Add resin slurry to filtered supernatant and incubate for an hour at 4  C on vertical rotator. 17. Load mixture to disposable column and let all liquid flow out. 18. Wash the resin with 20 mL resuspension buffer. 19. Wash the resin with 50 mL resuspension buffer containing 10 mM imidazole. 20. Wash the resin with 50 mL resuspension buffer containing 20 mM imidazole. 21. Wash the resin with 50 mL resuspension buffer containing 30 mM imidazole. 22. Wash the resin with 20 mL resuspension buffer containing 40 mM imidazole. 23. Elute bound protein by adding resuspension buffer containing 300 mM imidazole. 24. Collect eluted fraction and concentrate it with 10,000 MWCO Amicon® centrifugal filter units. 25. Change buffer to KRas protein storage buffer through PD-10 desalting column.

324

Du-San Baek et al.

3.5 GppNHp and GDP Incubation with Purified KRasG12D

1. For GppNHp incubation, change storage buffer of biotinylated and purified recombinant 8
His-Avi-KRasG12D to GppNHp reaction buffer by PD-10 column [5]. 2. Add tenfold molar excess of GppNHp solution and 2 U of alkaline phosphatase immobilized agarose per mg recombinant 8
His-Avi-KRasG12D. 3. Incubate the mixture for 1 h at room temperature with gentle vertical rotation. 4. Spin down the mixture at 12,000
g for 10 min at 4  C and collect supernatant. 5. Change the reaction buffer to KRas protein storage buffer through PD-10 column. 6. Concentrate GppNHp bound KRasG12D protein (KRasG12DGppNHp) with 10,000 MWCO Amicon® centrifugal filter. 7. Store at 80  C. 8. For GDP incubation, change the KRas protein storage buffer of non-biotinylated and purified recombinant 8
His-AviKRasG12D to GDP reaction buffer using a PD-10 column (see Note 10). 9. Add 20-fold molar excess of GDP to protein in GDP reaction buffer. 10. Incubate mixture for 30 min at 30  C. 11. Add 60 mM MgCl2 and further incubate for 30 min at 4  C. 12. Change reaction buffer to KRas protein storage buffer through PD-10 column. 13. Concentrate GDP bound KRasG12D protein (KRasG12D-GDP) with 10,000 MWCO Amicon® centrifugal filter.

3.6 MACS Screening of HC Library Against KRasG12D-GppNHp

1. Thaw frozen aliquots of HC haploid yeast library at room temperature. 2. Spin down cells at 2500
g for 5 min and discard yeast storage buffer. 3. Resuspend in 1 L SDU media (pH 6.4) and incubate cells on shaking incubator at 220 rpm and 30  C for overnight. 4. Passage 1
109 cells (assuming HC library size is approximately 1
108 diversity) into fresh 200 mL SDU media (pH 6.4), followed by incubation on shaking incubator at 220 rpm and 30  C for overnight. 5. Spin down at least 1
109 cells from a passaged library culture at 2500
g for 5 min. 6. Resuspend cells in 200 mL SGU media (pH 6.4, initial OD600 is 0.5) to induce expression of HCs.

Yeast Mating for Antibody Discovery and Engineering

325

7. Incubate cells at 30  C on shaking incubator at 220 rpm and 30  C for 18 h (see Note 11). 8. Spin down at least 1
109 freshly induced yeast cells at 2500
g for 5 min. 9. Discard supernatant. 10. Resuspend cells in 50 mL PBSM buffer (see Note 12). 11. Repeat steps 8 and 9. 12. Resuspend cells in 10 mL PBSM buffer. 13. Add 1 μM biotinylated KRasG12D-GppNHp and non-biotinylated 10 μM KRasG12D-GDP (1:10 competition ratio) to resuspended cells [5]. 14. Incubate cells and antigen mixture for an hour at room temperature with gentle rotation. 15. Place mixture on ice for 10 min. 16. Spin down cells at 4  C, 2500
g for 5 min. 17. Discard supernatant containing unbound antigen. 18. Wash cells with 10 mL ice-cold PBSM buffer. 19. Repeat steps 16 and 17. 20. Resuspend cells with 5 mL ice-cold PBSM buffer. 21. Add 200 μL streptavidin microbeads to the 5 mL suspension. 22. Incubate the mixture on ice for 15 min and gently invert periodically during incubation. 23. Repeat steps 16 and 17. 24. Resuspend cells evenly with 10 mL ice-cold PBSM buffer and keep on ice. 25. Place an LS column on the magnet and stand assembly. 26. Wash column with 5 mL PBSM buffer. 27. Load resuspended cells to the column. 28. Wash column with 10 mL PBSM buffer. 29. Remove the column from magnet and place over a 15-mL collection conical tube. 30. Add 10 mL SDU media (pH 6.4) to the column and push the plunger to elute the remaining cells on column. 31. Grow eluted HC library cells on shaking incubator at 220 rpm and 30  C for overnight. 3.7 FACS Screening of HC Library Against KRasG12D-GppNHp

1. Spin down at least 5
107 cells of HC library obtained after one round of magnetic sorting (first MACS pool) at 8000
g for 1 min. 2. Resuspend cells in 10 mL SGU media (pH 6.4, initial OD600 is 0.5) to induce expression of HCs.

326

Du-San Baek et al.

3. Incubate cells on shaking incubator at 220 rpm and 30  C for 18 h. 4. Spin down approximately 5
107 freshly induced yeast cells at 8000
g for 1 min. 5. Discard supernatant. 6. Wash cells with 1 mL PBSF buffer. 7. Repeat steps 4 and 5. 8. Resuspend cells in 1 mL PBSF buffer. 9. Add 500 nM biotinylated KRasG12D-GppNHp and non-biotinylated 5 μM KRasG12D-GDP (1:10 competition ratio) to resuspended cells [5]. 10. Incubate cells and antigen mixture for an hour at room temperature with gentle rotation. 11. Repeat steps 4 and 5, and then place pellet on ice. 12. Resuspend cells in 1 mL ice-cold PBSF buffer. 13. Add FITC-conjugated anti c-Myc antibody (5 μg/mL as final concentration) and R-phycoerythrin-conjugated streptavidin (20 μg/mL as final concentration). 14. Incubate cells and secondary labeling reagent mixture for 30 min on ice with occasional inversion. 15. Spin down yeast cells at 4  C, 8000
g for 1 min. 16. Discard supernatant. 17. Resuspend cells in 2 mL ice-cold PBSF buffer. 18. Run flow cytometry and cell sorting to collect the top 0.5–1% of target-binding cells into freshly prepared SDU media (pH 6.4). 3.8 Mating of the Enriched HC Library 3.8.1 Mating of the Enriched HC Library with a Fixed LC with CytosolPenetrating Ability

To continue antibody screening in Fab format, enriched MATa haploid yeast cells of HC library can be mated with MATα haploid yeast cells of either fixed LC (this section) or LC library (Subheading 3.8.2) to generate diploid yeast cells enabling to display Fab fragment on yeast surface (Fig. 2). 1. Clone TMab4 LC to pYDS-K plasmid via NheI and BsiWI restriction enzyme sites. 2. Transform DNA to YVH10 strain by using the Frozen-EZ Yeast Transformation II Kit and select transformants on SDT plate for 2 days in 30  C incubator. 3. Inoculate single yeast colony into SDT media (pH 6.4) and culture overnight on shaking incubator at 220 rpm and 30  C. 4. Passage 1
107 cells of TMab4 LC haploid yeast cells to 10 mL SDT media (pH 4.5).

Yeast Mating for Antibody Discovery and Engineering

A

Design of HC and LC library Construction of HC and LC library

B

Haploid HC library VH CH1 c-myc

HC library (Haploid yeast, JAR200)

Fixed LC or LC library

1st MACS

327

Haploid LC library

Heavy chain (HC)

Light chain (LC)

VL CL

Aga2p Aga1p

flag flag

c-myc Aga2p VH CH1

VL C L pYDS-K

pYDS-H

URA3

TRP1

(Haploid yeast, YVH10) JAR200 (MATa, URA-, TRP+)

1st FACS

YVH10 (MATα, URA+, TRP-)

Mating

Mating

Sub-Fab library (Diploid yeast, JAR200×Y VH10)

Diploid Fab library VH

VL

CH1

CL

Fragment antigen-binding (Fab)

1st FACS c-myc Aga2p

2nd

VH CH1

FACS

flag

pYDS-H VL CL

Characterization of Single clones Sequencing

TRP1

pYDS-K URA3

JAR200×Y VH10 (URA+, TRP+)

Fig. 2 Selection of Fabs via yeast mating. (a) Flowchart for isolation of Fab from a combinatorial HC library and either fixed LC or LC library. (b) Schematic process for generating combinatorial Fab antibody library on yeast cell surface via yeast mating between two haploid strains carrying HC and LC libraries

5. In parallel, passage 1
107 cells of HC library obtained after one round of flow cytometric enrichment (first FACS pool in Subheading 3.7) into fresh 10 mL SDU media (pH 4.5). 6. Incubate two haploid cells in respective selection media on shaking incubator at 220 rpm and 30  C until reaching mid-log phase growth (OD600 between 1.2 and 1.6). 7. Collect 1.2
107 cells from each culture (approximately ~1 mL) and spin down at 8000
g for 1 min. 8. Discard supernatant and resuspend both cells in 100 μL pre-warmed YPD media (pH 4.5) and mix thoroughly by continuous pipetting up and down to initiate mating process as depicted in Fig. 2. 9. Drop resuspended cells onto the center of two pre-warmed YPD plates (pH 4.5). Drops should make approximately 1-cm-radius round circle with a final cell density as 2.5–3
107 cells/cm2. Do not spread out mechanically [1].

328

Du-San Baek et al.

10. Incubate plate on the bench to allow liquid to soak into the plate for 30 min. 11. Relocate a plate to 30  C incubator without shaking and incubate it for ~5–6 h. Keep monitoring for dumbbell-shaped zygotes using microscopy. If it appears that less than 30–50% of yeast cells are in zygote formation, longer incubation may be required. 12. After ~5–6 h, collect mated yeast cells completely by pipetting with 1 mL SD media (pH 6.4). 13. Spin down mated cells at 8000
g for 1 min and subsequently wash with 1 mL sterilized ddH2O and repeat this step three times to remove the remaining YPD media completely. 14. Transfer mated cells to 20 mL SD media (pH 6.4, initial OD600 is 0.15) and incubate cells overnight to allow for outgrowing of diploid cells. 15. The next day, spin down 500 μL culture and resuspend cells in 10 mL SG media (pH 6.4, initial OD600 is approximately 0.5) to induce expression of Fabs. 16. Incubate cells on shaking incubator at 220 rpm and 30  C for 18 h. 17. After induction of sub-Fab library in SG media (pH 6.4), conduct subsequent FACS screening in more stringent condition. For second and third rounds of FACS, use 50 nM biotinylated KRasG12D-GppNHp and non-biotinylated 5 μM KRasG12D-GDP (1:100 competition ratio) to isolate highly selective Fab binders [5]. 18. After finishing screenings, spread cells to SD plate and inoculate yeast colony to 3 mL SG media for induction, followed by incubation on shaking incubator at 220 rpm and 30  C for 18 h. 19. Characterize single clones using flow cytometry with desirable antigen concentration. 3.8.2 Mating of the Enriched HC Library with the Initial LC Library

1. Thaw frozen aliquots of the LC haploid yeast library at room temperature. 2. Spin down yeast cells at 2500
g for 5 min and discard storage buffer. 3. Resuspend cells in 1 L SDT media (pH 6.4) and incubate on shaking incubator at 220 rpm and 30  C for overnight. 4. Passage 1
109 cells (assuming LC library size is approximately 1
108 diversity) to fresh 200 mL SDT media (pH 6.4) and incubate cells on shaking incubator at 220 rpm and 30  C for overnight.

Yeast Mating for Antibody Discovery and Engineering

329

5. Spin down 1
109 cells from a passaged LC library culture at 2500
g for 5 min and discard media. 6. Resuspend the cells in 200 mL SDT media (pH 4.5) and incubate for overnight on shaking incubator at 30  C, 220 rpm. 7. Spin down 2
108 cells from above LC library culture at 2500
g for 5 min. 8. During steps 4–7, in parallel, prepare 2
108 cells for HC library obtained after one round of flow cytometric enrichment (first FACS pool) in SDU media (pH 4.5). 9. Transfer grown haploid yeast cells (2
108 cells/each library) for enriched HC or initial LC library to 400 mL SDU or SDT media (pH 4.5, initial OD600 is 0.1) at the same time to synchronize cell growth. 10. Incubate both the libraries until reaching to mid-log phase (OD600 is between 1.2 and 1.6) on shaking incubator at 30  C, 220 rpm. 11. Mix 4
109 haploid cells (2
109 cells/each library, requires almost half of culture) and vortex at least 5 min to allow for thorough mixing and single cell suspension. 12. Spin down cells at 2500
g, 5 min, and wash one time with 40 mL sterilized ddH2O. 13. Spin down cells again and discard supernatant. 14. Resuspend mixed cells in 3 mL pre-warmed YPD media (pH 4.5) by vigorous pipetting. It should be completely resuspended with no clumps. 15. Drop resuspended cells onto the center of two pre-warmed YPD plates (pH 4.5, in 100pi dish; ~78.5 cm2 area) equally without mechanical spreading. Orbitally rotate the plate by hand gently and spread out evenly to make the equal density at every point of plate. Final cell density will reach out 2.5–3
107 cells/cm2 [1]. 16. Incubate the two plates in non-shaking incubator at 30  C for ~5–6 h. Keep monitoring dumbbell-shaped zygotes in microscopy. If necessary, allow for longer incubations. 17. After ~5–6 h, collect mated yeast cells completely by pipetting with 10 mL SD media. 18. Spin down mated cells at 2500
g, 5 min and subsequently wash with 40 mL sterilized ddH2O and repeat three times. 19. Check OD600 and dilute sample properly to calculate mating efficiency and library size (see Note 13). 20. Inoculate mated cells to 1 L SD media (pH 6.4, initial OD600 is 0.5) and incubate cells on shaking incubator at 30  C and 220 rpm for overnight to allow outgrowing of diploid cells.

330

Du-San Baek et al.

21. Spin down at least 5
109 cells at 2500
g for 5 min. 22. Resuspend cells in 1 L SG media (pH 6.4, initial OD600 is 0.5). 23. Incubate cells at 30  C on shaking incubator at 220 rpm and 30  C for 18 h for Fab induction. 24. After induction of Fab diploid library, conduct subsequent MACS and FACS screening in more stringent conditions. For second MACS, use 100 nM biotinylated KRasG12D-GppNHp and non-biotinylated 5 μM KRasG12D-GDP (1:50 competition ratio), and for second and third FACS, use 50 nM biotinylated KRasG12D-GppNHp and non-biotinylated 5 μM KRasG12DGDP (1:100 competition ratio) to isolate highly selective Fab binders (see Note 14) [5]. 20. After finishing screenings, spread cells to SD plate and inoculate yeast colony to 3 mL SG media for induction, followed by incubation on shaking incubator at 220 rpm and 30  C for 18 h. 25. Characterize single clones using flow-cytometry with desirable antigen concentration. 3.9 Sequencing of VH and VL

1. Rescue both HC and LC plasmid DNA from identified single yeast clone using yeast DNA mini-prep kit. 2. Prepare Q5® High-Fidelity DNA polymerase reaction mix for a typical 50–100 μl reaction per tube as following: (a) 25 ng rescued DNA (step 1) (VH3-23 cloned pYDS-H or Vκ1-16 cloned pYDS-K plasmid). (b) 200 μM each deoxyribonucleotide triphosphate (dNTP). (c) 0.5 μM sequencing forward primer (Tm: ~53  C). (d) 0.5 μM sequencing reverse primer (Tm: ~53  C). (e) 1
Q5® reaction buffer. (f) 0.02 U/μL of Q5® polymerase. (g) Bring volume to 50–100 μL with ddH2O. 3. Run PCR reaction on following program: (a) Initial denaturation: 98  C (1 min). (b) 30 cycles: 98  C (10 s), 55  C (30 s), and 72  C (15 s). (c) Final extension: 72  C (2 min). (d) Cooling: 16  C (5 min). 4. Purify amplified VH and VL DNA fragments after electrophoresis on 1.2% agarose gel. 5. Measure concentration and send samples with respective sequencing forward primers to identify nucleotide and amino acid sequence of selected Fab binders.

Yeast Mating for Antibody Discovery and Engineering

4

331

Notes 1. If had a plan to use other plasmid backbones instead of pYDSH and pYDS-K plasmid, the nucleotide sequences in primers (H1-For, K1-For, H4-Rev, and K4-Rev) for the homologous recombination sites (HMS) should be replaced by desirable sequences in other plasmids. 2. The annealing temperature with Q5® polymerase tends to be higher than with typical polymerases. If planning to use other polymerases, annealing temperature should be adjusted accordingly. 3. The optical density value (A260/280) for purity of prepared inset DNA should be approximately 1.8. To prepare highly pure insert DNA, wash the column more than twice with 70% ethanol during DNA purification. 4. 50 mL culture scale is for one time electroporation. If more electroporation is required to achieve the desired library size, scale-up is necessary. In general, use of 4 μg of digested vector and 12 μg of insert DNA generates ~1–3
108 library size. 5. This protocol was written based on ~1
107 yeast cells/mL corresponding to a value of “1” for absorbance at 600 nm wavelength (OD600). Before starting yeast work, titration should be performed to calculate how many yeast cells would correspond to absorbance at OD600 measured from your spectrophotometer. For example, based on OD600 measured by Multiskan GO Microplate Spectrophotometer (Thermo Fisher), the titration revealed the following cell densities: (a) JAR200 strain (MATa): OD600 ¼ 1 is corresponded to ~2
107 yeast cells/mL. (b) YVH10 strain (MATα): OD600 ¼ 1 is corresponded to ~1
107 yeast cells/mL. 6. The DNA mixture in water should be less than 50 μL. Reduce the volume by precipitation and resuspension in a smaller volume if necessary. 7. Among many electroporators in the market, only Bio-Rad Gene Pulser I or II should be in use for this transformation protocol. When we tested other latest models like Bio-Rad Gene Pulser Xcell, transformation efficiency was significantly lowered while we continued to obtain >108 transformants per electroporation with Bio-Rad Gene Pulser II [18]. Alternatively, the Frozen-EZ Yeast Transformation II Kit (Zymo Research) can be used for generating 106 to 107 library size without strict requirement of electroporator. 8. Typical time constant (TC) ranges from 3.0 to 4.5 ms [18].

332

Du-San Baek et al.

9. The selection of each of the libraries proceeds in the SD media (Ura-/Trp-) supplemented with uracil for the yeast containing HC-encoded plasmid (Trp+) and supplemented with tryptophan for the yeast containing LC-encoded plasmid (Ura+) [1]. 10. In absence of pBirAcm plasmid, E. coli strain can produce non-biotinylated protein. For production of non-biotinylated KRasG12D-GDP, only pET23m 8
His-Avi-KRasG12D plasmid should be used for transformation and cells can be selected in LBA plate or media. Also, supplement of D-biotin is not required. 11. The addition of 2 g/L D-(+)-Glucose to induction media (SG, SGU and SGT) can improve surface display amounts in general. 12. Conveniently, addition of 100
BSA solution to commercial DPBS is an alternative way to make 0.5% BSA (w/v) containing PBSM buffer. 13. To estimate the mating efficiency, serially dilute partial washed cells to SD media and spread out onto SD and SDT agar plates, respectively. Incubate plates at 30  C for 2 days and count the number of colonies. Percentage mating efficiency can be calculated as the number of diploid colonies in SD plates divided by the number of total colonies in SDT plates [1]. Mating efficiencyð%Þ ¼

number of colonies on SD plate number of colonies on SDT plate
100%

14. During the screening, binders against secondary reagents are likely to be enriched and can generate false-positive signal in flow cytometry analysis. Thus, alternative use of secondary reagents with changed labeling schemes should be applied properly. Instead of streptavidin, neutravidin or anti-biotin antibody can be applied. Alexa488, 647, or others can be used as alternatives for PE or FITC.

Acknowledgments This work was supported by a grant from the Priority Research Center Program (2019R1A6A1A11051471 to YSK) through the National Research Foundation of Korea (NRF), funded by the Ministry of Science, ICT & Future Planning, Republic of Korea, and a grant funded by Samsung Future Technology Center [grant number SRFC-MA1802-09 to YSK].

Yeast Mating for Antibody Discovery and Engineering

333

References 1. Baek DS, Kim YS (2014) Construction of a large synthetic human Fab antibody library on yeast cell surface by optimized yeast mating. J Microbiol Biotechnol 24(3):408–420. https://doi.org/10.4014/jmb.1401.01002 2. Baek DS, Kim YS (2015) Humanization of a phosphothreonine peptide-specific chicken antibody by combinatorial library optimization of the phosphoepitope-binding motif. Biochem Biophys Res Commun 463(3): 414–420. https://doi.org/10.1016/j.bbrc. 2015.05.086 3. Baek DS, Kim JH, Kim YJ, Kim YS (2018) Immunoglobulin Fc-fused peptide without C-terminal Arg or Lys residue augments neuropilin-1-dependent tumor vascular permeability. Mol Pharm 15(2):394–402. https:// doi.org/10.1021/acs.molphar maceut. 7b00761 4. Choi HJ, Kim YJ, Choi DK, Kim YS (2015) Engineering of immunoglobulin Fc heterodimers using yeast surface-displayed combinatorial Fc library screening. PLoS One 10(12): e0145349. https://doi.org/10.1371/journal. pone.0145349 5. Shin SM, Choi DK, Jung K, Bae J, Kim JS, Park SW, Song KH, Kim YS (2017) Antibody targeting intracellular oncogenic Ras mutants exerts anti-tumour effects after systemic administration. Nat Commun 8:15090. https://doi.org/10.1038/ncomms15090 6. Kim JE, Lee DH, Jung K, Kim EJ, Choi Y, Park HS, Kim YS (2021) Engineering of humanized antibodies against human interleukin 5 receptor alpha subunit that cause potent antibodydependent cell-mediated cytotoxicity. Front Immunol 11:593748. https://doi.org/10. 3389/fimmu.2020.593748 7. Boder ET, Wittrup KD (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol 15(6):553–557. https://doi.org/10.1038/nbt0697-553 8. Cohen-Khait R, Schreiber G (2016) Low-stringency selection of TEM1 for BLIP shows interface plasticity and selection for faster binders. Proc Natl Acad Sci U S A 113(52): 14982–14987. https://doi.org/10.1073/ pnas.1613122113 9. Lee CH, Park KJ, Sung ES, Kim A, Choi JD, Kim JS, Kim SH, Kwon MH, Kim YS (2010) Engineering of a human kringle domain into agonistic and antagonistic binding proteins functioning in vitro and in vivo. Proc Natl Acad Sci U S A 107(21):9567–9571. https:// doi.org/10.1073/pnas.1001541107 10. Sadio F, Stadlmayr G, Eibensteiner K, Stadlbauer K, Ruker F, Wozniak-Knopp G

(2020) Methods for construction of yeast display libraries of four-domain T-cell receptors. Methods Mol Biol 2070:223–248. https:// doi.org/10.1007/978-1-4939-9853-1_13 11. Schutz M, Batyuk A, Klenk C, Kummer L, de Picciotto S, Gulbakan B, Wu Y, Newby GA, Zosel F, Schoppe J, Sedlak E, Mittl PRE, Zenobi R, Wittrup KD, Pluckthun A (2016) Generation of fluorogen-activating designed ankyrin repeat proteins (FADAs) as versatile sensor tools. J Mol Biol 428(6):1272–1289. https://doi.org/10.1016/j.jmb.2016.01.017 12. Pepper LR, Cho YK, Boder ET, Shusta EV (2008) A decade of yeast surface display technology: where are we now? Comb Chem High Throughput Screen 11(2):127–134. https:// doi.org/10.2174/138620708783744516 13. Lim S, Glasgow JE, Filsinger Interrante M, Storm EM, Cochran JR (2017) Dual display of proteins on the yeast cell surface simplifies quantification of binding interactions and enzymatic bioconjugation reactions. Biotechnol J 12(5):1600696. https://doi.org/10. 1002/biot.201600696 14. Lou J, Geren I, Garcia-Rodriguez C, Forsyth CM, Wen W, Knopp K, Brown J, Smith T, Smith LA, Marks JD (2010) Affinity maturation of human botulinum neurotoxin antibodies by light chain shuffling via yeast mating. Protein Eng Des Sel 23(4):311–319. https:// doi.org/10.1093/protein/gzq001 15. Cappellaro C, Baldermann C, Rachel R, Tanner W (1994) Mating type-specific cell-cell recognition of Saccharomyces cerevisiae: cell wall attachment and active sites of a- and alphaagglutinin. EMBO J 13(20):4737–4744 16. Weaver-Feldhaus JM, Lou J, Coleman JR, Siegel RW, Marks JD, Feldhaus MJ (2004) Yeast mating for combinatorial Fab library generation and surface display. FEBS Lett 564(1-2): 24–34. https://doi.org/10.1016/S00145793(04)00309-6 17. Shin SM, Kim JS, Park SW, Jun SY, Kweon HJ, Choi DK, Lee D, Cho YB, Kim YS (2020) Direct targeting of oncogenic RAS mutants with a tumor-specific cytosol-penetrating antibody inhibits RAS mutant-driven tumor growth. Sci Adv 6(3):eaay2174. https://doi. org/10.1126/sciadv.aay2174 18. Benatuil L, Perez JM, Belk J, Hsieh CM (2010) An improved yeast transformation method for the generation of very large human antibody libraries. Protein Eng Des Sel 23(4):155–159. https://doi.org/10.1093/ protein/gzq002

Chapter 18 Humanization of Chicken-Derived Antibodies by Yeast Surface Display Jan P. Bogen, Adrian Elter, Julius Grzeschik, Bjo¨rn Hock, and Harald Kolmar Abstract Chicken-derived antibodies emerged as a promising tool for diagnostic and therapeutic usage. Due to the phylogenetic distance between birds and mammals, chicken immunization campaigns with human antigens result in a chicken antibody (IgY) repertoire targeting epitopes not addressed by rodent-derived antibodies. However, this phylogenetic distance accounts for a low homology of IgY molecules to human antibodies, resulting in potential immunogenicity and thus excluding IgYs from therapeutic applications. Herein, we describe a straightforward method to efficiently humanize chicken-derived antibodies by a CDR-graftingbased approach, including a simultaneous randomization of key residues (Vernier residues). Utilizing yeast surface display (YSD) and fluorescence-activated cell sorting (FACS), yeast cells displaying functional humanized scFvs and Fab variants are isolated, and subsequent next-generation sequencing (NGS) enables the identification of humanized antibody variants with restored affinity and beneficial protein characteristics. Key words Chicken antibody, Humanization, Yeast surface display, Fluorescence-activated cell sorting, CDR grafting, Vernier residues, Next-generation sequencing

1

Introduction Anti-drug antibodies (ADA), antibodies raised by the patients’ immune system against therapeutic molecules, are known since the clinical usage of the first approved monoclonal antibody (mAb) muromonab, diminishing its therapeutic efficacy [1, 2]. To overcome the obstacles of ADAs, chimerization technologies were introduced where the VH and VL domains are of rodent origin while the constant domains of the mAb are derived from the human germline [3]. The CD20-binding chimeric monoclonal antibody (mAb) rituximab was the first approved antibody generated using

Jan P. Bogen and Adrian Elter contributed equally to this work. Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_18, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

335

336

Jan P. Bogen et al.

this technology [4]. To further reduce the immunogenicity, in 1986, the Winter group transplanted the complementarity determining regions (CDRs) of a mouse mAb onto a human acceptor framework, yielding a humanized fragment variable (Fv) [5]. Alemtuzumab, humanized by CDR grafting, was the first of its kind to gain approval [4]. Chicken-derived antibodies became of interest in recent years due to their broad epitope coverage and options for straightforward library generation. Their phylogenetic distance to humans accounts for a strong immune response against epitopes on conserved mammalian target proteins, not addressable by rodent immunization approaches [6]. By now, Sym021, targeting PD-1, is the first humanized chicken-derived mAb in clinical development [7]. Even though in silico humanization technologies exist for chicken mAbs [8], CDR grafting remains the most straightforward method. Our group recently established a CDR-grafting-based humanization approach for chicken-derived mAbs, including the partial randomization of Vernier residues utilizing yeast surface display (YSD) [9]. Those Vernier residues are amino acids critical for the correct CDR orientation and therefore antigen binding [10]. Using FACS, a YSD library of humanized variants was screened, and the resulting repertoire of functional binders was analyzed by NGS, yielding humanized mAbs with parental affinity and improved biophysical characteristics. In this chapter, we describe this stepwise humanization process starting from the CDR-grafting in silico, including the randomization of Vernier residues. We discuss the library generation and screening. The resulting clones are analyzed by NGS, and we illuminate potential options to identify the best fitting humanized antibody.

