Animal Cell Biotechnology: Methods and Protocols [4th ed. 2020] 978-1-0716-0190-7, 978-1-0716-0191-4

Animal Cell Biotechnology: Methods and Protocols, Fourth Edition constitutes a comprehensive manual of state-of-the-art

1,519 142 11MB

English Pages XIV, 399 [396] Year 2020

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Animal Cell Biotechnology: Methods and Protocols [4th ed. 2020]
 978-1-0716-0190-7, 978-1-0716-0191-4

Table of contents :
Front Matter ....Pages i-xiv
Front Matter ....Pages 1-1
Near-Physiological Cell Cycle Synchronization with Countercurrent Centrifugal Elutriation (Johannes Möller, Katrin Korte, Ralf Pörtner, An-Ping Zeng, Uwe Jandt)....Pages 3-16
Turning Cells into Reagents: The Preparation of Assay Ready Cells for Routine Use in Bioassays (Oliver Wehmeier, Alexander Loa)....Pages 17-25
Screening of Media Supplements for High-Performance Perfusion Cultures by Design of Experiment (Patrick Mayrhofer, Renate Kunert)....Pages 27-39
Front Matter ....Pages 41-41
ambr® 15 Microbioreactors for CHO Cell Cultivation (Steve R. C. Warr)....Pages 43-67
Using a Parallel Micro-Cultivation System (Micro-Matrix) as a Process Development Tool for Cell Culture Applications (Vincent Wiegmann, Cristina Bernal Martinez, Frank Baganz)....Pages 69-81
HEK293 Cell-Based Bioprocess Development at Bench Scale by Means of Online Monitoring in Shake Flasks (RAMOS and SFR) (Tibor Anderlei, Michael V. Keebler, Jordi Joan Cairó, Martí Lecina)....Pages 83-103
Orbitally Shaken Single-Use Bioreactor for Animal Cell Cultivation: Fed-Batch and Perfusion Mode (Tim Bürgin, Juliana Coronel, Gerrit Hagens, Michael V. Keebler, Yvonne Genzel, Udo Reichl et al.)....Pages 105-123
Development of Mammalian Cell Perfusion Cultures at Lab Scale: From Orbitally Shaken Tubes to Benchtop Bioreactors (Moritz Wolf, Massimo Morbidelli)....Pages 125-140
Perfusion Control for High Cell Density Cultivation and Viral Vaccine Production (Alexander Nikolay, Thomas Bissinger, Gwendal Gränicher, Yixiao Wu, Yvonne Genzel, Udo Reichl)....Pages 141-168
How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200 L Scale (Cedric Schirmer, Jan Müller, Nina Steffen, Sören Werner, Regine Eibl, Dieter Eibl)....Pages 169-186
Front Matter ....Pages 187-187
Generic Workflow for the Setup of Mechanistic Process Models (Sven Daume, Sandro Kofler, Julian Kager, Paul Kroll, Christoph Herwig)....Pages 189-211
Estimation of Process Model Parameters (Sahar Deppe, Björn Frahm, Volker C. Hass, Tanja Hernández Rodríguez, Kim B. Kuchemüller, Johannes Möller et al.)....Pages 213-234
Efficient Optimization of Process Strategies with Model-Assisted Design of Experiments (Kim B. Kuchemüller, Ralf Pörtner, Johannes Möller)....Pages 235-249
Design, Optimization, and Adaptive Control of Cell Culture Seed Trains (Tanja Hernández Rodríguez, Björn Frahm)....Pages 251-267
Front Matter ....Pages 269-269
High-Throughput Quantification and Glycosylation Analysis of Antibodies Using Bead-Based Assays (Sebastian Giehring)....Pages 271-284
Surface Plasmon Resonance-Based Method for Rapid Product Sialylation Assessment in Cell Culture (Olivier Henry, Eric Karengera, Florian Cambay, Gregory De Crescenzo)....Pages 285-293
Analysis of Product Quality of Complex Polymeric IgM Produced by CHO Cells (Julia Hennicke, Renate Kunert)....Pages 295-302
Raman Trapping Microscopy for Non-invasive Analysis of Biological Samples (Hesham K. Yosef, Karin Schütze)....Pages 303-317
An Optical Biosensor for Continuous Glucose Monitoring in Animal Cell Cultures (Mario Lederle, Mircea Tric, Claudio Packi, Tobias Werner, Philipp Wiedemann)....Pages 319-333
Turbidimetry and Dielectric Spectroscopy as Process Analytical Technologies for Mammalian and Insect Cell Cultures (Lukas Käßer, Jan Zitzmann, Tanja Grein, Tobias Weidner, Denise Salzig, Peter Czermak)....Pages 335-364
Front Matter ....Pages 365-365
Continuous Chromatography Purification of Virus-Based Biopharmaceuticals: A Shortcut Design Method (Ricardo J. S. Silva, João P. Mendes, Manuel J. T. Carrondo, Paula M. Marques, Cristina Peixoto)....Pages 367-384
Single Step Purification of Glycogen Synthase Kinase Isoforms from Small Scale Transient Expression in HEK293 Cells with a Calcium-Dependent Fragment Complementation System (Gavin McGauran, Sara Linse, David J. O’Connell)....Pages 385-396
Back Matter ....Pages 397-399

Citation preview

Methods in Molecular Biology 2095

Ralf Pörtner Editor

Animal Cell Biotechnology Methods and Protocols Fourth Edition

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.

Animal Cell Biotechnology Methods and Protocols Fourth Edition

Edited by

Ralf Pörtner Institute of Bioprocess and Biosystems Engineering, Hamburg University of Technology (TUHH), Hamburg, Germany

Editor Ralf Po¨rtner Institute of Bioprocess and Biosystems Engineering Hamburg University of Technology (TUHH) Hamburg, Germany

ISSN 1064-3745 ISSN 1940-6029 (electronic) Methods in Molecular Biology ISBN 978-1-0716-0190-7 ISBN 978-1-0716-0191-4 (eBook) https://doi.org/10.1007/978-1-0716-0191-4 © Springer Science+Business Media, LLC, part of Springer Nature 2020 This work is subject to copyright. All rights are reserved 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, express 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: 233 Spring Street, New York, NY 10013, U.S.A.

Preface Mammalian cells are used in industry as well as in research for a variety of applications. Examples are the production of monoclonal antibodies or proteins for diagnostic or therapeutic use, production of viral vaccines as well as cultivation of tissue cells for artificial organs or for gene therapy. Beside the techniques required for establishing specific cell lines, techniques are required for optimized cultivation in small- and large-scale, cell characterization and analysis, purification of biopharmaceuticals and vaccines. Animal Cell Biotechnology: Methods and Protocols, Fourth Edition constitutes a comprehensive manual of state-of-the-art techniques for setting up mammalian cell lines and media for the development of biopharmaceuticals and optimizing critical parameters for cell culture considering the whole cascade from the lab to the final production. Special emphasis was put on model-assisted concepts. Scientists with long-refined expertise describe cuttingedge techniques for the production of therapeutic proteins and vaccines. Capturing the major advances that have occurred in both science and the technology of these biopharmaceuticals, this important book covers the powerful new techniques used in cell line and media development, optimizing process techniques and process strategies, use of modelassisted tools for process design and optimization, and in analysis. Topics include cell line and media development, techniques for process development, model-based techniques for process development, process analysis, and downstream techniques. The volume is divided into five parts that reflect the processes required for different stages of production. In Part I, basic techniques for establishment of near-physiological cell-cycle synchronization and for preparation of assay ready cells for routine use in bioassays are addressed. Part II addresses tools for process development, especially miniaturized bioreactor concepts, perfusion techniques, and application of single-use-bioreactors. Part III covers model-based techniques, especially the workflow for the setup of mechanistic process models, estimation of model parameters, model-assisted Design of Experiments, and seed train optimization. Part IV details analytical techniques, including quantification and glycosylation analysis of antibodies, Raman microscopy, optical sensors and turbidimetry, and dielectric spectroscopy. Part V describes downstream techniques, for example, continuous chromatography purification of virus-based biopharmaceuticals as well as purification of kinase Isoforms from transient expression. In summary, this volume constitutes a comprehensive manual of state of the art and new techniques for setting up mammalian cell lines for production of biopharmaceuticals, and optimizing critical parameters for cell culture considering the whole cascade from the lab to the final production. Inevitably, some omissions will occur in the test, but the authors have sought to avoid duplications by extensive cross-referencing to chapters in other volumes of

v

vi

Preface

this series and elsewhere. We hope the volume provides a useful compendium of techniques for scientists in industrial and research laboratories that use mammalian cells for biotechnology purposes. The editor is grateful for the support of all the contributors, the series editor Prof. John Walker, Hertfordshire, UK, and the publishers who have made this volume possible. Hamburg, Germany

Ralf Po¨rtner

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

PART I

CELL LINE AND MEDIA DEVELOPMENT

1 Near-Physiological Cell Cycle Synchronization with Countercurrent Centrifugal Elutriation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Johannes Mo¨ller, Katrin Korte, Ralf Po¨rtner, An-Ping Zeng, and Uwe Jandt 2 Turning Cells into Reagents: The Preparation of Assay Ready Cells for Routine Use in Bioassays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oliver Wehmeier and Alexander Loa 3 Screening of Media Supplements for High-Performance Perfusion Cultures by Design of Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Patrick Mayrhofer and Renate Kunert

PART II

v xi

3

17

27

TECHNIQUES FOR PROCESS DEVELOPMENT

4 ambr® 15 Microbioreactors for CHO Cell Cultivation. . . . . . . . . . . . . . . . . . . . . . . 43 Steve R. C. Warr 5 Using a Parallel Micro-Cultivation System (Micro-Matrix) as a Process Development Tool for Cell Culture Applications. . . . . . . . . . . . . . . . . 69 Vincent Wiegmann, Cristina Bernal Martinez, and Frank Baganz 6 HEK293 Cell-Based Bioprocess Development at Bench Scale by Means of Online Monitoring in Shake Flasks (RAMOS and SFR) . . . . . . . . . . 83 Tibor Anderlei, Michael V. Keebler, Jordi Joan Cairo, and Martı´ Lecina 7 Orbitally Shaken Single-Use Bioreactor for Animal Cell Cultivation: Fed-Batch and Perfusion Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 ¨ rgin, Juliana Coronel, Gerrit Hagens, Michael V. Keebler, Tim Bu Yvonne Genzel, Udo Reichl, and Tibor Anderlei 8 Development of Mammalian Cell Perfusion Cultures at Lab Scale: From Orbitally Shaken Tubes to Benchtop Bioreactors . . . . . . . . . . . . . . . . . . . . . . 125 Moritz Wolf and Massimo Morbidelli 9 Perfusion Control for High Cell Density Cultivation and Viral Vaccine Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 Alexander Nikolay, Thomas Bissinger, Gwendal Gr€ a nicher, Yixiao Wu, Yvonne Genzel, and Udo Reichl

vii

viii

10

Contents

How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200 L Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 ¨ ller, Nina Steffen, So¨ren Werner, Cedric Schirmer, Jan Mu Regine Eibl, and Dieter Eibl

PART III 11

12

13

14

Generic Workflow for the Setup of Mechanistic Process Models . . . . . . . . . . . . . . Sven Daume, Sandro Kofler, Julian Kager, Paul Kroll, and Christoph Herwig Estimation of Process Model Parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sahar Deppe, Bjo¨rn Frahm, Volker C. Hass, Tanja Herna´ndez Rodrı´guez, ¨ ller, Johannes Mo¨ller, and Ralf Po¨rtner Kim B. Kuchemu Efficient Optimization of Process Strategies with Model-Assisted Design of Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ¨ ller, Ralf Po¨rtner, and Johannes Mo¨ller Kim B. Kuchemu Design, Optimization, and Adaptive Control of Cell Culture Seed Trains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

PART IV 15

16

17

18

19

20

MODEL-BASED TECHNIQUES FOR PROCESS DEVELOPMENT 189

213

235

251

PROCESS ANALYSIS

High-Throughput Quantification and Glycosylation Analysis of Antibodies Using Bead-Based Assays. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sebastian Giehring Surface Plasmon Resonance-Based Method for Rapid Product Sialylation Assessment in Cell Culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Olivier Henry, Eric Karengera, Florian Cambay, and Gregory De Crescenzo Analysis of Product Quality of Complex Polymeric IgM Produced by CHO Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Julia Hennicke and Renate Kunert Raman Trapping Microscopy for Non-invasive Analysis of Biological Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ¨ tze Hesham K. Yosef and Karin Schu An Optical Biosensor for Continuous Glucose Monitoring in Animal Cell Cultures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mario Lederle, Mircea Tric, Claudio Packi, Tobias Werner, and Philipp Wiedemann Turbidimetry and Dielectric Spectroscopy as Process Analytical Technologies for Mammalian and Insect Cell Cultures . . . . . . . . . . . . . . . . . . . . . . Lukas K€ a ßer, Jan Zitzmann, Tanja Grein, Tobias Weidner, Denise Salzig, and Peter Czermak

271

285

295

303

319

335

Contents

PART V 21

22

ix

DOWNSTREAM TECHNIQUES

Continuous Chromatography Purification of Virus-Based Biopharmaceuticals: A Shortcut Design Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367 ˜ o P. Mendes, Manuel J. T. Carrondo, Ricardo J. S. Silva, Joa Paula M. Marques, and Cristina Peixoto Single Step Purification of Glycogen Synthase Kinase Isoforms from Small Scale Transient Expression in HEK293 Cells with a Calcium-Dependent Fragment Complementation System. . . . . . . . . . . . . . 385 Gavin McGauran, Sara Linse, and David J. O’Connell

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

397

Contributors TIBOR ANDERLEI • Adolf Ku¨hner AG, Birsfelden, Switzerland FRANK BAGANZ • Department of Biochemical Engineering, The Advanced Centre for Biochemical Engineering, University College London, London, UK THOMAS BISSINGER • Bioprocess Engineering, Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany TIM BU¨RGIN • Adolf Ku¨hner AG, Birsfelden, Switzerland JORDI JOAN CAIRO´ • Department of Chemical, Biological and Environmental Engineering, Universitat Auto`noma de Barcelona, Cerdanyola del Valle`s, Spain FLORIAN CAMBAY • Department of Chemical Engineering, Polytechnique Montre´al, Montre´al, QC, Canada MANUEL J. T. CARRONDO • Instituto de Biologia Experimental e Tecnologica, Oeiras, Portugal JULIANA CORONEL • Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany PETER CZERMAK • Institute of Bioprocess Engineering and Pharmaceutical Technology (IBPT), Technische Hochschule Mittelhessen (THM)—University of Applied Sciences, Giessen, Germany; Faculty of Biology and Chemistry, Justus-Liebig-University Giessen, Giessen, Germany; Division Bioresources, Fraunhofer Institute for Molecular Biology and Applied Ecology (IME), Giessen, Germany SVEN DAUME • Institute of Chemical, Environmental and Bioscience Engineering, TU Wien, Vienna, Austria GREGORY DE CRESCENZO • Department of Chemical Engineering, Polytechnique Montre´al, Montre´al, QC, Canada SAHAR DEPPE • Biotechnology & Bioprocess Engineering, Ostwestfalen-Lippe University of Applied Sciences and Arts, Lemgo, Germany DIETER EIBL • School of Life Sciences and Facility Management, Institute of Chemistry and Biotechnology, Zurich University of Applied Sciences, W€ a denswil, Switzerland REGINE EIBL • School of Life Sciences and Facility Management, Institute of Chemistry and Biotechnology, Zurich University of Applied Sciences, W€ a denswil, Switzerland ¨ BJORN FRAHM • Biotechnology & Bioprocess Engineering, Ostwestfalen-Lippe University of Applied Sciences and Arts, Lemgo, Germany YVONNE GENZEL • Bioprocess Engineering, Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany SEBASTIAN GIEHRING • PAIA Biotech GmbH, Ko¨ln, Germany GWENDAL GR€ANICHER • Bioprocess Engineering, Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany TANJA GREIN • Institute of Bioprocess Engineering and Pharmaceutical Technology (IBPT), Technische Hochschule Mittelhessen (THM)—University of Applied Sciences, Giessen, Germany GERRIT HAGENS • HES-SO University of Applied Sciences and Arts Western Switzerland, Sion, Switzerland VOLKER C. HASS • University of Applied Sciences, Hochschule Furtwangen University, Villingen-Schwenningen, Germany

xi

xii

Contributors

JULIA HENNICKE • Department of Biotechnology, VIBT, University of Natural Resources and Life Sciences (BOKU), Vienna, Austria OLIVIER HENRY • Department of Chemical Engineering, Polytechnique Montre´al, Montre´al, QC, Canada TANJA HERNA´NDEZ RODRI´GUEZ • Biotechnology & Bioprocess Engineering, OstwestfalenLippe University of Applied Sciences and Arts, Lemgo, Germany CHRISTOPH HERWIG • Institute of Chemical, Environmental and Bioscience Engineering, TU Wien, Vienna, Austria UWE JANDT • Institute of Bioprocess and Biosystems Engineering, Hamburg University of Technology, Hamburg, Germany LUKAS K€AßER • Institute of Bioprocess Engineering and Pharmaceutical Technology (IBPT), Technische Hochschule Mittelhessen (THM)—University of Applied Sciences, Giessen, Germany JULIAN KAGER • Institute of Chemical, Environmental and Bioscience Engineering, TU Wien, Vienna, Austria ERIC KARENGERA • Department of Chemical Engineering, Polytechnique Montre´al, Montre´al, QC, Canada; UCB Pharmaceuticals, Brussels, Belgium MICHAEL V. KEEBLER • Kuhner Shaker Inc., San Carlos, CA, USA SANDRO KOFLER • Institute of Chemical, Environmental and Bioscience Engineering, TU Wien, Vienna, Austria KATRIN KORTE • Institute of Bioprocess and Biosystems Engineering, Hamburg University of Technology, Hamburg, Germany PAUL KROLL • Securecell AG, Schlieren, Switzerland KIM B. KUCHEMU¨LLER • Institute of Bioprocess and Biosystems Engineering, Hamburg University of Technology, Hamburg, Germany RENATE KUNERT • Department of Biotechnology, VIBT, University of Natural Resources and Life Sciences (BOKU), Vienna, Austria MARTI´ LECINA • Department of Chemical, Biological and Environmental Engineering, Universitat Auto`noma de Barcelona, Cerdanyola del Valle`s, Spain; Bioengineering Department, IQS, Universitat Ramon Llull, Barcelona, Spain MARIO LEDERLE • Rentschler Biopharma SE, Laupheim, Germany; Department of Biotechnology, Mannheim University of Applied Sciences, Mannheim, Germany SARA LINSE • Department of Biochemistry and Structural Biology, Lund University, Lund, Sweden ALEXANDER LOA • acCELLerate GmbH, Hamburg, Germany PAULA M. MARQUES • Instituto de Biologia Experimental e Tecnologica, Oeiras, Portugal; Instituto de Tecnologia Quı´mica e Biologica, Universidade Nova de Lisboa, Oeiras, Portugal CRISTINA BERNAL MARTINEZ • Applikon-Biotechnology BV, Delft, The Netherlands PATRICK MAYRHOFER • Department of Biotechnology, VIBT, University of Natural Resources and Life Sciences (BOKU), Vienna, Austria GAVIN MCGAURAN • School of Biomolecular and Biomedical Science, Conway Institute of Biomolecular and Biomedical Science, University College Dublin, Dublin 4, Ireland JOA˜O P. MENDES • Instituto de Biologia Experimental e Tecnologica, Oeiras, Portugal JOHANNES MO¨LLER • Institute of Bioprocess and Biosystems Engineering, Hamburg University of Technology, Hamburg, Germany

Contributors

xiii

MASSIMO MORBIDELLI • Institute for Chemical and Bioengineering, ETH Zurich, Zurich, Switzerland JAN MU¨LLER • School of Life Sciences and Facility Management, Institute of Chemistry and Biotechnology, Zurich University of Applied Sciences, W€ a denswil, Switzerland ALEXANDER NIKOLAY • Bioprocess Engineering, Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany DAVID J. O’CONNELL • School of Biomolecular and Biomedical Science, Conway Institute of Biomolecular and Biomedical Science, University College Dublin, Dublin 4, Ireland; BEACON Bioeconomy Research Centre, University College Dublin, Dublin 4, Ireland CLAUDIO PACKI • Department of Biotechnology, Mannheim University of Applied Sciences, Mannheim, Germany CRISTINA PEIXOTO • Instituto de Biologia Experimental e Tecnologica, Oeiras, Portugal; Instituto de Tecnologia Quı´mica e Biologica, Universidade Nova de Lisboa, Oeiras, Portugal RALF PO¨RTNER • Institute of Bioprocess and Biosystems Engineering, Hamburg University of Technology, Hamburg, Germany UDO REICHL • Bioprocess Engineering, Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany; Bioprocess Engineering, Otto von Guericke University Magdeburg, Magdeburg, Germany DENISE SALZIG • Institute of Bioprocess Engineering and Pharmaceutical Technology (IBPT), Technische Hochschule Mittelhessen (THM)—University of Applied Sciences, Giessen, Germany CEDRIC SCHIRMER • School of Life Sciences and Facility Management, Institute of Chemistry and Biotechnology, Zurich University of Applied Sciences, W€ a denswil, Switzerland ¨ KARIN SCHUTZE • CellTool GmbH, Tutzing, Germany RICARDO J. S. SILVA • Instituto de Biologia Experimental e Tecnologica, Oeiras, Portugal NINA STEFFEN • School of Life Sciences and Facility Management, Institute of Chemistry and Biotechnology, Zurich University of Applied Sciences, W€ a denswil, Switzerland MIRCEA TRIC • Fraunhofer Institute IPA, Mannheim, Germany STEVE R. C. WARR • GSK Medicines Research Centre, Stevenage, UK OLIVER WEHMEIER • acCELLerate GmbH, Hamburg, Germany TOBIAS WEIDNER • Institute of Bioprocess Engineering and Pharmaceutical Technology (IBPT), Technische Hochschule Mittelhessen (THM)—University of Applied Sciences, Giessen, Germany SO¨REN WERNER • School of Life Sciences and Facility Management, Institute of Chemistry and Biotechnology, Zurich University of Applied Sciences, W€ a denswil, Switzerland TOBIAS WERNER • Department of Biotechnology, Mannheim University of Applied Sciences, Mannheim, Germany PHILIPP WIEDEMANN • Department of Biotechnology, Mannheim University of Applied Sciences, Mannheim, Germany VINCENT WIEGMANN • Department of Biochemical Engineering, The Advanced Centre for Biochemical Engineering, University College London, London, UK MORITZ WOLF • Institute for Chemical and Bioengineering, ETH Zurich, Zurich, Switzerland YIXIAO WU • State Key Laboratory of Bioreactor Engineering, East China University of Science and Technology, Shanghai, China

xiv

Contributors

HESHAM K. YOSEF • CellTool GmbH, Tutzing, Germany AN-PING ZENG • Institute of Bioprocess and Biosystems Engineering, Hamburg University of Technology, Hamburg, Germany JAN ZITZMANN • Institute of Bioprocess Engineering and Pharmaceutical Technology (IBPT), Technische Hochschule Mittelhessen (THM)—University of Applied Sciences, Giessen, Germany

Part I Cell Line and Media Development

Chapter 1 Near-Physiological Cell Cycle Synchronization with Countercurrent Centrifugal Elutriation Johannes Mo¨ller, Katrin Korte, Ralf Po¨rtner, An-Ping Zeng, and Uwe Jandt Abstract The bioreactor conditions and cell diversity in mammalian cell cultures are often regarded as homogeneous. Recently, the influence of various kinds of heterogeneities on production rates receives increasing attention. Besides spatial gradients within the cultivation system, the variation between cell populations and the progress of the cells through the cell cycle can affect the dynamics of the cultivation process. Strong metabolic up- and down-regulations leading to variable productivities, even in exponentially growing cell cultures, have been identified in CHO cell cultivations. Consequently, scientific studies of cell cycle-related effects and metabolic regulations require experiments utilizing cell cycle-enriched subpopulations. Importantly, the enrichment procedure itself must not strongly interfere with the cell culture under investigation. Such subpopulations can be generated by near-physiological countercurrent centrifugal elutriation, which is described in the following chapter. At first, a brief overview regarding the cell cycle, currently identified effects and commonly used methods, and their applicability is outlined. Then, the experimental setup and the synchronization itself are explained. Key words Population heterogeneities, Cell size, Separation, Centrifugation

Abbreviations CHO DNA DAPI FSC G1 phase G2 phase HEK M PBS S phase SSC

Chinese hamster ovary Deoxyribonucleic acid 40 ,6-Diamidino-2-phenylindole Forward scatter Gap phase 1 Gap phase 2 Human embryonic kidney Mitosis Phosphate-buffered saline Synthesis phase Side scatter

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_1, © Springer Science+Business Media, LLC, part of Springer Nature 2020

3

Johannes Mo¨ller et al.

4

Nomenclature %sync %non-sync YΩ

1

Percentage of cells in a cell cycle phase after synchronization Percentage of cells in a cell cycle phase in the preculture Enrichment factor

Introduction The main task of a bioreactor is to ensure stable and reproducible conditions, enabling efficient cell growth and high productivity. Contrary to this aim, spatial heterogeneities in bioreactor hydrodynamics and culture conditions such as pH or dissolved oxygen concentration and their impact on the bioprocess performance have been recently identified [1, 2]. At the same time, the dynamics of the cultivation process is also affected by heterogeneous cell populations in the bioreactor. For example, different protein expression levels can occur even in monoclonal-derived cell lines, leading to high- and low-producing cells [3]. In addition, one major reason for cell population heterogeneities is the progress of proliferating cell cultures through the cell cycle [4, 5].

1.1 Cell Cycle Definition

The progression of mammalian cells through the cell cycle is still not fully understood and is part of ongoing research [6, 7]. In general, the cell cycle is a series of specifically ordered events, which are necessary for cell proliferation. The cell genome and the cell organelles are doubled in quantity, and two genetically identical cells are created at the end of the cell cycle. It can be separated into four distinct phases [8]: 1. Gap phase 1 (G1): The cells are growing in this phase and increasing their cell mass. No transition to the synthesis phase occurs, if the growth conditions are not appropriate. 2. Synthesis phase (S): The chromosome is replicated in this phase, and protein expression is generally upregulated. 3. Gap phase 2 (G2): As in G1 phase, the cells grow and express proteins, which increases their cell mass. 4. Mitosis (M): The mitosis involves several steps. In brief, the replicated chromosome is distributed to the edges of the mother cell. Then, the cytoplasm is divided and two daughter cells are formed. G2 and M phase are mostly summarized into one G2/M phase in animal cell technology. A summary of the cell cycle phases including their distribution in exponentially growing cell cultures (in brackets) is shown in Fig. 1.

Cell Cycle Synchronization

5

Fig. 1 Representation of the cell cycle phases. Number in brackets represents cell cycle distribution in exponentially growing CHO K-1 cells. Existence of G0 phase is still controversial

It may happen that cells enter a quiescent state (G0 phase), in which they do not grow, but remain alive. But the existence of this phase and its characteristics are controversially discussed in literature [9]. 1.2 Cell Cycle Analysis by Synchronization

The effects of cell cycle dependencies in cell cultures can be conveniently studied by applying synchronization methods on proliferating cells, effectively leading to a time-dependent variable (oscillating) composition of the culture. Importantly, the synchronization procedure itself should have minimal interference with the cell growth and physiology, and it should allow to reliably enforce synchronization according to the following list of quality criteria for synchronization techniques [10]: 1. In synchronized culture, every cell parameter should have a similar value as cells in the corresponding cell cycle phase in an unsynchronized culture. 2. Unaffected cell growth after synchronization. Average kinetics should behave similarly in both non-synchronized and synchronized cultivations after integration over integer multiples of the cell cycle duration. 3. Minimal increase in cell number during the interdivision time. This corresponds to a short fraction of time for division, compared to the duration of the cell cycle. 4. Narrower DNA distribution and narrower size distribution compared to non-synchronized culture. The progress of both distributions must be coherent with the cell growth and the doubling times. A variety of synchronization methods, mainly consisting of chemical and physical methods, have been discussed in the past. Chemical methods are assumed to block the cell cycle distribution at specific points and synchronize the whole culture (see ref. 10 for

6

Johannes Mo¨ller et al.

discussion). Hence, they interact with the cell metabolism and were shown to not synchronize cells physiologically. Studies of cell cycle dependencies using chemical methods are controversially discussed [10] or categorically denied [7]. In short, they do not fulfill one or more of the abovementioned criteria. Contrarily, physical methods synchronize the culture by selecting only a part of it. They are assumed to not principally interfere with cell cycle control and the central metabolism, which makes them the only available methods for cell-culture studies under nearphysiological conditions. Nevertheless, the quality of physical synchronization needs to be controlled according to the mentioned criteria. Numerous physical synchronization methods have been summarized and discussed in [10]. The physical method achieving overall best reproducibility, yield, and cell cycle separation while minimally interfering with cell physiology is countercurrent centrifugal elutriation. 1.3 Physical Synchronization by Countercurrent Centrifugal Elutriation

Countercurrent centrifugal elutriation is a physical method originally developed for the separation of cells out of, for example, blood samples [10]. In our group, a workflow for the investigation of cell cycle-dependent subpopulation dynamics by nearphysiological countercurrent centrifugal elutriation combined with population-balanced modeling was developed in recent years [4, 11]. Strong metabolic regulations for the nonproducing industrial cell lines AGE1.HN and CHO-K1, antibody producing CHO DP-12 cells, and partly HEK293s have been identified [5, 11–13]. Other presumed cell cycle dependencies, for example, cell cycle effects in transfection efficiency, have been found to be weak or not present under near-physiological cultivation conditions [14]. The method utilizes a modified centrifuge including an elutriator in the rotor (Fig. 2). The elutriator has an inlet for a liquid stream (e.g., phosphatebuffered saline, medium), which passes through the funnel-shaped flow chamber in countercurrent flow and leaves the system. If non-synchronized cells are introduced into the flow chamber (Fig. 3a) and a centrifugal force is applied in addition to the liquid flow, the cells are separated due to their size and density (Fig. 3b). This is based on the opposite effects (centrifugal force and countercurrent flow) and allows the separation of populations with different cell diameters and densities through the adjustment of the countercurrent flow rate (Fig. 3c). In summary, the main advantages of this method are: 1. Near-physiological separation of cells based on size and density 2. High amount of synchronized cells (up to 20  109 cells per batch) 3. High reproducibility and cell cycle phase distinction

Cell Cycle Synchronization

7

Fig. 2 Setup of centrifuge equipped with an elutriator inside

Fig. 3 (a–c) Different states in elutriation chamber. (Modified from ref. 4)

4. Near-physiological synchronization to study biological effects 5. Relatively fast procedure (approx. 30 min), excluding preparation (see Note 1) 6. Multiple synchronization batches possible The synchronized cell population can be collected at the outlet and used for further analysis or experimentation. The efficiency of cell cycle synchronization can be evaluated based on the cell enrichment factor YΩ, which is the relationship of the percentage of cells in a cell cycle phase after synchronization %sync divided by the number of cells in the same phase in the preculture (%non-sync). YΩ ¼

2 2.1

%sync  100% %non‐sync

ð1Þ

Materials General

1. The animal cell line and medium. Preferably growing in suspension with chemically-defined medium. 2. A cultivation system for cultivation of cell cycle-enriched populations (e.g., shake flasks, bioreactor) (see Note 2).

8

Johannes Mo¨ller et al.

3. Hemocytometer or an automated system for determination of total, viable, and dead cell concentration (see Note 3). 4. A flow cytometer to determine the cell cycle distribution of cell cycle-synchronized cell populations (see Note 4). 2.2

3

Elutriation Setup

The specific devices are examples and were so used by the authors. l

Centrifuge (Beckman Avanti J-26S XP, modified for JE-5.0 Elutriator, Beckman Coulter, Krefeld, Germany).

l

JE-5.0 Elutriation system including standard and large chamber (Beckman Coulter). Choice depends on required cell numbers (see Note 5).

l

Silicon tubings, pressure gauge, and bubble trap.

l

Pump (preferably pulse-free), for example, 8-roller head pump (Medorex, No¨rten-Hardenberg, Germany).

l

Syringe pump for reproducible sample injection, for example, Syringe Pump AL1000 (World Precision Instruments GmbH, Friedberg, Germany).

l

Sodium hypochlorite solution (5% v/v in purified water) for sterilization of elutriation system, freshly prepared.

l

Sterile phosphate-buffered saline (pH ¼ 7.4, short: PBS).

l

Measuring cylinder for flow calibration.

l

Stopwatch.

l

PBS, not sterile.

l

Second centrifuge for centrifugation of preculture.

l

Sterile syringe and sterile single-use disposables (tubes, pipettes).

l

Sterile 3-way valve.

Methods The following methods were adapted from refs. 5, 11 and optimized for the near-physiological synchronization of CHO cells. The shown flow rates and centrifugation parameters are based on CHO DP-12 cells.

3.1

Preparation

1. Assemble elutriator based on manufacturer instructions for use. 2. Place elutriator in centrifuge as in manufacturer instructions for use. 3. Connect silicone tubing to inlet and outlet (see Fig. 4).

Cell Cycle Synchronization

9

Fig. 4 Setup of centrifuge equipped with an elutriator inside

4. Connect bubble trap, pressure gauge, tubing, and pump as shown in Fig. 4. 5. Tighten all tube connections with cable ties (especially between the inside and outside of centrifuge). 6. Sterilize PBS by autoclaving, and ensure bottle is equipped with sterile filter and tubing necessary for sucking PBS out. 7. Have same bottle (min 5 l volume) with sterilized or nonsterilized PBS for flow calibration (if calibration is necessary) (see Note 6). 8. Three-way valve is positioned that no syringe is connected and flow is possible (like a T). 3.2 Setup of Flow System

1. Connect nonsterile PBS to the flow system (buffer reservoir). 2. Start pump with a low pump rate to fill the system with buffer; release of air must be ensured. Care is necessary because presence of bubbles can disturb the synchronization and lead to overpressure and leakage (nonsterility). 3. Fill the bubble gauge to approx. half the volume with buffer by turning it around while the pump is running; repeat 3 times (see Note 7). 4. Fill the flow system continuously with buffer and wait approx. 10 min. 5. Watch for leakage. 6. Take out elutriator insert and turn it in different directions to release the remaining bubbles; no bubbles should remain in it. Softly knock on the elutriator to release bubbles.

10

Johannes Mo¨ller et al.

7. Start centrifugation (T ¼ 25  C; important: acceleration, min; deceleration, min) with 300  g as set point (comparable to standard centrifugation of CHO cells). 8. Watch the centrifuge carefully. If it starts to shake, turn off and remove bubbles again. The pressure shown by the pressure gauge should not increase drastically. If it does, turn off and remove bubbles again. 9. Proceed only if the flow system and centrifugation run smoothly with low pressure and continuous flow. Check flow at the outlet. 10. Leave system at least 30 min to ensure stable flow conditions. 3.3 Nonsterile Flow Calibration (Pump)

The calibration of the system flow rate is required in order to develop reproducible elutriation protocols with different pumps or tubings. The flow rates shown here were developed for the standard and large chamber. 1. Prepare flow system as described in Subheading 3.2. 2. Set flow rate to a low level; the authors used approx. 10–40 mL/min for the standard chamber and 80–160 mL/ min for the large chamber. 3. Wait at least one minute until the liquid flow is continuous. 4. Place outlet tubing (normally connected to collection flask) into a measuring cylinder and start the time measurement. 5. Wait 1 min and put outlet tubing into the collection flask; stop the time measurement. 6. Note the volume and time. 7. Empty the measuring cylinder. 8. Repeat steps 2–6 for the whole range of the investigated flow rates (3 times per flow rate). 9. Plot the volume flow rate (volume/time, y axis) over the set points of the pump (x axis) in a table calculator. 10. Determine the goodness of fit and slope (R2 should be at least >0.95) using a linear slope correlation.

3.4 Sterile Preparation Before Synchronization

In this part, the sterile preparation of the calibrated (Subheading 3.3) system is described after the setup (Subheading 3.2). Chemical sterilization is appropriate (see Note 8). The buffer reservoir and the outlet tube/waste container should be placed under a clean bench. 1. Stop liquid pump, centrifuge is turned on, and centrifugation is applied at 300  g.

Cell Cycle Synchronization

11

2. Connect the sodium hypochlorite solution to the inlet (buffer reservoir), and start pump again with a low flow rate. The outlet tube needs to be connected to a sterile waste container. 3. Wait approx. 10 min. 4. Put outlet tube in sodium hypochlorite solution so that it is pumped in a circle. 5. Wait at least 1 h. 6. Put the outlet tube in the sterile waste container. 7. Stop the pump. 8. Connect inlet tube to sterile PBS and start pump again. 9. Watch for bubbles and/or pressure increases, and ensure a bubble-free elutriator. 10. Flush the system at least for 30 min before starting the synchronization. 3.5 Sterile Synchronization

Important: If the synchronization is performed the first time, a further characterization of the flow chamber (Subheading 3.6) is recommended to determine the optimal procedure for the elutriation beforehand. This part describes a run with known parameters. All cell culture handling and collection of fractions need to be done in a clean bench. 1. Proof all connections and the current pressure (centrifuge and pump are turned on, 25  C and 300  g in a centrifuge). 2. Set flow rate to a set point lower than your first fraction (fractions will be washed out otherwise). 3. Measure cell density and viability in preculture (e.g., out of a shake flask, exponentially growing, viability >95%). 4. Centrifuge preculture (as in own protocol, or 10 min, 300  g). 5. Discard remaining medium. 6. Suspend cell pellet in a low volume of sterile PBS (e.g., 5–15 mL). 7. Transfer cell suspension into a sterile syringe (do not connect a needle and do not suck the cell suspension up) (see Notes 9 and 10). 8. Connect the syringe with the flow system (3-way valve still in T). 9. Turn the 3-way valve such that the syringe is connected (valve in T). 10. Put the syringe (connected to flow system) into the syringe pump. 11. Start the syringe pump with 1 mL/min.

12

Johannes Mo¨ller et al.

12. Start to collect 50 mL fractions at the outlet, and put them on ice (two people recommended). Fractions could be collected by holding the outlet over the 50 mL tubes. Wait until the 50 mL mark is reached, and hand it over on the next flask. 13. Watch pressure gauge; it should not increase (if it increases, your cells might be clumping). Look at the elutriation chamber if your centrifuge is equipped with a stroboscope and a watch window. 14. Wait until the cell suspension is loaded in the elutriator. 15. Leave syringe in syringe pump and collect one more fraction. 16. Start to increase the fluid flow stepwise and collect at least 50 mL fractions with each fluid flow. 17. Stop to increase the fluid flow after you reach the last fraction. 18. Switch to a new 50 mL tube and turn off the centrifuge (deceleration: min) to collect the remaining cells. 19. Stop pump after the last fraction. 20. Take a sample of fractions (e.g., 1 mL) for the determination of the cell number and cell cycle distribution. Centrifuge and suspend cells in lower volume if necessary. 21. Proceed with further studies (see Note 11). 22. Cleaning and disabling as suggested in the manufacturer’s protocol. 3.6 Characterization of the Flow Chamber

The characterization of the flow chamber is needed to know which cell cycle distribution is released at which flow rate. Therefore, the investigated flow rates should be separated in small parts, for example, 4–8 flow rates with equal distance. 1. Set up elutriator as in Subheadings 3.2 and 3.3. Characterization can be performed unsterile, but analytics should then be performed as soon as possible. 2. Perform elutriation as in Subheading 3.5. If it is performed unsterile, no clean bench and sterilization are needed. 3. Collect fractions and analyze the cell number and cell cycle distribution of each fraction. 4. Transfer results to a table calculator and evaluate cell cycle distributions. An example of the characterization of the large chamber (40 mL, Beckman Coulter) is shown in Fig. 5 with 7 fractions due to different flow rates (F1–F7) and one fraction (F8-Dec) while turning off the centrifuge. An increasing G2/M phase distribution with increasing fluid flow is typical due to the increasing size of the cell population. Remaining cells will stay in the system and are washed out in the last fraction.

Cell Cycle Synchronization

13

Fig. 5 Characterization of the large chamber (40 mL, Beckman Coulter) is exemplary shown in this figure with 7 fractions due to different flow rates (F1–F7) and one fraction (F8-Dec) while turning off the centrifuge. Preculture: CHO DP-12 cells

4

Notes 1. It has been shown that the inoculum can be pooled out of multiple elutriation runs [4], but a single run with higher cell numbers is recommended. 2. Cultivation of cell cycle-synchronized populations should be performed under substrate excess without substrate limitations or metabolite inhibitions. A repeated-batch setup is recommended. Furthermore, a high sampling frequency (approx. 8 samples per oscillation or every 3 h) is necessary to study cell cycle-dependent metabolism. 3. Measurement of cell concentration and cell size distribution can be performed with the Z2 particle counter (Z2, Beckman Coulter) as explained in [12]. Viability can be measured using the DAPI method. Therefore, the samples (1 mL) were taken and centrifuged at 300  g for 3 min, and the supernatant was frozen for further analysis. The cell pellet was suspended in 4  C PBS with 1 μg/mL 40 ,6-diamidino-2-phenylindole (DAPI) and measured with flow cytometry (e.g., Beckman Coulter, Cytoflex Laser: 405 nm, Signal: 450/50). 30,000 events should be recorded, debris and doublets can be excluded using SSC-A vs. FSC-A and FSC-H vs. FSC-A gating, and non-stained cells are gated as viable.

14

Johannes Mo¨ller et al.

4. The cell cycle was determined with DNA staining using DAPI. The remaining cell pellet (see Note 3) needs to be fixed in 70% ethanol at 20  C for at least 30 min. First, the cells are permeabilized (0.1% Triton X-100 in PBS, 30 min, 0.1 μg/mL DAPI) at room temperature. Then, the DNA fluorescence is measured with flow cytometry (Laser: 405 nm, Signal: 450/50). 50,000 events are recorded, and debris and doublets are excluded as described in Note 3. The flow cytometer data are exported (Kaluza 1.3, Beckman Coulter) to Microsoft Excel (Microsoft, USA) and applied for standardized and reliable quantification as published in [12]. 5. The standard chamber has a volume of 4 mL and the large chamber a volume of 40 mL. If studies are performed with CHO cells, the 40 mL chamber is recommended because it is possible to synchronize much more cells (2  107–109 mio cells) than with the standard chamber (2  108–1010 mio cells) in a single run. 6. PBS can lead to crystal forming if leakage occurs. Use ultrapure water in the beginning to ensure a leakage-free system. 7. It is important to observe the pressure gauge during the work with the elutriator. Mostly, leakages result from high pressure due to bubbles or clumping cells. 8. All parts of the elutriator are also heat sterilizable (watch manufacturer’s recommendation). Sterilization with sodium hypochlorite solution is relatively simple and effective, but the disinfectant must be removed carefully to not harm the cells. 9. Loading of the syringe can be performed as follows: Take out the plunger and put a plug on the Luer Lock connection. Then pipette the cell suspension into the syringe, put in the plunger, turn the syringe, open the plug a bit, and release the air by pressing on the plunger flange. 10. Loading of the system is a sensitive step. It can lead to blockage of the fluid system and to high pressure and leakage. If the optimal loading speed is unknown, careful loading is advisable. Loading could also be performed by pressing the syringe by hand but is much more difficult and hardly reproducible compared to a syringe pump. 11. Bioreactor cultivations with controlled cultivation parameters are recommended. Further knowledge could be drawn from model-based studies as proposed in [5, 11]. In brief, nearphysiological synchronization (as discussed in this chapter), population-resolved mechanistic modeling, and statistical analysis of the specific parameter distributions can be combined to understand cell cycle-dependent regulations in mammalian cell culture. Exemplary cultivation results of cell cycle-enriched populations could be seen in Fig. 6 with a typical stepwise growth profile (see ref. 12 for more details).

Cell Cycle Synchronization

15

Fig. 6 Exemplary growth curves for an oscillating culture (HEK293s cells; see ref. 12). Points represent measured normalized cell densities. Solid lines representing independent overlaying sliding window fits, which could be used to gain understanding of cell cycle-dependent metabolic regulations [5] References 1. Brunner M, Braun P, Doppler P, Posch C, Behrens D, Herwig C, Fricke J (2017) The impact of pH inhomogeneities on CHO cell physiology and fed-batch process performance – two-compartment scale-down modelling and intracellular pH excursion. Biotechnol J 12 (7):1600633. https://doi.org/10.1002/biot. 201600633 2. Lara AR, Galindo E, Ramirez OT, Palomares LA (2006) Living with heterogeneities in bioreactors: understanding the effects of environmental gradients on cells. Mol Biotechnol 34 (3):355–381. https://doi.org/10.1385/ MB:34:3:355 3. Pilbrough W, Munro TP, Gray P (2009) Intraclonal protein expression heterogeneity in recombinant CHO cells. PLoS One 4(12): e8432. https://doi.org/10.1371/journal. pone.0008432. 4. Platas Barradas O, Jandt U, Becker M, Bahnemann J, Po¨rtner R, Zeng A-P (2015) Synchronized mammalian cell culture: Part I—A physical strategy for synchronized cultivation under physiological conditions. Biotechnol Progr 31(1):165–174. https://doi.org/ 10.1002/btpr.1944 5. Mo¨ller J, Korte K, Po¨rtner R, Zeng A-P, Jandt U (2018) Model-based identification of cellcycle-dependent metabolism and putative autocrine effects in antibody producing CHO cell culture. Biotechnol Bioeng 115:2996–3008. https://doi.org/10.1002/ bit.26828

6. Kohrman AQ, Matus DQ (2017) Divide or conquer: cell cycle regulation of invasive behavior. Trends Cell Biol 27(1):12–25. https://doi. org/10.1016/j.tcb.2016.08.003 7. Cooper S, Gonzalez-Hernandez M (2009) Experimental reconsideration of the utility of serum starvation as a method for synchronizing mammalian cells. Cell Biol Int 33(1):71–77. https://doi.org/10.1016/j.cellbi.2008.09. 009 8. Alberts B, Johnson A, Lewis J, Raff M, Roberts K, Walter P (2004) Molekularbiologie der Zelle (German Edition). Wiley-VCH-Verlag GmbH, German. ISBN: 978-3-52734072-9 9. Cooper S (1998) On the proposal of a G0 phase and the restriction point. FASEB J 12 (3):367–373. https://doi.org/10.1096/ fasebj.12.3.367 10. Jandt U, Platas Barradas O, Po¨rtner R, Zeng A-P (2014) Mammalian cell culture synchronization under physiological conditions and population dynamic simulation. Appl Microbiol Biot 98(10):4311–4319. https://doi.org/10. 1007/s00253-014-5553-6 11. Jandt U, Platas Barradas O, Po¨rtner R, Zeng A-P (2015) Synchronized mammalian cell culture: Part II—population ensemble modeling and analysis for development of reproducible processes. Biotechnol Prog 31(1):175–185. https://doi.org/10.1002/btpr.2006 12. Castillo AE, Fuge G, Jandt U, Zeng A-P (2015) Growth kinetics and validation of

16

Johannes Mo¨ller et al.

near-physiologically synchronized HEK293s cultures. Eng Life Sci 15(5):509–518. https://doi.org/10.1002/elsc.201400224 13. Mo¨ller J, Bhat K, Riecken K, Po¨rtner R, Zeng A-P, Jandt U (2019) Process-induced cell cycle oscillations in CHO cultures: Online monitoring and model-based investigation. Biotechnol

Bioeng 116:2931–2943. https://doi.org/10. 1002/bit.27124 14. Fuge G, Zeng A-P, Jandt U (2017) Weak cell cycle dependency but strong distortive effects of transfection with lipofectamine 2000 in near-physiologically synchronized cell culture. Eng Life Sci 17(4):348–356. https://doi.org/ 10.1002/elsc.201600113

Chapter 2 Turning Cells into Reagents: The Preparation of Assay Ready Cells for Routine Use in Bioassays Oliver Wehmeier and Alexander Loa Abstract Assay ready cells are cryopreserved at a highly functional state and can be used in cell-based assay without prior cultivation or cell passaging. Basically, like any other reagent, the cells are applied to the assay instantly after thawing. Introduced initially in the drug discovery process where assay ready Frozen Instant Cells help to streamline cell-based high-throughput screening campaigns, the methodology now has been accepted for a much broader scope of applications and industries. The preparation of assay ready Frozen Instant Cells is not so much a piece of magic but a combination of good cell culture practice, careful handling, and individually optimized cryopreservation protocols. Here, a standard protocol is presented, how HepG2 cells are frozen in an assay ready quality to be used as a reagent in routine cell-based assays. Key words Assay ready cells, Cryopreservation, Bioassay, Good cell culture practice, Controlled-rate freezing

1

Introduction Mammalian cell lines serve as an in vitro model to simulate certain functional units of the human body [1, 2]. As such, they became an indispensable tool for multiple applications in various industries. Not only in pharmaceutical research but also for the identification of novel functional ingredients in the cosmetics and food industry [3–5], cell-based assays are used to screen chemical or biological entities for a desired biological function. For the proper safety assessment of newly designed chemicals but, according to REACH, also for chemicals which already have been marketed, mammalian cell lines can be used in an increasing number of alternative methods which have been approved to replace toxicity testing involving animals [6]. Finally, mammalian cell lines are used in bioassays to control the proper biological function of therapeutic proteins and antibodies which cannot be analyzed by biochemical methods alone [7].

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_2, © Springer Science+Business Media, LLC, part of Springer Nature 2020

17

18

Oliver Wehmeier and Alexander Loa

All three applications have their unique requirements for a cellbased assay. For the screening of new functional entities, miniaturization and scalability are important. Mostly, recombinant cell lines are used which need to respond to a stimulus in a very distinct and concentration-dependent manner. Large quantities of cells must be supplied at once to support the limited time frame of an HTS campaign. For in vitro toxicity testing, physiological relevance and robustness of the results are of equal importance. Very often, unmodified mammalian cells are used which have specific physiological properties that are addressed in the assay [8]. Therefore, a controlled cultivation of the cells is of the essence to ensure a reliable performance and significant results [9]. For the quality control of biologics that must be conducted under GMP, the bioassays are usually designed to reflect the correct mode of action of the biologic and must deliver the precise potency of the analyte at high specificity and unaffected by other elements of the assay [10]. The signal to background span and the full dose-response are in fact of secondary importance. Cells used for potency assays in manufacturing control must respond in a very robust and reliable way. Probably not so many assay runs are performed daily, but it is necessary that the cells still perform the same even after years. This put high requirements on the standardization and reproducibility of the cell culture process, in particular when considering critical but unsteady reagents like fetal bovine serum. For long, the common understanding has been—and sometimes still is—that cells need time to recover from cryopreservation until they regain their full functionality and proliferative capacity. After thawing, cells must be passaged two to three times before they can be used reliably in a cell-based assay. This is of course correct if cryopreservation is conducted with the sole idea of cell storage and a resuscitation of any kind whatsoever. If cryopreservation is considered as way to prepare functional reagents which cannot only be recovered but instantly perform as if they have never been frozen, different standards must be imposed. Based on the traditional understanding, one might expect that a unique and eventually proprietary technology is required to prepare assay ready Frozen Instant Cells. However, the sole secret is a combination of a good cell culture practice [9] and probably a mind shift to accept the idea that assay ready cells are in fact reagents, whose quality directly depends on how the cells have been handled. If they have been treated optimal, the assay ready cells will perform great, if not they don’t. However, as this sounds straight and easy, the devil is in the detail. Although the final process for the preparation of highly functional assay ready cells will probably not include any secret ingredient, it must be carefully adjusted for each individual cell type and application. In the following, a standard process for the

Turning Cells into Reagents

19

preparation of assay ready cell from HepG2 cells will be described, giving further hints recommendation for different cell types.

2

Materials

2.1 Cell Expansion of HepG2 Cells

1. Growth medium: Ham’s F12 (with phenol red), 2 mM Lglutamine, 20% fetal bovine serum with low endotoxin level (” and check the summary of fit graph to get a first impression for the quality of the established model (Fig. 3c, d). R2 describes the model fit and should be above 0.5. Q2 is the most critical and most sensitive indicator for a good model and is an estimate of future prediction precisions. It should be greater than 0.5 and better than 0.7  R2 for a good model. Model validity indicates the result of different statistical tests performed by the software to detect outliers, an incorrect model, or transformation problems. This value might also be small if the variability of the triplicates is very small, but still a valid model is obtained. The reproducibility describes the variability of the replicate cultures in comparison to total variability and should be above 0.5.

Fig. 3 Model output provided by the MODDE software. Using a central composite design, the quadratic model coefficients for mean titer (mTiter) (a) and mean total cell concentration (mTCC) (b) can be calculated. Those values are “center and scaled” by dividing the coefficients by the standard deviation of their respective response. Thereby, coefficients between responses of different ranges are comparable. Model quality for mTiter (c) and mTCC (d) is presented by R2 (green), Q2 (dark blue), the model validity (yellow), and reproducibility (light blue). Optimal factor level combinations for maximum predicted mTiter (e) and mTCC (f) can be illustrated by using contour plots

38

Patrick Mayrhofer and Renate Kunert

8. Click “Next>” to obtain the normal probability plot that can be used for the detection of outliers. 9. By clicking “Next>,” the graph for observed vs. predicted values is presented. 10. Repeat steps 4–9 for generating further models for additional investigated responses. 11. Determine the optimum level factors by using contour plots (Fig. 3e, f) or using the MODDE optimizer function that calculates the optimal factor levels for two or more simultaneously optimized responses.

4

Notes 1. To detect curvature of the response function, a triplicate center point can be used at this step already. 2. Here, we can freely control the different levels of the quantitative concentration factors using 1 as the low and +1 as the high limits. 3. Further responses can be added later. 4. RSM stands for “response surface methodologies.” 5. In this special case, α ¼ 1 was chosen comprising only three different factor levels (1, 0, and +1) in the final design to reduce the media preparation time and complexity. Choosing this design allows a reduction in the number of experiments to a total of 23 runs as opposed to 34 (¼81) for a full factorial design with four factors at three levels. A star distance α ¼ 1 is called a “face-centered design.” The advantage is that only three factors have to be investigated. 6. Using the total molar amino acid ratio is one option to normalize different feed supplements relative to each other as exemplified in this protocol. However, this requires the knowledge of the amino acid content of the (proprietary) feed supplements. The amino acid content of biological solutions is routinely analyzed by HPLC analysis. In our example, Cell Boost 1 has an amino acid molarity 1.8 times higher than Cell Boost 3, and therefore, we used Cell Boost 3 with a 1.8fold volumetric excess. 7. Phenol red might be added as pH indicator. Additionally, L-glutamine (4–8 mM) supplementation is needed for certain cell lines. 8. Alternatively, cells can be seeded at 0.5  106 c/mL directly in the culture tubes for expansion, and semi-perfusion is then started on day 3. This saves additional culture vessels for cell expansion. However, using this method, there might be

DoE-Based Perfusion Medium Development

39

increased variability between cultures because of slightly different cell numbers at the start of perfusion. This variability might be minimized by pooling cells and subsequent aliquoting before starting the semi-perfusion routine. 9. The preset ViCell instrument limit is set to 10  106 c/mL. Dilute if cell numbers above that limit are expected. 10. Additional supernatant can be saved for further analysis or protein A affinity chromatography. 11. Make sure that cells are fully resuspended and no cell clumps are present. For this, some cell lines can be vortexed shortly. Alternatively, homogeneous mixing can be achieved by pipetting the suspension up and down. 12. In the MODDE software, the “auto-tuning” function helps to optimize the model quality. References 1. Woodside SM, Bowen BD, Piret JM (1998) Mammalian cell retention devices for stirred perfusion bioreactors. Cytotechnology 28:163–175 2. Voisard D, Meuwly F, Ruffieux P-A et al (2003) Potential of cell retention techniques for large-scale high-density perfusion culture of suspended mammalian cells. Biotechnol Bioeng 82:751–765 3. Castilho LR, Medronho RA (2002) Cell retention devices for suspended-cell perfusion cultures. Adv Biochem Eng Biotechnol 74:129–169 4. Bielser J-M, Wolf M, Souquet J et al (2018) Perfusion mammalian cell culture for recombinant protein manufacturing – a critical review. Biotechnol Adv 36:1328–1340 5. Godawat R, Konstantinov K, Rohani M et al (2015) End-to-end integrated fully continuous production of recombinant monoclonal antibodies. J Biotechnol 213:13–19 6. Warikoo V, Godawat R, Brower K et al (2012) Integrated continuous production of recombinant therapeutic proteins. Biotechnol Bioeng 109:3018–3029 7. Rathore AS, Agarwal H, Sharma AK et al (2015) Continuous processing for production of biopharmaceuticals. Prep Biochem Biotechnol 45:836–849 8. Chen C, Wong HE, Goudar CT (2018) Upstream process intensification and

continuous manufacturing. Curr Op Chem Eng 22:191–198 9. Ritacco FV, Wu Y, Khetan A (2018) Cell culture media for recombinant protein expression in Chinese hamster ovary (CHO) cells: history, key components, and optimization strategies. Biotechnol Prog 34:1407–1426 10. Lin H, Leighty RW, Godfrey S et al (2017) Principles and approach to developing mammalian cell culture media for high cell density perfusion process leveraging established fed-batch media. Biotechnol Prog 33:891–901 11. Gomez N, Ambhaikar M, Zhang L et al (2016) Analysis of Tubespins as a suitable scale-down model of bioreactors for high cell density CHO cell culture. Biotechnol Progr 33:490–499 12. Villiger-Oberbek A, Yang Y, Zhou W et al (2015) Development and application of a high-throughput platform for perfusion-based cell culture processes. J Biotechnol 212:21–29 13. Wolf MKF, Lorenz V, Karst DJ et al (2018) Development of a shake tube-based scaledown model for perfusion cultures. Biotechnol Bioeng 115:2703–2713 14. Goletz S, Stahn R, Kreye S (2016) Small scale cultivation method for suspension cells. https://patents.google.com/patent/ WO2016193083A1/en 15. Parampalli A, Eskridge K, Smith L et al (2007) Development of serum-free media in CHO-DG44 cells using a central composite statistical design. Cytotechnology 54:57–68

Part II Techniques for Process Development

Chapter 4 ambr® 15 Microbioreactors for CHO Cell Cultivation Steve R. C. Warr Abstract The ambr 15 has become the industry’s standard automated microbioreactor system for mammalian cell culture. It has applications throughout the industry, most commonly for cell line screening and media/feed development. On each ambr 15 workstation, conditions in up to 48  15 mL bioreactors can be individually controlled while a liquid handler enables automated addition and removal of liquids during the process. Integrated cell counting, metabolite analysis and pH offset correction are also possible thereby reducing the operator interactions that are required. Extensive user and software manuals are supplied by the manufacturer, but in this chapter we describe additional ways of working that we have implemented in routine cell line screening using the ambr 15. Key words ambr® 15, Cell culture, Scale-down model, Automated bioreactor

1

Introduction The ambr 15 is an automated bioreactor system allowing 24 or 48 parallel cell culture cultivations in a cost-effective 10–15 mL microbioreactor format. This can provide an effective scale-down model of conventional bioreactors through closed loop pH and DO control, independent gassing of each bioreactor, and control of temperature and agitation combined with the automated addition of feeds and other reagents [1, 2]. Continuous improvement since its launch in 2010 has ensured the ambr 15 has become the industry’s standard for small-scale automated cell culture. Currently, integrated cell counting, pH offset correction and metabolite analysis are available, and integrated MODDE Design of Experiments software can be used to facilitate the execution of DoE on the ambr 15. Additionally, a microbial version (ambr 15 fermentation) was launched in 2017 [3], and a generation 2 system was launched in 2019. Across the BioPharma industry the ambr 15 is most commonly used for cell line screening and the development of media, feed and process [4, 5], but other applications include modeling perfusion

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_4, © Springer Science+Business Media, LLC, part of Springer Nature 2020

43

44

Steve R. C. Warr

cell cultures [6, 7], growing cells on microcarriers [8, 9] and the execution of automated transient transfections [10, 11]. Because the ambr 15 can be used to model larger-scale production vessels, it offers significant advantages over screening in traditional shake flask or well plate systems [12, 13]. Cells are grown in a well-controlled environment and feeding can closely mimic that in the final production process. The ability to simultaneously evaluate up to 48 cell lines in a consistent, automated manner can also increase throughput and reduce the time scientists need to spend in the lab. In most cases, labs will have a single or lead ambr 15 user(s) but individual experiments may be “owned” and carried out by other lab scientists. It is expected that users will have undertaken ambr 15 training from Sartorius and will therefore be competent in ambr 15 process writing and operation. Individual users will develop their own processes and operating protocols and therefore, rather than providing an operating protocol, this chapter will describe some specific ways of working that we have developed to facilitate successful cell culture in the ambr 15. Although we have implemented these primarily for cell line screening using a fed-batch process, many of the principles described are equally applicable to other experimental designs.

2

Materials General: Successful operation of the ambr 15 requires its installation in a Biological Safety Cabinet (BSC), meaning this should be regarded as a permanently installed piece of equipment. Depending on circumstances a user may install the ambr 15 in an existing BSC or may purchase a model specifically for the ambr 15. In either case, the location should be considered carefully in relation to consumable storage, inoculum supply and off-line sample processing. We have installed a UPS system to ensure continued operation in the event of disruption to the electricity supply. Gas: Nitrogen, oxygen and carbon dioxide are required for bioreactor gassing, DO and pH control. These should be supplied at between 0.5 and 2 bar pressure. ambr 15 Clamp Plates: A single clamp plate is required for each culture station and must be autoclaved before use. Although these are initially supplied with the ambr 15 workstation, we have found it useful to purchase additional clamp plates in advance so there is no delay in the experiment schedule in the event of a replacement clamp plate being required because of blocking etc. ambr 15 Specific Consumables: These include 1 and 5 mL pipette tips, well plates for reagent delivery and tubes or plates to receive samples. The specific well plates used (e.g. 1 well, 24 well etc.) will depend on the number and volume of reagents required.

CHO Cell Cultivation in the ambr 15

45

The ambr 15 can sample into a variety of labware including plates, microfuge tubes, Vicell cups etc. depending on user preference. In principle any plates, tubes, or cups can be used if the specific geometries are loaded into the ambr 15 software and they are held in a suitable rack to fit on the ambr 15 deck. Additional Analytical Equipment: This will be required for off-line sample analysis and will depend on specific user requirements. However, users should consider: 1. Sample volume: It is good practice to reduce the volume of samples removed from the ambr 15 (sample volumes of 200–500 μL are typical) but these should be compatible with the volumes required by the analyzers. 2. Sample throughput: Typically, the ambr 15 will generate 48 samples for analysis per sample point; users should consider the analysis time (a) if the analyzer is a shared resource and (b) if the data is required to action further steps in the experiment (e.g. glucose control). For our experiments, we typically 1. Correct pH offsets using the integrated Analysis Module. 2. Use the integrated Vicell for cell counting. 3. Measure metabolite levels off-line using a CEDEX Bio HT analyzer.

3

Methods The ambr 15 is supplied with an extensive user manual and Sartorius engineers will provide user training during the installation which can be tailored to the initial experimental designs to be used by the customer (e.g. cell line screening, process development through DoE etc.). This chapter will not reproduce the Sartorius User Manual but seeks to provide an overview of the various steps required to complete an ambr 15 experiment and will highlight specific ways of working that we have implemented in our ambr 15 protocol for cell culture screening. In summary, undertaking cell culture in the ambr 15 can be split into several different activities. These are 1. Generating the experimental process: This is the series of steps that instructs the ambr 15 to perform specific operations, e.g. add media, sample bioreactor etc., i.e. how to run the experiment. 2. Preparing the ambr 15: This includes preparing reagents and consumables used during the experiment, cleaning the ambr 15 etc.

46

Steve R. C. Warr

3. Running the experiment: This includes any operations that must be carried out by the user during the experiment, e.g. off-line analyses and importing the data into the ambr 15, reloading consumables etc. 4. After experiment activities: These include collating and analyzing the experimental data as well as end of experiment machine operations e.g. cleaning etc. 3.1 Experiment Process

It is expected that ambr 15 users will have undertaken the vendorsupplied training and will be sufficiently familiar with the ambr 15 software to write an experimental process to meet their requirements. This section is not a process writing tutorial but highlights some specific conventions and ways of working that we have found useful when generating ambr 15 experimental processes. Experimental design is not addressed here; this will vary from user to user and can range from simple screening experiments where the majority of bioreactor parameters are kept constant to complex multifactorial DoE designs (see Note 1).

3.1.1 Process Overview

The key to a successful experiment in the ambr 15 is the experimental process. This is the series of steps that instructs the ambr 15 to perform specific actions in a specific order at specific times. The process can be evaluated using the ambr 15 software using the various checking tools provided, but time invested in writing and checking the process should ensure a successful experimental run (see Subheading 3.1.17). Writing a process will often involve translating an existing process from shake flasks or bioreactors to the ambr 15 and in these cases, some compromises may be necessary to account for the small volumes and physical limitations of the ambr 15 (see Note 2). We have found it useful to install the ambr software on an additional laptop not connected to the ambr. This enables new processes to be written and evaluated away from the ambr and then loaded onto the ambr at the start of the experiment using either a network connection or USB stick (see Note 3).

3.1.2 Nomenclature

Consistent and appropriate nomenclature should be used to ensure ambr 15 experiment data aligns with individual user’s data solutions, particularly where ambr 15 experimental data is compared to data from other scales of operation. We have implemented a system whereby each experiment, ambr 15 bioreactor and sample have a unique database friendly identifier which is then used to link the ambr 15 data to precursor (e.g. cell line development) and successor (e.g. product quality) data (see Note 4).

CHO Cell Cultivation in the ambr 15

47

3.1.3 Templates

Cell culture screening will generally be undertaken using a standard process which will often be a scaled down version of the eventual proposed manufacturing process and will therefore be repeatedly run in the ambr 15. In this instance, it is useful to create a template process which can then be used as the basis for each subsequent experiment. Although a new experiment can be directly created from a previous experiment and renamed appropriately, this has the disadvantage of including any additional ad hoc process steps that may have been added in the original experiment (see Note 5).

3.1.4 Labware

Individual plates are defined by editing the labware configuration in the Mimic tab of the software. It is useful to give each plate a simple descriptive ID, e.g. glucose feed, and if multiple reagents are used in the same plate to specify the relevant wells, e.g. AF 1-4. Glucose 7 to 24 would describe a 24-well plate containing antifoam in wells A1 to A4 and glucose in B1 to D6. Each plate ID must be unique and therefore, if multiple plates containing the same reagents are required, a unique suffix (e.g. culture station number) can be added. For instance, in our cell line screening experiments, we would typically load 4 identical plates of media for the initial bioreactor filling steps and define these as Media CS1, Media CS2 etc. Plate definitions should be created before the process steps for which they are required otherwise the plate IDs will not appear in the drop-down menu during the Insert Step/Edit Parameters state. It is important to consider the overall location of plates throughout the experiment and to associate each plate with the appropriate deck position as plates cannot easily be assigned to a different deck position without renaming. Plates cannot be deleted once they are specified within a particular process step.

3.1.5 Plate Location

Efficient usage of the ambr 15 plate decks and appropriate well plate mapping can minimize plate and liquid handler tip usage, reduce liquid handler operation time and reduce potential contamination risk. It is good practice to define frequently replaced plates (e.g. labile feeds) to positions at the front of the deck (see Note 6).

3.1.6 Plate Mapping

Plate mapping becomes increasingly important with complex processes which may require several additions of distinct feeds in addition to standard glucose feeding, pH control etc. Final plate mapping and plate type will depend on the volume of reagent required, but it is generally good practice to minimize the number of bioreactors associated with specific wells although this can increase tip usage and step duration. In addition, users may choose to map plates according to rows or to columns; this will depend on personal preference but, for simplicity, this should remain consistent for all plates throughout the process (see Note 7).

48

Steve R. C. Warr

3.1.7 Process Step Timing

Each process step is scheduled at a unique time and inserting a new step will be scheduled for 5 s after the previous one. Insertion of a step with an identical time is not permitted. Should this be attempted, the scheduled time will default to 1 s after the previous step. It is good practice to leave at least 5 s between each step as this allows additional ad hoc process steps to be inserted during the experiment if required e.g. to resample a specific bioreactor (see Note 8).

3.1.8 Process Step Duration

In the experiment process the scheduled times for each step are considered absolute times, not as relative offsets from each other, and steps are carried out in chronological order (see Subheading 3.1.9). As each liquid handling step has a duration dependent on the specific liquid handler operations required, we have found it useful to reflect this in the specified process step times. This ensures that subsequent user interactions, e.g. “load plate,” are scheduled to occur at a realistic time rather than inadvertently outside working hours. This is also useful for ensuring an appropriate time difference between different steps (e.g. pH adjustment after a feed) and for visualizing the experiment using the calendar view and process groups (see Subheading 3.1.10).

3.1.9 Process Step Priority

The priority function can be used to override the chronological completion of process steps by assigning a higher priority to critical process steps. If a step has a higher priority than the preceding steps it will be carried out before the lower priority preceding steps but only (a) after the current step has completed and (b) the scheduled time has been reached. We have used this to ensure that specific steps occur as soon as possible after the desired time, e.g. a timedependent pH change or feeding event.

3.1.10

The process group function can be used to collate a number of steps together in a single group. This has several advantages:

Process Groups

1. Process groups can be named to identify their function, e.g. inoculation, nutrient feed, etc. This enables them to appear in the calendar which can provide a convenient overview of the entire process. 2. Reschedule an entire group (i.e. multiple steps) in one go. 3. Group “copy and paste” saves individually copying multiple steps, e.g. for multiple sample times. 4. Process groups can also be exported to a file and imported to other experiments which can ensure identical methodologies are used for specific steps across different experiments. We routinely use process groups to identify the majority of liquid handling operations during an experiment i.e. media fill,

CHO Cell Cultivation in the ambr 15

49

AM (Analysis Module) pH adjustment, inoculation, sampling, cell counting, glucose feed, nutrient feed etc. 3.1.11 Process Step Scheduling

Although the ambr 15 allows process steps to occur outside of working hours with no manual intervention (apart from loading consumables, etc.), the extent to which users take advantage of this is a matter of personal preference. Similarly, the relative scheduling of different types of process steps should be considered e.g. samples for off-line metabolite analysis should be taken before any addition of nutrient feed. Typically, we run AM pH adjustment steps and antifoam additions during the night and then schedule sampling to occur so samples are available as soon as the operator arrives in the lab. It is good practice to perform any off-line analyses as soon as possible after sampling to ensure an accurate representation of the bioreactor culture. We have found it useful to perform any cell counting steps during the working day during which time any off-line sample analysis can be completed. Off-line sample data is then imported into the experiment when or before cell counting has finished and prior to any further feeding steps. Thus, feeding steps are typically carried out during the afternoon with the aim of completion before the end of the working day so that the operator can replenish consumables as required before leaving the lab.

3.1.12

DoE Tags

Associating DoE tags with a specific step simplifies the uploading of data acquired off-line (e.g. inoculum cell counts) and can also be used for uploading DoE designs via integrated Umetrics MODDE software. We routinely use DoE tags to upload the media and culture volumes required for pre-inoculation volume adjustment (see Note 9) and inoculation (see Note 10 and Subheading 3.1.15).

3.1.13

Pause Steps

Pause steps can be inserted into an experiment process as appropriate and can be useful if multiple users are running the experiment or a single user is handing over the experiment to another user part way through the experiment. The pause dialogue step can be used as a reminder of a required user intervention e.g. “Load more tips” or “Change Vicell Reagent Pack,” etc. or as a check to ensure a manual intervention has been carried out before the experiment continues e.g. to import off-line data that is required for subsequent steps using equations that require this data. The user pause step has a time-out function in which the user can specify when the experiment will continue even if the pause step has not been acknowledged. This can be useful if the pause steps are reminders to check parameters but less so if the experiment requires off-line data entry prior to continuation (see Note 11).