2

Materials

2.1 Cloning of Humanized VH and VL Domains

1. Q5® High-Fidelity DNA polymerase (New England Biolabs). 2. 5
Q5® buffer (New England Biolabs). 3. dNTPs (New England Biolabs). 4. Oligonucleotides/primers for the amplification of VH/VL coding genes. 5. Oligonucleotides encoding for humanized VH or VL. 6. Nuclease-free water. 7. Thermocycler. 8. Device and reagents for agarose gel electrophoresis. 9. Wizard SV gel and PCR cleanup system (Promega) or an equivalent DNA purification system. 10. BioSpec Nano or equivalent instrumentation.

Humanization of Chicken-Derived Antibodies

2.2 Construction of Yeast Surface Display Library

337

1. Yeast strain: Saccharomyces cerevisiae strain EBY100. 2. YPD media: 20 g/L D-(+)-glucose, 20 g/L tryptone, 10 g/L yeast extract and 100 μg/ml ampicillin. 3. Electroporation buffer: 1 M Sorbitol, 1 mM CaCl2. 4. LiAc buffer: 0.1 M LiAc, 10 mM DTT. 5. 1 M Sorbitol. 6. PBS pH 7.4: 8.1 g/L NaCl, 1.13 g/L Na2HPO4, 0.75 g/L KCl, and 0.27 g/L KH2PO4. 7. Electroporation system GenePulser Xcell™ (Bio-Rad). 8. 0.2 cm Electroporation cuvettes (Bio-Rad). 9. Bacto™ Casamino acids (BD Biosciences). 10. SD-CAA: 8.6 g/L NaH2PO4 · H2O, 5.4 g/L Na2HPO4, 1.7 g/L yeast nitrogen base without amino acids, 5 g/L ammonium sulphate, 5 g/L Bacto Casamino Acids, 20 g/L glucose, and 100 μg/ml ampicillin (+12 g/L agar agar for agar plates). 11. SG-CAA: 8.6 g/L NaH2PO4 · H2O, 5.4 g/L Na2HPO4, 1.7 g/L yeast nitrogen base without amino acids, 5 g/L ammonium sulphate, 5 g/L Bacto Casamino Acids, 20 g/L galactose and 100 μg/ml ampicillin. 12. 9-cm Petri dishes.

2.2.1 Humanized scFv Library-Specific Materials

1. BamHI-HF® (New England Biolabs). 2. NheI-HF® (New England Biolabs). 3. 10
Cutsmart Buffer (New England Biolabs). 4. pCT plasmid [11]. 5. Wizard SV gel and PCR cleanup system (Promega).

2.2.2 Humanized Fab Library Specific Materials

1. BsaI-HFv2® (New England Biolabs). 2. T4 DNA Ligase® (New England Biolabs). 3. T4 DNA Ligase Buffer (New England Biolabs). 4. pDest (Lambda) [12]. 5. pEntry [12]. 6. Wizard SV gel and PCR cleanup system (Promega).

2.3 Library Screening of Humanized scFvs/ Fabs

1. PBS-B: PBS, 0.1% (w/v) bovine serum albumin. 2. Fc-tagged EGFR-extra cellular domain (ECD) (R & D Systems). 3. Goat anti-Human (Invitrogen).

IgG

Fc

Secondary

4. Anti-c-myc-Biotin antibody (Miltenyi Biotec).

Antibody,

PE

338

Jan P. Bogen et al.

5. Streptavidin-APC (ThermoFisher). 6. Goat Anti-Human 647 (SouthernBiotech).

Lambda-Alexa

Fluor®

7. BD Influx or comparable cell sorter. 2.4 Next-Generation Sequencing

1. Zymoprep Yeast Plasmid Miniprep (Zymo Research). 2. Q5® High-Fidelity DNA polymerase (New England Biolabs). 3. 5
Q5® buffer (New England Biolabs). 4. dNTPs (New England Biolabs). 5. NGS (barcode) primer. 6. Nuclease-free water. 7. Thermocycler. 8. Geneious Prime or comparable software.

2.5 Reformatting, Expression, Purification, and Characterization

1. Q5® High-Fidelity DNA polymerase (New England Biolabs). 2. 5
Q5® buffer (New England Biolabs). 3. dNTPs (New England Biolabs). 4. Nuclease-free water. 5. Thermocycler. 6. pTT5-derived destination vector [13]. 7. CH1-CH2-CH3 entry vector [13]. 8. Lambda entry vector [13]. 9. CH2-CH3 entry vector [9]. 10. T4 DNA ligase Buffer (New England Biolabs). 11. T4 DNA Ligase (New England Biolabs). 12. SapI (New England Biolabs). 13. Competent XL1 blue E. coli cells. 14. Ampicillin LB agar plates and media (100 μg/mL). 15. PureYield Plasmid Midiprep System (Promega). 16. ExpiHEK293F (Thermo Fisher). 17. Lipofectamin (Thermo Fisher). 18. 20% tryptone solution. 19. New Brunswick™ S41i (Eppendorf) or comparable cell culture shaker. 20. Chromatography system and Protein A purification columns.

Humanization of Chicken-Derived Antibodies

3

339

Methods In the following section, we describe the identification of CDR regions within chicken-derived antibodies, the in silico grafting onto a human acceptor framework, including the randomization of important Vernier residues as well as the subsequent PCR-based gene assembly of oligonucleotides. Based on these humanized VH and VL genes, yeast surface display libraries are generated by either homologous recombination for humanized scFv libraries or by Golden Gate Assembly for humanized Fab libraries. Isolation of functional variants by FACS yields a number of humanized variants that are subsequently analyzed by NGS, followed by selection of variants that show the strongest enrichment over the course of two sorting rounds. Production of the most enriched variants proved the successful humanization of chicken-derived antibodies without losing parental affinity and retaining or even improving basic biophysical properties. As an example, we demonstrate the humanization process based on a chicken-derived antibody clone (E1) specific for the epidermal growth factor receptor (EGFR), previously isolated from an immunized chicken [9, 14].

3.1 In Silico Humanization of Chicken-Derived Antibodies

1. For the successful humanization of chicken-derived antibodies by a CDR-grafting-based approach, the correct identification of the CDR regions is key. The CDR can be annotated utilizing the standard Kabat numbering scheme [15]. 2. The human acceptor framework is chosen based on its homology to the parental chicken antibody. The most homologous sequence was identified by comparing the chicken VH and VL sequences to the human antibody repertoire utilizing the Igblast algorithm (https://www.ncbi.nlm.nih.gov/igblast/). We identified the human germline genes IGHV3-23 and IGHJ4 for the VH and the IGLV3-25 and IGLJ2 for the VL to be the most homologous genes to our exemplary chicken antibody (see Note 1). These human acceptor germline sequences are shown in Table 1. 3. The coding sequences of the CDRs are grafted in silico in frame between the acceptor framework sequences (CDR1 between FR1 and FR2, CDR2 between FR2 and FR3, CDR3 between FR3 an FR4), resulting in a DNA sequence, which exhibits human framework and chicken CDR sequences. A sequence alignment of the parental E1 clone and the grafted humanized antibody is depicted in Fig. 1. This corresponds to a classic CDR-grafting approach. However, these sequences will most probably not encode a functional antibody [9].

Table 1 Amino acid and DNA sequences of human germline acceptor sequences are shown, divided by domain and corresponding framework region. Vernier residues and corresponding codons are marked in red

VH Amino acid GAGGTGCAGCTGTTGGAGTCTGGGGGAGGCTTGG DNA TACAGCCTGGGGGGTCCCTGAGACTCTCCTGTGCA sequence GCCTCTGGATTCACCTTTAGC Amino WVRQAPGKGLEWVS acid DNA TGGGTCCGCCAGGCTCCAGGGAAGGGGCTGGAGT sequence GGGTCTCA

EVQLLESGGGLVQPGGSLRLSCAASGFTFS Fr 1

IGHV 3-23

Fr 2

RFTISRDNSKNTLYLQMNSLRAEDTAVYYCAK Fr 3

CGGTTCACCATCTCCAGAGACAATTCCAAGAACA CGCTGTATCTGCAAATGAACAGCCTGAGAGCCGA GGACACGGCCGTATATTACTGTGCGAAA WGQGTLVTVSS

IGHJ4

Fr 4

TGGGGCCAAGGAACCCTGGTCACCGTCTCCTCA

Amino acid DNA sequence Amino acid DNA sequence

VL SYELMQPPSVSVSPGQTARITC Fr 1

TCCTATGAGCTGATGCAGCCACCCTCGGTGTCAGT GTCCCCAGGACAGACGGCCAGGATCACCTGC WYQQKPGQAPVLVIY

IGLV3 Fr -25 2

TGGTACCAGCAGAAGCCAGGCCAGGCCCCTGTGC TGGTGATATAT

IGLJ2

Fr 4

Amino acid DNA sequence

Amino acid GGGATCCCTGAGCGATTCTCTGGCTCCAGCTCAGG DNA GACAACAGTCACGTTGACCATCAGTGGAGTCCAG sequence GCAGAAGATGAGGCTGACTATTACTGT Amino FGGGTKLTVL acid DNA TTCGGCGGAGGGACCAAGCTGACCGTCCTA sequence GIPERFSGSSSGTTVTLTISGVQAEDEADYYC

Fr 3

Amino acid DNA sequence

Humanization of Chicken-Derived Antibodies

341

Fig. 1 Exemplary sequence alignment of the parental E1 clone to its CDR-grafted counterpart. CDRs are annotated in red. Vernier residues are depicted in blue

4. Framework sequences exhibit Vernier residues responsible for the correct orientation of the CDRs enabling antigen recognition. In order to identify Vernier residues of importance and back mutate them to their chicken counterpart, some Vernier residues are partially randomized using degenerated codons. These codons encode either the human or the chicken amino acid residue at their specific position. Due to the nature of degenerated codons, additional residues with neither human nor avian origin may be encoded (see Note 2). In case that these residues have a negative impact on the humanized antibody, their frequency within the library should decline over the course of sorting. A table of all Vernier residues, their human and chicken sequences, as well as the chosen degenerated codons are given in Table 2. Potential differences between any other parental molecule and the chosen Vernier residues depicted in Table 2 can, but do not have to, be taken into account (see Note 3). 5. The numbers of potential VH variants (1024) multiplied with the number of potential VL variants (288) results in a maximal theoretical library size of 2.94
105 Vernier combinations. 6. Based on the in silico grafted and randomized sequences, oligonucleotides are designed, exhibiting a length of 60–90 bp with overlapping sequences of 20–30 bp (see Note 4). Exemplary primers can be found in Table 3. The alignment of the VH oligos is depicted in Fig. 2, the alignment of the VL oligos is shown in Fig. 3. The exact cloning procedure is described in Subheading 3.2.

342

Jan P. Bogen et al.

Table 2 Vernier residues and respective degenerated codons for in silico randomization Germline residues Kabat no.

Human

Chicken

Additional residues

Degenerated codon

VH

47 49 67 75 76 78

Trp Ser Phe Lys Asn Leu

Phe Ala Ala Gln Ser Val

Cys, Leu Gly, Thr Val, Ser Thr/Pro / /

TKS RSC KYC MMA ARC STA

VL

46 66 69 71

Leu Ser Thr Val

Thr Ala Ser Ala

Ser, Met Asn, Thr, Tyr, Asp Asn Thr, Ile

WYG DMC AVC RYA

3.2 Cloning of Humanized VH and VL Domains

In this section, the cloning of the VL and VH genes, based on overlapping oligonucleotides encoding the acceptor framework, the grafted CDRs, and the randomized Vernier residues is described. 1. Fusion of oligonucleotides is based on PCR, utilizing the Q5® High-Fidelity DNA polymerase according to the manufacture’s protocol. 2. Two reactions, one for the VH and one for the VL, respectively, are performed. For the VH reaction, oligos 1–7 (Table 3) are utilized, for the VL reaction, oligos 12–17 are chosen. All respective oligonucleotides and the flanking primers are mixed for one reaction. 3. Flanking primers either introduce overhangs for golden gate cloning (Fab) or subsequent scFv construction. 4. If a Fab is to be constructed, utilize the primers 8 and 11 for the VH sequence and the primers 18 and 21 for the VL sequence (Table 3). 5. For the construction of a scFv, utilize the primers 9 and 10 for the VH and 19 and 20 for the VLs (Table 3). 6. An alignment of the utilized oligos, the flanking primers and the desired sequence is depicted in Fig. 2 for the VH and in Fig. 3 for the VL sequences. 7. In brief, one reaction comprises: (a) 10 μL 5
Q5 Reaction Buffer. (b) 1 μL dNTPs (10 mM).

Table 3 Oligonucleotides for library generation. Permutated Vernier residues are marked in red. Overhangs for subsequent Golden Gate Assembly are highlighted in blue, while overhangs for the fusion into an scFv format are shown in pink. Homologous sequences to the pCT vector utilized for homologous recombination are shown in green. The CDR encoding sequences are highlighted in yellow and bold, with red (CDR1), blue (CDR2), or green (CDR3) letters. Flanking primers for amplification of full-length VH and VL sequences, introducing overhangs for scFv-fusion or Golden Gate cloning, are written in italic Resulti ng templa te

Oligo Name

1

VH

E1 VH Fr1 for

2

VH

E1 VH Fr1 CDR1 FR2 for

3

VH

E1 VH CDR1 Fr2 CDR2 for

4

VH

E1 VH CDR2 for

5

VH

E1 VH CDR2 Fr3 for

6

VH

E1 VH Fr3 CDR3 for

7

VH

E1 VH CDR3 Fr4 rev

Fab

8

VH

VH chicken hum for GGA

scFv

9

VH

VH Chicken hum for scFv

scFv

10

VH

VH chicken hum rev scFv

Library type

Fab/scFv

Fab/scFv

Fab/scFv

Fab/scFv

Fab/scFv

Fab/scFv

Fab/scFv

Oligo number

Sequence (5'-3') GAGGTGCAGCTGTTGGAGT CTGGGGGAGGCTTGGTACA GCCTGGGGGGTCCCTGAGA CTCTCC GCCTGGGGGGTCCCTGAGA CTCTCCTGTGCAGCCTCTG GATTCACCTTTAGCAGCTT CAACATGCTCTGGGTCCG CC CAACATGCTCTGGGTCCG CCAGGCTCCAGGGAAGGG GCTGGAGTKSGTTRSCGAC ATTTACAGCACTGGTAGT TAC CAGCACTGGTAGTTACAC GAGATACGCGCCGGCGG TGGATGGCC CGCGCCGGCGGTGGATG GCCGGKYCACCATCTCCAG AGACAATTCCMMAARCAC GSTATATCTGCAAATGAAC AGCCTGAGAGCCG CTGCAAATGAACAGCCTGA GAGCCGAGGACACGGCCG TATATTACTGTGCGAAAAG TTCTACTAGTGGTTTTTG TGGTGGTGTTAGTTG TGAGGAGACGGTGACCAG GGTTCCTTGGCCCCATGCG TCGATGAGTCCGCCACAA CTAACACCACCACAAAAA CCACTAGTAGAACT GCGCGCGCGGTCTCAAGGT GAGGTGCAGCTGTTGGAGTC TGGGGG GGTGGTGGTGGTTCTGGTGG TGGTGGTTCTGCTAGCGAGG TGCAGCTGTTGGAGTCTGGG GG TCCGCCCCCCGACCCGCCG CCGCCTGAGCCGCCTCCCC CTGAGGAGACGGTGACCAG GGTTCCTTG

Concentrati on in the 1. PCR

20 nM

20 nM

20 nM

20 nM

20 nM

20 nM

20 nM

500 nM

500 nM

500 nM

Table 3 (Continued)

Fab

11

VH

12

VL

13

VL

14

VL

Fab/scFv

15

VL

Fab/scFv

16

VL

Fab/scFv

17

VL

Fab

18

VL

scFv

19

VL

scFv

20

VL

Fab

21

VL

Fab/scFv

Fab/scFv

Fab/scFv

GCGCGCTGGTCTCTTAGTAG VH chicken AAGCTGAGGAGACGGTGACC hum rev GGA AGGGTTCCTTG TCCTATGAGCTGATGCAGC E1 VL Fr1 CACCCTCGGTGTCAGTGTC for CCCAGGACAGACGGCCAG GATCACCTGCTCC GACGGCCAGGATCACCTGC E1 VL Fr1 TCCGGGGGTGTTAACAGC CDR1 Fr2 AACCACTATGGCTGGTAC for CAGCAGAAGCCAGGCCAG GCCC GCAGAAGCCAGGCCAGGC E1 VL Fr2 CCCTGTGWYGGTGATATAT CDR2 Fr3 GCTAACACCAACAGGCCC TCGGGGATCCCTGAGCGAT for TCTC GGGATCCCTGAGCGATTCT CTGGCTCCDMCTCAGGGAV E1 VL Fr3 CACARYAACGTTGACCATC for AGTGGAGTCCAGGCAGAA G CAGTGGAGTCCAGGCAGA AGATGAGGCTGACTATTAC E1 VL Fr3 TGTGGGAGTGGAGACAGC CDR3 for AGTGGTGCTGCATTCGGC GG TAGGACGGTCAGCTTGGTC E1 VL CDR3 CCTCCGCCGAATGCAGCA Fr4 rev CCACTGCTGTCTCCACTC CCAC GCGCGCGGTCTCTAAGCGTT VL Chicken CCTATGAGCTGATGCAGCCA hum for GGA CCC GGCGGCTCAGGCGGCGGCG VL Chicken GGTCGGGGGGCGGAGGGAG hum for scFv CTCCTATGAGCTGATGCAGC CACCC CAAGTCCTCTTCAGAAATAAG VL Chicken CTTTTGTTCGGATCCTAGGAC hum rev scFv GGTCAGCTTGGTCCCTC VL Chicken GCGCGCGGTCTCTGTCCTAG hum rev GGA GACGGTCAGCTTGGTCCCTC

500 nM

20 nM

20 nM

20 nM

200 nM

20 nM

20 nM

500 nM

500 nM

500 nM 500 nM

Fig. 2 Exemplary sequence alignment of the oligonucleotides encoding the VH sequence and the flanking primers. Oligos are named and numbered according to Table 3. Frameworks are depicted in gray, CDRs in red, blue, and green respectively. Vernier residues, encoding the human amino acid, are shown in violet. Created with SnapGene

Humanization of Chicken-Derived Antibodies 345

Fig. 3 Exemplary sequence alignment of the oligonucleotides encoding the VL sequence and the flanking primers. Oligos are named and numbered according to Table 3. Frameworks are depicted in gray, CDRs in red, blue and green respectively. Vernier residues, encoding the human amino acid, are shown in violet. Created with SnapGene

346 Jan P. Bogen et al.

Humanization of Chicken-Derived Antibodies

347

Table 4 PCR protocol for the fusion of oligonucleotides resulting in VH and VL genes as well as the subsequent assembly of VH and VL genes to the corresponding scFv genes Step

Duration

Cycles



30 s

1



20 s 30 s (1. and 2. PCR) 30 s (3. PCR scFvs only)

30

4. Elongation

98 C Overlap dependent (1. PCR) 72  C (2. PCR) 72  C (3. PCR scFvs only) 72  C

5. Final elongation

72  C

2 min

1. Initial denaturation 2. Denaturation 3. Anneling

Temperature 98 C

30 s (1. PCR) 1

(c) 2.5 μL of a 10 μM solution of flanking primers for either VH or VL and either scFv or Fab (500 nM end concentration, Table 3, italic). (d) 1 μL of a 1 μM solution of each oligonucleotide encoding either VH or VL (20 nM end concentration each, Table 3, plain). (e) 0.5 μL Q5® High-Fidelity DNA polymerase. (f) Add nuclease-free water to an end volume of 50 μL. 8. PCR conditions are described in Table 4. The annealing temperature for this first PCR is dependent on the length of the overlap between the oligonucleotides and the resulting TM values. 9. Based on their overlap and the excess of flanking primer, the PCR results in amplification of full-length VH or VL genes. Due to the random fusion of oligonucleotides, all possible Vernier combinations are generated. Agarose gel electrophoresis will help to verify successful amplification of full-length VH (~380 bp) and VL (~320 bp) genes. 10. In some cases, an additional DNA purification step is needed to ensure sufficient quality of the PCR product. Agarose gel extraction using Promega Wizard® SV Gel and PCR Cleanup System or a comparable kit will ensure the removal of unwanted PCR products. Repeat the first PCR, but instead of VH/VL encoding oligonucleotides, 1 μL of the gel extraction purified PCR product is utilized as template. Furthermore, the annealing temperature is dependent on the TM value of the flanking primers (Table 4). After applying the resulting product to a 1% (w/v) agarose gel, full-length VH and VL genes should be observable. If not, see Note 5. 11. Purify the full-length VH and VL genes via the Promega Wizard® SV Gel and PCR Clean-up System or a comparable

348

Jan P. Bogen et al.

kit and determine the DNA concentration. For the generation of Fab-libraries, continue with Subheading 3.3. 12. For the generation of scFvs, a subsequent PCR fuses the VH and VL genes by overlap extension PCR. Therefore, 10 ng of VH and 10 ng of VL PCR product are utilized together with the flanking primers “VH Chicken hum for scFv” (Oligo number 9) and “VL Chicken hum rev scFv” (oligo number 20) according to Table 3. The resulting scFv construct is ~750 bp in size, which can be verified by a 1% (w/v) agarose gel electrophoresis. 13. Purify the full-length scFv genes via the Promega Wizard® SV Gel and PCR Clean-up System or a comparable kit and determine the DNA concentration. 14. For library establishment, especially for the homologous recombination-based scFv libraries, larger quantities of PCR products might be needed. For five electroporations with 12 μg PCR insert each, 60 μg of scFv encoding amplicons are sufficient. Multiply or scale up the reaction until a sufficient amount of PCR product is generated. Store the PCR products at 20  C until needed. 3.3 Construction of Yeast Surface Display Library 3.3.1 Preparation for the Generation of Humanized scFv Libraries

Commonly, YSD libraries are generated by a process referred to as homologous recombination. In this process, a vector exhibiting the yeast ORI, a Trp auxotrophic marker, and the aga2 gene is linearized by restriction enzymes. Together with the respective insert DNA (scFv genes), the linearized vector is used for yeast cell transformation via homologous recombination. A detailed overview of the generation of scFv libraries for YSD is described here [16]. In brief: 1. Perform a double digestion in a 100 μL volume comprising 70 μg pCT plasmid [11] with 40 U BamHI-HF®, 40 U NheI-HF®, and 10 μL CutSmart buffer. Add nuclease-free water to a final volume of 100 μL and incubate the mixture for 1 h at 37  C. 2. Analyze the digestion via agarose gel electrophoresis to assure complete vector linearization. Purify the linear vector utilizing a PCR cleanup kit according to the supplier’s manual. Determine the DNA concentration and store the DNA at 20  C until needed.

3.3.2 Preparation for the Generation of Humanized Fab Libraries

For a straightforward Fab library generation, Rosowski and coworkers developed a Golden Gate-based system [12]. The destination vector (pDest Lambda) encodes a human lambda CL and a human CH1 fused with the aga2 gene in inverse directions comprising BsaI sites, as well as a yeast ORI and Trp auxothrophic marker (see Note 6). Furthermore, a second vector encoding the

Humanization of Chicken-Derived Antibodies

349

Table 5 Temperature protocol for Golden Gate Assembly of VH and VL genes Step

Temperature 

Duration

Cycles

1. 2.

37 C 16  C

1 min 1 min

30

3.

55  C

5 min

1

1

1

4.



4 C

bidirectional Gal1,10 promotor flanked by BsaI sites is utilized (pEntry). VH and VL genes are flanked by BsaI sites as well. In the course of the Golden Gate reaction, pDest Lambda is linearized, VH and VL genes are inserted into the vector as well as the Gal1,10 promotor. A detailed overview of the generation of Fab libraries for YSD is described here [12, 17]. In brief: 1. Perform eight Golden Gate reactions in parallel, each comprising of 1 μg pDest Lambda, 1.4 μg pEntry, 160 ng VH insert, 160 ng VL insert, 200 U BsaI-HFv2, 800 U T4 DNA ligase, 10 μL 10
T4 Ligase buffer and nuclease-free water in a final volume of 100 μL. 2. Incubate the reactions in a thermocycler following the protocol depicted in Table 5. 3. Purify the Golden Gate product via the Promega Wizard® SV Gel and PCR Clean-up System or a comparable kit by applying four reactions on one cleanup column and elute in 30 μL volume, resulting in two aliquots. Store the DNA at 20  C until needed. 3.3.3 Yeast Transformation

The following protocol is designed for two electroporation reactions, sufficient for establishment of the Fab library. Scale the protocol up for scFv libraries, where five electroporation reactions are sufficient. A detailed version of this protocol is described elsewhere [18] (see Note 7). In brief: 1. Incubate S. cerevisiae EBY100 cells overnight in YPD media at 30  C at 180 rpm. 2. Inoculate 100 mL fresh YPD media using the overnight culture aiming for an OD600 of about 0.3. 3. Incubate cells at 30  C and 180 rpm until OD600 reaches about 1.6. 4. Collect cells by centrifugation at 4000
g for 3 min and discard the supernatant.

350

Jan P. Bogen et al.

5. Wash cells twice (by resuspending) with 50 mL ice-cold water, followed by a washing step with 50 mL ice-cold electroporation buffer. 6. Resuspend the centrifuged cells in 20 mL LiAc buffer and incubate for 30 min at 30  C and 180 rpm. 7. Centrifuge cells and wash once with 50 mL ice-cold electroporation buffer. 8. Resuspend the cells in approximately 200 μL electroporation buffer, resulting in about 1 mL cell suspension. This volume is sufficient to perform two separate electroporation reactions utilizing 400 μL cell suspension each. Keep cells on ice until electroporation. 9. Depending on the generation of a scFv or Fab library, the protocol differs: (a) For a scFv library, mix 4 μg linear pCT (Subheading 3.3.1) and 12 μg scFv-encoding PCR product (Subheading 3.2) with 400 μL EBY100 cells. The volume of the DNA mixture should not exceed 50 μL. (b) For a Fab library, mix the 30 μL of the Golden Gate product (Subheading 3.3.2) with 400 μL EBY100 cells. 10. Transfer the yeast-DNA mixture into a pre-cooled electroporation cuvette (0.2 cm) and incubate for 10 min on ice. Electroporation is performed at 200 Ω, 2.5 kV, and 25 μF. The time constants should range from 3.0 to 4.5 ms. Transfer 400 μL of electroporated cells into about 8 mL of a 1:1 mixture of YPD and 1 M sorbitol and incubate for 1 h at 30  C and 180 rpm. 11. Collect cells by centrifugation and resuspend cells in 10 mL PBS. To enable subsequent verification of the number of transformed yeast cells, perform dilution plating on SD-CAA plates, followed by incubation at 30  C for approx. 72 h (see Note 8). Transfer remaining cells into 1 L SD-CAA media and incubate for 2 days at 30  C and 180 rpm. Typically, one electroporation step yields 5
107 to 1
108 transformed cells, dependent on the DNA quality. 12. Collect 1
1010 cells by centrifugation (OD600 of 1 are approximately 1
107 cells/mL) and resuspend the library in a final volume of 2 mL comprising of 5% (v/v) glycerol and 0.67% (w/v) yeast nitrogen base and store at 80  C until further use (see Note 9). 3.4 Library Staining and Sorting

The following protocol describes the staining and sorting of humanized scFv and Fab-libraries utilizing Fc-tagged EGFR-ECD and FACS. For this specific humanization campaign, two iterative sorting rounds were sufficient, even though additional sorting rounds might be necessary for the isolation of binders comprising parental affinities.

Humanization of Chicken-Derived Antibodies 3.4.1 Cell Staining

351

The described protocol is identical for both sorting rounds and only differ in the antigen concentration. The antigen concentration has to be adapted depending on the initial affinity of the parental chicken antibody. 1. Inoculate 50 mL SG-CAA media with an aliquot of the transformed library to a final OD600 of 1.0 and incubate overnight a 30  C and 180 rpm. 2. Collect 1
107 cells by centrifugation at 8000
g at 4  C for 3 min and wash cells with 1 mL PBS-B (see Note 10). 3. For scFvs: Cells are stained with a 1:20 dilution of an anti-cmyc-biotin antibody in the presence of 200 nM Fc-tagged EGFR-ECD (first round) or 50 nM Fc-tagged EGFR-ECD (second round), respectively. After 30 min incubation at 4  C, cells are washed with 1 mL PBS-B and stained with a 1:50 dilution (10 μg/mL) of Streptavidin-APC and a 1:50 dilution (10 μg/mL) of an anti-human Fc-PE antibody. For a negative control, perform a separate cell staining in the absence of antigen. 4. For Fabs: Cells are stained with a 1:75 dilution (13.3 μg/mL) of an anti-human Lambda-AF647 antibody in the presence of 200 nM Fc-tagged EGFR-ECD (first round) or 50 nM Fc-tagged EGFR-ECD (second round), respectively. After 30 min incubation at 4  C, cells are washed with 1 mL PBS-B and stained with a 1:50 dilution (10 μg/mL) of an anti-human Fc-PE antibody. For a negative control, perform a separate cell staining in absence of antigen. 5. Cells are washed twice with 1 mL ice-cold PBS-B and are resuspended in 1 mL PBSB for the following cell sorting procedure.

3.4.2 Screening by FACS

Here we described the sorting process utilizing a BD Influx cell sorter, but in general, all FACS devices with at least two different lasers might be used. The sorting process is identical for both sorting rounds and only differs in the utilized gates and the number of sorted cells. 1. A suitable sorting gate should include a suitable number of double-positive cells and not more than 0.3% of cells of the respective negative control. An example is shown in Fig. 4. For the first round, the “sort enrich mode” is utilized and the second round is performed using the “sort pure mode.” The number of sorted cells should at least exceed the initial library or the prior sorting round by a factor of 10. 2. Sorted cells are plated on SD-CAA agar plates and incubated for 48 h at 30  C. Subsequently, cells are rinsed and incubated in SD-CAA media overnight at 30  C and either induced in SG-CAA media for a subsequent sorting round or directed to

352

Jan P. Bogen et al.

Fig. 4 Exemplary two-dimensional FACS plots for screening of humanized antibodies by YSD. The upper right quadrants were sorted

the NGS analysis if sufficient enrichment of an antigen-binding population was achieved. 3.5 Next-Generation Sequencing

3.5.1 Sample Preparation

In this section, the preparation of NGS samples and the interpretation of the results of Illumina sequencing are described. The authors utilized barcoded NGS primers in order to combine several sequencing reactions. Implementing barcodes allows for the mixture of different samples, such as of the initial library and the final sorting round without loss of sequence data (see Note 11). 1. Yeast cells derived from the initial library, the first and the final sorting round are cultivated in SD-CAA media overnight at 30  C. Utilizing the Zymoprep Yeast Plasmid Miniprep kit according to the manufactures’ instructions, plasmids of yeast cells are isolated (see Note 12). 2. Resulting plasmids are utilized as templates for the subsequent amplification of VH and VL domains, according to Subheading 3.2 and Table 4 (“2. PCR”). As primers, oligonucleotides depicted in Table 6 are utilized. The PCR mix comprises: (a) 10 μL 5
Q5 Reaction Buffer. (b) 1 μL dNTPs (10 mM each). (c) 2.5 μL of a 10 μM solution of flanking primers for either VH or VL and either scFv or Fab and for the respective sorting round (Table 6).

Table 6 Barcode primers for NGS analysis. Stuffer sequences are highlighted in blue, barcodes are shown in green, encoding for clone (E1), antibody format (scFv or Fab), domain (VH or VL) and sorting round (initial, first round or second round). Depicted primers do not include Illumina adapter sequences Primer name E1 scFv VH Initial rev E1 scFv VH R1 rev E1 scFv VH R2 rev E1 scFv VL Initial rev E1 scFv VL R1 rev E1 scFv VL R2 rev E1 Fab VH Initial rev E1 Fab VH R1 rev E1 Fab VH R2 rev E1 Fab VL Initial rev E1 Fab VL R1 rev E1 Fab VL R2 rev E1 scFv VH Initial for E1 scFv VH R1 for E1 scFv VH R2 for E1 scFv VL Initial for E1 scFv VL R1 for E1 scFv VL R2 for E1 Fab VH Initial for E1 Fab VH R1 for E1 Fab VH R2 for E1 Fab VL Initial for E1 Fab VL R1 for E1 Fab VL R2 for

Primer number 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

Sequence (5'-3') AAATGACGTTCTCTGAAATGAGGAGACGGTGACCAGGG TTCCTTG AAATGACGTTCTTCTAAATGAGGAGACGGTGACCAGGG TTCCTTG AAATGACGTTCTGCGAAATGAGGAGACGGTGACCAGGG TTCCTTG AAATGAATCTCTCTGAAATAGGACGGTCAGCTTGGTCC CTC AAATGAATCTCTTCTAAATAGGACGGTCAGCTTGGTCCC TC AAATGAATCTCTGCGAAATAGGACGGTCAGCTTGGTCC CTC AAAGCTCGTTCTCTGAAATGAGGAGACGGTGACCAGGG TTCCTTG AAAGCTCGTTCTTCTAAATGAGGAGACGGTGACCAGGG TTCCTTG AAAGCTCGTTCTGCGAAATGAGGAGACGGTGACCAGGG TTCCTTG AAAGCTATCTCTCTGAAATAGGACGGTCAGCTTGGTCC CTC AAAGCTATCTCTTCTAAATAGGACGGTCAGCTTGGTCCC TC AAAGCTATCTCTGCGAAATAGGACGGTCAGCTTGGTCC CTC AAATGACGTTCTCTGAAAGAGGTGCAGCTGTTGGAGTC TGGGGG AAATGACGTTCTTCTAAAGAGGTGCAGCTGTTGGAGTC TGGGGG AAATGACGTTCTGCGAAAGAGGTGCAGCTGTTGGAGTC TGGGGG AAATGAATCTCTCTGAAATCCTATGAGCTGATGCAGCC ACCC AAATGAATCTCTTCTAAATCCTATGAGCTGATGCAGCCA CCC AAATGAATCTCTGCGAAATCCTATGAGCTGATGCAGCC ACCC AAAGCTCGTTCTCTGAAAGAGGTGCAGCTGTTGGAGTC TGGGGG AAAGCTCGTTCTTCTAAAGAGGTGCAGCTGTTGGAGTCT GGGGG AAAGCTCGTTCTGCGAAAGAGGTGCAGCTGTTGGAGTC TGGGGG AAAGCTATCTCTCTGAAATCCTATGAGCTGATGCAGCC ACCC AAAGCTATCTCTTCTAAATCCTATGAGCTGATGCAGCCA CCC AAAGCTATCTCTGCGAAATCCTATGAGCTGATGCAGCC ACCC

354

Jan P. Bogen et al.

(d) 1 μL of a 10 ng/μL concentration of plasmid DNA encoding the humanized VH/VLs. (e) 0.5 μL Q5® High-Fidelity DNA polymerase. (f) Add nuclease-free water to an end volume of 50 μL. 3. Depending on the library, the format, and the round, different primers are utilized for this reaction. (a) scFv library VH: initial, primer 22 and 34; round 1, primer 23 and 35; round 2, primer 24 and 36 (Table 6). (b) scFv library VL: initial, primer 25 and 37; round 1, primer 26 and 38; round 2, primer 27 and 39 (Table 6). (c) Fab library VH: initial, primer 28 and 40; round 1, primer 29 and 41; round 2, primer 30 and 42 (Table 6). (d) Fab library VL: initial, primer 31 and 43; round 1, primer 32 and 44; round 2 primer 33 and 45 (Table 6). 4. Amplicons are analyzed on an 1% (w/v) agarose gel. Purify the VH and VL PCR products utilizing a PCR clean-up kit according to the supplier’s manual. Determine the DNA concentration and store the DNA at 20  C until needed. 5. For NGS analysis, a DNA concentration of 20 ng/μL in 50 μL volume (1000 ng total DNA) for each sample is needed (see Note 13). Required DNA concentrations depend on the Illumina sequencing protocol. Pool all VH amplicons and all VL amplicons of one campaign in two separate reaction tubes with equal DNA amount per sorting round. Based on the preference of the researcher, a single Illumina run, consisting of a pool of all amplicons, is feasible as well. However, this will result in fewer reads per domain and sorting round. 6. Send out the tubes to a service provider for Illumina sequencing. 3.5.2 NGS Data Interpretation

In this section, the evaluation of the NGS results is discussed. Usually, fastaq files are provided. The authors used Geneious Primer 2019.0.4 for data interpretation, but other programs with comparable functionality can also be used. 1. Import the fastaq files into Geneious Prime. The program will recognize the NGS data derived from an Illumina run. The sequences will have a maximum length of 250 bp if the “Amplicon-EZ150-500bp” offer from Genewiz is used. Other NGS technologies can yield longer reads. 2. Use the “Match paired reads” function to generate the fulllength sequences. 3. By using the “Separate by barcode” option, VH and VL sequences can be separately analyzed according to the sorting

Humanization of Chicken-Derived Antibodies

355

round they derive from. Use the barcodes depicted in Table 6. Continue with the VH and VL sequences separately for the initial, the first, and the second sorting rounds. 4. Use the “Trim using BBDuk” function to delete the barcodes and stuffer sequences and end up with the plain VH or VL sequences, respectively. 5. Use the “De novo assemble” function to align the sequences. The sequences align perfectly to one another, and differences in the sequences are exclusively seen at the Vernier positions (see Note 14). 6. These files can be used to analyze the number of Vernier combinations. The initial library should contain all possible Vernier combinations to ensure a successful library generation. Furthermore, the frequency of each possible Vernier combination in round 1 and 2 needs to be normalized to their respective frequency found in the initial library. 7. Identify the three VH and VL sequences that show the strongest increment in frequency over the course of the two sorting rounds. Respective genes are ordered at a suitable supplier. 3.6 Reformatting, Expression, Purification, and Characterization

In this section, the authors describe the generation of heavy- and light-chain shuffled humanized chicken-derived full-length antibodies and scFv-Fc fusion proteins as well as their production. The resulting clones are ranked based on their biophysical properties and their homology to the human germline. This subsequent characterization is described in short and can be changed dependent on the intended function of each antibody.

3.6.1 Reformatting of Humanized scFv- and FabBased Antibodies

The selection of VH and VL domains is based on their increment of frequency compared to the initial library. However, the exact light chain–heavy chain pairing is unknown. Therefore, the three most frequent VH and the three most frequent VL sequences are shuffled, resulting in nine different combinations. This can be done in either the scFv- or the Fab-format. 1. The VH and VL genes are amplified by PCR as described in Table 4 utilizing primers depicted in Table 7: (a) scFv VH: primer 46 and 48. (b) scFv VL: primer 50 and 52. (c) Fab VH: primer 46 and 47. (d) Fab VL: primer 49 and 51. 2. For scFv-fragments, a second PCR is needed to fuse the VH and the VL sequence, resulting in a full-length scFv (Tables 4 and 7): primer 46 and 52.

356

Jan P. Bogen et al.

3. Fab-based VH and VL amplicons are inserted into a pTT5derived vector utilizing either an entry vector encoding the CH1-CH2-CH3 sequence of a human IgG1 or a Lambda CL, respectively, as described here [13] (see Note 15). 4. ScFv amplicons are inserted into a pTT5-derived vector utilizing the entry vector encoding the CH2-CH3 sequence of a human IgG1 as described here [9] (see Note 16). 5. Golden Gate is performed utilizing 75 ng destination vector, 75 ng VH or VL insert or 75 ng scFv insert, 2.5 μL T4 ligase buffer, 1000 U T4 Ligase, 20 U SapI, and up to 25 μL H2O. 6. Golden Gate samples are incubated in the thermocycler according to Table 5. 7. 5 μL of the reaction mixture is transformed into chemically competent E. coli XL1 blue cells and incubated on LB ampicillin agar plates overnight. 8. Resulting clones are sequenced utilizing standard Sanger sequencing. Subsequently, clones are inoculated in 50 mL LB ampicillin media, and plasmids are isolated utilizing PureYield Plasmid Midiprep System. 3.6.2 Production and Purification of Humanized Antibodies

1. ExpiHEK293F cells are transfected utilizing lipofectamine and 30 μg vector DNA according to the manufactures’ instructions. The following day, cells are fed using a 20% tryptone solution and are incubated for 5 additional days at 37  C and 8.0% CO2 and 110 rpm. 2. Cells are harvested by centrifugation, and the supernatant is sterile-filtrated. 3. Utilizing standard Protein A chromatography, humanized fulllength antibodies and scFv-Fc fusions are purified and subsequently dialyzed against PBS buffer.

3.6.3 Characterization of Humanized Antibodies

In order to identify the best humanized antibody for its individual purpose, different characterization steps are possible. The most important would be the determination of binding to its antigen and the affinity. Standard technologies like biolayer interferometry (BLI), surface plasmon resonance (SPR), enzyme-linked immunosorbent assay (ELISA), or flow cytometry-based approaches can be utilized for this purpose. Size exclusion chromatography (SEC) and dynamic light scattering (DLS) can be performed to assess the aggregation behavior of resulting mAbs. Nano differential scanning fluorimetry (NanoDSF) is suitable to investigate thermal stability. Furthermore, biological activity assays can be performed to compare the humanized antibodies to their parental counterparts. In

Humanization of Chicken-Derived Antibodies

357

the end, biophysical properties and the number of human Vernier residues should be used to choose the best variant for further development.

4

Notes 1. While some groups identified the same germline gene segments to be the most homologous to their chicken mAbs [19, 20], one report identified the germline IGLV3-19 as the most homologous to their respective VL sequence [7]. Even though we successfully humanized multiple chicken antibodies utilizing the described germlines, in some cases, other acceptor sequences might yield antibodies with preferable properties. 2. It is possible to circumvent the existence of non-human or avian Vernier residues in the final humanized mAb by simply design primers encoding only for the human or the chicken residue for each residue of interest. However, this will elevate the number of oligonucleotides needed since some oligonucleotides exhibit multiple Vernier residues, giving rise to multiple oligonucleotides, encoding all possible human-aviancombinations. Besides the additional molecular biological work needed for cloning to avoid non-human/avian residues, non-human/non-avian residues can positively affect the properties of the antibody and, therefore, contribute to a successful humanization [9]. 3. It is possible that Vernier residues of the parental chicken antibody are divergent from those stated in Table 2. Different pseudogenes utilized in gene conversion or somatic hypermutations can be the reason for this circumstance. It is possible to adjust the utilized degenerated codon to include the original residue of the parental antibody. However, we performed multiple successful humanizations utilizing the codons depicted in Table 2, irrespective of potential mutations in the parental molecule. This underlines that these randomized Vernier residues are sufficient to humanize any given chicken-derived antibody. 4. The length of utilized oligonucleotides, as well as their overlap, can be varied based on the researchers’ preferences. 5. For the rare event that the second PCR does not yield in fulllength VH and VL genes, it is possible to build the genes in multiple, iterative PCR cycles where two oligonucleotides are utilized as templates and two other oligonucleotides as primers.

358

Jan P. Bogen et al.

After agarose gel electrophoresis and PCR clean-up, the product is utilized as template for the subsequent reaction. In each following PCR cycle, new oligonucleotides are added to the reaction until a full-length gene is generated. However, this second PCR usually yields in full-length product, so this alternative protocol is only needed in a very limited number on cases. 6. Since chicken-derived antibodies have a high homology to the VL domains of the human lambda subtype, respective CL domains are chosen for YSD and subsequent expression. 7. Keep cells on ice whenever possible to assure a high transformation efficiency. 8. The calculated number of transformed cells should exceed the theoretical library size by at least tenfold. 9. An OD600 of 1.0 corresponds to approximately 1.0
107 yeast cells/mL. 10. The number of collected yeast cells should exceed the library size by at least the factor of 10. If one utilizes other Vernier residues or codons, the number of stained cells might need to be adjusted. 11. Since the expected library size is easy to handle, barcoded primers can be utilized to minimize the numbers of individual NGS runs. However, if the humanization strategy is changed, for example, by including additional Vernier residues or including other degenerated codons, separated, non-barcoded NGS runs might be preferred. 12. This step can be substituted by any other commercial yeast plasmid isolation kit or alternative protocols [21]. 13. This amount is recommended by Genewiz and might differ for other NGS service providers. 14. Other unmatching positions are probably derived from errors accruing in the polymerase chain reactions or in the course of the next-generation sequencing. They should only be found in single variants. 15. Other mammalian expression vectors can be used as well and only need an adjustment of primer overhangs (Table 7). 16. Other expression vectors, for example, for E. coli production, can be chosen as well.

Humanization of Chicken-Derived Antibodies

359

Table 7 Primer sequences for reformatting of full-length antibodies and scFv-Fc fusions. Overhangs encoding SapI sites are shown in red, overhangs for the fusion into an scFv format are shown in blue Primer name HumRe VH Fr1 for HumRe VH Fr4 Fab rev HumRe VH Fr4 scFv rev HumRe VL Fr1 Fab for HumRe VL Fr1 scFv for HumRe VL Fr4 Fab Lam rev HumRe VL Fr4 scFv rev

Primer Sequence (5'-3') number 46 AAAAAGCTCTTCAAGTGAGGTGCAGCTGTTGGAGTCTGGGGG 47

TTTTTTGCTCTTCTGGCTGAGGAGACGGTGACCAGGGTTCC

48

TCCGCCCCCCGAACCGCCGCCGCCTGAGCCGCCTCCCC CTGAGGAGACGGTGACCAGGGTTCC AAAAAGCTCTTCAAGTTCCTATGAGCTGATGCAGCCACCC

49 50 51

GGCGGCTCAGGCGGCGGCGGTTCGGGGGGCGGAGGGA GCTCCTATGAGCTGATGCAGCCACCC TTTTTTGCTCTTCACCCTAGGACGGTCAGCTTGGTCCCTC

52

TTTTTTGCTCTTCTTTCTAGGACGGTCAGCTTGGTCCCTC

References 1. Kung P, Goldstein G, Reinherz EL, Schlossman SF (1979) Monoclonal antibodies defining distinctive human T cell surface antigens. Science (New York, NY) 206(4416):347–349. https://doi.org/10.1126/science.314668 2. Smith SL (1996) Ten years of Orthoclone OKT3 (muromonab-CD3): a review. J Transpl Coord 6(3):109–119.; quiz 120–101. https:// doi.org/10.7182/pr tr.1.6.3. 8145l3u185493182 3. Morrison SL, Johnson MJ, Herzenberg LA, Oi VT (1984) Chimeric human antibody molecules: mouse antigen-binding domains with human constant region domains. Proc Natl Acad Sci 81(21):6851. https://doi.org/10. 1073/pnas.81.21.6851 4. Ulitzka M, Carrara S, Grzeschik J, Kornmann H, Hock B, Kolmar H (2020) Engineering therapeutic antibodies for patient safety: tackling the immunogenicity problem. Protein Eng Des Sel 33:gzaa025. https://doi. org/10.1093/protein/gzaa025 5. Jones PT, Dear PH, Foote J, Neuberger MS, Winter G (1986) Replacing the complementarity-determining regions in a human antibody with those from a mouse. Nature 321(6069):522–525. https://doi. org/10.1038/321522a0 6. Abdiche YN, Harriman R, Deng X, Yeung YA, Miles A, Morishige W, Boustany L, Zhu L,

Izquierdo SM, Harriman W (2016) Assessing kinetic and epitopic diversity across orthogonal monoclonal antibody generation platforms. MAbs 8(2):264–277. https://doi.org/10. 1080/19420862.2015.1118596 7. Gjetting T, Gad M, Fro¨hlich C, Lindsted T, Melander MC, Bhatia VK, Grandal MM, Dietrich N, Uhlenbrock F, Galler GR, Strandh M, Lantto J, Bouquin T, Horak ID, Kragh M, Pedersen MW, Koefoed K (2019) Sym021, a promising anti-PD1 clinical candidate antibody derived from a new chicken antibody discovery platform. MAbs 11(4): 6 6 6 – 6 8 0 . h t t p s : // d o i . o r g / 1 0 . 1 0 8 0 / 19420862.2019.1596514 8. Tsurushita N, Park M, Pakabunto K, Ong K, Avdalovic A, Fu H, Jia A, Va´squez M, Kumar S (2004) Humanization of a chicken anti-IL-12 monoclonal antibody. J Immunol Methods 295(1-2):9–19. https://doi.org/10.1016/j. jim.2004.08.018 9. Elter A, Bogen JP, Hinz SC, Fiebig D, Macarro´n Palacios A, Grzeschik J, Hock B, Kolmar H (2020) Humanization of chicken-derived scFv using yeast surface display and NGS data mining. Biotechnol J 16(3):2000231. https://doi.org/10.1002/biot.202000231 10. Foote J, Winter G (1992) Antibody framework residues affecting the conformation of the hypervariable loops. J Mol Biol 224(2):

360

Jan P. Bogen et al.

487–499. https://doi.org/10.1016/00222836(92)91010-m 11. Boder ET, Wittrup KD (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol 15(6):553–557. https://doi.org/10.1038/nbt0697-553 12. Rosowski S, Becker S, Toleikis L, Valldorf B, Grzeschik J, Demir D, Willenbu¨cher I, Gaa R, Kolmar H, Zielonka S, Krah S (2018) A novel one-step approach for the construction of yeast surface display Fab antibody libraries. Microb Cell Factories 17(1):3. https://doi.org/10. 1186/s12934-017-0853-z 13. Bogen JP, Storka J, Yanakieva D, Fiebig D, Grzeschik J, Hock B, Kolmar H (2020) Isolation of common light chain antibodies from immunized chickens using yeast biopanning and fluorescence-activated cell sorting. Biotechnol J 16:e2000240. https://doi.org/10. 1002/biot.202000240 14. Grzeschik J, Yanakieva D, Roth L, Krah S, Hinz SC, Elter A, Zollmann T, Schwall G, Zielonka S, Kolmar H (2019) Yeast surface display in combination with fluorescenceactivated cell sorting enables the rapid isolation of antibody fragments derived from immunized chickens. Biotechnol J 14(4):e1800466. https://doi.org/10.1002/biot.201800466 15. Wu L, Oficjalska K, Lambert M, Fennell BJ, Darmanin-Sheehan A, Shu´illeabha´in DN, Autin B, Cummins E, Tchistiakova L, Bloom L, Paulsen J, Gill D, Cunningham O, Finlay WJ (2012) Fundamental characteristics of the immunoglobulin VH repertoire of chickens in comparison with those of humans, mice, and camelids. J Immunol 188(1): 3 2 2 – 3 3 3 . h t t p s : // d o i . o r g / 1 0 . 4 0 4 9 / jimmunol.1102466

16. Bogen JP, Grzeschik J, Krah S, Zielonka S, Kolmar H (2020) Rapid generation of chicken immune libraries for yeast surface display. Methods Mol Biol 2070:289–302. https:// doi.org/10.1007/978-1-4939-9853-1_16 17. Krah S, Grzeschik J, Rosowski S, Gaa R, Willenbuecher I, Demir D, Toleikis L, Kolmar H, Becker S, Zielonka S (2018) A streamlined approach for the construction of large yeast surface display fab antibody libraries. Methods Mol Biol 1827:145–161. https://doi.org/10.1007/978-1-49398648-4_8 18. Benatuil L, Perez JM, Belk J, Hsieh CM (2010) An improved yeast transformation method for the generation of very large human antibody libraries. Protein Eng Des Sel 23(4):155–159. https://doi.org/10.1093/ protein/gzq002 19. Nishibori N, Horiuchi H, Furusawa S, Matsuda H (2006) Humanization of chicken monoclonal antibody using phage-display system. Mol Immunol 43(6):634–642. https:// doi.org/10.1016/j.molimm.2005.04.002 20. Baek DS, Kim YS (2015) Humanization of a phosphothreonine peptide-specific chicken antibody by combinatorial library optimization of the phosphoepitope-binding motif. Biochem Biophys Res Commun 463(3): 414–420. https://doi.org/10.1016/j.bbrc. 2015.05.086 21. Holm C, Meeks-Wagner DW, Fangman WL, Botstein D (1986) A rapid, efficient method for isolating DNA from yeast. Gene 42(2): 169–173. https://doi.org/10.1016/03781119(86)90293-3

Chapter 19 Engineering Tissue Inhibitors of Metalloproteinases Using Yeast Surface Display Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh Abstract Yeast surface display (YSD) has been extensively used for protein design, engineering, and directed evolution in the past two decades. Here, we describe methods for directed evolution of tissue inhibitors of metalloproteinase (TIMP), the natural inhibitors of matrix metalloproteinases (MMPs), through design and generation of a combinatorial library of TIMP mutants and screening the targeted TIMP library of variants toward MMP binding using YSD. This protocol can be adopted to other natural enzyme inhibitors and similar protein binders such as antibodies. Key words Yeast surface display, Tissue inhibitor of metalloproteinase, Matrix metalloproteinase, Fluorescent-activated cell sorting, Directed evolution, Rational design of proteins, Enzyme inhibitors

1

Introduction Choice of cell display platforms is critical in successful screening of protein mutant libraries and engineering proteins. Among all cell display platforms, yeast surface display (YSD) is proven to be one of the most applicable cell display technologies for engineering proteins to date, as it offers simplicity similar to phage and bacteria display systems, yet higher transformation efficiency, and eukaryotic posttranslational modifications similar to mammalian display platforms [1–4]. Antibodies were first to be engineered using YSD [2, 5, 6]; however, YSD has been soon expanded for use in engineering other non-antibody scaffolds, including natural inhibitors of enzymes [3, 4]. Among non-antibody scaffolds, enzyme inhibitors have specific importance due to their therapeutic potential. Matrix metalloproteinases (MMPs) are a family of zincdependent endopeptidases that play a critical role in the remodeling of the extracellular matrix (ECM) [7]. Overexpression of MMPs is related to several diseases such as cancer, neurological disorders, and cardiovascular diseases [8, 9]. Tissue inhibitor of

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_19, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

361

362

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

Fig. 1 Yeast surface display and immunolabeling platform for quantifying TIMP and MMP interactions. The schematic demonstrates how TIMP-1 variants are displayed on the yeast surface by N-terminal fusion to yeast cell wall protein Aga2. TIMP-1 expression is detected through immunolabeling of the c-myc epitope tag (red and green antibodies), and binding of biotinylated MMP-3 catalytic domain is detected via fluorescent conjugated streptavidin (orange star). (Illustration created with BioRender.com)

metalloproteinases (TIMP) are endogenous inhibitors of MMPs and are great candidates for developing protein therapeutics based on MMP inhibitors. YSD was previously used to engineer fulllength human TIMP-1 and the N-terminal domain of TIMP2 for binding affinity and selectivity toward specific MMPs [10– 13] (Fig. 1). Various methods have emerged and have been finetuned to generate, screen, and analyze various YSD libraries [5, 6, 14]. Here, we describe designing, generating, and screening of a random library of TIMP variants, largely focused on residues located in the five interacting loops between TIMPs and MMPs. This protocol also outlines a high-throughput screening method of the targeted TIMP library, including isolation and evaluation of TIMP variants with improved binding to a specific MMP, with TIMP-1 and MMP-3 catalytic domain (MMP-3cd) as an example pair. This method section can be easily adopted for other TIMPs, natural enzyme inhibitors, or other protein binders.