50

Steve R. C. Warr

3.1.14

Media Fill

Filling the bioreactor vessels with media should be undertaken at least 6 hours before inoculation to allow temperature and pH stabilization and to allow the pH and DO sensors to equilibrate. The initial media fill volume is dependent on user preference. Typically, we add 11 mL of medium (+ antifoam) to each bioreactor, assuming a 15 mL starting volume after inoculation, and then start stirring, temperature control, and “control DO pH” to enable the bioreactors to equilibrate overnight before inoculation. Typically, we monitor DO and pH in the bioreactors overnight which enables any faulty vessels or clamp plates to be easily identified prior to continuing the experiment. Volume corrections are carried out prior to inoculation if required (see Note 9).

3.1.15

Inoculation

Inoculation methodology will depend on user preference and also the type of experiment. For instance, a cell line screening experiment will require each bioreactor to be inoculated with a different cell line while a process development experiment using DoE will generally require a single cell line and therefore a single inoculum source (e.g. 1 well plate). Bioreactors can be inoculated with a fixed volume of inoculum or to a constant target VCC which will require the addition of different volumes to each bioreactor. The specific methodology will require different labware and process steps and should also take account of the total volume of each inoculum needed as well as the length of time the inoculum culture will remain on the deck prior to transfer to the bioreactors (see Note 10). Typically we inoculate bioreactors to a target iVCC having previously adjusted the medium volume as described in Subheading 3.1.14 (see Note 9) and usually inoculate groups of 6 bioreactors at a time to minimize the time the culture is exposed in the inoculum plate on the ambr 15 workstation deck.

3.1.16 Steps

Feed Addition

In larger vessels, the sample volumes removed during an experiment are usually small relative to the bioreactor working volume and feeding strategies are often gravimetric or % initial working volume based. In the ambr 15 the total sample volume removed during an experiment can be significant compared to the initial bioreactor working volume; this leads to divergent volume fluxes between the scales which should be accounted for in translating a large-scale feeding strategy to the ambr 15. We have compensated for this by implementing a volumetric feeding strategy based on the current bioreactor volume and further adjusted to compensate for the volume flux differences. This can be achieved using the “Add Liquid to Culture Vessel Percentage” or “Add Liquid to Culture Vessel Equation” steps (see Note 2).

Nutrient Feeds

CHO Cell Cultivation in the ambr 15

51

Glucose Feed

The trigger for glucose feeding will depend on the specific process but will often be when the glucose concentration falls below a defined level. Glucose measurement currently requires a sample to be removed from the bioreactor and the glucose level determined (either off-line or on the integrated Nova BioProfile Flex2 if available). To minimize sampling and analyses we have used the “Evaluate Equation” and “Add Liquid to Vessel Using Equation” steps to predict the current glucose level in a bioreactor from the previous glucose measurements and the consumption rate calculated from these. This enables closer control of glucose levels than would be achieved by feeding glucose after each sample point.

3.1.17

Once the user is satisfied with the process, it is important to check the process using the software tools provided. If the process has been written on a stand-alone laptop, this should be done prior to loading the experiment onto the control laptop. NB, if using a previously templated process, there should be no inconsistencies in the process although checking the process is still recommended. This can be done as follows:

Process Checking

1. Use the “Check Times” button in the Process Steps tab to confirm that the scheduled time of each individual process step is unique. 2. Use the “Check Layout” button in the Mimic tab to confirm that the correct plates have been defined. 3. Use the “Deck Map” button in the Mimic tab to confirm plate locations at specific times during the experiment. 4. Use the “Show Volume Prediction” option under “Advanced Options” in the Process Definitions tab to check that the bioreactor volume does not fall outside the recommended working range (e.g. 10–15 mL). 5. Use the “Process Step View” button in the Process Steps tab to visualize the ambr 15 deck over time by pressing the play button. This is a useful overview of the experiment and can be filtered to provide more detail on specific step types or well mappings etc. The above process checks can be completed off-line on any computer on which the ambr 15 software and specific process are loaded. Additional checks are automatically carried out by ambr 15 once the experiment is loaded onto the ambr 15 control laptop and initiated by pressing the “Start” button (see Subheading 3.3.1). 3.2 Preparing the ambr 15

Good laboratory and experimental practice should be followed at all times. It will be standard practice to record the batch numbers of all consumables and reagents used and to keep a record of any issues that occur on the machine during the experiment. It is sensible to

52

Steve R. C. Warr

ensure sufficient consumables and reagents to complete the experiment are available and ideally located in close proximity to the ambr 15. Lab layout obviously depends on the circumstances and preferences of each user but we have found it useful to segregate an area of the lab for ambr use only. This contains not only the ambr 15 but also consumable storage and dedicated ancillary equipment and analyzers required for sample analysis during the experiment. 3.2.1 Clamp Plates

Ensure the clamp plates have been cleaned immediately after their previous use following the recommended procedure described in the user manual (see Note 12). We have found it useful to confirm clamp plate operation by carrying out a pre-experiment gassing check with the clamp plates in situ (see Note 13). After completion of the gassing check, the clamp plates should be bagged, sealed and autoclaved prior to use. After autoclaving, place the sterile bagged clamp plates in an incubator at 37  C to ensure any condensation in the bags after autoclaving is removed (see Note 14).

3.2.2 Analysis Module

Check the levels of reagents and expiry date of electrodes and replace if necessary. It is good practice to recalibrate the AM prior to experiment start to ensure successful operation. In the AM tab under Current Status, it is good practice to tick the box “Automatic Retry pH Read” to ensure that, in the event of a pH read failure, a second sample will be analyzed immediately afterwards.

3.2.3 Vicell

Check the levels of Vicell reagent and replace if necessary.

3.2.4 Pre-experiment Cleaning

Clean the ambr 15 before use by wiping down with 70% ethanol (or equivalent). This should include the deck, undersurface of the plate nests, and LH and AM surfaces. The glass plate on the underside of each culture station should also be cleaned regularly between experiments using 70% ethanol (or equivalent) following the instructions in the user manual.

3.2.5 Load Experiment

The required experiment should be loaded onto the ambr 15. If the process is already present on the ambr 15 control laptop it can be selected from the list of experiments that will appear after selecting the “Load Experiment” button in the Experiment tab. The “Import Experiment” button allows the user to import an experiment generated elsewhere either via a network connection or a USB stick, alternatively the complete experiment file can be copied to the Experiment folder in the ambr directory on the C drive of the ambr 15 control laptop.

3.2.6 Initial Bioreactor and Consumable Loading

Bioreactors, pipette tips and some reagent plates can be physically loaded onto the ambr 15 before initiating the experiment. In this case, it is important to (a) refer to process steps to confirm deck

CHO Cell Cultivation in the ambr 15

53

loading and (b) retain the insert in the box of reagents as this is required after starting the experiment to enter the pH and DO spot calibration details. 3.2.7 Printed Plate Maps

It is useful to place hard copies of the plate maps adjacent to the ambr 15 so users can easily check the specific positions of reagents e.g. if reloading is required, without reference to the ambr 15 process. Reagent maps for each plate can be visualized and printed using the “Reagent Mapping” button in the Process Steps tab. However, where wells are used in multiple process steps (e.g. on different days), we have found it useful to print blank versions of the plate maps which then the users complete themselves. This is particularly useful for users not familiar with the process or feeding strategy etc.

3.3 Running an Experiment

This section describes some specific ways of working that we have implemented to facilitate smooth operation of our ambr 15s.

3.3.1 Starting the Experiment

After loading an experiment the initial screen presents a list of actions that the user is required to do during the course of the experiment e.g. process steps requiring user intervention, fill reagent plates with specific minimum volume of reagents etc. The user should complete any initial actions before starting the experiment. Typically, these will be loading the bioreactors, loading pipette tips and loading reagent plates required for the initial media fill. Once the initial actions are complete the experiment can be started by pressing the “Start” button. The user must then enter the pH and DO spot calibration details by scanning the QI Code insert as requested (details can also be entered manually). The user will then be asked to acknowledge that lids are in place on the appropriate labware and then the system will run through a series of checks to validate the process (see Note 15). Once these checks are completed the liquid handler will initialize after which the experiment will start at the first process step.

3.3.2 Loading Tip Boxes

For most processes it is likely that 1 mL pipette tip boxes will need to be replaced several times during the experiment. As tip boxes are used in the order in which they are loaded onto the ambr 15 we have found it useful to load the right-hand tip boxes first. These are used first and will therefore need replacing first which is easier to do for the right-hand ones. Users should ensure that sufficient tips are loaded to ensure all process steps can be completed until pipette boxes can be replaced. This can be easily confirmed as pausing and then resuming the experiment will generate a message with the time (i.e. process step) at which the tips will run out. (This is also useful for confirming when the Vicell reagent will run out.)

54

Steve R. C. Warr

3.3.3 Daily Checks

We have found it useful to implement a series of checks at the start and end of each day to ensure continued smooth operation of the process. These include reloading tip boxes if required, checking sufficient reagents and feeds are loaded and replenishing if necessary, ensuring no pause steps requiring user interaction are scheduled overnight etc. (see Note 16). End of day and start of day checks can be built into the process using a series of time-limited pause steps requiring user acknowledgment to continue or by hard copy sign-up sheets next to the instrument.

3.3.4 pH Offset Adjustment

Most ambr 15 systems will be supplied with an Analysis Module to allow the automatic adjustment of offsets of the pH spot readings. If the Analysis Module is not available, pH offset adjustment can be done through manual sampling (Paused Sample Liquid from Culture Vessel for pH), off-line pH check using a blood gas analyzer and then manually importing the pH offsets into the ambr 15 experiment. Due to pH spot drift, we have found that the pH control is improved by increasing the frequency of adjustment of the pH offsets. We typically use the AM to check and adjust the pH offsets each day.

3.3.5 Cell Counting

The use of an integrated Vicell enables automated cell counting; the data is recorded by the ambr 15 and can then be used in subsequent process steps e.g. to calculate cell-specific feed rates. Vicell operation can be monitored using remote desktop connection; however, to execute Vicell-specific actions, e.g. replacing the Vicell reagent pack, it is necessary to disconnect the Vicell from the ambr using the “Disconnect Cell Counter” button in the Cell Counts tab. During cell counting the Vicell can lose communication with the ambr 15; we have experienced cases where 1. Vicell stops counting (i.e. does not complete a cell count). Therefore, the process is effectively paused at this point. No further sampling occurs. 2. Cell count steps continue but no counts are recorded. 3. Cell count steps continue, counts are recorded on the Vicell computer and are stored in the text files, but the data is not transferred to the ambr 15 and the Vessel Data and Cell Count tabs are not updated. These issues are not immediately evident in the ambr 15 software screens. Therefore, during cell counting, we routinely monitor the Vicell images and text files using the remote desktop connection to the Vicell computer. We have found the most reliable way to rectify these issues when they occur is to

CHO Cell Cultivation in the ambr 15

55

1. Exit the ambr 15 software and switch off the control laptop, 2. Exit the Vicell software and switch off the Vicell computer, 3. Restart the Vicell computer, 4. Wait until the Vicell has initialized and then restart the ambr 15 control laptop and continue the experiment, and 5. Insert additional process steps to ensure all cell counts are performed for all bioreactors as required. Depending on the Vicell method, counting 48 bioreactors can take more than 5 h. We have found it useful to complete this during the working day where possible so that any Vicell disconnection issues can be addressed in a timely manner. 3.3.6 Importing Off-Line Data

Off-line data (e.g. glucose concentration) can be imported to the ambr 15 during the experiment. For an individual value this can be done by right-clicking on a vessel within the Vessel Data tab. Selecting vessel variables allows a biological value to be selected and then the value directly inputted by the user. If the biological parameter already exists in the Vessel Data table, additional values can be manually added by double-clicking in the appropriate column and entering the time and value as required. A more useful method for multiple bioreactors and/or parameters is to use the “Import Data” button in the Vessel Data tab. After clicking the “Import Data” button the user will be asked to load the relevant data from a csv file which can then be uploaded en masse to the ambr 15 experiment file. This requires the type of data in each cell to be defined as a bioreactor, a time, a parameter etc. We routinely do this using a template in which values in specific cells have been predefined. However, this can also be achieved manually by selecting the role in the “Assignments” key or by using the “Auto Detect” button. When the data is in the correct format, selecting the “Upload” button will transfer the data to the Vessel Data table.

3.3.7 Bioreactor Harvesting

We routinely retain 5 mL aliquots of culture supernatant at the end of an experiment to enable product quality and other analyses to be carried out. To streamline this process we include a harvest process group in the experiment process which is scheduled immediately after the final cell counting steps. This process group turns off pH control and vessel gassing to each culture station and then “stops stirring” to enable the clamp plate to be removed and bioreactors removed from the culture station. Using 3D printing we have manufactured centrifuge racks that hold up to six ambr 15 vessels. These fit in centrifuge well plate holders and enable direct centrifugation of the ambr 15 bioreactors without the need to transfer the culture to a separate centrifuge tube.

56

Steve R. C. Warr

After centrifugation vessel contents are decanted to 5 mL sample tubes and stored at 80  C prior to further analysis. 3.4 After Experiment Activities

Several activities must be completed when an ambr 15 experiment is completed so that the ambr 15 is ready for the next experiment. There will be other user-specific operations related to data analysis, collation, archiving etc.

3.4.1 Terminating an Experiment

After the last process step has been completed, the ambr 15 will terminate the experiment and the word “Stopped” will appear at the top of the software window. At this point, the bioreactors can be removed from the culture stations after removing the clamp plates (see Subheading 3.3.7).

3.4.2 Post-Experiment Cleaning

The ambr 15 workstation should be cleaned as in Subheading 3.2.4 above. It is important to clean the clamp plates as soon as possible after removal to ensure the removal of any cell culture debris collected during the experiment (e.g. after a foam out) using the procedure specified in the ambr 15 user manual (see Note 12).

3.4.3 Data Handling

The level of data and experiment archiving will depend on individual user requirements. The ambr 15 software does have built-in export functions allowing some or all of the data to be exported as required (see Note 17). Although the data is exported in a useable format (csv), the structure makes further analysis challenging without informatics input. We initially addressed this issue by developing a user interface that allowed the selection of experiments and association of metadata and was able to parse the instrument experiment folder directly. This provided a highly flexible set of raw data from the experiment on which customizable calculations could be performed, allowing time series and summary data to be produced on a per-experiment basis. This data could then be combined with configuration data and associated with additional analytical data (e.g. product quality) as required. Similarly, the data tables are amenable to DoE or PLS modeling through transformation and linear interpolation. Although this initial solution provided a viable work around for single experiments, further analysis across multiple experiments was problematical. Therefore, we have now developed an improved solution where ambr experiment files are parsed directly into a structured database where the calculation of derived variables is performed. Specific subsections are made available to a filtering and software package where cross experiment comparisons can easily be visualized.

CHO Cell Cultivation in the ambr 15

4

57

Notes 1. Experimental designs will vary from user to user and will depend on experimental objective. Simple screening experiments will generally alter a single factor e.g. cell line, base medium etc. while maintaining all other factors constant. DoE experiments are often used to facilitate process development or in other situations where it desired to alter multiple factors in the same experiment. Selection of the appropriate design can greatly increase the power of the experiment by simplifying subsequent data analysis. The ambr 15 software allows different process steps to be “DoE tagged” by right-clicking the appropriate process step and selecting DOE, Add Tag to Steps from the drop-down boxes. This allows the user to tag steps with different DoE identifiers. Clicking the “DOE Summary” button in the Process Steps tab shows a summary table of DoE tags for each bioreactor in which values can be manually edited by the operator. Alternatively, the user can create a DoE template in Excel, complete the design, and then import this to the ambr 15 as a csv file. We routinely use this method to adjust medium volumes prior to inoculation based on the cell counts of the inoculum (see Subheadings 3.1.14 and 3.1.15; see also Notes 10 and 11). More recently, Umetrics MODDE software has been integrated into the ambr 15 software which allows an ambr 15 process to be linked to an experimental design to create one or more experiments to be run on the ambr 15. After completion, the experimental results can be imported back into the MODDE software for analysis [14]. 2. Because of the small culture volume within the bioreactor, the following factors should be considered when transferring a process from conventional bioreactors or shake flasks to the ambr 15: (a) Frequency and volume of sampling (b) In a typical process the volume flux in the ambr 15 gradually diverges from that in larger bioreactors because of the sample volumes relative to the working volume and therefore feed volumes may need to be recalculated. For example, gravimetric-based feeding regimes must be simulated in the ambr 15. (c) The minimum addition volume using the liquid handler is 10 μL (20 μL for accuracy). This may necessitate the dilution of some reagents.

58

Steve R. C. Warr

(d) Headspace gassing is not possible in sparged ambr 15 vessels. (e) Foam accumulation is more rapid in the ambr 15, and therefore, more frequent antifoam addition may be required. 3. A new process (i.e. experiment) can be created within the ambr 15 software either by using the experiment wizard or by using the previously loaded experiment as a basis for the new experiment. We have found it more useful to create a template process outside of the ambr 15 control laptop and then load this onto the ambr 15 workstation (see Note 5). This can be done either by copying an entire experiment file to the ambr 15 control laptop or by separately importing the Process Definition and Process Layout as separate files. 4. We have implemented a system whereby each experiment, bioreactor and sample has a unique identifier. In this system, Experiments are identified by the scale and year and numbered sequentially, i.e. AM19-001. Bioreactors are identified by their ambr 15 culture station position, i.e. CS1 positions 1 to 12 ¼ bioreactors 1 to 12, CS2 positions 1 to 12 ¼ bioreactors 13 to 24, etc. Samples are also identified numerically e.g. sample 01 ¼ the first sample removed from a bioreactor, 02 is the second sample, etc. These are suffixed A, B, C, etc. to identify replicates. Sample type is identified by a further suffix, e.g. s ¼ supernatant, etc. Thus, in the identifier AM19-003-015-02A-s, AM19 ¼ Scale and year (ambr 15 2019) 003 ¼ Third ambr 15 experiment in 2019 015 ¼ Bioreactor 15 (i.e. CS2 position 3) 02A ¼ Second sample removed from this bioreactor (A denotes first replicate if required) s ¼ supernatant 5. Creation of template process ensures that the same experimental design and process steps are run consistently for each experiment. A master copy of the template can be created and used as the basis for each new experiment. At this point, it is useful to save the process with descriptive name (e.g. Production Process 1_Template). This is also useful to ensure process step timings, plate mapping etc. are consistent if the same process is run on different ambr 15 workstations. 6. An ambr 15 48-way workstation has 7 positions available for plates (see Fig. 1). It is important to consider the positioning of different plates during the experiment to facilitate user

CHO Cell Cultivation in the ambr 15

59

Fig. 1 ambr 15 48-way Deck Map

interventions for plate loading and reagent refilling. In our experimental processes we routinely position plates requiring minimum intervention (e.g. cell counter diluent) in Decks 1, 2, and 7; plates that require frequent changes are positioned at the front of the machine for ease of access e.g. the user can easily remove or load a plate to Deck 3 without reaching across the bioreactors or pipette tip boxes or interfering with the liquid handler. In our protocols, Deck 1 typically has a 1-well plate containing diluent for the Vicell, and Deck 2 is a 24-well plate containing antifoam and base. These plates can be loaded at the start of the experiment and do not require replacing during the experiment. Sample plates which must be removed and replaced at every sample point are typically positioned in Decks 5 and 6. 7. The exact plate mapping will be dependent on user preferences i.e. how many bioreactors to supply from the same reagent stock solution (i.e. plate well), whether to map using plate rows, columns etc. It is generally good practice to minimize the number of reactors supplied with media, feed or reagent from a particular plate or well although this may mean slightly more complex initial plate mapping to ensure the desired well/ bioreactor linkages. However, practical considerations such as well capacity, volume required, deck space, and reagent type will also influence the type of plate used and the associated mapping. Apart from initial media additions to the bioreactors and the cell counter diluent, we routinely use 24-well plates for the following additions: pH Control (Base): Although a single 1-well plate could be used to supply base to all 48 bioreactors, we use a 24-well plate and supply a single culture station from each well for approximately 25% of the experiment duration using the “Wells Open” parameter in the process step (i.e. 16 wells in total contain base).

60

Steve R. C. Warr

Antifoam: Volumes of antifoam used tend to be small and we use 4 wells of a 24-well plate (typically the same plate that is used for base) for antifoam with one culture station (i.e. 12 bioreactors) supplied per well. Glucose: This could also be supplied from a single stock solution in, e.g. a 1-well plate. However, to minimize any possible contamination issues we use a 24-well plate and use each well to supply 3 bioreactors (i.e. 16 wells in total). This leaves 8 spare wells that can be used for ad hoc additions if required. Nutrient Feed: A 24-well plate is also used for nutrient feeds. In processes requiring multiple feeds and depending on storage requirements/stability of the feeds, it may be necessary to reload the feed solutions immediately before each feed. In this case it is good practice to keep the same interactions between bioreactor and plate row or column. For instance, Day 1 feed: Well A1 ! Culture Station 1, Bioreactor 1–12 Well A2 ! Culture Station 2, Bioreactor 1–12 Well A3 ! Culture Station 3, Bioreactor 1–12 Well A4 ! Culture Station 4, Bioreactor 1–12 Day 2 feed: Well B1 ! Culture Station 1, Bioreactor 1–12 Well B2 ! Culture Station 2, Bioreactor 1–12 Well B3 ! Culture Station 3, Bioreactor 1–12 Well B4 ! Culture Station 4, Bioreactor 1–12 Day 3 feed: Wells C1 to C4, etc. For processes with more than 4-feed addition time points, it is necessary to reload the feed plate. In summary, our general plate usage philosophy is shown in Table 1 below. Table 1 Typical plate type usage for ambr 15 reagents Reagent

Plate type

Mapping

Media

1 well (or 4 well)

Maximum of 1 culture station per plate

Nutrient feeds

24 well

1 culture station per well. Separate wells for different feed time points (e.g., A1–A4 for day 1 feed, B1–B4 for day 2 feed, etc.)

Antifoam

24 well

1 culture station per well

Base

24 well

1 culture station per well/100 h

Glucose

24 well

3 bioreactors per well

CHO Cell Cultivation in the ambr 15

61

8. Although a process could be written with each step scheduled for 5 s after the previous one, this would mean each step being completed immediately in sequence. Because liquid handling steps take a finite time, we reflect these times (at least approximately) in the scheduling, i.e. to sample 48 bioreactors typically takes approximately 45–60 min; and although we would schedule sampling from each culture station at 5 s intervals to minimize any delays between steps, we would not schedule the next step (e.g. cell count) until at least 45 min after the start of sampling. We have found this particularly useful when grouping steps so they appear in the calendar view as an approximation of step (and therefore group) duration; this enables a realistic overview of the experiment and a good approximation of the timing of events relative to each other. 9. Adding media to bioreactors is quicker than removing it (because a 5 mL tip can be used to dispense medium to multiple bioreactors) and therefore we usually underfill the bioreactors and then volume correct upward prior to inoculation. As we typically require between 3000 and 4000 μL of inoculum culture, we fill the bioreactors with 11000 μL of medium and then volume adjust upward with new medium prior to inoculation. We have developed an “Inoculation Calculator” template in Excel which we use to calculate the required volume adjustments and inoculum volume; these are then imported to the ambr 15 using DoE tags for media addition, for removal and for inoculation. This requires the process to include “Add Liquid to Culture Vessel” and “Sample Liquid from Culture Vessel” steps prior to inoculation. These steps, as well as inoculation, should be DoE tagged Removal, Addition and Inoculation respectively. Thus, assuming a desired 15000 μL working volume and requiring an inoculum volume of 3700 μL, the media fill steps would be as follows: (a) Fill bioreactors with 11000 μL medium. (b) Leave overnight to equilibrate. (c) Perform a pH offset adjustment using the Analysis Module. (d) Measure the cell count of the inoculum culture. (e) Use the Inoculation Calculator to calculate the volume of inoculum and volume of media required, i.e. current bioreactor volume ¼ 11000 μL, inoculum vol required ¼ 3700 μL, therefore additional medium required ¼ 300 μL.

62

Steve R. C. Warr

(f) Import volumes to ambr 15 using DoE Summary in the Process Steps view. (NB ensure DoE tags in the process are identical to those in the template.) (g) Process will then insert the appropriate volumes into the media addition, removal and inoculation steps. This is useful where multiple cell lines have different cell counts and therefore, different bioreactors will require different volume adjustments and inoculation volumes. The use of DoE tags avoids the need to manually input the required volumes thereby saving time and avoiding transcription errors. 10. Typically we inoculate in groups of 6 bioreactors in which each cell line is loaded to a single well of a 24-well plate; this ensures adequate mixing of the culture prior to transfer to the bioreactor and ensures the cells are exposed to the uncontrolled conditions in the plate for approximately 10 min or less. Thus, in summary, our typical inoculation process would be as follows: (a) Load wells A1 to A6 of a 24-well plate with cell lines in a biosafety cabinet. (b) Physically place plate on the correct deck position in the ambr 15. (c) Acknowledge pause step confirming inoculum is in place (this should be in the process). (d) The ambr 15 will inoculate the first 6 bioreactors in turn (e.g. CS1-1 to CS1-6) (specify a new 5 mL tip for each bioreactor). (e) During inoculation load wells B1 to B6 of a second 24-well plate with the next 6 cell lines. (f) Once the first 6 bioreactors have been inoculated, physically replace the 24-well inoculum plate with the second plate. (g) Acknowledge pause step to confirm inoculum is in wells B1 to B6. (h) The ambr 15 will inoculate the next 6 bioreactors (e.g. CS1-7 to CS1-12). (i) Repeat steps 5–8 until all 48 bioreactors have been inoculated. The use of two 24-well plates in parallel reduces the overall inoculation time as one plate can be loaded with inoculum while the bioreactors are being inoculated from the other plate loaded on the ambr 15. It is important to include multiple pause steps in the inoculation process group to ensure that transfer of inoculum from plate to bioreactor only occurs once

CHO Cell Cultivation in the ambr 15

63

the inoculum plate is in place on the ambr 15 with the inoculum in the correct wells. For experiments in which a single cell line is to be inoculated into multiple bioreactors a 1-well plate could be used although users should ensure that the aspiration of the culture via the liquid handler prior to inoculating each bioreactor is sufficient to ensure a uniform culture suspension. Furthermore, inoculation of 48 vessels, even using the same pipette and a single dispense per bioreactor, will take a significant length of time (approximately 1 h) during which time the culture will be on the deck in a non-temperature-controlled, non-gassing environment. Therefore, users may prefer to use a 4- or 24-well plate and manually load the inoculum culture from the source flask immediately before inoculation (i.e. using the same procedure as described above for multiple cell lines but loading the same cell line to each well). 11. The frequency of pause steps will depend on user preferences. Judicious use of pause steps is a useful way to ensure user interactions (e.g. manually refilling a plate) are carried out when required. Typically, we use pause steps before each nutrient or glucose feed step as a reminder to reload the feed plate or to import off-line glucose data which is required by subsequent steps to calculate the glucose feed volume required. In these instances we set the time-out to zero, meaning the experiment would pause indefinitely until the step is acknowledged. We have found this more useful than setting a finite time-out (e.g., 1800 s) in which case the pause step would be overridden after 30 min as a user may be unavailable during that time and unable to load the most recent off-line data prior to glucose feeding which would result in feed volumes being calculated from previous (out-of-date) data. We also use pause steps as a check for users to confirm that required interactions have been completed. For example, immediately before inoculation, we include a pause step asking the user to confirm the inoculum plate containing cells is loaded onto the deck. Again, a time-out of 0 s ensures the experiment will not continue until the step is acknowledged. Some users may also use pause steps each day to confirm that user interventions such as tip loading, emptying the tip waste etc. have been completed. We do not typically use pause steps to instruct a user to load more tips because although tip usage for many steps is defined, steps such as pH control or those requiring equation-defined operations (e.g. based on a cell count or imported metabolite value) will use varying numbers of tips depending on the data.

64

Steve R. C. Warr

12. Thorough cleaning of the clamp plates immediately after each experiment is critical to ensure they do not become blocked with any media etc. during the experiment. The cleaning procedure detailed in the user manual and clamp plate cleaning guide [15, 16] involves using a syringe to pass water, 70% alcohol, and then air through each channel in the clamp plate. It is recommended this is done as soon as possible after the end of the experiment to prevent any liquid contamination drying within the channels. 13. To check clamp plate function in situ, a gassing check can be carried out before sterilization of the clamp plate. This should be carried out as follows: (a) Clean the clamp plate as described above (see Note 12). (b) Place 6 test bioreactors containing approximately 10 mL of water in positions 1 to 6 (i.e. back row) of each culture station. (c) Place the clamp plates on the culture stations and tighten the clamp plates nuts using the tightening tool provided. Ensure these are not overtightened—an additional half turn after finger tightening is usually sufficient. (d) Within the Setup tab, set vessel gassing for each clamp plate in turn. It is usually sensible to set the gas flow rate equivalent to the lowest flow rate to be used in the experiment. To access the Setup tab, it may be necessary to unlock the set up using the “Unlock Setup” button in the Configuration tab. (e) Observe gas bubbles in each of the bioreactor vessels. This is usually easier by reducing the ambient background light and by turning on the system lights using the “Turn Lights On” button in the Overview tab. Alternatively, a torch or mobile phone light can be used. (f) Once bubbles have been observed consistently in each bioreactor, remove the clamp plates and move the test bioreactors to positions 7 to 12 (i.e. front row) of each culture station. Resite the clamp plates using the tightening tool as above. NB it is not usually necessary to stop gassing during this stage. (g) Observe gas bubbles in each of the bioreactor vessels as above. (h) Consistent gas flow indicated by the presence of gas bubbles in the bioreactors indicates correct clamp plate function. 14. Clamp plate preparation—The pivot holes in the clamp plates must all be aligned in the same direction prior to installing on the culture station so that the stirrer plate can engage correctly

CHO Cell Cultivation in the ambr 15

65

with the clamp plate. Pivot hole alignment can be carried out before autoclaving using a 200 μL pipette tip to rotate the individual pivot assemblies. This procedure may require repeating after autoclaving to correct any movement that has occurred during autoclaving and unwrapping the clamp plate in the cabinet. 15. To ensure the “Starting Experiment Checklist” is completed without interruption, the user should: (a) Have completed the process checks described in Subheading 3.1.17. (b) Ensure that any integrated equipment (e.g. Analysis Module, Vicell) is switched on, connected and ready for use and that any required user interactions are specified for working hours. (c) Completion of the checklist requires acknowledging a number of tick boxes. (d) Although the process checks described in Subheading 3.1.17 will have been completed, the Starting Progress Checklist includes the physical connections to the ambr 15 workstation (e.g. gas pressure etc.). 16. We do not typically include a user pause step to confirm checks completed. Hard copies of our checklists are placed adjacent to the ambr 15 workstations (Table 2). 17. If a user-developed bespoke data solution is not available, the built-in export functions in the ambr 15 software can be used. An entire experiment can be exported to another computer running the ambr 15 software using the “Export Experiment” button in the Experiment tab. Alternatively, the relevant experiment file can be directly copied from the ambr 15 control computer C drive. However, these can be large files and contain all data associated with an experiment, most of which will not require analysis. Specific data can be exported as csv files using the “Export Settings” and “Export Data” buttons in the Results tab. File size can be significantly reduced by excluding data points that are identical or very close to previous values according to userspecified limits. However, the resulting table will usually require reformatting before further analysis. The Results tab enables specific data sets to be visualized within an experiment and, if required, can be copied using Print Screen or the Snipping Tool. Because none of these solutions is entirely satisfactory, we strongly recommend users to develop their own data solutions to facilitate ambr 15 data analysis.

66

Steve R. C. Warr

Table 2 Start of day and end of day checklist for ambr 15 Start of day checklist 1 Check for error messages using the Status tab and that the status is running. 2 Check individual bioreactor profiles to ensure they are operating within the specified control parameters. 3 Visually inspect the workstation to check if bioreactor, plate, and tip box lids are in the expected positions. 4 Using the Process Steps tab, check that previous steps have been completed as expected. 5 Empty the used tip container. 6 Check future process steps and reload tip boxes, plates and cell culture reagents that will be required during the day. Pipette tip boxes must be physically loaded and then reloaded in the ambr 15 software via the Mimic tab. 7 Check for any upcoming user pause steps and take appropriate action. End of day checklist 1 Check any steps that are scheduled to be performed overnight. 2 Reload tip boxes, plates and reagents as required. Pipette tip boxes must be physically loaded and then reloaded in the ambr 15 software via the Mimic tab. 3 Empty the used tip container. 4 Check that no user pause steps are scheduled during the night. 5 Check the ambr 15 workstation status is running.

References 1. Rameez S, Mostafa SS, Miller C, Shukla AA (2014) High-throughput miniaturized bioreactors for cell culture process development: reproducibility, scalability and control. Biotechnol Prog 30:718–727. https://doi.org/ 10.1002/btpr.1874 2. Moses S, Manahan M, Ambrogelly A, Ling WLW (2012) Assessment of AMBRTM as a model for high-throughput cell culture process development strategy. Adv BioSci Biotechnol 3:918–927. https://doi.org/10.4236/ abb2012.37113 3. Velez-Suberbie ML, Betts JPJ, Walker KL, Robinson C, Zoro B, Keshavarz-Moore E (2017) High throughput automated microbial bioreactor system used for clone selection and rapid scale-down optimization. Biotechnol Prog 34:58–68. https://doi.org/10.1002/ btpr2534

4. Priola JJ, Calzadilla N, Baumann M, Borth N, Tate CG, Betenbaugh MJ (2016) Highthroughput screening and selection of mammalian cells for enhanced protein production. Biotechnol J 11:853–865. https://doi.org/ 10.1002/biot.201500579 5. Velugula-Yellela SR, Williams A, Trunfio N, Hsu CJ, Chavez B, Yoon S, Agarabi C (2017) Impact of media and antifoam selection on monoclonal antibody production and quality using a high throughput micro-bioreactor system. Biotechnol Prog 34:262–270. https:// doi.org/10.1002/btpr.2575 6. Sartorius Application Note (2015) ambr 15 cell culture perfusion mimic. Available via Sartorius.com. https://www.sartoriusglobal. com/_ui/images/h4f/he6/ 8875009998878.pdf

CHO Cell Cultivation in the ambr 15 7. Sewell DJ, Turner R, Field R, Holmes W, Pradhan R, Spencer C, Oliver SG, Slater NKH, Dikicioglu D (2019) Enhancing the functionality of a microscale bioreactor system as an industrial process development tool for mammalian perfusion culture. Biotech and Bioeng 116(6):1315–1325. https://doi.org/ 10.1002/bit.26946 8. Sartorius Application Note (2015) ambr 15 cell culture microcarriers. Available at Sarto rius.com. https://www.sartoriusglobal.com/_ ui/images/haa/he5/8875816058910.pdf 9. Nienow AW, Hewitt CJ, Heathman TRJ, Glyn VAM, Fonte GN, Hanga MP, Coopman K, Rafiq QA (2015) Agitation conditions for the culture and detachment of hMSCs from microcarriers in multiple bioreactor platforms. Biochem Eng J 108:24–29. https://doi.org/10. 1016/j.bej.2015.08.003 10. Sartorius Application Note (2015) ambr 15 cell culture transient transfection. Available at Sartorius.com. https://www.sartoriusglobal. com/_ui/images/h7a/h0f/ 8876496945182.pdf 11. Umetrics Suite Blog (2018) Design of experiments enables the optimization of transfection

67

efficiency in line with QbD principles. Available via Umetrtics.com. https://blog.umetrics. com/design-of-experiments-enables-qualityby-design-for-bioprocess-development 12. Janakiraman V, Kwiatkowski C, Kshirsagar R, Ryll T, Huang YM (2015) Application of highthroughput mini-bioreactor system for systematic scale-down modelling, process characterization and control strategy development. Biotechnol Prog 31:1623–1632. https://doi. org/10.1002/btpr.2162 13. Sandner V, Pybus LP, McCreath G, Glassey J (2018) Scale-down model development in ambr systems: an industrial perspective. Biotechnol J 14(4):e1700766. https://doi.org/ 10.1002/biot.201700766 14. Sartorius Document: TAP-9670-06-020 (2014) ambr 15 DOE software reference manual 15. Sartorius Document: ambr 15 cell culture user manual (2016) TAP-9670-06-005 16. Sartorius Document: ambr 15 clamp plate cleaning guide (2018) TAP-9670-06-14

Chapter 5 Using a Parallel Micro-Cultivation System (Micro-Matrix) as a Process Development Tool for Cell Culture Applications Vincent Wiegmann, Cristina Bernal Martinez, and Frank Baganz Abstract Micro-bioreactors appear frequently in today’s biotechnology industry as screening and process development tools for cell culture applications. The micro-bioreactor’s small volume allows for a high throughput, and when compared to other small-scale systems, such as microtiter plates, its measurement and control capabilities offer a much better insight into the bioprocess. Applikon’s micro-Matrix is one of the microbioreactors that are commercially available today. The micro-Matrix system consists of shaken disposable 24 deep square well plates in which each well is controlled individually for pH, dissolved oxygen (DO), and temperature. Additionally, a feeding module supports automated additions of liquid to each well. This chapter describes how the micro-Matrix can be used for fed-batch cultivations of Chinese Hamster Ovary (CHO) cells. Key words Miniature shaken bioreactor, Micro-Matrix, Fed batch, GS-CHO cells, Scale-down

1

Introduction Throughput is a decisive factor in driving the development of upstream processes forward. Conventional stirred tank bioreactors (STRs) are unpractical for extensive optimization studies due to their slow turnover and high spatial demand. Typically, microwell systems are used for early-stage development, such as clone screening, before employing shake flask cultures for media development or to design feeding strategies [1]. Although both systems allow for moderate to high throughput, a lack of monitoring and control curtails the information density in comparison to STRs. Microbioreactors bridge this gap between bioreactors and microwell systems by offering a fully-controlled culture environment on a small scale. This allows for the quick accumulation of process knowledge early on in the development timeline, which can speed up development while keeping media-related costs to a minimum [2].

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_5, © Springer Science+Business Media, LLC, part of Springer Nature 2020

69

70

Vincent Wiegmann et al.

Several developments were instrumental in the recent advancement and success of micro-bioreactors. Monitoring and control of pH and DO at the small scale have been enabled by the availability of affordable, small, noninvasive sensors. Additionally, routine off-line analyses such as antibody titer, viable cell concentration, and metabolites can now be performed at submilliliter volumes, which allows to fully utilize the high degree of parallelization. Micro-bioreactors also frequently incorporate single-use materials, which substantially reduce the turnover time when compared to glass or stainless steel vessels and allow runs to be scheduled in rapid succession [3]. In general, micro-bioreactors can be divided into either shaken or stirred systems. The design of stirred systems (e.g., ambr15, bioREACTOR) follows similar operational principles to largerscale vessels where a stirrer is inserted into the liquid from above and a sparger provides aeration. Shaken systems (e.g., microMatrix, BioLector), on the other hand, rely on overhead aeration [4, 5]. However, to date, only the ambr15 and the micro-Matrix have been documented in the literature for cell culture applications [2, 5–8]. Micro-bioreactors exist with various degrees of automation. In many cases, liquid-handling robots can be integrated with the micro-bioreactor system (e.g., RoboLector, ambr15) to facilitate routine interactions such as sampling or the addition of feed medium [6, 7, 9]. However, alternative approaches have also been explored. For example, Applikon’s micro-Matrix enables automated liquid additions by connecting each culture compartment to individual liquid lines, allowing for near-continuous feeding without running the risk of overloading the schedule of a pipetting robot [10]. An overview of today’s commercially available microbioreactors options is given in Table 1. Applikon’s micro-Matrix (Applikon, Delft, the Netherlands) is a platform that holds 24 individual micro-bioreactors based on a 24 deep-well cassette with 4 independent gas inlets for overhead aeration and one liquid feed line. The built-in orbital shaker operates at shaking speeds between 0 and 400 rpm at an orbital throw of 25 mm. The cassette is placed in a temperature-controlled chamber at a defined set point. Additionally, the temperature can be controlled individually for each well, with a maximum difference of 1  C between adjacent wells [11, 12]. For cell culture processes, the DO is typically controlled with nitrogen and air. As evaporation is of particular concern in the small scale, pure oxygen can be used instead of air to further decrease gas flow rates and, by extension, minimize evaporation. Control of the pH is achieved through overlay with carbon dioxide. Furthermore, the liquid feed line can be used for automated additions of base. The parameters DO, pH, and temperature are maintained through PID control loops that can be customized for each well individually.

Using a Parallel Micro-cultivation System as Process Development Tool

71

Table 1 Comparison of currently available commercial micro-bioreactor options Microbioreactor

Working Agitation Aeration volume

Special Vessel format features

ambr 15

Stirred

Sparger

10–15 mL

Disposable microbioreactor vessels

BioLectora

Shaken

Overlay

0.8–2.4 mL 48 Can be FlowerPlate combined with liquid handler

m2p-labs

bioREACTORa Stirred

Overlay

8–15 mL

Disposable PS Can be vessels combined with liquid handler

2mag

micro-Matrix

Overlay

1–5 mL

24 deep square well plate

Shaken

Vendor

Automated liquid handling

Automated feeding module

Sartorius-Stedim

Applikon Biotechnology

a

Cell culture applications have not yet been documented for this device

The protocol outlined here describes how the micro-Matrix can be used for fed-batch cultivations of CHO cells, where the automated feeding module achieves the feed additions.

2 2.1

Materials Cell Culture

1. CHO cell line (see Note 1). 2. Chemically defined basal and feed medium (see Note 1). 3. Bicarbonate buffer: 250 mM Na2CO3 and 250 mM NaHCO3.

2.2

Preculture

1. Shake flask (see Note 2). 2. Certomat MO II shaker with 25 mm orbit (Sartorius). 3. MCO-18M O2/CO2 Incubator (Sanyo).

2.3 Micro-Matrix Cultures

1. micro-Matrix micro-bioreactor Biotechnology).

system

(Applikon

2. Disposable micro-Matrix cassette (Applikon Biotechnology) with pre-calibrated optical sensors for pH and DO. 3. Disposable micro-Matrix filter bars (Applikon Biotechnology).

72

2.4

Vincent Wiegmann et al.

Cell Count

1. Vi-CELL XR (Beckman Coulter).

2.5 Analysis of Nutrient and Metabolite Concentrations

1. BioProfile Flex (Nova Biomedical).

2.6 Quantification of Antibody Titer

1. High-performance liquid chromatography (HPLC) (see Note 3). 2. Protein G column (see Note 3). 3. Phosphate buffer at pH 7 for loading and washing step. 4. Glycine buffer at pH 2.8 for elution.

3 3.1

Methods Preculture

1. Remove a vial from the working cell bank stored in liquid nitrogen, and thaw in a water bath at 37  C (see Note 4). Transfer vial content into a falcon tube, and dilute 10 times with fresh medium. Centrifuge at 1000  g for 5 min, and discard the supernatant. Resuspend the cell pellet in CD-CHO with 25 μM Methionine sulfoximine (MSX), and transfer into a shake flask. Adjust the cell concentration to 0.3  106 cells mL1. 2. Expand the cells on a shaker in a humidified incubator at 37  C, 5% CO2, 95% air, and 160 rpm. Passage cells regularly, and observe growth for at least 7 days prior to starting the experiment (see Note 5).

3.2

Initial Setup

1. Turn on the system via the switch on the right-hand side of the machine. 2. Connect gas supply of oxygen, nitrogen, and carbon dioxide to three of the gas inlet connectors at the back of the device. Using regulators, the inlet pressure should be set between 2 and 6 bar (see Notes 6 and 7). 3. Connect the Ethernet RJ45 connector to a PC with the control software.

3.3 Calibrating the Feeding Module

The micro-valves of the Liquid Addition Top Plate need to be calibrated for every liquid individually. This procedure has to be performed prior to autoclaving the module (see Note 8). 1. Fill the Liquid Addition Feed Bottle (LAFB) with the liquid that the micro-valves need to be calibrated for, and connect the LAFB to the compressed gas outlet inside the micro-Matrix cabinet (see Note 9).

Using a Parallel Micro-cultivation System as Process Development Tool

73

2. Connect the Liquid Addition Top Plate to the LAFB, place it over a waste container, and connect the control box to the delivery module. 3. In the Human Machine Interface (HMI) software, click “Actuator and PID setup” ! “Liquids Library.” Create a new liquid, insert a name, and select the pulse time (see Note 10). Select the number of pulses and click “Start calibration” to start the delivery of liquid. Run the calibration until all lines have been primed. 4. Now, place the feeding module on a rack with pre-weighed centrifuge tubes and repeat step 3. 5. After all pulses have been delivered, weigh the centrifuge tubes and calculate the delivered volume by taking the density of the liquid into account. Dividing with the number of pulses, the volume per pulse can be calculated and must then be inserted in the “Pulse volume” section. 6. Save the values by clicking the green tick. 7. Before autoclaving the feeding module, it needs to be cleaned as outlined in Subheading 3.10 to avoid feed residues blocking the tubing or micro-valves. 3.4 Calibration and Recalibration of pH Sensors

1. After defining the Recipe, the calibration values can be entered by clicking “Calibrate the micro-Matrix” on the main screen and either scanning the QR code on the cassette, entering the batch number of the cassette, or manually entering the calibration values. 2. Before starting the experiment, the pH offsets should be determined by aseptically filling each well of the cassette that will be used for the experiment with at least 2 mL of medium and mounting the assembled cassette onto the micro-Matrix without connecting the gas lines (see Note 11). 3. Initiate pH measurements by starting the experiment without agitation, DO, pH, or temperature control. Let the probes equilibrate for 1–2 h or until the measurements reach a plateau (see Note 12). 4. Aseptically remove an appropriate amount of the medium from each to perform off-line pH measurements, and note the off-line value for each well. 5. Continue the run with the residual medium still in the wells. 6. While the device is running, click “Calibrate the microMatrix,” select the “pH measurement calibration” tab, and tick the box “Upload off-line pH values.” Enter the off-line values for each well and click OK (see Note 13).

74

3.5

Vincent Wiegmann et al.

Inoculation

1. Perform a cell count of the preculture. Based on the viable cell density, prepare the inoculum with a final cell count of 0.3  106 viable cells mL1 by diluting with fresh basal medium (see Notes 1 and 14). 2. In the laminar flow cabinet, remove the cassette carefully from the packaging. The cassette has a plastic cover, which has to be removed without compromising the sterility of the plate (see Note 14). 3. Aseptically take up the well-mixed inoculum (shake before use) using a pipette, and dispense the required volume into each well of the cassette (see Note 15). 4. Aseptically fill the LAFB with approximately 200 mL of the feed medium (this can also be done at a later stage of the process to avoid degradation of the medium). 5. Now, the lines of the Liquid Addition Top Plate need to be primed. To do so, pressurize the bottle using a syringe, and place the feeding module over a sterile container (e.g., the plastic lid of the cassette). Connect the priming module and turn the switch to the “On” position. All valves will now successively open and dispense some of the liquid. After two cycles, the priming module can be dismounted. 6. Then, place the Liquid Addition Top Plate onto the cassette and use the two metal clamps to hold it in place. Note that the cover only fits in one orientation. Row D should align with the “Front” written on the liquid feeding module. 7. Aseptically remove the filter bars from the packaging and fit them onto the Liquid Addition Top Plate. Small plastic fittings indicate the correct orientation of the filter bars. (Note: Force may be necessary to slot the gas bars in place.) 8. Mount the assembled cassette onto the micro-Matrix, use the transport clamps to hold the cassette in place, and attach the gas bars to the filter bars. Finally, fasten the control box to the Liquid Addition Top Plate (see Note 16).

3.6 Actuator Selection

1. Click “Actuator and PID setup” in the main screen to open the configuration menu as shown in Fig. 1. 2. Select “24 wells cassette” under “Cassette layout.” 3. Select the appropriate liquid from the drop-down menu (see Subheading 3.3 for calibration of the feeding module) under “Liquids configuration.” 4. Highlight one or more wells in the “Reactor selection” field. Any changes made in the following fields will be valid for all selected wells.

Using a Parallel Micro-cultivation System as Process Development Tool

75

Fig. 1 Actuator setup in the micro-Matrix HMI

5. Select a suitable PID setup from the drop-down menu under “Control configuration.” The PID setting can be customized for each well individually if necessary (see Notes 17–20). 6. Define a cutoff volume for the liquid delivery. 7. In the “Actuator configuration” field, tick the boxes of the gases that will be used in the experiment and assign them to their corresponding actions via drag and drop (see Note 7). 8. Levels can be populated if a cascade control chain is desired. 9. Click OK to save and return to the main screen.

76

Vincent Wiegmann et al.

Fig. 2 Setup of the experimental conditions in the micro-Matrix HMI. (a) Dissolved oxygen (dO2), pH, temperature and shaking speed. (b) Liquid additions 3.7

Recipe Setup

The micro-Matrix software allows for a range of process conditions and feeding regimes. The following instructions describe a fed-batch protocol in which feeding commences after 72 h of cultivation with a constant rate of 100 nL min1 (see Note 21). 1. Click “Recipe” in the main screen to open the configuration menu as shown in Fig. 2. 2. Define the desired set points for DO, temperature, pH, and shaking speed.

Using a Parallel Micro-cultivation System as Process Development Tool

77

3. Create a second process period by clicking “Add new period” in the top left corner, and define the duration of Period 1 to be 72 h (hhhh:mm). 4. Check that the set points are identical in both periods. 5. In the “Direct control” tab, define the flow rate of feed medium. In Period 1, the flow rate is set to be 0 nL min1, and in Period 2, it is set to be 100 nL min1. In the “Alarm” tab, upper and lower bounds of the process parameters can be defined. If one or more of the measured process parameters fall outside these specifications, the green light on the device will turn red. 3.8

Sampling

Sampling can be done either on a sacrificial basis or on an individual basis. In case of the sacrificial sampling, the entire volume of one well is removed, whereas in the case of the individual sampling, only a portion of the working volume is extracted. In this protocol, only individual sampling is considered. 1. Unclamp the plate, take it from the shaker, and weigh it to determine the plate-wide liquid loss due to evaporation. 2. Transfer the assembled cassette and feeding bottle to the laminar flow cabinet. Carefully remove the top plate from the cassette. The gravimetrically determined liquid loss can now be compensated with deionised H2O, before the required sample volume is extracted from each well and transferred to labeled Eppendorf tubes (see Note 22). 3. Optionally, bicarbonate buffer can be added for pH control. 4. Assemble and weigh the cassette before mounting it on the Orbiter Platform (see Subheading 3.5). 5. Use part of the cell suspension to determine the cell concentration (at appropriate dilution with phosphate-buffered saline (1PBS), take the limited volume into consideration) (see Note 23). 6. Centrifuge the remaining suspension at 16.1 g for 5 min to pellet the cells. 7. The supernatant can be stored at 18  C for later analysis.

3.9 Ending an Experiment and Export of Data

1. Terminate the running experiment by clicking “Stop Experiment.” 2. Now, select between micro-Matrix data file, Excel file, and Text file format. 3. Click “Export” to define the location and name of the exported data file (see Notes 24 and 25).

78

3.10

Vincent Wiegmann et al.

Cleaning

Although most of the materials such as filter bars and cassette are disposable, particularly the liquid feeding module needs to undergo a cleaning procedure after use to avoid blockage of tubing and micro-valves. 1. Remove the Liquid Addition Top Plate from the used cassette, position it over a container, and replace the residual feed medium in the LAFB with hot distilled water. 2. Pressurize the bottle using a syringe or compressed air, connect the priming module, and turn the switch to the “On” position. All valves will now successively open and dispense some of the liquid. Rinse the lines for at least 2 min. 3. Now, perform a backflush by immersing the top plate in hot distilled water and creating a vacuum in the LAFB using a syringe. Again, turn the switch of the priming module to the “On” position and continue the backflush for at least 2 min. 4. Before storing the liquid feeding module, perform an air flush by repeating step 2 with an empty LAFB.

4

Notes 1. The micro-Matrix was used by the authors to culture a GS-CHO cell line (Lonza) expressing a monoclonal antibody. CD-CHO (Life Technologies) was used as basal medium and Efficient Feed B (Life Technologies) as feed medium. MSX (Sigma-Aldrich) was used in routine passages to amplify antibody expression. 2. Routine passages were performed in 250 mL shake flask with vent cap (Corning Life Sciences). 3. Titer analysis was performed with an Agilent 1200 (Agilent Technologies) together with a 1 mL HiTrap Protein G HP column (GE Healthcare). 4. The high dimethyl sulfoxide content in the freezing medium causes the viability to decrease rapidly at room temperature. Aim to minimize the time between thawing and dilution in fresh medium. 5. Setting the working volume of shake flasks to 20% of their nominal volume ensures a reasonable surface to volume ratio in terms of aeration. A higher working volume may lead to limitations in the oxygen transfer, while lower fill volumes may induce out-of-phase motion of the liquid. 6. A liquid leak detector can be used to check gas connections. Gas leaks can lead to an unexpected depletion of the gas supply and should be avoided.

Using a Parallel Micro-cultivation System as Process Development Tool

79

7. Blends of gasses can be used as needed to accommodate alternative control strategies or process conditions. However, for most standard applications, pure gases can be used. In cell culture applications, compressed air or oxygen are typically used for up-control of the DO, while nitrogen is used for down-control of the DO. Down-control of the pH is achieved with carbon dioxide, while bicarbonate buffer or base is added for up-control of the pH. 8. Autoclaving cycles may affect the dosing rate of the microvalves. It is recommended to recalibrate the liquid feeding module after 5 autoclaving cycles. 9. Any liquid that will be delivered by the feeding module should be 0.45 μm filtered before use to increase the lifetime of the micro-valves. 10. The pulse time defines the quantity of liquid that is delivered with each pulse. The shorter the pulse time, the smaller the doses that the system is capable of delivering. 11. Instead of medium, sterile 1PBS can be used to determine the pH offset as it provides more stable pH values. However, the 1xPBS will have to be removed prior to starting the experiment. 12. As the pH readings are affected by temperature, online and off-line readings for the pH recalibration should be done at room temperature. By entering “X” as set point, the temperature control can be disabled. Similarly, the gas bars should not be connected to avoid inadvertent pH changes. 13. Recalibration of the pH cannot be transferred between experiments. Therefore, recalibration and the actual experiment have to be run within the same project. 14. Instead of mixing fresh medium and preculture inside the wells, a cell suspension with the final cell concentration should be prepared in a separate sterile vessel (e.g., shake flask), which is then used to fill the wells of the respective cell culture system. This procedure guarantees consistent seeding density. 15. A repeater pipette (e.g., Rainin or Eppendorf) can be helpful to dispense volumes that otherwise would have to be handled with serological pipettes (e.g., when inoculating the microMatrix cassette with >1 mL). 16. The LAFB is placed inside the micro-Matrix cabinet by default. However, the cabinet is usually kept at elevated temperatures, which can accelerate degradation of the feed medium. By using a longer piece of gas tubing, the LAFB can be routed outside the cabinet and placed inside a Styrofoam box filled with cooling pads. By replacing the cooling pads every 24 h, the temperature inside the box can be kept below 5  C.

80

Vincent Wiegmann et al.

17. The maximum gas flow rate can be adjusted by changing the value for “Volume per pulse” under “Configuration” ! “Device Settings” ! “Gas.” Lower gas flow rates can help to reduce evaporation. 18. The PID settings can be adjusted for each individual well under “Actuator and PID setup” ! “Control configuration.” Often, accurate control can be achieved with the Proportional Gain alone. If after adjustment of the Proportional Gain a continuous offset persists, a small Integral Gain can be added while reducing the Proportional Gain by 10%. In some cases, Derivative Gain can be useful to reduce oscillation. However, for the control needs of our cell culture processes, control configurations without Derivative Gain suffice. 19. To avoid confusion, each gas inlet can be given a name under “Configuration” ! “Device Settings” ! “Gas.” 20. If the feeding module is not used for upward pH control, the buffer can be added manually. By roughly fixing the molar quantities of buffer to the molar quantities of produced lactate in the medium, we achieve steady pH control without occupying the feed line with buffer. 21. Due to the large sample volumes relative to the working volume in the micro-Matrix, feed volumes should be adjusted to avoid overfeeding compared to larger-scale cultures. Instead of fixing the feed volume to the initial working volume, it can be helpful to feed based on the current working volume to match feeding strategies between scales more effectively. 22. Multichannel pipettes with adjustable spacer (e.g., Rainin) can facilitate interactions with the microtiter plate such as inoculation, sampling, and feeding. 23. High dilutions with 1x PBS can affect the cell size. If the average cell size is an important parameter, dilutions should be kept as low as possible. 24. To avoid excessive file sizes, the interval between each data point can be increased when exporting the data as Excel files (downsampling). 25. Exported Excel files contain diagrams for all logged data (pH, DO, temperature, liquid delivery, and gas delivery). The micro-Matrix files can be opened anytime using the HMI software and allow for review of the logged data within the software interface. The Text file contains the raw values of the logged data.

Using a Parallel Micro-cultivation System as Process Development Tool

81

References 1. Doig S, Baganz F, Lye G (2006) High throughput screening and process optimisation. In: Basic Biotechnology, 3rd edn. Cambridge University Press, Cambridge 2. Kelly W, Veigne S, Li X, Subramanian SS, Huang Z, Schaefer E (2018) Optimizing performance of semi-continuous cell culture in an ambr15TM microbioreactor using dynamic flux balance modeling. Biotechnol Prog 34 (2):420–431 3. Bareither R, Pollard D (2011) A review of advanced small-scale parallel bioreactor technology for accelerated process development: current state and future need. Biotechnol Prog 27(1):2–14 4. Hemmerich J, Noack S, Wiechert W, Oldiges M (2018) Microbioreactor systems for accelerated bioprocess development. Biotechnol J 13 (4):1700141 5. Wiegmann V, Martinez CB, Baganz F (2018) A simple method to determine evaporation and compensate for liquid losses in small-scale cell culture systems. Biotechnol Lett 40 (7):1029–1036 6. Hsu WT, Aulakh RPS, Traul DL, Yuk IH (2012) Advanced microscale bioreactor system: a representative scale-down model for bench-top bioreactors. Cytotechnology 64 (6):667–678

7. Nienow AW, Rielly CD, Brosnan K, Bargh N, Lee K, Coopman K, Hewitt CJ (2013) The physical characterisation of a microscale parallel bioreactor platform with an industrial CHO cell line expressing an IgG4. Biochem Eng J 76:25–36 8. Moses S, Manahan M, Ambrogelly A, Ling WLW (2012) Assessment of AMBRTM as a model for high-throughput cell culture process development strategy. Adv Biosci Biotechnol 7:918–927 9. Chen A, Chitta R, Chang D, Amanullah A (2009) Twenty-four well plate miniature bioreactor system as a scale-down model for cell culture process development. Biotechnol Bioeng 102(1):148–160 10. Applikon (2016) Brochure: micro-Matrix 24 bioreactors in a convenient microtiter forma [Online]. http://www.applikon-biotech nology.us/images/download/micro-matrix/ micro-Matrix-leaflet.pdf. Accessed 30 Jan 2017 11. Lattermann C, Bu¨chs J (2015) Microscale and miniscale fermentation and screening. Curr Opin Biotechnol 35:1–6 12. DePalma A (2014) Single-use bioreactors dare to scale. Gen Eng Biotechnol News 14:24–26–27

Chapter 6 HEK293 Cell-Based Bioprocess Development at Bench Scale by Means of Online Monitoring in Shake Flasks (RAMOS and SFR) Tibor Anderlei, Michael V. Keebler, Jordi Joan Cairo´, and Martı´ Lecina Abstract The platforms for bioprocess development have been developed in parallel to the needs of the manufacturing industry of biopharmaceuticals, aiming to ensure the quality and safety of their products. In this sense, Quality by Design (QbD) and Process Analytical Technology (PAT) have become the pillars for quality control and quality assurance. A new combination of Shake Flask Reader (SFR) and Respiration Activity Monitoring System for online determination of OTR and CTR (RAMOS) allows online monitoring of main culture parameters needed for bioprocess development (pH, pO2, OTR, CTR, and QR) as presented below. Eventually, a case study of the application of the combination of SFR-RAMOS system is presented. The case study discloses the optimization of HEK293 cells cultures through the manipulation of their metabolic behavior. Key words Ramos, SFR, OTR, Bioprocess optimization, HEK293 cells, Culture monitoring, Glucose and lactate co-consumption, Metabolic phases

1

Introduction Complex biomolecules, such as functional mAbs or highly glycosylated proteins, require posttranscriptional modifications, which are only performed by mammalian cells [1]. These functionally and pharmacokinetically relevant posttranslational modifications are highly human compatible, which is why mammalian cells are typically the preferred hosts for the production of complex biomolecules [2]. Economic concerns surrounding the low productivity of mammalian cells are ameliorated by the recent developments that allow yields on the production scale of up to 10 g/L [1]. Although 70% of biopharmaceuticals are currently produced with CHO cells [2], Human Embryonic Kidney 293 (HEK293) cell lines have been gaining importance during the last decade [3]. In addition to their unique human protein glycosylation pattern, HEK293 cells are easily genetically modified [4, 5], can be

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_6, © Springer Science+Business Media, LLC, part of Springer Nature 2020

83

84

Tibor Anderlei et al.

grown in suspension up to high cell densities, and are prolific in serum-containing or serum-free platforms [6]. The increasing interest in HEK293-based bioprocesses has expanded their application from viral vector-based gene therapy [7, 8] to protein expression [6, 8], in which human glycosylation patterns are needed [9]. Bioprocesses based on HEK293 cell culture, as the majority of mammalian cell lines, are limited by the inefficient metabolism, which are characterized by a large consumption of glucose and concomitant production of lactate (an inhibitory by-product that limits cell growth) [5]. The significant effort across disciplines to optimize the metabolism of mammalian cells, and particularly HEK293, has shown that only a reduction in lactate production [10] or even lactate consumption (when glutamine or glucose has been completely depleted) has been observed but negatively affecting cell growth [11]. Recently, works have disclosed that under certain culture conditions, HEK293 cells are able to co-metabolize glucose and lactate, even during the growth phase [12]. Extracellular pH and lactate concentration appear to be the key factors that trigger a metabolic shift from glucose consumption/ lactate production to concomitant consumption of lactate and glucose from the culture onset. The initial pH ranges explored were within the range of 6.6–6.8, combined with the addition of exogenous lactate (12–15 mM). Interestingly, cell growth was not affected under these conditions, and cells demonstrated a typical growth pattern. Others have reported a similar physiological behavior in mammalian cells and concluded that this is due to the active co-transportation of lactate and H+ into the cytosolic space through MCT transporters [13], and specifically in HEK293 cells [14]. A deeper analysis by means of metabolic flux balance analysis demonstrated that no differences in ATP production were observed when comparing conventional glucose consumption with concomitant glucose and lactate consumption [15]. In contrast, cell growth was drastically inhibited in lactate-free media at these low pH values. This novel strategy for the metabolic control of HEK293 cells opens the door to further development of mammalian cellbased applications but highlights the need for tools that are capable of monitoring the performance of a culture over time. Such online monitoring tools, besides being a pillar for optimizing the design and productivity of a bioprocess, will also play an important role in quality control. The platforms for bioprocess development have been developed in parallel to the needs of the manufacturing industry of biopharmaceuticals, aiming to ensure the quality and safety of their products. In this sense, Quality by Design (QbD) and Process Analytical Technology (PAT) have become the pillars for quality control and quality assurance. The concept of QbD is focused on

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

85

the idea that the quality of a product cannot solely be determined by the analysis of the final product since many of the critical parameters that define the quality of a final product are determined prior to initiating and along the process [16]. In parallel, PAT is a key element for manufacturing processes analysis and control, and its potential is built from online measurements made during the bioprocess, which feed into the implementation of QbD strategies [17]. In order to guarantee the quality of a given protein (including posttranslational modifications and glycosylation), it is absolutely necessary to monitor parameters beyond those that are typically observed (e.g., pH, pO2, or T ). By monitoring the physiological state and the metabolic activity of cultured cells in real time, investigators will be able to detect any deviation from their previously defined bioprocess and redirect cells toward the optimal physiological state [18]. Those techniques that offer direct and online (real-time) measurements, as well as off-line or indirect (calculated or estimated through other related variables) measurements, are the preferred options for a monitoring system. Over the last few decades, the methods used for monitoring the development of a culture have included: (a) Cell density measurements, which include turbidimetric principles [19], electrical principles, or image analysis [20] using commercially established probes (b) Monitoring the concentration of glucose and glutamine (main substrates), as well as lactate and ammonia (main by-products), directly by HPLC or indirectly by flow injection analysis (FIA) [21, 22] (c) Monitoring the oxygen uptake rate (O.U.R.), which directly correlates with metabolic cell activity [23] and indirectly correlates with cell density [24], is a tool that has attracted much attention from bioprocess engineers. Many different measurement methodologies have been developed from this dynamic technique [25], including mass balance applied in the gas phase [26] and the liquid phase [27, 28]. These monitoring tools, together with the software for process control, have facilitated the implementation of culturing strategies that yielded high cell densities (e.g., fed-batch or perfusion) by leading manufacturers of protein-based therapeutics. Moreover, the data provided by the monitoring tools provide insight on the performance of a culture that can be used to predict critical time points throughout the duration of a bioprocess [29]. A new combination of Shake Flask Reader (SFR) and Respiration Activity Monitoring System for online determination of OTR and CTR (RAMOS) is described in the present piece of work.

86

Tibor Anderlei et al.

The SFR from PreSens (Precision Sensing GmbH) is an online measurement device that determines valuable parameters such as dissolved oxygen and pH in shake flasks in a noninvasive manner from a cultured process. The two main components of the device are the ready-to-use vessels, which contain pre-calibrated sensor spots (in our case, 250 mL shake flasks (Corning, Inc.)), and the SFR main unit, which is responsible for exciting the fluorescence dyes embedded in the sensor spots. The fluorescence signal depends upon the dissolved oxygen concentration and the pH of the sample. The SFR is screwed directly onto the tray of Ku¨hner incubator shaker. Therefore, cells can be cultivated at 37 C in a humidified atmosphere (90–95%) and 5% CO2. RAMOS determines the oxygen transfer rate (OTR), the carbon dioxide transfer rate (CTR), and the respiratory quotient (RQ) of microbial, plant, and cell cultures online. During a cell culture process, the RAMOS will repeatedly interchange between a rinsing phase and measuring phase. The rinsing phase opens the air inlets and outlets of the system to allow for a full exchange of the headspace gas, while during the measuring phase the system is closed and the change in the partial pressure of oxygen and carbon dioxide over time is measured. Both the rinsing phase and the measuring phase operate for a predetermined period of time during each cycle, which may vary depending on the species being grown or the conditions of the culture. The partial pressure data collected from each measuring phase will be used to calculate the OTR, CTR, and RQ of the culture. These data can be used to monitor and track the metabolic activity of a culture over time, which can be then used to optimize the success of a bioprocess [26]. The following work illustrates how a Ku¨hner shaking incubator (LT-XC), equipped with the RAMOS and SFR modifications, can be utilized for bioprocess analysis and optimization. The data presented here are from a case study in which these tools were used to optimize the culturing process of HEK293 cells through manipulation of their metabolic behavior.

2

Materials 1. HEK293-F6 cell line (see Note 1). 2. HyQ SFMTransFx-293 supplemented with 5% FBS and 10% CB5 solution (see Note 2). 3. Lactate solution for media modification (see Note 3). 4. Trypan blue solution. 5. Neubauer hemocytometer. 6. Phase contrast microscope (see Note 4). 7. Automatic enzymatic analyzer (see Note 5).

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

87

8. Refrigerated bench centrifuge for conical 50-mL tubes. 9. Orbital shaker (see Note 6). 10. CO2 incubator. 11. SFR 250-mL shake flasks and 125 and 250 mL plain bottom polycarbonate shake flasks with filtered vent caps. Cells were cultured in 125 mL shake flasks. 12. RAMOS system (HiTec Zang, Germany, installed by Ku¨hner AG, Switzerland). 13. SFR system (PreSens). 14. Incubator shaker. 15. Gas supply.

3

Methods

3.1 Cell Passaging and Maintenance

1. Culture cells in polycarbonate Erlenmeyer shake flasks (125 mL, plain bottom) equipped with filtered ventilation caps. The incubating environment is passively humidified and maintained at 37  C with 5% CO2. Set the orbital shaken platform at 110 rpm. 2. Cells have to be routinely passaged at a frequency of twice a week at a cell seeding density of 0.25  106 cell/mL. 3. Cell passaging or scaling up must be done with exponentially growing cells. Mid-exponential growth phase culture is recommended to achieve a great enough cell density to ensure that cells maintain their physiological integrity (see Note 6). 4. Take a sample from the cell culture and count the number of cells in order to calculate the cell density, as described in the following Subheading 3.3. 5. Calculate the volume of the inoculum culture to reach the desired final concentration (0.25  106 cell/mL) of the passaging culture (see Note 7). V i ½mL ¼

V iþ1  C iþ1 Ci

Where: l

Vi is the volume of initial cell broth needed for subculture.

l

Vi + 1 is the final volume after cell passage (i.e., 13 mL).

l

Ci + 1 is the desired viable cell concentration for the new cell passage after subculture (i.e., 0.25  106 cell/mL).

l

Ci is viable cell density before passaging.

6. Remove the volume of cell culture from the shake flask until only Vi previously calculated is left in the shake flask.

88

Tibor Anderlei et al.

7. Add fresh cell media previously warmed up to 37  C to a final desired volume. 3.2 Scaling Up for the Onset of Experimental Designs

The procedure for scaling up is similar to the culture maintenance described in Subheading 3.1 but requires a larger concentration of cells during the onset of the experiment. As a result, 1L Erlenmeyer shake flasks (polycarbonate, plain bottom, equipped with filtered ventilation caps) from Corning, Inc. were implemented in place of the 125 mL flasks (see Note 8). 1. Proceed as is indicated in instructions 1–4 from the Subheading 3.1. 2. Estimate the number of cells needed to set the experimental design (VexpXini,exp), and add 20% as a safety precaution (designated as the “Safety Factor” below). Calculate the volume of the inoculum culture (Vinoc) using the following formula: V inoc ½mL ¼

V exp  C ini, exp  SF C inoc

where: l

Vinoc is the volume of the culture needed to seed all the experimental conditions (including replicates).

l

Vexp is the total volume of all the cultures conditions.

l

Cini,exp is the viable cell concentration desired for each of the experimental conditions.

l

Cinoc is viable cell density of the culture used for the seeding of each experimental condition (that should belong to the mid-exponential growing phase).

l

SF is the Safety Factor in order to produce more cells than what is strictly necessary.

3. When scaling up cells for an experiment, it is recommended to maintain a reserve seeding culture, in case any issues are encountered and the process needs to be repeated. 3.3 Cell Counting Using the Trypan Blue Exclusion Method

1. An aliquot of 50 μL from your cell culture sample is gently mixed by pipetting with 50 μL of trypan blue dye solution (0.2% v/v). 2. Double-check that the lid of the hemocytometer is correctly positioned, and load the mixture of the sample with trypan blue into the two loading chambers near the edge of the lid. Approximately, 10 μL of stained sample are required to fully load each of the two hemocytometer chambers. However, it is better to load the pipette with a larger volume to ensure that the chamber is entirely filled.