2

Materials

2.1 Expression of MMP-3 Catalytic Domain

1. pET3a-Hisx6-pro-MMP-3cd vector in Rosetta2(DE3)pLysS competent Escherichia coli cells. 2. MilliQ H2O.

Engineering TIMPs using YSD

363

3. Ampicillin (Amp) at 1000
stock: Dissolve 100 mg/mL of amp in MilliQ H2O and filter sterilize. Store stock at 20  C. 4. Chloramphenicol (Cam) at 1000
stock: Dissolve 34 mg/mL of cam in 100% ethanol and filter sterilize. Store stock at 20  C. 5. 2-L glass bottle. 6. Luria-Bertani (LB) broth + amp + cam: Add 25 g of dry LB media (10.0 g bacto tryptone, 5.0 g bacto yeast extract, 10.0 g NaCl) to a 2 L glass bottle, and dissolve in 1 L MilliQ H2O. Autoclave the broth, and add 100 μg/mL of amp and 34 μg/mL of cam once cooled to about 45  C. Store broth at 4  C. 7. 37  C shaking incubator at 250 rpm. 8. Sterile 1-L flask. 9. Spectrophotometer able to measure optical density at 600 nm. 10. Disposable cuvette. 11. 0.22-μm syringe filter. 12. 5-mL syringe. 13. 1 M isopropyl-β-D-thiogalactopyranoside (IPTG): Dissolve 2.38 g of IPTG in a final volume of 10 mL MilliQ H2O. Filter sterilize with a 0.22-μm syringe filter. Divide into 1 mL aliquots and store at 20  C. 14. Sterile 250-mL conical bottles. 15. Refrigerated centrifuge to spin 250-mL conical bottles. 16. Swing bucket rotor to hold 250-mL conical bottles. 2.2 Inclusion Body Extraction and Solubilization of MMP3 Catalytic Domain

1. Expressed, pelleted MMP-3 catalytic domain. 2. MilliQ H2O. 3. Sterile 250-mL conical bottles. 4. 1-L glass bottles. 5. Stir plate and stir bars. 6. 10 M Urea: In a 1 L flask, dissolve 300.3 g of urea in 600 mL of MilliQ H2O, stirring vigorously. Once solution has cleared, slowly add remaining 300.3 g of urea. Add MilliQ H2O to a final volume of 1 L. Do not heat or autoclave. Store at room temperature. Make the solution a day before use. 7. 1 M Tris-HCl, pH 8.0: Dissolve 121.1 g of Tris base in a final volume of 1 L MilliQ H2O. Use HCl to adjust to a final pH of 8.0. Store at 4  C. 8. 0.5 M Ethylenediaminetetraacetic acid (EDTA), pH 8.0: Add 186.1 g of EDTA to 800 mL MilliQ H2O, stirring vigorously. Adjust the solution to pH 8.0 (see Note 1). Bring solution to a final volume of 1 L using MilliQ H2O. Store at 4  C.

364

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

9. 1 M NaCl: Dissolve 58.44 g of NaCl in a final volume of 1 L MilliQ H2O. 10. 1 M Dithiothreitol (DTT): Dissolve 3.0 g of DTT to a final volume of 20 mL MilliQ H2O. Divide into 1 mL aliquots and store at 20  C. 11. Lysozyme. 12. Triton X-100. 13. Lysis Buffer: Contains a final concentration of 50 mM TrisHCl, pH 8.0, 1 mM EDTA, 100 mM NaCl, 0.133 g/mL lysozyme, and 49% v/v Triton X-100. Bring to a final volume of 1 L with MilliQ H2O. Adjust to a final pH of 8.0. Store at 4  C. 14. Inclusion body buffer: Contains a final concentration of 20 mM Tris-HCl, pH 8.0, 1 mM EDTA, 100 mM NaCl, 5 mM DTT, 2% v/v Triton X-100, and 0.5 M Urea. Bring to a final volume of 1 L with MilliQ H2O. Adjust to a final pH of 8.0. Store at 4  C. 15. Solubilization buffer: Contains a final concentration of 20 mM Tris-HCl, pH 8.0, 50 mM NaCl, 10 mM DTT, and 6 M Urea. Bring to a final volume of 100 mL with MilliQ H2O. Adjust to a final pH of 8.0. Store at 4  C. 16. Tabletop vortex. 17. Orbital shaker stationed at 4  C, or refrigerated shaking incubator at 150 rpm. 18. 10% (w/v) sodium deoxycholate. 19. DNase I. 20. Refrigerated centrifuge to spin 50-mL conical tubes. 21. Sterile 50-mL conical tubes. 22. Sonicator. 2.3 Purification of MMP-3 Catalytic Domain

1. Solubilized MMP-3 catalytic domain. 2. MilliQ H2O. 3. Gravity-flow column. 4. Ni-NTA resin. 5. 10 M Urea (see Subheading 2.2). 6. 1 M Tris-HCl, pH 8.0 (see Subheading 2.2). 7. 1 M NaCl (see Subheading 2.2). 8. 5 M Imidazole: Dissolve 5.1 g of imidazole in 10 mL of MilliQ H2O. 9. HT equilibration buffer: Contains a final concentration of 20 mM Tris-HCl, pH 8.0, 50 mM NaCl, and 6 M urea. Bring to a final volume of 100 mL with MilliQ H2O. Adjust to a final pH of 7.4. Store at 4  C.

Engineering TIMPs using YSD

365

10. Sterile 50-mL conical tube. 11. HT wash buffer: Contains a final concentration of 20 mM TrisHCl (pH 8.0), 50 mM NaCl, 6 M urea, and 25 mM imidazole. Bring to a final volume of 100 mL with MilliQ H2O. Adjust to a final pH of 7.4. Store at 4  C. 12. HT elution buffer: Contains a final concentration of 20 mM Tris-HCl (pH 8.0), 50 mM NaCl, 6 M urea, 250 mM imidazole. Bring to a final volume of 100 mL with MilliQ H2O. Adjust to a final pH of 7.4. Store at 4  C. 13. HT regeneration buffer: Contains a final concentration of 20 mM 2-(N-Morpholino) ethanesulfonic acid (MES), and 0.1 M NaCl. Bring to a final volume of 100 mL with MilliQ H2O. Adjust to a final pH of 5.0. Store at 4  C. 14. Sterile 15-mL conical tubes. 15. Spectrophotometer able to measure optical density at 280 nm. 16. Sterile 1.5-mL microfuge tube. 17. 20% v/v ethanol. 2.4 Refolding of MMP-3 Catalytic Domain

1. Purified MMP-3 catalytic domain. 2. Slide-a-Lyzer dialysis cassette with 10K molecular weight cutoff, Slide-a-Lyzer dialysis flask, or Snakeskin dialysis tubing (Thermo Scientific). 3. Snakeskin dialysis clips. 4. Buoy to hold dialysis cassette or Snakeskin tubing. 5. MilliQ H2O. 6. 10 M Urea (see Subheading 2.2). 7. 1 M Tris-HCl, pH 8.0 (see Subheading 2.2). 8. 1 M NaCl (see Subheading 2.2). 9. 1 M CaCl2: Dissolve 11.1 g of CaCl2 in a final volume of 100 mL MilliQ H2O. 10. 1 M ZnCl2. Dissolve 2.0 g of ZnCl2 in a final volume of 15 mL MilliQ H2O. 11. Dialysis Buffer 1: Contains a final concentration of 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10 mM CaCl2, 1 μM ZnCl2, and 4 M Urea. Bring to a final volume of 1 L with MilliQ H2O. Adjust to a final pH of 7.4. Store at 4  C. 12. Dialysis Buffer 2: Contains a final concentration of 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10 mM CaCl2, 1 μM ZnCl2, and 2 M Urea. Bring to a final volume of 1 L with MilliQ H2O. Adjust to a final pH of 7.4. Store at 4  C. 13. Dialysis Buffer 3: Contains a final concentration of 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10 mM CaCl2, and 1 μM

366

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

ZnCl2. Bring to a final volume of 1 L with MilliQ H2O. Adjust to a final pH of 7.4. Store at 4  C. 14. 5 mL syringe. 15. 18-guage, 1-inch beveled needle. 16. Stir plate stationed at 4  C and stir bar. 17. Sterile 1.5-mL microcentrifuge tubes. 2.5 Re-concentration of MMP-3 Catalytic Domain

1. Refolded MMP-3 catalytic domain. 2. Sterile 1.5-mL microcentrifuge tubes. 3. Refrigerated microcentrifuge. 4. 50 mL Amicon ultra centrifuge filter unit or 400 mL Amicon stirred cell with 10k MWCO membrane (EMD Millipore). 5. Refrigerated centrifuge to spin 50-mL conical tubes. 6. Swing bucket rotor to hold 50-mL conical tubes. 7. Disposable cuvette. 8. Spectrophotometer able to measure optical density at 280 nm.

2.6 APMA Activation of MMP-3 Catalytic Domain

1. Reconcentrated MMP-3 catalytic domain. 2. 4-aminophenylmercuric acetate (APMA). 3. 37  C shaking incubator at 250 rpm. 4. Refrigerated microcentrifuge. 5. Sterile 1.5-mL microcentrifuge tubes.

2.7 Desalting of MMP-3 Catalytic Domain

1. Activated MMP-3 catalytic domain. 2. Zeba spin desalting column (Thermo Scientific). 3. Sterile 15-mL conical tube. 4. Refrigerated centrifuge able to spin 15-mL conical tubes. 5. Sterile MilliQ water. 6. Sterile 1.5-mL microcentrifuge tubes.

2.8 Biotinylation of MMP-3 Catalytic Domain

1. EZ-Link NHS-PEG4 biotinylation kit (Thermo Scientific). 2. Zeba spin desalting columns (Thermo Scientific). 3. 40 -Hydroxyazobenzene-2-carboxylic (Thermo Scientific).

acid

(HABA)

assay

4. Phosphate-buffered saline (PBS), pH 7.4: Mix 100 mM sodium phosphate and 150 mM sodium chloride to a final volume of 20 mL. Adjust to a final pH of 7.4. 5. HABA/Avidin solution: Dissolve 10 mg of avidin into a solution containing 600 μL 10 mM HABA and 19.4 mL PBS. If a precipitate forms, the solution can be filtered. Solution can be stored up to 2 weeks at 4  C.

Engineering TIMPs using YSD

367

6. Sterile 15-mL conical tube. 7. Sterile 1.5-mL microcentrifuge tubes. 8. Disposable cuvettes. 9. Spectrophotometer able to measure optical density at 500 nm. 10. Microplate. 11. Microplate reader. 12. Orbital shaker. 2.9 Generation of TIMP-1 Library and Cell Growth

1. Sterilized MilliQ water. 2. Pellet Paint® Co-Precipitant (EMD Millipore). 3. A library of human TIMP-1 gene variants constructed via gene blocks. 4. pCHA-VRC01 DNA plasmid (Addgene). 5. pCHA-TIMP-1 DNA plasmid. 6. Restriction enzymes: NheI, BsrGI and BamHI (New England Biolabs). 7. Q5® High-Fidelity 2
Master Mix (New England BioLabs). 8. Forward and reverse primers with homologous 15-bp overhangs to the pCHA-VRC01 yeast display vector up- and down-stream of NheI and BamHI RE sites, respectively. The forward primer must contain a BsrGI RE site, replacing that of NheI. 9. Forward and reverse primers with homologous 50-bp overhangs to the pCHA-TIMP-1 yeast display vector up- and down-stream of BsrGI and BamHI RE sites, respectively. 10. 10
CutSmart reaction buffer (New England BioLabs). 11. LB broth + ampicillin (see Subheading 2.1). Do not add chloramphenicol. 12. LB + ampicillin agar plate: Follow the protocol for LB broth + ampicillin, and add 18 g of agar to the solution before autoclaving. Pour the agar plates under a flame, to maintain sterility. Store agar plates at 4  C. 13. The yeast Saccharomyces cerevisiae strain EBY100 (MATa AGA1::GAL1-AGA1::URA3 ura3–52 trp1 leu2-Δ200 his3Δ200 pep4::HIS3 prb11.6R can1 GAL). 14. YPD agar plates: Dissolve 10 g yeast extract, 20 g peptone, 20 g dextrose, and 15 g agar in 1 L MilliQ H2O. Autoclave media. Once media has cooled enough to touch (approximately 45  C), pour the agar into petri dishes while under flame, to maintain sterility. Allow for agar to cool and solidify, and store the YPD agar plates at 4  C.

368

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

15. YPD media: Dissolve 10 g yeast extract, 20 g peptone, and 20 g dextrose in 1 L MilliQ H2O. Autoclave media. Media may appear darker after autoclaving due to dextrose caramelizing—this should not affect the media. Store media at 4  C. 16. 1-L glass bottles. 17. 30  C shaking incubator at 250 rpm. 18. Sterile 150-mL flask. 19. Disposable cuvettes. 20. Spectrophotometer able to measure optical density at 600 nm. 21. Electroporation cuvettes. 22. 1 M Sorbitol: Dissolve 18.2 g of sorbitol in a final volume of 100 mL MilliQ H2O. 23. Refrigerated centrifuge to spin 50-mL conical tubes. 24. Gene pulser. 25. Sterile 50-mL conical tubes. 26. Ampicillin at 1000
stock (see Subheading 2.1). 27. SD-CAA media, pH 4.5: Dissolve 20 g dextrose, 6.7 g yeast nitrogen base, 5.0 g acid casein peptone, and citrate buffer salts (14.7 g sodium citrate, 4.3 g citric acid monohydrate) in 1 L MilliQ H2O. Adjust to a final pH of 4.5. Filter sterilize the solution. Store in sterilized container at 4  C. 28. SD-CAA agar plate, pH 6.0: Dissolve 20 g dextrose, 6.7 g yeast nitrogen base, and 5.0 g acid casein peptone in 100 mL of MilliQ H2O. Mix well and filter sterilize media. Dissolve phosphate buffer salts (10.2 g sodium phosphate dibasic heptahydrate, 8.6 g sodium phosphate monobasic monohydrate), 15.0 g agar, and 182 g sorbitol in 900 mL MilliQ H2O. Autoclave salt, agar, and sorbitol solution. Once autoclaved solution has cooled to approximately 45  C, add filtered solution and mix vigorously. Pour agar plates under a flame, to keep sterile. Store the agar plates at 4  C. 2.10 Yeast Surface Display and Cell Growth

1. MilliQ H2O. 2. Penicillin/Streptomycin (pen/strep) at 100
stock: Make 10,000 IU/mL and 10,000 μg/mL, respectively, in 100 mL MilliQ H2O. Filter sterilize solutions. Divide into 1 mL aliquots and store at 20  C. 3. SD-CAA media, pH 4.5 + pen/strep (see Subheading 2.9). Final concentration of pen/strep in SD-CAA media will be 1%. For instance, in 500 mL of SD-CAA media, add 5 mL of pen/step from 100
stock. 4. SG-CAA media, pH 6.0: Dissolve 20 g galactose, 6.7 g yeast nitrogen base, 5.0 g acid casein peptone, and phosphate buffer

Engineering TIMPs using YSD

369

salts (5.4 g sodium phosphate dibasic, anhydrous, and 8.56 g sodium phosphate monobasic monohydrate) in 1 L MilliQ H2O. Adjust to a final pH of 6.0. Filter sterilize the solution and store in sterilized container at 4  C. 5. Sterile 50-mL serological pipette. 6. Sterile 5-mL serological pipette. 7. Pipette controller. 8. Sterile 150-mL flasks. 9. Sterile glass 15-mL culture tubes. 10. 30  C shaking incubator at 250 rpm. 11. Sterile 1.5-mL microfuge tubes. 12. Tabletop microcentrifuge. 13. Disposable cuvettes. 14. Spectrophotometer able to measure optical density at 600 nm. 15. Bunsen burner and starter. 16. 70% ethanol. 17. Sterile 40% glycerol. 18. Cryovials. 2.11 Cell Preparation for Flow Cytometry

1. MilliQ H2O. 2. 10
PBS, pH 7.4: Dissolve 80 g NaCl, 2 g KCl, 27.2 g Na2HPO4  7H2O (dibasic heptahydrate), 2.4 g KH2PO4 (monobasic anhydrous) in 1 L MilliQ H2O. Adjust to a final pH of 7.4. Filter sterilize and store at room temperature. 3. 1
PBS + 0.1% bovine serum albumin (BSA), pH 7.4 (PBSA): Add 50 mL of 10
PBS to graduated cylinder. Bring to a volume of 500 mL with MilliQ H2O. Adjust to a final pH of 7.4. Add 0.5 g of BSA to solution and mix well. Store solution at 4  C. 4. Sterile 1.5-mL microfuge tubes. 5. Tabletop microcentrifuge. 6. Disposable cuvettes. 7. 70% ethanol. 8. Pasteur pipette. 9. Vacuum system used to aspirate supernatant. 10. Spectrophotometer able to measure optical density at 600 nm.

2.12

Flow Cytometry

1. Primary labels: Mouse anti-c-myc (9e10) (Sigma), and biotinylated MMP-3 catalytic domain.

370

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

2. Secondary labels: Anti-mouse Alexa Fluor-488 (Thermo Scientific), and Streptavidin Alexa Fluor-647 conjugate (4E3D10H2/E3) (Thermo Scientific). 3. Tabletop vortex. 4. Flow cytometer with sorting capability. 5. Ampicillin at 1000
stock (see Subheading 2.1). 6. SD-CAA media, pH 4.5 (see Subheading 2.9). 7. Sterile 1.5-mL microcentrifuge tubes. 8. Sterile 1.5-mL amber microcentrifuge tubes. 9. Sterile culture tube to collect sorted cells. 2.13 DNA Preparation and Evaluation of Individual Clones

1. Zymoprep Yeast Plasmid Miniprep II kit (Zymo Research). 2. YPD media (see Subheading 2.9). 3. SD-CAA plate, pH 6.0 (see Subheading 2.9). 4. Ampicillin at 1000
stock (see Subheading 2.1). 5. LB broth + ampicillin (see Subheading 2.1). Do not add chloramphenicol. 6. LB + ampicillin agar plate (see Subheading 2.9). 7. Chemically competent Escherichia coli for transformation. 8. Super Optimal broth with Catabolite repression (SOC): Often provided with chemically competent cells. To make SOC, add 2.0 g of 2% bactotryptone, 0.5 g of 0.5% yeast extract, 0.2 mL of 5 M NaCl, 1.0 mL of 1 M MgCl2, 1.0 mL of 1 M MgSO4, and 0.36 g of dextrose to a 200-mL bottle. Bring to a final volume of 100 mL, and autoclave or filter sterilize. 9. 1-L glass bottle. 10. Sterile toothpicks. 11. Sterile glass culture tubes. 12. 30  C shaking incubator at 250 rpm. 13. 37  C shaking incubator at 250 rpm. 14. Sterile 1.5-mL microfuge tubes. 15. Tabletop microcentrifuge. 16. 80% ethanol. 17. Water-bath set to 37  C. 18. Sorted TIMP-1 variant library. 19. Qiagen or Promega Miniprep kit. 20. SnapGene, or other DNA analysis software.

Engineering TIMPs using YSD

3

371

Methods

3.1 Expression, Purification, Solubilization, and Biotinylation of MMP-3 Catalytic Domain (MMP-3cd) 3.1.1 Expression of MMP-3cd Protein

1. Select single colonies of pET3a-Hisx6-pro-MMP-3cd transformants in Rosetta2(DE3)pLysS competent Escherichia coli to incubate and shake overnight (~16 h) in 5 mL of LB amp cam® media at 37  C and 250 rpm. 2. Per overnight culture, inoculate one 1-L flask containing 500 mL of LB amp cam® media. This should return the cells to logarithmic growth. 3. When the optical density at 600 nm (OD600) reaches between 0.4 and 0.5, which will take roughly 3–4 h, induce the cultures with 1 M isopropyl-β-D-thiogalactopyranoside (IPTG) to a final concentration of 1 mM. 4. Continue to incubate for approximately 3 more hours (see Note 2). 5. Centrifuge the induced culture in 250-mL conical bottles in a swing bucket rotor for 5 min at 10,000
g and 4  C. Discard the supernatant.

3.1.2 Extraction of Insoluble MMP-cd from Inclusion Bodies

1. Prepare 10 M fresh urea (see Note 3). 2. Measure the weight of MMP3-cd pellet. Resuspend MMP-3cd pellet in 3 mL per gram of pellet using Lysis Buffer, via vortexing or pipetting. 3. Shake overnight at 4  C, at a minimum of 150 rpm. 4. Add 1.25 mL of 10% (w/v) sodium deoxycholate per liter of culture grown. 5. Shake at room temperature for 30 min at a minimum of 150 rpm. 6. Add 10 μL of DNase I per liter of culture grown. 7. Shake at room temperature for 30 min at a minimum of 150 rpm. 8. Centrifuge for 15 min at 10,000
g and 4  C. Discard supernatant. 9. Resuspend pellet in 100 mL of Inclusion Body Buffer per liter of culture grown, through sonication. Transfer the samples into 50-mL conical tubes. 10. Sonicate each sample submerged in ice/water bath in a beaker for six cycles of 15 s, output 5, and 50% pulse. Allow 15 s rest periods for cooling between cycles. 11. Centrifuge for 30 min at 16,000
g and 4  C (see Note 4). 12. Resuspend each pellet in 5 mL Solubilization Buffer per liter of culture grown, by pipetting.

372

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

13. Incubate for at least 30 min on ice to allow proteins to solubilize. 14. Centrifuge for 30 min at 16,000
g and 4  C. Keep the supernatant. 3.1.3 Purify and Refold MMP-3cd

1. Pack a gravity-flow protein purification column with 2 mL of well-mixed Ni-NTA resin. 2. Equilibrate the column with five resin-bed volumes of HT Equilibration Buffer. 3. Add the soluble protein fraction to the column, which should be a clear supernatant. Collect the “HT Flowthrough.” 4. Wash the resin with 15 mL of HT Wash Buffer. Collect the flow-through labeled “HT Wash” (see Note 5). 5. Elute the MMP-3cd-Hisx6-tagged protein from the Ni-NTA column by adding 5 mL of HT Elution Buffer. Collect the eluted protein fractions in 0.5–1 ml collection tubes “HT Elution.” 6. Dilute the HT Elution fraction with HT Elution Buffer to a final concentration ranging from 0.3 to 0.6 mg/mL. 7. Following the manufacturer’s protocol, add the diluted protein solution to a Slide-a-Lyzer dialysis cassette with 10K molecular weight cutoff, Slide-a-Lyzer dialysis flask, or Snakeskin dialysis tubing. 8. Dialyze MMP-3cd against Dialysis Buffer 1. Stir on a magnetic stirrer at 4  C, for at least 8 h. 9. Repeat dialysis against Dialysis Buffer 2 and Dialysis Buffer 3. 10. Transfer the sample into new sterile 1.5-mL microcentrifuge tubes and label as “Dialyzed MMP”. If precipitate has formed, centrifuge the sample for 15 min at 17,000
g at 4  C (see Note 6). 11. Transfer the supernatant into new sterile 1.5-mL microcentrifuge tubes and label as “Refolded MMP.” Samples can be stored at 80  C or continue with re-concentration.

3.1.4 Re-concentration, Activation, and Desalting of MMP-3cd

1. Pour the “Refolded MMP” sample from the previous step into a 50-mL Amicon ultra centrifuge filter unit or 400 mL Amicon stirred cell with 10k MWCO membrane. 2. If using a 50-mL Amicon ultra centrifuge filter unit with 10k MWCO membrane, spin at 5000
g in a swing bucket rotor until a final concentration of 1 mg/mL is reached. 3. If using a 400 mL Amicon stirred cell with 10k MWCO membrane, wash the membrane with at least 25 mL of MilliQ H2O before adding the refolded MMP. Stir until a final concentration of 1 mg/mL is reached.