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

89

3. Count the number of living and dead cells (see Note 9) within each section of a chamber. Repeat for both chambers (see Note 10). Cell counting is performed using a phase contrast microscope at 100 magnification. 4. Discard the highest and the lowest counted values for both the number of living and the number of dead cells in each chamber. 5. The remaining values can be used to calculate the viable cell density using the following equation:   n1þn2þn3þn4 Cells ns Cell density ¼  DF mL dV   n1 þ n2 þ n3 þ n4  104 ¼ 2 where: l

n1, n2, n3, and n4 are the number of counted cells within the two accepted squares.

l

ns is the number of squares counted (i.e., four squares (two squares per chamber)).

l

d is the dilution of the sample in trypan blue dye (this value is 0.5).

l

V is the volume loaded in the hemocytometer chamber (i.e., 4  104 mL).

l

DF is the dilution factor in case that any dilution of the sample has been done (see Note 11).

6. Then cell viability can be estimated using the following equation: Viability ½% ¼

3.4 Setup and Cell Inoculation of the Experimental Design

Viable cell density  100 Total cell density

1. Anticipate and prepare all of the different media compositions needed to run the experiment (see Note 12). 2. Estimate the total volume of each media needed for the full experiment, and prepare an additional 10% to avoid running out. 3. Perform a sterility test by incubating an aliquot of each media in a 6- or 12-well plate for 24 h at 37  C. 4. Check under the microscope that the media is clean and free of any potential contamination. 5. Fill the 250 mL SFR shake flasks for each experimental condition (including the culture replicates) with the final culturing volume and incubate them in the CO2 incubator at 37  C. Conditioning the media for 20–30 min in the experimental

90

Tibor Anderlei et al.

environment prior to adding cells is important to help reduce cell stress from the shock of a new environment. 6. Estimate the inoculation volume required for each experimental condition (using the equation presented in Subheading 3.2), and dispense into a falcon tube. A single falcon tube can be used to pool the inoculum required for different replicates of the same experimental condition, or each replicate can be aliquoted into individual tubes (see Note 13). 7. Centrifuge the inoculum cultures at 300  g for 5 min at 4  C (Heraeus, benchtop centrifuge for conical 50 mL tubes). 8. Gently discard the supernatant, being careful not to disturb the cellular pellet. 9. Resuspend the pellet in the desired volume of media that is needed for the experiment. 10. Once the SFR flask has been equilibrated in the incubator, inoculate the cell suspension. 11. The experiment is ready to be run. 3.5 Setup of the RAMOS System

1. Make sure that the gas mixture humidification is set at 85% in order to avoid water condensation into the equipment (see Note 14). 2. Define the experimental conditions accordingly to the equipment capacity: up to 8 flasks only RAMOS system is required or up to 6 when RAMOS and SFR systems are desired (see Note 15 and Fig. 1). 3. Set the temperature of the LT-XC incubator shaker at 37.0  C.

3.6 Preparation of the RAMOS System

1. Start the RAMOS system, and follow the click sequence on left side (New experiment, Parameterize, Oxygen calibration, Leakage test) displayed in the home screen of the monitoring software (Fig. 2). 2. By clicking on the step “Parameterize,” some important points to be paid attention to are displayed (Fig. 3) in order to set up the RAMOS system for cell cultivation. 3. Introduce the flask volume: this is the total gas volume of the flask. 4. Max. measuring time should be set to 20–30 min. 5. Flow rinsing phase should be set to 60 mL/min. 6. Oxygen content in the inlet gas should be set at 20.95–5% CO2 concentration. 7. Before carrying out the oxygen and the leakage test, all flasks need to be correctly attached to the system, and the temperature of the incubator shaker has to be stable. The clamps have special positioning holes ensuring the right installation of the

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

91

Fig. 1 Picture of the LT-XC incubator shaker with RAMOS and SFR integrated

Fig. 2 RAMOS home screen. On the left side, the sequence of the different steps for starting a new experiment. Center and right areas are used to define the parameters of the experiment and to register the details of the experiment

modified Corning 250 mL flasks with positioning pins and pH and DO sensors (200001341, PresSens, Germany). 8. It is recommended to set up the experiment a full day in advance to allow the system to equilibrate overnight for a calibration the next morning (see also setup SFR and Note 16).

92

Tibor Anderlei et al.

Fig. 3 Screenshot of the RAMOS Parameterize menu where all the relevant experimental culturing parameters have to be defined before running the experiment

9. Connect the 250 mL Corning flask with integrated pH and DO sensors (optical) to the RAMOS system. A special adapter (white plastic part, shown in Fig. 4) is available to ease the assembly. 10. The flask is correctly assembled with the RAMOS system by connecting the oxygen sensor cable (black) and the air inlet and outlet lines to the respective ports. 11. After calibration and leakage test, the system is ready to inoculate. 12. By clicking the “start experiment” button, initiate the online measurement of OTR/CTR/RQ. 3.7 Setup of the SFR System

1. Before the SFR can be used, a Bluetooth connection between the PC and the SFR has to be established. A provided Bluetooth-USB stick has to be attached to the PC, and the push button at the front of the SFR has to be activated.

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

93

Fig. 4 Corning 250 mL flask equipped with RAMOS adapter harboring the pH and DO optical sensors

2. Start the SFR software and connect the SFR with the software (Fig. 5). 3. Ordering the modified Corning flasks from PreSens, calibration data for the DO and pH spot will be sent along and have to be entered in the software (see Fig. 6). 4. Fill media in the flasks and attach the flasks to the SFR. Positioning pins at the flasks will ensure that the optical sensor spots in the flask will be located optimal above the reading unit. 5. Let the system equilibrate overnight. (Analogue RAMOS shaking speed 100 RPM, 50 mm.) 6. Before starting inoculation, carry out a single-point calibration (see Fig. 7) by measuring off-line the pH. DO should be calibrated to 100%. 7. Start the experiment parallel in the RAMOS and SFR software. 3.8

Culture Sampling

1. Pause the monitoring software. 2. Stop the orbital shaker.

94

Tibor Anderlei et al.

Fig. 5 Screenshot of the PreSens home screen

Fig. 6 Screenshot of the Calibration Data menu where besides general information of the equipment, the virtual tray with the information of the calibration parameters are displayed for each culture (position)

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

95

Fig. 7 Screenshot of the Calibration Data menu once the single-point calibration is performed. Observe that for each position (culture), pH and Oxygen concentration are provided in real time

3. Carefully unplug the RAMOS O2 sensor system plugged on the adapting caps. 4. Carefully disconnect the gas tubes. 5. Remove the flasks with the white adapter from the clamps. 6. Unscrew the adaptors into the laminar hood, and proceed to sample an aliquot from the culture. Homogenize the culture just before sampling by softly shaking the flask. 7. The sample volume depends on the further sample analysis to be performed. Normally, 0.5 mL are enough for cell density and viability estimation and lactate and glucose enzymatic measurement. 8. Screw back the white adaptors to the flasks, and make sure that they are properly assembled. 9. Put the flasks into the shaker, and make sure that they are correctly positioned. 10. Carefully restore the black O2 sensors and the gas tubes. 11. Resume shaking, and change the actual liquid volume (liquid volume—sample volume) in the RAMOS software (parameterize).

96

Tibor Anderlei et al.

3.9 Off-Line Monitoring: Cell Density, Viability, and Lactate/Glucose Concentration

1. To estimate the cell density and viability, follow the instructions provided in Subheading 3.3 (cell counting using the trypan blue exclusion method). 2. After the cell density has been determined, centrifuge the remainder of the sample at 300  g for 10 min. A 1-min pulse at 1700  g can be used in the case of Eppendorf 1.5-mL tubes. 3. Filter the sample through a syringe filter (0.20 μm), and transfer the flow-through into a new tube. 4. Ensure that the bottles of the YSI analyzer are filled with the corresponding calibrating and washing buffers. Empty the waste buffer. 5. Switch the YSI 2100 to run mode. 6. Wait until the equipment successfully calibrated. A “Ready to sample” message will appear on the display screen. 7. Press “Sample” and wait for the needle to move to the aspirating position. 8. Introduce the needle into the sample and repress the Sample button. 9. If the sample is out of range, dilute the sample and repeat the analysis. 10. Perform each analysis in triplicates. 11. Wash the system once with Milli-Q water in between samples. 12. Put the equipment in Standby mode.

3.10

Study Case

With the aim of illustrating the potential applications of the SFR shake flasks equipped with the RAMOS system, a brief case study, which is subsection of a larger ongoing project, is presented below. As mentioned earlier, substantial differences in the metabolism of HEK293 cells have been observed with different culture conditions. At a physiological pH of ~7, media with little or no initial lactate concentration will observe an increase in the concentration of lactate until glucose is completely depleted. Thereafter, cells will begin to consume lactate and proliferate at a decreased rate, similar to that observed during the end of an exponential growth phase under normal metabolic conditions. During this period of increased lactate, cells will metabolically convert to consuming both glucose and lactate if the pH of the media is reduced to 6.8 or lower. This simultaneous consumption of glucose and lactate has no significant effect on cell growth [14]. In order to confirm these previous observations of how a shift in pH will trigger this type of metabolic shift in HEK293 cells, cultures at low pH (6.8) with different initial lactate concentrations were performed using identical media (HyQ SFMTransFx-293, supplemented with 5% FBS

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

97

Fig. 8 HEK293 cells cultivation data. Plots a1 and b1 show time profile of SFR (pH, pO2) and off-line measurements (cell density, glucose and lactate concentration, and fluorescence intensity). In plots a2 and b2, online measurements provided by RAMOS are represented (OTR, CTR, and RQ). In Condition a, no initial lactate was added into the media, and in Condition b, the initial lactate concentration was about 12 mM

and 10% CB5). In Condition A, no lactate was added to the media, and the initial pH was set at 6.6. In Condition B, the pH of the media was also set at 6.6, but the media was supplemented with 12 mM of lactate. In addition to the online variables measured by the culturing system (pH, pO2, OTR, CTR, and RQ), off-line measurements of cell density, glucose and lactate concentration, and fluorescence intensity were also collected. Results are shown in Fig. 8. The SFR-RAMOS (Ku¨hner AG) hybrid instrument provided the perfect cell culture platform for monitoring the correlation between pH and lactate metabolism in HEK293 cells. Moreover, the SFR-RAMOS enables live monitoring of pO2 and OTR in shake flask, which was needed to rule out the hypothesis that oxygen limitation was affecting cellular respiration and the effects of lactate consumption. The results show the negative effect of a low pH on cellular growth when lactate in not present in the media (Fig. 8a1). A lag phase about 24–30 h is observed, in which lactate was still being

98

Tibor Anderlei et al.

generated. Afterward, when lactate reached about 8 mM, cells were able to grow but at a much lower growth rate (Fig. 8a1 and b1). Such a lag phase is not observed when lactate was present in the media from the start (Fig. 8b1), confirming that a lactate concentration about 12 mM and pH of 6.6 can trigger the co-consumption of lactate and glucose with no significant negative effects on cell growth. This yielded a significant difference in the cell densities achieved by the end of the cultures (7.5  106 vs. 1.4  107 cells/mL). As observed in Fig. 8a1 and b1, the pO2 and pH profiles correlate with cell density and lactate concentration. On the one hand, this allows us to rule out the hypothesis of any potential oxygen limitation that could shift the metabolism toward lactate consumption, since oxygen is not limiting in any stage of the culture (neither glucose). On the other hand, the pH profile indicates that the pH was decreased when lactate was generated or consumed. As shown in Fig. 8a1, the pH dropped to 6.5 in Condition A, which was likely due to the amount of lactate generated, while Fig. 8b1 demonstrates that Condition B was able to main a constant pH of ~6.6. The initial decrease in pH that was observed in both cases is likely due to conditioning of the media with CO2, prior to equilibrium. The OTR and CTR also correlated well with cell density, drawing similar profiles and reporting on metabolic activity of the cells. Even more interesting was the evolution of the respiratory quotient (RQ), which indicates the ratio between the CO2 generated and O2 consumed. During the initial stages of lactate production, slightly higher values of RQ values were observed (about 1.1 mol CO2/mol O2). When lactate was significantly consumed together with glucose, RQ values were decreased to slightly below 1 (Fig. 8b2). The increase of oxygen consumption could be attributed to the fact that lactate is more reduced than pyruvate. Since lactate has to be oxidized to CO2 and H2O, more oxygen is needed to fully oxidize lactate in respect to pyruvate. In conclusion, the SFR-RAMOS system offers the opportunity to better understand the physiological and metabolic behavior of cell cultures in a regular shake flask. This bypasses the need for a bioreactor and provides the investigator with a new tool for optimizing the performance of their bioprocess.

4

Notes 1. The cell lines used in this work were the HEK293SF-3F6 cell line, which was adapted for growth in suspension culture and was kindly provided by Dr. A. Kamen (National Research Council of Canada). The HEK293SF-3F6 cell line was further

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

99

transfected with eGFP using the expression vector pIRESpuro2 (Clontech). 2. Cell maintenance and experimental setup was performed with SFMTransFx-293 medium (HyClone, Thermo Scientific) supplemented with 4mM GlutaMAX (Gibco, Invitrogen), 5% FBS (v/v) (Sigma-Aldrich), and 10% of Cell Boost 5 (80 g/L) (HyClone, Thermo Scientific). CB5 is a nutritional supplement designed to increase cell density and productivity. CB5 is a nonchemically defined source of amino acids, vitamins, lipids, cholesterol, growth factors, and glucose. CB5 (80 g) was dissolved in 1 L of MilliQ water under continuous stirring, and pH was adjusted to 7.2 by carefully adding NaOH (5M) (drop by drop). The solutions were sterile filtered using 0.22 μm filter units with filling bell (Millipore, Cat SVGVB1010). Solutions can be stored at 4  C for a maximum of 6 months. Eventually, 2 g L-1 Kolliphor® P 188 (Sigma-Aldrich) was added to the media (20 mL/L of 100 g Kolliphor/L). The solution was heat sterilized via autoclaving. 3. Adjusting the lactate concentration was performed as follows: A stock solution of sodium lactate, NaC3H5O3 1M (PanReac) was prepared in Milli-Q water and sterile filtered. The desired initial lactate concentration ([lactate]0) was added to media immediately prior to the experiment onset. 4. The cell number was determined by manual counting at 100 magnification using a phase contrast microscope (Nikon eclipse TS100) and a hemocytometer (Improved Neubauer Chamber, Brand). The hemocytometer is a thick microscope slide with two delimitated loading chambers. Each of these chambers is engraved with a laser-etched grid of perpendicular lines. Viability was assessed using the trypan blue dye exclusion method. Trypan blue dye (Sigma-Aldrich, T-8154) is diluted to 0.2% (v/v) in PBS and filtered. 5. Glucose and lactate were measured in an YSI automatic glucose and lactate analyzer (Yellow Springs Instrument, 2700 Select). YSI membranes contain three layers. Enzymes are immobilized on the second layer and are specific for the substrate of interest (glucose oxidase or L-lactate oxidase). Once the substrate is oxidized, hydrogen peroxide is produced and passes through the third layer, which is made of cellulose acetate and is only permeable to small molecules (like hydrogen peroxide). The permeated hydrogen peroxide reaches the platinum electrode and is oxidized to produce a current, which is proportional to the concentration of the substrate. The analyzer uses 25 μL of the sample and provides the values of glucose and lactate concentration (g/L), with an error of 0.1 g/L. The

100

Tibor Anderlei et al.

instrument range is 0.05–20.0 g/L for glucose and 0.05–2.00 g/L for lactate. When analyzing samples with a higher concentration of glucose or lactate, it is necessary to dilute them with Milli-Q water until the measure fits into the dynamic range of instrument. 6. For cell maintenance and the preparation of inoculum, cells were cultured in regular polycarbonate shake flasks. Shake flasks were placed on an orbital shaker (Stuart, SLL1) especially design to be placed into a humidified, CO2 environment. Be aware that only specially designed shaker platforms can be placed into a CO2 incubator, which are resistant to CO2 oxidation and produce unwanted heat. In the case of HEK293 cells, cell densities between 2 and 3  106 cells/mL were typically observed before a new passage was carried out. This was done to ensure that cells were midway through the exponential growth curve at the beginning of each passage or experiment. In either case, cells were seeded at a density of ~0.25  106 cell/mL. 7. The recommended range of working volume for plain bottom polycarbonate 125 mL shake flasks (Corning, Inc.) is between 13 and 35 mL when performing regular cell passaging (2–3 days of culturing). Lower volumes are more susceptible to osmolar inhibition resulting from evaporation, and higher volumes could limit the oxygen transfer from the gas phase to the liquid phase. 8. To control for heterogeneities between different cultures, it is strongly recommended that a single seeding culture be used for all experimental conditions. Only when different media is used should cells be previously adapted to grow under different conditions before seeding the experiment. Any scale-up should also be performed separately in each media. 9. Trypan blue dye only penetrates nonviable cells with a damaged membrane, which stains the cells and results in a bluish appearance under the microscope. Any viable cells in the culture will remain unstained and will emit a shiny white/yellowish color. Dead cells and living cells should be counted separately so that the cell density and % of viability can be accurately estimated. 10. The gridded area of the hemocytometer consists of nine 1  1 mm (1 mm2) squares. Only four of these squares are required for the cell counting. For ease of counting, each square is divided into four smaller squares of 0.25  0.25 mm (0.0625 mm2). The total combined area is 16 mm2. 11. When measuring cell density, counting 20–80 cells per square is recommended to yield a higher level of accuracy. If needed, dilute the sample with PBS prior to staining with trypan blue to

Animal Cell Cultures Monitoring Platform: RAMOS and SFR

101

achieve a cell density that fits this range. Manual cell counting with the trypan blue method should be restricted to an error of 10%. 12. When media at different pH levels is required to run an experiment, the pH will need to be modified in media that has already been equilibrated in a CO2 incubator. In other words, it is recommendable to equilibrate an aliquot of media by introducing it into the CO2 incubator, and when pH becomes stable, add the alkali or acid buffers required to set the experimental initial desired pH value. Once the desired pH of the media has been achieved, the media must be filter-sterilized in the laminar hood. 13. Using the same aliquot to seed multiple conditions and replicates is recommended to reduce any variation between replicates and simplify the experimental setup. This can be achieved by pooling cells into a single falcon tube prior to administering the aliquots. It is important to ensure that there is enough liquid in the “master pot” to account for any pipetting error. When conditioning media in the SFR flasks prior to adding the inoculum, it is important to consider what the final volume will be following inoculation. For example, an experiment with 3 replicates of 25 mL total volume (each with 2 mL of inoculum) will require a setup of 3 flasks (each with 23 mL) that will need to be conditioned prior to inoculation. 14. The RAMOS system [26] was initially developed for microbial cultivations. In order to use it for mammalian cell cultures, a humidified gas mixture (5% CO2, FlowCon, Ku¨hner AG, Switzerland, was used) is needed at the gas inlet port of the RAMOS system. The humidification must be below 85% RH in order no condensation can take place in the RAMOS controller (especially mass flow controller). 15. The RAMOS block is normally screwed onto the shaking tray (see Fig. 1), and up to 8 measuring flasks can be connected to the RAMOS block. If there is a need to measure the pH and DO in conjunction with the OTR, the RAMOS system is compatible with the SFR-Reader (PreSens) and is screwed directly onto the SFR-Reader via an adapter plate. In this case, 6 flasks can be used to simultaneously measure OTR/CTR/RQ and pH/DO. The SFR system is screwed onto a Ku¨hner universal tray (EU-tray), and 6 clamps (SM310250P, Ku¨hner AG, Switzerland) are placed on the open positions of the SFR system. 16. The software will not respond if the conditions are not stable and will stop the calibration test.

102

Tibor Anderlei et al.

References 1. Butler M (2005) Animal cell cultures: recent achievements and perspectives in the production of biopharmaceuticals. Appl Microbiol Biotechnol 68:283–291 2. Wurm FM (2004) Production of recombinant protein therapeutics in cultivated mammalian cells. Nat Biotechnol 22:1393–1398 3. Moreira A (2007) The evolution of protein expression and cell culture. BioPharm Int 20:10 4. Butler M, Spearman M (2014) The choice of mammalian cell host and possibilities for glycosylation engineering. Curr Opin Biotechnol 30:107–112 5. Dumont J, Euwart D, Mei B et al (2016) Human cell lines for biopharmaceutical manufacturing: history, status, and future perspectives. Crit Rev Biotechnol 36:1110–1122 6. Roma´n R, Miret J, Scalia F, Casablancas A, Lecina M, Cairo´ JJ (2016) Enhancing heterologous protein expression and secretion in HEK293 cells by means of combination of CMV promoter and IFNα2 signal peptide. J Biotechnol 239:57–60 7. Kim JY, Kim YG, Lee GM (2012) CHO cells in biotechnology for production of recombinant proteins: current state and further potential. Appl Microbiol Biotechnol 93:917–930 8. Liste-Calleja L, Lecina M, Cairo´ JJ (2014) HEK293 cell culture media study towards bioprocess optimization: animal derived component free and animal derived component containing platforms. J Biosci Bioeng 117:471–477 9. Delenda C, Chillon M, Douar AM, Merten OW (2007) Cells for gene therapy and vector production. In: Po¨rtner R (ed) Animal cell biotechnology, Methods in biotechnology, vol 24. Humana Press, Totowa, NJ 10. Durocher Y, Butler M (2009) Expression systems for therapeutic glycoprotein production. Curr Opin Biotechnol 20:700–707 11. Hassell T, Gleave S, Butler M (1991) Growth inhibition in animal cell culture. Appl Biochem Biotechnol 30:29–41 12. Gagnon M, Hiller G, Luan Y-T et al (2011) High-End pH controlled delivery of glucose effectively suppresses lactate accumulation in CHO Fed-batch cultures. Biotechnol Bioeng 108:1328–1337 13. Martı´nez VS, Dietmair S, Quek L-E et al (2013) Flux balance analysis of CHO cells before and after a metabolic switch from lactate production to consumption. Biotechnol Bioeng 110:660–666

14. Liste-Calleja L, Lecina M, Lopez-Repullo J et al (2015) Lactate and glucose concomitant consumption as a self-regulated pH detoxification mechanism in HEK293 cell cultures. Appl Microbiol Biotechnol 99:9951–9960 15. Halestrap AP, Price NT (1999) The protonlinked monocarboxylate transporter (MCT) family: structure, function and regulation. Biochem J 343:281–299 16. San Martı´n A, Ceballo S, Ruminot I et al (2013) A genetically encoded FRET lactate sensor and its use to detect the Warburg effect in single cancer cells. PLoS One 8(2):e57712 17. Martı´nez-Monge I, Albiol J, Lecina M et al (2019) Metabolic flux balance analysis during lactate and glucose concomitant consumption in HEK293 cell cultures. Biotechnol Bioeng 116:388–404. https://doi.org/10.1002/bit. 26858 18. Yu LX, Amidon G, Khan MA et al (2014) Understanding pharmaceutical quality by design. AAPS J 16:771–783 19. Rathore AS, Bhambure R, Ghare V (2010) Process analytical technology PAT for biopharmaceutical products. Anal Bioanal Chem 398:137–154 20. Zhao L, Fu H-Y, Zhou W et al (2015) Advances in process monitoring tools for cell culture bioprocesses. Eng Life Sci 15:459–468 21. Junker BH, Reddy J, Gbewonyo K et al (1994) On-line and in-situ monitoring technology for cell density measurement in microbial and animal cell cultures. Bioprocess Eng 10:195–207 22. Ho¨pfner T, Bluma A, Rudolph G et al (2010) A review of non-invasive optical-based image analysis systems for continuous bioprocess monitoring. Bioprocess Biosyst Eng 33:247–256 23. Ruffieux PA, von Stockar U, Marison IW (1998) Measurement of volumetric (OUR) and determination of specific (qO2) oxygen uptake rates in animal cell cultures. J Biotechnol 63:85–95 24. Casablancas A, Ga´mez X et al (2013) Comparison of control strategies for fed-batch culture of hybridoma cells based on on-line monitoring of oxygen uptake rate, optical cell density and glucose concentration. J Chem Technol Biotechnol 88:1680–1689 25. Sauer PW, Burky JE, Wesson MC et al (2000) A high-yielding, generic fed-batch cell culture process for production of recombinant antibodies. Biotechnol Bioeng 67:585–597 26. Ga´lvez J, Lecina M, Sola` C et al (2012) Optimization of HEK-293S cell cultures for the

Animal Cell Cultures Monitoring Platform: RAMOS and SFR production of adenoviral vectors in bioreactors using on-line OUR measurements. J Biotechnol 157:214–222 27. Anderlei T, Zang W, Papaspyrou M, Buechs J (2004) Online respiration activity measurement OTR, CTR, RQ in shake flasks. Biochem Eng J 17:187194 28. Fontova A, Lecina M, Lo´pez-Repullo J et al (2018) A simplified implementation of the

103

stationary liquid mass balance method for on-line OUR monitoring in animal cell cultures. J Chem Technol Biotechnol 93:1757–1766 29. Lecina M, Soley A, Gra`cia J et al (2006) Application of online OUR measurements to detect actions points to improve baculovirus-insect cell cultures in bioreactors. J Biotechnol 125:385–394

Chapter 7 Orbitally Shaken Single-Use Bioreactor for Animal Cell Cultivation: Fed-Batch and Perfusion Mode Tim Bu¨rgin, Juliana Coronel, Gerrit Hagens, Michael V. Keebler, Yvonne Genzel, Udo Reichl, and Tibor Anderlei Abstract Increasing the cultivation volume from small to large scale can be a rather complex and challenging process when the method of aeration and mixing is different between scales. Orbitally shaken bioreactors (OSBs) utilize the same hydrodynamic principles that define the success of smaller-scale cultures, which are developed on an orbitally shaken platform, and can simplify scale-up. Here we describe the basic working principles of scale-up in terms of the volumetric oxygen transfer coefficient (kLa) and mixing time and how to define these parameters experimentally. The scale-up process from an Erlenmeyer flask shaken on an orbital platform to an orbitally shaken single-use bioreactor (SB10-X, 12 L) is described in terms of both fed-batch and perfusion-based processes. The fed-batch process utilizes a recombinant variant of the mammalian cell line, Chinese hamster ovary (CHO), to express a biosimilar of a therapeutic monoclonal antibody. The perfusion-based process utilizes either an alternating tangential flow filtration (ATF) or a tangential flow filtration (TFF) system for cell retention to cultivate an avian cell line, AGE1.CR.pIX, for the propagation of influenza A virus, H1N1, in high cell density. Based on two example cell cultivations, processes outline the advantages that come with using an orbitally shaken bioreactor for scaling-up a process. The described methods are also applicable to other suspension cell lines. Key words Orbitally shaken bioreactor, Single-use bag, Scale-up parameters, Process development, Recombinant CHO cell line, Monoclonal antibody, Biosimilar, Glycosylation, Perfusion, Viral vaccine production

1

Introduction

1.1 Basic Working Principles of an Orbitally Shaken Bioreactor

An orbitally shaken bioreactor (OSB) is a culture vessel, which is placed on an orbitally shaken platform (e.g., a shaker table in an incubator shaker). The shaking diameter is different and often adjustable. The vessel is passively aerated (e.g., an Erlenmeyer flask with vented cap) or actively aerated (a single-use bag connected to a gas mixing device). The shaking motion induces a liquid wave that travels along the wall of the vessel and facilitates passive aeration and mixing of the medium. The energy needed to maintain this wave is referred to as the power input [1]. The total

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_7, © Springer Science+Business Media, LLC, part of Springer Nature 2020

105

106

Tim Bu¨rgin et al.

amount of this power, as well as its distribution in the liquid, is correlated with shear forces that can damage the cells. In the case of stirred tank bioreactors, the energy is transferred into the liquid through the use of an impeller, which can cause locally high shear stress on cells near the impeller blades. Conversely, OSBs distribute the power evenly by applying force to the entire vessel, limiting the majority of the shear stress to the surface of the vessel [2]. An OSB is aerated through gassing the headspace of the vessel (volume of gas located above the liquid) and relying on the principles of hydrodynamics to facilitate oxygen transfer into the liquid. The circulating wave leaves behind a thin film of liquid, which increases the surface area of the liquid and enhances the mass transfer between the gas and the liquid phase [3]. This type of surface aeration limits the formation of bubbles that can cause significant shear when bursting and reduces foam formation [4]. 1.2 General Design of Orbitally Shaken Single-Use Bioreactors

Orbitally shaken platforms can accommodate a wide range of vessel types and sizes, yielding a large degree of flexibility and independence from specific manufacturers. In the small scale (3 L), OSBs are cylindrically shaped vessels holding single-use bags (e.g., the SB10-X) [4, 5]. For the cultivation of shear-sensitive cells, round vessels are commonly used (baffles or square-shaped vessels cause high shear forces). At the small scale, the vessels are aerated passively through a vented cap or membrane. At the larger scale, single-use bags with sterile air filters are used to adjust the concentration of gases in the headspace of the OSB. This allows the user to define the gas composition of both the headspace and the liquid without sparging gas directly into the liquid and to adjust dissolved oxygen (DO) concentration. In a similar manner, the pH value of the cell culture medium can be controlled via the gas mixture. That is, when increasing the CO2 content of the gas mixture, the pH value of the medium will decrease (assuming a bicarbonate-based buffer). Similarly, a decrease in the CO2 content of the gas mixture will lead to an increase in the pH value. The single-use bags utilized with these OSBs come with a variety of tubing options that facilitate liquid transfer into and out of the bag. In addition to acid and base control, these additional tubing options can be utilized for supplementing the culture with nutrients like glucose or other chemicals. The bags are also equipped with optical sensors, which allow the user to noninvasively measure pH value and DO concentration of the culture medium. In contrast to small OSBs that are placed inside an incubator shaker, which do not allow for direct control of temperature and gas mixture inside the vessel, these large OSBs allow the direct control of the temperature and gas composition inside the bag.

Cultivation in Orbitally Shaken Bioreactor

107

Scale-Up

The phrase “scaling-up” refers to an increase in the cultivation volume from small to larger scales. For instance, the transfer of a culture from a 3 L Erlenmeyer flask to a 12 L OSB like the SB10-X. Maintaining a consistent level of productivity and product quality at the larger scale is crucial for the success of any scale-up. This makes the planning of a scalable process a rather complex task. Increasing the cultivation volume is often accompanied with a change in the bioreactor geometry and technology [6]. With OSBs, this task is comparably easy because the mixing and aeration principles, both influencing the hydrodynamics, are maintained throughout the whole scale-up when geometrically similar vessels are used; the shaking frequency and/or shaking orbit can be adapted to maintain the selected scale-up parameters throughout the different bioreactors. In the pharma industry, the most commonly used scale-up parameters are (1) volumetric power input (P/VL), which correlates with shear forces in the liquid; (2) the volumetric oxygen transfer coefficient (kLa), which represents the efficiency of gas transfer from the gas phase to the liquid phase in the bioreactor; and (3) the DO concentration in the cultivation medium [7]. Other parameters, like pH value, evaporation rate, and the concentration of leachables, can also influence the success of a scale-up. When scaling-up, the goal is to maintain all of the scaleup parameters at the larger scale, but in practice this is not possible; therefore, it is necessary to select and maintain those parameters, which have the greatest impact on product yields and process productivity [8]. For instance, aerobic processes rely heavily on the oxygen transfer rate (OTR) and kLa, and so optimization of both is a priority in the scale-up [9]. The mixing time is another crucial parameter that can weigh heavily on the success of a scale-up and should not be underestimated. In large-scale bioreactors, the mixing time can exceed significantly the values determined for smaller-scale bioreactors. These longer mixing times can lead to heterogeneities in the liquid with a diversity of local microenvironments, which will result in some cells experiencing greater nutrient and oxygen deprivation than others [10]. Typically, the manufacturer of a bioreactor will support customers with scale-up parameter data. Those data will help the user to select the optimal settings (shaking diameter, shaking orbit, aeration rate, etc.) for a scale-up or scale-down procedure.

1.4 Cultivation of CHO Cells for Recombinant Protein Production

All biosimilar developments begin with a thorough analysis of the initial drug in order to characterize the batch-to-batch variation in the physicochemical properties of the biologics. The developer may then consider which cell line to utilize and how to design the manufacturing process. Since glycosylation is species-specific, developers of a biosimilar will often prefer to use the same host cell line (e.g., CHO-K1, CHO-DG44, NS0, SP2/0, etc.) as the one used for the initial biologics [11]. Since the manufacturing

1.3

108

Tim Bu¨rgin et al.

process (i.e., the upstream process (USP) and the downstream process (DSP)) will have a significant impact on the physicochemical characteristics (glycosylation, charge species, deamination, etc.) of a biologic, biosimilar developers will often prefer to replicate the process modes and specific unit operations (e.g., fed-batch, perfusion, clarification, affinity or ion exchange chromatography, etc.) as much as possible [12]. However, since it is impossible to use the same recombinant cell line with exactly the same manufacturing process as the initial biologic, biosimilar developers will often evaluate several clones in parallel. Only those clones with the highest level of product quality (biosimilarity) and cell line productivity are selected for further development. Similarly, development of the manufacturing process for a biosimilar is a highly re-iterative endeavor and requires a considerable amount of expertise to end up with an economically viable process that delivers a high-quality product. Since product quantity is often inversely related to the product quality, strategic planning is crucial to ensure that each of the parameters is fully optimized. Some of the key parameters to consider include the cell seeding density, media composition, feeding strategy, media supplements, the choice of the bioreactor platform and settings, and the DSP platforms and settings. This is just a selection of the key aspects to consider when developing the manufacturing process of a biosimilar (for a review, see ref. 11). The glycosylation of a therapeutic protein will affect its reactivity with membrane receptors, as well as its interactions with other proteins, and its stability in circulation. Since abnormal glycan structures can have a strong negative impact on the safety of a therapeutic protein [12–14], glycan analysis is an essential step in developing manufacturing processes. In this report, we describe the seamless transfer of a biosimilar from a shake flask (200 mL working volume) to an orbitally shaken single-use bioreactor (SB10-X, 8 L working volume), without compromising product quality or product quantity of the biosimilar. 1.5 Cell CultureBased Viral Vaccine Production

A cell culture process for whole virus vaccine production can be characterized as a biphasic process regarding the time of infection (TOI). The cell growth phase is defined as the period from inoculation to the TOI, in which cells are actively growing, followed by the virus production phase. Depending on the characteristics of infection conditions, virus replication, and virus spreading (e.g., multiplicity of infection), a period of subsequent cell growth after infection can be observed that is often followed by virus-induced cell death [15]. Besides virus vaccine manufacturing in adherent cells in batch mode [16], successful suspension cell-based processes have been reported in batch, continuous, or perfusion modes [17]. Here we present an example for successful process intensification using orbitally shaken vessels with headspace operation.

Cultivation in Orbitally Shaken Bioreactor

109

Therefore, an OSB (SB10-X, 5–10 L working volume) was used that was either coupled to an alternating tangential flow filtration (ATF) or to a tangential flow filtration (TFF) perfusion system. In both cases, a maximum cell density of up to 5  107 cells/mL was reached before successful infection with influenza A virus.

2

Materials

2.1 Estimation of Scale-Up Parameters 2.1.1 Mixing Times

1. 1 % starch solution (stock solution): Add 10 g starch cold soluble to 100 mL dH2O. Mix it in a glass beaker for a few minutes at room temperature (see Note 1). Transfer the mixture into a 2 L Erlenmeyer flask with baffles. Increase the volume to 1000 mL with dH2O. Shake the mixture at a high shaking frequency (e.g., at 250 rpm with a shaking orbit of 50 mm) until the starch is completely emulsified (see Note 2). Then store the stock solution at 6  C or less. 2. 0.8 vol. % colored starch solution: Dilute the stock solution with dH2O and colorize it with Lugol’s solution (see Note 3). For a working volume of 1 L (as an example), apply the following mixture: 8 mL stock solution, 4 mL 0.33 % Lugol’s solution, and 988 mL dH2O. 3. Discoloration agent: 1 M sodium thiosulfate (approx. 0.33 mL per 1 L for the discoloration reaction). 4. High-speed camera Exilim (Casio). 5. Light source (e.g., a LED lamp). 6. Table-top shaker LS-X (Adolf Ku¨hner AG) with custom-made tray that accommodates a high-speed camera (mounted on holder, placed in front of the vessel, which has to be filmed) and a transparent cylindrical vessel holding the 12 L standard single-use bag (Hegewald Medizinprodukte GmbH) used with the SB10-X (see Subheading 2.1.2). 7. 25 mL syringe (optimal syringe volume depends on required sodium thiosulfate and tubing volume connected to the bag) and 1/800 tubing for connection to single-use bag.

2.1.2 Volumetric Oxygen Transfer Coefficient of Oxygen (kLa)

1. Pressurized nitrogen and air. 2. dH2O. 3. Bioreactor SB10-X with control unit and integrated data acquisition software Insight (Adolf Ku¨hner AG). 4. 12 L standard single-use bag with integrated optical DO sensor (bag, Hegewald Medizinprodukte GmbH; DO sensor, PreSens GmbH).

110

2.2

Tim Bu¨rgin et al.

Cell Cultures

2.2.1 CHO Cells for FedBatch Cultivation

1. Cell line: CHO-K1 recombinant cell line expressing the monoclonal antibody Alemtuzumab as an example protein. 2. Medium: BalanCD Growth A basal medium containing 6 mM L-glutamine (Irvine Scientific) and supplemented with Pluronic (0.1 % V/V) and glucose (4 g/L and see Note 4). 3. Feeds: ActiCHO feed A CD and ActiCHO feed B CD (both GE Healthcare).

2.2.2 AGE1.CR.pIX Cells for Perfusion Cultivation

1. Cell line: Avian AGE1.CR.pIX suspension cells (ProBioGen AG). 2. Medium: Chemically defined CD-U3 medium (BiochromMerck) supplemented with 2 mM L-glutamine, 2 mM alanine, and 10 ng/L LONG R3 IGF-I (Sigma).

2.3 Cultivation Systems for Cell Cultures

1. Disposable Erlenmeyer shake flasks at different volumes.

2.3.1 CHO Cells

3. Same bioreactor and single-use bag as previously specified (see Subheading 2.1.2 and Note 6).

2.3.2 AGE1.CR.pIX Cells

1. Disposable 125 mL and 1 L Erlenmeyer shake flasks with baffles.

2. Incubator shaker with shaking orbit of 25 mm and CO2 and humidity control (see Note 5).

2. Incubator shaker with shaking orbit of 50 mm and CO2 control. 3. Same bioreactor and single-use bag as previously specified (see Subheading 2.1.2 and Note 7). Perfusion Set-Up

1. Perfusion system: Either ATF2 with C24U-v2 controller or TFF with centrifugal pump with magnetic levitation (PuraLev® 200MU, Levitronix) for recirculation. 2. Hollow fibers of polysulfone and polyethersulfone (UFP-500E-4X2MA, GE Healthcare, and F2:RF02PES, Repligen). 3. Peristaltic pumps for feeding and harvesting. 4. Tubing assembled with coupling inserts (MPC connectors) complementary with those of the single-use bag, used for inoculation, feeding, and perfusion. 5. STT quick connect couplings (Sartorius Stedim Biotech GmbH) to enable subsequent couplings out of the biological safety cabinet (see Subheading 3.4.2).

2.4

Influenza A Virus

1. Human influenza A/PR/8/34 H1N1 virus adapted to MCDK cells (Robert Koch Institute, Amp. 3138, TCID50 ¼ 1.23  108 virions/mL): all handling in a class II biological safety cabinet. 2. Trypsin (see Note 8) from a 5000 U/mL stock solution prepared in PBS, for virus infection.

Cultivation in Orbitally Shaken Bioreactor

3

111

Methods

3.1 kLa Measurements

The kLa measurements in the SB10-X are performed in dH2O (see Note 9) using the gassing-out method for orbitally shaken bioreactors as described by Tissot et al. [18], with some modifications. 1. Install the bag in the bioreactor vessel and fill it with the desired amount of dH2O, for example, by using a calibrated peristaltic pump. 2. Fully inflate the bag with air. 3. Gently shake the water while heating it up to 37  C. 4. In parallel, sparge N2 directly into the water to facilitate a faster drop in DO with less N2 consumption than via headspace gassing. During the sparging, make sure that one of the tubing lines on top of the bag is open to avoid an overpressure. Continue until the DO is close to 0 %. Measure the DO with the calibrated optical DO sensor integrated in the bag (see Note 10). For the sparging, a thin rigid copper tube connected to a N2 gas bottle can be used. Insert the copper tube via a port on the top of the bag (see Note 11). 5. Stop the shaking and flush air into the bag’s headspace to remove all N2 (see Note 12). 6. Initiate the data acquisition with a sampling rate of 1 s for the DO sensor. Then set the shaking and air gas flow rate to the desired values and start aeration. 7. Stop the measurement after the DO reached 100 %. For the evaluation of the kLa value, use the calculation proposed by DECHEMA (plotting the logarithmical, relative DO concentration as a function of time) [19]. Utilize the data from the linear phase of the DO curve (~10–90 %) for the calculation. Calculate the y-axis values using the following (Eq. 1): ln ð100  DOðt ÞÞ

ð1Þ

Plot the y-axis (from Eq. 1) as a function of time. Then insert a linear trend line with the following trend line (Eq. 2): y ¼mxþb

ð2Þ

The absolute value of the slope of the trend line (“m”) represents the kLa value. 8. Multiple measurements are recommended. Use the same dH2O and initiate the repeats at step 3 (see Note 13).

112

Tim Bu¨rgin et al.

3.2 Mixing Time Measurements

Mixing time studies can be done with the decolorization method described by Lo¨ffelholz et al. [20], with some modifications. The decolorization is based on a redox reaction: reduction of iodine in the liquid and oxidation of the added thiosulfate, which leads to the decolorization of the liquid. The emulsified starch in the liquid acts as an indicator [19] (see Note 14). 1. Fix a transparent plastic cylinder with the same dimensions as the bioreactor vessel on a standard shaker tray, for example, by using pins (which pinch the plastic cylinder) screwed into the tray (Fig. 1). 2. Place the tray on a table-top shaker with the required orbital shaking diameter and install the single-use bag in the cylinder. 3. Mount the high-speed camera on the same tray, directly opposite the cylinder. Illuminate with a lamp the liquid inside the bag (see Note 15). 4. Connect a silicone tube with an inner diameter of approx. 3 mm to the bag using one of the available bag port flanges. 5. Connect a 25 mL syringe (or similar) to the 1/800 tubing on the bag. Fill the syringe with the required volume of sodium thiosulfate (see Subheading 2.1.1 for volume ratios). Fill the rest of the syringe with air (which is required to push the entire sodium thiosulfate solution through the tube into the bag).

Fig. 1 Set-up of the mixing time experiment

Cultivation in Orbitally Shaken Bioreactor

113

6. Fill the single-use bag with the 0.8 % starch solution containing the Lugol’s solution. Fully inflate the bag to keep it in place inside the vessel. 7. Switch on the lamp and initiate image acquisition at 210 or more frames per second (see Note 16). 8. Start the shaker and add the sodium thiosulfate via the syringe once the shaking frequency is reached and the wave motion is steady (see Note 17). 9. After the decolorization reaction is complete, stop the filming and shaking. 10. Visually evaluate the video at the computer to quantify the time needed to achieve decolorization of the liquid following the addition of sodium thiosulfate. 11. Multiple measurements are recommended. Prior to each repetition, thoroughly rinse the bag with dH2O to remove any remaining sodium thiosulfate. Then refill the bag and follow the steps outlined above (see Notes 18 and 19). 3.3 Routine Subculture and Inoculum Propagation

1. Thaw at 37  C a cryo-vial containing 1  106 cells/mL and resuspend the cells in a shake flask with 19 mL of basal medium (BalanCD CHO Growth A supplemented with L-glutamine with an end concentration of 6 mM).

3.3.1 CHO Cells

2. Place the shake flask in an incubator shaker at 37  C at 120 rpm with an orbital shaking diameter of 25 mm. 3. Maintain the incubator environment at 85 % relative humidity, with 5 % CO2 maintained in the chamber. 4. After 24 h, count the cells and replace the basal medium with fresh medium. Then return the cells to the incubator shaker for another 24 h. 5. Passage the cells once the culture reached a cell density between 1 and 2  106 cells/mL (usually every third day) and seed at 2  105 cells/mL in fresh medium (see Note 20). 6. For the preparation of inoculum, increase the culture volume 2  (to 1 L) as soon as the cell density reached 2  106 cells/ mL. This is required for the bioreactor seeding. 7. Install a bag containing 7 L basal medium (same as above) in the bioreactor and seed it with 2  105 cells/mL (see Note 20).

3.3.2 AGE1.CR.pIX Cells

1. Cultivate the cells in 125 mL shake flasks, shaking at 185 rpm at 37  C with 5 % CO2. 2. Passage the cells every 3–4 days by inoculation with 0.8  106 viable cells/mL in 50 mL working volume.

114

Tim Bu¨rgin et al.

3. For inoculum propagation, seed the cells in larger flasks (1 L flasks, up to 600 mL working volume) and shake at 150 rpm using the same conditions. 3.4 Cultivation in the SB10-X Bioreactor 3.4.1 CHO Cell Cultivation in FedBatch Mode

1. Connect the bottle containing the inoculum to the SB10-X bioreactor and introduce the seed culture via gravity flow. 2. Start the cultivation with the following parameters (optimal parameters were previously determined): DO at 40 %, shaking frequency at 85 rpm, pH of 7.0, and gas flow (O2, CO2, air, and N2 connected) on demand at 300 mL/min. 3. For sampling, extract 4 mL of cell culture using a syringe connected via Luer Lock connection to the sampling port of the bag. Daily sampling is needed to monitor cell growth and viability, as well as the levels of glucose, galactose, and lactate over the cultivation period. Daily sampling also provides the opportunity to monitor the production of the target antibody, which helps to track the performance of the process. 4. Add glucose from a stock solution (30 % W/V) when the concentration drops below 2 g/L in order to maintain a minimum concentration of 3 g/L. 5. Four days after inoculation, implement a temperature shift from 37 to 32  C. 6. Maintain the following feeding conditions for the duration of the cultivation: l

Feed A: Add daily from day 3 onward 1 % (V/V) of total culture volume

l

Feed B: Add daily from day 3 onward 0.5 % (V/V) of total culture volume

l

Supplements: Uridine, manganese chloride, and galactose were added every day from the third cultivation day onward to reach concentrations of 1 mM, 0.0046 mM, and 5 mM, respectively (see also ref.21).

7. When the cell viability is reduced to 70%, stop the bioreactor run and harvest the culture medium. 8. The cultivation is monitored during the entire bioreactor run as illustrated in Fig. 2. A comparison of viable cell densities for cultures grown in shake flasks and the SB10-X bioreactor is shown in Fig. 3 (see Note 21). 3.4.2 AGE1.CR.pIX Cell Cultivation and Virus Infection in Perfusion Mode

1. After sterilization of all materials (see Note 22), connect the lines to the bioreactor via MPC connectors under the hood of a class II biological safety cabinet. For ATF operation, connect the SB10-X and the hollow fiber (HF) via one port located at the bottom of the bag (with the HF positioned below the bioreactor). The TFF set-up requires two connections of the

Cultivation in Orbitally Shaken Bioreactor

115

Fig. 2 Online monitoring of process parameters of SB10-X bioreactor run of a fed-batch CHO cultivation. Recombinant CHO-K1 cell line expressing the monoclonal antibody alemtuzumab was cultivated during a 16-day period

Fig. 3 Comparison of viable cell density and cell viability of cultures grown in shake flasks (200 mL working volume) and SB10-X bioreactor (8 L working volume) of a fed-batch CHO cultivation

bioreactor bag to the HF unit to link the bioreactor outlet at the bottom to the centrifugal pump, the pump to the lower inlet of the HF unit, and finally, the upper port of the HF unit to the upper inlet of the bioreactor bag (with the HF unit positioned next to the SB10-X).

116

Tim Bu¨rgin et al.

2. Fill the bag with air, add medium, and start the controllers (temperature, shaking frequency, and flow rate of the gas mixture). Following the equilibrium phase (typically overnight), calibrate the sensors (see Note 10). 3. Inoculate the bioreactor with 1.0–1.2  106 cells/mL and add medium to the desired initial working volume (see Note 23). Settings for cultivations: 37  C, working volume of 5–10 L, shaking frequency of 70–90 rpm (see Note 24), and aeration with a flow rate of 300–500 mL/min of a 20–40 % O2 (see Note 25) and 0–5 % CO2 gas mixture in N2. For cultivations with constant 10 L working volume, adjust parameters accordingly (see Note 26). 4. Start the perfusion mode approximately 2 days after inoculation, maintaining a cell-specific perfusion rate of 0.06 nL/cell/ d to support their glucose demand [22]. 5. Infect the AGE1.CR.pIX cells with influenza H1N1 virus (see Subheading 2.4) with a multiplicity of infection (MOI) of 103 infectious virions/cell. To enable virus entry, interrupt the perfusion for approximately 1 h (while recirculation is maintained). Two doses of 107 U/cell trypsin are administered: (1) at the time of infection and (2) during the early infection phase (up to 24 h post-infection, hpi) (see Note 27). Since trypsin (24 kDa) is smaller than the molecular cut-off of the membranes used to retain proteins during the perfusion, it will likely be washed out of the bioreactor over time. 6. Medium replacement before infection is not strictly necessary, as the CD-U3 medium is used for both the cell growth and virus production phases, but virus titers can be enhanced with medium addition [22] (see Note 28). For SB10-X cultivations, add fresh medium at the beginning of the virus production phase (up to 24 hpi) to increase the bioreactor working volume from 5 L up to 8 L (see Note 29). During the dilution step, maintain perfusion mode. Typical kinetics of the influenza virus production by AGE1.CR.pIX cells in the SB10-X bioreactor with perfusion is shown in Fig. 4 (see Note 30). 3.4.3 Analytical Assays

Harvested cell culture broth is centrifuged at 1000  g and the supernatant is filtered using a 0.45 mm filtered syringe. Capture of Protein A is done using an NAbTM spin column (Thermo Fisher), according to the manufacturer’s instructions. HPLC analysis is performed using an Agilent 1100 and an Aminex HPX-87H column (300  7.8 mm). SEC analysis is performed using an Agilent 1100n HPLC and a Yarra SEC-3000 column. CE analysis for glycan analysis is performed using a CE ProteomeLab PA800 with the Beckman Coulter carbohydrate labeling kit, according to the manufacturer’s instructions. Glycans released from the

Cultivation in Orbitally Shaken Bioreactor

117

Fig. 4 Influenza A virus production using AGE1.CR.pIX cells cultivated in an SB10-X bioreactor in perfusion mode. (a) Viable cell density and cell viability of cultivations using ATF (●) and TFF (○) systems during the cell growth phase at 5 L scale and after infection (post-infection, pi; infection at t ¼ 0 d). During the virus production phase, the bioreactor volume was increased to 7 L (ATF) or 8 L with (TFF). (b) Influenza A virus HA titer (HA activity)

antibodies by PNGase F (Peptide: N-glycosidase F) are labeled with the APTS (8-aminopyrene-1,3,6-trisulfonate) and separated through CE with laser-induced fluorescence (LIF) detection. The time course of the product titer during the bioreactor run is illustrated in Fig. 5. Production of the antibody until day 14 led to a final concentration of 2.13 mg/mL of alemtuzumab. Glycosylation analysis confirmed the required product quality and is illustrated in Fig. 6. 3.4.4 AGE1.CR.pIX Cells

4

Cell viability: Vi-Cell®XR (Beckman Coulter), protocol specific for AGE1.CR.pIX cells with a maximum relative standard deviation of 5 % [23]. Hemagglutination assay (HA): hemagglutination activity was analyzed following previously established protocols and is expressed in log10 HA units/100 μL [24].

Notes 1. The chance of clump formation can be reduced by adding the starch slowly and carefully into the stirred or shaken water. Grinding the starch through a sieve can also help. 2. If there are still clumps in the emulsion, manually mash them with a spoon and/or boil the emulsion.

118

Tim Bu¨rgin et al.

Fig. 5 Time course of product titer in SB10-X bioreactor of a fed-batch CHO cultivation. The product titer was measured by size exclusion chromatography. Quantification was performed using commercially available Lemtrada (originator protein Alemtuzumab) as a standard

Fig. 6 Capillary gel electrophoresis (CE) of APTS-labeled glycans released from the antibodies by PNGase F treatment. Addition of supplements as per ref. 21 resulted in a glycan profile very similar to the antibody produced by the originator. In particular, high level of galactosylated glycan forms was found

Cultivation in Orbitally Shaken Bioreactor

119

3. Add Lugol’s solution to the 0.8 vol. % starch solution immediately prior to the experiment. Light is bleaching iodine in a starch emulsion. 4. Media, feeds, and supplements need to be evaluated for each recombinant cell line. The listed elements have been identified as optimal for the tested cell line. 5. The optimal shaking frequency needs to be determined for each recombinant cell line by running shake flask cultures at different shaking frequencies. Viable cell density is the determining factor for scale-up toward the bioreactor scale. 6. It is recommended to run the culture in a minimum of 6 L working volume. Lower volumes result in foam formation that is difficult to control. 7. Standard bags (# SMX760001) can be used for perfusion operation, as there they provide a high number of connection ports. Perfusion bags were more recently developed (# SMX760003), and they were used for the latest experiments with constant 10 L working volume. The perfusion bag has two ports in the bottom, which enable connection to two perfusion devices simultaneously. In some experiments, two ATF systems were coupled to the bioreactor (see Subheading 3.4.2 and Note 26). 8. Trypsin activity from different lots can be different; therefore, the optimal trypsin concentration should be determined before infecting cell cultures with influenza A virus. 9. 1 PBS buffer or a NaHCO3 buffer solution may represent a cell culture medium better than dH2O. Using buffers may allow the determination of kLa values closer to cultivation media. 10. Normally, the calibration parameters of the optical sensors integrated in the bag are delivered by the bag manufacturer (two-point calibration at 0 and 21% air saturation). Those calibration parameters must be set in the data acquisition software prior to using the sensor for the measurement. Additionally, optical sensors require a few hours to equilibrate within the liquid. 11. The silicone tube of the port can be shortened to simplify pushing the copper tube into the bag. Remember to keep a few cm of the silicone tube to close it afterward with a pinch clamp. 12. If N2 supply is not shut off and the N2 is not completely removed from the headspace of the bag, the remaining N2 will influence the measurement. The driving force and saturation concentration is modified, which results in a reduced kLa value. If exhaust gas analysis is not available, it is recommended

120

Tim Bu¨rgin et al.

to exchange the headspace volume (nominal bag volume when inflated in the bioreactor vessel minus liquid volume) at least three times to ensure that all N2 is replaced by air [19]. 13. Using this method kLa values between 3.7 h1 (12 L at 80 rpm) and 50.6 h1 (4 L at 140 rpm) were measured in the SB10-X. The shaking speed range of the SB10-X bioreactor is 80–140 rpm. 14. This decolorization method captures the overall mixing time, making it possible to discover dead zones with very weak mixing. This is in contrast to local mixing time methods (e.g., with pH sensors in liquid), which can only report on areas where the sensors are placed. These local mixing time methods allow dead zones to remain undiscovered, which could compromise the measured mixing time. The sensors used for these local mixing time measurements can also act as baffles, which could influence the hydrodynamics also compromising the measured mixing time. 15. To avoid light reflection on the vessel wall, some trial and error (different angles, etc.) is required. Furthermore, the contrast of the discoloration reaction in the video can be improved and the interpretability enhanced by creating a white background behind the cylinder. It is important to maintain consistency in the lighting environment to eliminate any effects of shading, etc., which could introduce adverse effects on the interpretability. 16. Set the automatic focusing mode to “off.” Manual focusing of the cylinder, with respect to liquid, is done without shaking. This ensures a sharp focus on the liquid during the entire shaking phase. 17. Transfer the sodium thiosulfate into the bag as fast as possible in a single motion (slowly adding the solution could compromise the mixing time). Furthermore, keep the location of the introduced sodium thiosulfate constant. For example, in the middle of the vessel or close to the vessel wall. 18. Alternatively, the remaining sodium thiosulfate excess can be oxidized by adding dropwise Lugol’s solution (under constant shaking) until the liquid is colored as it was prior to the decolorization. 19. Using this method mixing times between 27 s (12 L at 80 rpm) and 4 s (3 L at 140 rpm) were measured in the SB10-X. The shaking speed range of the SB10-X bioreactor is 80–140 rpm. 20. Optimal seed densities need to be determined beforehand. 21. Lower viable cell densities and quicker decline of viability in shake flask cultures are a frequently confirmed observation. With an overall feed of 1.35 L, cell densities of

Cultivation in Orbitally Shaken Bioreactor

121

2.3  107 cells/mL were reached after 8 days of cultivation in the SB10-X. 22. Rinse the HFs with Milli-Q water for 60 min, and autoclave (121  C, 30 min) with the other materials needed for perfusion (listed in Subheading 2.3.2.1). 23. Although the range of working volume in the SB10-X is 3–12 L, we recommend a minimum of 5 L when working with HF-based perfusion to avoid air in the HF. Because the connection of the bioreactor with the perfusion system is done via the bottom part of the bag, air can enter the HF due to wave movement when the working volume is low (e.g., 3 L). 24. Initial agitation is 70 rpm for cultivations with 5 L working volume. During virus production, after the bioreactor volume is increased to 7–8 L, increase the shaking frequency to 90 rpm. 25. After trypsin addition at time of infection, the demand for oxygen is increased, occasionally leading to a drop in DO levels. Therefore, the gas mixture is enriched with oxygen before infection by manually increasing its percentage (to 35–60 %), preventing a DO drop to critical levels that would negatively impact (the cell growth and) virus production. 26. Experiments were carried out to evaluate the performance of the system in perfusion mode at 10 L working volume. The bioreactor was operated at constant volume, since inoculation. Initial experiments with the settings of cultivations at 5 L scale (up to 8 L after infection), or with increased shaking frequency, resulted in unsatisfactory cell growth. It was necessary to increase the shaking frequency (100–120 rpm) and the gas flow rate (600–800 mL/min) to sustain cell growth at high cell densities in perfusion mode. The shaking frequency was increased from 100 to 110 rpm during the cell growth phase and to 120 rpm for the virus production phase. One ATF 2 system was enough for perfusion operation, even at 10 L scale. Nonetheless, a second ATF 2 system was coupled to the bioreactor as a backup. 27. Trypsin is necessary for efficient influenza propagation in cell culture, particularly for multi-cycle replication. The addition of trypsin leads to virus titer increase by different mechanisms, such as HA cleavage [25] and interference with the antiviral host cell defense [26]. 28. Different strategies can be applied depending on the process. Partial medium exchange at time of infection can be useful (e.g., influenza A virus production by CAP cells [27]). Or a hybrid fed-batch/daily medium exchange (perfusion) process, with medium added at defined periods of time after infection, followed by perfusion (e.g., MVA virus propagation in AGE1.

122

Tim Bu¨rgin et al.

CR.pIX cells [28]). The optimal time for operation in each mode should be determined depending on the virus replication dynamics. 29. For 10 L cultivations, an alternative method for adding fresh medium is used, as the working volume is already very high. The bioreactor working volume is decreased to 5 L and, subsequently, fresh medium added to a final volume of 10 L, leading to a 50% medium exchange. 30. Using this method, maximum values of 4.0–4.3  107 cells/ mL were obtained before infection, and maximum virus titers between 3.64 and 3.73 log10 (HA units/100 μL) were reached 1–2 days post-infection. Similar results were obtained for the 10 L cultivation using the ATF system, although the cells were infected at a lower cell density to reduce the process time and medium consumption (maximum of 3.5  107 cells/mL before infection and maximum titer of 3.75 log10 (HA units/ 100 μL) at 1 day post-infection).

Acknowledgments The authors thank V. Sandig (ProBioGen AG) for kindly providing the AGE1.CR.pIX cell line and I. Behrendt (MPI Magdeburg) for the excellent technical work. References 1. Raval K, Kato Y, Bu¨chs J (2007) Comparison of torque method and temperature method for determination of power consumption in disposable shaken bioreactors. Biochem Eng J 34(3):224–227 2. Bu¨chs J, Zoels B (2011) Evaluation of maximum to specific power consumption ratio in shaking bioreactors. J Chem Eng Japan 34 (5):647–653 3. Klo¨ckner W, Bu¨chs J (2012) Advances in shaking technologies. Trends Biotechnol 30 (6):307–314 4. Anderlei T, Cesana C, Bu¨rki C, De Jesus M, Ku¨hner M, Wurm F, Lohser R (2009) Shaken bioreactors provide culture alternative. Gen Eng Biotechnol News 29:19 5. Raval K, Liu C-M, Bu¨chs J (2006) Large-scale disposable shaking bioreactors. BioProcess Int 41(1):46–49 6. Selker M, Paldus B (2009) Single-use solutions for scale-up and technology transfer. Innov Pharm Technol:57–59 7. Margaritis A, Zajic JE (1978) Mixing, mass transfer, and scale-up of polysaccharide

fermentations. Biotechnology and bioengineering 20:939–1001. https://doi.org/10. 1002/bit.260200702 8. Junker BH (2004) Scale-up methodologies for Escherichia coli and yeast fermentation processes. J BiosciBioeng 97(6):347–364. https://doi.org/10.1016/s1389-1723(04) 70218-2 9. Garcia-Ochoa F, Gomez E (2009) Bioreactor scale-up and oxygen transfer rate in microbial processes: an overview. Biotechnol Adv 27 (2):153–176. https://doi.org/10.1016/j.bio techadv.2008.10.006 10. Kieran PM, Malone DM, MacLoughlin PF (2000) Effects of hydrodynamic and interfacial forces on plant cell suspension systems. Adv Biochem Eng/Biotechnol 67:139–177 11. Al-Sabbagh A, Olech E, McClellan JE, Kirchhoff CF (2016) Development of biosimilars. Semin Arthritis Rheumatism 45(5 Suppl): S11–S18. https://doi.org/10.1016/j. semarthrit.2016.01.002 12. Schiestl M, Stangler T, Torella C, Cepeljnik T, Toll H, Grau R (2011) Acceptable changes in

Cultivation in Orbitally Shaken Bioreactor quality attributes of glycosylated biopharmaceuticals. Nat Biotechnol 29:310–312. https://doi.org/10.1038/nbt.1839 13. Cymer F, Beck H, Rohde A, Reusch D (2018) Therapeutic monoclonal antibody N-glycosylation – structure, function and therapeutic potential. Biologicals 52:1–11. https://doi.org/10.1016/j.biologicals.2017. 11.001 14. Higel F, Seidl A, Sorgel F, Friess W (2016) N-glycosylation heterogeneity and the influence on structure, function and pharmacokinetics of monoclonal antibodies and Fc fusion proteins. EurJPharmBiopharm 100:94–100. https://doi.org/10.1016/j.ejpb.2016.01.005 15. Tapia F, Va´zquez-Ramı´rez D, Genzel Y, Reichl U (2016) Bioreactors for high cell density and continuous multi-stage cultivations: options for process intensification in cell culture-based viral vaccine production. Appl MicrobiolBiotechnol 100:2121–2132. https://doi.org/10. 1007/s00253-015-7267-9 16. Gallo-Ramı´rez LE, Nikolay A, Genzel Y, Reichl U (2015) Bioreactor concepts for cell culturebased viral vaccine production. Expert RevVaccines 14:1181–1195. https://doi.org/10. 1586/14760584.2015.1067144 17. Gutie´rrez-Granados S, Go`dia F, Cervera L (2018) Continuous manufacturing of viral particles. Curr Opin Chem Eng 22:107–114. https://doi.org/10.1016/j.coche.2018.09. 009 18. Tissot S, Mickel P, Hacker DB, De Jesus M, Wurm F (2012) kLa as a predictor for successful probe-independent mammalian cell bioprocesses in orbitally shaken bioreactors. New Biotechnol 29(3):387–394. https://doi.org/ 10.1016/j.nbt.2011.10.010 19. Meusel W, Lo¨ffelholz C, Husemann U, Dreher T, Greller G, Kauling J, Eibl D, Kleebank S, Bauer I, Glo¨ckler R, Huber P, Kuhlmann W, John GT, Werner S, Kaiser SC, Po¨rtner R, Kraume M (2016) Recommendations for process engineering characterisation of single-use bioreactors and mixing systems by using experimental methods. DECHEMA 20. Lo¨ffelholz C, Husemann U, Greller G, Meusel W, Kauling J, Ay P, Kraume M, Eibl R, Eibl D (2012) Bioengineering parameters for single-use bioreactors: overview

123

and evaluation of suitable methods - Lo¨ffelholz - 2013 - Chemie Ingenieur Technik - Wiley Online Library. https://doi.org/10.1002/ cite.201200125 21. Gramer MJ, Eckblad JJ, Donahue R, Brown J, Shultz C, Vickerman K, Priem P, van den Bremer ET, Gerritsen J, van Berkel PH (2011) Modulation of antibody galactosylation through feeding of uridine, manganese chloride, and galactose. Biotechnol Bioeng 108 (7):1591–1602. https://doi.org/10.1002/ bit.23075 22. Va´zquez-Ramı´rez D, Genzel Y, Jordan I, Sandig V, Reichl U (2018) High-cell-density cultivations to increase MVA virus production. Vaccine 36:3124–3133. https://doi.org/10. 1016/j.vaccine.2017.10.112 23. Lohr V (2014) Characterization of the avian designer cells AGE1.CR and AGE1.CR.pIX considering growth, metabolism and production of influenza virus and Modified Vaccinia Virus Ankara (MVA) 24. Kalbfuss B, Knochlein A, Krober T, Reichl U (2008) Monitoring influenza virus content in vaccine production: precise assays for the quantitation of hemagglutination and neuraminidase activity. Biologicals 36(3):145–161 25. Klenk HD, Rott R, Orlich M, Blodorn J (1975) Activation of influenza A viruses by trypsin treatment. Virology 68(2):426–439 26. Seitz C, Isken B, Heynisch B, Rettkowski M, Frensing T, Reichl U (2012) Trypsin promotes efficient influenza vaccine production in MDCK cells by interfering with the antiviral host response. Appl MicrobiolBiotechnol 93 (2):601–611. https://doi.org/10.1007/ s00253-011-3569-8 27. Genzel Y, Vogel T, Buck J, Behrendt I, Ramirez DV, Schiedner G, Jordan I, Reichl U (2014) High cell density cultivations by alternating tangential flow (ATF) perfusion for influenza A virus production using suspension cells. Vaccine 32:2770–2781 28. Va´zquez-Ramı´rez D, Jordan I, Sandig V, Genzel Y, Reichl U (2019) High titer MVA and influenza A virus production using a hybrid fed-batch/perfusion strategy with an ATF system | SpringerLink.https://doi.org/10.1007/ s00253-019-09694-2

Chapter 8 Development of Mammalian Cell Perfusion Cultures at Lab Scale: From Orbitally Shaken Tubes to Benchtop Bioreactors Moritz Wolf and Massimo Morbidelli Abstract This chapter introduces the necessary concepts to develop mammalian cell perfusion cultures for the expression of therapeutic proteins at lab scale. We highlight the operation of the orbitally shaken tubes and of a classical glass vessel reactor system coupled to an external alternating tangential flow (ATF) device. Two different experiments can be performed in the shake-tube system: (1) the VCDmax experiment exploring the maximum achievable viable cell density at a given medium exchange rate and (2) the VCDSS experiment for the prediction of process performance at constant viable cell density and a given medium exchange rate for the design of the benchtop bioreactor process. In addition, the operation of the benchtop system is discussed containing start-up and control procedures for a long-term production run. Key words Scale-down models, Shake tubes, CHO cells, Perfusion process development, Benchtop bioreactors

1

Introduction

1.1 Continuous Production of Biopharmaceuticals

In the last decade, the idea to move toward the integrated continuous manufacturing (ICB) of biopharmaceuticals has become one of the dominating topics within the community accompanied by advances in continuous up- and downstream technologies as well as process automation [1–5]. Continuous operations do not only decrease capital and operating costs but enable better process and product quality control, process flexibility, and, in particular, process intensification [6–13]. In this context, mammalian cell perfusion cultures have not only been reconsidered as the essential first step within an integrated continuous end-to-end production facility but even more as a general production platform for therapeutics. The use of perfusion processes is not a new technology in bioprocessing. In fact, it has been known and applied in biotechnology for many years, especially for the production of fragile proteins, such as factor VIII, interferon beta-1a, or infliximab [14, 15]. Nevertheless, as a result of recent technology advances,

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_8, © Springer Science+Business Media, LLC, part of Springer Nature 2020

125

126

Moritz Wolf and Massimo Morbidelli

such as the development of the alternating tangential flow (ATF) filtration devices for cell retention, combined with increasing research interest, intensified perfusion processes with volumetric productivities of above 1 g/L/day have been established that outperform state-of-the-art fed-batch processes [16, 17]. Independent of the concept of process integration, the recent advances pave the way toward a more general application of continuous cell cultures. 1.2 Perfusion Process Performance

The development and optimization of a perfusion process requires the screening of multiple operating conditions. The target is to develop a stable and robust perfusion process. Such a process is aimed to be operated at a fixed viable cell density and perfusion rate set point in order to enable steady-state operation. This requires to keep the cells in a growing state for the entire run and to balance the nutrient supply to prevent nutritional limitations and inhibitions potentially inducing cell death. Operation at constant viable cell density requires to balance the proliferation of cells in a growing state by the removal of additional cells from the reactor [18]. Therefore, a cell-containing bleed stream is necessary [18]. As the bleed stream contains cells and the protein of interest, this stream has to be regarded as product loss since further processing of this cellcontaining stream is challenging. As a result, cellular growth should be minimized to a sustainable minimum, such that the bleed stream constitutes 5–10% of the overall medium exchange rate [19, 20]. The evaluation and comparison of the performance of perfusion processes requires the introduction of process-specific parameters. The overall medium exchange or dilution rate, often called perfusion rate, P, equals the sum of both outlet streams, the cellfree harvest rate, H, and the cell-containing bleed rate, B.   RV P ¼H þB ð1Þ Day The cell-specific perfusion rate (CSPR) characterizes the amount of medium fed to a single cell per day and is defined by the ratio of the perfusion rate (RV/Day) and the viable cell density, XV, (106 cells/mL).   P pLmedium CSPR ¼ ð2Þ X V Cell day The CSPR indicates the performance of the used culture medium, the so-called medium depth. For the operation at a process-specific optimum in terms of viable cell density, medium consumption, and productivity, it is important to assess the potential of the given expression system (combination of cell clone and medium), typically referred to as minimum CSPR (CSPRmin) [21, 22]. At CSPRmin, the ratio of perfusion rate and viable cell density cannot be further decreased without compromising cellular

Development of Mammalian Cell Perfusion Cultures at Lab Scale

127

viability or productivity. Extensive studies of perfusion processes, cell lines, and medium formulation in the last two decades resulted in essential improvements of the process performance indicated by the decrease of accomplished CSPRs that have been reduced from 100 pL/cell/day down to 15 pL/cell/day [23–25]. 1.3 Process Development at Lab Scale

The large number of cultivations needed, in general, for the development of a perfusion process requires systems that can be operated at sufficiently small volume (μL or mL scale). In addition, these systems should allow running several experiments in parallel in order to reduce media consumption and to save time, especially when considering clone selection, media optimization, and the prescreening of suitable operating conditions for perfusion. Due to the continuous nature of perfusion processes, moreover, the system needs to be able to mimic the continuous operation of the reactor including cell removal and cell-free harvesting. Promising tools include (1) the use of the ambr®15 system by applying cell sedimentation to separate cells from culture medium followed by partially replacing the spent medium without the use of a cell retention device; (2) the use of shake-tube (ST) bioreactor system, a less expensive small-scale reactor system, by applying centrifugation to separate cells from the medium followed by medium replacement; and (3) the application of deep well plates (DWP) by also applying sedimentation for the separation of cells from medium and the consecutive medium exchange [26–30]. The screening of steady-state conditions, however, requires to not only exchange cell-free medium but also to include a cell-removing bleed step, where a portion of the entire culture is removed from the system in order to operate close to a target VCD set point and to provide a closer representation of the actual perfusion bioreactor. In addition, there have been recent demonstrations of high cell density perfusion cultures at up to 100  106 cells/mL at the 250 mL scale in more automated systems, such as the ambr® 250 perfusion that can be used to design benchtop liter-scale systems [31, 32]. Such benchtop bioreactor systems allow, in a next step, the design of process operating conditions in order to maximize reactor productivity, minimize medium consumption, and produce the product at desired product quality. Further scale-up to the commercial scale can be done based on classical engineering procedures for three-phase stirred tank reactors by applying, e.g., computational fluid dynamics [33–35]. Within this chapter, we present the strategies for the operation of the abovementioned shake-tube and benchtop bioreactors to design and develop mammalian cell perfusion processes at lab scale.