Engineering TIMPs using YSD

373

4. Make 1 mL aliquots in sterile 1.5-mL microcentrifuge tubes labeled as “Concentrated MMP” (see Note 7). 5. Make a fresh stock solution of 20 mM APMA in a fume hood right before activation. 6. Per 1 mL aliquot of MMP-3cd (with a concentration of 1 mg/ ml), add 50 μL of 20 mM APMA. APMA is an organomercurial agent that facilitates the loss of the enzyme propeptide domain through a cysteine switch, which causes the activation of MMPs. APMA is highly toxic, so work with it under a hood and discard remaining APMA in its own waste container. 7. Incubate MMP-3cd/APMA mixture overnight in 37  C shaking incubator at 250 rpm. 8. If a precipitate forms after MMP activation, centrifuge at maximum speed for 10 min at 4  C. 9. Store supernatant in sterile 1.5-mL microcentrifuge tube labeled as “Activated MMP,” and discard precipitate in APMA waste container. 10. Retrieve a Zeba spin desalting column and twist off the bottom closure. 11. Place Zeba spin column in sterile 15-mL conical tube and loosen cap. 12. Centrifuge column at 1000
g for 2 min to remove the storage solution. 13. Place a mark on the side of the column where the compacted resin is slanted outward (see Note 8). 14. Wash column with sterile MilliQ H2O, per manufacturer’s protocol. 15. Place column in new sterile 15-mL conical tube. 16. Remove the cap and slowly add 0.5–2 mL of sample to the column, carefully as to not disturb the resin. 17. Centrifuge at 1000
g for 2 min to collect the sample. 18. Repeat until all of the activated MMP-3cd has been run through the column. 3.1.5 Biotinylation of MMP-3cd

1. Follow the EZ-Link NHS-PEG4 biotinylation kit protocol to biotinylate purified MMP-3cd protein. 2. Add biotin in a 1:10 (protein:biotin) molar ratio. 3. Incubate the MMP:biotin solution at room temperature for 30 min. 4. Purify the biotinylated MMP-3cd via Zeba spin desalting column, following the manufacture’s protocol. 5. Test for degree of biotinylation using the HABA assay, following the kit protocol (see Note 9).

374

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

3.2 Design and Generation of TIMP-1 Variant Library 3.2.1 Inserting the TIMP1 Gene in the pCHA Yeast Display Vector

1. Construct primers that have homologous 15-bp overhangs to the pCHA-VRC01 yeast display vector up- and down-stream of NheI and BamHI RE sites, respectively, replacing the NheI restriction enzyme site with BsrGI restriction enzyme site (see Note 10). 2. Amplify the TIMP-1 gene via PCR, using the primers constructed from the previous step. 3. Digest the VRC01 gene from the pCHA-VRC01 vector using NheI and BamHI restriction enzymes and 10
CutSmart reaction buffer. 4. Run the PCR products of the TIMP-1 gene and the digested pCHA vector on a 1% agarose gel. 5. Gel purify the PCR and double digest products, following the kit protocol. 6. Assemble the pCHA digested vector and TIMP-1 gene using HiFi DNA assembly, and transform into chemically competent E. coli cells, following the manufacturer’s protocol. 7. Plate the transformed cells on LB + amp agar plates. Grow at 37  C overnight. 8. Pick and grow transformed colonies in LB + amp broth in shaking incubator overnight at 37  C and 250 rpm. 9. Extract and purify the TIMP-1 variant plasmid from the bacteria using a Miniprep kit, following the manufacturer’s protocol. 10. Sequence the variants using primers up- and down-stream of the TIMP-1 gene, to confirm proper assembly of the TIMP-1 gene in the pCHA vector.

3.2.2 Design Targeted Library of TIMP-1 Random Mutants

1. Choose the amino acids at the interface of TIMP/MMP interaction loops based on protein crystal structures (Fig. 2). For TIMP-1/MMP-3cd, the crystal structures include WT-TIMP1/MMP-3cd (PDB ID: 1UEA), TIMP-1-L34G (PDB ID: 6MAV), and TIMP-1-L34G/L133P/L151C/G154A (PDB ID: 6N9D). The 17 amino acids selected from five interacting loops (AB, C-connector, EF, MTL, and GH) of human TIMP-1 are highlighted in Fig. 2. 2. Design and order the gene blocks with selected residues marked for random mutations (NNS degenerate codon incorporation where N ¼ any nucleotide and S ¼ G or C) from gene synthesis companies that offer gene blocks with random mutations for selected residues.

3.2.3 Construct a Library of Human TIMP-1 Gene Variants

1. Using primers that have homologous 50-bp overhangs to the pCHA-TIMP-1 yeast display vector up- and down-stream of BsrGI and BamHI RE sites, respectively, amplify the yeast

Engineering TIMPs using YSD

375

MMP-3cd

C-connector GH

AB Loop MTL

C-TIMP-1

EF

N-TIMP-1

Fig. 2 Design of targeted random library for engineering TIMP protein scaffolds. Protein crystal structure of length human TIMP-1-mutant L34G in complex with MMP-3cd (PDB ID ¼ 6MAV) [10]. MMP-3cd is shown in cyan color, TIMP-1 N-terminal domain in light yellow, and C-terminal domain in gray. The residues in the five interacting loops of TIMP-1 (AB, C-connector, EF, GH, and MTL) at the interface of MMP-3cd that were chosen for random mutation are highlighted with their side chains in magenta

surface display TIMP-1 variant library through PCR using the targeted TIMP-1 random mutant gene block as the PCR template. 2. Digest the TIMP-1 gene from the pCHA-TIMP-1 vector using BsrGI and BamHI restriction enzymes. 3. Run the PCR products of the TIMP-1 gene variants and the digested pCHA vector on a 1% agarose gel. 4. Gel purify the PCR and double digest products, following the kit protocol. 5. Repeat PCR and double digest until 5 μg of cut vector and 25 μg of PCR product have been purified. 1 μg of cut vector and 5 μg of PCR product will be used for each electrotransformation, for a total of five electrotransformations.

376

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

3.2.4 Preparation of EBY100 Culture for Electrotransformation

1. Streak yeast strain EBY100 on a YPD plate and grow in a 30  C incubator for 2–3 days until single isolated colonies grow on the agar plate. 2. Inoculate a single isolated EBY100 colony in 5 mL YPD media and shake overnight at 30  C and 250 rpm. 3. Inoculate 50 mL YPD media with overnight culture, to an absorbance of 0.1 at 600 nm. 4. Grow the culture shaking at 30  C and 250 rpm, until the culture reaches an OD600 reading of 1.3–1.5, which will take approximately 6 h (see Note 11).

3.2.5 Electrotransformation of the TIMP1 Variant Library into the Yeast Strain EBY100 Cells

1. As the EBY100 cells grow, use Pellet Paint® Co-Precipitant to precipitate the pCHA digested vector and purified PCR product. 2. Resuspend each DNA pellet with sterile MilliQ H2O, to a final concentration of 1 μg/μL of cut vector and 5 μg/μL for TIMP1 variant insert for each electroporation cuvette that will be used. 3. Pellet the cell culture at 2500
g for 3 min at 4  C and aspirate the supernatant. 4. Resuspend the cells with 50 mL ice-cold sterile MilliQ H2O. 5. Repellet the cells, aspirate the supernatant, and repeat with 25 mL of ice-cold sterile MilliQ H2O. 6. Pellet the cells again and resuspend pellet in 2 mL of 1 M ice-cold sterile sorbitol. 7. Centrifuge as in step 3, and discard supernatant. 8. Resuspend cell pellet to a final volume of 150 μL in 1 M ice-cold sterile sorbitol. 9. Aliquot 50 μL of yeast cells into sterile 1.5-mL microcentrifuge tubes. 10. Add 1 μL of cut vector and 1 μL of TIMP-1 variant amplified gene insert with 50-bp overhangs to each 50 μL cell suspension, and flick gently to mix. 11. Incubate on ice for at least 5 min, without mixing. 12. Aliquot 50 μL of sample into a pre-chilled 2 mm electroporation cuvette and keep samples on ice. 13. Set up electroporation parameters on Bio-Rad Gene Pulser as follows: turn on micropulser and select “pre-set protocol,” choose “fungal” protocol, and select “S. cerevisiae, 2 mm” (see Note 12). 14. Load ice-cold electroporate.

cuvette

into pre-set

gene

pulser

and

15. Immediately aliquot 1 mL of pre-warmed to 30  C YPD media to the cuvette and mix by pipetting.

Engineering TIMPs using YSD

377

16. Record time constant and electroporate remaining cells. Time constant for these conditions usually fall between 4.9 and 5.1 ms (see Note 13). 17. Transfer all electroporated samples into a 50-mL conical tube and rinse each electroporation cuvette with 1 mL of YPD media to recover any remaining cells. 18. Incubate the 50-mL conical tube, shaking at 30  C and 250 rpm for 1 h. 19. Pellet the cells at 2500
g for 5 min. 20. Aspirate supernatant and resuspend cells in 10 mL of SD-CAA media, pH 4.5 + pen/strep. 21. Make several serial dilutions and plate on SD-CAA pH 6.0 agar plates. 22. Allow plates to grow at 30  C for 2–3 days and count colonies to determine the library size (see Note 14). 23. Add 10 mL of resuspended cells to 140 mL of SD-CAA, pH 4.5 + pen/strep media in a sterile 250 mL flask. 24. Grow cells for 24–48 h at 30  C with shaking at 250 rpm. The TIMP-1 variant library can be plated on SD-CAA, pH 6.0 agar plates, made into glycerol stocks, and/or passaged to be analyzed via flow cytometer. 3.3 Preparation of TIMP-1 Variant Yeast Surface Display Library 3.3.1 Passage of the TIMP-1 Variant Library

3.3.2 Preparation of Frozen Glycerol Stocks

1. Inoculate 50 mL of SD-CAA pH 4.5 + pen/strep with the TIMP-1 variant yeast library. 2. Grow the culture shaking overnight at 30  C and 250 rpm, or until culture has become saturated. The culture is considered saturated at an OD600 reading between 8 and 12. 3. Add 10 mL of culture into new 100 mL aliquot of SD-CAA pH 4.5 + pen/strep, and passage the cells. Based on your library diversity, estimate the volume of the cells needed to passage the next round (usually at 1:10 dilution) without losing the mutant diversity. 1. After three passages of the yeast library, aliquot 750 μL of culture into a sterile 2-mL cryovial, and add 750 μL of sterile solution of 40% glycerol to the same cryovial, and mix gently. 2. Store in a Styrofoam box for 4–6 h at 20  C, before moving to the library archive at 80  C freezer for longer storage.

3.3.3 Induction of the TIMP-1 Variant Library

1. Passage the library as described in Subheading 3.3.1. 2. After the final passage, inoculate the library in a 5 mL culture of SG-CAA pH 6.0 to a final OD600 of 1. 3. Induce yeast cultures for 18–22 h shaking at 30  C and 250 rpm (see Note 15). 4. Grow and induce pCHA-TIMP-1 transformants as the control.

378

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

3.4 FACS Screen of TIMP-1 Variant Library Toward MMP-3cd Binding 3.4.1 Preparation of Induced TIMP-1 Variant Library for Immunolabeling

1. Perform passage and induction of cells as outlined in Subheadings 3.3.1 and 3.3.3. 2. Measure the absorbance at 600 nm of the induced library. 3. Pellet the induced library cells to a final OD600 of approximately 1.5 in a 1.5-mL microcentrifuge tube. Depending on sorting percentage and number of cells aimed to be collected, multiple tubes need to be labeled, usually 3–5. 4. Aspirate supernatant using Pasteur pipette vacuum system. 5. Wash cells by resuspending them in ice-cold 0.1% PBSA. Keep cells on ice. 6. Pellet cells, aspirate supernatant, and wash them again, for a total of two washes. 7. After final wash, aspirate supernatant and place cells on ice. 8. Resuspend one pellet in 1500 μL of PBSA, and do not label with any antibodies or fluorophores. This will serve as a negative control.

3.4.2 Bind Biotinylated MMP-3cd to the TIMP-1 Variant Displayed Protein, and Strep-AF647 Labeling

1. Prepare all solutions on ice. 2. Prepare a 0.1 mg/mL stock solution of mouse anti-c-myc in filtered 0.1% PBSA. 3. Prepare a separate stock solution of 0.5 mg/mL of biotinylated MMP-cd in filtered 0.1% PBSA. 4. Prepare the first immunolabeling solution so that the final concentration of anti-c-myc is 1–2 μg/mL, respectively, in 0.1% PBSA. Prepare enough solution so that each pellet can receive a 100 μL aliquot. 5. Resuspend the washed pellet in 100 μL of the first immunolabeling solution and incubate on ice for 1 h. 6. After first immunolabeling incubation is complete, pellet the cells and aspirate the supernatant. 7. Wash the cells thrice in 0.1% PBSA, as outlined above. 8. Prepare second immunolabeling solution in opaque 1.5-mL microfuge tube, away from light. 9. Prepare the second immunolabeling solution so that the final concentration of anti-mouse Alexa Fluor-488 and streptavidin Alexa Fluor-647 conjugate are each 10 μg/mL in 0.1% PBSA. Prepare enough solution so that each pellet can receive a 100 μL aliquot. 10. After the final wash, resuspend the pellet using 100 μL of second immunolabeling solution. Incubate on ice, away from light, for 30 min.

Engineering TIMPs using YSD

379

11. After incubation, pellet the cells and aspirate the supernatant. 12. Wash the cells thrice in 0.1% PBSA, as outlined above. 13. Resuspend cells in 1500 μL of 0.1% PBSA. 3.4.3 Run Flow Cytometry and Screen for TIMP-1 Variant Populations of Interest

1. Set the flow cytometer cell sorter (e.g., BD FACS Aria) with sorting abilities to collect data on 25,000–750,000 cell events per sample, with medium fluidics. 2. Open a window for a scatter plot with FSC-A on the x-axis and SSC-A on the y-axis, a histogram with FITC-A as the x-axis, a histogram with APC-A as the x-axis, and two-dimensional plot with FITC and APC as the axis. 3. Briefly vortex the negative control sample and run it on the flow cytometer.

3.4.4 Set the Sorting Gate

1. Draw a polygon gate on the scatter plot to encompass the healthy cells that will be analyzed. 2. Gate the remaining plots so that only the data collected within the polygon is accounted for. 3. Use the bisector or range tool on the cytometer software to create a range gate on the FITC-A and APC-A histograms. 4. Draw a quadrant on the two-dimensional FITC-A/APC-A plot that separates Alex Fluor-488 positive, Alexa Fluor-647 positive, and negative samples. 5. Briefly vortex and run the labeled TIMP-1 variant library samples. Adjust the voltages/quadrants as necessary. 6. Draw a polygon gate that captures the higher ratio of binding to expression of double-positive cells on the two-dimensional FITC-A/APC-A plot (upper right quadrant) (Fig. 3).

3.4.5 Screen the TIMP-1 Variant Library Population

1. Set up a sterile 5-mL falcon tube, washed thoroughly with PBSA to discourage non-specific binding of yeast cells to the collection tube, containing 1 mL SD-CAA pH 4.5 + pen/strep in the collection area. 2. Specify the double-positive TIMP-1 variant population to be collected, and initiate the sort. 3. Do not collect more than 1 mL of sample per collection tube— another sterile culture tube with SD-CAA can be placed in the collection area. 4. After one tube has collected the desired volume of cells, cap the tube, vortex briefly, and place on ice (see Note 16).

380

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

Fig. 3 Flow cytometry two-dimensional scatter plots of dually labeled TIMP-1 wildtype and variant libraries. The polygonal P1 sort gate on the naı¨ve library (left panel) represents a population of cells that have a high binding ratio of MMP-3cd to TIMP-1/c-myc expression. After three rounds of FACS sorting the TIMP-1 variant library toward MMP-3cd, the screened yeast cells are significantly enriched for MMP-3 binding: TIMP-1 expression (right panel) [10]

3.4.6 Recover the Sorted TIMP-1 Variant Library

1. Wash the sides of the collection tubes with additional 4 mL SD-CAA pH 4.5 + pen/strep. 2. Grow the cells shaking overnight at 30  C and 250 rpm. 3. Induce the TIMP-1 variant library by culturing cells in SG-CAA media pH 6.0, overnight at 30  C. 4. Make glycerol stocks Subheading 3.3.2.

3.4.7 Test the Sorted TIMP-1 Variant Library for Improved MMP-3 Binding

of

rescued

cells,

following

1. Passage, induce, label, and run cells on flow cytometer following Subheadings 3.4.1 and 3.4.2. 2. Compare sorted population to population before sort, to determine if improved binding has occurred. If so, prepare a 1:10,000 dilution of the cells and spread on SD-CAA pH 6.0 plates. 3. Allow to grow in dry 30  C incubator for 2–3 days. Continue to Subheading 3.5. 4. Additional sorts can be conducted. Here, a total of five sorting rounds were completed, while incrementally decreasing MMP-3cd concentration from 50 to 5 nM. If additional sorts are required, repeat Subheading 3.4.

Engineering TIMPs using YSD

3.5 DNA Preparation and Evaluation of Individual TIMP-1 Variant Clones 3.5.1 Extraction of DNA Plasmids from the Sorted TIMP-1 Variant Clones and Transformation into E. coli Cells

381

1. Prepare a grid plate using an SD-CAA pH 6.0 agar plate. 2. Pick sorted colonies from Subheading 3.4.7 using sterile toothpicks and touch on the prepared labeled grid plate. 3. Using the same toothpick, inoculate in 1 mL YPD media in a 30  C incubator shaking at 250 rpm. 4. Allow the culture to grow to an OD600 of 0.2–0.6 (see Note 17). 5. Pellet the culture for 2 min at 600
g and aspirate the supernatant. 6. Use the Zymoprep yeast plasmid miniprep II kit to extract the plasmid DNA, following the standard protocol with YPD culture. Incubation with zymolyase at 37  C water bath should be performed for 60 min. 7. Transform the extracted TIMP-1 variant plasmids into chemically competent E. coli cells, following the manufacturer’s protocol (see Note 18). Plate the transformed cells on LB-Amp agar plates.

3.5.2 Purification of Extracted TIMP Variant DNA from Yeast Library and Sanger Sequencing

1. Pick the transformed colonies from Subheading 3.5.1, step 7 and inoculate 10 mL of LB + amp broth. 2. Grow the cultures overnight in a 37  C incubator shaking at 250 rpm. 3. Extract and purify the TIMP-1 variant plasmid from the bacteria using a Miniprep kit, following the manufacturer’s protocol. 4. Sequence the variants using primers up- and down-stream of the TIMP-1 gene. 5. Analyze the sequences using the software SnapGene, or any similar gene sequencing analysis software. Make note of the most common mutations after sequence alignment.

3.6 Evaluating Isolated TIMP Variants for Improved MMP Binding

1. Inoculate one colony from the labeled SD-CAA grid plate prepared in Subheading 3.5.1, step 2 in 5 mL of SD-CAA media pH 4.5 + pen/strep. 2. Grow the culture for 16–20 h shaking at 30  C and 250 rpm.

3.6.1 Growth and Induction of TIMP-1 Variant Clones

3. For induction, follow the same protocol outlined in Subheading 3.3.3.

3.6.2 Labeling the Yeast Displayed TIMP-1 Variant Clones and Testing for Improved MMP-3cd Binding via Flow Cytometry

1. Follow the protocol outlined in Subheadings 3.4.1–3.4.4. 2. Test TIMP-1 variants displayed on the yeast surface for binding to MMP-3cd using titration against different concentrations of biotinylated MMP-3cd between 0.5 nM and 10 μM (see Note 19).

382

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

3.6.3 Drawing the Binding Curve

1. Use the median fluorescent intensity values from flow cytometry software (e.g., FlowJo) to draw the sigmoidal binding curve for WT-TIMP-1 and TIMP-1 variants. 2. Set a gate for the healthy yeast cells for each clone and negative control, as outlined in Subheading 3.4.4. 3. Within the statistics band on FlowJo, click on “Median.” 4. Apply the median statistic for each filter (FL1 and FL4) of each TIMP-1 variant. The software will calculate the median fluorescence intensity (MFI) for each sample. 5. Subtract the MFI of the negative control from each clone. 6. A sigmoidal binding curve can be drawn for each clone, by using the MFI values as the y-axis, and the different concentration of biotinylated MMP-3cd between 0.5 nM and 10 μM as the x-axis.

3.6.4 Comparison of TIMP-1 Variants and WTTIMP-1 for Binding to MMP-3cd Binding

1. Calculate the MFI for binding (strepAF647/FL4) to expression (c-myc/FL1) for each clone as described in Subheading 3.6.3. For this analysis, MMP-3cd concentration of 1 μM was used to ensure saturated MMP-3cd binding to the displayed TIMP-1 variants. 2. Analyze statistical significance of improvement for binding/ expression normalized to WT-TIMP-1 values.

4

Notes 1. EDTA will not dissolve completely until the solution is at pH 8.0. Even then, it may take a long time to solubilize. Continue to stir vigorously, leaving the solution to stir overnight. 2. Induction can also be left overnight, which may increase protein yields. 3. Make 10 M urea solution fresh and not more than 1 day in advance. Dissolve completely and stir thoroughly before use. Do not heat or autoclave urea. Keep at room temperature. 4. Sonication in Inclusion Body Buffer can be repeated to recover more protein from lysed cell debris. However, too much sonication can harm MMP yield. 5. Keep 50–100 μL of each sample from each protein purification step for analysis using SDS-PAGE. 6. It is common for precipitation to form during MMP dialysis. Precipitation can be recovered after dialysis against all three buffers is completed. After centrifugation and transferring the supernatant into sterile 1.5-mL microcentrifuge tubes, dissolve

Engineering TIMPs using YSD

383

the precipitant in HT Equilibration Buffer. Store resuspended protein at 80  C, or repeat dialysis in new cassette or Snakeskin tubing. 7. Precipitate may form after concentration of MMP-3cd. Recover the precipitate if desired, as outlined in MMP re-solubilization section. 8. For all subsequent centrifugation steps, place column in centrifuge with the mark facing outward. 9. Use a HABA calculator to determine the degree of MMP biotinylation. If biotinylation was unsuccessful, such as there is less than 1 biotin per molecule, biotinylate the protein again. Increase the amount of biotin being added if necessary. 10. The “CXC” motif at the N-terminus of TIMPs has a key role in binding to the catalytic site of MMPs and efficient inhibition of MMPs through interaction with the MMP catalytic zinc ion. Thus, it is important to display TIMP proteins and their variants at the N-terminus of the Aga2p protein for an accessible N-terminus of yeast displayed TIMPs. The pCHA yeast vector which provides an N-terminal yeast surface display of protein of interest was used for this purpose. For further posttranslational processing of an intact “CTC” motif at the N-terminus of TIMP-1, a BsrGI restriction site (TGTACA) was cloned which introduced scarless “CT” followed after a dibasic (-KR-) sequence that is a canonical site for cleavage by the intracellular yeast endopeptidase Kex2 in pre-pro signal sequence. 11. It is important to cultivate the cells at an OD600 reading of 1.3–1.5. Cells outside of the desired range may not be able to take up DNA as easily, which will decrease transformation efficiency. If an OD above 1.5 is reached, it is recommended to inoculate a fresh 50 mL of YPD to an OD of 0.1. 12. If a different micropulser is being used for electrotransformation, set the parameters as: 1500 V, 25 μF, 200 Ω, and 2 mm cuvette. 13. Time constants should range between 4.9 and 5.1 ms. If time constants fall outside of that range, too much DNA or salt contaminants may be present. This can affect transformation efficiency and library size. Make sure cell pellet is clear of any noticeable debris and contamination and introduce additional wash steps with MilliQ H2O if necessary. 14. If the desired library size is not reached, there are a few variables to consider. If the 50 mL YPD cultures of EBY100 were cultivated in lag or stationary phase, the cells are not under optimal conditions for electrotransformations—ensure EBY100 cells are being cultivated once the culture reaches an

384

Mari R. Toumaian and Maryam Raeeszadeh-Sarmazdeh

OD of 1.3–1.5. Inspect the pellet after each wash step—is there any signs of contamination or debris? The pellet should be white, and if that is not the case, the reagents may need to be further filtered/sterilized to ensure there are no contaminants within the YPD media, MilliQ H2O, and/or 1 M sorbitol. Once pelleted, the cells must be kept on ice. Deviations in temperature can change membrane permeability, and effect transformation efficiency. If the time constant for the electroporation falls outside of the 4.9–5.1 ms range, this can indicate that contaminants were present in the sample, too much DNA was added, and/or EBY100 cells were cultivated in the lag or stationary phase. 15. Induced cells can be stored at 4  C for 2 weeks, without significant hindrance of yeast display or loss of cell viability. 16. Adding 500 μL of SD-CAA media to the collection tube can be used to recover the yeast cells in the collection tube that are not pelleted. This can also be used when number of cells collected are lower, or there are issues with yeast attaching to the collection tube. 17. Older yeast cells, or cells in lag phase, can be used as well. However, double the specified zymolyase must be added for sufficient cell lysis, and yields may not be as high as compared to using cells in log phase. 18. The gene of interest can also be amplified using PCR and purified using Promega PCR purification kit. Send the purified PCR product for gene sequencing using Sanger or nextgeneration sequencing. 19. The titration range was chosen to cover the extreme high and low concentrations. TIMP-1 and MMP-3cd in solution bind with mid-picomolar affinity. The titration range needs to be optimized for other binding partners with different affinities.

Acknowledgments M.R.-S thanks the NIH-P20 GM103650-COBRE Integrative Neuroscience grant. References 1. Boder ET, Raeeszadeh-Sarmazdeh M, Price JV (2012) Engineering antibodies by yeast display. Arch Biochem Biophys 526(2):99–106. https://doi.org/10.1016/j.abb.2012.03.009 2. Boder ET, Wittrup KD (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol 15(6):553–557. https://doi.org/10.1038/nbt0697-553

3. Pepper LR, Cho YK, Boder ET, Shusta EV (2008) A decade of yeast surface display technology: where are we now? Comb Chem High Throughput Screen 11(2):127–134. https:// doi.org/10.2174/138620708783744516 4. Cherf GM, Cochran JR (2015) Applications of yeast surface display for protein engineering.