128

2

Moritz Wolf and Massimo Morbidelli

Materials The list of materials provided below is adapted from the one for upstream processing presented in [36] and contains all the equipment necessary to perform the presented experiments.

2.1 Shake-Tube Experiments

1. Shake-tube bioreactors (TubeSpin® bioreactor 50, working volume 1–35 mL, TPP, Switzerland) for all experiments performed in shake tubes. 2. A shaking incubator with heating system, humidity, and CO2 control (e.g., ISF1-X or LT-X, Adolf Kuhner AG, Switzerland). 3. Laminar flow cabinet for sterile working equipped with a Pipetteboy, 200 and 1000 μL pipettes and pipette tips, as well as a rack for shake-tube bioreactors. 4. A centrifuge for the shake-tube system (e.g., Centrifuge 5810/ 5810 R, Vaudaux-Eppendorf AG, Switzerland) for the separation of cells and medium.

2.2 Benchtop Bioreactor Experiments

1. A benchtop bioreactor system (e.g., Vaudaux-Eppendorf AG, Switzerland) consisting of a set of glass vessels (1.0–2.0 L working volume); a reactor control system with implemented modules for monitoring and control of pH and DO, heating and stirring, gas flow rate, and mixing; as well as additional peristaltic pump modules is used for the realization of a lab-scale perfusion system. Alternatively also wave-bag bioreactors can be used [37]. Additionally, the process control computer needs to allow the implementation of a biocapacitance probe (e.g., ABER Instruments Ltd, UK) and the connection of balances (e.g., Mettler Toledo, USA) for online cell density and reactor weight control, respectively. 2. A cell retention device is necessary to operate at high cell densities and to harvest cell-free supernatant continuously. Currently, the integration of cross-flow filtration devices using PES hollow fiber modules (e.g., 25 cm length, pore size of 0.5 μm, filtration area of 1570 cm2, 1 mm fiber diameter) represents the state of the art, typically coupled with either an alternating tangential flow (ATF) system or a tangential flow filtration (TFF) device. For lab scale, an ATF2 device (Repligen Cooperation, USA) or a bearingless centrifugal pump for the TFF operation (PuraLev MU200, Levitronix GmbH, Switzerland) is the proper system.

2.3 Cell Line and Medium

1. A cryopreserved CHO cell line (5–10  106 cells/mL) producing a therapeutic protein stored in culture medium including 10% DMSO in liquid nitrogen is typically used. For the cultivation, a chemically defined medium for CHO cell culture

Development of Mammalian Cell Perfusion Cultures at Lab Scale

129

is necessary, preferably a commercially available or supplied medium formulation. 2.4 Additional Equipment

1. Additional types of containers (e.g., 600 mL TubeSpin bioreactor, TPP, Switzerland, or roller bottles, Corning®, USA) are needed for the cell expansion when the volume of the expansion is exceeding 35 mL. 2. For in-process control (IPC), a mammalian cell counting device is necessary, e.g., a trypan blue exclusion methodbased instrument, such as the Cedex HiRes (Roche Diagnostics, Switzerland). 3. An analyzer for bioprocesses needs to be used for daily offline measurements of key metabolites (glucose, lactate, ammonia), e.g., Cedex Bio (Roche Diagnostics). The protein concentration can be either analyzed with the Cedex Bio (Roche Diagnostics) or using ProtA affinity chromatography. 4. For preparation of cell culture medium in large amounts, suitable bags equipped with connections to the bioreactor system are recommended (e.g., BPC, 100L, Life Technologies). 5. In the case of sterile filtration of the prepared media, suitable filters are required, e.g., Sartobran® 300 (Sartorius, Germany). 6. The sterile connection of different systems can be performed by a sterile tubing welder (Terumo, Japan) in combination with appropriate tubing (e.g., PharMed® BPT tubing, SaintGobain, France).

3

Methods In this section, we will present the operation of shake-tube (Fig. 1) and benchtop (Fig. 2) bioreactors in perfusion mode. In the case of the orbitally shaken tubes (OrbShake tubes), we distinguish two different experiments referred to as VCDmax and VCDSS experiment. The first one evaluates the maximum achievable VCD for a given perfusion rate (without cell-containing bleed), while the second one targets to mimic steady-state perfusion processes (with cell-containing bleed). In the case of the benchtop system, we present the operation of a perfused production reactor at constant cell density.

3.1 Medium Preparation

1. Medium preparation has to be performed according to the supplier’s protocol. For the experiments in the shake-tube system (working volume: 10 mL) operated at a medium exchange volume of one reactor volume per day, we require 100–200 mL medium per experiment depending on the run duration. For the operation of the bench-scale system, typically,

130

Moritz Wolf and Massimo Morbidelli

ml 45 40 35 30 25 20 15 10

Fig. 1 Schematic of an orbitally shaken bioreactor. Oxygenation and carbon dioxide removal occur via passive gas ventilation through the holes in the reactor head (dashed two-sided arrow) Bioreactor setup

Retention device

Bleed Feed Harvest

Fig. 2 Schematic of the bioreactor setup for bench-scale experiments. The setup consists of a bioreactor part containing all the relevant probes (pH, DO, biomass), stirrer, and aeration, as well as ports for media addition and cell removal, and a cell retention device (such as the alternating tangential flow filtration system) coupled to a hollow fiber device for cell-free harvest. A detailed configuration is explained in [36]

a perfusion run requires large media amounts. Operated at a working volume of 1.5 L at a perfusion rate of one reactor volume per day during 30 days, a perfusion process consumes at least 45 L medium. It is recommended to prepare at least 20–25 L of medium at a time.

Development of Mammalian Cell Perfusion Cultures at Lab Scale

131

After preparation and pH adjustment (see Note 1), the medium can be sterile filtered into the corresponding medium bag or bottle. 3.2

Cell Expansion

1. Cells stored in a working cell bank in liquid nitrogen need to be thawed and expanded in expansion medium containing an additional selection chemical. 2. The cultivation is performed in shake tubes at 36.5  C, 5% CO2, and 320 rpm in a humidified incubator for one week, passaging every second day. 3. At each passage state, cells are diluted to 0.3  106 cells/mL in incrementally larger containers (shake tubes or roller bottles) to accumulate a sufficiently high cell number for the inoculation of either the shake-tube-based perfusion bioreactors, an N1 perfusion bioreactor, or direct inoculation to the perfused production bioreactor for the operation in the benchscale system (see Note 2).

3.3 Shake-Tube Bioreactor Operation 3.3.1 VCDmax Experiment

1. Shake tubes are inoculated from the expansion at an initially low cell density, e.g., 0.5  106 cells/mL, and placed inside the  humidified incubator (36.5 C, 5% CO2, and 300–320 rpm) for 24 h. The working volume should range between 10 and 20 mL. 2. After 24 h, 500 μL are withdrawn from the reactor for a cell count and metabolite analysis. 3. Then, the shake tube is placed in the centrifuge for 5 min at 800 rpm in order to separate medium and cells. 4. The cell-free supernatant is removed/decanted and after filtration stored for further analysis. 5. The remaining cell pellet is resuspended in fresh cell culture medium. 6. Steps 2–5 can be repeated on a daily basis (every 24 h). The shake tube should be exchanged every 3 days to prevent the accumulation of cell debris [28]. The procedure is illustrated in Fig. 3. By repeating this procedure on a daily basis, the maximum achievable viable cell density (VCDmax) at the given medium exchange rate can be evaluated. After this peak is achieved, the amount of fresh nutrients supplied per cell per day becomes insufficient, and cell density stops to increase or eventually starts to decrease as illustrated in Fig. 4. Given the VCDmax and the applied medium exchange rate, we can calculate the critical CSPR, CSPRcrit, according to Eq. 2. As we remove a proper fraction of cells for the daily cell count, a small bleed is applied every day. Overall, the bleed rate depends on both, the removed cell volume and the working volume.

132

ml 45 40 35 30 25 20 15 10

Moritz Wolf and Massimo Morbidelli

Sample VCD

ml 45 40 35 30 25 20 15 10

Spin down

ml 45 40 35 30 25 20 15 10

Harvest

ml 45 40 35 30 25 20 15 10

Resuspend

ml 45 40 35 30 25 20 15 10

Fig. 3 Procedure for VCDmax experiments in the shake-tube bioreactors. After inoculation, the shake tube is placed for 24 h in the shaking incubator. On a daily basis the following steps are performed: (1) prepare 500 μL sample for VCD count and metabolite analysis, (2) spin-down the cells, (3) harvest the cell-free supernatant, and (4) resuspend cells in fresh medium. Steps 1–4 are repeated every 24 h

Fig. 4 Viable cell density (squares) and viability (circles) as a function of time in the VCDmax experiment. After day 3, a daily medium exchange of 1 RV/day is performed. After reaching the VCDmax (red dashed line), the cell density starts to drop slowly



V Sample 1 V tot Δt

ð3Þ

The harvest rate can be calculated according to Eq. 1. In most of the cases, the actual bleed rate will constitute between 5 and 10% within this procedure. 3.3.2 VCDss Experiment

1. The cell expansion is used to prepare a high cell density seeding solution. For this, the last step of the expansion should be carried out in containers that fit in a centrifuge (e.g., 600 mL TubeSpin bioreactor) 2. Cells are spun down, the supernatant is removed, and cells are resuspended with a defined volume of fresh medium to achieve the desired cell concentration. For example, for an expansion cell culture at 1.2  106 cells/mL in 400 mL media, cells can be resuspended in 15 mL fresh medium to obtain a high seeding solution of 32  106 cells/mL.

Development of Mammalian Cell Perfusion Cultures at Lab Scale

133

3. Shake tubes are inoculated from the high seeding solution at a high cell density, the VCD set point, XV, SP, close to the targeted steady-state cell density (see Note 3), and placed inside the humidified incubator (36.5  C, 5% CO2, and 300–320 rpm) for 24 h. The working volume should range between 10 and 20 mL (see Note 4). 4. After 24 h, 500 μL are withdrawn from the reactor for a cell count and metabolite analysis. 5. In order to keep the VCD close to a desired steady-state value, a cell-removing bleed step is introduced. Based on the measured VCD value, XV, meas, the bleed volume VBleed can be calculated.  X V ,meas  X V ,SP V tot V Bleed ¼ ð4Þ X V ,meas where Vtot represents the working volume. The VCD set point has to be chosen carefully so that the average of the given set point, XV, SP, and the measured VCD values, XV, meas, is close to the targeted steady-state value. 6. In order to remove the proper amount of cells, the sample volume has to be taken into consideration. V Remove ¼ V Bleed  V Sample

ð5Þ

7. After removing the additional cells, the shake tube is placed in the centrifuge for 5 min at 800 rpm in order to separate medium and cells. 8. Based on the chosen medium exchange rate or volume, VExchange, the harvest volume VHarvest can be calculated. V Harvest ¼ V Exchange  V Bleed

ð6Þ

where VExchange represents the targeted volume that should be exchanged. In the case VExchange is one full reactor volume, VExchange equals Vtot. The calculated cell-free supernatant is removed/decanted and after filtration stored for further analysis. 9. The remaining cell pellet is resuspended in fresh cell culture medium. 10. The corresponding flow rates are then calculated as follows: B¼ H ¼

V Bleed 1 V tot Δt

ð7Þ

V Harvest 1 V tot Δt

ð8Þ

11. Steps 4–10 can be repeated on a daily basis (every 24 h). As for the VCDmax experiment, it is recommend to transfer the

134

ml 45 40 35 30 25 20 15 10

Moritz Wolf and Massimo Morbidelli

Sample VCD

ml 45 40 35 30 25 20 15 10

Bleed, Spin down

ml 45 40 35 30 25 20 15 10

Harvest

ml 45 40 35 30 25 20 15 10

Resuspend

ml 45 40 35 30 25 20 15 10

Fig. 5 Procedure for VCDSS experiments in the shake-tube bioreactors. After inoculation close to the target viable cell density, the shake tube is placed for 24 h in the shaking incubator. On a daily basis the following steps are performed: (1) prepare 500 μL sample for VCD count and metabolite analysis; (2) cells are removed to decrease the viable cell density down to the set point, XV, SP; (3) spin-down the cells; (4) harvest the cell-free supernatant; and (5) resuspend cells in fresh medium. Steps 1–5 are repeated every 24 h

culture every 3 days to a new shake-tube reactor. The procedure is illustrated in Fig. 5. By repeating this procedure on a daily basis, the steady-state performance at a targeted viable cell density (VCDSS) and given medium exchange rate can be evaluated. This includes the prediction of cellular activities at a given target VCD, such as the growth rate, the metabolic production (mAb, AMM) and consumption (GLC) rates, as well as the prediction of the ratio of the harvest and the bleed rate [29]. A suitable duration of this experiment is 7–10 days as illustrated in Fig. 6. We can perform several experiments at a time, evaluating the steady-state performance of different target viable cell densities at a given perfusion rate, or of a single viable cell density at various perfusion rates. 3.4 Benchtop Bioreactor Operation

3.4.1 Production Bioreactor

The procedure for the bioreactor operation has been previously discussed in the context of the continuous expression and production of therapeutic proteins [36], where the authors describe a detailed preparation and inoculation procedure for mammalian cell perfusion cultures. The different calibration procedures for the different reactor probes (pH, DO, pumps) are further discussed in [38]. 1. The production bioreactor is inoculated from either the expansion at a low viable cell density or at a high cell density close to the viable cell density set point from an N1 seed bioreactor after additional 7–10 days, depending on the growth properties of the cell line. Cells can be transferred via sterile connection (see Note 5). 2. Upon inoculation, the essential control loops have to be initiated: either (1a) constant feed/perfusion rate or (1b) constant harvest rate, (2a) gravimetric feedback loop of the harvest pump or (2b) gravimetric feedback loop of the feed pump, or (3) cell density control of the bleed pump (see Note 6).

Development of Mammalian Cell Perfusion Cultures at Lab Scale

135

Fig. 6 Process performance in the VCDSS experiments in shake-tube bioreactors. (a) Viable cell density (squares) and viability (circles), (b) flow rates (feed, harvest, bleed), (c) GLC and LAC, and (d) titer as a function of time. The dashed lines represent (a) the targeted VCD of 20  106 cells/mL, (b) the daily medium exchange rate of 1 RV/day, and (c) the glucose concentration in the feed. The red lines in (a) represent the additional cell bleed

3. The cell density set point should be selected carefully based on preliminary studies on the medium depth/CSPRmin performed in the shake-tube experiments (Subheading 3.3). A biocapacitance probe can be used to monitor the biomass/viable cell density inside the bioreactor. The daily offline VCD measurements can be compared to the online biomass/VCD value, which allows a direct adjustment of the biomass conversion factor according to the offline VCD measurements. This can be necessary in order to account for changes in cell diameter at the beginning of the culture when cells transition from the growing to the steady-state phase. The combination of an N1 perfusion seed bioreactor that allows direct inoculation close to the cell density set point and immediate start of perfusion operation facilitates a fast transition to a steady-state bioreactor operation. The operation can be maintained for the desired process duration (usually 20–60 days).

136

Moritz Wolf and Massimo Morbidelli

4. At the end of the culture, all control loops can be stopped including disconnecting the probes, turning off cooling of the condenser, clamping off and disconnecting the different tubes, and a careful cleaning of all probes and all the parts of the reactor vessel. 3.4.2 Bioreactor Monitoring

1. Daily in-process control (IPC) includes the measurement of all cell-related characteristics (viable cell density, viability, diameter, and aggregation rate) preferably using the trypan blue exclusion method (e.g., Cedex HiRes). Additionally, offline pH measurements of the cell culture broth enable to guarantee a proper function of the online pH probe. 2. Samples for metabolite analysis are withdrawn from the reactor and the harvest to analyze main metabolites (glucose, lactate, potentially ammonia, and IgG titer) using a proper cell culture analyzer. Samples for further analysis of product quality can be kept frozen. For analytical methods (see Note 7). 3. Additionally, in-process control involves a daily check of the different flow rates and the corresponding pumps and tubing. The tubing should be adjusted every 3–4 days in order to prevent physical tubing damage. Bleed and harvest rate can be calculated and compared based on the weight change of the different hold-up tanks capturing the outlet flows. In the case of a constant harvest rate, the pump rate has to be adjusted accordingly. With illustrative purposes the time profiles of the daily offline measurements as well as N-linked glycosylation and charge variant distributions are shown in Figs. 7 and 8. In this case, the targeted viable cell density set point was 20  106 cells/mL and the harvest rate was set to 0.65 RV/ day to operate at an overall perfusion rate of 1.0 RV/day. In case of strong deviations between online biomass reading and offline viable cell density count, the probe’s biomass conversion factor has to be corrected. 4. In a last step of the IPC, online monitored values (reactor weight, pump flow rates, gas flow rate and gas fractions, pH, DO, temperature) shall be checked as they are indicative for the current status of the operation (stable operation, steady state, pump status).

4

Notes 1. In order to avoid excessive addition of base during the perfusion process, a slightly basic medium is prepared (pH adjustment to pH 7.2). During the process, pH is maintained by balancing the CO2 fraction in the inlet gas and the produced lactate in the culture.

Development of Mammalian Cell Perfusion Cultures at Lab Scale

137

Fig. 7 Process performance in the benchtop bioreactor for an experiment targeting steady-state operation at 20  106 cells/mL. (a) Viable cell density (squares) and viability (circles), (b) flow rates (feed, harvest, bleed), (c) GLC and LAC, and (d) titer as a function of time. The dashed lines represent (b) the fixed harvest rate of 0.65 RV/day and (c) the glucose concentration in the feed

Fig. 8 Resulting product quality in the benchtop bioreactor at 20  106 cells/mL. (a) N-linked glycosylation patterns and (b) charge variant distribution as a function of time

138

Moritz Wolf and Massimo Morbidelli

2. The implementation of an N1 perfusion bioreactor enables the split of the cell accumulation phase and the production phase at constant cell density. However, the split into two reactors requires additional equipment. 3. The inoculation of the steady-state shake-tube runs should be close, but below the targeted VCD. The former guarantees to operate directly around the targeted VCD set point and thus reduces the time required for the experiment. The latter is necessary, as an inoculation too close or above the target VCD could result in overgrowing of the cells and such induced cell death. 4. In the case of the steady-state operation in shake tubes, in particular at higher viable cell densities (above 20  106 cells/mL), oxygenation can become limiting. Therefore, a suitably low working volume guarantees that sufficient air can be supplied to the cells [29, 39]. 5. For sterile connection of the reactor to the N1 seeding reactor or the bottle containing the cells from the expansion, sterile welding is recommended. In addition, the transfer is facilitated by applying in either the seed reactor or the inoculation bottle not exceeding 1.5–2 bar. 6. The implemented control loops result in different pump flow rates. In the case of controlling the harvest rate, the harvest pump is set to a constant value and consequently the pump profile shows a constant pattern over time. The bleed pump is activated when the online monitored viable cell density exceeds the VCD set point. As a result, the pump is switched on and off to a certain pump flow rate. The gravimetric feedback control loop of the feed pump results in an oscillating pump profile of the feed pump. The reactor weight also fluctuates around the set point due to the implemented control structure. 7. Critical quality attributes can be monitored by various techniques. Exclusion chromatography is typically applied for aggregates; weak cation exchange analytics or microfluidic devices are available for fragments and charge isoforms; a standard method is capillary gel electrophoresis with laser-induced fluorescence detection (CGE-LIF) for N-linked glycosylation.

Acknowledgments This work has been supported by the KTI (CTI) Program of the Swiss Economic Ministry (Project 19190.2 PFIW-IW). The authors declare that they have no conflicts of interest pertaining to the contents of this article.

Development of Mammalian Cell Perfusion Cultures at Lab Scale

139

References 1. Lute S, Agarabi C, Johnson S, Chavez B, Brorson K (2017) FDA/OBP laboratory research to support continuous bioprocessing.In: Integrated Continuous Biomanufacturing III. Suzanne Farid, University College London, United Kingdom Chetan Goudar, Amgen, USA Paula Alves, IBET, Portugal Veena Warikoo, Axcella Health, Inc., USA Eds, ECI Symposium Series 2. Zydney AL (2015) Perspectives on integrated continuous bioprocessing – opportunities and challenges. Curr Opin Chem Eng 10:8–13. https://doi.org/10.1016/j.coche.2015.07. 005 3. Konstantinov KB, Cooney CL (2015) White paper on continuous bioprocessing. May 20–21, 2014 continuous manufacturing symposium. J Pharm Sci 104:813–820. https:// doi.org/10.1002/jps.24268 4. Croughan MS, Konstantinov KB, Cooney C (2015) The future of industrial bioprocessing: batch or continuous? Biotechnol Bioeng 112:648–651. https://doi.org/10.1002/bit. 25529 5. Subramanian G (2018) Continuous biomanufacturing: innovative technologies and methods. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 6. Farid SS, Pollock J, Ho SV (2014) Evaluating the economic and operational feasibility of continuous processes for monoclonal antibodies. In: Continuous processing in pharmaceutical manufacturing. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, pp 433–456 7. Steinebach F, Ulmer N, Wolf M, Decker L, Schneider V, W€alchli R, Karst D, Souquet J, Morbidelli M (2017) Design and operation of a continuous integrated monoclonal antibody production process. Biotechnol Prog 33:1303–1313. https://doi.org/10.1002/ btpr.2522 8. Sawyer D, Sanderson K, Lu R, Daszkowski T, Clark E, Mcduff P, Astrom J, Heffernan C, Duffy L, Poole S, Ryll T, Sheehy P, Strachan D, Souquet J, Beattie D, Pollard D, Stauch O, Bezy P, Sauer T, Boettcher L, Simpson C, Dakin J, Pitt S, Boyle A (2017) Biomanufacturing technology roadmap – overview 9. Walther J, Godawat R, Hwang C, Abe Y, Sinclair A, Konstantinov K (2015) The business impact of an integrated continuous biomanufacturing platform for recombinant protein production. J Biotechnol 213:3–12. https://doi.org/10.1016/j.jbiotec.2015.05. 010

10. Xenopoulos A (2015) A new, integrated, continuous purification process template for monoclonal antibodies: process modeling and cost of goods studies. J Biotechnol 213:42–53. https://doi.org/10.1016/j.jbiotec.2015.04. 020 11. Xu S, Gavin J, Jiang R, Chen H (2017) Bioreactor productivity and media cost comparison for different intensified cell culture processes. Biotechnol Prog 33:867–878. https://doi. org/10.1002/btpr.2415 12. Reay D, Ramshaw C, Harvey A (2013) Process intensification: engineering for efficiency, sustainability and flexibility. ButterworthHeinemann, Oxford 13. Moulijn JA, Stankiewicz A, Grievink J, Go´rak A (2008) Process intensification and process systems engineering: a friendly symbiosis. Comput Chem Eng 32:3–11. https://doi.org/10. 1016/j.compchemeng.2007.05.014 14. Bo¨deker BGD, Newcomb R, Yuan P, Braufman A, Kelsey W (1994) Production of recombinant factor VIII from perfusion cultures: i. Large-scale fermentation. In: Spier RE, Griffiths JB, Berthold W (eds) Animal cell technology. Butterworth-Heinemann, Oxford, pp 580–583 15. Langer ES (2011) Trends in perfusion bioreactors. Bioprocess Int 9:10 16. Bonham-Carter J, Shevitz J et al (2011) A brief history of perfusion biomanufacturing. BioProcess Int 9:24–30 17. Barrett S, Chang A, Bandow N (2017) Intensification of a multi-product perfusion platform through medium and process development. In: Farid S, Goudar C, Alves P, Warikoo P (eds) Integrated continuous biomanufacturing III. Engineering Conferences International, Cascais, Portugal 18. Descheˆnes J-S, Desbiens A, Perrier M, Kamen A (2006) Use of cell bleed in a high cell density perfusion culture and multivariable control of biomass and metabolite concentrations. AsiaPacific. J Chem Eng 1:82–91. https://doi. org/10.1002/apj.10 19. Lin H, Leighty RW, Godfrey S, Wang SB (2017) Principles and approach to developing mammalian cell culture media for high cell density perfusion process leveraging established fed-batch media. Biotechnol Prog 33:891–901. https://doi.org/10.1002/btpr. 2472 20. Wolf MKF, Closet A, Bzowska M, Bielser J-M, Souquet J, Broly H, Morbidelli M (2018) Improved performance in mammalian cell

140

Moritz Wolf and Massimo Morbidelli

perfusion cultures by growth inhibition. Biotechnol J 14(2):e1700722. https://doi.org/ 10.1002/biot.201700722 21. Konstantinov K, Goudar C, Ng M, Meneses R, Thrift J, Chuppa S, Matanguihan C, Michaels J, Naveh D (2006) The “push-tolow” approach for optimization of high-density perfusion cultures of animal cells. Adv Biochem Eng Biotechnol 101:75–98. https://doi.org/ 10.1007/10_016 22. Bausch M, Schultheiss C, Sieck JB (2018) Recommendations for comparison of productivity between fed-batch and perfusion processes. Biotechnol J 14(2):e1700721. https:// doi.org/10.1002/biot.201700721 23. Dowd JE, Jubb A, Kwok KE, Piret JM (2003) Optimization and control of perfusion cultures using a viable cell probe and cell specific perfusion rates. Cytotechnology 42:35–45. https:// doi.org/10.1023/A:1026192228471 24. Xu S, Chen H (2016) High-density mammalian cell cultures in stirred-tank bioreactor without external pH control. J Biotechnol 231:149–159. https://doi.org/10.1016/j. jbiotec.2016.06.019 25. Karst DJ, Serra E, Villiger TK, Soos M, Morbidelli M (2016) Characterization and comparison of ATF and TFF in stirred bioreactors for continuous mammalian cell culture processes. Biochem Eng J 110:17–26. https://doi.org/ 10.1016/j.bej.2016.02.003 26. Goletz S, Stahn R, Kreye S (2016) Small scale cultivation method for suspension cells 27. Gomez N, Ambhaikar M, Zhang L, Huang C-J, Barkhordarian H, Lull J, Gutierrez C (2017) Analysis of Tubespins as a suitable scale-down model of bioreactors for high cell density CHO cell culture. Biotechnol Prog 33:490–499. https://doi.org/10.1002/btpr. 2418 28. Villiger-Oberbek A, Yang Y, Zhou W, Yang J (2015) Development and application of a highthroughput platform for perfusion-based cell culture processes. J Biotechnol 212:21–29. https://doi.org/10.1016/j.jbiotec.2015.06. 428 29. Wolf MKF, Lorenz V, Karst DJ, Souquet J, Broly H, Morbidelli M (2018) Development of a shake tube-based scale-down model for perfusion cultures. Biotechnol Bioeng 115:2703–2713. https://doi.org/10.1002/ bit.26804 30. Sieck JB, Schild C, von Hagen J (2017) Perfusion formats and their specific medium requirements. In: Continuous biomanufacturing – innovative technologies and methods, pp

171–200. https://doi.org/10.1002/ 9783527699902.ch7 31. Chotteau V (2017) Process development in screening scale bioreactors and perspectives for very high cell density perfusion. In: Integrated Continuous Biomanufacturing III. Suzanne Farid, University College London, United Kingdom Chetan Goudar, Amgen, USA Paula Alves, IBET, Portugal Veena Warikoo, Axcella Health, Inc., USA Eds, ECI Symposium Series 32. Zoro B, Tait A, Carpio M, McHugh K (2018) Development of a novel automated perfusion mini-bioreactor ambr® 250 perfusion. In: A. Robinson, PhD, Tulane University R. Venkat, PhD, MedImmune E. Schaefer, ScD, J&J Janssen,editors.Cell Culture Engineering XVI, ECI Symposium Series, 2018 33. Villiger TK, Neunstoecklin B, Karst DJ, Lucas E, Stettler M, Broly H, Morbidelli M, Soos M (2018) Experimental and CFD physical characterization of animal cell bioreactors: from micro- to production scale. Biochem Eng J 131:84–94. https://doi.org/10.1016/j.bej. 2017.12.004 34. Garcia-Ochoa F, Gomez E (2009) Bioreactor scale-up and oxygen transfer rate in microbial processes: an overview. Biotechnol Adv 27:153–176. https://doi.org/10.1016/j.bio techadv.2008.10.006 35. Russell TWF, Robinson AS, Wagner NJ (2008) Mass and heat transfer: analysis of mass contactors and heat exchangers. Cambridge University Press, New York 36. Vogg S, Wolf MKF, Morbidelli M (2018) Continuous and integrated expression and purification of recombinant antibodies. In: Hacker DL (ed) Methods in molecular biology. Springer, New York, NY, pp 147–178 37. Clincke M, Mo¨lleryd C, Zhang Y, Lindskog E, Walsh K, Chotteau V (2013) Very high density of CHO cells in perfusion by ATF or TFF in WAVE bioreactorTM. Part I. Effect of the cell density on the process. Biotechnol Prog 29:754–767. https://doi.org/10.1002/btpr. 1704 38. Fan Y, Ley D, Andersen MR (2018) Fed-Batch CHO Cell Culture for Lab-Scale Antibody Production. Methods Mol Biol 1674:147–161. https://doi.org/10.1007/ 978-1-4939-7312-5 39. Zhu L, Song B, Wang Z, Monteil DT, Shen X, Hacker DL, De Jesus M, Wurm FM (2017) Studies on fluid dynamics of the flow field and gas transfer in orbitally shaken tubes. Biotechnol Prog 33:192–200. https://doi.org/10. 1002/btpr.2375

Chapter 9 Perfusion Control for High Cell Density Cultivation and Viral Vaccine Production Alexander Nikolay, Thomas Bissinger, Gwendal Gr€anicher, Yixiao Wu, Yvonne Genzel, and Udo Reichl Abstract The global demand for complex biopharmaceuticals like recombinant proteins, vaccines, or viral vectors is steadily rising. To further improve process productivity and to reduce production costs, process intensification can contribute significantly. The design and optimization of perfusion processes toward very high cell densities require careful selection of strategies for optimal perfusion rate control. In this chapter, various options are discussed to guarantee high cell-specific virus yields and to achieve virus concentrations up to 1010 virions/mL. This includes reactor volume exchange regimes and perfusion rate control based on process variables such as cell concentration and metabolite or by-product concentration. Strategies to achieve high cell densities by perfusion rate control and their experimental implementation are described in detail for pseudo-perfusion or small-scale perfusion bioreactor systems. Suspension cell lines such as MDCK, BHK-21, EB66®, and AGE1.CR.pIX® are used to exemplify production of influenza, yellow fever, Zika, and modified vaccinia Ankara virus. Key words Viral vaccine, High cell density cultivation, Perfusion bioreactor, Pseudo-perfusion, Perfusion rate, Control strategy, On-line sensors, Alternating tangential flow filtration

1

Introduction To prevent the spread of infectious diseases, vaccination of both humans and animals (pets, livestock, wildlife) remains the most effective solution. Administered viral vaccines contain antigenic structures (such as live-attenuated virus, inactivated virus, or their subunit virus components) and stimulate the adaptive immune system to elicit an effective immune response against specific pathogens [1]. This turns vaccine development and supply to be crucial for worldwide disease control and eradication [2]. However, growing vaccine markets, recurrent pandemics, and emerging diseases pose a major challenge for current vaccine manufacturing processes

Alexander Nikolay and Thomas Bissinger contributed equally to this work. Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_9, © Springer Science+Business Media, LLC, part of Springer Nature 2020

141

142

Alexander Nikolay et al.

toward the establishment of new strategies for process intensification. A typical process for cell culture-derived virus production comprises two phases in which cells are first propagated in a cultivation vessel and then infected with the virus. The virus enters the host cell, uses the cell machinery, and replicates to—ideally—high virus titers. Virus replication and release is usually associated with cell lysis, so that the virus material is harvested batch-wise at its peak concentration. In order to intensify a virus production process with suspension cells, both phases can be optimized: the host cells can be cultivated to higher cell concentrations before infection and process parameters at time point of infection can be changed for improved virus replication [3]. 1.1 Perfusion Rate Control Strategies for the Cell Growth Phase

To obtain high cell densities, perfusion bioreactor systems are required to replace spent medium with fresh medium while suspension cells are retained typically by membranes, sedimentation, or centrifugation [4]. Over the last decades, different options for perfusion rate control have been established to supply cells continuously with sufficient amounts of substrates and to remove waste products. This includes manual and automated control of shake flasks (pseudo-perfusion) or stirred tank bioreactors (perfusion), respectively: 1. Pseudo-perfusion For high-throughput screening as well as initial processes development and optimization, shake flask cultivations are a convenient tool. Using shake flasks or spin tubes, total medium usage is reduced, while cell concentrations similar to bioreactor perfusion systems can be achieved (see Note 1). This allows more comprehensive studies at higher cell concentrations than typically obtained in conventional batch cultivations. Therefore, cells are centrifuged and the cell-free supernatant is partially or fully exchanged. Due to the lack of on-line probes and the limited working volume, media exchanges usually follow a regime that considers pre-defined growth and metabolite uptake rates. Either time intervals decrease or exchange volume fractions increase to maintain sufficient substrate levels. This batch-wise medium exchange results in typical profiles of (over-)feeding with abrupt cell-specific perfusion rate (CSPR) increases (Fig. 1a). To prevent strong gradients, the perfusion medium can be pre-heated and added to a remaining fraction of old medium. The idea is to push the cell to its maximum growth rather than optimization of medium utilization. 2. Reactor volume exchange regimes The reactor volume (RV) exchange rate per day (RV/day) describes the perfusion flow rate in terms of the total working volume. Perfusion rates are often set to a fixed flow rate or

Perfusion Control Strategies

143

Fig. 1 Perfusion rate control strategies in shake flasks and bioreactors with their respective impact on substrate concentration (red line), reactor volume exchange rate (RV/day; blue dotted line), and cell-specific perfusion rate (CSPR; green dashed line). (a) Pseudo-perfusion cultivation with an RV/daybased perfusion rate control by batch-wise medium exchange, which results in temporary overfeeding and CSPR peaks. Media exchange time points are typically defined by time intervals or metabolite levels. (b) In contrast to shake flasks, perfusion bioreactors enable a constant medium supply following an RV/ day-based regime. Perfusion rates increase step-wise at fixed time points to meet nutrient demands. This results in fluctuating metabolite concentrations and CSPRs until maximum cell concentrations are reached due to medium limitations. (c) The CSPR-based control strategy starts with inoculation and maintains a constant CSPR. The RV/day increases according to the demands of the growing cells. On-line capacitance probes enable process automation. As metabolite uptake rates typically decrease at lower metabolite levels, metabolite concentrations may vary during the cultivation, i.e., a later cultivation time. (d) Substrate-based perfusion control strategies start when a certain set point is reached and maintain the metabolite level. The RV/day does not necessarily follow the cell concentration, and due to decreasing metabolite uptake rates, CSPR may also vary. On-line measurement devices enable automated closedloop process control

increased step-wise with progressing cultivation time, but not necessarily coupled to the actual cell concentration. RV/day’s are typically chosen based on previous data on cell growth and metabolite uptake rates. As pre-defined perfusion rates do not take into account the actual state of the biological system, the regime is not robust enough to cover for biological variations (e.g., cell growth, uptake rates). In consequence, perfusion rates may not be optimally selected for cell growth resulting in under- or oversupply of substrates. In addition, metabolic by-products may accumulate to inhibiting concentrations,

144

Alexander Nikolay et al.

affecting cell growth, productivity, or product quality. Thus, metabolite concentrations and CSPRs (pL/cell/day) may vary during the cultivation run (Fig. 1b). Based on physiological changes (e.g., specific cell growth rate) and process alterations (e.g., timing, working volume), batch-to-batch variations may occur. If such changes are only moderate, especially in a wellcharacterized and robust production system, the RV/day regime is a successful strategy. It is easily applicable and requires minimal technical effort. The cells typically grow to certain cell concentrations and maintain steady-state conditions in dependence of the substrate supply. In combination with a cell bleed, this is particularly advantageous for the continuous expression of recombinant proteins [5]. 3. Cell-specific perfusion rate control The perfusion rate can be also determined by measuring the viable cell concentration (VCC), which results in a constant CSPR and metabolic environment for the cell [6]. Based on manual counting or the use of automated cell counters (Vi-Cell), perfusion rates can be adjusted manually, depending on the growth performance. To improve the CPSR control between sample points, the growth rate can be used to predict perfusion rates until the next sampling time. The perfusion pump rate can be programmed to correlate to the perfusion rate with the expected cell growth. Automation is possible by the use of on-line biomass probes measuring the cell volume or cell concentration in the bioreactor vessel and forwarding the signal to a perfusion pump [7]. This open-loop control system automatizes the perfusion process and runs with high precision based on the actual growth performance (Fig. 1c). This contributes to a higher reproducibility in case on-line biomass measurements are reliable. Nevertheless, a CSPR-Based control strategy enables an adequate supply of medium at different cell concentrations and therefore contributes to better defined cultivation conditions. Thus, cellular behavior and productivity can be studied, e.g., to investigate a decrease in cell-specific virus yield at higher cell concentrations [8]. 4. Metabolite-based perfusion rate control Changes in key substrate concentrations such as glucose or glutamine are good indicators for the overall substrate consumption. Off-line metabolite measurements (e.g., BioProfile) are typically performed with regular bioreactor sampling, and perfusion rates can be adjusted to maintain (at least near) steady-state substrate concentrations. For better manual perfusion rate control, the last specific cell growth and substrate uptake rates can be considered to predict the required perfusion rate until the next sampling time point. Closed-loop control systems can be established with on-line metabolite

Perfusion Control Strategies

145

measurement devices (e.g., BioPAT Trace), where the output signal is used to control a perfusion pump. Substrate-based perfusion rate control requires good knowledge on the nutritional demand of the cells and the identification of key metabolites to avoid over- or under-feeding (Fig. 1d). Besides substrates, measurements of by-product levels can be also used for perfusion control. A typical by-product is lactate. Its formation is promoted by an excessive glucose uptake. It decreases the pH value of the medium, alters glycosylation structures of products, and potentially inhibits cell growth. A low lactate level is typically desired and its concentration can be estimated from on-line pH measurements. At the beginning of a cultivation, lactate concentrations in the medium increase and pH values decrease. As glucose levels further decrease, many cells start to consume lactate and pH values increase again. With a well-chosen pH set point, perfusion with fresh medium can be started to supply the cells with glucose to maintain optimal growth conditions [9]. If cells do not consume extracellular lactate, fresh medium dilutes the constantly formed lactate, and constant pH levels correspond to constant lactate concentrations. This strategy is rather challenging and assumes a good correlation of metabolite consumption rates and formation of lactate to avoid controller failure. In addition, overall medium consumption has to be taken into account. Overall, this control principle is easily adaptable to every state-of-the-art bioreactor system with pH control. 1.2 Perfusion Rate Control Strategies for the Virus Production Phase

In the second phase of vaccine production, virus replication is initiated by the addition of a virus seed, when a desired cell concentration is reached. Various options are given to improve virus yields. With respect to perfusion control strategies, cells may be first concentrated and perfusion rates reduced or paused to facilitate virus adsorption to the cell. The perfusion rate could be adjusted during the initial virus production phase to avoid virus wash-out and to favor virus replication. However, the selection of the control strategy may also depend on specific details of virus-host cell interaction, as cellular metabolite demands, uptake rates, or by-product formation rates can change after infection. Thus, hybrid perfusion rate strategies may be required, e.g., switching to a constant RV/day value based on the last perfusion rate before infection [8]. Nevertheless, the infection phase has to be investigated in detail and optimized for each virus-host cell system. Other process parameters, such as temperature or pH shift, volume expansion, batch-wise medium exchange, and different virus harvest strategies, can be additionally applied, but are not further discussed here (see reviews such as [3, 10]).

146

Alexander Nikolay et al.

Table 1 Vi-Cell settings for different cell lines Parameter

BHK-21

EB66®

AGE1.CR

MDCK

Assay cells

Min. diameter (μm)

10

8

8

8

5

Max. diameter (μm)

35

30

30

25

28

Number of images

100

100

100

100

50

Aspirate cycle

3

3

3

3

3

Trypan blue mixing cycle

3

2

3

3

3

Cell brightness (%)

85

90

90

85

85

Cell sharpness

80

100

100

80

80

Viable cell spot brightness (%)

90

85

85

90

90

Viable cell spot area (%)

4

3

3

4

4

Min. circularity

0

0.5

0.5

0

0

Decluster degree

Medium

High

High

Medium

Medium

Comment: Assay cells encompass Vero, PS, and adherent MDCK cells

The cell-specific virus yield (CSVY; in virions/cell) can be considered a benchmark for the production system and should be maintained or increased to optimize productivity of the perfusion systems. The CSVY is defined as the amount of virus produced per cell and is expressed as: CSVY ¼

N vir, max  N vir,init N cell, max

ð1Þ

where Nvir, max is the maximum number of virus particles (virions), Nvir, init is the number of virus particles used for infection (virions), and Ncell, max is the maximum number of cells (cells) (see Notes 2 and 3).

2 2.1

Materials Cells and Media

1. BHK-21 suspension cells (baby hamster kidney cell, from IDT Biologika, Dessau, Germany), maintained in serum-free medium. 2. AGE1.CR.pIX® suspension cells (avian designer cell line, ProBioGen, Berlin, Germany), cultivated in CD-U3 medium (ProBioGen AG) supplemented with 2 mM L-glutamine (stock solution: 200 mM, sterile filtered (0.22 μm), stable at 4  C for 3 months), 2 mM L-alanine, and recombinant insulinlike growth factor (LONG-R3 IGF, 10 ng/mL final concentration, Sigma) [11].

Perfusion Control Strategies

147

3. EB66® suspension cells (duck embryo-derived, Valneva, Lyon, France) [12], cultivated in CDM4Avian medium (GE Healthcare) supplemented with 2.5 mM L-glutamine. 4. MDCK (Madin-Darby canine kidney) suspension cells derived from adherent MDCK cells (ATCC, CCL-34) [13], cultivated in Xeno-CDM (Shanghai BioEngine Sci-Tech, Shanghai, China) supplemented with 8 mM L-glutamine. 5. Assay cell lines such as adherent MDCK cells (ECACC, #84121903), Vero cells (WHO seed ECACC 134th), and porcine kidney stable epithelial (PS) cells (thankfully provided by M. Niedrig, Robert Koch Institute, Berlin, Germany) cultivated in GMEM (Gibco) supplemented with 5.5 g/L glucose, 10% fetal calf serum, and 2 g/L peptone (hereafter referred to as Z-medium). For infection with influenza virus, adherent MDCK cells are cultivated in serum free Z-medium, supplemented with trypsin (5 U/mL) (hereafter referred to as V-medium). 2.2

Viruses

1. Human influenza A virus (IAV) strain A/PR/8/34 of the subtype H1N1 (Robert Koch Institute, Amp. 3138) propagated in adherent MDCK cells (ECACC). Seed virus (1.8  109 virions/mL based on TCID50 assay) was adapted in five passages at an MOI of 105 and porcine trypsin (Gibco) to MDCK suspension cells. 2. Live-attenuated yellow fever virus (YFV) 17D-204, produced in specific pathogen-free eggs (kindly provided by M. Niedrig, Robert Koch Institute, Berlin, Germany) and expanded in adherent Vero cells with a titer of 8  107 infectious virions/ mL based on plaque assay. 3. Zika virus (ZIKV) isolated from blood specimens of PCR-positive patient in Rio de Janeiro, Brazil, expanded in C6/36 insect cells (thankfully provided by T. S. Moreno, Fiocruz, Rio de Janeiro, Brazil) and Vero cells. Serial blind passages for five iterations in EB66® cells with a titer of 3  106 infectious virions/mL based on plaque assay. 4. Recombinant modified vaccinia Ankara (MVA) virus isolate MVA-CR19.GFP (provided by ProBioGen, Berlin, Germany) containing green fluorescent protein insertion cassette and adapted to suspension cells with a titer of 4.1  108virions/ mL based on TCID50 assay.

2.3 Virus Quantification

1. Hemagglutination assay: Microplate reader Infinite® M200 (Tecan) and round-bottom 96-well plates (lid for active virus samples). 2. TCID50 assays: IAV staining with primary antibodies antiinfluenza A/PR/8/34 H1N1 HA serum (#03/242, NIBSC)

148

Alexander Nikolay et al.

and secondary antibody Alexa Fluor donkey anti-sheep IgG (#A11015, Thermo Fisher Scientific); gentamicin (Thermo Fisher Scientific) and ultrasonic bath (USC600D, VWR). 3. Plaque assay: naphthalene black solution (1 g naphthol blue black, 13.6 g sodium acetate, 60 mL glacial acetic acid; add to 1 L ddH2O). 2.4

Cell Cultivation

1. Shake flask and pseudo-perfusion cultivations: BHK-21 and AGE1.CR.pIX® cells cultivated in 125/250 mL baffled polycarbonate Erlenmeyer flasks (Corning). EB66® and MDCK cells in 125/250 mL non-baffled polycarbonate Erlenmeyer flasks (Corning). 2. Working volumes: 30–50 mL (125 mL shaker) and 60–100 mL (250 mL shaker). 3. Cultivation in a Multitron Pro incubator, with 50 mm shaking throw (Infors HT). 4. Cell counting: Vi-Cell XR (Beckman Coulter) with cell linespecific settings as single measurement (see Note 5). EB66® agglomerates treated with porcine trypsin (Gibco) and stopped with fetal calf serum (FCS, Gibco).

2.5 Perfusion Bioreactor System

1. Doubled-jacket UniVessel1 L glass bioreactor with digital control unit (DCU) BioStat B2 plus controller (both Sartorius Stedim) with PT100 thermometer, pO2 probe (InPro 6100/ 150/S/N, Mettler Toledo), and pH probe (405-DPAS-SCK8S/150, Mettler Toledo) control. Gas aeration through a micro-sparger (Sartorius, pore diameter ~20 μm). 2. Dead-volume free sampling port assembled by a T-piece connector, a one-way sampling valve (Eppendorf), an air filter (0.2 μm, 25 mm), and a short silicon tube connected to the stainless steel dip tube of the bioreactor. 3. Perfusion cell culture performed with an alternating tangential flow filtration unit (ATF2 system with C24U–V2.0 controller, Refine Technology) using 0.2 μm pore size polyethersulfone membranes (Spectrum, #S02-P20U-10-N) or 500 kDa pore size polysulfone membranes (GE Healthcare, #UFP-500-E4X2MA). Feed and permeate flow rate controlled through peristaltic pumps (120 U, Watson-Marlow) and peristaltic pump tubes (PharMed BPT NSF-51, ID 0.76 mm). 4. Gravimetric balance to determine working volume of the perfusion system (Midrics 1, Sartorius Stedim Biotech).

2.6 Perfusion Control Equipment

1. CSPR-based control: Multi-frequency capacitance probe with pre-amplifier (Incyte Hamilton) connected with a M12 cable to the controller (ArcView controller 265, Hamilton). A 4–20 mA output box (Hamilton) with an open-end AUX

Perfusion Control Strategies

149

M12 to the output port of the ArcView controller and an openend cable and 15-pin D-SUB male connector to the WatsonMarlow pump. 2. Metabolite-based control: BioProfile 100 Plus (Nova Biomedical) to measure glucose, lactate, glutamine, ammonium, glutamate. BioPAT Trace (Sartorius Stedim) on-line-monitoring tool for glucose and lactate with volume-free sampling dialysis sampling probe (12 mm), transport buffer solution (#BPT0006) glucose and lactate calibration solutions (#BPT0007-11, both Sartorius Stedim).

3

Methods

3.1 Cell Culture Maintenance

Suspension cells are grown routinely in shake flasks. Depending on the population doubling time, BHK-21 and AGE1.CR.pIX® cells are passaged every 3 to 4 days and inoculated at 5  105 cells/mL or 8  105 cells/mL, respectively. MDCK and EB66® cells are passaged every 2 to 3 days and inoculated at 5  105 cells/mL or 3  105 cells/mL, respectively. All cells are maintained in an orbital shaking incubator at 37  C and 5% CO2 with the exception of EB66® cells, which are incubated at 7.5% CO2. Shaking speeds vary from 100 rpm for MDCK and 150 rpm for EB66® cells to 185 rpm for BHK and AGE1.CR.pIX® cells.

3.2 Cell Counting and Metabolite Measurements

Concentration and viability of individual suspension cells can be measured directly. However, EB66® cells show a tendency to form cell clumps in bioreactors. Thus, add 200 μL of the cell broth to 200 μL trypsin (5000 U/mL) and incubate at 600 rpm for 10 min and 37  C. The reaction is stopped with 200 μL FCS before cell counting (see Note 4). Final samples are measured over 100 images at about 1  104 to 1  107 cells/mL (dilute with PBS or FCS if necessary) using a Vi-Cell. Linear measurement range and cell-specific parameters for data acquisition need to be determined before (see Note 5). Metabolite samples are taken from the culture and centrifuged for 1 min at 700  g; virus-containing samples are inactivated additionally for 3 min at 80  C (see Note 6) before optional storage at 80  C. All measurements are performed in the validated working range of the BioProfile 100 Plus.

3.3 Virus Titration Assays

The total number of IAV particles is quantified with the hemagglutination (HA) assay. In brief, fill round-bottom 96-well plates with 100 μL PBS except for column 1. Add 100 and 70.7 μL of cell-free virus sample alternatingly to each row of the plate in the first and second column, and fill up every second row to 100 μL with 29.3 μL PBS. Serial dilute the sample with PBS by pipetting

3.3.1 Hemagglutination Assay (IAV)

150

Alexander Nikolay et al.

100 μL from one column to the next one. Use an internal standard for every second 96-well plate and treat it similar to the sample. Add 100 μL of chicken erythrocyte solution (adjusted to 2  107 erythrocytes/mL) to each well, and incubate plates for 3–8 h at room temperature (RT). Hemagglutination is evaluated with a plate reader at 700 nm extinction. Based on a curve fitting function, the titer was calculated in reference to an external standard. The virus titer is expressed as the common logarithm (log10) of the hemagglutination unit (HAU) per analysis volume (100 μL): log10(HAU/100 μL). 3.3.2 Tissue Culture Infection Dose (IAV)

The 50% tissue culture infection dose (TCID50) assay for quantification of the infectious IAV is performed as follows: Prepare a tenfold serial dilution (101 to 108) of each virus sample in V-medium containing 1% gentamicin. Infect confluent MDCK cells in flat-bottom 96-well plates with 100 μL of diluted sample (eight replicates per dilution). After 24 h of incubation (37  C, 5% CO2) add 100 μL of V-medium and incubate cells again for 24 h. Before the staining remove the cultivation medium and fix the cells with 100 μL cold acetone (80% acetone; 30 min at 4  C). Stain cells with 50 μL of an HA-specific primary antibody (1:200 diluted) and 50 μL of a fluorescence-labeled secondary antibody (1:500 diluted). Incubate both antibodies for 1 h at 37  C, and wash cells with PBS twice between every step. Count fluorescencepositive (infected) and fluorescence-negative (non-infected) wells. Calculate infectious virus titers (virions/mL) from eight replicates according to the Spearman-K€arber method [14, 15]: log 10 ðvirions=100 μLÞ ¼ d þ 0:5 þ

n 8

ð2Þ

where d is log10 of the highest dilution with completely (eight) infected wells (e.g., log10(105)) and n is the cumulative number of infected wells in all dilutions which contain non-infected wells (dilutions > d). 3.3.3 Tissue Culture Infection Dose (MVA)

The TCID50 titer for MVA is performed with Vero cells growing in DMEM supplemented with 10% (v/v) FCS and 0.1% (v/v) gentamicin. Seed cells at 5  105 cells/mL in flat-bottom 96-well plates and 100 μL/well. After standard incubation for 24 h, infect confluent cells with 100 μL of a tenfold serial dilution of virus sample in DMEM with 10% FCS and 1% gentamicin. Therefore, treat the virus sample for 1 min in a sonication water bath at 45 kHz. After 48 h post-infection, count fluorescent (infected) Vero cells and calculate the infectious virus titer (virions/mL) according to the Spearman-K€arber method as described above.

Perfusion Control Strategies 3.3.4 Plaque Assay (YFV, ZIKV)

3.4 Perfusion Bioreactor Assembly and Operation

151

Infectious virus titers of YFV and ZIKV in the cell culture supernatant are determined by a plaque assay. Therefore, add 200 μL of tenfold serial diluted virus sample and 200 μL porcine kidney (PS) cells (2  106 cells/mL) simultaneously to 24-well plates and incubate for 4 h at 37  C. Add 1.6% (w/v in Z-medium) carboxylmethyl cellulose as overlay to the cell/virus mixture and incubate for 3 to 4 days. Fix cells with 3.7% (v/v in PBS) formalin for 15 min and stain with naphthalene for 30 min. Count plaques from duplicates in respective dilutions; titers are expressed as plaque-forming units per volume (PFU/mL). 1. Equip the perfusion bioreactor (700 mL) with a single microsparger unit (pore size ~20 μm), a 6 mm dip tube for sampling, a plastic septum for inoculation and infection, stainless steel sterile connectors for bioreactor filling, perfusion medium addition, and optional pH adjustment. Connect an ATF2 system with a 12 mm dip tube to the vessel. Permeate is extracted from the upper half of the vertically positioned hollow fiber membrane (Fig. 2). Attach air filters (0.2 μm, 50 mm) to the sparger tube and to the air exhaust outlet. 2. Calibrate the pH probe (stored in 3 M KCl solution) at pH 7 and pH 9.21 at RT, and check function of the dissolved oxygen electrode according to the manufacturer’s recommendation.

Fig. 2 Schematic illustration of the perfusion bioreactor assembly. A balance (balance 1) is placed under the bioreactor/ATF system to maintain a constant working volume by replacing removed permeate volume with feed medium. Cell-free permeate flow is controlled either manually or via on-line measurements (pH, substrate, biomass). Balance 2 is used to confirm permeate flow rates

152

Alexander Nikolay et al.

3. If required, condition the hollow fiber membrane by flushing with non-sterile PBS for 1 h. In the meantime, calibrate external peristaltic pumps of the feed and harvest line. 4. For 30 min of autoclaving at 121  C, cover sensitive connectors and filters with aluminum foil (see Note 7). After cooling, fill flasks for pH adjustment under sterile conditions with 0.5 M NaOH and 1 M HCl, respectively; if required select a pH set point of 7.2 (see Note 8). 5. Position the bioreactor and the coupled ATF unit on a balance. Depending on hollow fiber length, it may be required to place the bioreactor systems on a stable pedestal (see Note 9). 6. Fill sterile growth medium (between 0.6 and 1 L) into the bioreactor and heat up to 37  C. The vessel typically has a pitched blade impeller. Start stirring in accordance to the requirements for the cell line (about 80–180 rpm). 7. Calibrate the pO2 probe after polarization at 100% air saturation and at 0% air (either by degassing or cable disconnection). Choose a set point between 40% and 80% which is maintained by pulse-wise oxygen sparging. 8. Inoculate seed cells through a syringe and a wide 19G needle. 9. Typical cultivation parameters are monitored on-line and recorded with the Sartorius digital control unit (DCU). Check the performance of the pH probe with externally measured pH values and adjust if required (avoid CO2 degassing of the sample). 10. When the perfusion process is started, initialize to weight control of the bioreactor in accordance to the perfusion strategy. Turn the ATF unit on and set flow rates to 0.8–0.9 L/min (see Notes 10 and 11). 11. In order to confirm perfusion rates, the weight of the permeate bottle can be monitored with a second balance. Based on the weight increase over time, calculate the effective permeate flow rate and re-calibrate the permeate harvest pump if necessary (Fig. 2). 12. At the end of the virus production phase, harvest the bioreactor (and in some cases permeate) and process the virus material for purification. Drain the bioreactor vessel completely and fill it with PBS. Similar to the sterilization process, autoclave the system for 30 min at 121  C to inactivate remaining virus. Then, wash the bioreactor system and ATF unit and, if required, re-condition the ATF membrane (see Note 12). 3.5 CSPR-Based Control Strategy

The CSPR is defined as perfused medium volume rate (Dperf) per viable cell and is expressed in pL per cell per day. In general, CSPRs can vary strongly between cell lines and applications and are

Perfusion Control Strategies

153

typically chosen in the range of 50–500 pL/cell/day [16]. Lower values are usually achieved with better cell growth media. The CSPR is normally set constant for the cell growth phase, but may need to be changed for optimum virus replication. 3.6 Pseudoperfusion

Pseudo-perfusion, as a scale-down model for high cell density cultivation, mimics “real” (scalable) perfusion systems. To maintain metabolite levels for extended cell growth (or virus propagation), media exchanges are performed manually (see Note 13). Pseudoperfusions cannot only be established following a rigid RV/daybased exchange regime but also based on cell concentrations and resulting CSPRs or substrate and by-product concentrations. Accordingly, this small-scale screening tool is very versatile (see Note 14) and suitable for many cell lines (see Note 1). The following example describes a two-step CSPR-based pseudo-perfusion control to achieve high cell concentrations for virus production (Fig. 3). First, partial media exchange volumes (VE in mL) are calculated according to a fixed time schedule (Eq. 3). When 60% of the total medium needs to be exchanged, the time intervals (Δti in h) are shortened (Eq. 4): VE ¼

 X μ∙Δt  1 ∙V W  CSPR ∙ e μ  ln X 0:6∙μ ∙CSPR þ 1 Δt ¼ μ

ð3Þ ð4Þ

where X is the viable cell concentration (in cells/mL), μ the specific cell growth rate (in 1/h), Δt the time interval to the previous sampling (in h), and Vw the working volume (in mL).

Fig. 3 Small-scale pseudo-perfusion cultivation of MDCK suspension cells to produce influenza A virus in shake flasks. (a) Cell concentration (blue dots) and accumulated virus titer (red triangles). (b) Perfusion rate (in RV/day, blue line) increased step-wise with increasing cell concentrations. Cell-specific perfusion rates (green squares) varied around the process value of 60 pL/cell/day. Vertical dotted line indicates time point of infection

154

Alexander Nikolay et al.

3.6.1 Cell Growth Phase

1. Inoculate MDCK suspension cells at 1  106cells/mL in 125 mL baffled shake flasks with a working volume of 50 mL, and cultivate cells in an orbital shaking incubator at standard conditions. Monitor cell growth behavior. 2. After 2 days, cells reach the exponential growth phase and medium replacement can be initiated. First, measure cell concentrations with the Vi-Cell and calculate the specific growth rate. Depending on the viable cell concentration, VE is calculated according to Eq. 3 for a constant CSPR of 60 pL/cell/ day (see Note 15). 3. Transfer the whole cell culture to a 50 mL sterile Falcon tube and centrifuge at 300  g for 5 min at RT. Remove the calculated volume VE and refill with fresh pre-warmed medium (see Note 16). 4. Gently re-suspend the cell pellet and transfer cells to the old shake flask and continue incubation. Cell concentrations can be measured again. 5. Within 4 days, the perfusion exchange volume reached 60% for time intervals as short as 8 h (see Note 16). Accordingly, the strategy toward higher exchange frequencies is initiated following Eq. 4.

3.6.2 Virus Production Phase

1. For IAV infection, measure the cell concentration and remove as much supernatant as possible by centrifugation. Fill up with fresh medium supplemented with 30 units/mL trypsin. 2. Infect with the adapted IAV at multiplicity of infection (MOI) of 0.1 (see Note 17). 3. After 5 h, perform a total medium replacement by centrifugation. The perfusion rate control continues with the previous CSPR-based strategy and maximum perfusion volume of 60% with varying exchange intervals Δtn (Eq. 4). 4. Supplement feed medium for volume exchange with 30 units/ mL trypsin and collect harvest (see Note 13). 5. When virus-induced cell lysis begins, last μ is set as a constant value for perfusion rate control (see Note 20).

3.7 Bioreactor with Manual CSPR Control

Based on previous data on metabolite uptake rates, a CSPR of 60 pL/cell/day was selected for AGE1.CR.pIX® cultivations (see Note 18). After each sampling, the perfusion rate (Qperf in mL/h) is adjusted manually based on the viable cell concentration (Fig. 4). To take into account the further increase in cell concentration and to maintain the CSPRs, the expected time interval (Δt in h) until next sampling point and the specific cell growth rate (μ in1/h) (see Notes 19 and 20) is considered resulting in: Q perf ¼ X i  V w  CSPR  eμΔt

ð5Þ

Perfusion Control Strategies

155

Fig. 4 Bioreactor with manual CSPR control for AGE1.CR.pIX®cultivation. Increased medium exchange is coupled with volume expansion during the infection phase to improve MVA virus titers. (a) Cell concentration (blue dots), virus titer (red triangles), and relative working volume (green line). (b) Perfusion rate (in RV/day) based on initial reactor volume (blue line) and CSPR (green squares). Vertical dotted line indicates time point of infection

3.7.1 Cell Growth Phase

1. Inoculate the bioreactor with AGE1.CR.pIX®at 6 2.5  10 cells/mL from a shake flask pre-culture. 2. When the cell concentration reaches 7  106 cells/mL, start recirculation with the ATF diaphragm pump and set the perfusion rate corresponding to a CSPR of 60 pL/cell/day. 3. Take samples every 12 to 24 h, double-check perfusion rates, and re-adjust peristaltic permeate pump rates if required to maintain the CSPR.

3.7.2 Virus Production Phase

In this example, the perfusion rate strategy is switched to an RV/ day-based perfusion for virus propagation (Fig. 4). The medium is exchanged at the time of infection and the working volume expanded during infection. Both actions aim at improved virus titers (see Note 21) [8]. 1. Exchange bioreactor volume 3 h before inoculation by applying maximum perfusion rates of about 12 RV/day. The working volume remains constant at 580 mL. 2. Infect the bioreactor at 6  107 cells/mL with MVA-CR19. GFP at MOI 0.04. 3. 15 min post-infection, expand the working volume with fresh growth medium from 580 to 800 mL and increase the perfusion rate to 1.8 RV/day (based on initial working volume). 4. 24 h post-infection, increase the working volume from 800 to 1000 mL and set the perfusion rate to 2.6 RV/day. 5. Three days post-infection maximum MVA titers of 2.1  109 infectious virions/mL are achieved.

156

Alexander Nikolay et al.

Fig. 5 Automated cell concentration-based CSPR control to cultivate EB66® cells and produce Zika virus. (a) On-line permittivity signal (blue solid line), off-line measured viable cell concentration (blue circles), and virus titer (red triangles). On- and off-line cell concentrations correlated until cell death phase. (b) Perfusion rate (in RV/day, solid line) increased with cell concentrations, and CSPR (black triangle) remained constant at about 34 pL/cell/day [17]

3.8 Bioreactor with Automated CSPR Control

3.8.1 Setting Up the Control Unit

The CSPR control can be further improved by the use of an on-line capacitance probe, which forwards the biomass signal to a perfusion pump (Fig. 5). In particular, the probe signal of the Incyte sensor is processed to estimate the viable cell volume per mL or the viable cell concentration (assuming cell diameter changes during the cultivation run are negligible). The corresponding data was published in Nikolay et al. [17] (see Note 22). 1. The correlation of the on-line permittivity signal (pF/cm) to the cell concentration (106 cells/mL) is described by a so-called specific cell factor. To calculate this factor, it requires Incyte probe measurement data from previous cultivations. Cell concentrations determined off-line are plotted over the on-line permittivity signal and the slope of a linear regression line (first order) determined the respective factor. Insert this number as Cell Density unit/Factor in the channel Incyte and field Measure of the ArcView controller (see Note 23). 2. Connect the Incyte probe to the pre-amplifier and then with a M12 cable to the ArcView controller. The output signal from the controller is forwarded via an open-end AUX M12 cable to the +24 V, GND, RS-485 A and B ports of the 4–20 mA output box. Connect the output ports 1 and 2 (channel 1) of the output box with an open-end cable and 15-pin D-SUB male connector to the Watson-Marlow pump (see Note 24). 3. For the configuration of the controller output signal, open Advanced settings and select 4–20 mA Configuration. Select

Perfusion Control Strategies

157

the channel of the input signal (typically Incyte probe channel 1) and insert configuration values for the 4 and 20 mA current output. Configuration values have to be identified empirically and link the on-line cell concentration to the peristaltic pump current (see Note 25). 4. The perfusion rate control system is initiated with cell inoculation and runs automatically. 3.8.2 Cell Growth Phase

1. Inoculate EB66® cells at 3  105 cells/mL, turn the ATF unit on, and start the CSPR-based perfusion rate control at 34 pL/ cell/day. 2. Take samples on a regular basis and confirm specific cell factors with off-line Vi-Cell measurements (see Note 23). If cell factor varies, the parameter should be re-adjusted in the control unit.

3.8.3 Virus Production Phase

1. Infect cells at 7.3  107 cells/mL with EB66®-adapted ZIKV and MOI 103. The CSPR-based perfusion rate control is not affected by the switch to virus replication and does not require any re-adjustments. 2. Three days post-infection, maximum 1.0  1010 PFU/mL are achieved.

ZIKV

titers

of

3.9 Substrate Concentration-Based Control

The metabolite concentration-based perfusion rate control aims at optimum metabolite supply for cell growth and virus production. The perfusion rate is set and re-adjusted by prospected growth rates and metabolite consumption rates. Sensors for on-line measurement of metabolites can be used in a simple closed-loop control for media perfusion to maintain a specific metabolite set point. In the following, examples for manual and automated perfusion rate control are given.

3.9.1 Manual Perfusion Rate Control

The perfusion rate is based on off-line measured concentrations of glutamine (Fig. 6). In the following experiment (published in [17]), glutamine levels at about 1 mM were targeted. Therefore, viable cell concentrations (Xi in cells/mL) and substrate consumption rates of the previous sampling interval (qS in pmol/cell/h) are determined to calculate the present perfusion rate (Qpres in mL/h) as: Q pres ¼

X i ∙V W ∙q S S 0  S br

ð6Þ

with the working volume (VW in mL), the substrate concentration of the perfusion medium (S0 in mM), and the target substrate concentration in the bioreactor Sbr. To meet the expected metabolite demand until next sampling point, the last specific cell growth rate (μ in1/h) is determined and prospective perfusion rate (Qprosp in mL/h) is calculated:

158

Alexander Nikolay et al.

Fig. 6 Metabolite concentration-based control to cultivate EB66® cells for YFV production. (a) Cell concentrations (blue dots) and virus titer (red triangles). (b) Fluctuating perfusion rate (in RV/day, solid blue line) to maintain glucose levels (green open squares) and glutamine at 1 mM (green solid squares).Vertical dotted line indicates time point of infection [17]

Q prosp ¼

X i ∙eμt ∙V W ∙q gln S 0  S br

ð7Þ

with initial parameters for EB66® of μ ¼ 0.035 h1 and qgln ¼ 6  103 pmol/cell/h. After each sampling, perfusion rate Qpres is set in the DCU with a time-dependent linear increase to Qprosp. Cell Growth Phase

1. Inoculate EB66® cells with 3  105 cells/mL. 2. After 3 days, the glutamine level reaches the set point of 1 mM. Initiate the perfusion process by starting the ATF unit to maintain glutamine concentrations. 3. Take samples on a regular basis and re-adjust Qpresand Qprosp. Note that variations of μ and qS, which are due to biological variations, can result in strong fluctuations of the perfusion rate (see Note 21).

Virus Production Phase

1. The perfusion rate strategy is directly applicable to the virus production phase. 2. After 12 days, cell concentrations increased to 9.1  107 cells/ mL and cells are infected with YFV at MOI 103. After 36 h post-infection, the cell death phase becomes visible and metabolite levels increase in the bioreactor. 3. Two days post-infection, maximum 7.3  108 PFU/mL are achieved.

YFV

titers

of

Perfusion Control Strategies 3.9.2 Automated Perfusion Rate Control

159

The on-line BioPAT Trace analyzer contains a diffusion membrane at the tip for volume-free sampling. The analyte diffuses from the cell broth into the internal pump-driven buffer stream and is then transported to the measuring cell, where glucose and lactate are quantified (linear quantification range given by the supplier: 0.01–40 g/L for glucose, 0.05–5 g/L for lactate). The analyzer is connected to a laptop for calibration settings and data recording. The calibration is performed with calibration solutions or is adjusted manually after measuring external samples. It has multiple analog output options for closed-loop control strategies which allows to connect the BioPAT Trace analyzer to a permeate pump. However, those have not been tested here. 1. Install the single-use tubing set (#BPT0003) to the BioPAT Trace. Therefore, insert valve tubing into the valve slots following the color code (top, red; middle, yellow; bottom, green) and then click the pump heads onto the motor shaft. Insert the measuring cell to the contact plug with measurement amplifier and the diffusion module with the two drip chambers in vertical positions to the holder. 2. Exchange the dialysis membrane in the dialysis probe with every new run, flush and fill the system with buffer solution, and autoclave the probe in the bioreactor under standard conditions. 3. Attach the tubing system to the dialysis probe and start the program Trace_Mon (v1.0.0.3). Prime the tubing set with the buffer solution and wait multiple measurements before performing the calibration with 1 and 10 g/L glucose and 0.5 and 5 g/L lactate standards, respectively. The resulting reference factor compensates for the diffusion cell and diffusion membrane performance, as well as for temperature differences.

3.10 By-ProductBased Perfusion Rate Control

Lactate (lac) as the secreted by-product is used for perfusion rate control. It is produced by viable cells (Xi in cells/mL) with a cellspecific lactate production rate (qlac in pmol/cell/h) and washedout via the ATF system at a certain perfusion rate (Qperf in mL/h). This results in the following mass balance: Q perf dplac ¼ X i  q lac   plac Vw dt

ð8Þ

If no additional pH control (such as CO2 or base addition) is applied, lactate concentrations ( plac in mM) influence the pH value in the bioreactor. Assuming a direct correlation, the pH value can be kept constant if the lactate concentration remains constant. With a constant lactate production rate at steady state, the pH value can be solely controlled by the dilution of lactate using fresh medium (see Note 26). In the following example, AGE1.CR.pIX® cells are

160

Alexander Nikolay et al.

Fig. 7 Lactate-based perfusion rate control to cultivate AGE1.CR.pIX® to high cell concentrations. For influenza A virus production, the strategy was changed to a hybrid strategy based on a constant RV/day profile. (a) Cell concentration (blue dots), virus titer (red triangles), and relative working volume (initially 600 mL) of the bioreactor (green line). (b) The perfusion rate (in RV/day, blue line) was controlled by different pH set points (green line) from 72 to 6 hpi (orange area). Vertical dotted line indicates time point of infection

used, which show a stable cell-specific lactate production rate and no lactate uptake (Fig. 7). Previous perfusion processes revealed stable pH values at 7.2 without pH control so that this set point is initially chosen (see Note 27). RV/day’s, and more extensively CSPRs, can be adjusted by adaptation of the pH set point (Fig. 7). 3.10.1

Cell Growth Phase

1. Inoculate AGE1.CR.pIX® cells at 1  106 cells/mL. Start ATF recirculation 60 h after inoculation (see Note 16) and leave the pH control of the bioreactor off (see Note 28). 2. After 96 h, the pH value reaches the set point and the perfusion flow rates are controlled by pH value changes. Therefore, use the base pump of the DCU to withdraw permeate from the outlet of the hollow fiber unit. 3. To increase the perfusion flow rate and by that the CSPR, the pH set point can be increased up to 7.35 (see Notes 29 and 30)

3.10.2 Phase

Virus Production

Due to virus infection, the cell metabolism changes dramatically and lactate concentrations may not be correlated to the pH value anymore. To avoid a control catastrophe by an increased lactate production rate and to reach high virus titers (see Note 21), an RV/ day-based perfusion rate strategy coupled with a reduction and expansion of the initial working volume of 600 mL is applied [18]. 1. 3 h before infection, dilute the cell culture medium in the bioreactor by increasing the perfusion rate temporarily to 12 RV/day.