Engineering TIMPs using YSD Methods Mol Biol 1319:155–175. https:// doi.org/10.1007/978-1-4939-2748-7_8 5. Boder ET, Wittrup KD (2000) Yeast surface display for directed evolution of protein expression, affinity, and stability. Methods Enzymol 328:430–444. https://doi.org/10.1016/ s0076-6879(00)28410-3 6. Chao G, Lau WL, Hackel BJ, Sazinsky SL, Lippow SM, Wittrup KD (2006) Isolating and engineering human antibodies using yeast surface display. Nat Protoc 1(2):755–768. https://doi.org/10.1038/nprot.2006.94 7. Egeblad M, Werb Z (2002) New functions for the matrix metalloproteinases in cancer progression. Nat Rev Cancer 2(3):161–174. https://doi.org/10.1038/nrc745 8. Raeeszadeh-Sarmazdeh M, Do LD, Hritz BG (2020) Metalloproteinases and their inhibitors: potential for the development of new therapeutics. Cells 9(5):1313. https://doi.org/10. 3390/cells9051313 9. Radisky ES, Raeeszadeh-Sarmazdeh M, Radisky DC (2017) Therapeutic potential of matrix metalloproteinase inhibition in breast cancer. J Cell Biochem 118(11):3531–3548. https://doi.org/10.1002/jcb.26185 10. Raeeszadeh-Sarmazdeh M, Greene KA, Sankaran B, Downey GP, Radisky DC, Radisky ES (2019) Directed evolution of the metalloproteinase inhibitor TIMP-1 reveals that its N-

385

and C-terminal domains cooperate in matrix metalloproteinase recognition. J Biol Chem 294(24):9476–9488. https://doi.org/10. 1074/jbc.RA119.008321 11. Arkadash V, Yosef G, Shirian J, Cohen I, Horev Y, Grossman M, Sagi I, Radisky ES, Shifman JM, Papo N (2017) Development of high affinity and high specificity inhibitors of matrix metalloproteinase 14 through computational design and directed evolution. J Biol Chem 292(8):3481–3495. https://doi.org/ 10.1074/jbc.M116.756718 12. Shirian J, Arkadash V, Cohen I, Sapir T, Radisky ES, Papo N, Shifman JM (2018) Converting a broad matrix metalloproteinase family inhibitor into a specific inhibitor of MMP-9 and MMP-14. FEBS Lett 592(7):1122–1134. https://doi.org/10.1002/1873-3468.13016 13. Yosef G, Arkadash V, Papo N (2018) Targeting the MMP-14/MMP-2/integrin alphavbeta3 axis with multispecific N-TIMP2-based antagonists for cancer therapy. J Biol Chem 293(34):13310–13326. https://doi.org/10. 1074/jbc.RA118.004406 14. Van Deventer JA, Wittrup KD (2014) Yeast surface display for antibody isolation: library construction, library screening, and affinity maturation. Methods Mol Biol 1131:151– 181. https://doi.org/10.1007/978-162703-992-5_10

Chapter 20 Discovery of Cyclic Peptide Binders from Chemically Constrained Yeast Display Libraries Kaitlyn Bacon, Stefano Menegatti, and Balaji M. Rao Abstract Cyclic peptides with engineered protein-binding activity have great potential as therapeutic and diagnostic reagents owing to their favorable properties, including high affinity and selectivity. Cyclic peptide binders have generally been isolated from phage display combinatorial libraries utilizing panning based selections. As an alternative, we have developed a yeast surface display platform to identify and characterize cyclic peptide binders from genetically encoded combinatorial libraries. Through a combination of magnetic selection and fluorescence-activated cell sorting (FACS), high-affinity cyclic peptide binders can be efficiently isolated from yeast display libraries. In this platform, linear peptide precursors are expressed as yeast surface fusions. To achieve cyclization of the linear precursors, the cells are incubated with disuccinimidyl glutarate, which crosslinks amine groups within the displayed linear peptide sequence. Here, we detail protocols for cyclizing linear peptides expressed as yeast surface fusions. We also discuss how to synthesize a yeast display library of linear peptide precursors. Subsequently, we provide suggestions on how to utilize magnetic selections and FACS to isolate cyclic peptide binders for target proteins of interest from a peptide combinatorial library. Lastly, we detail how yeast surface displayed cyclic peptides can be used to obtain efficient estimates of binding affinity, eliminating the need for chemically synthesized peptides when performing mutant characterization. Key words Cyclic peptides, Yeast surface display, Protein engineering, Combinatorial library screening, Ligand discovery

1

Introduction Cyclic peptides combine the desired properties of protein ligands— high affinity and selectivity [1]—without the limitations of protein ligands—cost [2], biochemical compatibility [3], and immunogenicity [4, 5]. Not surprisingly, cyclic peptides have gained significance as a class of alternative synthetic affinity ligands for diagnostic and therapeutic applications as well as basic research [6–9]. Unlike larger protein ligands, cyclic peptides are ideal for inhibiting protein–protein interactions; they can easily intercalate between protein–protein complexes due to their small size and modularity

Michael W. Traxlmayr (ed.), Yeast Surface Display, Methods in Molecular Biology, vol. 2491, https://doi.org/10.1007/978-1-0716-2285-8_20, © The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2022

387

388

Kaitlyn Bacon et al.

[10]. During chemical synthesis, cyclic peptides can also be modified with labels (e.g., fluorescent or radioactive probes) [11–13] or fixed to other biomolecules [14] to bestow additional biochemical functionalities. Several combinatorial approaches have been employed for the isolation of peptide binders, including genetic-based platforms, like phage-display [15–18], mRNA-display [19–21], ribosomal-display [22–24], and bacterial-display [25, 26], as well as synthetic platforms, like one-bead-one compound [27]. A unique feature of genetic-based platforms is the linkage of the expressed peptide (phenotype) to its encoded sequence (genotype). This allows high-affinity mutants to be isolated from combinatorial libraries using convergent steps of selection (i.e., directed evolution). Genetic-based platforms typically encode a linear peptide sequence. Various strategies are available for cyclizing the expressed linear peptide precursors, including chemical [28] and enzymatic approaches [29–31]. However, disulfide bond formation remains the most popular [32–34]. Interest has grown in engineering cyclic peptides fused to cell-penetrating peptide sequences to target intracellular proteins for therapeutic and research applications [14, 35– 37]. For intracellular applications, peptides cyclized via disulfide bond formation cannot be used, as the peptides are unlikely to maintain a cyclic structure within the reducing environment of the cytosol [38, 39]. Consequently, a broadly applicable cyclization method that utilizes stable bond formation is desired to increase the efficiency of identifying cyclic peptide binders for both intracellular and extracellular applications. For example, cyclization strategies relying on the formation of thioether and amide bonds have been used to stably cyclize peptides expressed using phage [28, 40] and mRNA display [39, 41–43]. When isolating peptide binders from a combinatorial library, a “panning” method is typically utilized to identify which mutants from the library have affinity for the target protein of interest. Specifically, the library is incubated with a surface that is decorated with the target protein; any mutants that interact with the surface are isolated for additional screening or identification. Despite the ease of isolating binders, panning selections do not discriminate based on affinity; both low- and high-affinity binders are isolated in a single population [44]. However, some display platforms, like yeast surface display, are amenable to fluorescence-activated cell sorting (FACS) that can be used to discriminate mutants based on affinity. While phage display has been frequently employed for generating libraries of cyclic peptides, phage display libraries cannot be sorted using FACS due to the small size of phage particles [45]. Peptides identified from libraries generated using platforms that are not amenable to FACS must be chemically synthesized for biochemical characterization (e.g., binding affinity). In contrast,

Chemically Cyclized Yeast Surface Display Peptide Libraries

389

peptide mutants expressed on the yeast surface can be biochemically characterized using immunofluorescent assays that depend on flow cytometry detection. Accordingly, we have developed a platform utilizing yeast surface display for the assembly of chemically cyclized peptide libraries that can undergo FACS to isolate the highest affinity library mutants. To date, limited examples exist of using yeast surface display to express cyclic peptide libraries [31, 46–48]; most instances have taken advantage of disulfidebond formation to constrain linearly expressed peptides rather than chemical crosslinkers that can afford stable cyclization [49– 52]. In our platform, a linear peptide precursor is expressed as an N-terminal fusion to the Aga2p subunit of the yeast mating protein a-agglutinin. The linear peptide is tethered to the yeast cell surface as the Aga2p subunit is disulfide bonded to the cell wall-associated Aga1p subunit of a-agglutinin. The displayed linear peptide precursor is cyclized after incubation with disuccinimidyl glutarate (DSG), which crosslinks the peptide’s N-terminal amine and an ε-amine of an included lysine residue (Fig. 1). When generating a combinatorial library, the amino acid residues between the N-terminal residue and a C-terminal lysine are randomized. The linear peptide is expressed with a N-terminal prepro secretory signal sequence that directs the nascent polypeptide through the secretory pathway [53–60]. As the polypeptide passes through the Golgi apparatus, the secretory sequence is cleaved between its C-terminal lysine and arginine residues [61, 62]. This results in an arginine residue at the N-terminus of the displayed linear peptide sequence. Thus, DSG crosslinks the α-amine of the N-terminal arginine with the ε-amine of the lysine residue located at the C-terminus of the peptide sequence. A HA epitope tag, which does not include any lysine residues, is fused to the C-terminus of the peptide to quantify peptide surface expression. Here, we describe a method to isolate cyclic peptide binders from a chemically crosslinked yeast display library of peptides. First, we give specific recommendations on how to display linear peptide sequences as N-terminal Aga2p fusions. Next, we provide detailed protocols on how to perform DSG crosslinking of yeast-displayed linear peptide precursors. Subsequently, we describe how to generate a yeast display library of linear peptide precursors followed by suggestions for performing magnetic and fluorescence based selections of the chemically crosslinked library to isolate high-affinity cyclic peptide binders. Finally, we include detailed discussion on estimating the affinity of individual cyclic peptide binders by titrating peptide-displaying yeast cells with soluble target protein.

390

Kaitlyn Bacon et al.

Fig. 1 Display of chemically crosslinked cyclic peptides on the yeast cell surface. Using the yeast surface display platform, a linear peptide precursor is expressed as a N-terminal fusion to Aga2p. The peptide–Aga2p fusion is tethered to the yeast wall by disulfide bonds formed between Aga2p and the cell wallassociated protein, Aga1p. A disuccinimidyl glutarate (DSG) crosslinker is used to cyclize the displayed linear peptide precursors between the α-amine of an arginine residue and the ε-amine of a lysine. The amino acids between the crosslinking residues can be randomized to generate a peptide combinatorial library

2

Materials All media and buffers should be made with deionized water at room temperature. Media for yeast growth and induction can be held at 4
C for 2 months. We recommend examining media for contamination before each use. All other buffers can be held at room temperature, unless otherwise noted. Yeast cultures can be stored at 4
C for short-term storage but should be held at 80
C for long-term storage.

2.1 Yeast Strains and Plasmids

1. Saccharomyces cerevisiae yeast strain EBY100 [63]. This yeast strain is leucine (LEU) and tryptophan (TRP) deficient. Plasmids with applicable selectable markers can be transformed into EBY100.

Chemically Cyclized Yeast Surface Display Peptide Libraries

391

Fig. 2 pCTCON-Nterm-expression cassette. This plasmid affords the expression of a linear peptide as a N-terminal fusion to Aga2p under the control of a GAL1/GAL10 bidirectional promoter. The prepro secretory sequence is cleaved as the fusion travels through the yeast’s secretory pathway resulting in the presence of an arginine residue at the N-terminus of each surface displayed peptide. A HA epitope tag is expressed downstream of the linear peptide sequence and can be used to quantify expression of peptide-Aga2p fusions on the yeast surface. A glycine/serine (GS) linker is included to increase the separation of the peptide from the yeast surface

2. Plasmid pCTCON-Nterm-Peptide (Fig. 2). This plasmid was constructed from the yeast surface display pCTCON plasmid to express a peptide sequence as a N-terminal fusion to a GS Linker, a HA tag, and Aga2p [64]. A prepro secretory sequence precedes the encoded peptide. Peptide expression is under the control of a galactose inducible promoter, GAL1/10. This plasmid contains a Trp selectable marker. 2.2 Yeast Media and Plates

1. Yeast Peptone Dextrose (YPD) medium (see Note 1): Mix 10 g/L yeast extract, 20 g/L peptone, and 20 g/L dextrose with an appropriate amount of water. Autoclave the solution. When preparing plates, also add 12 g/L agar and autoclave. 2. SDCAA medium (see Note 2): Dissolve 20 g/L dextrose, 6.7 g/L Difco™ yeast nitrogen base without amino acids (Becton Dickinson), 5 g/L casamino acids, 5.4 g/L Na2HPO4, and 8.6 g/L NaH2PO4 ∙ H2O in an appropriate amount of water. Use a 0.22-μm bottle top filter to sterilize. 3. SDCAA plates: Add 5.4 g Na2HPO4, 8.6 g NaH2PO4  H2O, 182 g sorbitol, and 12 g agar to 900 mL of water and autoclave. Separately, combine 20 g dextrose, 6.7 g Difco™ yeast nitrogen base without amino acids, and 5 g casamino acids in 100 mL of water followed by sterilization with a 0.22-μm filter. Add the filtered solution to the cooled autoclaved solution (temp 2 h. 8. Prepare the bead suspensions for both positive and negative sorts following Subheading 3.1.2, steps 16–18, using the bead incubation mixtures from steps 6 and 7, respectively. 9. Pellet (2000
g, 3 min) 20
library diversity. Resuspend cells in PBSF. Use 1.0 mL/1.4
109 cells. 10. Aliquot cell suspension (1.0 mL per 2.0-mL microcentrifuge tube) into an appropriate number of 2.0-mL microcentrifuge tubes. 11. Pellet cells (2000
g, 3 min), discard supernatant, then resuspend cells in 950 μL PBSF. 12. Pipette 50 μL of Dynabeads suspension (negative sort beads) from step 8 into each 2.0-mL microcentrifuge tube. 13. Incubate cells/Dynabeads suspension for >2 h at 4  C while rotating. 14. Place tubes on magnetic rack and collect supernatant. 15. Repeat steps 11–14 one or two additional times (optional). Naked beads, or off-target coated beads may be used in these negative sorts. Alternatively, proceed straight to step 15. 16. Perform the positive sort procedure outlined in Subheading 3.1.2 with the following adjustments: Adjust the volume of SDCAA medium with Pen-Strep to be appropriate for the current library diversity (typically 20 mL outgrowth is acceptable), perform at least two PBSF washes (Subheading 3.1.2, step 23), when plating cells to determine library density, factor in that it will be decreased with each sort, so less drastic dilution factor(s) may be more appropriate than those described in Subheading 3.1.3. 3.1.7 Continued Magnetic Sorting (Variable)

1. Repeat the positive and negative sorting processes described in Subheading 3.1.6 until library diversity has been reduced to a density of 3 h, or until OD600 ¼ 4.0 whichever is sooner (see Note 23). 5. Pellet 20
library diversity (2000
g, 3 min). 6. Resuspend pellet in SGCAA medium with Pen-Strep. Use an appropriate volume of medium to obtain a theoretical OD600 ¼ 1.0 density (see Note 24). 7. Incubate at 20  C for 16–48 h.

3.2.2 FACS Sample Preparation (1 Day)

1. Measure OD600 of induced cells. 2. To prepare the sub-library to be sorted, pellet (2000
g, 3 min) 20
diversity (typically 1 x 10 cells) in an appropriately sized conical tube. 3. To prepare the control samples, pellet (2000
g, 3 min) 1 x 107 cells for each of four control samples: negative control (naked yeast cells), single-color controls (one each for each fluorophore utilized, two in this case), three-component control (cells exposed to all antibodies and fluorophores, but not to the antigen). 4. Resuspend cells in 1.0 mL PBSF. 5. Pellet cell suspensions (2000
g, 3 min). 6. Resuspend cells in 1.0 mL PBSF, and pellet cell suspensions once again (2000 x g, 3 min), store on ice unless incubating with antigen or antibody solutions. 7. Resuspend cells from step 2 in 1 mL of antigen solution (for first FACS sort, ~100 nM, or 15 μL from 6.6 μM stock) and incubate for 20–30 min at ambient temperature while rotating (see Note 25). 8. Pellet (2000
g, 3 min) cells. 9. Resuspend cells in 1.0 mL PBSF.

434

Elliot Corless et al.

10. Pellet (2000
g, 3 min) cells, store on ice unless incubating with antibody solution. 11. Resuspend remaining cells in PBSF so that cell density is held constant at 1 x 108 cells/mL. (Antigen-labeled cells will be in 1 mL and controls will be in 100 μL). 12. Incubate cells (including control cells, except negative) with primary antibodies for 20–30 min at ambient temperature. Add 1.0 μL (1:1000 dilution) primary antibodies per 1 mL of cell suspension (Final concentration of all samples: 1 x 108 cells/mL. Incubate single color controls with only their relevant primary antibodies, incubate the three-component control and sample with both primary antibodies. 13. Repeat steps 5 and 6 twice. 14. Incubate cells with secondary antibodies for 15–30 min at 4  C (15 min for first FACS sort, longer incubation times are recommended for subsequent sorts with lowered antigen concentrations to maximize fluorescence of antigen-binding cells). Add 1.0 μL (1:1000 dilution) secondary antibodies to cell suspensions(Final concentration: 1 x 108 cells/mL). Incubate single color controls with only their relevant secondary antibodies, incubate the three-component control and sample with both secondary antibodies. 15. Repeat steps 5 and 6 then pellet (2000
g, 3 min) cells and store on ice until moving on to Subheading 3.2.3. 16. Just before FACS appointment, resuspend cells in 1.0 mL PBSF+, and transfer samples to Falcon 12
75 mm polypropylene test tubes. 3.2.3 FACS Sorting

1. Power up the FACS machine, and perform any relevant configuration steps for the particular machine. 2. Load the negative control and apply an appropriate gate based on the forward and side scattering profile to select for individual (singlet) cells. 3. Load a single-color control to calibrate the photomultiplier tube (PMT) voltage and axis for the first fluorophore. 4. Load the second single-color control to calibrate the PMT voltage and axis for the second fluorophore. 5. Load the three-component control to verify the axis calibration in previous steps (see Note 26). 6. Load the sample, and apply a gate that will select for cells with high fluorescence in both axes (see Fig. 3). 7. Sort all remaining cells, or a sufficient number to safely cover the library diversity, e.g., 2000–20,000 events, depending on gating, and round of sorting/current binder efficiency, into 5.0 mL SDCAA with Pen-Strep.

Affinity Pairs for Immunoassays

435

Fig. 3 Fluorescence intensity plots of three successive FACS operations. AlexaFluor488 fluorescence is plotted against AlexaFluor647 fluorescence indicating protein expression and antigen binding, respectively. Dots in plot represent single sorting events. Gates are shown as solid lines, and %/% numbers indicate the percentage of events in the top right quadrant gate and the polygon gate, respectively. Antigen concentration is denoted above each plot

8. Incubate (30  C, 220 RPM) suspended cells until the growth is opaque (OD600 ¼ 0.6–1.0) (see Note 27). 9. Pellet cells. 10. Resuspend cells in SGCAA media with Pen-Strep to induce for subsequent sort(s). 11. Incubate (20  C, 220 RPM) for 40–48 h. 3.2.4 Additional FACS Sorts

Library diversity must be reduced to ~200–1000 FACS events corresponding to a maximum of ~200–1000 unique clones (see Note 28), preferably with the highest possible binding affinity and specificity. This necessitates additional FACS processes. Additional sorts are performed identically to Subheadings 3.2.2 and 3.2.3, with some exceptions. Additional sorts implement reduced concentrations of antigen, decreased twofold for each subsequent sort. To increase specificity, nonspecific proteins are introduced in the form of Fetal Bovine Serum (FBS) which contains multiple competitive proteins (see Note 29). Additionally, to remove variants that bind to the poly-Histidine tag used to purify hCRP, Imidazole is incorporated during the antigen incubation step in subsequent sorts (see Note 30). The composition of three additional sorts for hCRP are shown in (Table 2). Sorts can be repeated if needed to meet assay performance criteria, and antigen concentration can be reduced to 1.0 nM or lower, as per assay requirements.

436

Elliot Corless et al.

Table 2 FACS sort conditions Sort number 2 hCRP

50.0 nM

Imidazole

250 mM

FBS

50% v/v

PBSF

1


Sort number 3 hCRP

25.0 nM

Imidazole

250 mM

FBS

50% v/v

PBSF

1


Sort number 4

3.3 Variant Identification

3.3.1 Variant Isolation and Sequence Determination

hCRP

12.5 nM

Imidazole

250 mM

FBS

50% v/v

PBSF

1


Once the rcSso7d library has been reduced to ~200–1000 FACS events, individual variants must be isolated, sequenced, and later characterized. Plasmids are isolated by mini-prep from yeast cells, then transformed into competent E. coli to allow facile sequencing of individual variants. The binding affinities of selected clones are characterized by transformation back into yeast, followed by flow cytometry titration assays to determine apparent dissociation constants. 1. Incubate (30  C, 220 RPM) reduced library produced from Subheading 3.2 in SDCAA medium with Pen-Strep. 2. Monitor the OD600 until culture reaches 50% or the WT amino acid is Arg, His, Lys, or Cys, exclude this site from diversification. If the frequency of Tyr and Asp are at least 2% at this position, force inclusion of

Alec A. Desai et al.

A 256

B Gate R6

192

1419

Count

Side scatter

128 64

946 473

Gate R6 0

0 0

64

128

192

256

101

Forward scatter

102

103

104

Fluorescence (RFU) EC50 (nM)

C Antibody Binding (RFU)

3 2

C1 4.8±1.3 C2 6.5±1.3 C3 8.4±0.6

1

WT 90.0±1.78

0 0.1

1

10

100

[Antibody] (nM)

D % Bound antibody

482

120 WT C2 C3 C1

100 80 60 40 20

Nonconformational antibody

1

10

100

1000

[Disaggregated Aβ] (nM)

Fig. 2 Flow cytometry-based analysis of antibody affinity and conformational specificity. Amyloid fibrils are immobilized on micron-sized magnetic beads and binding of soluble antibodies is evaluated using flow cytometry. (a) Flow cytogram displaying forward scatter (488 nm) versus side scatter (488 nm) results, with the singlet population of fibril-coated beads highlighted in the R6 gate. (b) Histogram of the counts of binding events in the R6 gate as a function of the antibody binding signal detected via fluorescence measurements (647 nm). (c) Relative binding of antibodies to Aβ fibrils as a function of the concentration of wild-type (WT) and affinity-matured (C1, C2 and C3) antibodies. The results are mean signals from the histogram shown in (b). (d) Antibody binding to fibrils in the presence of disaggregated Aβ. Each antibody (30 nM) was pre-incubated with different concentrations of disaggregated Aβ, and then the antibodies were evaluated for their ability to recognize Aβ fibrils. A control (nonconformational) antibody displays decreased binding to immobilized fibrils as the concentration of disaggregated Aβ is increased

Anti-Amyloid Antibody Discovery and Maturation

483

Fig. 3 Overview of the design of antibody sub-libraries for affinity maturation. The design of sub-libraries for affinity maturation involves three major steps. First, the CDR positions to mutate are chosen based on the variability of each CDR position in natural antibody repertoires. The sites are ranked from most variable (most attractive for mutagenesis) to least variable (least attractive for mutagenesis), and highly conserved positions (>50% WT on average in natural antibody repertoires) are eliminated from consideration. Moreover, a subset of CDR sites are also excluded from consideration if their WT residues are Arg, Lys, His or Cys. Second, for the selected ~6–10 CDR sites with the highest variability, degenerate codons are chosen that encode the WT residue as well as 3–5 other amino acids based on maximizing the sum of the average site-specific frequencies of each encoded residue in natural antibody repertoires (referred to as natural diversity coverage). Degenerate codons with Arg, Lys and His are excluded. The libraries are designed to typically encode 106 to 108 variants. If there are multiple possible degenerate codons that encode the same set of amino acid mutations, codon selection is based on species-specific codon usage (e.g., S. cerevisiae codon usage). Third, mutagenic primers with degenerate codons and amplification primers without mutations are designed. One mutagenic primer is designed for each mutated CDR that contains the site-specific degenerate codons and 18–30 base pairs of framework DNA on both ends of the mutated CDR. Four additional primers are required for generating sub-libraries with a single mutated CDR. Typically two or three CDRs are mutated when generating sub-libraries for affinity maturation

484

Alec A. Desai et al.

these amino acids in the diversification. If the frequency of Tyr or Asp are at least 2% at this position, force inclusion of the more common amino acid in the diversification (see Note 11). Rank the resulting positions amenable to diversification in order of increasing variability. 6. Selection of degenerate codons: To generate a library of 106 to 108 variants, six to ten positions need to be mutated to five different amino acids (wild type and five mutations). For each of the nine most variable sites, choose the degenerate codons that encode the required amino acids (e.g., WT, Tyr, Asp). Remove any codons that encode for more than six amino acids (see Note 12). Remove any codons that encode for Arg, His, Lys or a stop codon. For the remaining codons, calculate the natural diversity coverage, which is defined as the sum of the average frequencies of the encoded amino acids for a given position in natural antibody repertories. For example, if position H52 is being mutated and Tyr is the wild type, the codon DMT would encode for Ala, Tyr, Ser, Thr, Asn, and Asp and would give a natural diversity coverage of 78.5%. Choose the codon with the maximal natural diversity coverage. If there are multiple codons, use codon usage for the host organism to choose the most favorable codon. 7. Designing primers for generating libraries: For each mutated CDR, design a primer that encodes for the entire CDR with the degenerate codons. Include 18–30 base pairs of DNA on each end of the CDR. Use the 18–30 base pair regions before and after each CDR as primers to amplify and connect each fragment of DNA. 3.3.2 Preparing Libraries

The following protocol for library transformation (beginning with step 11) is adapted from a previous publication [14] with minor changes. 1. Prepare plasmid for the wild-type antibody in the yeast surface display plasmid. 2. Amplify regions of antibody that will not be diversified using standard PCR. The use of proof-reading enzymes like Q5 DNA polymerase or an equivalent proof-reading DNA polymerase is recommended. 3. Run the PCR product on a 1% agarose gel and purify the desired band by gel extraction. 4. Using amplified constant regions as templates, perform PCR with primers containing designed degenerate codons for each diversified CDR loop separately. The region of the primer containing degenerate codons should overhang the template DNA, such that the region complementary to the degenerate

Anti-Amyloid Antibody Discovery and Maturation

485

codons is not present during the PCR. A proof-reading polymerase and PCR cycles identical to that used in step 2 may be used (see Note 13). 5. Repeat step 4 as necessary to obtain diversity in all desired CDRs. The PCR product from a previous PCR incorporation of degenerate codons or from amplification of the constant regions may be used as the template as necessary. 6. Run the PCR product on a 1% agarose gel and purify the desired band by gel extraction. 7. Perform an overlap PCR to assemble DNA encoding all regions of the antibody as well as 30–50 base pairs of overlap with the vector at both the 50 and 30 ends. This overlap enables assembly of the plasmid via homologous recombination after transformation into yeast. 8. Run the PCR product on a 1% agarose gel and purify the desired band by gel extraction. 9. Double-digest the backbone, incubate with calf intestine alkaline phosphatase (CIP), and purify by gel electrophoresis (see Note 14). 10. Mix 12 μg of PCR insert with 4 μg of linear backbone in a 1.5 mL tube. Ethanol precipitate the DNA mixture with pellet paint co-precipitant. Uncap the tube and wrap it in aluminum foil to dry by placing it under vacuum or in a chemical hood. 11. Two days prior to transformation, start a 5 mL culture of EBY100 in YPD supplemented with 100 μg/mL ampicillin, 100 μg/mL kanamycin, and 100 pen-strep. Grow cells overnight at 30
C at 225 RPM. 12. The day before transformation, inoculate a 50 mL culture of EBY100 in YPD supplemented with 100 μg/mL ampicillin, 100 μg/mL kanamycin, and 100 pen-strep from the 5 mL culture grown the previous day. Grow cells overnight at 30
C at 225 RPM. 13. The day of transformation, inoculate an EBY100 culture in YPD supplemented with 100 μg/mL ampicillin, 100 μg/mL kanamycin, and 100 pen-strep from the 50 mL culture grown overnight. This culture should be seeded at an OD of 0.3. The volume of this culture may be scaled up depending upon the number of libraries that will be transformed. For every library that will be transformed, prepare 50 mL of EBY100. 14. Grow the cells in a shaker incubator at 30
C and 225 RPM until the OD reaches ~1.6. This step will take approximately 5–6 h. 15. Add 50 mL of EBY100 culture to 50 mL centrifuge tube. Centrifuge at 2500  g for 5 min and discard supernatant.