Perfusion Control Strategies

161

2. Concentrate cells by reducing the working volume from 600 to 300 mL and infect with IAV at MOI 103 and a trypsin activity of 107 units/cell. 3. Expand the working volume to 800 mL with a perfusion rate of 2 RV/day (based on 600 mL) for 12 h. 4. Set the perfusion rate to 1.3 RV/day (based on initial working volume). 5. The highest IAV titer of 3.3 log10(HAU/100 μL) is reached 60 h post-infection.

4

Notes 1. A variety of suspension cell lines such as MDCK, AGE1.CR. pIX®, BHK (Fig. 8), and EB66® cells [17] can grow in pseudoperfusion cultivations to very high cell concentrations. This enables high-throughput media and infection studies beyond cell concentrations typically obtained in batch cultivations. If initial small-scale approaches fail, the cell growth should be additionally investigated in scalable perfusion bioreactor cultivations. 2. Cell-specific virus yields (CSVYs) are based on total cell concentrations (derived from Vi-Cell measurement) as the virus production is a lytic process. In particular, it is assumed that dead (trypan blue-positive) cells also contributed to the final virus titer. This rather underestimates the CSVY but seems a fair approximation.

Fig. 8 Cell growth of different cell lines in pseudo-perfusion cultivations. MDCK cells (red circles), AGE1.CR.pIX® cells (orange triangles), and BHK cells (blue squares) achieve cell concentrations up to 5  107 cells/mL

162

Alexander Nikolay et al.

3. As virus particles accumulate in the bioreactor and only a minor percentage passes through the membrane, the equation for CSVY estimation considers only the virus concentration in the bioreactor. Virus retention has been observed for influenza A, MVA, and flaviviruses at membrane pore sizes even larger than the virus particle diameter. The equation can be simply extended in case virus titers in the permeate stream are significant. 4. Proteolytic enzymes, such as trypsin, typically digest trypan blue-positive cells within trypsinization incubation times. Thus, viabilities should not be considered for data plotting or calculations after virus infection with trypsin addition. 5. Vi-Cell measurement settings must be individually identified for all cell lines. Therefore, begin data acquisition with standard parameters as given by the manufacturer’s instructions. Then, check the cell annotation manually, adjust parameters if necessary, and re-analyze the sample. Set the cell diameter range relatively narrow to exclude false-positive counts (e.g., small particles or cell debris). Also consider the decluster degree as important parameter to distinguish cells in aggregates. Confirm correct annotations after every parameter change. Shearsensitive cells may require less aspiration and trypan blue mixing cycles. 6. Complete virus inactivation should be verified in permissive adherent or suspension cells over at least two passages (T75 tissue culture flask or 125 mL shake flask, 1 mL sample in 25 mL working volume). All relevant virus activity assays need to be negative after the last passage; appropriate controls need to be included. General cell substrate and virus works are subject to laboratory biosafety guidelines, and biosafety levels are defined by health authorities such as the CDC, the NIH, the ZKBS, and other authorities. 7. To avoid vessel bursting during autoclaving, keep water connectors for double-glass heating jacket open. Furthermore, keep the exhaust gas valve of the ATF chamber open. 8. Even if the pH value may not be critical for cell growth, it influences strongly the trypsin activity during influenza virus production and is therefore essential to achieve high titers. The pH value may need to be adjusted for the virus production phase. 9. Additional safety precautions may be required when handling biosafety level 2 viruses in a bioreactor environment. In this application, bioreactor systems are placed in an autoclavable drip tray to avoid spillage of liquids. Bioreactor setups and additional equipment are mounted to safety stands to prevent accidental overturning. The use of glassware should be reduced

Perfusion Control Strategies

163

to the minimum and replaced by plasticware. Additional safety measures may need to be considered and should be discussed with local health authorities. 10. Minimal exchange flow rates should be chosen to minimize mechanical shear stress on the cell. As a rule of thumb, the specific flow velocity can be set to a moderate shear stress (γ) of 2000 s1 for mammalian cells. However, operational minimal of 0.5 L/min for the ATF2 system resulted in poor aspiration. Accumulating gasses in the tube led to a recirculation stop. Thus, it is advisable to perform the ATF2 recirculation with 0.8 L/min. 11. In certain experiments, it is wise to de-couple the start of the recirculation with the ATF2 diaphragm pump from the perfusion process feeding fresh medium. Thus, the impact of each single step can be investigated for cell growth. 12. Certain hollow fiber membranes might be reusable after a perfusion run. For reconditioning, use the ATF2 system to wash the membrane with non-sterile PBS for 1 h. A washing solution of 0.1–0.5 M NaOH is then used for 45 min at 30–50  C. NaOH is then removed by washing with Milli-Q water or PBS for 30 min. Disconnect the membrane from the ATF system and dry the membrane by flushing air at low flow rates through the membrane before storing. 13. Due to periodic media removal during the infection phase, initial virus particle concentration can drop due to membrane passing/membrane binding and dilution with fresh medium. Thus, a higher multiplicity of infection (MOI) may be necessary to initiate virus production phase, especially for slowly propagating viruses such as yellow fever virus. For final calculations, the total amount of accumulated virus and the final harvest can be considered. 14. Pseudo-perfusion cultivations can be controlled by various perfusion rate control strategies. Limitations, however, are given by increasing sampling measurement volumes (especially for spin tube cultivations) and by the lack of on-line measurement tools. 15. A cell-specific perfusion rate (CSPR) of 60 pL/cell/day was chosen based on experimental data. A CSPR should be chosen for which substrate limitations can be excluded as far as possible. 16. As the accumulating cell pellet takes a major volumetric proportion at higher cell concentrations, a maximum volume exchange between 60 and 90% can be chosen. Higher media removal can increase the risk to lose cells during passaging. Thus, it may be advisable to wash used pipette tips with fresh perfusion medium. Furthermore, maximizing exchange

164

Alexander Nikolay et al.

volumes may lead to strong erratic environmental changes (e.g., pH, osmotic pressure, metabolites) impairing cell growth. 17. The optimum multiplicity of infection (MOI, based on infectious titer) must be screened for each new virus isolate and potentially adjusted. Good starting points are MOIs between 103 and 102 for flaviviruses, whereas promising MOIs for MVA virus range between 0.1 and 1. As MVA forms virus aggregates, the virus has to be sonicated for 1 min before infection. IAV can be tested in the range of 102and 106, and HA needs to be activated for membrane fusion and viral genome release by the addition of trypsin to the medium in the range of 5–50 units/mL. Perfusion cultivations may require additional trypsin additions throughout the cultivation time. Trypsin can be therefore added batch-wise with infection and additionally supplemented to the perfusion medium or added 12–18 h post-infection again. If low MOIs are required, it may be advisable to pre-dilute the virus stock in sterile PBS or growth medium by a factor of 10 or even 100. 18. A simple way to determine a suitable CSPR and to reduce the risk of metabolite limitations is the identification of key metabolites and their specific uptake rates. The CSPR (pL/cell/ day) can be calculated based on the cell-specific limiting metabolite consumption rate (qmet in pmol/cell/h), the metabolite concentration in the medium (s0, met in mM), and the targeted metabolite concentration in the bioreactor (sbr, met in mM), as follows: CSPR ¼

q met s 0,met  s br,met

19. The manual CSPR-based perfusion rate control considers constant changes in cell concentrations and specific cell growth rates. Although growth rates may alter during the cultivation, it may be advisable to fix this parameter to avoid controller fluctuations which may result in strong environmental changes for the cell. The CSPR typically drops throughout the cultivation time. This may lead to metabolite accumulation. 20. After virus infection, a specific cell death rate might be taken into account and has to be subtracted from the specific cell growth rate (μ) used in Eq. 5. 21. When cells are infected at certain cell concentrations (sometimes already at 5  106 cells/mL), a decrease in cell-specific virus yields can be observed for many viruses such as influenza virus [19], MVA virus [8], or adenovirus [20]. Previous studies have shown that limitation in yields can be overcome by replacing spent cell growth medium by fresh medium at the time of

Perfusion Control Strategies

165

infection and expanding the bioreactor working volume during the infection phase [8]. Thus, the decrease in CSVYs may derive from nutritional limitations or unknown inhibitors secreted in the medium. 22. The CSPR control is based on the use of the biomass probe with a Sartorius DCU. In other trials, the ArcView controller was also connected successfully to the Eppendorf DasGip DCU. Thus additional OPC licenses may be required. All read-outs from the ArcView were transferred to the DASware control system and cell concentrations could be easily used to estimate pump rates. This enabled a user-friendly and fast integration of biomass data without time-consuming data calibration (see Notes 23 and 24). 23. Cell factors may vary during the cultivation run due to cell morphology changes. It is therefore advisable to double-check cell concentration by off-line measurements and re-adjust the cell factor. To avoid such issues, a control strategy based on the viable cell volume is conceivable resulting in a cell volumespecific perfusion rate (CVSPR) control. Furthermore, it was observed that cell factors varied between cultivation runs with exactly the same setup; a good explanation is still missing. 24. The cable to connect the ArcView Controller to the analog output box and the cable from the analog output box to the peristaltic pump were built in the in-house electronic workshop. 25. The identification of calibration values must be performed empirically. The challenge is to find values describing the linear correlation of output signal to pump rate and the linear correlation of pump rate to volume flow. First, the planned CSPR and working volume of the bioreactor vessel have to be defined. Then, calibrate the peristaltic pump and calculate theoretical pump rates for low and high cell concentration scenarios at the defined working volume. The perfusion control unit is set up with water. Start the ArcView controller, go to Unit Settings Menu, select the unit for Incyte, click on Measure, and edit Zero Cell Density to a manual value (type low or high cell concentration scenario). Thus, different cell concentrations can be simulated and peristaltic pump rates can be read. Use an Excel spreadsheet to visualize the correlation of pump rate to cell concentration, note the process output values, and add the desired correlation line into the graph. Now, try different configuration values to overlay both correlation lines. Unfortunately, 0 as lower (4 mA) configuration value does not necessarily mean 0 rpm. As an advice, fix the upper (20 mA) configuration value and change the lower (4 mA) configuration value step-wise. Be aware that this

166

Alexander Nikolay et al.

changes the slope as well as the y intersection. For our application and setup, configuration values of –12.5 for 4 mA and 1740 for 20 mA fitted the best. Once, configuration values have been identified and bioreactor working volumes remain constant for further experiments, CSPRs can be simply changed by editing Cell Density unit/Factor in channel Incyte and field Measure, which multiplies the used CSPR for calibration, and the product describes the new process CSPR. 26. Increasing the pH and buffer capacity of the feeding medium can be used to optimize the perfusion rate for the desired pH set point or to reduce the CSPR at later phases of the cultivation with CO2 accumulation. Additionally, buffer systems independent of CO2 (e.g., HEPES) can be added to support stable pH values. 27. Used cell lines have to be well characterized both in batch and perfusion cultivations to determine feasible pH set points and appropriate lactate concentrations. 28. It was observed that pH values increased strongly with re-circulation of the medium by the ATF diaphragm pump. This is probably due to degassing of CO2 in the medium during the alternating flow changing between over- and under-pressure. The pH value typically decreases after 2 to 3 days to normal values. Otherwise, the pH increase can be countered by CO2 sparging. This phenomenon was not observed for cross-flow filtration such as tangential flow filtration (TFF). 29. pH set point values may be reduced to achieve higher cell concentrations. Insufficient CO2 stripping interferes with the control as it decreases the pH. In consequence, CSPRs elevate by constant lactate concentrations. Additional sparging of nitrogen or air can help to remove CO2 from the cultivation broth. As another alternative, open-tube or macro-sparger can be used instead of micro-sparger. 30. Increasing CSPR over the cultivation is probably unavoidable with this control system. However, the ease of this application together with a base-free pH control is making this strategy interesting, if overfeeding is not a big issue.

Acknowledgments We would like to thank our former colleague Daniel V. Ramirez for experimental data on the lactate-based perfusion rate control and for the provision of the ATF bioreactor illustration. Furthermore, we would like to express our gratitude to our collaboration partners for the allowance to work with the cell lines (ProBioGen AG,

Perfusion Control Strategies

167

AGE1.CRpIX; Valneva, EB66®; IDT, BHK-21; East China University of Science and Technology, MDCK). Additionally, we thank Hamilton Bonaduz AG for providing the analog output box and Eppendorf for the implementation to the DasGip control system. Our gratitude also goes to Sartorius AG for providing the BioPAT Trace system. References 1. Plotkin SA (2003) Vaccines, vaccination, and vaccinology. J Infect Dis 187(9):1349–1359. https://doi.org/10.1086/374419 2. Andre FE, Booy R, Bock HL, Clemens J, Datta SK, John TJ, Lee BW, Lolekha S, Peltola H, Ruff T (2008) Vaccination greatly reduces disease, disability, death and inequity worldwide. Bull World Health Organ 86:140–146 3. Tapia F, Va´zquez-Ramı´rez D, Genzel Y, Reichl U (2016) Bioreactors for high cell density and continuous multi-stage cultivations: options for process intensification in cell culture-based viral vaccine production. ApplMicrobiolBiotechnol 100(5):2121–2132. https://doi.org/ 10.1007/s00253-015-7267-9 4. Kompala DS, Ozturk SS (2005) Optimization of high cell density perfusion bioreactors. In: Ozturk SS, Hu WS (eds) Cell culture technology for pharmaceutical and cell-based therapies, vol 1. CRC Press, Boca Raton, pp 387–411 5. Karst DJ, Steinhoff RF, Kopp MRG, Serra E, Soos M, Zenobi R, Morbidelli M (2017) Intracellular CHO cell metabolite profiling reveals steady-state dependent metabolic fingerprints in perfusion culture. BiotechnolProgr 33 (4):879–890. https://doi.org/10.1002/btpr. 2421 6. Ozturk SS (1996) Engineering challenges in high density cell culture systems. Cytotechnology 22(1):3–16. https://doi.org/10.1007/ bf00353919 7. Dowd JE, Jubb A, Kwok KE, Piret JM (2003) Optimization and control of perfusion cultures using a viable cell probe and cell specific perfusion rates. Cytotechnology 42(1):35–45. https://doi.org/10.1023/A:1026192228471 8. Vazquez-Ramirez D, Genzel Y, Jordan I, Sandig V, Reichl U (2018) High-cell-density cultivations to increase MVA virus production. Vaccine 36(22):3124–3133. https://doi.org/ 10.1016/j.vaccine.2017.10.112 9. Hiller GW, Ovalle AM, Gagnon MP, Curran ML, Wang W (2017) Cell-controlled hybrid perfusion fed-batch CHO cell process provides significant productivity improvement over

conventional fed-batch cultures. BiotechnolBioeng 114(7):1438–1447. https://doi.org/ 10.1002/bit.26259 10. Gallo-Ramirez LE, Nikolay A, Genzel Y, Reichl U (2015) Bioreactor concepts for cell culturebased viral vaccine production. Expert Rev Vaccines 14(9):1181–1195. https://doi.org/10. 1586/14760584.2015.1067144 11. Jordan I, Vos A, Beilfuß S, Neubert A, Breul S, Sandig V (2009) An avian cell line designed for production of highly attenuated viruses. Vaccine 27(5):748–756. https://doi.org/10. 1016/j.vaccine.2008.11.066 12. Brown SW, Mehtali M (2010) The Avian EB66 (R) cell line, application to vaccines, and therapeutic protein production. PDA JPharmaceutSciTechnol/ PDA 64(5):419–425 13. Huang D, Peng W-J, Ye Q, Liu X-P, Zhao L, Fan L, Xia-Hou K, Jia H-J, Luo J, Zhou L-T, Li B-B, Wang S-L, Xu W-T, Chen Z, Tan W-S (2015) Serum-free suspension culture of MDCK cells for production of influenza H1N1 vaccines. PLoS One 10(11):e0141686. https://doi.org/10.1371/journal.pone. 0141686 14. Spearman C (1908) The method of ‘right and wrong cases’(‘constant stimuli’) without Gauss’s formulae. Br J Psychol 2(3):227–242 15. K€arber G (1931) BeitragzurkollektivenBehandlungpharmakologischerReihenversuche. Naunyn-SchmiedebergsArchivfu¨rexperimentellePathologie und Pharmakologie 162 (4):480–483. https://doi.org/10.1007/ bf01863914 16. Konstantinov K, Goudar C, Ng M, Meneses R, Thrift J, Chuppa S, Matanguihan C, Michaels J, Naveh D (2006) The“push-tolow” approach for optimization of high-density perfusion cultures of animal cells. AdvBiochemEng/Biotechnol 101:75–98 17. Nikolay A, Le´on A, Schwamborn K, Genzel Y, Reichl U (2018) Process intensification of EB66® cell cultivations leads to high-yield yellow fever and Zika virus production. ApplMicrobiolBiotechnol 102(20):8725–8737. https://doi.org/10.1007/s00253-018-9275-z

168

Alexander Nikolay et al.

18. Va´zquez-Ramı´rez D, Jordan I, Sandig V, Genzel Y, Reichl U (2019) High titer MVA and influenza A virus production using a hybrid fed-batch/perfusion strategy with an ATF system. ApplMicrobiolBiotechnol 103 (7):3025–3035. https://doi.org/10.1007/ s00253-019-09694-2 19. Genzel Y, Vogel T, Buck J, Behrendt I, Ramirez DV, Schiedner G, Jordan I, Reichl U (2014) High cell density cultivations by alternating tangential flow (ATF) perfusion for

influenza A virus production using suspension cells. Vaccine 32(24):2770–2781. https://doi. org/10.1016/j.vaccine.2014.02.016 20. Ferreira TB, Ferreira AL, Carrondo MJT, Alves PM (2005) Effect of refeed strategies and non-ammoniagenic medium on adenovirus production at high cell densities. J Biotechnol 119(3):272–280. https://doi.org/10.1016/j. jbiotec.2005.03.009

Chapter 10 How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200 L Scale Cedric Schirmer, Jan Mu¨ller, Nina Steffen, So¨ren Werner, Regine Eibl, and Dieter Eibl Abstract Single-use bioreactors have increasingly been used in recent years, for both research and development as well as industrial production, especially in mammalian cell-based processes. Among the numerous singleuse bioreactors available today, wave-mixed bags and stirred systems dominate. Wave-mixed single-use bioreactors are the system of choice for inoculum production, while stirred single-use bioreactors are most often preferred for antibody expression. For this reason, the present chapter describes protocols instructing the reader to use the wave-mixed BIOSTAT® RM 50 for cell expansion and to produce a monoclonal antibody (mAb) in Pall’s Allegro™ STR 200 at pilot scale for the first time. All methods described are based on a Chinese hamster ovary (CHO) suspension cell line expressing a recombinant immunoglobulin G (IgG). Key words CHO cell expansion, Fed-batch, IgG production, Stirred single-use bioreactor, Wavemixed single-use bioreactor

1

Introduction Single-use (also referred to as disposable) bioreactors whose cultivation containers are fabricated from plastics instead of glass or stainless steel offer advantages over their reusable counterparts. Because these plastic containers are typically provided pre-sterilized by the vendors, they can be quickly brought into operation. There is no need for sterilization of the cultivation containers, which are decontaminated and discarded after completion of process operations. Thus, time-consuming cleaning procedures with corrosive chemicals and water for injections can be dispensed with. Moreover, the time and cost of cleaning validation and the risk of product cross-contaminations in multiproduct facilities are reduced. For this reason, many users of single-use bioreactors confirm the results of studies demonstrating that these allow for safer, greener, more flexible, and cheaper production processes

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_10, © Springer Science+Business Media, LLC, part of Springer Nature 2020

169

170

Cedric Schirmer et al.

than their reusable counterparts. However, this presupposes that the bioreactor type was appropriately selected and that the container, which is either a rigid polystyrene or polycarbonate container or a flexible bag, was installed and operated correctly. Various types of bag designs can be differentiated: two-dimensional, pillowlike and three-dimensional, and cylindrical or cube-shaped. Due to their fragility, the bags have to be shaped and fixed by a bag holder (tray or tank) which is also used for tempering and cooling of the bag content. The bioreactor bag consists of multilayer films, the preferred contact layers of which are polyethylene and ethylene vinyl acetate. Weaknesses resulting from single-use bioreactors arise mainly from chemical, biological, and physical properties and the processing of the plastics used. For example, bioreactor bag scalability is currently limited to 6 m3. Furthermore, the replacement of the bioreactor container constitutes an increase in operating costs and storage capacity. However, the primary risk associated with singleuse bioreactors is the possible migration of leachables in addition to leaks. Both leaks and leachables in particular concern bioreactor bags. If a single-use bioreactor is to be implemented successfully, prior risk analyses and studies of the production cells and the culture medium are required. A strong cooperation with the manufacturer of the bioreactor may contribute to a reduction in experiments on the part of the user [1–4]. 1.1 Predominance of Stirred and WaveMixed Single-Use Bioreactors

As described by Eibl et al. [2], nowadays the user can choose between a multitude of different single-use bioreactors offered by different manufacturers. They differ in size, film material, design and instrumentation of the bioreactor container, type of power input, and control unit. When not concerning static systems, single-use bioreactor types available on the market are mechanically driven, hydraulically driven, pneumatically driven, or hybrid. The majority of existing single-use bioreactors are mechanically driven, namely, stirred, wave-mixed, and orbitally shaken. Stirred singleuse bioreactors available from 10 mL to 6 m3 working volume form the largest group, followed by wave-mixed systems. Wave-mixed single-use bioreactors have replaced spinner flasks and reusable stirred bioreactors in inoculum productions over the past 10 years. Although wave-mixed bioreactors with one-, two-, and three-dimensional motion are obtainable, those with one-dimensional motion are predominant, such as the representatives of the BIOSTAT® RM family used to produce the inoculum for the subsequently described Allegro™ STR study. In wavemixed bioreactors with one-dimensional motion, the mixing of the culture broth is achieved by moving the rocker with the bag back and forth. The intensity of mixing can be varied via rocking angle, rocking rate, and filling volume of the bag, which is a maximum of 50% of the total volume. By moving the bag

How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200. . .

171

Table 1 Overview of available stirred single-use bioreactors exceeding benchtop scale for mammalian cellbased processes Bioreactor brand

Working volume (L)

Vendor

CSR™ series

50–6000

ABEC

Xcellerex™ XDR

4.5–2000

GE Healthcare

Mobius® CellReady

10–2000

Merck Millipore

Allegro™ STR

10–2000

Pall Life Sciences

12.5–2000

Sartorius Stedim Biotech

10–2000

Thermo Fisher Scientific

BIOSTAT STR

®

HyPerforma™ S.U.B.

containing culture medium and cells, a wave is induced, while the medium surface is permanently renewed. Gas is exchanged via medium surface, and the generated foam is worked into the culture broth. This usually makes the addition of an antifoam agent, as is required for stirred single-use bioreactors, unnecessary. [2, 3, 5] An overview of stirred single-use bioreactors exceeding benchtop scale for mammalian cell-based processes is given in Table 1. These have typical H/D ratios of about 2:1, lower specific power inputs (5–200 W m3), smaller gassing rates (up to 0.2 vvm), and smaller oxygen transfer rates (0.5–8 mmol O2 L1 h1) compared to microbial versions [5–7]. The three-dimensional bags come already equipped with gassing devices, sampling ports, and further hose connections and partly also with single-use sensors. The temperature is controlled in the simplest way by using a heat blanket installed in the bag holder. For applications where the temperature reduction is of importance, systems with a double jacket are available. In order to prevent blocking of the exhaust air filters, filter heaters are used instead of exhaust air coolers [5, 7, 8]. It is also worth mentioning that centrically mounted, rotating impellers in cylindrical bags dominate, which are either bottom- or top-driven by magnetic force or a single mechanical seal. Cube-shaped bags as installed in the Allegro™ STR family are rather unique, although they ensure gentle and intensive mixing. 1.2 The Allegro™ STR Family and Its Characteristics

The Allegro™ STR system (see Fig. 1) is currently available in four vessel sizes: 50 L, 200 L, 1000 L, and 2000 L. An overview of the main important dimensions of the complete bioreactor family can be seen in Table 2 [7, 9]. A large impeller with a diameter of half the vessel width is mounted at the bottom. The impeller is driven by a shaft, mounted with a seal through the sterile barrier. The bottommounted shaft is short, and thus, packaging size can be reduced. The impeller can be rotated in clockwise or anti-clockwise direction in order to generate an upward or downward directed fluid flow,

172

Cedric Schirmer et al.

Fig. 1 Scheme of the biocontainer of the Allegro™ STR bioreactor family. W width, C clearance from the bioreactor bottom to the middle of the impeller, Di impeller diameter, HL liquid height Table 2 Overview of Allegro™ STR bioreactor dimensions Allegro STR

50

200

1000

2000

Working volume (L)

10–50

60–200

300–1000

400–2000

Liquid height HL (mm)

363

625

1090

1360

Width  depth (mm)

383  353

585  585

1090  1025

1360  1270

Impeller diameter Di (mm)

185

290

490

490

Liquid height HL/width

1

1

1

1

Impeller diameter Di/width

0.5

0.5

0.5

0.4

Sparger hole size (mm)

1

1

2

2

Sparger hole number

8

6

28

28

respectively. The bioreactor system comprises of the holding frame with heating and/or cooling capability and the control tower with a user interface (see Fig. 2). Besides controlling temperature and agitation speed, pH and dissolved oxygen concentration (DO) can be measured and controlled. The measurement of pH and DO can be executed by classical probes, which have to be

How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200. . .

173

Fig. 2 Pall’s Allegro™ STR 200 bioreactor (image provided courtesy of Pall Corporation)

introduced into the bag under sterile conditions by the user, or with pre-installed single-use sensors. The Allegro™ STR 200 bioreactor applied in the investigations is well-characterized. The specific power input, the mixing time, the oxygen mass transfer coefficient, the CO2 stripping rate, and the heating capability were determined. The power number of the large single impeller is 1.9 for upward directed flow and 2.2 for downward directed flow, which was determined by measuring the torque with a torque transducer fitted to the shaft. Measurements with empty vessels were accomplished to account for friction because of bearings and seals. Determination of the mixing time was done by changing the conductivity of a completely mixed system with a small volume of a high concentrated salt solution (15% NaCl, 100 mL) and, subsequently, measuring the time at which homogeneity was reached. Conductivity was measured with two probes at a temperature of 37  C, one at the bottom and one in a corner of the bag. Depending on the impeller speed, the mixing time ranged between 10 and 45 s (impeller speed of 25–150 rpm). The gassing-out method was used to determine the volumetric oxygen mass transfer coefficient kLa. The measurements were conducted with two probes at 37  C. The kLa value was about 5 h1 and below for an impeller speed of 75 rpm and below, 13 h1 for 100 rpm, 23 h1 for 125 rpm, and 33 h1 for 150 rpm at an aeration rate of 20 L min1 and at 200 L working volume. In addition, it could be

174

Cedric Schirmer et al.

shown that sufficient CO2 stripping may be achieved with the bioreactor system, depending on the air flow rate. The stripping rate was between 1 mol L1 day1 at 2 L min1, 3 mol L1 day1 at 11 L min1, and 5 mol L1 day1 at 20 L min1[10, 11]. Heating and cooling capability for small-scale laboratory bioreactors is often assumed to play a minor role. However, for larger systems, the temperature stability should be validated. Measurements with nine thermocouples at different locations showed a good temperature stability in a range of 0.2  C, measured over a time period of 1.5 h for the Allegro™ STR family. The heating time of chilled water from 4  C to 37  C took about 9 h [10, 11]. 1.3 CHO-Cell-Based IgG Production in the Allegro™ STR 200

About 70% of the recombinant proteins produced today are based on recombinant CHO cells [12, 13]. Over the last few decades, various efforts have been made to increase the productivity of CHO cells. Today, mAb titers exceeding 5 g L1 are already a reality in fed-batch processes based on chemically defined culture media [13–17]. The high product titers have resulted in shrinking bioreactor sizes in many mAb productions. In fact, the 1 or 2 m3 stirred single-use bioreactor is often sufficient at production scale, whereby the inoculum expansion is usually realized in shake flasks and wavemixed bioreactors as described by Kaiser et al. [18]. A similar approach, but for the 200 L production scale, was used by the present authors in order to produce IgG1 with a CHO-S suspension cell line established by Cobra Biologics. The cells grow in suspension in the chemically defined ActiPro™ medium from GE Healthcare and have average doubling times of 23 h in orbitally shaken, stirred, and wave-mixed bioreactor systems operated in feeding mode. We applied the set-up shown in Fig. 3 to study the IgG production in the cube-shaped Allegro™ STR at 200 L scale. Because this IgG production process was originally developed for a 200 L stirred single-use bioreactor with cylindrical bag, the process parameters had to be modified. The Newton or power number Ne (Eq.1), relating the resistance force to the inertia force, offers the possibility of process transfer to different bioreactor systems with a desired specific power input P/V [19] of 50 W m3. The total power input P can be obtained from the specific power input and the medium volume V present in the bioreactor (Eq. 2). Ne ¼

P ρ   Di 5 n3

P ¼ P=V  V

ð1Þ ð2Þ

Using a Newton number of 2.2 with a downward flow (see Subheading 1.2), the total power input P, an impeller diameter Di of 0.29 m, and a fluid density ρ of 1000 kg m3, we calculated the impeller speed n after conversion of Eq. 1. The parameters required to ensure sufficient gas supply were transferred from the process

How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200. . .

175

Fig. 3 Set-up of the inoculum and IgG production process. P1 passage 1, P2 passage 2, P3 passage 3, P4 passage 4

development based on the cylindrical vessel. In other words, the overlay gas flow with air was kept constant at 5 L min1. The gassing with air by the sparger was adapted according to the volume present in the vessel. A constant gassing with 0.02 vvm, which corresponds to an airflow of 3.2 L min1 at 160 L bioreactor starting volume (Eq. 3), was aimed at. The gassing with oxygen and CO2 was realized depending upon the need for the regulation of the DO and pH value over the sparger by the adjusted control. This presupposed the setting of appropriate control parameters for air, oxygen, and CO2 gassing with the proportional and integral (PI) controller parameters (see Note 1). 0:02 vvm  160 L ¼ 3:2 L  min 1

ð3Þ

Figure 4 and Table 3 summarize the results achievable when growing the IgG1-secreting CHO-S suspension cell line in the Allegro™ STR 200. The results are comparable to those achieved using a single-use stirred bioreactor operated with cylindrical bag of same size and under identical fluid flow conditions (data not shown). The average peak cell density at the end of the growth phase was 2.15  107 cells mL1, which correlated with a decrease in viability from this point. The specific growth rate during the exponential growth phase was 0.028 h1. During this period, lactate production up to a peak level of 2.3 g L1 with subsequent rapid consumption and low lactate levels (metabolic shift) were observed. The glucose concentration of 6.4 g L1 present in the medium at the beginning of the cultivation showed a continuous decrease during the first 60 h. However, with the subsequent

176

Cedric Schirmer et al.

Fig. 4 Cell growth profile (top left), lactate and glucose profiles (top right), as well as mAb quantities (bottom left) achieved in the Allegro™ STR 200. In addition, results of the main glycan analysis of the harvested mAb are presented (bottom right). G galactose units, F fucose units Table 3 Charge variant and antibody composition of harvested mAbs in Pall’s Allegro™ STR 200 Aggregate (%)

Low molecular weight species (%)

Monomer (%)

Acidic variant Basic variant Target mAbs (%) (%) (%)

3.0  0.5

1.7  0.1

95.3  0.4

32.5  0.7

19.5  1.4

48.0  0.7

feeding procedure and further glucose consumption, a fluctuating concentration profile appeared. The antibody concentration in the Allegro™ STR 200 started at 0.03 g L1 on the third day (start of feeding). As expected, the mAb titer increased to an average level of 2.18 g L1 in the culture harvest (day 11). The main peak of the main glycan analysis at G0F showed a predominantly fucosylated species, as is typical of this IgG antibody. From the charge variant analysis, the mAb composition in the form of acidic and basic variants was obvious and indicated that the target mAb remained unchanged. To summarize, a very close variant profile to the reference antibody was detected, and only small amounts of aggregates or low molecular weight species were found. This allows us to conclude that the shear stress acting on the cells under the selected process parameters does not impair the IgG quality in the Allegro™ STR 200.

How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200. . .

2

177

Materials The materials listed below were necessary for the IgG production in the Allegro™ STR 200. In addition to these materials, the following were routinely used: standard cell culture equipment (e.g., microscope, safety cabinet class II, centrifuge, water bath, magnetic stirrer, CO2 incubators, pH meter, pipettes, 0.2 μm filters, sample vials, shake flasks, medium storage bags, peristaltic pumps, balances, tubes, clamps) as well as a welder (see Note 2) and a tube sealer (see Note 3). The following analyzers were also required for process monitoring: cell counter (see Note 4) and analyzer for medium compounds (see Note 5) and IgG (see Note 6).

2.1 CHO Cell Expansion (Inoculum Production) in Shake Flasks and in the BIOSTAT® RM 50

1. CHO cell line of interest (see Note 7). 2. Corresponding chemically defined cultivation medium prepared according to the manufacturer’s instructions and stored at 4  C in the dark (see Notes 8–11). 3. Shake flasks with filtered venting caps in sizes of 125 mL, 500 mL, and 1 L. 4. BIOSTAT® RM 50 with FlexSafe® RM 50 L basic (Sartorius Stedim Biotech).

2.2 IgG Production in the Allegro™ STR 200

1. Prepared inoculum in sufficient quantity (see Subheadings 3.1 and 3.2). 2. Chemically defined cultivation medium prepared according to the manufacturer’s instructions and stored at 4  C in the dark (see Notes 8–11). 3. Chemically defined feed media prepared according to the manufacturer’s instructions and stored at 4  C in the dark (see Note 8). 4. Glucose solution 450 g L1 for additional feed (see Subheading 3.3 and Note 12). 5. Antifoam C (see Note 13). 6. Sodium hydroxide solution 1 M (see Note 14). 7. Kleenpak™ sterile connectors (Pall). 8. Allegro™ STR 200 with a standard Allegro™ STR 200 biocontainer (Pall) and conventional pH and DO probes.

3

Methods The following protocols, developed for cell expansion and IgG expression with a CHO cell line, include preliminary production of media, inoculum production, preparation of bioreactors,

178

Cedric Schirmer et al.

inoculation, sampling, as well as analytics. As outlined in Fig. 3, the process starts from cryovial via inoculum production in shake flasks and the BIOSTAT® RM 50 where a volume expansion strategy was applied. The final step was the IgG production in the Allegro™ STR at pilot scale. With some modifications, protocols can easily be transferred to other animal suspension cell lines. 3.1 Shake FlaskBased Inoculum Production

1. Prepare the cell culture medium for the preculture in the shake flasks (see Note 8). 2. Expand CHO suspension cells in shake flasks in order to produce the preculture for the inoculum production in the BIOSTAT® RM 50 (see Subheading 3.2). (a) Thaw CHO cells 15 days before the start of the IgG production run. Remove one cryovial containing approximately 1 mL cell suspension with a concentration of 1  107 cells mL1 from the cryotank and immediately thaw the vial in a pre-warmed water bath (37  C). The lid should not touch the water to avoid risk of contamination. As soon as only small ice crystals are left, the cryovial is wiped with a tissue and alcohol. The whole content of the cryovial is directly transferred into a 125 mL shake flask containing 29 mL medium in a safety cabinet. (b) Incubate the cells in a shaking incubator at 37  C, 108 rpm (25 mm shaking diameter), 80% relative humidity, and 8% CO2. (c) 24 h post-thawing, take a 1 mL sample in order to measure the cell density and viability (see Note 4). If the cells did not proliferate as expected, centrifuge the cell suspension (300 g, 3 min) and exchange the media. (d) Take a 1 mL sample every 24 h in order to check cell density and viability (see Note 4). (e) Expand the culture in two 500 mL shake flasks with 100 mL working volume 3 days post-thawing (passage 2). The required cell suspension volume (inoculation density of 0.3–0.5  106 cells mL1) is added to the two shake flasks prepared with cell culture media. In so doing, use 50 and 5 mL serological pipettes. (f) Seven days post-thawing, repeat the expansion with passage 3 into six 1 L shake flasks with a working volume of 250 mL. (g) Feed 100 mL culture medium 3 days post-expansion (1 day before the start of the inoculum production in the BIOSTAT® RM 50).

How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200. . .

3.2 Inoculum Production in the BIOSTAT RM 50 L

179

In order to reach a high cell density with a small inoculation volume and number of passages, a volume expansion strategy is used for inoculum production in the BIOSTAT® RM 50 with a FlexSafe® RM 50 L basic. Start with 12.5 L culture volume (1/2 of the maximum working volume) and an inoculation cell density of 0.3–0.5  106 cells mL1. 1. Prepare at least 25 L medium and pour it into a medium bag ensuring it is sterile (see Note 8). The preparation and sterility test of the media should be completed at least 5 days before IgG production in the Allegro™ STR 200 (1 day before inoculum production in the BIOSTAT® RM 50). 2. Prepare the bioreactor for inoculation and cultivation (5 days before IgG production in the Allegro™ STR 200). (a) Unpack the FlexSafe® RM 50 L basic in the safety cabinet and close all pinch clamps. (b) Install the bag on the rocker platform, install the exhaust filter heater, and connect the gassing tubes. (c) Place the media bag on a balance next to the BIOSTAT® RM 50 and connect it to the FlexSafe® RM 50 L by welding it (see Note 2). (d) Transfer approximately 10 L medium with a peristaltic pump into the cultivation bag using the weight indicated by the balance and then close the pinch clamp. (e) Fill the bag with process air and start the heat controller (37  C). (f) Cover the cultivation bag with the cover hood and start the rocking motion (start parameters 7 , 20 rpm). 3. Inoculation of the BIOSTAT® RM 50 (4 days before IgG production in the Allegro™ STR 200). (a) Prepare the inoculum cell suspension. Transfer the shake flasks into the safety cabinet and measure the cell density and cell viability (95% is necessary; see Note 4). Determine the number of shake flasks (total volume) required for inoculation. (b) Take the necessary number of shake flasks and transfer the cell suspension into a 1 L shake flask. Once again, measure the cell density and viability (see Note 4), and calculate the required cell suspension volume in order to achieve an inoculation density of 0.3–0.5  106 cells mL1. (c) Remove the tray of the BIOSTAT® RM 50 from the rocking unit and transfer it into the safety cabinet. (d) Insert a sterile 50 mL syringe (only the cylindrical jacket is used as funnel) into the bag’s Luer Lock port and transfer the calculated volume of cell suspension into the

180

Cedric Schirmer et al.

Table 4 Bag volume and parameter adaptions during the inoculum production in the BIOSTAT® RM 50 Day Working volume Volume expansion Parameter adaptions 0

12.5 L

None

20 rpm, 7 , 0.875 L/min (0.07 vvm), 8% CO2

1

12.5 L

None

25 rpm

2

12.5 L

6.25 L

30 rpm, 8 , 1.31 L/min (0.07 vvm)

3

18.75 L

6.25 L

35 rpm, 8 , 1.75 L/min (0.07 vvm), reduce CO2 to 6.6%

4

25 L

None

Inoculation of the Allegro™ STR 200

FlexSafe® RM 50 L. Afterward, remove the syringe and close the port. (e) Replace the tray with the bag on the rocking unit. (f) Add fresh medium by using a pump and balance to reach the start volume of 12.5 L. 4. Put the bioreactor into operation (start heat controller with 37  C, gassing with 0.07 vvm overlay and 8% CO2, rocking motion, and exhaust filter heater). 5. Take a sample and determine the cell density and viability 30 min after inoculation (see Note 4). Measure the concentrations of metabolites (see Note 5). Repeat this step every 24 h as well as prior and post-feeding. 6. Perform a feeding step of 6.25 L (1/4 reactor working volume, with fresh medium) 48 h post-inoculation (a) Open the pinch clamp and start the peristaltic pump of the connected media bag located on the balance next to the BIOSTAT® RM 50 system. (b) Control the feeding volume by weight change of the media bag. (c) Close the pinch clamp after feeding. (d) Repeat step 5. (e) Increase the rocking motion and the gassing in relation to the present cultivation volume in order to reach an appropriate amount of DO and maintain a pH value in the range of pH 7.0  0.2 (Table 4). 7. 24 h after the first feeding step, repeat steps 5 and 6. After this step, the maximum working volume of the FlexSafe® RM 50 L (25 L) is reached. 8. Four days post-inoculation, repeat step 5. The following criteria must be fulfilled in order to inoculate the Allegro™ STR 200. Cells should be in the middle of the exponential phase and

How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200. . .

181

should have a viable cell density of 4–5  106 cells mL1 and a viability of 95% or higher. 3.3 IgG Production in the Allegro™ STR 200

The following procedure describes the IgG production in the Allegro™ STR 200 using a fed-batch process: 1. Prepare at least 150 L medium, a necessary amount of feed medium, and pour it into medium bags ensuring it is sterile (see Note 8). The preparation and sterility test of the media should be completed at least 2 days before inoculation of the Allegro™ STR 200. 2. Prepare the additives required for the cultivation at least 2 days before inoculation of the Allegro™ STR 200. (a) Prepare 5 L of a glucose solution with a concentration of 450 g L1 in a bottle with two tube connections (see Note 12). (b) Prepare 2 L of an Antifoam C solution by diluting 1:1000 with water in a bottle with two tube connections (see Note 13). (c) Prepare 5 L of a 1 M sodium hydroxide solution in a bottle with two tube connections (see Note 14). (d) Connect the inlet tubes of the three prepared bottles with sterile filters and the outlet tubes with Kleenpak® connectors. (e) Sterilize the bottles with the additives at 121  C for 20 min in the autoclave. 3. Prepare the conventional pH and DO probes 2 days before inoculation of the Allegro™ STR 200. (a) Connect the probes to the control unit and wait until the DO probe is polarized. (b) Calibrate the pH probe with pH 4 and 7 buffer solutions. (c) Calibrate the zero point of the DO probe with nitrogen. (d) Fit the probes in probe bellows with Kleenpak® connectors and sterilize them by autoclaving at 121  C for 20 min. 4. One day before inoculation, prepare the bioreactor. (a) Unpack the Allegro™ STR 200 bioreactor and close all pinch clamps. (b) Install the bag in the bioreactor system following the set-up instructions on the bioreactor control unit. (c) Insert the temperature probe. (d) Insert the pH and DO probes by connecting the bellows with the Kleenpak® connectors, pushing the probes into the reactor, and locking them into position.

182

Cedric Schirmer et al.

(e) Start the filter heaters. (f) Fill the bioreactor with 150 L of the medium by connecting the medium bags one by one with the welder (see Note 2) and pumping the medium with a peristaltic pump into the bioreactor. (g) Set the agitation speed (downflow) and the overlay airflow to the start parameters (73 rpm and 5 L min1). The agitation speed is calculated by using a constant specific power input per volume of 50 W m3 and a power number of 2.2 (see Subheading 1.3) and has to be changed with increasing volume during the cultivation. The overlay gas flow is kept constant. (h) Set the temperature setpoint to 37  C. (i) Connect the bags with the feed medium (see Note 8) to the bioreactor by welding it (see Note 2). (j) Connect the bottles with Antifoam C, glucose, and sodium hydroxide to the bioreactor using the Kleenpak® sterile connectors. (k) Insert the tubes of the Antifoam C and the sodium hydroxide bottles into the peristaltic pumps of the control unit. In this way, the base can be given into the bioreactor automatically using a pH control loop and the antifoam manually using the control unit. (l) Prime the tubes of the feeding media and the other additives. 5. Final preparations before the inoculation. (a) When the temperature is reached and the DO value is stable, recalibrate the DO probe to 100%. (b) Take a sample using two syringes. Discard 5 mL with the first syringe and take a 5 mL sample with the second. Measure the pH externally and recalibrate the pH probe. (c) Set the sparger airflow to 3.2 L min1 (0.02 vvm of the start volume of 160 L) and keep a constant value of 0.02 vvm during the cultivation. (d) Start the pH control loop (pH 7.0; see Note 1) and the DO control loop (40% air saturation; see Note 1). 6. Inoculate the Allegro™ STR 200. (a) Check the viable cell density and viability (see Note 4) and calculate the necessary volume of inoculum to reach a starting cell density of 0.4  106 cells mL1. (b) Place the tray including FlexSafe® RM 50 L from the rocking unit on a balance next to the Allegro™ STR 200.

How to Produce mAbs in a Cube-Shaped Stirred Single-Use Bioreactor at 200. . .

183

(c) Connect the FlexSafe® RM 50 L to the Allegro™ STR 200 by welding and transfer the calculated volume of cell suspension by gravitation. Disconnect the FlexSafe® bag by sealing (see Note 3). (d) After 30 min of stirring, take a sample using two sterile syringes. Discard 5 mL with the first syringe, and take 6 mL of the suspension with the second. Determine the viable cell density, viability (see Note 4), and metabolite concentrations (see Note 5). Freeze 2 2 mL in tubes for the product (IgG) analysis (see Note 6). 7. Take samples in the same manner every 12 h and before and after feeding. 8. Feed the cells from day 3 to day 9 post-inoculation with feed media (see Note 8). Should the glucose concentration remain under 4 g L1 after feeding, add glucose from the 450 g L1 solution to reach 4 g L1. 9. After the last feeding of the cells on day 9 post-inoculation, samples are taken and analyzed until a viability of F) should fall below the statistical significance level (α ¼ 0.05).

244

Kim B. Kuchemu¨ller et al.

16. The Lack of Fit Tests table shows the compared residual error with pure error from replicated design points. The model should not be used as a response predictor, if there is a significant lack of fit. 17. Check the Model Summary Statistics for standard deviation, which should be low; the R2, which should be high; and the PRESS, which also should be low. 18. The regression model recommended by the program or preferred by user could be selected in the tab Model. There are some automatic reduction algorithms like Forward, Backward, and Stepwise. 19. Click the ANOVA tab to confirm the adequacy of the chosen model by significance of the model and individual term probability (Prob > F). In our approach, probability values greater than 0.10 could be deleted. Next click to Fit Statistics to see various statistics, whereby an R2 near 1 is good. 20. The best way to capture the diagnostic details is to view the chart available on the tab Diagnostics. In the Normal Plot, the data points should be approximately linear. A non-normality in the error term could be corrected by a transformation. Points that are far away from the line could be marked and checked in Residuals vs. Run. 21. In Model Graphs the Contour plot represents the factors in graduate color shading, and the 3D Surface Plot shows the variety of the responses as a function of the chosen factors. Click under View at Show Crosshairs Window to see data points. 3.6 Experimental Design Optimization

In total, the experimental design and the data are evaluated and controlled by the evaluation of the experimental designs (see Subheading 3.5). With this the optimization of the process is executable. The optimization of the experimental design is necessary to determine the optimal ranges of the sensitive parameters. 1. Therefore, choose the desired Goal in the Optimization node, which could apply to the response, the factor, or both response and factor, for each factor and response from the menu. The criteria, which could be chosen, are shown in Table 5 (see Note 22). l The shape of its particular desirability function could be adjusted by choosing a Weight assigned to each goal. The Importance of each goal can also be changed in relation to the other goals. Normally, a setting of three pluses is chosen (corresponds to a weighting). Increase the importance during changing it to five pluses. 2. The goals are combined into an overall desirability function D with the aim to maximize the optimization problem. Using

Model-Assisted Design of Experiments

245

Table 5 Criteria for desired goals Desired goal

Impact

Maximize

The upper limit is the desired result, while the lower limit is the lowest acceptable result

Minimize

The lower limit is the desired result, while the upper limit is the highest acceptable result

Target

The best result is between acceptable limits, and there is only one best result

In range

Specifies a range for acceptable results

None

Excludes the response model for optimization

Process capability index

Brings specifications into the optimization. It calculates the number of standard errors, in which the predicted response is within the specification limits

Equal to

Sets the factor equal to a single value to limit the optimization search

the steepest slope, the program chooses different random starting points to find the best local maximum. 1

1

D ¼ ðd 1 ∙d 2 ∙ . . . ∙d n Þn ¼ ∏ni¼1 d i n

where di is the individual desirability for each response, D is the objective function, and n the number of factors and responses [21]. 3. Cubes are calculated that represent the desirability at different positions of the experimental design, taking into account the previously defined goals. One is exemplary shown in Fig. 2b. Three factors, the ranges of the specific factors, and the desirability of the combinations are depicted. 4. With multiple responses, regions in which the requirements simultaneously meet the critical properties are sought. The best compromise by overlaying critical response contours can be found on Contour plot and at the 3D Surface (see Fig. 2c, d ). Regions that do not fit the optimization criteria in the contour plot are shaded gray. 5. Based on the optimization process, an optimal range can be selected, and a new experimental design can be created. With this the experimental space can be subsequently evaluated, and the experimental boundaries are narrowed down. Then, a few experiments could be performed in laboratory to evaluate the reduced experimental space (see Note 23).

246

4

Kim B. Kuchemu¨ller et al.

Notes 1. A factor is a potentially influencing variable, while a confounding factor is an external factor with an undesirable effect on the controlled system. The responses are variable measures, for which changes based on the factors need to be evaluated. The range of the factors depends on knowledge and potential compounds of, e.g., medium. 2. The mathematical description of a medium containing serum is difficult because of diverse serum charges and their undefined composition. 3. Further information are provided in [11]. 4. Concentrations of glucose, glutamine, and lactate can be measured with YSI 2900 (Xylem Inc., USA). Ammonia can be determined with an enzymatic kit, such as ammonia UV method from NZYTech (NZYTech, LDA. Genes and Enzymes, Portugal) by photometric method with measurements at a wavelength of 340 nm. Antibody concentration could be determined using an ELISA or preferably with bio-layer interferometry (Octet RED; Pall ForteBio, USA). 5. Software examples for solving ordinary differential equations are, e.g., the programs MATLAB (The MathWorks Inc., USA) or R (R Core Team, New Zealand). 6. Statistical software, in which the planning and evaluation of experimental designs are possible, are, e.g., the programs DesignExpert (Stat-Ease, Inc., USA), JMP (SAS Institute, USA), R (R Core Team, New Zealand), or Modde (Sartorius Stedim biotech, Sweden). 7. First data should be generated based on prior knowledge, literature, or first test experiments, e.g., for medium tests. The evaluated data should be used to cover a wide range of requirements, e.g., inhibitions or limitations. However, only few experiments should be performed in small scale, such as shaking flasks or deep well plates. Sometimes, it is useful to harness a traditional screening design to detect unknown influencing variables. In total, the data should be used to model the growth, metabolism, and productivity. 8. An example of an algorithm for solving differential equations is the fourth order Runge-Kutta algorithm with variable step size [13]. The algorithm is implemented, e.g., in MATLAB (The MathWorks Inc., USA) as well as in R (R Core Team, New Zealand). 9. An example for an algorithm to identify optimal model parameters is the simplex algorithm by Nelder and Mead. Based on given starting values, n-points in an n-dimensional parameter

Model-Assisted Design of Experiments

247

space, called simplex, are calculated. Starting from a start simplex S0, a sequence Sk of simplexes is constructed, which move in the permissible quantity in the direction of a local solution x∗ and become smaller in the environment of x∗ [21]. 10. The used cultivation data consist of offline data. 11. The values can be determined during experiments or through research in literature. For example, Monod kinetic constant for glucose uptake will be determined during glucose-limited phase of an experiment [22]. 12. Examples for performance functionals are the coefficient of determination or the root-mean-square error [23, 24]. 13. The choice of a weighting factor ensures that the individual simulations are adapted to the data. Relative to the focus, the weighting can be set, e.g., to the living cell number and the antibody concentration, so that these are weighted with 100, while the other concentrations are weighted with 1. 14. For more information, see Chapter 12, this edition [25]. 15. Focus on the area, which should be optimally displayed. For example, the presented model can be used for adaptation up to the transition into the stationary phase, while the stationary and the dying phase cannot be represented sufficiently. However, the goal is a high cell count; therefore, the representation of the dying phase is not important. This must be taken into account during the evaluation. 16. Screening designs will be used for determination of factors, which influence the process. The application of screening designs in a biosystem is facilitated by the good definition of the cultivation components. Limitations occur due to the dependence on the knowledge of the experimenter, who determines the range of factors to be investigated [26]. 17. Higher regression models than quadratic could lead to an overfitting. Therefore, start with quadratic regression model and only change it if model significance is not given [3]. 18. Since not all experiments are performed in the laboratory, it is not necessary to reduce the points or decrease the amount of blocks to reduce the costs or to make the amount of experiments tradable. 19. A set of design points will be chosen by the program and exchanged repetitive until achieving a set that meets the specifications established in the model selection screen. 20. The recommended factor combinations and the simulated responses can be exported into a spreadsheet, such as Microsoft Excel (Microsoft Corporation, USA).

248

Kim B. Kuchemu¨ller et al.

21. More information of statistical design evaluation could be found in the DesignExpert (Stat-Ease, Inc., USA) manual. 22. The constraints can be based on knowledge and literature research. For example, the antibody concentration and the viable cell density should be maximized as a high product titer is the target. The constraints for the metabolites should include the cell growth and product quality as they may inhibit or limit the process. 23. Based on the chosen constraints, an area describing the desirability can be prepared. Only the area, in which the desirability function is greater than zero, would increase the process understanding. Therefore, use the suggested area to create a new experimental design and narrow your experimental boundaries. The new experimental design could then be performed to evaluate the mDoE. References 1. Mo¨ller J, Po¨rtner R (2017) Model-based design of process strategies for cell culture bioprocesses: state of the art and new perspectives. In: Gowder S (ed) New insights in cell culture technology. InTech, London. ISBN: 978-95351-3133-5 2. Sandadi S, Ensari S, Kearns B (2006) Application of fractional factorial designs to screen active factors for antibody production by Chinese hamster ovary cells. Biotechnol Prog 22:595–600. https://doi.org/10.1021/ bp050300q 3. Kleppmann W (2013) Versuchsplanung. Produkte und Prozesse optimieren, 8th edn. Hanser, Mu¨nchen, Wien. ISBN: 978-3446437524 4. Zhang H, Wang H, Liu M, Zhang T, Zhang J, Wang X, Xiang W (2013) Rational development of a serum-free medium and fed-batch process for a GS-CHO cell line expressing recombinant antibody. Cytotechnology 65:363–378. https://doi.org/10.1007/ s10616-012-9488-4 5. Nasri NMR, Razavi SH (2010) Use of response surface methodology in a fed-batch process for optimization of tricarboxylic acid cycle intermediates to achieve high levels of canthaxanthin from Dietzia natronolimnaea HS-1. J Biosci Bioeng 109:361–368. https://doi.org/ 10.1016/j.jbiosc.2009.10.013 6. Siebertz K, Bebber DV, Hochkirchen T (2010) Statistische Versuchsplanung. In: Design of Experiments (DoE). Springer, Heidelberg, ISBN: 978-3-642-05493-8 7. Mandenius C, Graumann K, Schultz T, Premstaller A, Olsson M, Petiot E,

Clemens C, Welin M (2009) Quality-by-design for biotechnology-related pharmaceuticals. Biotechnol J 4:600–609. https://doi.org/10. 1002/biot.200800333 8. Mo¨ller J, Kuchemu¨ller KB, Herna´ndez Rodrı´guez T, Frahm B, Hass VC, Po¨rtner R (2018) Model-assisted design of process strategies for cell culture processes. Am Pharm Rev 21(3):39–41 9. Mo¨ller J, Kuchemu¨ller K, Po¨rtner R (2018) Model-assisted DoE – A concept study for cell culture process development. Chemie Ingenieur Technik 90(9):1235–1235. https://doi. org/10.1002/cite.201855228 10. Abt V, Barz T, Cruz N, Herwig C, Kroll P, Mo¨ller J, Po¨rtner R, Schenkendorf R (2018) Model-based tools for optimal experiments in bioprocess engineering. Curr Opin Chem Eng 22:244–252. https://doi.org/10.1016/j. coche.2018.11.007 11. Mo¨ller J, Kuchemu¨ller KB, Steinmetz T, Koopmann KS, Po¨rtner R (2019) Model-assisted design of experiments as a concept for knowledge-based bioprocess development. Bioprocess Biosyst Eng 42(5):867–882. https://doi.org/10.1007/s00449-01902089-7 12. Kern S, Platas-Barradas O, Po¨rtner R, Frahm B (2014) Model-based strategy for cell culture seed train layout verified at lab scale. Cytotechnology 68(4):1019–1032. https://doi.org/ 10.1007/s10616-015-9858-9 13. Frahm B, Lane P, M€arkl H, Po¨rtner R (2003) Improvement of a mammalian cell culture process by adaptive, model-based dialysis

Model-Assisted Design of Experiments fed-batch cultivation and suppression of apoptosis. Bioprocess Biosyst Eng 26:1–10. https://doi.org/10.1007/s00449-003-0335z 14. Mohler L, Bock A, Reichl U (2008) Segregated mathematical model for growth of anchoragedependent MDCK cells in microcarrier culture. Biotechnol Prog 24:110–119. https://doi. org/10.1021/bp0701923 15. Ling WLW, Bai Y, Cheng C, Padawer I, Wu C (2015) Development and manufacturability assessment of chemically-defined medium for the production of protein therapeutics in CHO cells. Biotechnol Prog 31:1163–1171. https:// doi.org/10.1002/btpr.2108 16. Liu C, Chang T (2006) Rational development of serum-free medium for Chinese hamster ovary cells. Process Biochem 41:2314–2319. https://doi.org/10.1016/j.procbio.2006.06. 008 17. Torkashvand F, Vaziri B, Maleknia S, Heydari A, Vossoughi M, Davami F, Mahboudi F (2015) Designed amino acid feed in improvement of production and quality targets of a therapeutic monoclonal antibody. PLoS One 10:e0140597. https://doi.org/10. 1371/journal.pone.0140597 18. Castro PML, Hayter PM, Ison AP, Bull AT (1992) Application of a statistical design to the optimization of culture medium for recombinant interferon-gamma production by Chinese hamster ovary cells. Appl Microbiol Biotechnol 38(1):84–90. https://doi.org/10. 1007/BF00169424 19. Chun C, Heineken K, Szeto D, Ryll T, Chamow S, Chung JD (2003) Application of factorial design to accelerate identification of CHO growth factor requirements of CHO growth factor requirements. Biotechnol Prog 19:52–57. https://doi.org/10.1021/ bp025575

249

20. Liu C, Chu I, Hwang S (2001) Factorial designs combined with the steepest ascent method to optimize serum-free media for CHO cells. Enzym Microb Technol 28:314–321. https://doi.org/10.1016/ S0141-0229(00)00346-X 21. Alt W (2011) Nichtlineare Optimierung, Eine Einfu¨hrung in Theorie, Verfahren und Anwendungen, 2nd edn. Vieweg+Teubner Verlag, Wiesbaden. ISBN: 978-3-8348-1558-3 22. Frahm B, Lane P, Atzert H, Munack A, Hoffmann M, Hass VC, Po¨rtner R (2002) Adaptive, Model-Based Control by the OpenLoop-Feedback-Optimal (OLFO) controller for the effective fed-batch cultivation of hybridoma cells. Biotechnol Prog 18:1095–1103. https://doi.org/10.1021/ bp020035y 23. Cameron A, Windmeijer F (1997) An R-squared measure of goodness of fit for some common nonlinear regression models. J Econ 77:329–342. https://doi.org/10.1016/ S0304-4076(96)01818-0 24. McHugh MJ (2005) Multi-model trends in East African rainfall associated with increased CO2. Geophys Res Lett 32:2068. https://doi. org/10.1029/2004GL021632 25. Deppe S, Kuchemu¨ller KB, Herna´ndez Rodrı´guez T, Po¨rtner R, Mo¨ller J, Frahm B (2019) Estimation of process model parameters. In: Po¨rtner R (ed) Methods in molecular biology – animal cell biotechnology, 4th edn. Springer, New York 26. Mandenius CF, Brundin A (2008) Bioprocess optimization using design-of-experiments methodology. Biotechnol Prog 24:1191–1203. https://doi.org/10.1002/ btpr.67

Chapter 14 Design, Optimization, and Adaptive Control of Cell Culture Seed Trains Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm Abstract For the production of biopharmaceuticals, a procedure called seed train or inoculum train is required to generate an adequate number of cells for the inoculation of the production bioreactor. This seed train is time- and cost-intensive but offers potential for optimization. A method and a protocol are described for seed train mapping, directed modeling, and simulation as well as its optimization regarding selected optimization criteria such as optimal points in time for cell passaging. Furthermore, the method can also be applied for the transfer of a seed train to a different production plant or the design of a new seed train, for example, for a new cell line. Another application is to support the selection of the optimal clone for a new process. Seed train prediction can be performed for different clones, and so it can be analyzed how the seed train protocol would look like and for which clones a suitable seed train protocol could be found. Although the chapter is directed toward suspension cell lines, the method is also generally applicable, e.g., for adherent cell lines. Key words Seed train, Inoculum train, Optimization, Modeling, Simulation, Prediction, Computational biotechnology, Cell culture, Animal cells, Production

1

Introduction The production of biopharmaceuticals for diagnostic and therapeutic applications based on suspension cell culture in bioreactor scales from a few hundred liters up to 20 m3 is state of the art. The generation of an adequate number of cells for the inoculation of a production bioreactor is time- and cost-intensive. From volumes used for cell thawing or cell line maintenance, the cell number has to be increased while passaging usually into larger cultivation systems. Examples are T-flasks, (disposable) shake flasks, spinner flasks, falcon tubes, roller bottles or bench-top bioreactor systems (e.g., stirred tanks, wave motion bioreactors, or orbital shaken bioreactors), and subsequently larger bioreactors. The production bioreactor is inoculated out of the largest seed train scale. Figure 1 illustrates a seed train example.

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_14, © Springer Science+Business Media, LLC, part of Springer Nature 2020

251

252

Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

Fig. 1 Seed train example from a frozen vial into a 10,000 L bioreactor using shake flasks for 5–200 mL scale, a wave cell bag for 1 L scale and bioreactors for 5–10,000 L scale

More and more, disposable technology is also applied for larger (bioreactor) scales of the seed train [1]. In order to reduce the number of passages within the seed train, also bioreactors are used which can be inoculated at volumes that are very small in relation to the maximum filling volume, e.g., around 17% [2, 3]. Afterward, culture volume is increased in this so-called inoculation bioreactor by medium addition. Therefore, such an inoculation bioreactor can replace two conventional bioreactors in a seed train. However, seed trains using disposable technology or “inoculation” bioreactors are also subject to optimization. In addition to familiar requirements of seed trains, it can be determined when medium is supplemented at which volume and if concentrated feed should be preferred over medium. Regardless of different approaches, a cell culture seed train lasts for a significant period of time and generates corresponding costs. For example, a seed train from cell thawing until inoculation of a 3,000 L scale requires around 3 weeks. Delays caused by unusual low growth rates, contamination of a scale, etc. can further increase this time span. Moreover, deviation from standard growth rates can enforce the personnel to adapt the typically used seed train protocol. There are numerous parameters determining a seed train. Examples are apparent (effective) growth rate and viability, selection of vessel and filling volumes of the seed train scales, differences in bioprocess engineering parameters between scales, inoculation cell densities, ratio of fresh medium to passaged medium, substrate and metabolite concentrations, point in time for cell passaging, and corresponding viable cell density. Therefore, the seed train offers space for optimization, for example, via the choice of the optimal points in time for passaging from one scale into the larger one or via the choice of inoculation density and culture volume at inoculation or, if possible (e.g., when

Seed Train Optimization

253

different bioreactor volumes are available), via the choice of bioreactor volumes. The first seed train steps are often inhomogeneous due to different applied culture systems (e.g., T-flasks, shake flasks, spinner flasks, falcon tubes, roller bottles, small scale single-use bioreactors), due to a different seed train protocol for these culture systems and the transitions between them in comparison to the larger bioreactor scales. For the later seed train steps (in bioreactor scales), scale-up steps and seed train protocol are often similar, but usually not yet identical since the same scale-up factor between all bioreactor scales has not been realized during the design of the cell culture plant. The experimental effort to test all the seed train steps is correspondingly high. At the same time, it is known that all cultivation steps, also the early stages during a seed train, have a significant impact on cell performance in production scale [4]. The method presented in this chapter can be applied for the analysis and optimization of existing seed trains as well as for the transfer of a seed train to a different production plant. It is possible to investigate to what extent existing bioreactor scales may be used for the seed train or which modifications, e.g., regarding filling volume or additional scales, are necessary. Furthermore, the method can be applied to support the selection of the optimal clone for a new process. Seed train simulation can be performed for different clones, and so it can be analyzed how the seed train protocol would look like and for which clones a suitable seed train protocol could be found. Another application is the design and the protocol development of a new seed train, for example, for a new cell line or for a new cell culture plant. Moreover, the method can be applied for ongoing seed trains, enabling seed train monitoring, control, and optimization. This chapter presents a method and the corresponding protocol to model, simulate, and optimize a seed train. Figure 2 presents the scheme of the method. The method is divided into two main parts, (a) adaption to cell line and (b) seed train. The adaption to a cell line is realized by a model building process resulting in a mathematical model including specific model parameters adapted to the cell line and the applied cultivation data. This means that, first of all, a model has to be chosen or created, and cultivation data have to be added in order to perform a priori model parameter identification (the terms parameter identification and parameter estimation are used in this work interchangeably). Depending on the evaluation of the results (e.g., identifiability of model parameters), it may be the case that the model has to be adapted (e.g., adaption of model equation terms). Depending on the data base, the adaption is performed using more than one cultivation data set.

254

Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

Fig. 2 Scheme of the method for seed train simulation and optimization, divided into two blocks, (a) adaption to cell line and (b) seed train. Input information is shown in form of rectangles; calculations are shown in form of ovals. Part (b) contains two loops leading to different applications (described in the two parallelograms)

After finishing the adaption to the cell line and to the corresponding cultivation data, the model including the identified model parameters is applied to the simulation of the first seed train scale. Additionally, the initial concentrations of the variables included in the model (e.g., viable and total cell density, important substrate and metabolite concentrations when starting the seed train) are required as inputs for the simulation of the first seed train scale. In order to simulate the temporal courses of the next seed train scales, the following inputs are required: Passaging strategy, seed train vessels, and conditions (e.g., minimum viable inoculation cell density, minimum filling volumes, working volumes) as well as medium concentrations (relevant substrate concentrations which are included in the model). Therewith, the point in time for passaging and initial concentrations of the next scale can be computed. Repeating this procedure for every scale (repeating loop 1) leads to the simulation of the whole seed train. In practice, this method can be applied for seed train analysis, optimization of existing seed trains, seed train transfer to another production plant or to a new cell line or product, clone selection, as well as protocol development (design of a new seed train) (as listed in the upper parallelogram in Fig. 2b). Furthermore, seed train simulation and prediction can be applied in form of monitoring an ongoing seed train. Therefore, a second loop has to be performed in parallel during a real ongoing seed train integrating current measured data. This means, while or after running the cultivation in the first scale, measured cultivation

Seed Train Optimization

255

data are used for identification of model parameters and their adaptation, if necessary (performing parameter estimation based on the collected data in Fig. 2a), and for computation of starting concentrations of the next scale. This way, the prediction of the future (the next scale(s)) includes information from the past (current scale). Seed train prediction performing this loop (loop 2) in parallel to the first loop (loop 1) allows, among others, adaptive seed train control (e.g., adaptation of the current seed train, handling to unusual cell growth) or, if feed is applied during the seed train, feed optimization (see second parallelogram in Fig. 2b).

2

Materials 1. The cell line of interest and an appropriate medium, preferred a chemically defined, serum-free medium. 2. A batch or fed-batch cultivation system. 3. Hemocytometer or an automated system such as the CEDEX (Roche, Germany) for determination of total and viable cell concentration. 4. Analyzer for determination of substrate and metabolite concentrations such as glucose, glutamine, lactate, and ammonia concentration (e.g., CuBiAn (OPTOCELL technology, Germany) or YSI Biochemistry Analyzer (Xylem Inc., USA) or equivalent) (see Note 1).

2.1

Software

1. A computer program and a corresponding software that is able to: (a) Calculate courses of an underlying model that may contain differential equations. An example of an algorithm for solving differential equations is the pair of embedded explicit Runge-Kutta methods of orders 2 and 3 with error control and variable step size by Bogacki and Shampine [5] (see Note 2). (b) Adjust model parameters in order to find the best fit of modeled courses to cultivation data. An example of an algorithm for parameter identification is the simplex algorithm by Nelder and Mead [6, 7] (see Note 3). (c) Calculate a complete seed train including the passaging from one scale to another based on the input of model parameters for the used cell line, medium information, initial concentrations (starting state) of the seed train as well as volumes, and inoculation cell densities, of course, based on the model itself.

256

Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

(d) Perform calculations regarding optimization criteria. This is described in more detail in Subheading 3.5. (e) Determine corresponding optima. A software example for implementation is the program MATLAB® [8].

3

Methods

3.1 Seed Train Analysis

1. Gather all relevant information concerning your seed train from thawing to inoculation of the production bioreactor or the specific part of the seed train that you like to optimize. 2. Document your seed train from thawing to inoculation of the production bioreactor or the specific part of the seed train that you like to optimize. Take down: (a) Inoculation cell densities and volumes. (b) Cell densities, viabilities, apparent (effective) growth rates, and volumes when passaging into the next scale. (c) Volume ratios of fresh medium to used medium containing cells (this information also results from the two previous points). (d) The initial concentrations (starting state) of the seed train—viable cell density, viability, volume, and concentrations of all important substrates and metabolites. (e) Applied medium including supplements and the corresponding concentrations of important substrates. 3. Analyze your seed train concerning restraints—do limitations occur, e.g., glucose, glutamine, or oxygen, or are inhibiting concentrations reached, e.g., regarding ammonia, lactate, and carbon dioxide? 4. If yes, are restricting circumstances inevitable or can, for example, medium composition or culture conditions such as power input, shear stress, and gas transfer rates be modified? Are there deviations of maximum apparent (effective) growth rate between different cultivation systems (see Note 4)? 5. What is your current indicator that triggers the cell passaging from one scale to the next? Is this, for example, an experimentally determined time span or is, for example, viable cell density monitored and used to passage cells, e.g., before the end of the log phase? 6. Which parameters are important for monitoring the seed train (see Note 5)?