486

Alec A. Desai et al.

16. Wash the cells by resuspending in 25 mL of filtered cold DI water. Centrifuge at 2500  g for 5 min and discard supernatant. 17. Wash the cells by resuspending in 25 mL of cold electroporation buffer. Centrifuge at 2500  g for 5 min and discard supernatant. 18. Resuspend cells in 25 mL of lithium acetate/DTT buffer. Incubate cells in a shaker incubator at 30
C at 225 RPM for 10–15 min. 19. Centrifuge cells at 2500  g for 5 min and discard supernatant. 20. Wash the cells by resuspending in 25 mL of cold electroporation buffer. Centrifuge at 2500  g for 5 min and discard supernatant. 21. Resuspend cell pellet in 100–200 μL of cold electroporation buffer such that the total volume of the cells and buffer does not exceed 400 μL. 22. Transfer cells to tube containing DNA prepared by ethanol precipitation and resuspend the DNA completely in the cell solution by through mixing. 23. Transfer cells to a 2 mm electroporation cuvette and keep them on ice for 5 min. 24. Electroporate cells at the following settings: 2500 V, 200 Ω, 25 μF. 25. Immediately after electroporation, resuspend cells in 1 mL of 1:1 YPD:sorbitol (using 1 M sorbitol). Return cuvette to ice, and repeat electroporation process for remaining libraries. 26. Transfer electroporated cells to a 14 mL culture tube, and add 1:1 YPD:sorbitol to reach a total volume of 8 mL. 27. Allow cells to recover in a shaker incubator at 30
C and 225 RPM for 1 h. 28. Centrifuge cells at 2500  g for 5 min. 29. Resuspend cells in 1 mL of SDCAA. Take a small aliquot of cells at this stage to prepare serial dilutions for plating. Plate 104 and 105 serial dilutions on dropout plates and allow to grow at 30
C for 2 days. Library size can be determined from the number of colonies. Typical transformational efficiencies vary between ~106 and ~107 transformants per μg of linearized vector. 30. Transfer remaining cells in SDCAA to a 1 L culture flask containing 200 mL of SDCAA supplemented with 100 μg/ mL ampicillin, 100 μg/mL kanamycin, and 100 pen-strep. 31. Grow library for 36–48 h at 30
C at 225 RPM.

Anti-Amyloid Antibody Discovery and Maturation

487

32. Regrow library in 200 mL of SDCAA for 16–18 h at 30
C and 225 RPM. Library can be stored at this stage by centrifuging at 2500  g, resuspending cells in 30% glycerol and 0.67% yeast nitrogen base, aliquoting into cryovials, and freezing at 80
C. 33. Induce library in 200 mL of SDGCAA. 3.3.3 Affinity Maturation Library Sorting

1. Prepare magnetic beads as described in Subheading 3.1.1. For affinity maturation, it is recommended to reduce the antigen concentration during sorting. This can be achieved in two ways. One approach is to reduce the antigen loading on the beads. Five- to 20-fold reductions in the mass of antigen immobilized on beads is recommended. Reduced antigen loading is recommended in the case where streptavidin beads are used. A stock of beads at lower antigen concentration may be prepared prior to beginning affinity maturation sorting as described in Subheading 3.1.1 and prepared for sorts as described in Subheading 3.1.2. A second approach is to reduce the number of beads. It is recommended that 107 beads be used in the first few positive selections, and in subsequent rounds, the number of beads may be decreased incrementally to as low as 1  106 beads. Reducing the number of beads used during sorting is recommended in cases in which enrichment for background binding (i.e., antibodies which bind to streptavidin beads or glycine-blocked tosyl beads) has been noted to be an issue during initial sorting or preliminary affinity maturation sorting. This has been observed more frequently in cases in which tosyl beads have been used. A stock of beads may be prepared for this strategy following the same protocol as described in Subheading 3.1.1. 2. Depending on the antigen type and library designs, one technique might work better than the other. It is recommended to try both approaches and see which one works the best. 3. Perform positive and negative selections as described in Subheading 3.1.3 (see Note 15).

3.3.4 Clone Evaluation of Affinity and Conformational Specificity

The evaluation of the affinity and conformational specificity of affinity-matured clones is evaluated generally as described in Subheading 3.2.3. However, the multivalency of fibril antigens can mask intrinsic differences in affinity for affinity-matured variants. Therefore, it is recommended to reduce the antigen concentration on the beads for affinity evaluation if required. Five- to 20-fold reductions in antigen loading may be tested in order to determine the appropriate loading condition. At an appropriate loading concentration, a sigmoidal binding curve should be obtained when a range of antibody concentrations are examined.

488

Alec A. Desai et al.

1. Prepare beads at low antigen loading, as described in Subheading 3.3.3. 2. Take desired number of beads (105 beads/sample) and block with blocking buffer (PBSB with 10% milk) for streptavidin beads or 10 mM glycine (pH 7.4) for tosyl-activated beads. 3. Post-blocking, wash the beads once with washing buffer (PBSB). 4. Thaw antibody solution and centrifuge at maximum speed for 5 min in a table-top centrifuge. Move the supernatant to a fresh tube and measure the concentration. 5. Incubate beads with antibody at desired concentrations. It is recommended that a broad range of antibody concentrations, such as from 0.1 to 300 nM, be tested during affinity maturation. Examined concentrations should display a range of signals, from minimal to no detectable binding at the lowest concentrations to saturated binding signal at the highest concentrations (see Fig. 2c). If a sigmoidal curve cannot be obtained with this range of antibody concentrations, it is recommended the mass of antigen on the beads be reduced. 6. From this point, sample preparation and flow cytometry analysis is performed as described in Subheading 3.2.3, beginning with step 4.

4

Notes 1. Cells can be induced in either SDGCAA at 20
C for ~36–40 h or SGCAA at 30
C for ~20–24 h. SGCAA contains the same composition as SDGCAA without glucose. Induction in SDGCAA increases the percentage of yeast cells in the library which express antibodies in some cases. 2. It is important to use freshly prepared milk solution for each sort. Centrifugation of the milk solution prior to use removes insoluble aggregates, which improves experimental reproducibility. 3. The anti-Myc antibody is used to detect antibody expression on the yeast surface. It is preferred that antibody expression be measured via detection of a peptide tag that is on the opposite terminus of the antibody from the linker (i.e., linker-antibodytag or tag-antibody-linker). This ensures that only full-length antibodies are selected. 4. Neutralization should be performed only for IgGs. For scFv-Fc fusion proteins, it is recommended to buffer exchange the protein rather than neutralize immediately after elution from

Anti-Amyloid Antibody Discovery and Maturation

489

the Protein A beads. The isoelectric points of scFv-Fc fusion proteins are typically lower than those for IgGs, and neutralization may cause precipitation of the fusion proteins. 5. Filtered electroporation buffer and DI water should be prepared fresh on the day that the libraries are prepared. These solutions can be prepared immediately after starting the EBY100 culture and kept on ice until use. It is important that these buffers are cold at the time of use. 6. It is important that lithium acetate/DTT buffer is prepared immediately before use to prevent the degradation of DTT. 7. It is best to perform a titration to evaluate the concentration of fibrils immobilized on magnetic beads. For initial antibody discovery, it is preferred to saturate the beads with the antigen to avoid enriching for antibodies binding to streptavidin or the magnetic bead surface. 8. It is important to perform multiple positive and negative selections for selection of conformational antibodies. However, the order in which they are performed is flexible and both orders of operation are useful. For initial antibody discovery, negative selections should be emphasized to isolate lead clones with highest conformational specificity. 9. It is recommended to always centrifuge antibody solutions after thawing to remove aggregates formed during freezing and/or thawing. This improves reproducibility in antibody binding assays. 10. HCDR1 is considered to be a combination of the Kabat and Chothia definitions, H26 to H35B. HCDR3 is defined as H93 to H102, which includes two additional residues on the N-terminus of the CDR. All other CDRs follow the Kabat scheme. 11. A simple diversification strategy would be to mutate positions with WT frequency 99.5%) (Sigma Aldrich). 14. HisPur Ni-NTA resin (ThermoFischer scientific). 15. Wash 1 buffer: PBS complemented with 20 mM imidazole, pH 7.4. 16. Wash 2 buffer: PBS complemented with 40 mM imidazole, pH 7.4. 17. Elution buffer: PBS complemented with 0.5 M imidazole, pH 7.4.

Engineering Fluorescence-Activating Proteins

3

603

Methods

3.1 Construction of the Library 3.1.1 Saturation Mutagenesis

1. Prepare purified plasmid DNA containing the gene that needs to be mutagenized. Measure its concentration spectrophotometrically. 2. Saturation mutagenesis is performed by site-directed mutagenesis PCR using primers with degenerate codons (see Note 1). If several PCR fragments have to be prepared to generate the fulllength gene, they must contain overlapping sequences to allow assembly by PCR in a second step. 3. Prepare the PCR reagents in a thin-walled, 0.2 mL tube: x μL Plasmid DNA template; 1 μL Q5 polymerase; 2.5 μL Forward primer (from 10 μM stock); 2.5 μL Reverse primer (from 10 μM stock); 1 μL 10 mM dNTP mix; 10 μL 5 Reaction buffer, x μL Ultrapure deionized H2O to a final volume of 50 μL. 4. Mix the sample and place the tube in the thermocycler. 5. Run the PCR program: Step 1: 95  C (2 min); Step 2: 95  C (30 s); Step 3: annealing temperature for primers (30 s); Step 4: 72  C (1 min); Step 5: repeat step 2–4 for an additional 35 cycles; Step 6: 72  C (10 min); Step 7: 4  C (hold). 6. Characterize the PCR product by running 2 μL of the PCR reaction on a 1% SYBR-green-stained agarose gel, alongside a DNA ladder. 7. Purify the PCR product using QIAquick PCR Purification kit according to the manufacturer’s protocol. 8. Measure the concentration spectrophotometrically.

of

the

sample

9. For assembly of several PCR fragments, repeat steps 3–8 using the fragments to be assembled as template for the PCR. 3.1.2 Error-Prone PCR

1. Prepare purified plasmid DNA containing the gene that needs to be mutagenized (see Note 2). Measure its concentration spectrophotometrically. 2. For the construction of library by error-prone PCR, the mutation rate is highly dependent on the amount of template DNA used for the amplification. Various amounts of DNA template should be tested to reach the desired mutation rate (see Note 2). 3. Prepare the error-prone PCR reagents in a thin-walled, 0.2 mL tube: x μL Plasmid DNA template (see Note 2); 1 μL Mutazyme II; 2.5 μL Forward primer (from 10 μM stock); 2.5 μL Reverse

604

Lina El Hajji et al.

primer (from 10 μM stock); 1 μL 10 mM dNTP mix; 5 μL 10 Reaction buffer, x μL Ultrapure deionized H2O to a final volume of 50 μL. 4. Mix the sample and place the tube in the thermocycler. 5. Run the error-prone PCR program: Step 1: 95  C (2 min); Step 2: 95  C (30 s); Step 3: annealing temperature for primers (30 s); Step 4: 72  C (1 min); Step 5: repeat step 2–4 for an additional 35 cycles; Step 6: 72  C (10 min); Step 7: 4  C (hold). 6. Characterize the PCR product by running 2 μL of the PCR reaction on a 1% SYBR-green-stained agarose gel, alongside a DNA ladder. 7. Purify the PCR product using QIAquick PCR Purification kit according to the manufacturer’s protocol. 8. Measure the concentration spectrophotometrically. 3.1.3 DNA Shuffling

of

the

sample

1. Prepare purified plasmid DNA containing the homologous genes that need to be shuffled. Measure their concentration spectrophotometrically. 2. Amplify the homologous genes separately by PCR. Set up a mixture of the forward and reverse primers with each parental gene as a template and use a high-fidelity DNA polymerase (see Notes 3 and 4). 3. Prepare the PCR reagents in a thin-walled, 0.2 mL tube: x μL Plasmid DNA template (see Notes 2–4); 1 μL Q5 polymerase; 2.5 μL Forward primer (from 10 μM stock); 2.5 μL Reverse primer (from 10 μM stock); 1 μL 10 mM dNTP mix; 10 μL 5 Reaction buffer, x μL Ultrapure deionized H2O to a final volume of 50 μL. 4. Mix the sample and place the tube in the thermocycler. 5. Run the PCR program: Step 1: 95  C (2 min); Step 2: 95  C (30 s); Step 3: annealing temperature for primers (30 s); Step 4: 72  C (1 min); Step 5: repeat step 2–4 for an additional 35 cycles; Step 6: 72  C (10 min); Step 7: 4  C (hold). 6. Characterize the PCR product by running 2 μL of the PCR reaction on a 1% SYBR-green-stained agarose gel, alongside a DNA ladder. A single sharp band of the expected size should be observed. 7. Purify the PCR product using QIAquick PCR Purification kit according to the manufacturer’s protocol. 8. Measure the concentration spectrophotometrically.

of

the

sample

Engineering Fluorescence-Activating Proteins

605

9. For the shuffling step, mix the PCR products to be shuffled for 5 min at 15  C. Up to 5 μg of template can be used. Digest the mix with 0.02 unit of DNase I (20 U/mL) for 25 min at 37  C followed by 10 min at 80  C for complete denaturation of the enzyme (see Note 5). 10. Run the digested product on a 2% SYBR-green-stained agarose gel, alongside a DNA ladder and cut DNA bands out of the agarose gel. Purify the DNA sample using the QIAquick gel extraction kit according to the manufacturer’s protocol (see Note 6). 11. Assemble the fragments in a primerless PCR-like reaction. 12. Prepare the reagents in a thin-walled, 0.2 mL tube: x μL 1–2 μg of purified fragments; 0.5 μL Taq polymerase (5 U/μL); 2.5 μL MgCl2 (from 50 mM stock); 1 μL 10 mM dNTP mix; 5 μL 10 ThermoPol buffer (Mg2+ free); x μL Ultrapure deionized H2O to a final volume of 50 μL. 13. Mix the sample and place the tube in the thermocycler. Run the primerless PCR program: Step 1: 90  C (30 s); Step 2: 42  C for reassembly (see Note 7) (30 s); Step 3: 72  C (1 min/kb); Step 4: repeat step 2–3 for an additional 45 cycles; Step 5: 72  C (10 min); Step 6: 4  C (hold). 14. Characterize the shuffled PCR product by running 2 μL of the PCR products on a 1% SYBR-green-stained agarose gel, alongside a DNA ladder. A smear should be observed. 15. Amplify the shuffled product by PCR with primers containing the appropriate restriction enzymes for subsequent digestion. Set up a mixture of the forward and reverse primers with the primerless PCR product as a template and use a high-fidelity DNA polymerase: x μL 20–100 ng of primerless PCR product; 0.5 μL Q5 DNA polymerase (5 U/μL); 2.5 μL Forward primer (from 10 μM stock); 2.5 μL Reverse primer (from 10 μM stock); 1 μL 10 mM dNTP mix; 10 μL 5 Reaction buffer, x μL Ultrapure deionized H2O to a final volume of 50 μL. 16. Mix the sample and place the tube in the thermocycler. Run the PCR program: Step 1: 98  C (30 s); Step 2: 98  C (10 s); Step 3: annealing temperature for primers (30 s); Step 4: 72  C (30 s/kb); Step 5: repeat step 2–4 for an additional 20 cycles; Step 6: 72  C (10 min); Step 7: 4  C (hold). 17. Characterize the amplified PCR product by running 2 μL of the PCR reaction on a 1% SYBR-green-stained agarose gel, alongside a DNA ladder. 18. Purify the PCR product using QIAquick PCR Purification kit according to the manufacturer’s protocol. Measure the concentration of the sample spectrophotometrically.

606

Lina El Hajji et al.

3.1.4 Vector and Insert Digestion

1. Prepare high quantity of purified pCTCON2 plasmid using the HISpeed Plasmid Maxi kit according to the manufacturer’s protocol. 2. Digest 3 μg of pCTCON2 plasmid with NheI-HF and BamHIHF restriction enzymes for 2.5 h at 37  C in 1 CutSmart buffer. Do the same with the insert prepared in Subheadings 3.1.1, or 3.1.2 or 3.1.3 (see Note 8). 3. Add 5 μL of calf intestinal alkaline phosphatase (CIP) to the plasmid sample to prevent plasmid recircularization. Incubate at 37  C for 2.5 h. 4. Purify the digested samples using QIAquick PCR Purification kit according to the manufacturer’s protocol. 5. Measure the concentration spectrophotometrically.

of

the

samples

6. Store the purified DNA at 20  C, if necessary. 3.1.5 Construction of a Test Library

1. Prepare two ligation reactions: one with the digested plasmid only (“plasmid only” ligation) and one regular ligation with the digested plasmid and the digested insert DNA (“test library” ligation). For the test library, the ratio plasmid:insert must be optimized to maximize transformation efficiency (see Note 9). We typically use 250 ng of digested plasmid and 50 ng of insert (of 375 bp) for ligation (which correspond to a molar plasmid: insert ratio of 1:3). Each reaction should contain the following: 1 ligation buffer; 250 ng of vector DNA; 1.5 μL T4 DNA ligase (400,000 U/mL); with or without the digested insert DNA; and ultrapure deionized H2O to a final volume of 25 μL. 2. Incubate the two ligation reactions at 16  C overnight. 3. Purify the two ligation reactions using MinElute PCR Purification kit according to the manufacturer’s protocol. 4. Measure the concentration spectrophotometrically.

of

the

samples

5. Use 50 ng of each ligation product to transform 50 μL of DH10β E. coli electrocompetent cells by electroporation (see Note 10). After electroporation, add 500 μL of LB medium and incubate the bacterial culture for 45 min at 37  C in an orbital shaker at 220 rpm. 6. For determining the number of transformants, resuspend 5 μL of the bacterial culture in 45 μL of LB (¼S0 ¼ 5/550 of the initial suspension). Dilute S0 by a factor of 10 to 100,000 (S1 to S5) in a similar fashion. Spread 45 μL of each dilution (S0 to S5) on LB-agar plates supplemented with 0.1 mg/mL ampicillin and incubate at 37  C overnight. The next day, count the number of bacterial colonies on each plate to determine the number of transformants (see Note 11). The “plasmid-only”

Engineering Fluorescence-Activating Proteins

607

ligation should lead to very few colonies (negative control). The number of transformants of this test library enables to estimate the number of ligation reactions and transformations that will be necessary for constructing the final library (see Note 12). 3.1.6 Analysis of the Mutagenesis Quality of the Test Library

1. Randomly pick 12 colonies and inoculate each colony in 2 mL LB supplemented with 0.1 mg/mL ampicillin. Grow bacterial cultures at 37  C overnight in an orbital shaker at 220 rpm for subsequent extraction of the DNA plasmids. 2. Purify the plasmids using QIAprep Spin Miniprep kit according to the manufacturer’s protocol. Measure the concentrations of the samples. 3. Sequence with adequate primers (custom primer ag761 or M13-Forward), and evaluate the quality of the mutagenesis step (see Notes 13–15).

3.1.7 Construction of the Large-Scale Library in Bacteria

1. Prepare n ligation reactions with the optimized conditions determined in Subheading 3.1.5. Prepare a ligation mix for n reactions containing the following: 1 ligation buffer; n  250 ng of digested plasmid DNA; n  optimized quantity of digested insert DNA; and ultrapure deionized H2O to a final volume of n  23.5 μL. Aliquot the mix in n  0.2 mL thin wall tubes. Each tubes contains 23.5 μL. Add 1.5 μL T4 DNA ligase (400,000 U/mL) to each tube. Prepare a “plasmid only” ligation reaction as described in Subheading 3.1.5. Incubate at 16  C overnight. 2. Purify each ligation reaction using the MinElute PCR purification kit according to the manufacturer’s protocol. 3. Gather all the ligation reactions and homogenize. 4. Measure the concentration spectrophotometrically.

of

the

sample

5. Prepare a fresh stock of electrocompetent DH10β E. coli bacteria. We recommend to prepare electrocompetent bacteria starting from 1 L of LB (or more) according to the desired size of the library. Save one transformation unit to estimate the competency (see Note 16). 6. Perform N transformations by electroporation (We typically do 40 to 50 transformations). Each transformation is performed by electroporating 50 ng of ligation reaction in 50 μL competent DH10β E. coli cells. Each transformation is resuspended in 500 μL of LB medium and incubated for 45 min at 37  C in an orbital shaker at 220 rpm. Pool all tubes together and measure the total volume Vtot (in μL) of bacterial cell suspension.

608

Lina El Hajji et al.

7. For determining the number of transformants, resuspend 10 μL of the final bacterial cell suspension in 90 μL of LB (¼S0 ¼ 10/Vtot of the initial suspension). Dilute S0 by a factor of 10 to 100,000 (S1 to S5) in a similar fashion. Spread 90 μL of each serial dilutions (S0 to S5) on regular LB-agar plates supplemented with 0.1 mg/mL ampicillin. Incubate plates at 37  C overnight. The next day, count bacterial colonies for estimating the number of transformants (see Note 17). 8. Plate the bacterial suspension on N/2 140  140 mm LB-agar plates supplemented with 0.1 mg/mL ampicillin. Plate 1 mL of bacterial suspension per plate. Incubate plates at 37  C overnight. 9. The following day, scrape the bacterial cell lawn with a T-shape cell scraper after adding 2  1 mL of LB on each plate. Gather all collected cells in a 50 mL screw-capped plastic centrifuge tube. 10. Save 10–15 mL of cell suspension to extract the DNA plasmid using the HISpeed Plasmid Maxi kit according to the manufacturer’s protocol. 11. Add 10% (v/v) glycerol to the remaining cell suspension, and make aliquots in 1.5 mL cryogenic vials. Store at 80  C. 3.1.8 Construction of the Large-Scale Library in Yeast

1. Inoculate one large S. cerevisiae EBY100 yeast colony from a fresh plate in a 1 L Erlenmeyer’s flask containing 100 mL YPAD 2 medium. Grow the yeast culture overnight at 30  C in an orbital shaker at 250 rpm. Meanwhile, warm 650 mL of YPAD 2 overnight at 30  C for the next day. 2. The next day, measure the concentration of the yeast culture spectrophotometrically. We typically find an OD600nm around 10 to 13 for overnight cultures, corresponding to ~108 cells/ mL (see Note 18). 3. Centrifuge a volume of yeast culture containing 3  109 cells at 2500  g for 5 min. Discard the supernatant and resuspend yeast cells in 20 mL prewarmed YPAD 2 (see Note 19). 4. Dispense the 20 mL of resuspended yeast cells in 600 mL of prewarmed YPAD 2. The final yeast cell suspension should contain 5  106 cells/mL (corresponding to an OD600nm ¼ 0.5). 5. Grow the yeast culture at 30  C in an orbital shaker at 250 rpm until OD600nm reaches 2, corresponding to a concentration of 2  107 cells/mL. This step can take between 4 to 6 h (see Note 20). 6. Meanwhile, denature 7 mL of a 2 mg/mL ss carrierDNA solution in boiling water for 5 min and chill on ice (see Note 21). Prepare the mix for 120 transformations in a 50 mL

Engineering Fluorescence-Activating Proteins

609

screw-capped plastic centrifuge tube on ice as follows: 28.8 mL PEG 50% (v/v), 4.32 mL Lithium acetate (LiAc) 1 M, 6 mL ss carrierDNA, x μL (~100–500 μg) of plasmid, sterile water to a final volume of 43.2 mL. 7. Pour the yeast culture into twelve 50 mL screw-capped plastic centrifuge tubes. Centrifuge at 2500  g for 5 min at 20  C. 8. Discard the supernatant and resuspend the yeast cell pellet of each tube in 25 mL of sterile water by vortexing. Centrifuge at 2500  g for 5 min at 20  C. 9. Discard the supernatant and resuspend the yeast cell pellet of each tube in 10 mL of 100 mM lithium acetate (LiAc) by vortexing. Combine the yeast cell suspensions in three 50 mL screw-capped plastic centrifuge tubes (3  40 mL). Centrifuge at 2500  g for 5 min at 20  C and discard the supernatant, taking care to remove the residual supernatant. 10. Add 14.4 mL of the transformation mix in each tube. Mix and vortex vigorously for ~1 min. Keep on ice during the transfer. 11. Incubate at 42  C for 45 min. We recommend to use a clean water bath or an incubator without shaking at 42  C. Mix the samples by inverting the tubes every 10 min. 12. Centrifuge the tubes at 2500  g for 5 min at 20  C. Discard the supernatant. 13. For each tube, gently resuspend the yeast cell pellet in 5 mL of sterile water by pipetting up and down. Mix the 3 tubes together (total volume ¼ 15 mL). 14. For determining the number of transformants, resuspend 15 μL of the total yeast cell suspension in 135 μL sterile water (S0 ¼ 1/1000 of the initial suspension). Dilute S0 by a factor 10 to 105 (S1 to S5) in a similar fashion. Spread 135 μL of each dilution (S0 to S5) on regular SDDO-agar plates (see Note 22). Incubate the plates at 30  C for 2–3 days. Count the number of yeast colonies for estimating the number of transformants (see Note 23). The final library size is given by: Library size ¼ Min ð# of transformants of the bacterial library; # of transformants of the yeast libraryÞ

15. To grow homogeneously the yeast library, plate the yeast cell suspension on 30 large 140  140 mm SDDO-agar plates. Plate 0.5 mL of yeast cell suspension per plate. Incubate the plates at 30  C for 2–3 days. We recommend to place a container filled with sterile water in the incubator to avoid plate drying. 16. Scrape the yeast cell lawn with a T-shape cell scraper adding 2  1 mL of SDDO on each plate. Gather the whole yeast cell suspension (about 60 mL) in a 250 mL sterile bottle and

610

Lina El Hajji et al.

homogenize. Add 20 mL of sterile glycerol to reach 30% (v/v) and aliquot in 1.5 mL cryogenic vials (~80  1 mL). After amplification, each aliquot should contain 109 to 1010 cells. Store the aliquots at 80  C. 17. Use one tube to quantify the cell viability after freezing. Resuspend 1 μL of the initial aliquot in 99 μL sterile water (S0 corresponding to 1/1000 of the initial suspension). Dilute S0 by a factor of 10 to 10n (S1 to Sn) in a similar fashion and spread the 100 μL of S0 to Sn on regular SDDO-agar plates. Incubate the plates at 30  C for 2–3 days prior to counting the number of colonies (see Note 24). 3.2 Library Expression in Yeast 3.2.1 Amplification and Induction

1. Resuspend 1 mL (about 1010 cells) of the yeast library (stored at 80  C) in 1 L SDDO medium supplemented with 1% (v/v) of penicillin/streptomycin. Homogenize and split the culture in four 1 L Erlenmeyer’s flasks (each flask contains 250 mL of culture). Grow the yeast culture overnight at 30  C in an orbital shaker at 250 rpm. 2. The next day, measure the concentration of the culture spectrophotometrically. We typically find an OD600nm around 5 to 10 for overnight cultures, corresponding to ~5–10  107 cells/mL (see Note 18). 3. Centrifuge a volume of yeast culture containing 1010 cells at 2500  g for 5 min. Discard the supernatant and resuspend the yeast cells in 40 mL SDDO. 4. Add the yeast suspension to 1 L SDDO medium supplemented with 1% (v/v) of penicillin/streptomycin. At this step, the OD600nm of the yeast suspension is 1. Grow the yeast culture at 30  C in an orbital shaker at 250 rpm until reaching OD600nm between 2 to 5, corresponding to a concentration of 2–5  107 cells/mL. This step can take between 3 to 5 h. 5. Measure the concentration of the yeast culture spectrophotometrically. Centrifuge a volume of yeast culture containing 5  109 cells at 2500  g for 5 min. Discard the supernatant and resuspend the yeast cells in 40 mL SGDO (see Notes 22 and 25). 6. Add the yeast suspension to 1 L SGDO medium supplemented with 1% (v/v) of penicillin/streptomycin. At this step the OD600nm of the yeast suspension is 0.5. Grow the yeast culture at 23  C in an orbital shaker at 250 rpm for 36 h (see Note 25).