Seed Train Optimization

3.2 Cell Line Cultivation Data

3.3 Modeling: Model Selection, Adaption, and Model Parameter Identification

257

1. Select reliable cultivation data for the used cell line. If not available (e.g., in case of a new cell line), perform a few experiments to create sufficient data. If cell line behavior differs, e.g., between static scales, e.g., T-flasks and shaken/stirred scales, more experiments are needed to describe the differences. Usually, an extensive cell line characterization is not necessary. A few directed batches using different initial concentrations of limiting and inhibiting substances are adequate so that different limiting and inhibiting conditions are generated separately from each other. Another possibility is a continuous cultivation (chemostat) with different states, but in this case the applied model equations have to cover also continuous cultivation. The goal here is to identify a model which enables the calculation of time courses for relevant process variables. For this, a kinetic model (also called “mechanistic model”) is required, and model parameters have to be adjusted in order to find the best fit of modeled courses to cultivation data, according to Fig. 2a). 1. Select a model from literature or set up your own model. An example is given in Tables 1 and 2 or see [9–11]. The model has to include all circumstances from Subheading 3.1, e.g., important parameters for monitoring the seed train such as cell growth and cell death as well as limiting and inhibiting substances. Furthermore, the model should be able to incorporate medium addition or feed addition if performed during the seed train or if considered for the future. 2. Program the model into suitable software (see Note 6). 3. Enter the cultivation data of the first cultivation. 4. Select a starting set of model parameters for your model that could be realistic. Since the optimal model parameters will be identified later on, rough values are sufficient. 5. Select boundary values for each model parameter just as a safety measure to impede the algorithm from selecting unrealistic parameter values during model parameter identification. 6. Model parameters that cannot be identified by fitting a certain cultivation should be fixed. Examples are model parameters for glutamine uptake at low glutamine concentrations and model parameters for the influence of glutamine concentration on growth rate at low glutamine concentrations. These model parameters cannot be identified during a glucose limited batch cultivation, for example, which does not exhibit low glutamine concentrations. An exception is a situation, where the algorithm leaves model parameters that cannot be identified, more or less at their starting values.

dV dt

¼ F Glc þ F Gln þ F Medium

¼  F Glc þF GlnVþF Medium  c LS  q LS  X v

qMAb ¼ qMAb, max

LS q LS ¼ q LS, max  c LScþk LS

qAmm ¼ YAmm/Gln  qGln

¼  F Glc þF GlnVþF Medium  c MAb þ q MAb  X v

dc MAb dt

dc LS dt

μ q Lac ¼ Y Lac=Glc  q Glc  q Lac,uptake, max  μmax μmax

¼  F Glc þF GlnVþF Medium  c Lac þ q Lac  X v

dc Lac dt

q Gln ¼ K p,Gln  q Gln, max  cGlc > cGlc,thr: Kp,Gln ¼ 1 cGlc  cGlc,thr: Kp,Gln ¼ 1 + a  (cGlc, thr  cGlc) c Gln c Gln þkGln

¼  F Glc þF GlnVþF Medium  c Amm þ q Amm  X v

¼ FVGln  c Gln,F þ F Medium  c Gln,Medium  F Glc þF GlnVþF Medium  c Gln  q Gln  X v V

dc Gln dt

Substrate uptake/metabolite production   Glc q Glc ¼ q Glc, max  c Glccþk  μþμμ þ 0, 5 Glc

dc Amm dt

¼ FVGlc  c Glc,F þ F Medium  c Glc,Medium  F Glc þF GlnVþF Medium  c Glc  q Glc  X v V

dc Glc dt

c LS c LS þK S,LS

max

μd ¼ μd,min

¼ μ  X v  F Glc þF GlnVþF Medium  X t  K Lys  ðX t  X v Þ

dX t dt

Liquid phase

c Glc c Gln μ ¼ μmax  c Glc þK  c Gln þK  S,Glc S,Gln

¼ ðμ  μd  F Glc þF GlnVþF Medium Þ  X v

Cell growth/death

Bio phase

dX v dt

Kinetics

Balances

Table 1 Equations of a model example

258 Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

Seed Train Optimization

259

Table 2 Abbreviations used in model example equations Parameter

Units

Starting value 1

a

Parameter which describes increased glutamine uptake at low glucose concentrations

[L mmol

cAmm

(total) ammonia concentration

[mmol L1]

cGlc

Glucose concentration

[mmol L1]

cGlc,F

Glucose concentration in glucose/nutrient concentrate feed

[mmol L1]

cGlc,Medium

Glucose concentration in medium

[mmol L1]

cGlc,thr

Threshold value of glucose concentration for increased glutamine-uptake at low glucose concentrations

[mmol L1]

cGln

L-glutamine concentration

[mmol L1]

cGln,F

L-glutamine concentration in glutamine concentrate feed

[mmol L1]

cGln,Medium

L-glutamine concentration in medium

[mmol L1]

cLac

Lactate concentration

[mmol L1]

cLS

Concentration of limiting substrate

[mmol L1]

cMab

Concentration of monoclonal antibodies

[mg L1]

FGlc

Feed rate of glucose/nutrient concentrate

[L h1]

FGln

Feed rate of glutamine concentrate

[L h1]

FMedium

Feed rate of medium

[L h1]

kGlc

Monod kinetic-constant for glucose uptake

[mmol L1]

kGln

Monod kinetic-constant for glutamine uptake

[mmol L

]

1 1

6 (usually fixed)

0.5 (usually fixed)

0.19

]

0.3

]

0.1

kLS

Monod kinetic-constant for the uptake of limiting substrate

[mmol L

KLys

Cell lysis constant

[h1]

0.01

Kp,Gln

Correction factor (considers dependence of glutamine uptake rate on low glucose concentrations)

[]

1 (usually fixed)

KS,Glc

Monod kinetic-constant for glucose

[mmol L1] 1

0.03

KS,Gln

Monod kinetic-constant for glutamine

[mmol L

]

0.03

KS,LS

Monod kinetic-constant for limiting substrate

[mmol L1]

0.01 (continued)

260

Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

Table 2 (continued) Parameter

Units

qAmm

Cell-specific ammonia production rate

[mmol cell1 h1]

qGlc

Cell-specific glucose uptake rate

[mmol cell1 h1]

qGlc,max

Maximum cell-specific glucose uptake rate

[mmol cell1 h1]

qGln

Cell-specific glutamine uptake rate

[mmol cell1 h1]

qGln,max

Maximum cell-specific glutamine uptake rate

[mmol cell1 h1]

qLac

Cell-specific lactate production rate

[mmol cell1 h1]

Starting value

0.45

0.085

qLac,uptake,max Maximum cell-specific lactate uptake rate

[mmol cell1 h1]

qLS

Cell-specific uptake rate of limiting substrate

[mmol cell1 h1]

qLS,max

Maximum cell-specific uptake rate of limiting substrate

[mmol cell 1 1 h ]

qMab

Cell-specific antibody production rate

[mg cell1 h1]

qMAb,max

Maximum cell-specific antibody production rate

[mg cell1 h1]

t

Time

[h]

V

Culture volume

[L]

Xt

Total cell concentration

[cells L1]

Xv

Viable cell concentration

[cells L1]

YAmm/Gln

Kinetic-production-constant (Stoichiometric ratio of ammonia production and glutamine uptake)

[]

0.4

YLac/Glc

Kinetic-production-constant (Stoichiometric ratio of lactate production and glucose uptake)

[]

1.6

μd

Cell-specific death rate

[h1]

μd,min

Minimum cell-specific death rate

[h1]

μ μmax

Cell-specific growth rate Maximum cell-specific growth rate

[h [h

0.4

0.014

0.5

0.002

1

]

1

]

0.055

Seed Train Optimization

261

7. Select or define a performance function that compares the modeled courses to the cultivation data, e.g., least squares. An example is given in [10]. 8. Start the model parameter identification algorithm, e.g., the simplex algorithm by Nelder and Mead, so that the optimal model parameters are identified based on their starting values and the applied cultivation data set. 9. Evaluate parameter identification results performing the following techniques: (a) Calculate temporal courses based on the identified optimal model parameters. Judge the quality of the model fit by plotting modeled courses and cultivation data and by calculation of a goodness of fit score like the normalized root-mean-squared deviation (NRMSD). (b) Judge the identified model parameters, for example, performing an identifiability analysis [12] and/or a sensitivity analysis [13]. (c) Plot and judge the courses of calculated rates of the model that change over time (e.g., the cell growth rate depending on substrate and metabolite concentrations). (d) If necessary, adjust the model, model parameter starting values, and/or model parameter boundaries. 10. Repeat steps 3–9 for all cultivations that you would like to use for model parameter identification. The obtained sets of identified model parameters may be used to define a starting parameter set for further identifications in order to have improved starting values. 11. Evaluate the results of the different parameter identification runs and deduce either (a) an optimal model parameter vector (if possible) or (b) a set of parameter vectors in order to perform simulations for every parameter vector (see Note 7). An optimal set could be generated by combining the identified model parameter values from the different runs, e.g., by averaging model parameter values identified by fitting different cultivations and adding model parameter values only identified by one fit (that have been at fixed values during the other fits). 3.4 Seed Train Modeling and Simulation

The goal here is to calculate (simulate) a complete seed train including the passaging from one scale to another within a software environment/program according to the structure, illustrated in Fig. 2b. The step-by-step approach is explained in this chapter. In terms of program implementation, it should be noted that input

262

Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

information (as indicated by rectangles) has to be defined or selected by the user and integrated in the programming via appropriate facilities. Calculation functions (as indicated by ovals) have to be implemented in the program and are executed in the background during simulation of the seed train (see Note 8). Subsequently, the workflow for seed train modeling and simulation is presented, which includes (after performing the first step in Fig. 2a) two loops, the first loop describing seed train simulation of performed or new seed trains without feedback (loop 1) and a second loop describing seed train simulation of ongoing seed trains including prediction and integration of feedback (loop 2). 1. In addition to a cell culture model and identified model parameters for the used cell line (compare to Fig. 2a, parameters obtained in Subheading 3.3), the following inputs have to be defined: initial concentrations (starting state) of the seed train, meaning the initial state of the first scale such as initial volume, inoculation cell density, and initial substrate and metabolite concentrations. 2. Loop 1: Using these inputs, the temporal courses of the variables described by the model can be simulated/predicted for a user-defined simulation time span in the current scale. 3. Next, the cell passaging from the current scale into the next scale has to be modeled. Therefore, a passaging strategy has to be selected. This could be, for example, a fixed time span for cell growth, the achievement of a certain viable cell density or other criteria, which will be described in Subheading 3.5, step 1. Moreover, seed train vessels and conditions have to be defined, comprising at least the number of scales, corresponding working volumes, and minimum filling volumes as well as lower and upper limits for inoculation cell density. Concerning the medium, relevant substrate concentrations are needed. Based on the simulated temporal courses (see step 2) and the added inputs described above, the point in time for cell passaging as well as the initial concentrations of the next scale can be calculated. 4. In case that the specifications defined for the seed train cannot be fulfilled, the simulation has to be stopped and specifications have to be changed. This could happen, for example, if at the end of the current scale, there are not enough cells for inoculating the next scale at the required minimum filling volume. In this case, a smaller scale is required. In case of having too many cells for the inoculation of the next scale (i.e., the inoculation cell density would be higher than the allowed upper limit), the volume of the inoculum (cultivation broth from the current scale) could be reduced, or it could be tested if the planned next scale could be jumped over.

Seed Train Optimization

263

5. Repeating steps 2–4 of this section for every seed train scale leads to the simulation/prediction of the whole seed train according to Fig. 2b. The initial state of each cultivation scale except for the first scale results from the state of the cells in their corresponding cultivation broth when passaging from the last scale. It also results from the applied inoculation cell density of the current scale and the amount of medium used to achieve this cell density. Moreover, it results from the medium composition regarding substrates (and metabolites, if glutamine, for example, may be partially degraded to ammonia). 6. Loop 2: Optionally, a second loop can be performed in parallel, integrating current measured data. This means, after running the cultivation in the current scale (step 2), measured cultivation data are used for identification of model parameters and their adaptation, if necessary (performing parameter estimation based on the collected data). This way, the prediction of the future (the next scale(s)) includes information from the past (current and eventually preceding scales). Seed train prediction performing this loop (loop 2) in parallel to the first loop (loop 1, steps 2 and 3) can be performed, if seed train prediction should be applied in form of monitoring a running seed train. This allows, among others, seed train monitoring and control (e.g., adaptation of the current seed train, handling to unusual cell growth), optimization of prediction, and, if feed is applied during the seed train, feed optimization. 7. For model validation, the complete simulated seed train can be compared to the real (performed) seed train. This step allows further alignment in addition to the cell culture modeling consisting of model parameter identification and model optimization (Subheading 3.3). 3.5 Seed Train Optimization and Further Applications

The seed train modeling and simulation performed in the previous Subheading 3.4 can be used within different contexts and following diverse goals. Loop 1 can be applied for seed train simulation of performed or new seed trains without feedback. This workflow enables, among others, the following applications: 1. If cultivation data from small and large scales of performed seed trains are available, seed train simulation can be performed for the following purposes: (a) Seed train analysis: Seed train analysis provides process knowledge, which is important in order to ensure a robust process or for process optimization. (b) Seed train optimization: Existing seed trains could be optimized regarding specific optimization criteria through simulation.

264

Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

(c) Seed train transfer: This could be transfer of an existing seed train protocol to another production plant and eventually adaption to different vessel volumes. 2. If only small scale data are available and new circumstances are considered, seed train simulation can be performed for the following purposes: (a) Clone selection: For every clone, the seed train will be simulated according to the clone-specific characteristics (e.g., specific maximum growth rate, substrate uptake rates, and metabolite production rates). This way it is possible to test if the clone fits into the planned seed train protocol or if changes are required. Furthermore, the simulations show which clone should be preferred regarding a specific goal. (b) Protocol development (¼seed train design): If a protocol for a new clone or cell line should be developed, seed train simulation based on small scale data could be applied in form of a supporting tool (e.g., for the decision of the number of vessels, etc.). Loop 2 can be applied for seed train simulation of ongoing seed trains including prediction and integration of feedback. This workflow enables, among others, the following applications: (a) Seed train monitoring and control (open-loop, closedloop, adaptive): Through integration of current data from the ongoing seed train, possible changes in cell behavior can be considered through adaption of model parameters, leading to probably more precise and accurate predictions. The gained information can be used in form of decision support (e.g., changing the point in time for cell passaging or changing certain process parameters). (b) Seed train optimization: Starting at a specific point in time, seed train prediction for the remaining seed train provides necessary information for the application of optimization strategies (e.g., feed optimization) concerning a defined goal. An example for seed train optimization of existing seed trains (Loop 1 (I)) is the optimization regarding a selected cell passaging criterion, as described subsequently. 1. Define your optimization criterion: (a) An example of an optimization criterion is to screen for the latest point in time showing a high apparent (effective) cell growth rate. Like this, cells are taken directly out of the exponential growth phase. Usually, the apparent

Seed Train Optimization

265

(effective) growth rate (¼growth rate minus death rate) is at its maximum during the exponential growth phase until a substrate starts to limit or a metabolite starts to inhibit. Therefore, the apparent growth rate can be plotted over time, and the point in time when the apparent growth rate starts to decline, e.g., 90% of the maximum, can be selected. (b) A second example of an optimization criterion is to screen for the point in time when one of the nutrient substrates falls below a defined threshold value (e.g., when glucose falls below 1 mmol/L). This criterion ensures that cells are passaged before a substrate starts to limit. (c) Another example is the optimal space-time yield (STY) regarding viable cells (viable cells per filling volume and time) or produced viable cells (new viable cells per filling volume and time (¼viable cells minus initial number of viable cells at inoculation per filling volume and time)). Plot the STY over time for each scale and find its optimum (which is the maximum, omitting the values within the first hour). This optimum can be used for cell passaging (see Note 9). (d) A fourth example of an optimization criterion is to combine two mentioned criteria, e.g., determining the average of the latest point in time showing a high apparent growth rate (example 1) and of optimal STY (example 3). This averaging results in an optimization criterion that combines a high STY with a high apparent growth rate. 2. Apply your selected optimization criterion to the modeled seed train. Keep in mind that medium or feed addition and therefore their flow rates and the starting volume of a scale as fed-batch may also be parameters that can be open for optimization in order to optimize the seed train. Other potential parameters are the filling states in a batch seed train (not using the total filling volume) or the initial cell densities. 3. Perform a seed train according to your optimization. 4. Compare the optimized seed train to your previously used seed train. If necessary, investigate and adjust the applied optimization criterion.

4

Notes 1. Further analysis could include specific assays for the product as well as amino acid concentrations. 2. A software example is the program MATLAB. Algorithms for solving ordinary differential equations can be selected in such

266

Tanja Herna´ndez Rodrı´guez and Bjo¨rn Frahm

programs. Several algorithms are available, such as the already mentioned pair of embedded explicit Runge-Kutta methods of orders 2 and 3 with error control and variable step size by Bogacki and Shampine [5]. Other examples are variable-order (VSVO) solvers based on the numerical differentiation formulas (NDFs) [14] or the fourth order Runge-Kutta algorithm [6] with variable step size. 3. A code for the simplex algorithm by Nelder and Mead is also available for MATLAB®. 4. If yes, apply common process transfer criteria in order to reduce them [15]. 5. For example, viable and total cell density, glucose and glutamine concentration as substrate concentrations, lactate and ammonia concentration as potentially inhibiting factors, another limiting/inhibiting medium component, pH-value, or an unknown factor. 6. For support concerning seed train optimization or programming, the author can be contacted. 7. In case that the obtained model parameter vectors are showing high variation for some specific model parameters due to non-identifiability, the generation of a model parameter vector by averaging the parameter values from the different runs may lead to inaccurate simulation results. In this case, it could be more informative to bundle the obtained parameter vectors in order to perform simulations based on every parameter vector. This would lead to a set of simulations, expressing the possible outcomes including the degree of uncertainty. 8. If there are input and output masks for everything described in this chapter, seed train analysis and optimization is much more user friendly and easy to apply for the next seed train to be optimized than entering information directly into the programming language. 9. Attention has to be paid to the phenomenon that the optimal STY can coincide with the end of the exponential growth phase of the cells. Passaging at that point in time means that the cells are already passing in the stationary phase. If cells out of this state are not favored for passaging, another optimization criterion has to be selected. Attention has also to be paid to the risk that actual cell growth may differ from model calculation due to deviation in cell behavior. In case of faster cell growth compared to model calculation but cell passaging according to model calculation, cells are then already in the stationary

Seed Train Optimization

267

phase. A possibility to avoid the described situations is to subtract a certain “safety” time span from the point in time showing the optimal STY or to multiply by a factor, e.g., 0.9. References 1. Rao G, Moreira A, Brorson K (2008) Disposable bioprocessing: the future has arrived. Biotechnol Bioeng 102:348–356 2. Heidemann R, Mered M, Wang DQ et al (2002) A new seed-train expansion method for recombinant mammalian cell lines. Cytotechnology 38:99–108 3. Heidemann R, Mered M, Michaels JD et al (2003) Device and method for seed-train expansion of mammalian cells. US Patent 2003/0113915 A1 4. Le H, Kabbur S, Pollastrini L, Sun Z, Mills K, Johnson K, Karypis G, Hu WS (2012) Multivariate analysis of cell culture bioprocess data–lactate consumption as process indicator. J Biotechnol 162(2–3):210–223 5. Bogacki P, Shampine LF (1989) A 3(2) pair of Runge-Kutta formulas. Appl Math Lett 2:321–325 6. Press WH, Teukolsky SA, Vetterling WT et al (1992) Numerical recipes in C. Cambridge University Press, Cambridge 7. Nelder JA, Mead R (1965) A simplex method for function minimization. Comp J 7:308–313 8. MATLAB Release (2017) The MathWorks, Inc., Natick, MA. USA. https://www.mathworks.com/products/matlab.html. [Last access: 10.10.2019] 9. Po¨rtner R, Sch€afer T (1996) Modelling hybridoma cell growth and metabolism – a comparison of selected models and data. J Biotechnol 49:119–135

10. Frahm B, Lane P, Atzert H et al (2002) Adaptive, model-based control by the open-loopfeedback-optimal (OLFO) controller for the effective fed-batch cultivation of hybridoma cells. Biotechnol Prog 18:1095–1103. https://doi.org/10.1021/bp020035y 11. Frahm B, Lane P, Po¨rtner R et al (2003) Improvement of a mammalian cell culture process by adaptive, model-based dialysis fed-batch cultivation and suppression of apoptosis. Bioprocess Biosyst Eng 26:1–10 12. Kroll P, Hofer A, Stelzer IV, Herwig C (2017) Workflow to set up substantial target-oriented mechanistic process models in bioprocess engineering. Process Biochem 62:24–36. https:// doi.org/10.1016/j.procbio.2017.07.017 13. Sin G, Meyer AS, Gernaey KV (2010) Assessing reliability of cellulose hydrolysis models to support biofuel process design—identifiability and uncertainty analysis. Comput Chem Eng 34(9):1385–1392. https://doi.org/10.1016/ j.compchemeng.2010.02.012 14. Shampine LF, Reichelt MW (1997) The MATLAB ODE Suite. SIAM J Sci Comput 18:1–22 15. Platas OP, Jandt U, Da Minh Phan L et al (2012) Evaluation of criteria for bioreactor comparison and operation standardization for mammalian cell culture. Eng Life Sci 12:518–528

Part IV Process Analysis

Chapter 15 High-Throughput Quantification and Glycosylation Analysis of Antibodies Using Bead-Based Assays Sebastian Giehring Abstract A novel version of bead -based assays with fluorescence detection enables the high-throughput analysis of antibodies and proteins. The protocols are carried out in special 384-well plates, require very few manual interventions, and are easy to automate. Here we describe how the technology can be used to determine antibody titers and screen for product glycosylation, a critical quality attribute, early in cell line and bioprocess development. Key words Product quality, Monoclonal antibodies, Glycosylation, Titer assay, High-throughput assays, Bead-based assays, Bioprocess development, Cell line development, Lectins

1

Introduction In this chapter, we introduce a relatively new bead-based immunoassay technology and describe how it is applied for the quantification and the glycosylation analysis of antibodies during the development and for the monitoring of bioprocesses. The quantification of product, frequently called titer analysis, is one of the most common tasks during developing and monitoring bioprocesses. Product titer is one of the main optimization criteria of bioprocess development together with cell density and viability. Product titers, sample numbers, and the available sample volume differ drastically during the different stages of the development process, and according to the development stage, one can choose from different methods for antibody quantification. In the early selection of clones and in media optimization where high throughput is required, technologies like biolayer interferometry are common, but also HTRF (homogeneous time-resolved fluorescence) and the Gyros technology are in use. Lowerthroughput methods include Protein A-HPLC, which is often used as a benchmark and QC method because of its high reproducibility and accuracy.

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_15, © Springer Science+Business Media, LLC, part of Springer Nature 2020

271

272

Sebastian Giehring

ELISA (enzyme-linked immunosorbent assays) are also still in use, especially if throughput is not an issue. It is, however, a timeconsuming method, and its main advantage, high sensitivity, is usually not required in bioprocess development. Glycosylation is a posttranslational modification (PTM) and a critical quality attribute for all therapeutic proteins including antibodies. Glycosylation has received increasing attention in recent years, raising awareness on how glycosylation impacts the quality of a monoclonal antibody (Mab). Two types of glycosylation exist: N-linked glycans, which are attached to the Asn in a Asn-X-Ser/ Thr sequence, and O-linked glycans, which can be attached to serine or threonine. Antibodies are exclusively N-glycosylated, but O-glycosylation may play an important role in Fc fusion proteins and other glycoproteins. Glycosylation affects many aspects of a glycoprotein, such as stability [1], aggregation [2], and their characteristics as therapeutics such as serum half-life [3–5], immunogenicity [6], and efficacy [7]. In addition, the N-glycans shape the function of Mabs. The presence of α1,6 linked (core) fucose reduces the interaction of the Fc domain with the FcγIIIa receptor and reduces the ability to evoke an antibody-dependent cell-mediated cytotoxicity (ADCC) [8, 9]. Galactosylation of the Mab is important for complementmediated cytotoxicity, CDC [10, 11]. Cell culture conditions such as pH, oxygen, temperature, and media composition affect the cellular glycosylation machinery and finally Mab glycosylation. High producer cells need to process high amounts of proteins, and therefore it is crucial to carefully optimize media components and cell culture conditions. Keeping the nutrient balance is essential to avoid short-term depletion of necessary nutrients which may cause decreased sialylation and increased high mannose glycans [12, 13]. The current understanding is that Mab glycosylation is largely determined by the cell line, and optimizing of the cell culture conditions can only modify the glycosylation of the Mab within the capacity of the given cell line. This highlights the importance of careful cell line selection including glycosylation as a selection criterium. Successful modulation of Mab glycosylation during media development has been widely studied, and there is increasing knowledge about which media supplements help to increase the degree of galactosylation [14] and mannosylation [15] and to decrease fucosylation [16]. Ehret [17] recently examined the effect of several of these known modulators on product glycosylation in two different industrial CHO cell lines. Currently, product glycosylation is usually analyzed at the later stages of the bioprocess development, mainly because of a lack of methods that allow a high sample throughput.

High-Throughput Quantification and Glycosylation Analysis of Antibodies. . .

273

Recently, first publications examined the feasibility of including glycan analysis in early process development and demonstrated that glycosylation profiles that were obtained from fed-batch cell cultures in deep well plates are predictive for the behavior of the cell line in bioreactors [18–20], highlighting the potential of integrating early glyco-screening into bioprocess development. Most analytical techniques for glycan analysis rely on enzymatic removal of the glycans, labeling and finally separating the glycans either by an HPLC (HILIC) or by electrophoresis. The detection or identification can be done by fluorescence detection (2-AB method) or by mass spectrometry. Despite some attempts to speed up the mainly manual sample preparation steps, these methods are still labor-intensive. They do therefore have only limited suitability for early and high-throughput studies. However, these methods remain the gold standard when it comes to QC and release testing. In contrast to these methods, the bead-based assays introduced in this chapter work with intact Mabs and use lectins to detect glycosylation profiles present on the Mab, resulting in much shorter assay times and higher throughput that allow screening multiple samples or timely monitoring of bioprocesses. In addition, the proposed protocols allow to measure cell culture supernatants without sample purification. The assays presented in this chapter are based on the technology developed by PAIA Biotech, Cologne, Germany. They offer high throughput for both Mab quantification and glycosylation analysis, need very little sample volumes, and are able to measure cell culture supernatants directly. In addition, they require little hands-on time, are easy to automate and do not require dedicated instruments. 1.1 Principle of Bead-Based Immunoassays

Bead-based immunoassays are a well-known alternative for classical ELISA formats. Woolley and Hayes [21] provide a good overview on the different variants of bead-based assays that have been developed over the last 15 years. Many of these technologies involve magnetic beads which are easy to wash and separate. A downside of this approach is however that it requires several handling steps which lead to more complicated assay protocols, except if the protocol can be performed in an automated and reliable fashion. All bead-based assays share the principle that the Mab in the sample interacts with functionalized beads and with a sandwiching fluorescence or luminescence marker which allows to determine the quantity of the Mab in the sample. PAIA bead-based assays use very similar assay ingredients, but the read-out has been reversed: instead of measuring the fluorescence marker on the beads, the unbound fluorescence marker in solution is measured. This is accomplished by the special bottom

274

Sebastian Giehring

Fig. 1 The principle of the competitive PA-104 Fc low titer assay

structure of the wells in the 384-well PAIA microplates (Fig. 1a) that separate the beads from the assay solution once the assay reaction is finished. This way, the amount of free unbound fluorescence marker can be measured directly, and the amount of fluorescence marker bound to the beads can be measured indirectly (if the initial marker concentration is known), facilitating both sandwichtype and displacement (or competitive) assay formats. In this article, we will focus on the competitive assay PA-104 for quantification and the sandwich-type assay PA-201 assay for glycosylation analysis. 1.2 High-Throughput Quantification of IgG

Figure 1 depicts the principle of the PA-104 Fc low titer assay, a competitive assay. The assay contains beads functionalized with an IgG (depicted as gray circle) and a fluorescence-labeled Protein A (blue squares with orange asterisk) for detection. During the reaction, the antibody from the sample (the analyte, shown in magenta) releases fluorescence-labeled Protein A from the beads (Fig. 1a) in a concentration-dependent manner (Fig. 1b). The PA-104 Fc low titer kit is designed for the rapid quantification of all IgGs and Fc proteins that interact with Protein A, for example, human IgG1, IgG2, and IgG4, but also IgGs from, e.g., rabbit or mouse. The validated concentration range of the assay is 10–100 μg/mL for human IgG, according to the EMEA guideline for bioanalytical method validation [22], when using a sample volume of 5 μL. The concentration range for non-human IgG can differ from these specifications, depending on their affinity toward Protein A. Samples with concentrations higher than the assay range can be diluted in any typical buffer, e.g., PBS or Tris. In order to obtain the most accurate quantification results, the standards for the generation of the standard curve should contain the same or a homologous IgG as the sample. For accurate

High-Throughput Quantification and Glycosylation Analysis of Antibodies. . .

275

determination of Mab titers in supernatant samples, it is recommend to prepare the standards in conditioned media, or, if not available, in fresh media. It is also possible to use calibration curves that have been generated with a different PA-104 kit. This allows to use all 384-wells on the plate so that samples from four entire 96-well plates can be combined onto one PAIAplate, e.g., using a 96 channel pipet head. A different strategy is leaving one column empty on the 96-well plates which leaves enough wells available for running the standards on the same PAIAplate. The standard curves in the PA-104 (Fig. 1b) have a sigmoidal shape. The quantification of Mab is most accurate in the concentration range in which the standard curve shows the highest slope. Sample concentrations in the flattened parts of the standard curve will be less reliable, and errors in, e.g., pipetting will introduce a somewhat larger error. Standard curves may be generated using 4- or 5-parameter fit models for sigmoidal curves which are available in the software of most fluorescence readers. PAIA Biotech also offers a very lean software tool which provides all essential functions for analyzing data from PAIA quantification assays. 1.3 High-Throughput Glycosylation Analysis of IgG

The PAIA assay technology is also a useful tool for screening the glycosylation of IgGs and Fc fusion proteins using lectins which bind specifically to certain types of glycosylation. Lectins have been used as a research tool for several decades, and more than 100 lectins from different hosts have been characterized and used as analytical tools [23]. Databases exist in which the binding specificities and affinities of an important number of lectins have been compiled [24]. Lectins are used for detecting proteins in blots, in several kinds of microarrays [25], in ELISAlike assay formats [26], and in sensor technology, e.g., surface plasmon resonance [27]. Microarrays offer the possibility to test a large number of lectins at the same time, whereas the other methods use only one lectin at a time. These approaches are all rather slow and labor-intensive; especially, microarrays are a low-throughput method. PAIA glycosylation assays use a customizable number of lectins. Here we show a setting with five lectins testing for the most important types of glycosylation in Mabs: fucosylation, mannosylation, galactosylation, and sialylation. The assay format is a sandwich assay, in which the capture beads bind the Mab analyte and fluorescence-labeled lectins bind to the glycans that are present on the Mab. The amount of lectin that binds to the beads depends on (1) the amount of IgG present, (2) the degree of glycosylation on the IgG, and (3) the affinity of the lectin to the glycan. The results are lectin binding profiles for the selected lectins (Fig. 2). 2,6 linked sialic acid is the type of sialylation which is found

276

Sebastian Giehring

Fig. 2 Lectin binding profiles of different therapeutic antibodies and one Fc fusion protein using galactose, fucose, high mannose, 2,3-linked sialic (2,3 SA), and 2,6 linked sialic acid (2,6 SA) binding lectins

in human and, e.g., mouse cell lines, whereas the 2,3 linkage is found in CHO cells. In the native state, the glycans on the Fc part of Mabs and Fc fusion proteins are buried inside the two heavy chains and are not available for lectin binding. Glycans on the protein part of Fc fusion proteins and on the Fab domain of antibodies (e.g., in Erbitux) are readily accessible for lectins, and these glycans can be detected in the native state. In order to detect Fc glycosylation with lectins, the Fc domain has to be denatured mildly so that the interaction with the capture beads is maintained. This is best achieved by thermal denaturation using the sample prep solution from the PAIA assay kits. Typically, temperatures between 65 and 80  C have to be applied for 5 min. This has to be determined only once for a given IgG. The optimal temperature is the one that shows the best lectin binding rates. This optimization should be performed in the same matrix in which the final assay will be performed (e.g., in diluted conditioned media). The amount of bound lectin is dependent on the amount of Mab in the sample. Therefore, it is necessary to bring all samples to the same concentration to allow for a direct comparison of glycosylation profiles. We recommend starting with an IgG concentration of 100 or 200 μg/mL. These concentrations work well for most antibodies of the IgG1 subtype. IgG2 and IgG4 are known to have significantly lower degrees of glycosylation IgG1. Analysis of these molecules therefore will require the use of higher concentrations. The analyte concentration may also need to be modified depending on the purpose of the assay. If, for example, the samples contain a high degree of fucosylation and one wants to resolve small

High-Throughput Quantification and Glycosylation Analysis of Antibodies. . .

277

Fig. 3 Lectin binding rates of dilution series of different antibodies and one Fc fusion protein (Enbrel) with different degrees of mannosylation and fucosylation

differences in a given set of samples, it is necessary to dilute the IgG sample. If only low amounts of, e.g., high mannose species are expected, it is advisable to increase the concentration of IgG in the assay. To this end we recommend to initially measure concentration series of reference samples with known glycosylation profiles and determine the optimal concentrations for the assays. Figure 3 shows an example of this optimization process for different molecules. The measurement of non-purified samples is possible with the PAIA glycan assays, but the assay performance can be negatively impacted by the culture supernatant. These effects are usually less pronounced if the analyte can be measured in the native state (non-Fc glycosylation) and if the supernatant contains only low concentrations of host cell proteins compared to the product. In general, it is also very helpful if the supernatant can be diluted by a factor of 1:10 before the assay, which means that the titers should be higher than 1 mg/mL in the supernatant. It is recommended to thoroughly test these effects using conditioned media spiked with reference IgG. There are some cell culture supplements that will definitely impact lectin binding because they directly interfere, e.g., galactose. In these cases, the determination of galactosylation is impossible without sample purification. However, a buffer exchange is sufficient to remove excess galactose.

2

Materials All assay kits in this section are from PAIA Biotech GmbH, Germany.

278

Sebastian Giehring

2.1 Material for Quantification Assays

PA-104 assay kit including a 384-well PAIAplate preloaded with dried-in capture beads and PAIAmix reagent. Storage conditions: 4  C (see Note 1).

2.2 Materials for Glycosylation Assays

PA-201 assay kit including 384-well PAIAplates pre-loaded with dried-in capture beads, sample preparation buffer, and lectin reagents for fucosylation, galactosylation, mannosylation, and sialylation. Storage conditions: 4  C (see Note 1).

2.3

Equipment

1. Orbital microplate shaker. The optimal shaker speed is 2300 rpm for shakers with an orbit of 2 mm (e.g. the BioShake from Q-Instruments) and 1600 rpm for shakers with 3 mm orbit (e.g., a Thermomixer from Eppendorf). (a) Slower shakers can also be used but may require significantly longer incubation times (see Notes 2 and 3). 2. Fluorescence microplate reader with bottom reading capability and suited for 384-well plates (e.g., from Tecan, Molecular Devices, Perkin-Elmer, BioTek and BMG Labtech) or automated fluorescence cell imager (e.g., Cellavista/NyONE from SynenTec). 3. Heating device for vials with accurate temperature control (0.1  C), e.g., Thermomixer (Eppendorf) or PCR heating block (only needed for glycosylation assays). 4. Multichannel pipets for the 2–10 μL and 50–100 μL range. 5. Optional: microplate centrifuge.

3

Methods Please also refer to the assays manuals that are available online.

3.1 Quantification Assay Protocol

1. Prepare the IgG standards, e.g., by a serial dilution of a suitable standard IgG in the same matrix as your samples, e.g., mock cell culture supernatant, medium, or buffer. The standards should cover the concentration range of 2–200 μg/mL. We recommend using at least 10 standards with at least two replicates to obtain good calibration data. 2. Dilute the samples, so that the final expected concentrations match the assay range, if necessary. 3. Remove sealing tape from the PAIAplate. 4. Add 35 μL of the ready-to-use PAIAmix into the wells to be used. 5. Add either 5 μL of standard or negative control or sample into the wells containing PAIAmix.

High-Throughput Quantification and Glycosylation Analysis of Antibodies. . .

279

6. Seal the plate with adhesive film. 7. Incubate the plate on the orbital shaker for 15 min at 1600 rpm (for 3 mm orbit) or 2300 rpm (for 2 mm orbit) on the orbital shaker. 8. After the incubation the beads need to settle outside the detection zone. Therefore, let the plate stand without movement for 10 min, and then shake the plate again for 5 min at 900 rpm or spin down the beads on a microplate centrifuge at 500  g for 1 min. 3.1.1 Measurement

PAIAplates are in the standard 384-well format, but they have special detection areas on the bottom of the plate which means that proper plate definition in the reader software is necessary for optimum sensitivity and precision. The plate definition only has to be done once, starting from a pre-defined format, e.g., for Greiner 384-well plates. PAIA offers test plates and dye solution for this purpose. Measurement with a plate reader: 1. When using a fluorescence plate reader, the bottom read mode has to be selected. 2. Excitation wavelength, 640 nm; emission wavelength, 665 nm for monochromators, Cy5 filter sets or similar for filter-based instruments. 3. The detection parameters should be optimized once (excitation time, lamp energy, number of flashes, gain, etc.) to ensure that the generated signals fit well into the dynamic range of the reader. Measurement with a fluorescence cell imager: 1. The fluorescence imager must be capable of imaging the well center and to measure the intensity. We recommend taking a single image with a 10x or 20x objective at a defined distance to be determined by the user. 2. Filter settings for dyes like Cy5, Alexa 647, or excitation source: red/emission filter: deep red for Cellavista and NyONE (SynenTec GmbH). 3. Optimization of the plate definition (see Notes 4–6) and imager settings (excitation time and energy, camera gain, etc.) are done by inspection of the images. It is important that the detector/the camera chip is not saturated.

3.1.2 Data Analysis

Data analysis can be performed with any software capable of generating calibration curves with 4 or 5 parameter fits and calculating concentrations of unknown samples. PAIA Biotech offers a software tool which is easy to operate and which is compatible with the data output format (Excel, csv format) of all types of readers.

280

Sebastian Giehring

3.2 Glycosylation Assay 3.2.1 Optimization of Sample Preparation (Denaturation)

This step is only necessary if you want to measure Fc glycans. Please note that the given volumes are sufficient for assays using four lectins. The volumes may be reduced/increased if less/more lectins are used. 1. Dilute your samples in the sample matrix (buffer or conditioned media) to a concentration of 200 μg/mL (see Note 7). 2. Mix 100 μL of each diluted sample with 100 μL sample prep solution in a vial that fits exactly into your heating device (see Note 8). 3. Prepare as many vials as conditions to be tested (e.g., RT as a control, 65  C, 70  C, 75  C, and 80  C). Please see Note 9. 4. Prepare the negative control accordingly: a sufficient volume of a 1:1 mixture of buffer and sample prep solution for all control wells (10 μL/well). 5. Please make sure the vials are properly closed. 6. Bring your heating device to the correct temperature and insert the samples into the device (see Note 10). 7. After completion of the denaturation (5 min), let the samples cool down to room temperature on the bench. 8. In the meantime, add 50 μL of the lectin reagent column by column into the wells of the PAIAplate according to the plate layout in Fig. 4 (e.g., lectin marker L1 into columns 1, 5, 9, 13, 17, and 21). 9. Add 10 μL of the pretreated samples and controls to the wells according to the plate layout. 10. Seal the plate with adhesive film. 11. Incubate the plate on the orbital shaker for 45 min at 1400 rpm (for 3 mm orbit) or 1800 rpm (for 2 mm orbit) on orbital shaker. 12. Centrifuge the plate for 1 min at min 500  g in a microplate centrifuge (e.g. swing-out). 13. Afterward the plate is ready for measurement. Interpretation of results from optimization experiments. The lectin binding patterns for the different sample prep conditions can be quickly generated by the Excel spreadsheets provided by PAIA Biotech. The optimal conditions are those that show the highest lectin binding rates.

3.2.2 Standard Assay Protocol

1. Dilute your samples in sample matrix to a suitable concentration, typical between 100 and 200 μg/mL. 2. Mix 100 μL of each diluted sample with 100 μL sample prep solution in a vial that fits exactly into your heating device.

High-Throughput Quantification and Glycosylation Analysis of Antibodies. . .

281

Fig. 4 Plate layout for an entire plate with a maximum of 24 samples in triplicates. L1 is lectin No. 1, L2 is lectin No. 2, etc. Controls are the sample matrix and sample prep mixtures without IgG

3. Prepare the negative controls accordingly: (a) A sufficient volume of a 1:1 mixture of buffer and sample prep solution for all control wells needed (10 μL/well). 4. Please make sure the vials are properly closed. 5. Bring your heating device to the optimal temperature (determined according to Subheading 3.2.1), and insert the samples into the device. 6. After completion of the denaturation, let the samples cool down to room temperature on the bench. 7. In the meantime, add 50 μL of the labeled lectins column by column into the wells of the PAIAplate according to the plate layout in Fig. 4. 8. Add 10 μL of the pretreated samples and controls to the wells according to the plate layout. 9. Seal the plate with adhesive film. 10. Incubate the plate on the orbital shaker for 45 min at 1400 rpm (for 3 mm orbit) or 1800 rpm (for 2 mm orbit) on orbital shaker. 11. Centrifuge the plate for 1 min at min 500  g in a microplate centrifuge (e.g., swing-out). 12. Afterward the plate is ready for measurement.

282

Sebastian Giehring

3.2.3 Measurement

According to Subheading 3.1.2.

3.2.4 Data Analysis

The generation of the lectin binding profiles from raw data can be easily performed in Excel spreadsheets available from PAIA Biotech for plate layouts like the one in Fig. 4. Fluorescence intensity raw data are pasted into the data input area of the respective Excel spreadsheet, and lectin binding profiles are automatically calculated and plotted. Lectin binding (in %) is calculated as follows:   Mean intensity of sample Lectin binding ¼ 1  Mean intensity of negative control An export table for the different samples is automatically generated so that the results of different plates can be easily combined for further analysis (e.g., clone ranking).

4

Notes 1. If you have only used part of the plate, please seal the plate with adhesive film, wrap it into the silver envelope and keep it in the fridge at 4  C. Please store the remaining PAIAmix in the fridge as well. 2. Efficient mixing is essential for PAIA assays, but you want to avoid spilling over. It is recommended to seal the plate for the shaking step and have a suitable rpm number for your type of shaker. After the shaking, there is typically some condensation on the sealing, which is not problematic. If you observe big droplets on the sealing, your shaking speed may be too high. This may also lead to higher CVs of replicates. 3. If you are not sure that your shaking is efficient enough (e.g., because the calibration curves look different than in the manual), you may increase the shaking speed or time. You can also shake and measure the assay several times in order to determine the shaking time that is necessary to reach equilibrium on your particular shaker model. If you feel that the reproducibility of your assays is not good, then you should check the pipetting accuracy for both buffer and sample and the plate definition in your reader, and you may also increase the number of flashes on the reader. 4. Data quality depends on the read-out instrument and its configuration. Accurate plate positioning for the 384-well plate is required. The dimensions of the PAIAplate are identical with the Greiner 384-well plate format (greiner384ft) which is available as a template in the software of most plate readers.

High-Throughput Quantification and Glycosylation Analysis of Antibodies. . .

283

5. If you use a fluorescence imager, you may have to define the PAIA plate given the dimensions in the assay manual. 6. We recommend to use a PAIA test plate (without beads) and fill all wells with 50 μL test dye solution, shake for 5 min at 1400 rpm, and read the plate on your reader/imager. You should achieve a CV for the whole plate of not higher than 4% and you shouldn’t observe any systematic errors reading the plate. Some reader software offer the possibility of scanning the positions of, e.g. corner wells and allow you to adjust plate layouts accordingly. 7. For the glycosylation assay, we recommend to perform the assay with a reference molecule with known glycosylation and also measure dilution series of the molecule when setting up the assay. 8. The vials for the denaturation step should fit tightly into the device so that efficient and reproducible heat transfer into the solution is accomplished. If this is not the case, you may get unreliable results. 9. If you are not sure about the accuracy of your heating device, you can check this with a temperature sensor which can be introduced into the solution in the vial. 10. If you observe an increase of turbidity in your sample during the optimization of the denaturing conditions for the glycan assays, then the applied temperature is too high. References 1. Zheng K et al (2011) The impact of glycosylation on monoclonal antibody conformation and stability. MAbs 3(6):568–576 2. Onitsuka M et al (2014) Glycosylation analysis of an aggregated antibody produced by Chinese hamster ovary cells in bioreactor culture. J Biosci Bioeng 117(5):639–644 3. Liu L (2018) Pharmacokinetics of monoclonal antibodies and Fc-fusion proteins. Protein Cell 9(1):15–32 4. Yu M et al (2012) Production, characterization and pharmacokinetic properties of antibodies with N-linked Mannose-5 glycans. MAbs 4 (4):475–487 5. Alessandri L et al (2012) Increased serum clearance of oligomannose species present on a human IgG1 molecule. MAbs 4(4):509–520 6. Chung CH et al (2008) Cetuximab-induced anaphylaxis and IgE specific for Galactose-α-1,3-Galactose. N Engl J Med 358:1109–1117

7. Sola RJ, Griebenow K (2010) Glycosylation of therapeutic proteins: an effective strategy to optimize efficacy. BioDrugs 24(1):9–21 8. Liu SD et al (2015) Afucosylated antibodies increase activation of FcγRIIIa-dependent signaling components to intensify processes promoting ADCC. Cancer Imunol Res 3 (2):173–183 9. Junttila TT et al (2010) Superior in vivo efficacy of Afucosylated Trastuzumab in the treatment of HER2-amplified breast cancer. Cancer Res 70(11):4481–4489 10. Peschke B et al (2017) Fc-Galactosylation of human immunoglobulin gamma Isotypes improves C1q binding and enhances complement-dependent cytotoxicity. Front Immunol 8:646 11. Hodoniczky J et al (2005) Control of recombinant monoclonal antibody effector functions by Fc N-glycan remodeling in vitro. Biotechnol Prog 21(6):1644–1652 12. Wong D et al (2005) Impact of dynamic online fed-batch strategies on metabolism,

284

Sebastian Giehring

productivity and N-glycosylation quality in CHO cell cultures. Biotechnol Bioeng 89 (2):164–177 13. Fan Y et al (2015) Amino acid and glucose metabolism in fed-batch CHO cell culture affects antibody production and glycosylation. Biotechnol Bioeng 112(3):521–535 14. Gramer MJ et al (2011) Modulation of antibody galactosylation through feeding of uridine, manganese chloride, and galactose. Biotechnol Bioeng 108(7):1591–1602 15. Bruehlmann D et al (2017) Cell culture media supplemented with raffinose reproducibly enhances high mannose glycan formation. J Biotechnol 252:32–42 16. Okeley NM et al (2013) Development of orally active inhibitors of protein and cellular fucosylation. PNAS 110:5404–5409 17. Ehret J et al (2019) Impact of cell culture media additives on IgG glycosylation produced in Chinese hamster ovary cells. Biotechnol Bioeng 116(4):816–830 18. Rouiller Y et al (2016) Screening and assessment of performance and molecule quality attributes of industrial cell lines across different fed-batch systems. Biotechnol Prog 32 (1):160–170 19. Mora A et al (2018) Sustaining an efficient and effective CHO cell line development platform by incorporation of 24-deep well plate

screening and multivariate analysis. Biotechnol Prog 34(1):175–186 20. Loebrich S et al (2019) Comprehensive manipulation of glycosylation profiles across development scales. MAbs 11(2):335–349 21. Wooley CF, Hayes MA (2013) Recent developments in emerging microimmunoassays. Bioanalysis 5(2):245–264 22. European Medicines Agency 2012: Guideline on bioanalytical method validation 23. Hendrickson OD, Zherdev AV (2018) Analytical applications of lectins. Crit Rev Analyt Chem 48(4):279–292. https://doi.org/10. 1080/10408347.2017.1422965 24. Lectin Frontier DataBase (LfDB), Glycoscience and Glycotechnology Research Group, National Institute of Advanced Industrial Science and Technology, Japan. https:// acgg.asia/lfdb2/index 25. Wang L et al (2014) Cross-platform comparison of glycan microarray formats. Glycobiology 24(6):507–517 26. Thompson R et al (2011) Optimization of the enzyme-linked lectin assay for enhanced glycoprotein and glycoconjugate analysis. Anal Biochem 413(2):114–122 27. Geuijen KP et al (2015) Label-free glycoprofiling with multiplex surface plasmon resonance: a tool to quantify sialylation of erythropoietin. Anal Chem 87:8115–8122

Chapter 16 Surface Plasmon Resonance-Based Method for Rapid Product Sialylation Assessment in Cell Culture Olivier Henry, Eric Karengera, Florian Cambay, and Gregory De Crescenzo Abstract To streamline cell culture process development, surface plasmon resonance (SPR) biosensors offer a versatile platform for the rapid quantification and quality analysis of recombinant proteins. As a representative case study, the present chapter details a procedure employing a SPR biosensor for determining the differential sialylation levels of recombinant interferon α2b contained in cell culture samples, using immobilized Sambucus nigra lectin. Of interest, this semiquantitative approach can be adapted to work with other lectins with unique carbohydrate-binding specificities, enabling a wide range of product characterization analysis. Key words Product quality assessment, Surface plasmon resonance, Product sialylation, Recombinant glycoprotein

1

Introduction Despite the critical implications of product heterogeneity encountered in biomanufacturing, a large part of process development activities still concentrate on cell growth and productivity rather than product quality. It is now widely documented that process development steps and scale-up may lead to undesired changes to product attributes and consistency [1]. In this context, the ability to rapidly monitor and control product glycosylation, from earlyto late-stage process development, is of salient interest to reduce the time and cost to market. Many factors are known to affect protein glycosylation and its inherent heterogeneity, including the host cell, the culture medium, the mode of operation, and the operating conditions [2–7]. For these reasons, regulatory agencies like the FDA and the EMA demand a rigorous characterization of glycoprotein quality, thereby making the control and the optimization of glycosylation a major concern for the pharmaceutical industries.

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_16, © Springer Science+Business Media, LLC, part of Springer Nature 2020

285

286

Olivier Henry et al.

Different approaches exist to assess the product bioactivity and glycosylation state of glycoproteins, including NMR, capillary electrophoresis, and liquid chromatography coupled to mass spectrometry to cite a few [8]. Whether the methods involve the analysis of intact proteins, glycopeptides, or released glycans, they are generally very sensitive but difficult to streamline as they require cumbersome sample pretreatment steps (e.g., chemical or enzymatic cleavage). Hence, the development of high-throughput analytical tools that can support process development efforts or allow realtime monitoring for product quality assessment remains a challenge [9]. SPR biosensor technology presents a versatile approach that bears the potential to enable rapid product quality assessment. By exploiting lectins affinity for specific glycans, SPR biosensing has proved an effective analytical platform for the differentiation and identification of glycoproteins [10]. Using immobilized receptors, this approach has also shown its value for the simultaneous quantification and bioactivity evaluation of recombinant antibodies, as well as for correlating various binding kinetic signatures with distinct antibody N-glycosylation patterns and aggregation states [11]. Of salient interest, previous studies have shown that an SPR biosensor can be harnessed to a bioreactor to allow at-line determination of product concentration and bioactivity [12, 13]. The present chapter describes a SPR-based approach for the semiquantitative assessment of sialylation changes in cell culture samples. Purified samples of IFNα2b were analyzed via an SPR assay to measure the protein interactions with a Sambucus nigra (SNA) lectin. SNA is one of the most widely used lectins for sialylation analysis because of its relatively good specificity. It binds to sialic acid attached to a galactose in α2,6- and, to a lesser degree, in α2,3-linkage. Control experiments performed with preparations of IFNα2b harboring different degrees of sialylation indicated that the shape of the normalized SNA-IFNα2b sensorgram is a signature of a given IFNα2b sialylation pattern, while the SNA/IFNα2b complex stability is linked to the level of sialylated IFNα2b. For many therapeutics biologics, the terminal sialic acid content is a critical quality attribute, and it is generally desired to achieve a high and consistent sialylation level, in order to increase the glycoprotein half-life in the circulating blood and to maximize therapeutic efficacy [14, 15]. The method described herein proved to be robust and sensitive enough for comparing the sialylation state of recombinant interferon produced in transformed and parental HEK293 cells, as well as to evaluate the impact of different nutrient substitution and fed-batch strategies [16–18].

SPR-Based Method for Product Sialylation Assessment

2

287

Materials

2.1 Analytical Instruments

¨ kta Explorer System (GE healthcare). 1. A

2.2 Product Purification by Cation Exchange Chromatography

1. 0.1 M citric acid: dilute 21 g of citric acid monohydrate in 1 L of ultrapure water.

2. Biacore T100 biosensor (GE healthcare).

2. 0.1 M Trisodium citrate: dilute 29.4 g of Trisodium citrate dihydrate in 1 L of ultrapure water. 3. Equilibration buffer: solution of 0.1 M Trisodium citrate pH 3.5 containing 0.35 M NaCl. Mix 300 mL of 0.1 M Trisodium citrate with 700 mL of 0.1 M citric acid monohydrate. Add 20.5 g of NaCl. Mix and adjust the pH to 3.5 using HCl and NaOH. 4. Elution buffer: solution of 1 L of 0.1 M Trisodium citrate pH 6 containing 0.35 M NaCl (elution buffer). Mix 885 mL of 0.1 M Trisodium citrate with 115 mL of 0.1 M citric acid monohydrate. Add 20.5 g of NaCl. Mix and adjust the pH to 6 using HCl and NaOH. 5. Tricorn™ Mono-S 10/100 GL column (GE healthcare) PBS buffer. 6. 3K centrifugal filter (Pall Corporation). 7. 50 mL Falcon™ tubes.

2.3 Sensor Chip Preparation for Product Quantification

1. Monoclonal anti-IFNα2b antibody, KT5 (Thermo Fisher Scientific, Waltham, MA). 2. CM5 sensor chips (GE healthcare). 3. 100 mM N-hydroxysuccinimide (NHS). 4. 400 mM 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC). 5. Acetate buffer pH 5. 6. Ethanolamine 1 M pH 8.5. 7. HBS-EP buffer: 0.01 M HEPES pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% v/v Surfactant P20. 8. HCl 50 mM.

2.4 Sensor Chip Preparation for Product Sialylation Evaluation

1. Unconjugated lectin: Sambucus nigra agglutinin (SNA) (Vector Laboratories, Burlingame, CA). 2. HBS-EP buffer: 0.01 M HEPES pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% v/v Surfactant P20. 3. Neuraminidase from Arthrobacter ureafaciens (Sigma–Aldrich, St. Louis, MO).

288

3

Olivier Henry et al.

Methods

3.1 Product Purification

1. Centrifuge the cell culture samples at 2000  g for 5 min and keep the supernatant. 2. Mix 10 mL of culture supernatant with the 30 mL of equilibration buffer in 50 mL Falcon™ tube and store the sample to 4  C. 3. Equilibrate the Tricorn™ Mono-S 10/100 GL column by injecting the equilibration buffer at 1 mL/min until stabilization of the pH (to around pH 3.5) and the conductivity. 4. Inject at 1 mL/min the culture sample previously diluted in the equilibration buffer onto the Tricorn™ Mono-S 10/100 GL column. During the loading, when there is around 5 mL of sample left in the Falcon™ tube, add 10 mL of equilibration buffer in the tube to ensure that most of the sample has been injected. Continue the loading until all the sample has been injected into the column. 5. Wash the column with the equilibration buffer at 3 mL/min for at least 5 min. 6. Elute the column at a flow rate of 3 mL/min with an increasing pH gradient from pH 3.5 to pH 6.0 for 30 min using the equilibration and the elution buffer. Monitor the gradual increase of the pH. The elution of IFNα2b should occur between pH 5.0 and pH 6.0. 7. Desalt the purified sample and buffer exchange by filtration against PBS using 3K centrifugal filter.

3.2 Product Quantification

1. The first step consists in immobilizing the monoclonal antiIFNα2b antibody on a CM5 sensor chip surface by amine coupling (Fig. 1). To this end, prepare the NHS/EDC solution by mixing 200 μL of 100 mM N-hydroxysuccinimide (NHS) with 200 μL of 400 mM 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC). 2. Inject the NHS/EDC solution at 5 μL/min for 7 min on two surfaces of the CM5 sensor chip to activate carboxymethylated groups. 3. Dilute the anti-IFNα2b antibody at 50 μg/mL in acetate buffer pH 5.0. 4. Inject the antibody in acetate buffer at 5 μL/min for 10 min on the capture surface. For the mock surface, no protein is injected. 5. Inject ethanolamine 1 M pH 8.5 at 5 μL/min for 7 min to block unreacted carboxymethylated groups on both surfaces.

SPR-Based Method for Product Sialylation Assessment

289

Fig. 1 Immobilization of anti-IFNa2b antibody on the capture surface: the sensorgram shows the surface activation (a) and the injection of the antibody (b), followed by the blockage of the unreacted carboxymethylated surface (c)

Fig. 2 Calibration curve for IFNa2b quantification by SPR

6. Prepare serial dilution of pure IFNα2b of known concentration in HBS-EP buffer (see Note 1). 7. Inject each concentration of pure IFNα2b at 50 μL/min for 1 min on both the capture and the mock surface of the CM5 sensor chip. After each injection, regenerate the surfaces using short injections (10 s) of 50 mM of HCl followed by at least 1 min of running buffer flowing on the surfaces. 8. Plot the signal at 2 s after samples injections against IFNα2b concentration to construct the calibration curve (Fig. 2). 9. Clarified supernatants collected from the cultures and diluted in HBS-EP can be then analyzed in duplicate to quantify final IFNα2b titer.

290

Olivier Henry et al.

3.3 Product Sialylation Analysis

1. To immobilize the lectin by the amine coupling method (see Note 2), start by injecting the NHS/EDC solution at 5 μL/min for 7 min on two surfaces of the CM5 sensor chip to activate carboxymethylated groups. 2. Dilute the SNA lectin at 50 μg/mL in acetate buffer pH 5.0. Other lectins recognizing different glycan residues of interest can be used in a similar way (see Note 3). 3. Inject the SNA lectin in acetate buffer at 5 μL/min for 10 min on the capture surface. On the mock surface, no protein is injected. 4. Inject ethanolamine 1 M pH 8.5 at 5 μL/min for 7 min to block unreacted carboxymethylated groups on both surfaces. 5. To prepare desialylated IFNα2b, mix 2 mU of neuraminidase from Arthrobacter ureafaciens with 100 μg of purified IFNα2b in 50 mM sodium phosphate buffer pH 5.0. For the untreated control, instead of neuraminidase, 50 mM sodium phosphate at pH 5.0 is added. 6. Incubate the samples at 37  C for 1 h. 7. Stop the enzymatic reaction by storing the sample at 4  C. 8. Prepare IFNα2b samples with different levels of sialylation by mixing untreated (hence sialylated) IFNα2b and totally desialylated IFNα2b to generate samples desialylated at 0, 25, 50, 75, and 100% (Fig. 3). Mix 25 μL of untreated IFNα2b with 75 μL of desialylated IFNα2b to get the sample desialylated at 75%. Mix 50 μL of untreated IFNα2b with 50 μL of desialylated IFNα2b to get the sample desialylated at 50%. Mix

Fig. 3 (a) Raw sensorgram data of the interaction of SNA lectin with pure IFNα2b at 50 mg/L at different desialylation levels. (b) Normalized sensorgram data of the interaction of SNA lectin with pure IFNα2b at 50 mg/L at different desialylation levels. An increase in sialylation percentage leads to an increase in the amplitude of the recorded SPR signal for the SNA-IFNα2b interactions. When normalized, these sensorgrams are superposed confirming that the shape of the normalized SNA-IFNα2b sensorgram is a distinct signature corresponding to a given IFNα2b sialylation pattern, while the amplitude of the raw sensorgram is linked to the concentration of sialylated IFNα2b

SPR-Based Method for Product Sialylation Assessment

291

75 μL of untreated IFNα2b with 25 μL of desialylated IFNα2b to get the sample desialylated at 25%. 9. For the detection by SPR, inject purified samples of IFNα2b diluted at 100 mg/L in HBS-EP at 50 μL/min for 1 min on the surfaces harboring the SNA lectin to evaluate the sialylation of IFNα2b in culture samples (see Note 4). 10. Between each sample injection, regenerate the surfaces using short injections (10 s) of 50 mM of HCl followed by at least 1 min of running buffer flowing on the surfaces. 11. Normalize the SPR sensorgrams to allow comparison between the proteins produced in the different culture samples (see Note 5). 12. Signal values at 180 s after the end of sample injection (300 s after the start of the cycle) are used for comparison (see Note 6).

4

Notes 1. All buffers should be filtered through a 0.22 μm filter before use. If the SPR biosensor does not contain a built-in degasser, the buffers must be degassed under vacuum and sonication. 2. Proper immobilization of the lectins is a crucial step. Depending on the lectin employed (e.g., if the lectin does not have enough accessible amine groups), other immobilization methods can be considered as an alternative to amine coupling (e.g., biotin/avidin). Biotinylated lectins are widely commercially available and can be used in conjunction with a streptavidincoated sensor surface. As an alternative approach, the recombinant protein can be fused with a poly-histidine or biotin tag to allow its capture with an immobilized anti-histidine antibody or streptavidin, followed by lectin injection. Finally, in the case of an antibody product, we refer the reader to [11, 19] for quality control and concentration assessment. 3. The SNA lectin is widely used for the detection of α2,6-linked sialylation in N- and O-linked glycoproteins because of its relatively high specificity [20, 21]. Fetuin from fetal bovine serum can be used as a standard sialoglycoprotein (containing approximately 8% by weight of bound sialic acid) to assess the capacity of the lectin to bind to sialic acid. Other lectins with known specific affinities may be employed, depending on the glycan residue of interest. For instance, Maackia Amurensis Lectin II (MALII) is specific for α-2,3 linked terminal sialic acids. Jacalin has a strong affinity for galactosyl (β-1,3) N-acetylgalactosamine occurring widely in O-glycosylated proteins but will bind both sialylated and non-sialylated glycans.

292

Olivier Henry et al.

Fig. 4 Analysis of the interaction between IFNα2b and SNA lectin. (a) Raw sensorgram data of the interaction of SNA lectin with pure IFNα2b from three different cell cultures conditions. (b) The corresponding normalized sensorgram data. The SPR response corresponding to the 300 s time point in panel (b) (see arrow) was used as an indication of the SNA-IFNα2b complex stability and thus of the type of sialylation pattern on IFNα2b. Hence, in this dataset, culture condition 1 yielded the highest degree of sialylation, as evidenced by the higher SPR signal at the end of the dissociation phase

Other lectins can be useful for detecting terminal galactose (Erythrina cristagalli lectin) or mannose residues (Concanavalin A). The combined use of multiple lectins can thus provide complementary information. 4. All analysis should be minimally performed in duplicate. This procedure can be repeated at different product concentrations (e.g., 50 and 200 mg/L) to confirm the results. 5. SPR sensorgrams must be normalized to allow a direct comparison of the lectin/protein kinetics of dissociation between tested samples. The sensorgrams shown in Fig. 4 were normalized by setting the maximum value of the dissociation phase to 100%. The sharp decrease at the sample/buffer injection transition is due to a refractive index difference between the sample medium and buffer. The more diluted the sample, the less pronounced the refractive index change. 6. Depending on the specific lectin/glycoprotein system under investigation, the comparison can be based on the maximum binding capacity (Rmax), the residual response after dissociation, or the slope during association. It is instrumental to calibrate the approach to ensure that the observed differences from SPR sensorgrams are due to distinct sialylation or glycosylation pattern. References 1. Li F, Vijayasankaran N, Shen AY, Kiss R, Amanullah A (2010) Cell culture processes for monoclonal antibody production. mAbs 2 (5):466–479

2. St Amand MM, Radhakrishnan D, Robinson AS, Ogunnaike BA (2014) Identification of manipulated variables for a glycosylation control strategy. Biotechnol Bioeng 111:10

SPR-Based Method for Product Sialylation Assessment 3. Shi HH, Goudar CT (2014) Recent advances in the understanding of biological implications and modulation methodologies of monoclonal antibody N-linked high mannose glycans. Biotechnol Bioeng 111(10):1907–1919 4. Ivarsson M, Villiger TK, Morbidelli M, Soos M (2014) Evaluating the impact of cell culture process parameters on monoclonal antibody N-glycosylation. J Biotechnol 188:88–96 5. Gawlitzek M, Estacio M, Furch T, Kiss R (2009) Identification of cell culture conditions to control N-glycosylation site-occupancy of recombinant glycoproteins expressed in CHO cells. Biotechnol Bioeng 103(6):1164–1175 6. Majid FA, Butler M, Al-Rubeai M (2007) Glycosylation of an immunoglobulin produced from a murine hybridoma cell line: the effect of culture mode and the anti-apoptotic gene, bcl-2. Biotechnol Bioeng 97(1):156–169 7. Hossler P, Khattak SF, Li ZJ (2009) Optimal and consistent protein glycosylation in mammalian cell culture. Glycobiology 19 (9):936–949 8. Zhang L, Luo S, Zhang B (2016) Glycan analysis of therapeutic glycoproteins. Mabs-Austin 8(2):205–215 9. Tharmalingam T, Wu CH, Callahan S, Goudar T (2015) A framework for real-time glycosylation monitoring (RT-GM) in mammalian cell culture. Biotechnol Bioeng 112 (6):1146–1154 10. Shinohara Y, Furukawa J (2014) Surface plasmon resonance as a tool to characterize lectincarbohydrate interactions. Methods Mol Biol 1200:185–205 11. Dorion-Thibaudeau J, Raymond C, Lattova E, Perreault H, Durocher Y, De Crescenzo G (2014) Towards the development of a surface plasmon resonance assay to evaluate the glycosylation pattern of monoclonal antibodies using the extracellular domains of CD16a and CD64. J Immunol Methods 408:24–34 12. Chavane N, Jacquemart R, Hoemann CD, Jolicoeur M, De Crescenzo G (2008) At-line quantification of bioactive antibody in bioreactor by surface plasmon resonance using epitope detection. Anal Biochem 378(2):158–165

293

13. Jacquemart R, Chavane N, Durocher Y, Hoemann C, De Crescenzo G, Jolicoeur M (2008) At-line monitoring of bioreactor protein production by surface plasmon resonance. Biotechnol Bioeng 100(1):184–188 14. Tejwani V, Andersen MR, Nam JH, Sharfstein ST (2018) Glycoengineering in CHO cells: advances in systems biology. Biotechnol J 13 (3):e1700234 15. Lalonde ME, Durocher Y (2017) Therapeutic glycoprotein production in mammalian cells. J Biotechnol 251:128–140 16. Karengera E, Robotham A, Kelly J, Durocher Y, De Crescenzo G, Henry O (2018) Concomitant reduction of lactate and ammonia accumulation in fed-batch cultures: impact on glycoprotein production and quality. Biotechnol Prog 34(2):494–504 17. Karengera E, Durocher Y, De Crescenzo G, Henry O (2017) Combining metabolic and process engineering strategies to improve recombinant glycoprotein production and quality. Appl Microbiol Biotechnol 101 (21):7837–7851 18. Karengera E, Robotham A, Kelly J, Durocher Y, De Crescenzo G, Henry O (2017) Altering the central carbon metabolism of HEK293 cells: impact on recombinant glycoprotein quality. J Biotechnol 242:73–82 19. Dorion-Thibaudeau J, Durocher Y, De Crescenzo G (2017) Quantification and simultaneous evaluation of the bioactivity of antibody produced in CHO cell culture-the use of the ectodomain of FcgammaRI and surface plasmon resonance-based biosensor. Mol Immunol 82:46–49 20. Haseley SR, Talaga P, Kamerling JP, Vliegenthart JF (1999) Characterization of the carbohydrate binding specificity and kinetic parameters of lectins by using surface plasmon resonance. Anal Biochem 274(2):203–210 21. Safina G, Duran Iu B, Alasel M, Danielsson B (2011) Surface plasmon resonance for realtime study of lectin-carbohydrate interactions for the differentiation and identification of glycoproteins. Talanta 84(5):1284–1290

Chapter 17 Analysis of Product Quality of Complex Polymeric IgM Produced by CHO Cells Julia Hennicke and Renate Kunert Abstract Immunoglobulin M (IgM) antibodies are considered as promising biopharmaceutical drugs in the future despite recombinant production is quite challenging as incomplete polymer formation or nucleic acid adherence can decrease the quality of the IgM preparation. Therefore, we defined densitometric and chromatographic methods as suitable tools to analyze the polymer distribution and the remaining nucleic acid content after initial IgM purification. Additionally, the quality of the glycosylation pattern is an important parameter for biological safety and efficacy. Key words Immunoglobulin M (IgM), Quality attributes, Polymer distribution, Nucleic acid contamination

1

Introduction IgM antibodies are the first immunoglobulin subtype that is produced during B-cell response. They defend various infections and additionally ensure homeostasis by clearance of apoptotic cells, misfolded proteins, or altered (tumorigenic) cells. Thus, IgMs are thought to have therapeutic potential for a wide range of diseases [1, 2]. The IgM monomer comprises two μ-heavy chains and two light chains of the κ or λ subtype. Human secretory IgM consists predominantly of five monomeric subunits and one additional J-chain. In addition, other forms like monomers and hexamers are found to some extent in serum [3]. The amount of incomplete IgMs (polymers smaller than pentamer) might be considerably increased during recombinant production [4, 5]. The lower number of assembled monomeric subunits reduces the avidity of the molecule, but also the functionality since effector proteins in the immune systems such as complement components or the polymeric immunoglobulin receptor (pIgR) specifically recognize the pentameric structure [6, 7]. Therefore, determination of the oligomerization

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_17, © Springer Science+Business Media, LLC, part of Springer Nature 2020

295

296

Julia Hennicke and Renate Kunert

degree of IgMs is a critical quality attribute, which influences the biologic activity of the IgM preparation. Purification of IgMs is quite challenging as they are sensitive to low pH, low ionic strength, and hydrophobic surfaces [8, 9]. Furthermore, IgMs tend to interact with chromatin catabolites, which render the purification even more difficult [10]. The remaining DNA could interfere with assays performed in fundamental research but could also lead to severe side effects when administered as therapeutic drugs. As many chromatographic systems failed to remove the DNA from an IgM preparation, it is important to monitor the remaining DNA content after the purification [11]. The glycosylation pattern is frequently investigated since it has an impact on pharmacokinetic and structural properties and is therefore also regarded as a quality attribute [12]. The glycan sites Asn171 (GS1), Asn332 (GS2), and Asn395 (GS3) are mainly occupied by complex type glycans, whereas exclusively oligomannose type is attached to Asn402 (GS4) and Asn563 (GS5) [13]. The J-chain contains an additional glycan site. The glycosylation pattern is important for the interaction with effector proteins like the complement component [14]. To study complex posttranslational modifications, recombinant Chinese hamster ovary (CHO) cell lines producing IgMs were developed [15], but also other mammalian systems or even plant host cell lines can be used [16, 17]. In this chapter, different methods for the analysis of IgM quality attributes are described.

2 2.1

Materials Densitometry

1. Crude supernatant of an IgM producing CHO cell line (see Note 1) and a reference standard, e.g., NativeMark™ Unstained Protein Standard (Thermo Fisher Scientific) or (preferably) the purified IgM (see Note 2). 2. 5 SDS-Loading dye (250 mM Tris–HCl pH 6.8, 19% SDS, 50% glycerol, 0.5% Bromophenol blue). 3. NativePAGE™ 3–12% Bis-Tris protein gels (Thermo Fisher Scientific). 4. NuPAGE™ Tris-Acetate SDS Running Buffer (20, Thermo Fisher Scientific). 5. XCell SureLock® Mini-Cell (Thermo Fisher Scientific). 6. Powerease 500 Electrophoresis Power supply (Invitrogen). 7. Fixing solution (50% methanol, 7% acetic acid). 8. Containers for gel staining. 9. A common bench rocker. 10. SYPRO® Ruby Protein Gel Stain (Invitrogen).