3.2.2 Immunolabeling

1. The day of sorting, measure the concentration of the yeast culture spectrophotometrically. For a library of 106 to 107 clones, we usually pellet 5  108 cells for one round of selection (see Note 26). Transfer the corresponding volume of yeast culture into a 50 mL screw-capped plastic centrifuge tube.

Engineering Fluorescence-Activating Proteins

611

2. Centrifuge at 2500  g for 5 min. Discard the supernatant and resuspend the cell pellet in 10 mL of PBS-BSA (see Note 27). 3. Centrifuge again at 2500  g for 5 min. Discard the supernatant and incubate yeast cells with the chicken primary antibody anti-c-myc diluted at 1/250 (4 μg/mL) in PBS for 30 min at room temperature on a rotative tube holder. We usually use 40 μL per 108 cells. 4. Centrifuge at 2500  g for 5 min. Discard the supernatant and resuspend the cell pellet in 10 mL of PBS-BSA (see Note 27). 5. Centrifuge at 2500  g for 5 min. Discard the supernatant and incubate yeast cells with the goat anti-chicken fluorescently labeled secondary antibody diluted at 1/100 (20 μg/mL) in PBS for 20 min at 4  C on a rotative tube holder. We usually use 40 μL per 108 cells. The secondary antibody must be labeled with a fluorophore spectrally orthogonal to the fluorescenceactivating protein you want to select. We usually use secondary antibody labeled with Alexa Fluor® 488 or Alexa Fluor® 647. 6. Centrifuge at 2500  g for 5 min at 4  C. Discard the supernatant and resuspend the cell pellet in 10 mL of PBS-BSA (see Note 27). 7. Centrifuge at 2500  g for 5 min at 4  C. Discard the supernatant and resuspend the cell pellet in n mL of PBS to reach a final concentration of 108 cells/mL. For a selection, we typically resuspend 5  108 cells in 5 mL of PBS. 8. Filter the yeast cell suspension through a 40 μm cell strainer to avoid sorting aggregated cells. 9. Add the appropriate volume of fluorogen to reach the desired concentration just prior to FACS sorting (see Note 28). We recommend not to exceed 0.1% (v/v) DMSO for solubilizing the fluorogen. 3.2.3 Sorting and Iterative Rounds of Flow Cytometry

1. Before analysis, set up a flow cytometry protocol using positive and negative controls to select for adequate laser excitation and fluorescence detector parameters (see Note 26). The yeast population is normally gated on forward- and side-scatter channels to remove debris and aggregated cells. Draw a sorting gate in the double-positive quadrant to isolate yeast cells that are doubly fluorescent because of the anti-myc labeling and the fluorogen labeling. In the first round of sorting, we typically collect 5% of the population of yeast cells to avoid loss of unique clones. As the population becomes enriched, narrower sort windows can be drawn to collect down to 0.1–0.5%. 2. Collect the selected yeast cells in a 50 mL screw-capped plastic centrifuge tube containing 10 mL of SDDO. For the first round, we typically analyze 5  108 cells (see Note 29).

612

Lina El Hajji et al.

3. Pour the collected yeast cell suspension (x mL) into (250  x) mL SDDO medium supplemented with 1% (v/v) of penicillin/ streptomycin. Grow the yeast culture overnight at 30  C in an orbital shaker at 250 rpm. 4. The next day, measure the concentration of the yeast culture spectrophotometrically. 5. The yeast population is further amplified on solid media for homogeneous growth. Transfer a volume of yeast culture containing 109 cells in a 50 mL screw-capped plastic centrifuge tube. Centrifuge at 2500  g for 5 min. Discard the supernatant, resuspend the cell pellet in 5 mL of SDDO medium and spread the yeast cell suspension on two large 140  140 mm SDDO-agar plates. Incubate the plates at 30  C for 2–3 days (see Note 30). Scrape the cell lawn with a T-shape cell scraper after adding 5 mL of SDDO medium (or sterile water) on each plate. Gather all collected cells in a 50 mL screw-capped plastic centrifuge tube (10 mL). 6. For subsequent round of flow cytometry sorting, grow 500 μL of the yeast cell suspension in 500 mL SDDO medium. Store the remaining yeast cell suspension in 30% (v/v) sterile glycerol and make 1 mL aliquots in 1.5 mL cryogenic vials. Store the aliquots at 80  C. 7. Repeat steps 1–6 of Subheading 3.2.1 and steps 1–9 of Subheading 3.2.2 prior to the next sorting round (see Notes 31 and 32). 3.3 Screening of Clones After FACS Rounds 3.3.1 Fluorescence Analysis of Selected Clones

1. Once satisfied with the sorted population, dilute a small amount of the yeast cell suspension from step 4 (Subheading 3.2.3) to plate on SDDO agar plates to isolate individual clones (similarly to what was done for colony counting). Incubate the plates at 30  C for 2–3 days. 2. Pick 48 yeast colonies and inoculate each colony in 5 mL of SDDO. Grow yeast cultures overnight at 30  C in an orbital shaker at 250 rpm. 3. The following day, measure the concentration of the yeast cultures spectrophotometrically (see Note 33). 4. For each clone, centrifuge a volume of yeast culture containing 5  107 cells at 2500  g for 5 min. Discard the supernatant and resuspend the cell pellet in 5 mL SDDO. At this step the OD600nm of the yeast suspension is 1. 5. Grow the yeast cultures at 30  C in an orbital shaker at 250 rpm until OD600nm reaches 2 to 5, corresponding to a concentration of 2–5  107 cells/mL. This step can take between 3 and 5 h.

Engineering Fluorescence-Activating Proteins

613

6. Measure the concentration of the yeast cultures spectrophotometrically. Centrifuge a volume of yeast culture containing 2.5  107 cells at 2500  g for 5 min. Discard the supernatant and resuspend the cell pellet in 5 mL galactose-containing SGDO. At this step the OD600nm of the suspension is 0.5. 7. Grow yeast cultures at 23  C in an orbital shaker at 250 rpm for 36 h. 8. The day of fluorescence measurements, measure the concentration of the yeast cultures. Collect 2  108 cells in 2 mL microcentrifuge tubes. 9. Centrifuge at 2500  g for 5 min. Discard the supernatant and resuspend the cell pellet in 2 mL of PBS-BSA. 10. Centrifuge at 2500  g for 5 min. Discard the supernatants and incubate the cell pellet with the chicken anti-c-myc primary antibody diluted at 1/250 (4 μg/mL) in PBS for 30 min at room temperature on a rotative tube holder. We usually use 40 μL per 108 cells. 11. Centrifuge at 2500  g for 5 min. Discard the supernatant and resuspend the cell pellet in 2 mL of PBS-BSA. 12. Centrifuge at 2500  g for 5 min. Discard the supernatants and incubate the cell pellet with the goat anti-chicken fluorescently labeled secondary antibody diluted at 1/100 (20 μg/mL) in PBS for 20 min at 4  C on a rotative tube holder. We usually use 40 μL per 108 cells. The secondary antibody must be labeled with a fluorophore spectrally orthogonal to the fluorescence-activating protein you want to select. We usually use secondary antibody labeled with Alexa Fluor® 488 or Alexa Fluor® 647. 13. Centrifuge at 2500  g for 5 min at 4  C. Discard the supernatants and resuspend the cell pellets in 2 mL of PBS-BSA. 14. Centrifuge at 2500  g for 5 min at 4  C. Discard the supernatants and resuspend the cell pellets in 2 mL of PBS to reach a final concentration of 108 cells/mL. 15. Prepare a fluorogen solution in PBS at the appropriate (2) concentration. 16. The fluorescence measurements are performed in triplicate using a fluorescence plate reader. Fill each well with 100 μL of the fluorogen solution prepared in step 15 (see Note 34). 17. For each clone, add 100 μL of immunolabeled yeast cell suspension prepared in step 14 per well. Prepare controls with yeast cells only, fluorogen only, and yeast cells expressing the parental clone
fluorogen. 18. For each well, measure the fluorescence (1) using the excitation/emission settings for the protein:fluorogen assembly, and (2) using the excitation/emission settings for the fluorescent

614

Lina El Hajji et al.

antibody. The fluorescence signal of the protein:fluorogen assembly is then divided by the fluorescence signal of the fluorescent antibody to normalize the fluorescence signal by the expression level. 19. Plot the fluorescence of each clone as a ratio of fluorescence increase in comparison with the parental clone. 3.3.2 DNA Analysis of the Selected Clones

1. Use the remaining volume of each yeast culture of step 4 (Subheading 3.3.1) to extract the plasmid DNA. 2. Centrifuge the yeast cell suspensions at 2500  g at room temperature for 5 min in 15 mL screw-capped plastic centrifuge tube. 3. Discard the supernatant and resuspend the cell pellets in water in 1.5 mL microcentrifuge tubes. 4. Centrifuge the cell suspensions at 2500  g at room temperature for 5 min. Discard the supernatants and resuspend each cell pellet in 200 μL of P1 buffer (from QIAprep Spin Miniprep kit). 5. To each tube, add 100 μL of lyticase buffer. Mix thoroughly and incubate at 37  C for at least 30 min (see Note 35). 6. Add 300 μL of P2 buffer (from QIAprep Spin Miniprep kit) to each tube and invert the tubes 4–6 gently to mix. Incubate at room temperature for 10 min. 7. Add 420 μL of N3 buffer (from QIAprep Spin Miniprep kit) to each tube and invert immediately to mix. 8. Centrifuge for 10 min at 10,000  g. 9. Apply the supernatants to Qiaprep spin columns according to the manufacturer’s protocol. 10. Measure the DNA concentrations spectrophotometrically. 11. For each yeast miniprep, transform 3 μL in 50 μL aliquots DH10β E. coli electrocompetent cells by electroporation. After electroporation, add 500 μL of LB medium and incubate the bacterial suspensions for 45 min at 37  C at 220 rpm. 12. Spot 20 μL of each transformation on a large ampicillin LB-agar plate and incubate at 37  C overnight. 13. For each clone, scrape cells and inoculate 2 mL LB supplemented with 0.1 mg/mL ampicillin. Grow bacterial cultures at 37  C overnight for subsequent extraction of DNA plasmids. 14. Purify the plasmids using QIAprep Spin Miniprep kit according to the manufacturer’s protocol. Measure the concentrations of the samples spectrophotometrically. 15. Sequence with adequate primers (M13-Forward primer or custom-made ag761) (see Note 13).

Engineering Fluorescence-Activating Proteins

3.4 Characterization of the Selected Clones 3.4.1 Cloning into pET28a

615

1. Subclone the DNA coding for the selected clones into the pET28a for bacterial expression. The described procedure will enable to introduce an N-terminal His-tag. 2. Digest 3 μg of pET28a plasmid with NheI-HF and XhoI-HF restriction enzymes for 2.5 h at 37  C in 1 CutSmart buffer. 3. Add 5 μL of calf intestinal alkaline phosphatase (CIP) to prevent vector recircularization. Incubate at 37  C for 2.5 h (see Note 36). 4. Purify using QIAquick PCR Purification kit according to the manufacturer’s protocol. 5. Similarly, digest 1.5 μg of isolated pCTCON2 plasmids containing the sequence of the selected clones with NheI-HF and XhoI-HF restriction enzymes for 2.5 h at 37  C in 1 CutSmart buffer. Purify the inserts by gel extraction using the QIAquick gel extraction kit according to the manufacturer’s protocol. 6. Measure the concentration spectrophotometrically.

of

the

samples

7. Store the purified DNA at 20  C, if necessary. 8. Prepare ligation reactions as follows: 1 ligation buffer; 100 ng of digested plasmid DNA; 1.5 μL T4 DNA ligase (400,000 U/ mL); with or without the digested insert DNA (25 ng); and ultrapure deionized water to a final volume of 25 μL. 9. Incubate the ligation reactions at 16  C overnight. 10. Purify the ligation reactions using the Qiagen MinElute PCR Purification kit according to the manufacturer’s protocol. 11. Measure the concentration spectrophotometrically.

of

the

samples

12. For each ligation, electroporate 50 ng of ligation reaction into 50 μL of electrocompetent DH10β E. coli cells. After electroporation, add 500 μL of LB medium and incubate the bacterial suspensions for 45 min at 37  C in an orbital shaker at 220 rpm. 13. Spread 50 μL of each bacterial suspension on LB-agar plates supplemented with 0.05 mg/mL kanamycin and incubate at 37  C overnight. 14. Randomly pick 3 colonies from each plate and inoculate each colony in 2 mL LB supplemented with 0.05 mg/mL kanamycin. Grow bacterial cultures at 37  C overnight in an orbital shaker at 220 rpm for subsequent extraction of DNA plasmids. 15. Purify the plasmids using QIAprep Spin Miniprep kit according to the manufacturer’s protocol. Measure the concentrations of the samples spectrophotometrically.

616

Lina El Hajji et al.

16. Run a test digestion to check for correct insert. Sequence with adequate primers (T7-promoter primer) (see Note 13). 3.4.2 Protein Production and Purification

1. Electroporate 50 ng of the generated pET28a plasmid in 50 μL of electrocompetent Rosetta (DE3) pLys E. coli cells. After electroporation, add 500 μL of LB medium and incubate the bacterial suspension for 45 min at 37  C in an orbital shaker at 220 rpm. 2. Spread 50 μL of the bacterial suspension on LB-agar plates supplemented with 0.05 mg/mL kanamycin and 0.034 mg/ mL chloramphenicol and incubate at 37  C overnight. 3. Pick one colony and inoculate 5 mL LB supplemented with 0.05 mg/mL kanamycin and 0.034 mg/mL chloramphenicol. Grow bacterial culture at 37  C overnight in an orbital shaker at 220 rpm. 4. The following day, add the saturated bacterial suspension to 500 mL of LB supplemented with 0.05 mg/mL kanamycin and 0.034 mg/mL chloramphenicol. 5. Grow bacterial culture at 37  C in an orbital shaker at 220 rpm until OD600nm reaches 0.6. Decrease temperature to 16  C. Induce protein production by adding β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 1 mM (see Note 37). Incubate the bacterial suspension at 16  C in an orbital shaker at 220 rpm overnight. 6. Harvest bacterial cells by centrifuging at 6000  g, 4  C for 20 min. Freeze the cell pellets at 20  C if necessary. 7. On ice, softly resuspend the cell pellets in 5 mL of lysis buffer per 500 mL of bacterial culture with a pipette. 8. Sonicate in 50 mL screw-capped plastic centrifuge tubes for 5 min: 20% amplitude; cycles of 3 s of pulsation and 1 s pause. Successful sonication can be assessed by a viscosity change of the liquid medium (see Note 38). 9. Incubate on ice for 2 h for DNA digestion on a shaking platform. 10. Centrifuge at 9300  g for 1 h at 4  C and recover the supernatants. 11. In the meantime, pellet 1 mL of Ni-NTA beads suspension in EtOH (per 500 mL of culture) and wash three times with water and three times with equilibration buffer (resuspend the beads in 0.75 mL of liquid each time, centrifuge at 2000  g, 4  C for 3 min, discard the supernatant and repeat) (see Note 39). 12. Incubate the supernatants from step 11 with the prepared Ni-NTA beads in a 50 mL screw-capped plastic centrifuge overnight at 4  C on a rotative tube holder.

Engineering Fluorescence-Activating Proteins

617

13. Wash the purification columns with 10 mL of water and 10 mL of equilibration buffer. 14. Gently mix the supernatant and beads mix and pack the column with it. Keep the flow-through. 15. Rinse the tube which contained the Ni-NTA beads with 10 mL of wash 1 buffer and pour into the column. 16. Wash the column with 10 mL of wash 1 buffer (keep the solution running out of the column). 17. Repeat step 16, but with 10 mL of wash 2 buffer (keep the solution running out of the column). 18. Elute the protein with 10 mL of elution buffer, and collect as 1 mL fractions. 19. Measure the protein spectrophotometrically.

concentration

in

the

fractions

20. Load the protein fractions on the adequate columns for buffer exchange, according to the manufacturer’s instructions (see Note 40). 21. Elute the proteins using PBS buffer in 500 μL fractions. 22. Measure the proteins concentration in the fractions and store at 30  C if necessary. 23. Run a SDS-PAGE to assess the protein size and purity. 3.4.3 Determination of the Dissociation Constant of the Tag:Fluorogens Assembly

1. The dissociation constant KD of the fluorogen:protein assembly is determined by titration experiment in a multi-well format using a fluorescence plate reader. As the fluorogen:protein assembly is strongly fluorescent, one can directly use fluorescence as a readout to determine the fraction of complex. Titrations are performed varying the fluorogen concentration while keeping the protein concentration constant. Measurement of the fluorescence intensity at each fluorogen concentration enables to quantify the fraction of complex. For simple analysis, these experiments are usually done in large molar excess of fluorogen to be able to neglect the variation of free fluorogen concentration upon complex formation. We typically start with [protein]0 ¼ 50 nM, and eventually lower the concentration for high affinity systems. To determine the specific signal from the complex, the contribution of the free fluorogen (baseline) should be measured and subtracted. We usually use twelve different concentrations of fluorogen (concentrations should be chosen in order to enable proper determination of the KD). One titration occupies thus one row of a 96-well plate. Measurements are done in triplicate. For one KD determination, one needs six rows of a 96-well plate, three rows for the actual titration and three rows for baseline determination. The total solution volume is set to 120 μL. For the titration, we use

618

Lina El Hajji et al.

60 μL of 2 fluorogen solution and 60 μL of 2 protein solution. For the baseline determination, we use 60 μL of 2 fluorogen solution completed with 60 μL of buffer. 2. For one KD determination, prepare a 2 protein stock solution (2.2 mL) and twelve 2 fluorogen stock solutions at different concentrations (0.4 mL) (see Note 41). Fill the plates in this order: buffer, then fluorogen solutions, then protein solution (see Note 42). 3. Acquire the fluorescence intensity of each well. Acquisition parameters should be optimized. Use the maximal absorption wavelength of the fluorescent protein:fluorogen assembly as excitation wavelength, and collect emission in a spectral window centered around the maximal emission wavelength of the fluorescent assembly. 4. Analyze the fluorescence titration data by fitting with a one-site specific binding model using Graphpad Prism software (see Note 43). 5. If necessary, after a rough determination of the dissociation constant, repeat the experiment adjusting the range of fluorogen concentrations. 3.4.4 Determination of the Fluorescence Quantum Yield of the Tag:Fluorogens Assembly

1. Using the dissociation constant previously determined, prepare a solution of protein and fluorogen with an excess of protein to obtain >95% bound fluorogen in order to measure absorbance and emission spectra resulting only from the protein:fluorogen assembly (see Note 44). Fluorogen concentration should be chosen so that maximal absorbance of the solution remains below 0.05 to minimize inner filter effect. For FAST:fluorogen assembly, we usually start with a solution containing 40 μM protein and 6 μM fluorogen (160 μL), when using 0.3  0.3 cm cuvettes. Prepare also 1 mL solution containing only the protein at the chosen concentration to perform reciprocal dilutions (see Note 45). 2. Preliminary measurements should be done to assess the best excitation wavelength, adjust the reading window, integration time, and band pass. Once set, these parameters should remain constant between all acquired data during an experiment. 3. The quantum yield is determined relatively to a reference of known fluorescence quantum yield. Choose a reference with spectral properties similar to your fluorescent protein:fluorogen assembly. 4. Acquire the absorbance spectrum of the protein solution without fluorogen. This spectrum is ultimately subtracted to the spectra of the different complex solutions to remove the contribution of the protein itself.

Engineering Fluorescence-Activating Proteins

619

5. Acquire the absorbance and emission spectra of the solution containing the protein:fluorogen assembly prepared in step 1. 6. From the cuvette from step 5, remove half the total volume and replace with the same volume of fluorogen-free protein solution. This way, the protein concentration remains constant, and the fluorogen concentration (and thus the protein:fluorogen assembly concentration) is divided by two. Acquire the absorbance and emission spectra. Repeat this step until you reach a fluorogen dilution factor of 1/16. 7. Without changing the acquisition parameters, acquire the absorbance and emission spectra of the reference solution. Perform reciprocal dilution as in step 6. 8. Data analysis: (a) Absorbance spectra: correct the absorbance spectra by subtracting the absorbance spectrum of the protein alone. Report the corrected absorbance A at the excitation wavelength for each acquired spectrum in a table. (b) Fluorescence spectra: compute the integrated area under the emission curves. Report the integrated fluorescence intensities I in the previously created table. (c) Plot the integrated fluorescence intensities I in function of the absorbance A. Fit the obtained graph with a linear regression intercepting the origin, and report the I/A slope value. (d) Perform the same analysis for the reference to determine the Iref/Aref slope value. (e) The fluorescence quantum yield ϕ is obtained with the simplified formula: 

I A ref ϕref ϕ¼ I ref A where ϕref is the fluorescence quantum yield of the reference.

4

Notes 1. We usually used partially randomized primers with degenerate NNK codons. N ¼ C, T, A or G and K ¼ C or T. 2. Initial amount of template is crucial in error-prone PCR as mutation frequency is dependent on DNA polymerase error rate and number of duplications. If the amount of template is high, it will undergo few duplications in the epPCR. On the other hand, a low amount of template will result in a greater number of duplications and more mutations will be

620

Lina El Hajji et al.

introduced. We recommend to test various template amounts (typically 0.1 ng, 1 ng and 10 ng). Note that the amount of template corresponds to the amount of target DNA instead of the amount of purified plasmid DNA. For example, for a plasmid of 5700 bp in total, 10 ng of target DNA (of 375 bp) corresponds to 152 ng of total plasmid DNA. 3. In family (DNA) shuffling procedure the parental DNA fragments are first amplified, fragmented, and randomly recombined to allow for random recombination based on homology sequence. Hence, sequences with more than 70% of homology sequence should be used for proper recombination. 4. For family shuffling procedure, template DNAs should allow proper amplification of the PCR products without introducing additional mutation as this method produces a point mutagenesis rate of 0.7% [18]. Choose in priority high-fidelity DNA polymerase and initial amount of target DNA allowing proper amplification of the PCR products (typically 50–100 ng of plasmid DNA). Note that error-prone PCR conditions could be performed for additional mutation rate at this step (see Note 2). 5. PCR products are first mixed together in equal amounts (5 μg in total) and then digested by DNase I (20 U/mL). For example, for five different templates, mix 1 μg of each PCR products for DNase I digestion. 6. For PCR products of less than or equal to 500 bp, PCR fragments can be isolated in a 2% agarose gel. A smear should be observed. Fragments between 50 and 250 bp can be extracted and purified using PCR QIAquick Gel Extraction kit according to the manufacturer’s protocol. The excised size bands and agarose percentage can be adjusted for further optimization. 7. The reassembly temperature can be adjusted according to the fragment sizes for further reassembly optimization. 8. For error-prone PCR conditions, digest the different PCR products initially obtained using various template amounts (typically 0.1 ng, 1 ng and 10 ng). 9. Ligate the plasmid and insert by varying the plasmid:insert molar ratio for further ligation optimization (typically 1:1 to 1:10 ratio). 10. Ideally, DH10β electrocompetent cells should be freshly prepared. To prepare electrocompetent cells, we typically use 1 L of bacterial suspension in mid-exponential phase growth (OD600nm ¼ 0.6). Successive cycles are performed as following: bacterial cells are harvested, centrifuged (6000  g, 4  C, 15 min) and resuspended in 1 L of sterile water, in 500 mL of sterile water and in 50 mL of 10% (v/v) sterile glycerol. In the

Engineering Fluorescence-Activating Proteins

621

final stage, the cells are resuspended in 2 mL of 10% (v/v) sterile glycerol (3 mL in total) and stored in 60  50 μL aliquots in 1.5 mL microcentrifuge tubes. Each aliquot should contain ~109 cells. We typically use one aliquot for one transformation. Store the aliquots at 80  C. 11. To estimate the number of transformants per electroporation of 50 ng of ligation product, count the number of bacterial colonies on each plate Si (i ¼ 0 to 5): Number of transformants ¼

550 50   number of colonies on plate Si ði¼05Þ  10i 5 45

12. Before scaling up to a full-sized library, we find useful to construct a library test. This allows to validate the mutagenesis and ligation protocols, to estimate the number of transformants for anticipating the final library size, and to analyze the mutagenesis quality. 13. For sequencing, we typically use the following primers: Custom made ag761 primer

50 -AAG GAC AAT AGC TCG ACG ATT GAA GG-30

Commonly used M13-Forward primer

50 -GTA AAA CGA CGG CCA GT-30

Commonly used T7 promoter 50 -TAA TAC GAC TCA CTA TAG GG-30 primer

14. For error-prone PCR conditions: we usually seek around 7–8 nucleotide mutations and 3–5 amino acid mutations. Choose the adequate amount of DNA template accordingly. 15. For family shuffling procedure, we seek to find the appropriate recombination of the different fragments. Note that additional mutations, insertions or deletions can be observed with shuffling procedure. The protocol introduced below was optimized for five different fragments of 375 bp each. Therefore, the amount of templates, amount of DNase I and incubation time, percentage of agarose gel, and size of excised bands (or smear) can be adjusted as necessary. For further optimization, we recommend to consult detailed DNA shuffling protocols [32–34]. 16. To estimate the competency of freshly prepared electrocompetent DH10β cells: count the number of bacterial colonies on each plate Si (i ¼ 0 to n) (as in Note 11) to have the number of transformants per electroporation per ng of control plasmid. We typically find a transformation efficiency of 5  105 transformants/ng of control plasmid.

622

Lina El Hajji et al.

17. The regular LB-agar plates of the dilutions and the “vector only” control should be used to estimate the scale-up size of the bacterial library and to verify that the control background is