Analysis of IgM Quality Attributes

297

11. Aluminum foil. 12. Washing solution (10% methanol, 7% acetic acid). 13. Typhoon FLA 9500 Laser Scanner (GE Healthcare, see Note 3). 2.2

SEC-HPLC

1. Purified IgM sample (see Note 4) and a reference standard, e.g., gel filtration standard (Bio-Rad) (see Note 5). 2. Ultrafree® HV centrifugal filters (Sigma-Aldrich), Nalgene™ bottle top filter (VWR), Supor 450 Membrane Disc Filters, 0.45 μm (Pall Laboratory). ˚ (Waters). 3. ACQUITY UPLC Protein SEC Column 450 A 4. Running buffer (0.1 M phosphate pH 5.5, 0.2 M NaCl). 5. A Shimadzu prominence LC20 high-performance liquid chromatography (HPLC) system coupled to a diode array detector (SPD-M20A, Shimadzu), a refractive index detector (RID-10A, Shimadzu), and a multiangle light scattering (MALS) detector (WYATT Heleos Dawn8+ plus QELS).

2.3 Released N-Glycan Assay

3

The glycosylation was analyzed by the company ProGlycAn (http://www.proglycan.com). Purified IgM samples (~100 μg) were sent for in solution analysis.

Methods

3.1 Characterization of the Polymer Distribution from the Crude Supernatant by Densitometry

1. Dilute 500 ng of IgM and the 5 SDS loading dye to a volume of 25 μL with ddH2O (see Note 6). Prepare at least duplicates for each sample. 2. Prepare 800 mL 1 NuPAGE™ Tris-Acetate SDS Running Buffer by dissolving 40 mL of the 20 stock solution in ddH2O. 3. Place the NativePAGE™ 3–12% Bis-Tris protein gels into the XCell SureLock® Mini-Cell, and fill the running buffer in both chambers. 4. Remove the gel comb and load the samples and a reference (see Note 2) into the pockets. 5. Connect and switch on the power supply. Apply a voltage of 200 V for 75 min. 6. Take the gel and remove the plastic covers. Be careful as the gel could break. A broken gel can make the troubles during the evaluation. 7. Put the gel in fixing solution and incubate for 30 min (see Note 7) on a bench rocker.

298

Julia Hennicke and Renate Kunert

Fig. 1 Polymer distribution in the crude supernatant. SDS-Gel used for densitometry. The gel was stained with SYPRO Ruby and scanned using the Typhoon FLA biomolecular imager. L-Ladder NativeMark™ Unstained Protein Standard, C-control IgM012_GL (affinity purified), SN1 + SN2- duplicate of one representative cell culture supernatant cultivated at 37.0  C, pH 7.05. IgM pentamers and dimers are framed in a blue box

8. Exchange the fixing solution by a fresh one and incubate for 30 min on a bench rocker. 9. Incubate the gel into the SYPRO® Ruby Protein Gel Stain overnight on a bench rocker. Wrap the staining container with aluminum foil or incubate in dark containers (see Note 8). 10. Wash the gel in washing solution for 30 min and protect the gel from light. 11. Wash the gel in ddH2O for 10 min and protect the gel from light. 12. Use the Typhoon FLA 9500 Laser Scanner to read out of the fluorescence at 610 nm after excitation at 450 nm. An example gel can be found in Fig. 1. 13. Process the image with the ImageQuant TL software and follow the instructions of the software. (Define lanes, proteins bands, and baseline correction; export the intensity of the protein bands for each IgM polymer). 14. Set the sum of all intensities to 100% and calculate the fraction of each polymer.

Analysis of IgM Quality Attributes

3.2 Characterization of the Polymer Distribution of the Purified Product by SEC-HPLC

299

1. Filter all solutions: Use, for example, the Nalgene™ bottle top filter with Supor 450 Membrane Disc Filters for the buffer solution and the Ultrafree® HV centrifugal filters for the samples. 2. Connect the ACQUITY UPLC Protein SEC Column to the HPLC system. 3. Wash the column with filtered ddH2O. 4. Equilibrate the column with running buffer. 5. Increase the flow rate slowly to 0.3 mL/min. 6. Inject 20 μL running buffer for baseline correction; wait for 20 min (see Note 9). 7. Inject 20 μL sample (or reference) and separate the protein sample for 20 min. 8. Use the software LabSolutions (Shimadzu) for evaluation. 9. Assign the peaks of the reference standard to the corresponding molecular weight to estimate the molecular weight of the IgM polymers in the IgM sample (Fig. 2). 10. Depending on the molecular weight, estimate the type of polymer (monomer, dimer, trimer, tetramer, pentamer, hexamer, etc.) (see Notes 10 and 11). 11. Calculate the peak area of each polymer peak with the software LabSolutions, set the sum of all peak areas to 100%, and calculate the fraction of each polymer.

3.3 Analysis of Remaining DNA Content

1. Use the chromatograms obtained with SEC-HPLC and the evaluation software LabSolutions (Shimadzu). Monitor the UV-signal at 280 and 254 nm.

Fig. 2 Polymer distribution in the purified IgM preparation. SEC chromatogram at 280 nm of purified, CHO-derived IgM012_GL cultivated (continuous line) and Bio-Rad Gel filtration standard (dashed line)

300

Julia Hennicke and Renate Kunert

Fig. 3 Nucleic acid content in the purified IgM preparation. (a) SEC chromatogram at 280 nm (solid) and 254 nm (dashed) of purified, CHO-derived IgM012_GL. (b) Peak spectra at the peak maxima after SEC-HPLC for identification of nucleic acid and protein fractions. Solid line, peak spectrum at retention time 6.63 min identified as HMWA NA; dotted line, peak spectrum at retention time 9.39 min identified as pentamer; dashed line, peak spectrum at retention time 11.98 min identified as dimer

2. If there are nucleic acids in the preparation, either peaks for high molecular weight aggregates (low retention volume) or peaks for small fragments (high retention volume) are observed. In Fig. 2, high molecular weight aggregates are visible at a retention volume of 6–8 min. To differentiate between the peaks of IgM polymers and nucleic acid contamination, compare the ratio of the absorption at 280 and 254 nm at the peak maximum (see Note 12). For nucleic acid contamination, the A254/A280 ratio is ~2, whereas the IgM peaks have a ratio of ~0.5 (Fig. 3a). Additionally, the peak spectra can identify nucleic acid contamination (Fig. 3b). 3. Calculate the peak areas of nucleic acid peaks (use the 254 nm chromatogram) and IgM peaks (use the 280 nm chromatogram) with the software LabSolutions. 4. Correct the peak areas by the extinction coefficient (see Notes 13 and 14). A (a) A corr ¼ εmass

5. Calculate the fraction of nucleic acids (NA) according the following formula: AcorrNA (a) c ðNAÞ ¼ AcorrNA þA corrIgM

3.4 Analysis of the Glycosylation

The glycosylation was analyzed by ProGlycAn (http://www. proglycan.com). Briefly, the glycans were enzymatically deglycosylated and analyzed following the instructions of the “released N-glycan assay” (Supelco® Analytical).

Analysis of IgM Quality Attributes

4

301

Notes 1. Recombinant IgM production was reported for other cell lines than CHO, e.g., hybridomas, NS0, Per.C6, or N. benthamiana [16–19]. Therefore, IgM containing supernatant of other host cell lines could be used. 2. The reference is used to identify the protein bands as particular IgM polymers. Therefore, previous knowledge about the expected polymer pattern of the IgM of interest is advantageous. Depending on the origin and composition of the sample, also the host cell proteins will be visible with this technique. 3. Other fluorescence laser scanners can be used alternatively. 4. The IgM can be purified with affinity chromatography. We published the affinity purification strategy in Hennicke et al. [8]. 5. Alternatively, other gel filtration standards such as BEH450 SEC Protein Standard Mix (Waters) can be used. 6. The sensitivity of the SYPRO® Ruby Protein Gel Stain is comparable to silver staining [20]. Therefore, 100–1000 ng IgM sample should give an evaluable result. 7. Alternatively, the fixing step could be performed once for 60 min. 8. The SYPRO® Ruby Protein Gel Stain is a fluorescence dye and should be protected from light to prevent photo bleaching. Use dark staining containers or wrap the container with aluminum foil. 9. We would recommend performing a buffer run between all samples. 10. It is useful to have a hint how many and which polymers are expected. To obtain this, a SDS-PAGE (as described in Subheading 3.1) can be performed as pre-experiment. 11. If a MALS detector is coupled to the HPLC system, the correct molecular weight can be measured with the software ASTRA 6. 12. The absorbance of nucleic acids was measured at 254 nm since aromatic amino acids absorb less at 254 nm than 260 nm while maintaining nearly equivalent nucleic acid absorbance. 13. The extinction coefficient of nucleic acids and proteins differ. The peak areas need to be corrected by the extinction coefficient, because nucleic acids absorb more light compared to a similar amount of protein. 14. The extinction coefficient of ε ¼ 20 g1 L cm1 was used for nucleic acids. The extinction coefficient of the IgM can be

302

Julia Hennicke and Renate Kunert

calculated with online tools such as the ExPASy tool [21], if the primary sequence is known. Alternatively, the extinction coefficient of polyclonal IgM is often estimated with ε ¼ 1.18 g1 L cm1. References 1. Vollmers HP, Br€andlein S (2009) Natural human immunoglobulins in cancer immunotherapy. Immunotherapy 1:241–248 2. Kaveri SV, Silverman GJ, Bayry J (2012) Natural IgM in immune equilibrium and harnessing their therapeutic potential. J Immunol 188:939–945 3. Brewer JW, Randall TD, Parkhouse RME et al (1994) Mechanism and subcellular localization of secretory IgM polymer assembly. J Biol Chem 269:17338–17348 4. Azuma Y, Ishikawa Y, Kawai S et al (2007) Recombinant human hexamer-dominant IgM monoclonal antibody to ganglioside GM3 for treatment of melanoma. Clin Cancer Res 13:2745–2750 5. Vorauer-Uhl K, Wallner J, Lhota G et al (2010) IgM characterization directly performed in crude culture supernatants by a new simple electrophoretic method. J Immunol Methods 359:21–27 6. Brandtzaeg P, Prydz H (1984) Direct evidence for an integrated function of J chain and secretory component in epithelial transport of immunoglobulins. Nature 311:71–73 7. Taylor B, Wright JF, Arya S et al (1994) C1q binding properties of monomer and polymer forms of mouse IgM mu-chain variants. Pro544Gly and Pro434Ala. J Immunol 153:5303–5313 8. Hennicke J, Lastin AM, Reinhart D et al (2017) Glycan profile of CHO derived IgM purified by highly efficient single step affinity chromatography. Anal Biochem 539:162–166 9. Gagnon P, Hensel F, Richieri R (2008) Purification of IgM monoclonal antibodies. Bio Pharm Int. RP0038 10. Gan HT, Lee J, Latiff SMA et al (2013) Characterization and removal of aggregates formed by nonspecific interaction of IgM monoclonal antibodies with chromatin catabolites during cell culture production. J Chromatogr A 1291:33–40 11. Gagnon P, Hensel F, Lee S et al (2011) Chromatographic behavior of IgM:DNA complexes. J Chromatogr A 1218:2405–2412 12. Higel F, Seidl A, So¨rgel F et al (2016) N-glycosylation heterogeneity and the

influence on structure, function and pharmacokinetics of monoclonal antibodies and Fc fusion proteins. Eur J Pharm Biopharm 100:94–100 13. Arnold JN, Wormald MR, Suter DM et al (2005) Human serum IgM glycosylation: identification of glycoforms that can bind to Mannan-binding lectin. J Biol Chem 280:29080–29087 14. Muraoka S, Shulman MJ (1989) Structural requirements for IgM assembly and cytolytic activity. Effects of mutations in the oligosaccharide acceptor site at Asn402. J Immunol 142:695–701 15. Wolbank S, Kunert R, Stiegler G et al (2003) Characterization of human class-switched polymeric (Immunoglobulin M [IgM] and IgA) anti-human immunodeficiency virus type 1 antibodies 2F5 and 2G12. J Virol 77:4095–4103 16. Tchoudakova A, Hensel F, Murillo A et al (2009) High level expression of functional human IgMs in human PER.C6® cells. MAbs 1:163–171 17. Loos A, Gruber C, Altmann F et al (2014) Expression and glycoengineering of functionally active heteromultimeric IgM in plants. Proc Natl Acad Sci 111:6263–6268 18. Sugahara T, Yano S, Sasaki T (2003) High IgM production by human-human hybridoma HB4C5 cells cultured in microtubes. Biosci Biotechnol Biochem 67:393–395 19. Gilmour JEM, Pittman S, Nesbitt R et al (2008) Effect of the presence or absence of J chain on expression of recombinant anti-Kell immunoglobulin M. Transfus Med 18:167–174 20. Berggren K, Chernokalskaya E, Steinberg TH et al (2000) Background-free, high sensitivity staining of proteins in oneand two-dimensional sodium dodecyl sulfatepolyacrylamide gels using a luminescent ruthenium complex. Electrophoresis 21:2509–2521 21. Gasteiger E, Hoogland A, Gattiker MR et al (2005) Protein identification and analysis tools on the ExPASY server. In: The proteomics protocols handbook. Humana Press, New York, pp 571–607

Chapter 18 Raman Trapping Microscopy for Non-invasive Analysis of Biological Samples Hesham K. Yosef and Karin Schu¨tze Abstract Raman microscopy is an emerging tool in biomedicine. It provides label-free and non-invasive analysis of biological cells. Due to its high biochemical specificity, Raman spectroscopy can be used to acquire spectral fingerprints that allow characterizing cells types and states. Here, we present a methodological approach for implementing Raman microscopy in skin cell measurements. Raman spectra can clearly identify keratinocytes, fibroblasts, and melanocytes cells that are involved in the production of autologous skin grafts. Consequently, Raman microscopy is a promising tool that can be used to analyze single cells and to test the quality of therapeutic cell products. Key words Raman trapping microscopy (RTM), Label-free analysis, Raman spectrum, Non-invasive analysis, Skin cell products, Principal component analysis

1

Introduction Raman scattering is a phenomenon that occurs when incident photons of light hit a biomolecule and integrate within its chemical bonding. Molecular vibrations are induced and the photons are emitted with a different energy as they become in-elastically scattered. These scattered photons are carrying the specific chemical information of the targeted biomolecule [1]. Each molecular vibration contributes to a spectral sum of the entire cell that is consequently as characteristic as a fingerprint. The successive instrumental developments over the last decades have granted the Raman spectroscopy access into the microscale biomedical applications such as analysis of cells and tissues [2–5]. By introducing the confocal Raman microscopy, Raman spectra can be collected from 1 μm3 volume of the analyzed sample, which enables the 3D measurements of cells and tissues [6, 7]. The integration of concurrent laser trapping with Raman microscopy has provided the potential to capture and control small cells and particles that are suspended in solutions,

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_18, © Springer Science+Business Media, LLC, part of Springer Nature 2020

303

304

Hesham K. Yosef and Karin Schu¨tze

simultaneously while conducting the Raman analysis. This implementation has offered new opportunities in cell sorting and selective biochemical analysis [8–11]. Despite the numerous advantages of Raman spectroscopy such as fast cell analysis and diseased cell detection, it has not been yet introduced into routine biological applications and medical diagnosis. Since Raman spectrometers are complex optomechanical instruments, it requires expert knowledge to measure, analyze, and interpret the acquired Raman spectra. An important step toward routine Raman spectroscopy was the recent development of an easy-to-handle Raman microscopic platform containing the complex Raman module working in the back as a “black box” (BioRam®, CellTool GmbH, Tutzing, Germany). Additionally, the availability of a simple data analysis software that implements routine statistic algorithms for depicting the spectral results was an unmet need. The newly emerged Raman Statistical Evaluation Software (CT-RamSES®, CellTool GmbH) has implemented a user-friendly graphical interface. This software can import, process, and display the spectral results at the touch of a button. Thus, it helps the biologists and physicians to get fast and meaningful biological Raman results. Herewith, unique spectral information can easily be compared with data from other standard biological assays such as immunofluorescence, DNA microarrays, and mass spectroscopy. In this chapter, we will demonstrate the potential of Raman trapping microscopy to characterize different skin cell types, which are used for the manufacturing of skin graft products. Skin grafts are used as a replacement after surgical interventions or acute skin trauma. They are produced by extracting healthy cells from the donor to develop a cell-matrix based on keratinocytes and fibroblasts mingled in the hydrogel to construct epidermis and dermis layers, respectively [12–14]. These cells have to be expanded in separate in vitro cultures prior to application on the skin graft matrix. Furthermore, the cultured cells have to be examined prior and post the application on the skin graft matrix to detect possible cross-contamination, which can affect the integrity of the skin graft product. Currently, traditional histology and immunohistochemistry (fluorescence-activated cell sorting (FACS) analysis and DNA count) are performed during the graft production to discriminate cell populations, quantifications, and cross-contamination of the skin grafts. However, these techniques are invasive and cannot be applied without affecting the cell viability [15]. Conversely, Raman microscopy is a non-invasive and label-free technique that can provide clear differentiation between different cell types, without applying antibodies and biomarkers or affecting the cell viability. In addition, Raman spectra can provide information about the state of the cells, i.e., whether they are healthy and functional or started to decay [3, 16]. Furthermore, a statistically reliable Raman data can

Raman Trapping Microscopy for Biological Samples

305

Fig. 1 Bright field micrograph of keratinocytes (a), fibroblasts (b), and melanocytes (c). The Raman mean spectra (d) of keratinocytes (1), fibroblasts (2), and melanocytes (3) are collected from 300 keratinocytes (3 donors), 300 fibroblasts (3 donors), and 100 melanocytes cells (1 donor). Raman measurements were collected using 785-nm laser of the Raman Trapping Microscope BioRam® (CellTool GmbH, Tutzing, Germany). The spectra were smoothed, baseline corrected, and unit-vector normalized

be acquired using a smaller number of cells (less than 100 cells) compared to other techniques such as mass spectrometry and FACS analysis, which consume a larger amount of cells to acquire reliable results. Raman spectral bands can reveal significant chemical information of the macromolecules in the biological cells. Each spectral band can be assigned to a specific vibration of a molecule in the cell. This can be demonstrated in the Raman mean spectra collected from suspended keratinocytes, melanocytes, and fibroblasts cells as depicted in Fig. 1. Raman bands are showing differences in spectral intensities and positions between the three cell types as follows: around 489 (glycogen), 542 (protein: cysteine), 719 (DNA/RNA: adenine; lipids), 756 (proteins: tryptophan), 876 (collagen; lipids), 912 (RNA; carbohydrates), 1002 (proteins: phenylalanine), 1044 (collagen: proline), 1130 (lipids), 1254 (lipids; protein; DNA/RNA), 1345 (proteins; carbohydrates) (see Note 1), 1453 (protein; lipids), 1495 (proteins; DNA/RNA: adenine, guanine), 1618 (proteins: phenylalanine, tyrosine, tryptophan), and 1657 cm 1 (protein) [17, 18]. To visualize these differences, the

306

Hesham K. Yosef and Karin Schu¨tze

Fig. 2 PCA score plot of keratinocytes, fibroblasts, and melanocytes

principal component analysis (PCA) is conducted on the Raman spectral data. PCA is a useful multivariate analysis tool to identify the similarities and differences between the analyzed data sets and can group similar entities based on its PCA score. Moreover, the calculated loadings of principal components (PCs) can illustrate the significant spectral variations that were used to differentiate the data sets [19]. By applying PCA on the Raman spectral results, a clear separation of keratinocytes, melanocytes, and fibroblasts is achieved, as depicted in the score plot (Fig. 2). Furthermore, the loadings of PC1, PC2, and PC3 are displaying the magnitude of spectral variation between the data sets. In case of PC1, for example, positive variations in bands around 719, 756, 876, 912, 1002, 1044, 1130, 1254, 1495, and 1654 cm 1 are contributing to the classification (Fig. 3a), which are based on changes of protein, nucleic acids, carbohydrates, and lipid expressions across the three cell types. These results demonstrate the great potential of Raman microscopy to discriminate and classify the type and biochemical status of fixed cells, which can be extended to measure live cells directly within the skin graft matrix [15]. In this chapter, a comprehensive description of the typical Raman measurement protocols of biological cells is presented, starting from materials and cell preparations, followed by Raman microscopic measurements, and, finally, data analysis using the multivariate analysis of PCA.

2 2.1

Materials Cells Culture

1. Complete culture medium (Dulbecco’s modified Eagle medium) supplemented by 10% fetal calf serum, 4 mM L-alanyl-L-glutamine, 1 mM sodium pyruvate, and 5 μg/ml gentamicin. 2. 1 Phosphate-buffered saline (PBS), and Trypsin-EDTA (TE).

Raman Trapping Microscopy for Biological Samples

307

Fig. 3 PCA loading plot of PC1 (a), PC2 (b), and PC3 (c), indicating major spectral variation between the data sets of keratinocytes, fibroblasts, and melanocytes

3. Chemical fixation solution of 4% paraformaldehyde (PFA) in 1 PBS, 7.2 pH: Weigh 40 g paraformaldehyde in 1 l flask and add 1 PBS till 800 ml is reached. Heat the solution while steering to 60  C. Add 30 μL of 5 M NaOH. If necessary, increase the volume of NaOH until paraformaldehyde is completely dissolved (see Note 2), adjust the pH of the solution to 7.2 by adding drops of 5 M HCl. Fill up the flask to the final volume of 1 l with 1 PBS. Filter the solution through a 0.45μm pore-size membrane filter. Cool on ice to 4  C prior using and at 20  C for longer storage (see Note 2). 4. Skin cells of keratinocytes, fibroblasts, and melanocytes, isolated from donors (see Note 3). 5. Centrifuge, cell incubator, and fridges (4  C and 2.2

Raman Substrate

20  C).

1. Micro-wells plates or microfluidic chips such as micro- channels slides equipped with glass bottom (0.17-μm-thick borosilicate cover glass) fitted with the dimensions of the stage (Fig. 4c) of the Raman microscope (see Note 4) for suspended cell measurements. 2. Silicon substrate for verifying the spectrometer calibration (see Note 5).

308

Hesham K. Yosef and Karin Schu¨tze

Fig. 4 Setting up the microscope for Raman measurement: (a) opening the microscope chamber, (b) adding 100 μL of water on the selected 60 water immersion objective, (c) fixing the micro-wells plate of glass bottom on the microscope stage, (d) removing the micro-wells plate cover and adding 200 μL of fixed cell suspension into a selected micro-well, (e) covering the micro- wells plate and the stage is steered in XY to the selected well and steered in Z direction till cells are in focus. (f) Closing the microscope chamber

2.3 Raman Microscope

Raman microscope consisting of: 1. An inverted digitized light microscopic system with visualization camera (see Note 6). 2. Objective lenses of low (10, 20, or 40) and high (60 water immersion/numerical aperture of 1–1.2 at least) magnifications (see Note 7). 3. Microscopic motorized scanning stage with x/y range: 180  120 mm, repeatable relocation accuracy of 1 μm + 0.01% of the running distance, and z-focus range of 45 mm with repeatable relocation accuracy of 0.5 μm + 0.01% of the running distance. 4. Raman excitation laser of 785 nm with at least 100-mW output power (see Note 8), laser coupling, and optical components (gratings and mirrors). 5. Highly sensitive Raman spectrograph with a frequency area of 100–3500 wavenumbers (or 100–2000 wavenumbers), optimized for the laser wavelength of 785 nm. 6. Ultrasensitive CCD detection camera with 1024  127 pixels and pixel size of 26  26 μm, working with Peltier cooling at 80  C. 7. Microscope control interface and software.

Raman Trapping Microscopy for Biological Samples

2.4

3

Data Analysis

309

1. A commercial software package (CT-RamSES®, CellTool GmbH, Tutzing) is used as a straightforward statistical analysis software based on the Python platform. The CT-RamSES was used for processing of the Raman spectra (smoothing, background correction, and normalization) as well as for PCA multivariate analysis and data visualization (see Note 9).

Methods 1. Cells are cultured and handled under sterilized conditions until the Raman measurement is conducted (see Note 10), including all solutions, culture plates, and microscopic slides. All solutions and chemicals should be preserved in the temperature conditions that are recommended by the manufacturer. 2. To conduct Raman analysis, cells should be subcultured, chemically fixed, and seeded on the Raman measurement-substrate.

3.1

Cell Preparations

1. Maintain the isolated keratinocytes, fibroblasts, and melanocytes cells in separate cultures (37  C/5% CO2) in 10 mL of the culture medium till they reach a confluence of 70–80%, discard the medium and add 3 mL of TE for 3 min/37 C, add 7 mL of medium on the cells, and harvest them from the plastic culturing dish into a Falcon tube and centrifuge at 250 g/3 min followed by discarding the supernatant. Next, suspend the cell pellets in 1 PBS and dilute it to 2  106 cells/ml. 2. Centrifuge the cells at 250 g/3 min, decant the supernatant, and then fix the cells by adding 2–5 mL of 4% PFA for 5 min, followed by 3 times of 5 min washing steps using 10 mL PBS, separated by centrifuge steps (250 g/3 min) and decantation of the PBS. Then store the fixed cells immersed in PBS at 4  C until conducting the Raman measurements. 3. Add the cells suspended in PBS buffer directly on the microwell plates or fluidic chips (see Note 11).

3.2 Raman Measurements

1. Turn on the excitation laser source and let it for a few minutes to stabilize. 2. Turn on the Raman microscopic system, attached computer, and the control software using the sequence recommended by the manufacturer. 3. Before starting the Raman measurements, check the spectrometer calibration and laser power, by acquiring the specific Raman band position and intensity of the silicon substrate (at 521 cm 1 band), using the same steps of the cell measurements (described in the next parts). Record the intensity and position of the silicon band for future correspondences (see Note 12).

310

Hesham K. Yosef and Karin Schu¨tze

3.3 Measurement Parameters

1. Carefully disinfect and clean the Raman stage and objective using lens cleaning tissues dipped in 70% ethanol (see Note 13). 2. Select 60 water immersion objective, adjust its glass correction collar to 0.17 μm (see Note 14), and add 100 μl of deionized water on the top of the objective using micropipette (Fig. 4b). 3. In the case of acquiring single spectra of the cells, optimize the laser exposure time and accumulation/repetitions until a clear Raman spectral band of the lowest possible exposure time is acquired. In the case of 80 mW/785 nm laser, adjust the exposure to 3 s of 10 accumulations (see Note 15). In the case of Raman imaging, select the size of the image, pixel resolution, and laser exposure. It is a compromise between image resolution and time of the measurement. 4. In the case of live cell measurements, an environmental controlling setup (incubator) is fitted on the microscope stage, adjust the setup to 37  C, and use 785-nm excitation laser for the measurement.

3.4 Raman Measurements

1. Open the microscope chamber and select the 60 water immersion objective, add 100 μL of water on the top of the objective lens, fix the micro-wells plate of glass bottom on the microscope stage (see Note 16), remove the micro-wells plate cover and add 200 μL of fixed cells/PBS suspension into a micro-well (see Note 17), cover the wells plate, and steer the stage in XY directions to center the selected well and close the microscope chamber (Fig. 4) (see Note 18). 2. Switch the camera mode to bright field and steer the stage in vertical Z direction until bringing the cells in focus (Fig. 5a). 3. Use the bright field mode of the camera to select the point of interest such as cell cytoplasm (Fig. 5b1), or a raster of points to conduct automatic pre-defined Raman measurements, by marking many measurement points of different cells using the pin-marking option of the software (Fig. 5b3). 4. Use XY control, bring the cell or the desired measurement point to the center of the camera view to be aligned with the laser illumination point, then switch to the dark field mode, open the laser shutter, and activate the live mode spectral acquisition by the controlling software. 5. Move the stage in the vertical Z direction while monitoring the live feed spectral changes until acquiring the highest intensities possible of Raman signals of the cell (see Note 19). 6. Press start Raman measurements. 7. After the measurement is finished, repeat the previous steps for further measurements or other samples (see Note 20).

Raman Trapping Microscopy for Biological Samples

311

Fig. 5 (a) Raman microscope controlling software in bright field mode showing keratinocytes cells in focus. (b) Different measurement points of keratinocytes, by centering the cytoplasm (1), or the nucleus (2) under the laser illumination point, or pin-marking several points (3)

312

Hesham K. Yosef and Karin Schu¨tze

8. Save the spectral results with the date and the cell type, and export it to the data analysis platform to conduct data processing. 3.5 Automated Spectral Acquisition

1. Use the automated Raman spectral acquisition to collect a large number of measurements in case of high cell density. 2. Choose the automated scanning mode, select the starting and endpoint (upper right and lower right edges of the region of interest) (Fig. 6a), and determine the distance between measurement points of about one cell size (10–20 μm). 3. Steer the stage control in XY to the central region of the point raster, open the laser shutter, and activate the live mode spectral acquisition. 4. Move the stage in the vertical Z direction while monitoring the live feed spectral changes until acquiring the highest intensities possible of Raman signal of the cell in the central area of the raster. 5. Run the measurement (Fig. 6b); the spectra are collected from each measurement point in the selected area, subsequently, and are displayed in the work space of the software (Fig. 6c). 6. Save the spectral results with the date and the cell type (see Note 21), and export it to the data analysis platform to conduct data processing.

Fig. 6 Raman microscope controlling software in the dark field mode for collecting automated spectral acquisition: (a) selected starting and endpoint (upper right and lower right edges of the region of interest), (b) running the measurements, and (c) all points are measured and the corresponding Raman spectra are displayed in the work space marked by the red rectangle

Raman Trapping Microscopy for Biological Samples

3.6

Data Analysis

313

1. Import the Raman data sets using the analysis software to conduct typical processing steps of Raman spectra. The following steps are implemented in the CT-RamSES software and are automatically performed. 2. Correct the spectral baseline by an asymmetric least squares fit, followed by removing spikes from the spectra using the smoothing median filter (window 5), and then finally normalize the spectra using a unit-vector normalization as displayed in Fig. 7a (see Note 22). 3. Calculate Raman mean spectra as well as the error bar of each data set, using the averaging function of the analysis software (Fig. 7b). 4. Apply PCA for a clearer representation of the spectral data and to identify similarities or differences between all data sets acquired from Raman measurements of two or more cell samples (Figs. 2 and 3). The histogram of the PC1 can clearly display the differences between the three cell types (Fig. 8).

4

Notes 1. The glass substrate can contribute to the band around 1345 cm 1. Carefully adjust the z layer focus during the measurements on the cell away as much possible from the glass substrate, without reducing the Raman signal of the cell. 2. Some particles of PFA may not fully dissolve and needed to be filtered afterward. PFA should be always freshly prepared or thawed from 20  C stock before fixation. 3. Other types of pathological cells or cell lines can also be measured by Raman trapping microscopy using the same protocol. 4. Different kinds of substrates and/or well plates can be used; commonly CaF2, glass, and quartz are used based on the application and the desired detection spectral bands. 5. Other substrates can be used to check the spectrometer calibrations such as polystyrene. 6. Inverted microscopes are more convenient in biological applications than upright microscopes, allowing easy access to the sample from above to mount a cell incubator and permitting laser trapping of cells from below. Furthermore, it reduces bacterial contaminations that can arise in an upright setup, in which it requires removing the cover of culture or cell plate in order to immerse the objective. Moreover, an optional fluorescence module can be added for comparison with Raman results.

314

Hesham K. Yosef and Karin Schu¨tze

Fig. 7 Processing of Raman spectra of keratinocytes: (a) uncorrected (1) and corrected (2) baseline spectrum; (b) processed 100 Raman spectra (1) and calculated mean spectrum with error bar represented by shading (2)

7. Using the 60 objective of the high numerical aperture (NA > 1) can enhance laser trapping of suspended cells. Despite that large cells are not completely trapped, the trapping laser can stabilize the suspended cells during the Raman measurements. Moreover, water immersion objective can acquire better Raman spectral and image quality than the dry counterpart. This is because water immersion can minimize refractions and reflections of light photons released from the cells. 8. Other laser excitation wavelengths such as 633 and 532 nm can be used, based on the application. In the case of laser trapping

Raman Trapping Microscopy for Biological Samples

315

Fig. 8 PC1 histogram revealing the different distribution of spectral variables between fibroblasts, keratinocytes, and melanocytes cells

and live cell measurements, 785-nm excitation laser is preferred, as shorter wavelength lasers can induce cell death [20]. 9. Other multivariate analysis methods can be used such as hierarchical cluster and K-means analysis. 10. Contamination of cells with biological or chemical substances such as bacteria or dirt will add artifacts to the Raman measurements of the cells. 11. Non-adherent cells can be also attached to the substrate surface using cytospinning setup [6]. 12. The calibration check does not have to be conducted before every measurement, but on a weekly or monthly basis based on the spectrometer stability. After recording Raman band intensity and position of the Si substrate, if there is no significant deviation in the values compared with previous records, proceed with the Raman measurements. Otherwise, calibration of the spectrometer must be conducted based on the manufacturer guidelines. Alternative substrates can be also used such as polystyrene (bands at 621, 1004, 1036, and 1609 cm 1). 13. The cleaning tissue dipped in 70% ethanol should be used with great care that no ethanol is dripping, so it does not harm any electrical or polymeric components of the microscopic system. 14. The glass correction collar is adjusted to a value equal to the thickness of the glass substrate to compensate the image

316

Hesham K. Yosef and Karin Schu¨tze

degradation because of the laser aberration produced by the difference of refractive indices between the air and glass. 15. Setting exposure time and accumulation values depends on the expected difference between the measured cell types and states. 16. Fix the slide firmly on the stage in a perfect horizontal orientation (Fig. 4c); any vertical tilting can produce variations of the confocal focus layers across the slide and can be disturbing in case of automatic acquisition of raster measurements. 17. Since Raman spectra of dried cells show distorted features, the fixed cells are preferably measured in PBS buffer [21]. 18. Keep tracking the PBS level on the cell sample between the Raman measurements, since it can evaporate because of the light exposure and room temperature leading to crystallization of the buffer salt on the cells. If needed, add more PBS carefully using micropipette. 19. An important step in confocal Raman systems is to determine the best Z focus layer that shows the highest Raman signal of the cell and the lowest possible contribution of the substrate spectral bands. 20. It is recommended to acquire at least 60 measurements/sample to collect a suitable sample-representative Raman data of significant statistical value. 21. The Raman data are displayed on the work space after measurements; the data is saved as Excel or other file formats, containing all the measurements and instrumental parameters, such as objective type, excitation wavelength, and laser accumulation time. 22. Other types of normalization can be used for in-depth comparison of the spectra, for example, normalization on Amid I (1660 cm 1), CH deformation (1453 cm 1), and phenylalanine (1002 cm 1). By applying the normalization on a specific band, the relative intensities of other bands in the Raman spectra can be compared between the different normalized data sets by calculating the difference spectrum between two mean spectra.

Acknowledgment We would like to thank CUTISS AG (Zu¨rich, Switzerland) for providing the skin cell samples. We would like also to thank Sarvesh Ghorpade, Dr. Heidi Kremling, and Florian Zunhammer for the support.

Raman Trapping Microscopy for Biological Samples

317

References 1. Raman CV, Krishnan KS (1928) A new type of secondary radiation. Nature 121:501–502. https://doi.org/10.1038/121501c0 2. Steinke M, Gross R, Walles H et al (2014) An engineered 3D human airway mucosa model based on an SIS scaffold. Biomaterials 35:7355–7362. https://doi.org/10.1016/j. biomaterials.2014.05.031 3. Buchacher T, Wiesinger-Mayr H, Vierlinger K et al (2014) Human blood monocytes support persistence, but not replication of the intracellular pathogen C. pneumoniae. BMC Immunol 15:60. https://doi.org/10.1186/s12865014-0060-1 4. Yosef HK, Frick T, Hammoud MK et al (2018) Exploring the efficacy and cellular uptake of sorafenib in colon cancer cells by Raman micro-spectroscopy. The Analyst 143:6069–6078. https://doi.org/10.1039/ C8AN02029A 5. Krafft C, Dietzek B, Popp J (2009) Raman and CARS microspectroscopy of cells and tissues. The Analyst 134:1046. https://doi.org/10. 1039/b822354h 6. Yosef HK, Krauß SD, Lechtonen T et al (2017) Noninvasive diagnosis of high-grade urothelial carcinoma in urine by raman spectral imaging. Anal Chem 89:6893–6899. https://doi.org/ 10.1021/acs.analchem.7b01403 7. Charwat V, Schu¨tze K, Holnthoner W et al (2015) Potential and limitations of microscopy and Raman spectroscopy for live-cell analysis of 3D cell cultures. J Biotechnol 205:70–81. https://doi.org/10.1016/j.jbiotec.2015.02. 007 8. Harvey TJ, Hughes C, Ward AD et al (2009) Classification of fixed urological cells using Raman tweezers. J Biophotonics 2:47–69. https://doi.org/10.1002/jbio.200810061 9. Huang WE, Ward AD, Whiteley AS (2009) Raman tweezers sorting of single microbial cells. Environ Microbiol Rep 1:44–49. https://doi.org/10.1111/j.1758-2229.2008. 00002.x 10. De Luca AC, Rusciano G, Ciancia R et al (2008) Spectroscopical and mechanical characterization of normal and thalassemic red blood cells by Raman Tweezers. Opt Express 16:7943. https://doi.org/10.1364/OE.16. 007943 11. Steinke M, Zunhammer F, Chatzopoulou EI et al (2018) Rapid analysis of cell-nanoparticle interactions using single-cell raman trapping

microscopy. Angew Chem Int Ed 57:4946–4950. https://doi.org/10.1002/ anie.201713151 12. Larouche D, Cantin-Warren L, Desgagne´ M et al (2016) Improved methods to produce tissue-engineered skin substitutes suitable for the permanent closure of full-thickness skin injuries. BioRes Open Access 5:320–329. https://doi.org/10.1089/biores.2016.0036 13. MacNeil S (2007) Progress and opportunities for tissue-engineered skin. Nature 445:874–880. https://doi.org/10.1038/ nature05664 14. Marino D, Luginbuhl J, Scola S et al (2014) Bioengineering dermo-epidermal skin grafts with blood and lymphatic capillaries. Sci Transl Med 6:221ra14. https://doi.org/10.1126/ scitranslmed.3006894 15. Pudlas M, Koch S, Bolwien C et al (2011) Raman spectroscopy: a noninvasive analysis tool for the discrimination of human skin cells. Tissue Eng Part C Methods 17:1027–1040. https://doi.org/10.1089/ ten.tec.2011.0082 16. Koch S, Walles H, Krause KH et al (2013) Novel cell identification: markerfree and suitable for living cells. In: Deckert V, Ramanujam N (eds) . Munich, Germany, p 87980J 17. Lyng FM, Traynor D, Ramos IRM et al (2015) Raman spectroscopy for screening and diagnosis of cervical cancer. Anal Bioanal Chem 407:8279–8289. https://doi.org/10.1007/ s00216-015-8946-1 18. Talari ACS, Movasaghi Z, Rehman S, Rehman I (2015) Raman spectroscopy of biological tissues. Appl Spectrosc Rev 50:46–111. https:// doi.org/10.1080/05704928.2014.923902 19. Bonnier F, Byrne HJ (2012) Understanding the molecular information contained in principal component analysis of vibrational spectra of biological systems. The Analyst 137:322–332. https://doi.org/10.1039/C1AN15821J 20. Notingher I, Verrier S, Romanska H et al (2002) In situ characterisation of living cells by Raman spectroscopy. Spectroscopy 16:43–51. https://doi.org/10.1155/2002/ 408381 21. Short KW, Carpenter S, Freyer JP, Mourant JR (2005) Raman spectroscopy detects biochemical changes due to proliferation in mammalian cell cultures. Biophys J 88:4274–4288. https://doi.org/10.1529/biophysj.103. 038604

Chapter 19 An Optical Biosensor for Continuous Glucose Monitoring in Animal Cell Cultures Mario Lederle, Mircea Tric, Claudio Packi, Tobias Werner, and Philipp Wiedemann Abstract Biosensors for continuous glucose monitoring in bioreactors could provide a valuable tool for optimizing culture conditions in biotechnological applications. We have developed an optical biosensor for long-term continuous glucose monitoring and demonstrated a tight glucose level control during cell culture in disposable bioreactors. The in-line sensor is based on a commercially available oxygen sensor that is coated with cross-linked glucose oxidase (GOD). The dynamic range of the sensor was tuned by a hydrophilic perforated diffusion membrane with an optimized permeability for glucose and oxygen. The biosensor was thoroughly characterized by experimental data and numerical simulations, which enabled insights into the internal concentration profile of the deactivating by-product hydrogen peroxide. The simulations were carried out with a one-dimensional biosensor model and revealed that, in addition to the internal hydrogen peroxide concentration, the turnover rate of the enzyme GOD plays a crucial role for culture monitoring is an integral part of animal cell cultivation. For several culture parameters, in situ sensors exist; others are predominantly monitored off-line. One important cell culture parameter is glucose concentration. Despite many efforts, there is still a lack of in situ sensors for continuous glucose monitoring. Such biosensors could provide a valuable tool for optimizing culture conditions in biotechnological applications. In this contribution, the manufacture of a long-term stable optical glucose sensor is described which is used to demonstrate glucose level monitoring during cell culture in disposable bioreactors. The in situ sensor is based on a commercially available oxygen sensor that is coated with cross-linked glucose oxidase and a hydrophilic perforated diffusion membrane. Glucose was measured in shake flasks and wave bags with only minor drifts of the sensor sensitivity during batch and fed-batch fermentations. Key words Enzyme-based optical biosensor, Glucose monitoring, Process monitoring, Cell culture

1

Introduction Culture monitoring is essential for the continuous characterization of cultivated animal cells both in the laboratory and in large scale applications. Monitoring can be carried out off-line or at-line, i.e., samples are withdrawn from the culture and analyzed with appropriate techniques in a—closer or more distant—laboratory. Here, care has to be taken to not obscure the original conditions of the

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_19, © Springer Science+Business Media, LLC, part of Springer Nature 2020

319

320

Mario Lederle et al.

sample as they were in the cultivation vessel (e.g., gassing out of CO2 or O2 between sampling and analysis) [1]. Alternatively, process parameters can be monitored online, e.g., via flow injection analysis or in a bypass configuration. Often, it is preferable to monitor parameters in situ, i.e., with sensors directly immersed in the culture broth. This configuration has several advantages; for example, monitoring data is created quickly, at best in real time, and less sampling means less risk of contamination of the culture [1, 2]. Furthermore, it opens up possibilities of process automation—closed control loops rather than manual interaction with the culture as a result of off-line monitoring results. Additionally and in a wider context, extensive process monitoring is crucial in the context of Quality by Design (QbD) approaches, in which analytical capabilities are the cornerstone of measuring process parameters and quality attributes in a validated environment [3]. For several parameters, proven and robust in situ sensors are available, particularly for physicochemical parameters like temperature, dissolved oxygen (DO), and pH value. On the other hand, many physiological parameters like substrate and metabolite concentrations as well as cell density, viability, and recombinant product concentration are still predominantly analyzed off-line or at-line: Although in situ sensors for such parameters would be advantageous, their development has proven to be challenging [2]. The need for development of reliable in situ sensors for preferably all key culture parameters is further emphasized by the now established field of single use bioreactors [4]—with the additional pressure of “low cost” for sensors in this environment. Glucose is often the predominant substrate of mammalian cells in culture [5]. Avoiding glucose limitations and preferably optimizing substrate levels and feed rates is a very important part of process development and optimization. All this relies on feasible methods to monitor the glucose content of a culture. To that end, several techniques have been developed, most of them fall into the category off- or at-line monitoring. Often, they include enzymatic analytical chemistry methods, e.g., employing glucose dehydrogenase or hexokinase and photometric or colorimetric detection of NADPH (commercial examples: glucose assays from standard laboratory providers like Promega or the Cedex Bio Analyzer, Roche). In other systems, glucose oxidase is used to convert glucose to gluconolactone and H2O2 which is amperometrically analyzed (commercial examples: YSI analyzers, Yellow Springs Instrument; or online methods, e.g., from C-CIT Sensors AG and coupled to a dialysis probe: TRACE Analytics). An alternative to the detection methods mentioned above is optical biosensors [6]. They have several advantages, as, e.g., proven by commercially available optical oxygen sensors: They can be pre-integrated in bioreactors—large, e.g., cell bags, as well as small, e.g., shake flasks; they transduce their signal in a

Optical Glucose Sensor

321

contactless manner through transparent surfaces omitting the need for wires leading into the reactor, thereby also reducing contamination risks; and they are reasonably priced, so they can be employed as single use material. We have previously reported the development of a disposable optical glucose biosensor for continuous glucose measurement in suspension cell cultures [7]. The sensor is based on a commercially available optical oxygen sensor (PreSens, Regensburg, Germany) and standard off-the-shelf material. Optical oxygen sensors are usually based on the quenching of luminescence in the presence of oxygen. In fact, they are more attractive than other conventional devices, such as amperometric electrodes because, in general, their design is relatively simple and they do not consume oxygen [8]. In addition to that, luminescence lifetime of a quenchable indicator is an intrinsic property that is independent of fluctuations in light intensity, detector sensitivity, and light path of the optical system. This makes optical sensors superior to conventional sensors that often suffer from baseline drift due to light fluctuation, variation of sensor positioning, or indicator leaching and photo-bleaching. Most of these problems can be overcome by using state-of-the-art luminescence decay time-based sensors [9]. Therefore, we have used this type of sensor to monitor glucose concentrations in several different formats including cell bags and shake flasks. In this contribution, we will describe the production and use of the sensor in detail.

2

Materials

2.1 Materials for Biosensor Fabrication

1. Bovine serum albumin powder. 2. Perforated diffusion barrier membranes (DBM) for limiting glucose diffusion (PreSens GmbH, Regensburg, Germany). 3. Glucose oxidase (from Aspergillus niger; powder). 4. Glutaraldehyde. 5. Glycerine. 6. Silicone glue (e.g., RS692-542; RS Components GmbH, Mo¨rfelden-Walldorf, Germany). 7. Oxygen-sensitive sensor spot type SP-PSt3 (3 and 5 mm; optically isolated; PreSens GmbH, Regensburg, Germany). 8. Optical fiber (POF, 2.5 m; SMA connector; PreSens GmbH, Regensburg, Germany). 9. Silicone tubes (outer diameter preferably slightly larger than 3/5 mm; inner diameter suitable for optical fibers; see Subheading 2.4). 10. Male luer lock plug.

322

Mario Lederle et al.

11. Double sided acrylate adhesive foil (see Note 1). 12. Neodymium mini magnets (e.g., 5  3 mm discs). 13. Micro pipette, tweezers, 1.5 mL Eppendorf cups, scissors. 2.2

Assay Kits

1. Glucose Bio, Glutamine V2 Bio, Lactate Bio, NH3 Bio (assays for Cedex Bio; Roche Diagnostics GmbH, Mannheim, Germany (see Note 2)). 2. Hexokinase Kit.

2.3 Specific Cell Culture Material, Consumables, and Cell Line

1. Disposable cell bag bioreactors (with screw cap). 2. Erlenmeyer flasks with O2 and pH sensors—Sensor Flask SFS-HP5-PSt3 (125 mL polycarbonate; PreSens GmbH, Regensburg, Germany). 3. Suspension-adapted CHO cell and suitable medium (see Note 3). 4. Glucose (glucose monohydrate powder). 5. Penicillin/streptomycin (100  concentrated solution).

2.4 Specific Instrumentation, Equipment, and Software

1. Sensor equipment for shake flask experiments: (a) Shake Flask Reader (SFR) with SFR software (PreSens GmbH, Regensburg, Germany). (b) Four-channel oxygen meter OXY-4 mini with OXY-4 mini version 2.3FB software and polymer optical fibers (POFs) (PreSens GmbH, Regensburg, Germany). 2. Sensor equipment for cell bag experiments: (a) Coaster system (PreSens GmbH, Regensburg, Germany). (b) Fibox 3 with Fibox 3 version 6.02 software (PreSens GmbH, Regensburg, Germany). 3. Shaker. 4. WAVE-type bioreactor system.

3

Methods

3.1 Sensor Fabrication

An overview of the biosensor setup is shown in Fig. 1. It is advisable to carry out all work described below in a laminar flow. Materials should be autoclaved (silicone tubes, Eppendorf cups, pipet tips) or—except the adhesive tape—placed in/disinfected with 90% ethanol (sensors, diffusion membranes, foil for cell bag sensor support, luer lock plugs, scissors, tweezers, etc.) before use. Make sure all materials are dry before you use them.

3.1.1 For Shake Flask Experiments

At this stage, sensors for shake flasks are attached to the top of a silicone tube. The silicone tube is slid over an optic fiber which is

Optical Glucose Sensor

A

323

B Diffusion membrane Biosensor

Enzyme layer Oxygen sensor Silicon glue Silicon tube

Oxygen Sensor

Polycarbontate Foil

Magnet CounterMagnet Bag Foil

Fiber optics Ex

Em

Ex

Em

Two-channel PreSens Coaster Ex

Em

Fig. 1 Biosensor setup (see text for details). In (a), the assembly of a biosensor for the application in shake flasks is shown. An oxygen sensor is placed on a piece of silicone tube by means of adhesive foil and coated with an enzyme layer. A diffusion membrane is mounted on top. The biosensor is sealed with silicone glue. The silicone tube with the sensor attached to the end is put onto an optical fiber installed in a shake flask. In (b), the assembly of biosensors for cell bag experiments is shown. A biosensor and an oxygen sensor are fixed on a piece of transparent foil with silicone glue. For fixation of the sensors inside the bag, two neodymium magnets were mounted on the plastic with silicone glue. After insertion of the sensors inside the bag, two counter magnets were placed outside for fixation. The signal can be read through the transparent bag foil

lead through the screw cap of shake flasks. The flasks with fiber are pre-sterilized by gamma irradiation (18 kGy). Sensor spots that can directly be fixed in bioreactors are described in the section for wave bags below. l

l

l

l

Step 1: A ca. 1 cm piece of silicone tube is placed on a luer lock plug. The top end of the tube is covered with double sided acrylic adhesive foil. A 3-mm optical oxygen sensor is centrally fixed on the tape (Fig. 2a; see Note 4). Step 2: An enzyme solution is produced by solubilizing 10 mg of glucose oxidase and 50 mg of bovine serum albumin (BSA) in 250 μL of PBS. The solution is thoroughly mixed with a micro pipette before the immobilization process is initiated by adding 50 μL of 2.5% glutaraldehyde to the enzyme solution (see Note 5). The solution then has to be used immediately. Step 3: An adequate proportion of the polymerizing enzyme solution (1.5 μL for 3-mm sensors; 2–3 μL for 5-mm sensors) is quickly spread over the carbon black side of the oxygen sensor surface by means of a micro pipette (Fig. 2b). Step 4: A perforated hydrophilic diffusion membrane stamped out with a punch and with a diameter slightly bigger than the oxygen sensor is placed on top of the polymerizing enzyme layer on the oxygen sensor using tweezers (Fig. 2c). The membrane has to cover the sensor fully and must touch the adhesive foil around the edges.

324

Mario Lederle et al.

Fig. 2 (a) Oxygen sensor (black side up) fixed on the top end of a piece of silicone tube by means of double sided acrylic foil. The tube is held in place with a luer lock plug. (b) Polymerizing enzyme solution is pipetted onto the oxygen sensor. (c) The enzyme solution is covered with diffusion membrane. (d) The rim of the biosensor is sealed with silicone glue

l

l

l

3.1.2 For Wave-Type Bag Experiments

l

l

l

l

Step 5: After the polymerization process stopped (5 min), the biosensor rim is sealed with silicone glue to prevent glucose from lateral influx (Fig. 2d). The silicone glue is allowed to dry for 15 min (see Note 6). Step 6: The biosensor is subsequently stored in PBS buffer at 4  C until further usage. The storage stability under these conditions exceeds 4 months. Step 7: Prior to usage, the biosensor can be sterilized with beta irradiation, gamma irradiation (18 kGy), or UV (15 W at 253.7 nm for 1 h) and subsequently applied to the optic fiber in a shake flask under laminar flow (see Note 7). Step 1: Two circular perforated hydrophilic diffusion membranes with a diameter of 6 mm are stamped out with a punch. Step 2: A clean quadratic plastic foil (e.g., material from a cell bag) is prepared. It will be used as a support for four optical oxygen sensor spots that are fixed on the plastic surface by a single drop of silicone glue for each. Step 3: An enzyme solution is produced by solubilizing 10 mg of glucose oxidase and 50 mg of bovine serum albumin (BSA) in 250 μL of PBS and 10 μL of glycerine. The solution is thoroughly mixed with a micro pipette before the immobilization process is initiated by adding 50 μL of 2.5% glutaraldehyde to the enzyme solution (see Note 5). The solution then has to be used immediately. Step 4: 2.5 μL of solution from step 3 is directly pipetted onto two of the O2 sensors (see Note 4) carefully.

Optical Glucose Sensor

325

Fig. 3 Installation of the biosensor in a cell bag. A transparent plastic foil with two biosensors and two oxygen sensors on top was placed in a cell bag through a screw cap port aseptically and fixed with two magnets. With the PreSens coaster seen on the right, the signal was measured through the transparent bag foil l

l

l

l

Step 5: Immediately place 2 diffusion membranes prepared in step 1 onto the enzyme layer. Step 6: After the polymerization process stopped (5 min), the biosensor rim is sealed with silicone glue to prevent lateral glucose influx. The silicone glue is allowed to dry for 15 min. Step 7: Two or four magnets are placed at the rims or in the four corners of the plastic foil piece and fixed by silicone glue. Step 8: The glucose sensor is then irradiated by UV light (15 W, 253.7 nm) for 1 h for sterilization.

In preparation for cell cultivation, the sensors are placed in the cell bag through the screw cap and fixed from the outside using additional magnets (see Fig. 3 and Note 7). The sensors are placed directly over a PreSens CFG coaster that is connected via optical fiber to the respective meter (Fibox 3 or a 4-channel oxygen meter OXY-4 mini; see Note 8). 3.2 Brief Description of PreSens GmbH Equipment Necessary for Glucose Biosensors 3.2.1 Fibox 3

The Fibox 3 is a single channel fiber optic oxygen transmitter with temperature compensation. It is designed for small fiber optic (POF) oxygen sensors with sensor coating type PSt3 (limit of detection 0.03% oxygen, 15 ppb dissolved oxygen). The small dimensions and low power consumption make it suitable for portable use. For operation, a PC/notebook is required that is connected via a serial COM port. The Fibox 3 is controlled using a comfortable software, which also saves and visualizes the measured values. The data output is programmable to deliver oxygen, temperature, or the raw values (phase or amplitude). The data

326

Mario Lederle et al.

are retrieved via PC/notebook and USB. The POF can be attached to a transparent foil or glass surface and held in place by applying different accessories, such as the two-channel coaster. A detailed operational manual of Fibox 3 is available on the PreSens website (accessed 20.12.2018) at https://www.presens.de/products/ detail/fibox-3.html. 3.2.2 OXY-4 Mini

The OXY-4 mini possesses similar specifications as the Fibox 3 but is a multi-channel fiber optic oxygen meter for simultaneous readout of up to 4 sensors. It is used with sensors based on a 2-mm optical fiber. A PC is connected to run the user-friendly software. OXY-4 mini is compatible with sensor type PSt3 (detection limit 15 ppb, 0–100% oxygen). In contrast to the Fibox 3, the OXY-4 mini does not contain temperature sensors; temperature changes during the measurement are not compensated by the software. The POFs are connected to the 4 SMA fiber connectors. A detailed operational manual of OXY-4 mini is available on the PreSens website (accessed 20.12.2018) at https://www.presens.de/ products/detail/oxy-4-mini.html.

3.3 Sensor Calibration and General Operation

In a basic setting of the sensor, monitoring data is saved in a text file and imported into “MS Excel” for further analysis. The pO2 values from the glucose sensor are subtracted from the values of the O2 sensor to calculate the difference (dO2), which is proportional to the glucose concentration. Typically, before inoculation with cells and glucose, i.e., using a glucose-free medium or, alternatively, water/PBS as a test medium, data are acquired for approximately 45 min to obtain a baseline dO2 representing a glucose concentration of 0 mM. For calculation of the glucose concentration with the obtained dO2 values, a two-point calibration is performed. For this, two dO2 values are plotted over their respective glucose concentrations (for calibrating the oxygen sensor, see Note 9). An example of a calibration is shown in Table 1 and Fig. 4.

Table 1 Example of a two-point calibration of a glucose biosensor: given are calculated dO2 values and the respective glucose concentrations Glucose concentration [mM]

dO2

0.00

1.83

9.78

36.75

The dO2 values were calculated using the average mean values of 40 values around the respective time point

Optical Glucose Sensor

327

40 35

dO2 [% air sat.]

30 25 y = 3,5715x + 1,8252 R2 = 1

20 15 10 5 0 0

2

4

8

6

10

12

glucose concentration [mM] dO2

Linear (dO2)

Fig. 4 Example of a two-point calibration curve for the glucose biosensor: Black square dots represent the two values of the calibration, and the dotted line represents the linear curve fit

Glucose concentrations can then be calculated using the equation: x¼

ðdO2  1:8252Þ 3:5715

The calculated glucose concentrations can be plotted over the fermentation time. With increasing levels of oxygen partial pressure in the medium, higher dO2 are possible and thus higher glucose levels can be measured. Only at relatively high oxygen levels (e.g., 80% air saturation), glucose concentrations as high as 60 mM can be measured. The dO2 value at 0 mM glucose was usually close to zero. Typical response times for sensors with a dynamic range of 0–20 mM glucose were t90  10 min (t90 ¼ time until 90% of the final sensor signal was reached). When using sensors with a less permeable membrane for higher dynamic measurement ranges, response times typically increased (t90 ¼ 20–30 min). Response times in shake flasks and cell bags were comparable. Since sensors tend to change in sensitivity over cell cultivation time, it is advisable to perform offline control measurements at regular intervals (e.g., daily, as in the case of pH probes). In case of discrepancies, a corrective calibration has to be performed. For this, a glucose concentration value of the offline analysis and the respective dO2 value, at this time point, are used as the second point in a new two-point calibration. From then on, glucose concentrations are calculated using the new equation.

328

Mario Lederle et al.

3.4 Sensor Working Range Test

It is advisable to test the working range of biosensors in a cell-free environment before embarking on cell culture experiments. The method presented below is comprehensive and lengthy and uses small glucose increments. With growing experience, it can be abbreviated or omitted. In that case, it might be preferable to perform sensor calibration with more than one glucose concentration. In the experiment presented here, the biosensor was aseptically mounted in a 2-L cell bag; the bag was filled with 500 mL water (or PBS; including Pen/Strep) and temperature set to 37  C. The oxygen transducer inside the glucose biosensor and the reference sensor were calibrated according to the calibration sheet of the manufacturer (see Subheading 3.3). The dissolved oxygen level (DO) in the bag was 90%. Then the baseline dO2 was determined for 18 h. Following this, glucose was added in 5 mM steps leading to a change in the pO2 signal of the glucose biosensor. After the drop in pO2 signal eased out, the next 5 mM of glucose was fed. This was done until the glucose concentration reached 75 mM. The raw data of the O2 reference sensor, the glucose biosensor, and the resulting dO2 are displayed in Fig. 5. Throughout the experiment, the signal of the O2 reference sensor remained constant. At a glucose concentration of 0 mM, the signal of the glucose biosensor showed no difference to the O2 reference sensor signal. After each glucose feeding step, the signal of the glucose sensor decreased stepwise, leading to a stepwise

100 90 80

O2 [% air sault.]

70 60 50 40 30 20 10 0 18

19

20

21 O2 sensor

22 time [h] Glucose sensor

23

24

25

dO2

Fig. 5 Raw data of the O2 reference sensor (black dots), the glucose biosensor (red dots), and the resulting dO2 (gray dots) over time for a working range test

Optical Glucose Sensor

329

60

50

dO2 [% air sat.]

40 y = 0,8519x + 2,9691 R2 = 0,999 30

20

10

0 0

10

20

30

40

50

60

70

glucose concentration [mM] dO2

Linear (dO2)

Fig. 6 Plot of the determined dO2 over glucose concentration (0–60 mM). The dotted line denotes the linear regression

increase in the dO2 values. At a glucose concentration of 75 mM, the signal of the glucose sensor reached 20% air saturation, while the dO2 reached 73.32% air saturation. In order to evaluate linearity of the sensor, dO2 is plotted over glucose concentration (i.e., the increments of the experiment). It is advisable to subsequently perform several regressions in order to find the best possible linearity range of the sensor. Here, it is 0–60 mM, resulting in a regression coefficient of 0.999 (see Fig. 6). 3.5

Examples

After the steps described above, sensors can be applied in cell culture experiments in different formats. The cell culture experiments described here were performed with a recombinant CHO cell in 125-mL shake flasks and 2-L cell bags. A cell culture experiment usually lasted 6–9 days, and sensor drifts within this period were typically low. In the best case, relative average signal errors were below 5% throughout a run. The inoculation density in shake flasks was 2.5  105 cells/mL in 50 mL working volume. Flasks were incubated at 37  C, 130 RPM, and 5% CO2. For disposable bag experiments, 2-L cell bags were filled with 0.5 L of medium and supplemented with antibiotics. Cell bags were operated at 37  C, 3.5% CO2, and 80% air saturation and inoculated with 5  105 cells/mL. Concentrated

Mario Lederle et al.

glucose solution (0.5 M) was added manually in the case of the cell bag. Samples were taken periodically and centrifuged at 400  g for 5 min to determine the glucose concentration in the supernatant offline by means of a hexokinase assay or the Cedex Bio [7]. In Fig. 7, the sensor signal of a typical batch experiment is shown. Flasks were inoculated with 40 mM of glucose. As cells consumed glucose, the sensor signal dropped with progressing culture time. The drop in the biosensor signal was confirmed with two different offline measurements. Figure 8 shows a typical fed-batch experiment in a cell bag. Glucose was added on days 4, 7, 8, and 10 which is readily reflected both by the glucose sensor and offline analytics. In the two experiments shown here, no drift of

Fig. 7 Glucose level monitoring in a shake flask with CHO cells (50 mL working volume). The experiment was carried out at 37  C, 5% CO2, and 130 RPM. The initial glucose concentration was 40 mM; no further glucose was added throughout the experiment. Offline analysis was performed via Cedex Bio Analyzer (Roche) and a hexokinase assay 12

Glucose [mM]

330

10 8 6 4 2 0 0

1

2

3

4

5

6

7

8

9

10

11

Time [d] Biosensor signal

Offline

Fig. 8 Glucose level monitoring in a cell bag bioreactor with CHO cells (500 mL working volume). The experiment was carried out at 37  C, 3.5% CO2, and 80% air saturation and the reactor operated at 10 tilting angle and 20 rocking cycles per minute. Glucose was added manually on days 4, 7, 8, and 10. Samples for offline control were drawn periodically and analyzed in a hexokinase assay in triplicate. Error bars correspond to the standard deviation [7]

Optical Glucose Sensor

331

the sensors was observed. It is advisable to periodically check for such drifts by external glucose analysis and if needed recalibrate the sensor (see Subheading 3.3).

4

Notes 1. In this study, acrylate adhesive foil from PreSens GmbH, Regensburg, Germany, was used. Other acrylate adhesive foils can be tried as well. 2. This applies only if a Cedex Bio Analyzer (Roche) is used for external glucose measurements. 3. In this study, PowerCHO 2 (serum-free CHO medium without glucose and L-glutamine, with 0.1% Pluronic-F68®; Biozym, Hessisch Oldendorf, Germany) was used to be able to define specific glucose concentrations. Other media would work as well. 4. Optical oxygen sensors can be pre-treated with 20% 3-aminopropyltriethoxysilane (APTES) for 2 h at 40  C in order to improve the enzyme layer adhesion to the hydrophobic oxygen sensor surface. 5. The process to dissolve the components takes some time. It is advisable to start with preparing the solution. To do so, weigh the components into a 1.5-mL Eppendorf cup. The ingredients can now dissolve at room temperature. 6. To apply the silicone glue, it is advisable to use a pipet tip. Make sure the glue is all around the tube and covers all edges but leaves out the area with the sensor spot. 7. Sensors can exposed to glucose for 2–4 days before starting an experiment, to avoid significant sensor drifts during the experiment [7]. The nature of the diffusion membrane determines the dynamic range of the biosensor. Biosensors with three different types of membranes were produced with dynamic ranges of 0–2 mM, 0–20 mM, and 0–60 mM of glucose. The production of reproducible biosensors takes patience and experience. It is possible that even within a batch of sensors produced at the same time, differences in the functional stability and the dynamic range are observed. A loss of functional stability is mostly observed when the sensor is operated near or outside its dynamic range. Drifts of the sensor signals were observed in both directions, loss as well as increase of sensitivity. 8. It is important to exactly fix the position of the sensor pad inside the bag with respect to that of the coaster outside the bag since changing positions cause signal changes. If necessary, it is advisable to fixate the coaster and bag with adhesive tape.

332

Mario Lederle et al.

9. Although oxygen sensors normally come pre-calibrated from the supplier, they may also be calibrated by the user (e.g., when using material from a different source): two-point calibration according to the instruction manual “Oxygen Sensor Spots PSt3/PSt6”: https://www.presens.de, downloaded 20.12.2018. (a) First calibration point: oxygen-free water Dissolve 1 g of sodium sulfite and 50 μL cobalt nitrate of a standard solution (ρ(Co) ¼ 1000 mg/L; in nitric acid 0.5 mol/L) in 100 mL water. Use a suitable vessel with a tightly fitting screw top. Make sure there is only little headspace in your vessel. Due to a chemical reaction of oxygen with the sulfite, the water becomes oxygen-free. Close the vessel with the screw top and shake it for approximately 1 min to dissolve sodium sulfite and to ensure that the water is oxygen-free. Fill the calibration solution in the vessel you have mounted the sensor spot in. Make sure the sensor spot surface is covered completely with the liquid. After recording the first calibration point, remove the calibration solution and fill the vessel with distilled water and stir it for 1 min. Repeat this procedure at least 5 times to clean the sensor spot from sodium sulfite. (b) Second calibration point: air-saturated water Add 100 mL water to a suitable vessel. To obtain air-saturated water, blow air into the water using an air pump with a glass frit (air stone), creating a multitude of small air bubbles, while stirring the solution. After 20 min, switch off the air pump and stir the solution for another 10 min to ensure that the water is not supersaturated. Fill the calibration solution in the vessel you have mounted the sensor spot in. Make sure the sensor spot surface is covered completely with the liquid. To minimize the response time, slightly stir the solution. Then follow the instructions in the respective transmitter manual for calibration.

Acknowledgment We dedicate this manuscript to Mario Lederle, our student and colleague, who worked on this project throughout his PhD and who unfortunately passed away last year.

Optical Glucose Sensor

333

References 1. Schwamb S, Puskeiler R, Wiedemann P (2015) Monitoring of cell culture. In: Al-Rubeai M (ed) Cell engineering: animal cell culture. Springer International, Basel, pp 185–221 2. Beutel S, Henkel S (2011) In situ sensor techniques in modern bioprocess monitoring. Appl Microbiol Biotechnol 91:1493–1505 3. Rathore AS (2009) Roadmap for implementation of quality by design (QbD) for biotechnology products. Trends Biotechnol 27:546–553 4. Eibl R, Kaiser S, Lombiser R, Eibl D (2010) Disposable bioreactors: the current state-ofthe-art and recommended applications in biotechnology. Appl Microbiol Biotechnol 86:41–49 5. Tayi V, Butler M (2015) Physiology and metabolism of animal cells for production. In: Hauser H, Wagner R (eds) Animal cell

biotechnology. De Gruyter, Berlin/Mu¨nchen/ Boston, pp 301–325 6. Borisov SM, Wolfbeis OS (2008) Optical biosensors. Chem Rev 108(2):423–461 7. Tric M, Lederle M, Neuner L, Dolgowjasow I, Wiedemann P, Wo¨lfl S, Werner T (2017) Optical biosensor optimized for continuous in-line glucose monitoring in animal cell culture. Anal Bioanal Chem 409:5711–5721 8. Demas JN, DeGraff BA, Coleman PB (1999) Oxygen sensors based on luminescence quenching. Anal Chem 71(23):793A–800A 9. Wolfbeis OS, Klimant I, Werner T, Huber C, Kosch U, Krause C, Neurauter G, Du¨rkop A (1998) Set of luminescence decay time based chemical sensors for clinical applications. Sensors Actuators B Chem 51:17–24

Chapter 20 Turbidimetry and Dielectric Spectroscopy as Process Analytical Technologies for Mammalian and Insect Cell Cultures Lukas K€aßer, Jan Zitzmann, Tanja Grein, Tobias Weidner, Denise Salzig, and Peter Czermak Abstract The production of biopharmaceuticals in cell culture involves stringent controls to ensure product safety and quality. To meet these requirements, quality by design principles must be applied during the development of cell culture processes so that quality is built into the product by understanding the manufacturing process. One key aspect is process analytical technology, in which comprehensive online monitoring is used to identify and control critical process parameters that affect critical quality attributes such as the product titer and purity. The application of industry-ready technologies such as turbidimetry and dielectric spectroscopy provides a deeper understanding of biological processes within the bioreactor and allows the physiological status of the cells to be monitored on a continuous basis. This in turn enables selective and targeted process controls to respond in an appropriate manner to process disturbances. This chapter outlines the principles of online dielectric spectroscopy and turbidimetry for the measurement of optical density as applied to mammalian and insect cells cultivated in stirred-tank bioreactors either in suspension or as adherent cells on microcarriers. Key words Process analytical technology, Online process monitoring, Optical density, Dielectric spectroscopy, Drosophila melanogaster S2 cells, Vero cells, Measles virus, Cell cultivation

1

Introduction For pharmaceutical products, quality is defined by the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) in ICH Q6A [1] as “The suitability of either a drug substance or drug product for its intended use. This term includes such attributes as the identity, strength, and purity.” According to ICH Q8(R2) [2], quality must be built into the product by design as part of the Quality by Design (QbD) initiative, which requires a profound understanding of the manufacturing process. Process analytical technology (PAT) is one of several instruments that can be used to gain such understanding.

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_20, © Springer Science+Business Media, LLC, part of Springer Nature 2020

335

336

ۧer et al. Lukas Ka

Critical process parameters (CPPs) affecting process stability and therefore the critical quality attributes (CQAs) of the product need to be identified during early process development. Once identified, the online monitoring of CPPs using PAT enables the process to be adjusted in real time in order to maintain process stability and ensure CQAs are achieved, resulting in a product with consistent quality and safety. In cell cultures, traditional offline analytics require recurrent manual sampling, thus providing information with a limited resolution. In contrast, online measurements involve the automatic acquisition of culture-related data without manual intervention. This reduces the risk of contamination [3], avoids any delay between sampling and analysis, and improves the accuracy of the manufacturing documentation by providing almost gapless process recording [3]. Current manufacturing processes based on cell cultures already incorporate the monitoring of basic physical and chemical conditions, such as the pH, temperature, concentration of dissolved oxygen (DO), and stirrer speed [4, 5]. Although these parameters provide valuable information about the environmental conditions in a bioreactor, they fail to provide direct information about the cell population [6]. However, because the accumulation of the product (which may be a small molecule, protein, virus, or the cells themselves) is strongly coupled to cell growth and viability, direct realtime monitoring of biological parameters is essential to ensure process efficiency and meet the QbD requirements discussed above. In formal terms, cell growth and viability can be defined as CPPs that must be kept under tight control to ensure the final product falls within the prescribed CQAs. Many direct and indirect measurement principles have been evaluated for this purpose (Table 1). Due to their robustness and practical simplicity, dielectric spectroscopy and the online measurement of optical density (OD) by near-infrared (NIR) turbidimetry are widely used methods, even in industrial applications [6, 21]. This chapter describes the use of turbidimetry and dielectric spectroscopy to monitor insect and mammalian cell lines, emphasizing their suitability for both suspension cell cultures and adherent cell lines. 1.1 Measurement of OD Using an NIR Turbidity Sensor

Two types of turbidity sensors have been commercially available for many years (Fig. 1), and their technical principles have been discussed extensively [10, 22, 23]. Transmission probes detect the attenuation of light as it passes a fixed distance through the culture broth (Fig. 1a), whereas backscatter probes detect light scattered by suspended particles at an angle of 180 (Fig. 1b). Depending on the specific design, backscatter probes have a broad detection range and are compatible with very high cell densities. Transmission probes require the selection of a suitable fixed optical path length for the anticipated cell concentration. For a certain range of lower

Process Analytical Technologies for Cell Cultures

337

Table 1 Commercially available products for the measurement of biological variables Method

Measurement principle

Systems

References

Dielectric spectroscopy

Proportionality between measured Aber Futura, Hamilton permittivity and concentration of viable Incyte cells

[7, 8]

Near infra-red (NIR) turbidimetry

Proportionality between light attenuation Exner ExCell 230, Hamilton or backscattering (optical density) and Dencytee, total cell concentration Optek ASD, Mettler InPro8000, Buglab proBE 3000, Cerex Wedgewood BT65

[7, 9, 10]

NIR spectroscopy

Detection of NIR-active groups and Sartorius BioPAT Spectro measurement of biomass and metabolites via multivariate data analysis

[7, 11, 12]

Twodimensional fluorescence spectroscopy

Measurement of cellular fluorophores such Delta Light and Optics as NAD(P)H or aromatic amino acids, BioView multivariate data analysis

[13]

Raman spectroscopy

Measurement and multivariate analysis of Raman scattering spectra

Kaiser Raman Rxn3

[14]

In situ microscopy

Cell counting by conventional or holographic microscopy

OVIZIO iLine F

[4, 15, 16]

Focused beam reflectance (FBR)

Measurement of particle size distribution using the FBR principle

Mettler Particle Track

[17]

Biocalorimetry

Proportionality between metabolic heat output and biomass

Mettler Toledo eRC1

[18]

Off-gas analysis

Proportionality between respiratory activity and biomass

Blue Sense Cell in One

[19]

Soft sensors

Process parameters calculated from other measured process parameters or retrieved from correlations or models

MATLAB

[20]

cell densities, the transmission signal shows a linear correlation with the cell density, and the Beer–Lambert law is formally applicable [23, 24]. At higher cell densities, a nonlinear response is often observed due to the prevalence of multiple scattering effects [10]. Importantly, the turbidity signal represents all suspended particles, not only living cells but also dead cells, cell debris, and gas bubbles [4, 9, 25]. Therefore, calibration should be always carried out under final process conditions. Some OD sensors can directly distinguish between gas bubbles and cells in highly aerated and agitated environments, including the proBE 3000 fiber optic

338

ۧer et al. Lukas Ka

Fig. 1 Principle of turbidimetry based on (a) transmission and (b) backscatter measurements

sensor (Bug Lab, Concord, CA, USA). Others such as the EXcell 230 sensor (Exner, Ettlingen, Germany) are especially compatible with a bubble-free aeration system. The latter sensor, which we used in our experiments presented below, emits light at 880 nm to exclude interactions with colored media components [10] and works with a fixed path length of 5 mm. 1.2 Dielectric Spectroscopy

Dielectric spectroscopy is mainly used to investigate the passive electrical properties of materials. The theory of dielectric spectroscopy dates back more than 100 years and has been widely reviewed [6, 9, 26–30]. Therefore, the following explanation is a simplified description of the underlying principles, and the reader is referred to the cited review articles for more information. As an online monitoring method for cell cultures, a weak alternating current is applied to the cell suspension, forcing ions in the culture medium to move toward their counter-electrode. Whereas ions in the medium can move freely, ions inside the cells are restricted by the plasma membrane. This leads to a charge separation at the membrane and the cells become polarized, essentially resembling quasispherical capacitors. At low frequencies, these capacitors can become fully polarized, and dielectric spectroscopy detects a high permittivity signal. With increasing frequency, polarization is incomplete because the time between current inversions is insufficient for the ions to move through the whole cell. The permittivity therefore declines with increasing frequency, and in the range of radio frequencies used for cell culture experiments, a sigmoidal relationship between permittivity and log(frequency) can be observed. This behavior is known as β-dispersion and is generally described by the empirical Cole–Cole equation (Eq. 1) [26, 27, 30, 31].

Process Analytical Technologies for Cell Cultures

339

Fig. 2 Illustration of β-dispersions, where Δεcc ¼ 10–20 pF/cm, fc ¼ 250–1000 kHz, and α ¼ 0.1–0.3. In this depiction, the values of ε include the electric constant

h i  1α Δε 1 þ f =f c sin ðαπ=2Þ þ εvac εf end ð1Þ εðf Þ ¼  2ð1αÞ  1α 1 þ f =f c þ 2 f =f c sin ðαπ=2Þ   ð2Þ Δεcc ¼ εvac ε1,2,3  εf end f c ¼ f at ε ¼ Δε=2

ð3Þ

Here, fc is the critical frequency at Δε/2, α is the Cole–Cole factor, εvac is the permittivity of free space (electric constant), ε1,2,3 and εfend are the permittivity plateaus at low and high frequencies, and Δεcc is the difference in permittivity of these plateaus (Eqs. 2 and 3). An increase in α causes the β-dispersion slope to decrease (Fig. 2, red), whereas an increase in Δεcc causes the magnitude of the β-dispersion to increase (Fig. 2, blue), and an increase in fc leads to a shift of the point of infliction of the β-dispersion (Fig. 2, violet). The values of Δεcc and fc can be predicted using Eqs. 4 and 5 for a suspension containing a certain volume fraction of spherical, membrane-enclosed cells [29]. Whereas σ i and σ o represent the static conductivity of medium inside and outside the cell, respectively, Cm represents the membrane capacitance per area and Gm represents the membrane conductance per area. Here, r denotes the mean cell radius and p the biomass volume fraction. The volume

340

ۧer et al. Lukas Ka

fraction can also be expressed as a function of r considering the quasi-spherical shape of the cells. Assuming that rGm is small, Eqs. 4 and 5 can be simplified, and if the radius of cells remains constant throughout the cultivation, Δεcc is proportional to the viable cell density N [27]. Δεcc ¼ ¼ fc ¼

9prC m 9prC m with p  2 4 4½1 þ rG m ð1=σ i þ 1=2σ o Þ 4πr 3 N 3

2σ i σ o þ rG m ðσ i þ σ o Þ 1  2πrC m ðσ i þ 2σ o Þ 2πrC m ð1=σ i þ 1=2σ o Þ

ð4Þ ð5Þ

As well as fitting the entire Cole–Cole model to derive the cell concentration by rearranging Eq. 4 [27], it is also possible to use simple dual frequency measurements in combination with linear regression techniques [9, 27, 32]. Furthermore dielectric spectra can also be treated as raw data for multivariate analysis via partial least squares regression [27]. The Incyte Sensor (Hamilton, Bonaduz, Switzerland), which we used in our experiments presented below, operates at 17 frequencies between 300 and 10,000 kHz. This enables the reconstruction of the entire β-dispersion curve and thus provides potentially deep insight into cell physiology. Because only cells with an intact membrane become polarized, dielectric spectroscopy signals exclusively provide information about intact cells. Other components, such as cell debris, gas bubbles, or solid particles, are only relevant insofar as they can replace a volume fraction of the biomass in the measuring region [6, 8]. 1.3 Insect Cell Suspension Cultures for Protein Expression

As a first example for the application of PAT, we discuss the production of recombinant proteins in insect cells. Recombinant proteins can be produced in various insect cell lines originating from Drosophila melanogaster (Schneider 2, S2), Trichoplusia ni (High Five), or Spodoptera frugiperda (Sf9, Sf21). In this context, the baculovirus expression vector system (BEVS) is the most common platform technology, and it has been extensively discussed [33, 34]. For the production of recombinant proteins using BEVS, a bioreactor is inoculated with proliferating uninfected cells. The baculovirus is added when the cells reach the optimal density and the target protein is produced during the course of the infection. To avoid product degradation, the cells should be harvested before the massive cell lysis caused by the late-stage infection. A second approach is the production of recombinant proteins in insect cell lines stably transfected with a plasmid carrying the expression construct. Selection pressure leads to the growth of those transformants that have stably integrated the expression construct. Because the production process itself does not lead to cell lysis, this platform technology allows more process mode diversity,

Process Analytical Technologies for Cell Cultures

341

including batch, fed-batch, continuous, and perfusion processes [33]. NIR turbidimetry and dielectric spectroscopy can be applied to both expression systems to coordinate key transitional events. For stable cell lines with high viability, the technologies can be used interchangeably to determine the optimal timing for induction and harvest [9]. However, because baculovirus-infected cells undergo major physiological changes such as swelling and lysis, dielectric spectroscopy may yield more information for the monitoring of BEVS-based processes [35–38]. Finally, the combination of NIR turbidimetry and dielectric spectroscopy is advantageous because different measurement principles are exploited to gain process information, allowing the online monitoring of cell viability [9]. 1.4 Adherent Mammalian Cell Cultures for the Production of Viruses and CellBased Products

2

Adherent mammalian cells such as Vero cells are often used for the production of so-called advanced therapy medicinal products (ATMPs), which include therapeutic viruses and even the cells themselves, such as therapeutic stem cells. This part of the chapter deals with the application of dielectric spectroscopy to Vero cells that are used for the production of infectious measles virus while growing as adherent cells on microcarriers. Oncolytic measles virus is potentially suitable for the treatment of many cancers, following the initial observation of complete regression in a boy with lymphoblastic leukemia after a measles virus infection [39, 40]. Like the BEVS system discussed above, Vero cells need to be expanded before infection. Doses of up to 1010 measles virus particles are required for oncolytic therapy, so the time of infection (TOI) and time of harvest (TOH) are CPPs which must be optimized to ensure sufficient virus titers [41]. Because the cells are growing on microcarriers, the number of particles in the suspension is relatively stable, and therefore, NIR turbidimetry is inappropriate for monitoring, but dielectric spectroscopy is ideal [42].

Materials

2.1 Cell Suspension Culture 2.1.1 Materials for Thawing Cells

This section lists the materials required for the methods described in Subheading 3.1. 1. Cryo-vial containing 1.5 mL of D. melanogaster S2 cell suspension (e.g., DMSZ no. ACC130, Braunschweig, Germany). For the generation of recombinant transformants, use the DES® system (Thermo Fisher Scientific, Waltham, MA, USA). 2. T-25 flask for the cultivation of cells in suspension (e.g., TC flask, Sarstedt, Nu¨mbrecht, Germany). 3. Serum-free insect cell culture medium 420, Sigma-Aldrich, Hamburg, Germany).

(e.g.,

ExCell

342

ۧer et al. Lukas Ka

4. Sterile single-use pipette tips (e.g., serological pipettes, Sarstedt, Nu¨mbrecht, Germany). 5. Centrifuge (e.g., Sigma 1-16K, Sigma-Aldrich, Hamburg, Germany). 6. Pipette holder/rubber bulb (e.g., pipetus, Hirschmann Laborger€ate, Eberstadt, Germany). 7. Aspiration system Germany).

(e.g.,

Vacusafe,

Integra,

Biebertal,

8. 15-mL and 50-mL centrifuge tubes (e.g., Sarstedt, Nu¨mbrecht, Germany). 9. Incubator (e.g., B-line, Binder, Tuttlingen, Germany) with orbital shaker (e.g., Celltron, Infors HT, Basel, Switzerland). 2.1.2 Materials for Cell Counting Using Trypan Blue Stain

1. 0.4% trypan blue (e.g., Sigma-Aldrich, Hamburg, Germany). 2. Incident light microscope (e.g., DM1i, Leica, Wetzlar, Germany). 3. Neubauer-improved counting chamber (e.g., Marienfeld, Ko¨nigshofen, Germany). 4. Phosphate-buffered saline (PBS, Biochrom, Berlin, Germany). 5. Piston pipettes (e.g., Research Plus, Eppendorf, Hamburg, Germany). 6. Pipette tips (e.g., Sarstedt, Nu¨mbrecht, Germany).

2.1.3 Materials for Cell Counting by Flow Cytometry

1. Flow cytometer (e.g., Guava easyCyte Flow Cytometer, Merck, Darmstadt, Germany). 2. Propidium iodide (PI) 0.005 g/L in PBS (e.g., Sigma-Aldrich, Hamburg, Germany). 3. 96-well plate (e.g., TC Plate 96, Sarstedt, Nu¨mbrecht, Germany). 4. Piston pipettes (e.g., Research Plus, Eppendorf, Hamburg, Germany). 5. Pipette tips (e.g., Sarstedt, Nu¨mbrecht, Germany). 6. Deionized water for instrument cleaning. 7. Guava Instrument Cleaning Fluid (ICF, Merck, Darmstadt, Germany). 8. 0.5-mL and 1.5-mL tubes for waste, water, and ICF (e.g., Sarstedt, Nu¨mbrecht, Germany).

2.1.4 Materials for Cell Passaging

1. Baffled shake flasks. 2. Serum-free insect cell culture medium 420, Sigma-Aldrich, Hamburg, Germany).

(e.g.,

ExCell

Process Analytical Technologies for Cell Cultures

343

3. Pipette holder/rubber bulb (e.g., pipetus, Hirschmann Laborger€ate, Eberstadt, Germany). 4. Sterile single-use pipette tips (e.g., serological pipettes, Sarstedt, Nu¨mbrecht, Germany). 5. Incubator (e.g., B-line, Binder) with orbital shaker (e.g., Celltron, Infors HT, Basel, Switzerland). 2.1.5 Materials for the Preparation of a Stirred-Tank Bioreactor for the Cultivation of Insect Cells

1. Bioreactor with 1-L working volume (e.g., 2L-Labfors, Infors HT, Basel, Switzerland) equipped with a bubble-free aeration system (e.g., an internally wound silicon tube). 2. Bioreactor control unit (e.g., Labfors 5 cell, Infors HT, Basel, Switzerland). 3. Pre-culture of D. melanogaster S2 cells. 4. Serum-free insect cell culture medium 420, Sigma-Aldrich, Hamburg, Germany).

(e.g.,

ExCell

5. Temperature probe (e.g., Labfors 5 cell, Infors HT, Basel, Switzerland). 6. pH probe (e.g., EasyFerm, Hamilton, Bonaduz, Switzerland). 7. DO probe (e.g., VisiFerm DO, Hamilton, Bonaduz, Switzerland). 8. Glass bottles with connector caps for acid, base, and harvest. 9. Glass bottles with connector caps and bottom drain for medium and inoculation. 10. Luer lock adapters (male and female) and lids. 11. Sterile air filters (e.g., Midisart, Sartorius, Go¨ttingen, Germany). 12. Sterile, hydrophobic air filter with Luer lock adapter (e.g., Minisart, Sartorius, Go¨ttingen, Germany). 13. Silicone tubing. 14. Y-tube connector. 15. Tubing clamps. 16. Sterile, single-packed, 10-mL syringe with Luer lock adapter (e.g., B. Braun, Melsungen, Germany). 17. 1 M phosphoric acid. 18. 1 M sodium hydroxide. 19. 1 M CuSO4. 20. 70% (v/v) ethanol. 21. Compressed air and oxygen.

344

ۧer et al. Lukas Ka

2.1.6 Additional Materials for Online Turbidimetry

1. Turbidity probe (e.g., EXcell 230 NIR, Exner, Ettlingen, Germany). 2. Documentation software (e.g., Expert software, Exner, Ettlingen, Germany). 3. USB connection cable.

2.1.7 Additional Materials for Online Dielectric Spectroscopy

1. Dielectric spectroscopy probe (e.g., Incyte, Hamilton, Bonaduz, Switzerland). 2. Signal transformation box (e.g., Cell Density ComBox, Hamilton, Bonaduz, Switzerland). 3. Documentation software (e.g., Cell Density software, Hamilton, Bonaduz, Switzerland). 4. Cleaning solution (15 g/L Na2SO3).

2.2 Adherent Cell Culture 2.2.1 Materials for Cell Thawing and Expansion

This section lists the materials required for the methods described in Subheading 3.2. 1. Cryo-vial with Vero cells (#CCL-81, ATCC, Manassas, VA, USA). 2. T-75 flask for the cultivation of adherent cells (e.g., TC flask, Sarstedt, Nu¨mbrecht, Germany) (see Note 1). 3. Sterile serological pipettes (e.g., Sarstedt, Nu¨mbrecht, Germany). 4. Pipette holder/rubber bulb (e.g., pipetus, Hirschmann Laborger€ate, Eberstadt, Germany). 5. Virus Production Medium Serum Free (VP-SFM) supplemented with 4 mM glutamine (Thermo Fisher Scientific, Waltham, MA, USA). 6. CO2 incubator (e.g., HERAcell 240, Thermo Fisher Scientific, Waltham, MA, USA) with compressed CO2 supply. 7. Water bath at 37  C.

2.2.2 Materials for Cell Passaging

1. T-75 or T-175 flask for the cultivation of adherent cells (e.g., TC flask, Sarstedt, Nu¨mbrecht, Germany) (see Note 1). 2. VP-SFM supplemented with 4 mM glutamine (Thermo Fisher Scientific, Waltham, MA, USA). 3. 0.25% (w/v) trypsin/0.02% EDTA solution (e.g., Biochrom, Berlin, Germany). 4. Trypsin inhibitor (e.g., Sigma-Aldrich, Hamburg, Germany). 5. Sterile serological pipettes (e.g., Sarstedt, Nu¨mbrecht, Germany). 6. Pipette holder/rubber bulb (e.g., pipetus, Hirschmann Laborger€ate, Eberstadt, Germany).

Process Analytical Technologies for Cell Cultures

7. Aspiration system Germany).

(e.g.,

Vacusafe,

Integra,

345

Biebertal,

8. 50-mL centrifuge tube (e.g., Sarstedt, Nu¨mbrecht, Germany). 9. Centrifuge (e.g., Heraeus Megafuge X1R, Thermo Fisher Scientific, Waltham, MA, USA). 10. PBS (Biochrom, Berlin, Germany). 11. CO2 incubator (e.g., HERAcell 240, Thermo Fisher Scientific, Waltham, MA, USA) with compressed CO2 supply. 2.2.3 Materials Used to Determine the 50% Tissue Culture Infective Dose (TCID50)

1. 96-well plates (e.g., TC Plate 96, Sarstedt, Nu¨mbrecht, Germany). 2. Vero cells (e.g., #CCL-81, ATCC, Manassas, VA, USA). 3. Measles virus (e.g., MVvac2 GFP (P), Paul-Ehrlich-Institut, Langen, Germany [43]). 4. VP-SFM supplemented with 4 mM glutamine (Thermo Fisher Scientific, Waltham, MA, USA). 5. 12-channel piston pipette (e.g., Research Plus, Eppendorf, Hamburg, Germany). 6. Single-use pipette reservoirs (e.g., 25-mL pipette reservoir, Argos, Vernon Hills, IL, USA). 7. Single-use pipette tips (e.g., Sarstedt, Nu¨mbrecht, Germany). 8. CO2 incubator (e.g., HERAcell 240, Thermo Fisher Scientific, Waltham, MA, USA) with compressed CO2 supply.

2.2.4 Materials for Measles Virus Production in a StirredTank Bioreactor

1. Bioreactor with 1-L working volume (e.g., Z611000110, Applikon, Delft, The Netherlands) equipped with a porous sparger for DO control. 2. Bioreactor control unit. 3. pH probe (e.g., Z001023551, Applikon, Delft, Netherlands). 4. Temperature probe Netherlands).

(e.g.,

PT-100,

Applikon,

Delft,

5. DO probe (e.g., Optical Oxygen Sensor, PreSens, Regensburg, Germany). 6. Vero cells (e.g., # CCL-81, ATCC, Manassas, VA, USA). 7. Measles virus (e.g., MVvac2 GFP (P)). 8. Glass bottles with connector caps for base and harvest (see Note 2). 9. Glass bottles with bottom drain and connector caps for medium and inoculation (see Note 2). 10. Sterile air filters (e.g., Midisart, Sartorius, Go¨ttingen, Germany).

346

ۧer et al. Lukas Ka

11. Sterile, hydrophobic air filter with Luer lock adapter (e.g., Minisart, Sartorius, Go¨ttingen, Germany). 12. Tubing clamps. 13. Sterile, single-packed, 10-mL syringe with Luer lock adapter (e.g., B. Braun, Melsungen, Germany). 14. Luer lock adapters (male and female). 15. Silicone tubing. 16. Y-tube connector. 17. 1 M sodium hydroxide. 18. Microcarriers (e.g., Cytodex, 4400 cm2/g, GE Healthcare, Chicago, IL, USA). 19. Compressed air. 20. Compressed CO2. 21. 70% (v/v) ethanol or suitable disinfection solution (see Note 3). 2.2.5 Materials for Cell Counting Using Crystal Violet Stain

1. Crystal violet staining solution (see Note 4). 2. 1-mL micro tubes (e.g., Eppendorf, Hamburg, Germany). 3. Thermomixer (e.g., ThermoMixer C, Eppendorf, Hamburg, Germany). 4. Neubauer-improved counting chamber (e.g., Marienfeld, Ko¨nigshofen, Germany). 5. Piston pipettes (e.g., Eppendorf, Hamburg, Germany). 6. Pipette tips (e.g., Sarstedt, Nu¨mbrecht, Germany). 7. Vortexer.

2.2.6 Additional Materials Used to Determine the TOH by Dielectric Spectroscopy

1. Dielectric spectroscopy probe (e.g., Incyte, Hamilton, Bonaduz, Switzerland). 2. Signal transformation box (e.g., Cell Density ComBox, Hamilton, Bonaduz, Switzerland). 3. Documentation software (e.g., Cell Density software, Hamilton, Bonaduz, Switzerland). 4. Cleaning solution (15 g/L Na2SO3).

3

Methods

3.1 Cell Suspension Culture

This chapter briefly describes methods for the maintenance and cultivation of D. melanogaster S2 cells in suspension.

3.1.1 Thawing the Cells

1. Thaw a 1.5-mL cryo-vial containing 1.5  107 S2 cells/mL. 2. Gently add 10 mL of medium to the cells and pipette up and down to ensure proper mixing (see Note 5).

Process Analytical Technologies for Cell Cultures

347

3. Centrifuge the suspension for 5 min at 200  g. 4. Aspirate the supernatant to remove residual DMSO (see Note 6). 5. Resuspend the cells in 5 mL fresh medium. 6. Transfer the 5 mL of cell suspension into a T-25 suspension flask. 7. Incubate the cells at 27  C. 8. Cultivate the cells at 27  C on the orbital shaker at a suitable rate (e.g., 80 rpm when using the Celltron shaker). 9. Passage after 3–4 days of cultivation to a concentration of 1.5  106 cells/mL (see Note 7). 3.1.2 Cell Counting by Trypan Blue Staining

1. Assemble the counting chamber as instructed by the manufacturer. 2. Take a sample of the cells (typically 100 μL). 3. Mix PBS with the sample to adjust the concentration so it is suitable for counting (typical ratios range from 1:1 to 1:20). 4. Take 75 μL of the diluted cell suspension and add 75 μL trypan blue (see Note 8). 5. Incubation is not necessary. Pipette the stained cell suspension immediately into the Neubauer counting chamber and count the cells under the incident light microscope. 6. Calculate the cell density according to the instructions provided by the manufacturer of the counting chamber (see Note 9).

3.1.3 Cell Counting by Flow Cytometry Basic Flow Cytometry Method

Measurement Procedure

1. Gate S2 cells in the FSC/SSC scatter plot. 2. Define a histogram (PI fluorescence channel) gated on the cell population. 3. Define histogram markers for viable (PI negative) and nonviable (PI positive) cells using corresponding cell samples (typically an exponentially growing cell population exhibits >95% viability, whereas nonviable cells can be generated, e.g., by treatment with 30% ethanol for 10 s). 1. Take a sample of the cells to be counted (typically 100 μL). 2. Define a worklist and select the wells of the 96-well plate that will be used for sample preparation. 3. Perform a Quick-Clean. 4. Select the analysis method created before. 5. Start the worklist.

348

ۧer et al. Lukas Ka

6. Place waste, water, and ICF tubes in the correct places (see Note 10). 7. Gently mix the sample by pipetting. 8. Add 20 μL of the sample into a well. 9. Add 180 μL of 0.005 g/L PI (see Note 11), and if the counts per second exceed 500, use a dilution of 1:20 (10 μL per 190 μL) instead of 1:10. 10. Place and load the plate into the flow cytometer. 11. Select the first well for the settings-check and readjust the settings or start analyzing the whole worklist. 12. Save the data file. 3.1.4 Passaging the Cells

1. Count the cells using the Neubauer chamber or by flow cytometry (see Note 12). 2. Calculate the volume of cell suspension and medium needed to adjust the new passage to 1.5  106 cells/mL (see Note 7). 3. First pipette the calculated volume of fresh medium into a new shake flask; then add the calculated volume of cell suspension from the previous passage (see Note 5). 4. Keep the previous passage as a backup. 5. Cultivate the cells at 27  C on the orbital shaker at a suitable rate (e.g., 80 rpm when using the Celltron shaker).

3.1.5 Preparation of a Stirred-Tank Bioreactor for the Cultivation of Insect Cells Preparation of the Bioreactor

1. Attach air filters and tubing with Luer lock adapters to the connector caps of all bottles (see Note 13). 2. Assemble the bottles for medium and inoculation and close the bottom drain tubing with Luer lock lids. 3. Fill the acid and base bottles with phosphoric acid and sodium hydroxide solution, respectively. 4. Screw the connector caps on all bottles (medium, inoculation, acid, base, and harvest). 5. Make sure to attach tubing clamps to all tubes. 6. Connect tubes for sampling (Subheading 3.1.6) and pH regulation to the bioreactor head plate and attach sterile air filters to the gas inlets and outlets. 7. Connect the pH probe and calibrate according to the instructions provided by the manufacturer. 8. Connect the DO probe and calibrate according to the instructions provided by the manufacturer. 9. Equip the reactor with the desired probes (DO, pH, OD, dielectric spectroscopy) and check the installation depth and orientation of the probes (especially the dielectric spectroscopy amplifier).

Process Analytical Technologies for Cell Cultures

349

10. Cover the probe connectors and sterile air filters with aluminum foil. 11. Autoclave bottles and the bioreactor with opened tubing clamps to allow the steam to reach all surfaces, but keep acid and base tubing clamps closed to avoid spillage. 12. After autoclaving, close tubing clamps and let the equipment cool to room temperature. Initiation of the Cultivation

1. Determine the cell concentration in the shake flask culture (Subheading 3.1.2). 2. Calculate the volume of cell suspension needed to inoculate the bioreactor at 1.5  106 cells/mL (inoculation volume). 3. Transfer insect cell medium (working volume minus inoculation volume) into the sterile medium bottle (work under a clean hood). 4. Remove the lids of the Luer lock adapters from the medium bottle drain tube and bioreactor medium inlet tube. 5. Disinfect the Luer lock adapters with 70% (v/v) ethanol and quickly connect the male to the female adapter. 6. Open both tubing clamps. 7. Use gravity or a peristaltic pump to transfer the sterile medium from the medium bottle to the bioreactor. 8. Use the process control system to set the start temperature (27  C) and stirrer speed (70 rpm, depending on stirrer and vessel geometry). 9. After the desired temperature is reached and the system is mixed, calibrate the DO probe (100% O2 saturation) according to the instructions provided by the manufacturer and start DO control. 10. Pump acid and base until the connection tube is filled up to the reactor inlet and start pH control (pH 6.4). 11. Transfer the inoculum into the sterile inoculum flask (work under a clean hood). 12. Remove the lids of the Luer lock adapters from the inoculum bottle drain tube and the bioreactor inoculum inlet tube. 13. Disinfect the Luer lock adapters with 70% (v/v) ethanol and quickly connect the male to the female adapter. 14. For inoculation, open the tubing clamps and transfer the inoculum to the bioreactor under gravity. 15. Inoculate and simultaneously start data acquisition with the biomass sensors according to “Turbidity Measurement” and “Dielectric Spectroscopy” below.

350

ۧer et al. Lukas Ka

Turbidity Measurement

1. Connect the turbidity probe to the computer and start data acquisition (e.g., the EXcell 230 sensor and the corresponding Expert software). 2. When a stable baseline is established, proceed with inoculation. 3. Use a sterile syringe to periodically take samples for offline cell density determination (Subheading 3.1.6). 4. Record the sampling time for data evaluation. 5. Cell expansion is usually performed for 5–10 days until the stationary phase is reached. 6. Correlation between the OD signal and the offline cell density can be analyzed, e.g., by simple linear regression or advanced methods such as linear mixed effects models (Fig. 3a) [9]. 7. For subsequent processes, the time of induction and harvest can be guided by the online signal (Fig. 3b) (see Notes 14 and 15).

Dielectric Spectroscopy

1. Before setting up the bioreactor, ensure the dielectric spectroscopy probe is properly cleaned and calibrated in accordance with the manufacturer’s instructions. 2. Once a stable temperature of 27  C has been reached, follow up with the settings for the dielectric probe (see Note 16). 3. Connect the amplifier to the probe, and choose the following settings: measure frequency ¼ 1000 kHz, frequency maximum ¼ 10,000 kHz, and signal integration ¼ high (animal cell culture profile). 4. Enable the frequency scan (f scan). 5. The instrument output “Cell density” (ε) equals the signal difference of the frequencies mentioned above, while “Delta Epsilon” (Δε) equals the maximal signal difference of the high and low frequency plateau. 6. Apply “mark zero” for the dual frequency measurement and for the f scan. 7. Start recording and type in your file name. Recording every 30 min is sufficient to follow cell growth. 8. Press “Inoculate” in the control software and proceed with inoculation (coordinate a synchronous start with the OD sensor if both sensors are used in parallel). 9. Use a sterile syringe to periodically take samples for offline cell density analysis (Subheading 3.1.6). 10. For data evaluation, apply a comment to every sample in the control unit. 11. Analyze the correlation between the offline cell density and permittivity (Fig. 3c).

Process Analytical Technologies for Cell Cultures

351

Fig. 3 Scatter plot of the offline cell count and (a) online OD or (c) permittivity based on 11 or 13 different cultivations of recombinant D. melanogaster S2 cell lines. (b) Example of using online signals for process guidance in a typical batch culture of S2 cells for the production of a secreted recombinant protein. (Adapted from ref. 9, published under open access CC-BY license 4.0)

12. Cell expansion and protein expression are usually performed for 5–10 days until the stationary phase is reached. 13. Use the online signals to determine the time of induction and harvest (Fig. 3b) (see Notes 14 and 15). Combination of Turbidimetry and Dielectric Spectroscopy

3.1.6 Offline Sampling

Because turbidity measurements capture information about the total cell concentration and dielectric spectroscopy captures information only about viable cells, both measurement techniques can be combined to obtain information about the culture’s viability. As previously described [9], the ratio between the permittivity and OD correlates with cell viability. Diverging signals therefore show changes in the viability of cultivated cells (Fig. 4). 1. During cultivation, close tubing clamps (1), (2), and (5) as shown in Fig. 5. 2. Remove the lid of the sampling Luer lock adapter (3). 3. Disinfect the sampling Luer lock adapter by applying disinfection spray and connect a sterile syringe. 4. Connect a separate syringe (air filled) to the air filter (4). 5. Open tubing clamps (2) and (5).

352

ۧer et al. Lukas Ka

Fig. 4 Cell viability determined by the parallel measurement of turbidity and permittivity in a sample batch cultivation. After entering the stationary phase, the cell population lost viability. This transition is visible in the corresponding ΔεOD880 phase trajectory. (Adapted from ref. 9, published under open access CC-BY license 4.0)

6. Flush sampling tube with air using the syringe connected to the air filter. 7. Close tubing clamp (2). 8. Open tubing clamp (1) and draw syringe to take a sample. 9. Close tubing clamp (1) and open tubing clamp (2). 10. Flush air through the tubing until bubbles appear at the tip of the sampling pipe in the bioreactor. 11. Close tubing clamp (5) and open tubing clamp (1). 12. Flush air through the tubing to push the remaining liquid into the sampling syringe. 13. Close tubing clamps (1) and (2). 14. Remove the sampling syringe and disinfect the sampling connector with disinfection spray. 15. Connect a new, sterile syringe to the sampling Luer lock adapter (3). 16. Proceed with offline analysis.

Process Analytical Technologies for Cell Cultures

353

Fig. 5 Schematic illustration of the offline sampler for the stirred-tank reactor (STR). Silicone tubes are connected to the Y-tubing connector. A Luer lock adapter is attached to the end of the sampling tube, and an air filter is attached to the end of the backflush tube 3.2 Adherent Cell Culture

3.2.1 Thawing the Cells

This chapter briefly describes methods for the maintenance and cultivation of adherent growing Vero cells for measles virus production. 1. Thaw a 1.5-mL cryo-vial containing 1106 Vero cells/mL in a water bath (37  C). 2. Transfer cells into a T-75 flask containing 25 mL fresh medium (4–8  C). 3. Incubate the cells at 37  C in a 5% CO2 atmosphere (6–12 h). 4. Remove the spent medium and replace it with fresh medium. 5. Incubate cells at 37  C in 5% CO2 atmosphere until they reach 80% confluence (see Note 17).

3.2.2 Passaging the Cells (Seed Train)

1. Once the cells are ready for passaging, aspirate the spent medium. 2. Wash the cells once with PBS without Ca2+ and Mg2+ (4–8  C). 3. Incubate with 0.012 mL/cm2 trypsin for 8 min at 37  C. 4. Add 0.012 mL/cm2 trypsin inhibitor. 5. Use 0.12 mL/cm2 medium or PBS to wash the cells off the bottom of the flask.

354

ۧer et al. Lukas Ka

6. Take a 50-μL sample for cell counting as described in Subheading 3.1.2. 7. Transfer the cell suspension into a centrifuge tube and centrifuge for 5 min at 300  g. 8. Discard the supernatant. 9. Resuspend the cell pellet in fresh medium. 10. Seed the cells in several new T-flasks for cell expansion (see Note 18). 11. Proceed with inoculation. 3.2.3 Determination of the TCID50

TCID50

determination

or

bioreactor

1. Harvest cells from T-flasks as described in Subheading 3.2.2. 2. For each virus sample, prepare 20 mL of cell suspension with a density of 5  104 cells/mL. 3. Transfer 20 mL of the cell suspension into the reagent reservoir. 4. Fill all wells of a 96-well plate with 200 μL cell suspension using a multichannel pipette (one seeding plate per virus sample is required). 5. Incubate at 37  C in 5% CO2 for 4 h (see Note 17). 6. In the meantime: (a) Take another 96-well plate (virus dilution plate) for virus dilution (12 wells per virus sample): Add 270 μL of fresh medium (4–8  C) to each well (see Note 19). (b) Add 30 μL of the virus sample to the first well and mix with the pipette. (c) Replace the pipette tips with fresh ones. (d) Take 30 μL from the first well and transfer it to the second well (see Note 20). (e) Mix the suspension. (f) Replace the pipette tips with fresh ones. (g) Repeat the last five steps until the twelfth well is reached. 7. Use a 12-channel pipette and add 30 μL of the diluted virus samples (from the virus dilution plate) to the cell seeding plate to generate eight replicates for each dilution (the pipetting scheme is shown in Fig. 6). 8. Incubate the cells at 37  C in 5% CO2 for 120 h. 9. Determine whether the cells in each well are infected by detecting the reporter protein GFP or the induction of syncytia.

Process Analytical Technologies for Cell Cultures

355

Fig. 6 Pipetting scheme for virus dilution. Infected wells are denoted (+), uninfected wells (–). The TCID50 is calculated as follows: log(TCID50)¼ 5 + 1(2.5–0.5) ¼ 7. Therefore, TCID50 ¼ 107. Divided by the sample volume of 30 μL ¼ 0.03 mL, the TCID50/mL value is 33.3  107/mL

10. Mark infected wells with (+) and count the infected wells per column (Fig. 6). 11. Calculate the TCID50 using Eq. 6 [44], where x ¼ log of the last column with 100% infected wells (Fig. 6, red circle), D ¼ log of the dilution factor (30 μL virus suspension in a total volume of 300 μL leads to a dilution factor of 10; hence, log(10) ¼ 1), and Sp ¼ sum of proportion of the first column with 100% infected wells to the column with 0% infected wells (Fig. 6, red box). log ðTCID50 Þ ¼ jx j þ D ðSp  0:5Þ

ð6Þ

356

ۧer et al. Lukas Ka

Fig. 7 Bioreactor setup for measles virus production 3.2.4 Preparation of a Bioreactor for the Cultivation of Vero Cells Growing on Microcarriers

1. Prepare a Cytodex 1 stock solution with a concentration of 40 g/L in PBS as recommended by the manufacturer. 2. Add microcarriers to the bioreactor to achieve a final concentration of 3 g/L (the beads will be autoclaved together with the reactor).

Preparation of Microcarriers Preparation of the Bioreactor Setup

1. Prepare the bioreactor as shown in Fig. 7 according to the instructions in Subheading “Preparation of the Bioreactor,” noting that the pH is adjusted with NaOH and CO2 due to the bicarbonate buffer system of the VP-SFM. 2. Add an additional virus suspension inlet tube and close the Luer lock adapter with a corresponding lid.

Initiation of the Cultivation

1. Preheat the culture medium to 32  C and add L-glutamine to a final concentration of 4 mM. 2. Calculate 60% of the final cultivation volume and transfer this volume of VP-SFM to a medium bottle, saving the remaining medium for cell resuspension. 3. Remove the lids of the Luer lock adapters from the medium bottle drain tube and bioreactor medium inlet tube.

Process Analytical Technologies for Cell Cultures

357

4. Disinfect the Luer lock adapters with 70% (v/v) ethanol and quickly connect the male to the female adapter. 5. Open both tubing clamps. 6. Use gravity or a peristaltic pump to transfer the sterile medium from the medium bottle to the bioreactor. 7. Set the start temperature controller to 32  C (see Note 21). 8. When the system reaches 32  C, calibrate the oxygen probe according to the manufacturer’s instructions. 9. Start all remaining controllers (pH ¼ 7.4, DO ¼ 50–70%) and set the minimum agitation rate (see Note 22). 10. Connect the amplifier to the dielectric spectroscopy probe. 11. Start dielectric spectroscopy as described in Subheading “Dielectric Spectroscopy,” but set the temperature to 32  C and the reading interval to 12 min. 12. Harvest the cells from the T-flasks as described in Subheading 3.2.2. 13. Resuspend the cells in the remaining culture medium (see Note 18). 14. Transfer the inoculum into the sterile inoculum flask (work under a clean hood). 15. Remove the lids of the Luer lock adapters from the inoculum bottle drain tube and the bioreactor inoculum inlet tube. 16. Disinfect the Luer lock adapters with 70% (v/v) ethanol and quickly connect the male to the female adapter. 17. For inoculation, open the tubing clamps and transfer the inoculum to the bioreactor under gravity. 18. Select “Inoculate” in the dielectric spectroscopy control software. 19. Let the cells attach to the microcarriers for at least 4 h (e.g., 70 rpm, no aeration). 20. Infect the cells with a multiplicity of infection (MOI) of 30 TCID50 per cell. 21. Therefore, fill a sterile syringe with the required volume of virus suspension under a clean hood (see Note 23) and close the syringe with a Luer lock lid. 22. Remove the lids of the Luer lock adapters from the bioreactor virus suspension inlet tube and the syringe. 23. Disinfect the Luer lock adapters with 70% (v/v) ethanol and quickly connect the male to the female adapter. 24. Inject the virus suspension into the bioreactor and add a comment in the dielectric spectroscopy control software.

358

ۧer et al. Lukas Ka

25. Use a sterile syringe to periodically take samples for the offline analysis of cell density (Subheading 3.2.5) and virus titer (Subheading 3.2.3) (see Note 24). 26. For data evaluation, apply a comment to every sample in the control unit. 3.2.5 Crystal Violet Staining to Count Cells Attached to Microcarriers

1. Take a 1-mL sample of microcarrier cell suspension as described in Subheading 3.1.6. 2. Centrifuge the microcarriers for 5 min at 300  g to settle them. 3. Remove 0.9 mL of the supernatant (see Note 25). 4. Add 0.9 mL crystal violet solution. 5. Incubate the suspension for 20 min at 37  C and 500 rpm in a thermomixer. 6. Pipette the stained suspension into the Neubauer counting chamber and count the cell nuclei under the incident light microscope (see Note 26). 7. Calculate the cell density according to the instructions provided by the manufacturer of the counting chamber.

Fig. 8 Permittivity signal (black line) and offline measles virus titer (gray boxes) during measles virus production in VP-SFM. Information about the time of harvest (TOH) can be obtained by dielectric spectroscopy because the permittivity signal drops with falling cell vitality, which is directly related to virus release

Process Analytical Technologies for Cell Cultures 3.2.6 Establishing the TOH

4

359

The measles virus is temperature sensitive, and should be harvested as soon as most of the virus is released into the medium. Chill the harvested measles virus suspension to 4  C as quickly as possible. After infection, the permittivity signal is primary influenced not by the cell concentration, but by the cytopathic effect (see Note 27). This effect can be used to determine the optimal TOH because the dielectric spectroscopy signal changes with cell morphology. About 40 h after the permittivity drops, the highest infective virus dose can be detected (Fig. 8) [42].

Notes 1. Carefully select a suitable T-flask or growth surface for sensitive, adherent growing cells. If slow growth or poor cell attachment is observed, and substrate limitation can be excluded as possible cause, different T-flasks should be tested. If a serumcontaining medium is replaced with serum-free medium, the growth surface should be re-evaluated: sometimes, the cells will need a different growth surface after changing the culture medium. 2. The usage of Cytodex 1 microcarriers requires the coating of all glass vessels and bottles with pharmagrade silicone oil. Wet all surfaces with the oil, then remove excess oil from the vessel, and incubate the vessel at >100  C for >6 h. Carry out this procedure at least six times. To remove the silicone from the vessel, fill the vessel with 1 M NaOH and boil it for 1 h. 3. Do not use virus-inactivating disinfection solutions during virus production because this will also inactivate the therapeutic virus. 4. Prepare a solution of 1 g/L citric acid containing 0.1% (w/v) crystal violet. After passing through a 0.2-μm syringe filter, the solution can be stored for one year at room temperature. 5. Do not preheat insect cell culture medium to 37  C before use. Although this may be appropriate for other cell culture media, insect cell culture media may be damaged because they are optimized for a cultivation temperature of 27  C. 6. After aspiration, add fresh medium promptly to avoid cell damage. 7. S2 cells do not grow properly when seeded at concentrations below 5  105 cells/mL. 8. Only cells with intact cell membranes remain unaffected by the blue dye. These cells appear bright and are counted as living cells. Dead cells with permeable membranes appear blue. The viability of the S2 cells used for inoculation should be >95%.

360

ۧer et al. Lukas Ka

9. Do not forget the prior dilution with PBS and the second dilution with trypan blue in the calculation. The cell concentration can be calculated by multiplying the number of counted cells, the chamber factor (usually 10,000), and the dilution factor, divided by the number of counted greater squares. 10. Make sure the waste bottle is empty and the ICF bottle is filled sufficiently. Do not prepare the samples before defining the worklist and equipping the flow cytometer with ICF and water. PI is toxic to cells and early sample preparation may lead to cell death and therefore to misleading results. 11. PI is a fluorescent dye (excitation wavelength ¼ 585 nm, emission wavelength ¼ 617 nm) used for the determination of cell viability. DNA from dead cells with ruptured or porous membranes is stained, whereas the dye cannot pass through the membrane of viable cells. 12. When inoculated at 1.5–2  106 cells/mL, cell densities exceeding 1  107 cells/mL can be expected after 3–4 days of incubation. 13. To prevent confusion, attach male Luer lock adapters to all tubes connected to the vessel. Attach female Luer lock adapters to all tubes connected to the bottles for medium, inoculum, acid, and base. 14. If the recombinant protein expression is controlled by the inducible D. melanogaster metallothionein promoter, copper sulfate (CuSO4) must be added during cultivation in order to start the production phase. Induce at the mid-exponential growth phase with 600 μM CuSO4. This procedure is not necessary if a constitutive promoter is used, such as the D. melanogaster Actin 5C distal promoter. 15. The harvesting procedure should be initiated before cell death during the stationary phase causes the release of host cell protein and product degradation. 16. After sterilization, do not connect the amplifier to the probe while the system is still hot. Wait until the system has cooled down to room temperature. 17. The CO2 concentration depends on the sodium bicarbonate concentration in the culture medium. Check the manufacturer’s recommendations. 18. An appropriate cell concentration for Vero cells in serum-free medium is 1  104 cells/cm2. A lower concentration can inhibit proliferation or even induce apoptosis. For serumcontaining medium, a seeding density of 5  103 cells/cm2 is appropriate.

Process Analytical Technologies for Cell Cultures

361

19. Measles virus is temperature sensitive, so working at low temperatures is important. If possible, use a cooling pad for the microtiter plate. 20. When using an eight-channel pipette for the virus dilutions 102 to 1012, eight samples can be processed simultaneously. 21. Vero cells grow best at 37  C and this temperature should be used to generate the inoculum. However, because measles virus is temperature sensitive, the cultivation temperature should be reduced to 32  C in the bioreactor. 22. The minimal agitation rate is the lowest rate at which all microcarriers are suspended. It depends on the reactor configuration and must be determined empirically. 23. Use a blunt needle to fill the syringe with virus suspension. Draw up the syringe up to the maximum. To ensure the complete injection of the virus suspension, the air in the syringe must drive the virus suspension from the syringe through the tube into the bioreactor. 24. To determine the TOH based on the permittivity signal, take samples during the first cultivation. For further cultivation, no offline analysis is needed. 25. The supernatant can be used for further measurements, e.g., glucose, lactate, or amino acid levels. 26. Two methods are widely used for the offline determination of cell concentration. One is cell detachment from the microcarriers, similar to cell detachment from static cultures in T-flasks, followed by conventional cell counting. The other is nuclear staining and counting [45, 46]. Crystal violet is a nuclear stain, and the stained nuclei are therefore counted rather than the cells. 27. If the host cells show indications of structural changes following virus infection, the virus is described as cytopathogenic. Common phenomena indicating a cytopathic effect include swelling, rounding, and lysis. Syncytia are formed if several infected cells fuse to create a multinucleate symplasm.

Acknowledgments We thank the Federal Ministry of Education and Research (BMBF) for financial support (Grant No. 13FH001IX5), the Hessen State Ministry of Higher Education, Research and the Arts for financial support within the Hessen initiative for scientific and economic excellence (LOEWE Center for Insect Biotechnology and Bioresources), and Richard M. Twyman for professional editing of the manuscript.

362

ۧer et al. Lukas Ka

References 1. International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (2009) ICH Q8(R2): pharmaceutical development. https://www.ich.org/ fileadmin/Public_Web_Site/ICH_Products/ Guidelines/Quality/Q8_R1/Step4/Q8_R2_ Guideline.pdf 2. International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (1999) ICH Q6A: specifications: test procedures and acceptance criteria for new drug substances and new drug products: chemical substances. https://www.ich. org/fileadmin/Public_Web_Site/ICH_ Products/Guidelines/Quality/Q6A/Step4/ Q6Astep4.pdf 3. Beutel S, Henkel S (2011) In situ sensor techniques in modern bioprocess monitoring. Appl Microbiol Biotechnol 91(6):1493–1505. https://doi.org/10.1007/s00253-011-34705 4. Biechele P, Busse C, Solle D et al (2015) Sensor systems for bioprocess monitoring. Eng Life Sci 15(5):469–488. https://doi.org/10. 1002/elsc.201500014 5. Eibl R, Eibl D, Po¨rtner R et al (eds) (2009) Cell and tissue reaction engineering. Principles and practice. Springer, Berlin, Heidelberg 6. Carvell JP, Dowd JE (2006) On-line measurements and control of viable cell density in cell culture manufacturing processes using radiofrequency impedance. Cytotechnology 50 (1-3):35–48. https://doi.org/10.1007/ s10616-005-3974-x 7. Kiviharju K, Salonen K, Moilanen U et al (2008) Biomass measurement online: the performance of in situ measurements and software sensors. J Ind Microbiol Biotechnol 35 (7):657–665. https://doi.org/10.1007/ s10295-008-0346-5 8. Justice C, Brix A, Freimark D et al (2011) Process control in cell culture technology using dielectric spectroscopy. Biotechnol Adv 29(4):391–401. https://doi.org/10.1016/j. biotechadv.2011.03.002 9. Zitzmann J, Weidner T, Eichner G et al (2018) Dielectric spectroscopy and optical density measurement for the online monitoring and control of recombinant protein production in stably transformed Drosophila melanogaster S2 cells. Sensors (Basel) 18:3. https://doi. org/10.3390/s18030900 10. Wu P, Ozturk SS, Blackie JD et al (1995) Evaluation and applications of optical cell density probes in mammalian cell bioreactors.

Biotechnol Bioeng 45(6):495–502. https:// doi.org/10.1002/bit.260450606 11. Cervera AE, Petersen N, Lantz AE et al (2009) Application of near-infrared spectroscopy for monitoring and control of cell culture and fermentation. Biotechnol Prog 25 (6):1561–1581. https://doi.org/10.1002/ btpr.280 12. Qiu J, Arnold MA, Murhammer DW (2014) On-line near infrared bioreactor monitoring of cell density and concentrations of glucose and lactate during insect cell cultivation. J Biotechnol 173:106–111. https://doi.org/10.1016/ j.jbiotec.2014.01.009 13. Marose S, Lindemann C, Scheper T (1998) Two-dimensional fluorescence spectroscopy: a new tool for on-line bioprocess monitoring. Biotechnol Prog 14(1):63–74. https://doi. org/10.1021/bp970124o 14. Abu-Absi NR, Kenty BM, Cuellar ME et al (2011) Real time monitoring of multiple parameters in mammalian cell culture bioreactors using an in-line Raman spectroscopy probe. Biotechnol Bioeng 108(5):1215–1221. https://doi.org/10.1002/bit.23023 15. Joeris K, Frerichs J-G, Konstantinov K et al (2002) In-situ microscopy: online process monitoring of mammalian cell cultures. Cytotechnology 38(1-3):129–134. https://doi. org/10.1023/A:1021170502775 16. Bluma A, Ho¨pfner T, Lindner P et al (2010) In-situ imaging sensors for bioprocess monitoring: state of the art. Anal Bioanal Chem 398(6):2429–2438. https://doi.org/10. 1007/s00216-010-4181-y 17. Whelan J, Murphy E, Pearson A et al (2012) Use of focused beam reflectance measurement (FBRM) for monitoring changes in biomass concentration. Bioprocess Biosyst Eng 35 (6):963–975. https://doi.org/10.1007/ s00449-012-0681-9 18. Cole H, Demont A, Marison I (2015) The application of dielectric spectroscopy and biocalorimetry for the monitoring of biomass in immobilized mammalian cell cultures. Processes 3(2):384–405. https://doi.org/10. 3390/pr3020384 19. Aehle M, Kuprijanov A, Schaepe S et al (2011) Simplified off-gas analyses in animal cell cultures for process monitoring and control purposes. Biotechnol Lett 33(11):2103–2110. https://doi.org/10.1007/s10529-011-06865 20. Chen LZ, Nguang SK, Li XM et al (2004) Soft sensors for on-line biomass measurements.

Process Analytical Technologies for Cell Cultures Bioprocess Biosyst Eng 26(3):191–195. https://doi.org/10.1007/s00449-004-03508 21. Ducommun P, Kadouri A, von Stockar U et al (2002) On-line determination of animal cell concentration in two industrial high-density culture processes by dielectric spectroscopy. Biotechnol Bioeng 77(3):316–323. https:// doi.org/10.1002/bit.1197 22. Konstantinov K, Chuppa S, Sajan E et al (1994) Real-time biomass-concentration monitoring in animal-cell cultures. Trends in Biotechnology 12(8):324–333. https://doi.org/ 10.1016/0167-7799(94)90049-3 23. Junker BH, Reddy J, Gbewonyo K et al (1994) On-line and in-situ monitoring technology for cell density measurement in microbial and animal cell cultures. Bioprocess Eng 10 (5-6):195–207. https://doi.org/10.1007/ BF00369530 24. Myers JA, Curtis BS, Curtis WR (2013) Improving accuracy of cell and chromophore concentration measurements using optical density. BMC Biophys 6(1):4. https://doi. org/10.1186/2046-1682-6-4 25. Zhao L, Fu H-Y, Zhou W et al (2015) Advances in process monitoring tools for cell culture bioprocesses. Eng Life Sci 15 (5):459–468. https://doi.org/10.1002/elsc. 201500006 26. Yardley JE, Kell DB, Barrett J et al (2000) On-line, real-time measurements of cellular biomass using dielectric spectroscopy. Biotechnol Genet Eng Rev 17(1):3–36. https://doi. org/10.1080/02648725.2000.10647986 27. Opel CF, Li J, Amanullah A (2010) Quantitative modeling of viable cell density, cell size, intracellular conductivity, and membrane capacitance in batch and fed-batch CHO processes using dielectric spectroscopy. Biotechnol Prog 26(4):1187–1199. https://doi.org/10. 1002/btpr.425 28. Braasch K, Nikolic-Jaric M, Cabel T et al (2013) The changing dielectric properties of CHO cells can be used to determine early apoptotic events in a bioprocess. Biotechnol Bioeng 110(11):2902–2914. https://doi. org/10.1002/bit.24976 29. Schwan HP (1957) Electrical properties of tissue and cell suspensions, vol 5. Elsevier, Amsterdam, pp 147–209 30. Asami K (2002) Characterization of heterogeneous systems by dielectric spectroscopy. Progr Polym Sci 27(8):1617–1659. https://doi.org/ 10.1016/S0079-6700(02)00015-1 31. Markx GH, Davey CL (1999) The dielectric properties of biological cells at

363

radiofrequencies: applications in biotechnology. Enzyme Microbial Technol 25 (3-5):161–171. https://doi.org/10.1016/ S0141-0229(99)00008-3 32. Ansorge S, Esteban G, Schmid G (2007) On-line monitoring of infected Sf-9 insect cell cultures by scanning permittivity measurements and comparison with off-line biovolume measurements. Cytotechnology 55 (2-3):115–124. https://doi.org/10.1007/ s10616-007-9093-0 33. Zitzmann J, Sprick G, Weidner T et al (2017) Process optimization for recombinant protein expression in insect cells. In: Gowder SJT (ed) New insights into cell culture technology. InTech Open, London 34. Kost TA, Condreay JP, Jarvis DL (2005) Baculovirus as versatile vectors for protein expression in insect and mammalian cells. Nat Biotechnol 23(5):567–575. https://doi.org/ 10.1038/nbt1095 35. Negrete A, Esteban G, Kotin RM (2007) Process optimization of large-scale production of recombinant adeno-associated vectors using dielectric spectroscopy. Appl Microbiol Biotechnol 76(4):761–772. https://doi.org/10. 1007/s00253-007-1030-9 36. Zeiser A, Elias CB, Voyer R et al (2000) On-line monitoring of physiological parameters of insect cell cultures during the growth and infection process. Biotechnol Prog 16 (5):803–808. https://doi.org/10.1021/ bp000092w 37. Zeiser A, Bdard C, Voyer R et al (1999) On-line monitoring of the progress of infection in Sf-9 insect cell cultures using relative permittivity measurements. Biotechnol Bioeng 63 (1):122–126. https://doi.org/10.1002/( SICI)1097-0290(19990405)63:13.0.CO;2-I 38. Ansorge S, Esteban G, Schmid G (2010) Multifrequency permittivity measurements enable on-line monitoring of changes in intracellular conductivity due to nutrient limitations during batch cultivations of CHO cells. Biotechnol Prog 26(1):272–283. https://doi.org/10. 1002/btpr.347 39. Bluming A, Ziegler J (1971) Regression of Burkitt’s lymphoma in association with measles infection. The Lancet 298(7715):105–106. https://doi.org/10.1016/S0140-6736(71) 92086-1 40. Gopisankar MG, Surendiran A (2018) Oncolytic virotherapy – a novel strategy for cancer therapy. Egyptian J Med Hum Genet 19 (3):165–169. https://doi.org/10.1016/j. ejmhg.2017.10.006

364

ۧer et al. Lukas Ka

41. Ungerechts G, Bossow S, Leuchs B et al (2016) Moving oncolytic viruses into the clinic: clinical-grade production, purification, and characterization of diverse oncolytic viruses. Mol Ther Methods Clin Dev 3:16018. https://doi.org/10.1038/mtm.2016.18 42. Grein TA, Loewe D, Dieken H et al (2018) High titer oncolytic measles virus production process by integration of dielectric spectroscopy as online monitoring system. Biotechnol Bioeng 115(5):1186–1194. https://doi.org/ 10.1002/bit.26538 43. Devaux P, von Messling V, Songsungthong W et al (2007) Tyrosine 110 in the measles virus phosphoprotein is required to block STAT1 phosphorylation. Virology 360(1):72–83.

https://doi.org/10.1016/j.virol.2006.09. 049 44. K€arber G (1931) Beitrag zur kollektiven Behandlung pharmakologischer Reihenversuche. Archiv f experiment Pathol u Pharmakol 162(4):480–483. https://doi.org/10.1007/ BF01863914 45. Levine DW, Wang DIC, Thilly WG (1979) Optimization of growth surface parameters in microcarrier cell culture. Biotechnol Bioeng 21 (5):821–845. https://doi.org/10.1002/bit. 260210507 46. Blu¨ml G (2007) Microcarrier cell culture technology. In: Po¨rtner R (ed) Animal cell biotechnology, Methods in biotechnology. Humana Press, Totowa, NJ

Part V Downstream Techniques

Chapter 21 Continuous Chromatography Purification of Virus-Based Biopharmaceuticals: A Shortcut Design Method Ricardo J. S. Silva, Joa˜o P. Mendes, Manuel J. T. Carrondo, Paula M. Marques, and Cristina Peixoto Abstract Novel biopharmaceutical products, such as vaccines and viral vectors, play a significant role in the development of innovative therapeutic, prophylactic, and clinical applications. However, several challenges are posed when manufacturing these products. The diversity of cell lines and the different physical and chemical properties of these biologicals require the use of different production and processing technologies. Alternative purification strategies that can improve the purification yield, such as continuous chromatography, are regarded nowadays as enabling technologies to overcome some of the bottlenecks in biomanufacturing. This chapter offers a shortcut approach to implement a semi-continuous chromatography purification of hepatitis C virus-like particles produced in insect cells with recombinant baculovirus. Although the purification is based on ion exchange chromatography, the present methodology can be extended to other types of chromatography. Key words Continuous chromatography, Downstream processing, Purification, Vaccines, Virus-like particles

1

Introduction Biopharmaceuticals are one of the fastest growing segments in the pharmaceutical industry having gained significant traction in the past decade. It is expected that by 2020 these products will represent more than a quarter of the pharmaceutical market [1]. Virusbased biologics are one of the most promising biopharmaceuticals of the twenty-first-century medicine and play a significant role in the development of innovative therapeutic, prophylactic, and clinical applications. These biologicals share a high degree of complexity and offer various challenges that require innovative technologies for their manufacturing. With the progression of programs from early phase to commercialization, manufacturing capacity usually needs to be increased by one to two orders of magnitude. Moreover, with the extensive optimization of the upstream production

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_21, © Springer Science+Business Media, LLC, part of Springer Nature 2020

367

368

Ricardo J. S. Silva et al.

BIOREACTION CLARIFICATION ULTRA/DIAFILTRATION INTERMEDIATE PURIFICATION POLISHING CONCENTRATION/FORMULATION STERILE FILTRATION DRUG SUBSTANCE

Fig. 1 Standard flow chart of unit operations in virus purification

observed in the last years, the harvested volumes and yields increased significantly, leading to additional challenges in the downstream process of virus-based biologics manufacturing. Chromatography separation is commonly regarded as the bedrock of downstream processing. The yield, purity, and biological potency of the final viral product resulting from chromatographybased purification strategies surpass that of conventional gradient ultracentrifugation purifications. A series of optimized chromatographic steps are required to obtaining virus preparations of high yield and purity as seen in Fig. 1. Two-step purification protocols, including two orthogonal chromatographic steps or a combination of chromatography with other unit operations such as filtration, are usually performed [2–5]. Chromatography purification can be performed using three different physical arrangements of the stationary phase: packed beds containing particles or beads, membrane adsorbers, and monoliths [6]. Preparative chromatography purification of biomolecules is widely performed using packed porous beds of shaped adsorbent particles. Nevertheless, for the purification of virusbased biopharmaceuticals, they suffer from two main disadvantages: a limited flow rate imposed by the compromise between pressure drop and mass transfer resistances and a low dynamic binding capacity, mainly due to pore exclusion of the viruses limiting adsorption to the outer surface of the beads. Convective chromatography media, such as membranes and monoliths, present

Continuous Chromatography Purification of Virus-Based Biopharmaceuticals. . .

369

considerable improvements in capacity, recovery, and productivity and can overcome the limitations previously discussed of particulate chromatography stationary phases. Such media have larger pore size and use convective transport, making them suitable for processing large molecules at higher flow rates without compromising binding capacities. In addition to the increased productivity, the reduced processing time achieved with these supports is also beneficial for live and attenuated viruses that are labile under the purification process conditions [3, 7]. Different types of chromatographic techniques and supports are available to target the physiochemical properties of the product and related impurities. The selection of the stationary phase also determines the nature of the mobile phase; thus, optimization of both phases is required in order to maximize product potency, quality, and recovery while minimizing the complexity of the separation. In addition, the optimization of a bioseparation is constrained by the product stability under the selected operating conditions of pH, salt concentration, or presence of organic solvents. Reported examples of chromatographic separation of virus particles found in the literature make use of different chromatographic techniques such as size-exclusion chromatography, ion exchange chromatography, affinity chromatography, hydrophobic interaction chromatography, and mixed-mode chromatography [2, 8–10]. Regardless of the stationary phase, chromatography purification can be performed under one of two possible modes: flow through or bind-and-elute chromatography. In the first mode, flow-through purification, impurities are retained in the stationary phase, while the product of interest is collected in the flow-through pool. A prominent example of this application is the polishing step of monoclonal antibodies by anion exchange chromatography. At near neutral pH and low ionic strength, impurities such as DNA, endotoxins, and a great percentage of host cell proteins and viruses are negatively charged. Under these conditions, these species are strongly bound, whereas the positively charged antibody typically flows through the resin bed. In the second operation mode, bind and elute, the product of interest is predominantly retained in the stationary phase, being preferably more strongly adsorbed than the impurities; afterward, desorption occurs either by changing the solvent’s ionic strength or pH. The optimal operation of a chromatography step under each of the aforementioned modes requires knowledge about the adsorption behavior of the product and associated impurities (Fig. 2). For flow-through applications, the length of the loading step is limited by the impurities’ dynamic binding capacity (DBC), since feeding beyond this point will result in co-elution of the impurities with the product. In bind-and-elute chromatography the product’s DBC will determine the length of

Ricardo J. S. Silva et al.

Cout/Cin

1,0

Product

1,0

Fast eluting impurities

Cout/Cin

370

Fast eluting impurities

Product

Slow eluting impurities

0,1

0,1 V*

V*

Fig. 2 Example of flow-through (left) and bind-and-elute chromatography (right); V∗ denotes the volume corresponding to the impurities and product DBC for each case

the feeding step, as loading beyond the defined critical DBC results in product loss. When the inlet stream of a chromatographic column is subject to a step change in the composition, a mass transfer zone (MTZ) develops along the bed (Fig. 3a). From then on, it is possible to distinguish two different regions in the packed bed, until the bed is fully equilibrated with the fluid phase. In the first region, closer to the upstream of the column, the adsorbed phase is in equilibrium with the injected fluid phase. The second region, the MTZ is a transition zone wherein the concentration in the fluid phase changes from the feed value to the previous state before the step change. After a certain period of time, the MTZ reaches the column outlet, and as it is desirable to reduce the product loss, the column feeding is often stopped shortly after a defined breakthrough percentage. As a consequence, a significant part of the static binding capacity (SBC) is left unused. A simple way to avoid product loss and to increase the usage of column capacity is to divide the original column into smaller packed beds connected in series (Fig. 3b). With this column configuration, the effluent of a column can be directed to the downstream adjacent column, thus capturing the MTZ of the upstream column. Once the first column of the setup is fully loaded, it can undergo the typical steps of washing, elution, regeneration, and equilibration. At the end of these steps, the column is reconnected to the end of the train. This cyclic procedure is applied to the next column of the train. In a multicolumn setup, it is possible to achieve higher loadings per unit volume of stationary phase because the MTZ moves along the column train and never exits the system. The multicolumn strategy briefly described above is the basis for the development of various designs and systems that differ in flexibility, number of packed beds, and equipment design [9, 11–15].

Continuous Chromatography Purification of Virus-Based Biopharmaceuticals. . .

371

Unused capacity

a)

t=t2

t=t1

t=0

Z=0

Z=0

Z=0

Z=L

C/C0

C/C0

0

time

0

time

t1 1

b) Z=0

1

3

Z=0

1

2

0

t1

t2

time

3

Z=0

Z=L

2

2

Z=L

Wasted product

C/C0

Z=L

Z=L

3

1

2

3

Z=L

1

2

3

2

3

1

Fig. 3 Batch single-column chromatography (a); multicolumn approach (b). The previous column (a) is subdivided in smaller beds to capture the movement of the MTZ

2

Materials The chromatographic mode selected for purification will define the buffers as well as solutions for column regeneration, cleaning, sanitization, and storage to be used. All solutions should be prepared using ultrapure water and analytical grade reagents and filtered using a 0.2-μm filter (see Notes 1–3). Table 1 lists commonly used buffers for each chromatography mode described that can be used as a starting point for purification. All appropriate safety precautions (see the material safety data sheets of handled chemicals and information provided by the local institution for environmental health and safety) should be followed. Disposal of all reagents and biological material should be performed according to the waste disposal regulations.

Buffers at physiologic pH and ionic strength

Isocratic elution, usually NaCl up to 0.2 M may be used to reduce non-specific adsorption

Affinity (see Note 8)

SEC (see Note 9)

Mixed-mode 5 mM of phosphate buffer pH 7.0 0.5 M of phosphate buffer (hydroxyapatite) (see Note 10)

Buffer solutions with high/ low pH or ionic strength

NaOH

1 M NaCl, dilute alkaline solutions

High salt solutions

Water; 0.5–1 M of NaOH, aqueous solutions of chaotropic salts

Equilibration buffer with a reduced concentration of salt

10–50 mM of buffer with 1–2 M ammonium sulfate or 3 M NaCl

HIC (see Note 7)

Regeneration solution Alkaline (1 M NaOH) or acidic solutions, high concentration salt solutions (2 M NaCl)

Elution buffer

IEX Buffer concentration: 10–50 mM; Equilibration buffer + 1M NaCl (see Notes 4–6) AEX starting pH: 8.0 CEX starting pH: 6.0

Equilibration buffer

Table 1 Starting buffers, regeneration, and storage solutions for different chromatography modes typically used for bioseparation

0.1 M NaOH + 10 mM phosphate

20% ethanol

20% ethanol or binding buffer

20% ethanol or in alternative 0.01 M NaOH

20% ethanol

Storage solution

372 Ricardo J. S. Silva et al.

Continuous Chromatography Purification of Virus-Based Biopharmaceuticals. . .

2.1 Production of Hepatitis C VirusLike Particles in Insect Cells with Recombinant Baculovirus

373

1. Cell culture Erlenmeyer (3 L). 2. Cultivation medium Sf-900 II SFM (Invitrogen). 3. Subculture of Sf9 cells (Life Technologies, Cat# 11496-015). 4. GAG-MLV and HCV-E1E2 recombinant baculovirus stock. 5. Orbital shaker with temperature control unit or suitable for operation in incubators. 6. Serological pipettes (10–100 mL). 7. Pipettes and tips for pipetting small volumes (10–1,000 μL). 8. Cell count system (hemocytometer in combination with trypan blue staining). 9. Trypan blue staining solution.

2.2 Clarification, Concentration, and Buffer Exchange

1. Silicone tubing. 2. Peristaltic pump and pressure transducers. 3. Schott flasks (2 L and 1 L). 4. Clarification filters with 5 and 0.3 μm pore size (Opticap XL 5, Polygard® CN membrane material, EMD Millipore). 5. Ultrapure water and buffer solution (50 mM HEPES, 150 mM NaCl, pH 7.4). 6. Tangential flow filtration cassette of regenerated cellulose with pore rating of 300 kDa and 0.1 m2 (Pellicon® XL, EMD Millipore) with holder.

2.3 Chromatography Method Development and Purification

1. Slurry packed xk16 columns (GE Healthcare) with ion exchange resin (Fractogel® EMD TMAE (M), EMD Millipore) to a final bed volume of 4 mL. 2. Ultrapure water and solutions. Equilibration buffer: 50 mM HEPES, 150 mM NaCl, pH 7.4; elution buffer: 50 mM HEPES, 150–1000 mM NaCl, pH 7.4; storage solution: 20% ethanol. Regeneration solution: 50 mM HEPES, 2 M NaCl, pH 7.4. 3. Virus-like particles feedstock prepared in Subheading 2.2. 4. 50 mL Superloop (GE Healthcare). ¨ kta Explorer 10s, GE Healthcare) 5. Chromatography system (A for batch chromatography. The same system was modified to support simultaneous operation of two columns. Suppliers such as GE Healthcare, Pall Life Sciences, Novasep, Knauer, and Semba Biosciences offer different solutions to perform continuous chromatography (see Note 11).

2.4 Analytical Methods

1. Serological pipettes (for 10–100 mL). 2. Pipettes and tips for pipetting small volumes (10–1000 μL).

374

Ricardo J. S. Silva et al.

3. 96-well microplates. 4. ds-DNA assay Technologies).

kit

(Quant-iT™

PicoGreen®,

Life

5. ELISA kit for p30 quantification (QuickTiter MuLV, Cell Biolabs). 6. Bicinchoninic acid protein assay kit (Pierce™, Thermo Scientific). 7. Immunoenzymetric assay kit for SF9 host cell protein quantification (Cygnus Technologies). 8. Plate reader (Infinite® 200 PRO NanoQuant, Tecan).

3

Methods

3.1 Production of Hepatitis C VirusLike Particles in Insect Cells with Recombinant Baculovirus

Insect cells and viruses are handled inside a sterile laminar flow hood. According to the number of cells counted in the previous subculture step, the cell culture is diluted to the desired density in the desired culture flask. In order to allow good aeration, the volume of the medium should not exceed 1/3 of the flask volume and should not be less than 1/5 to avoid drying out. The main steps to expand SF9 cells and produce Hepatitis C VLPs using recombinant baculovirus are summarized below and should be used to produce approximately 2.5 L of cell culture. 1. Pre-warm the cell culture media for 15 min in 27 water bath.



C

2. Determine viable and total cell counts of the Sf9 subculture and transfer 6.3  108 cells to each 3 L Erlenmeyer. 3. Transfer the pre-warmed media to the Erlenmeyers to a final volume of 1200 mL. 4. Determine viable and total cell counts. Viable cell concentration should be in the range of 3–5  105 cells/mL. 5. Place the cell culture in an incubator shaker with a temperature setpoint of 27  C and an agitation rate of 120 RPM. 6. Monitor the cell viability daily; the doubling time of Sf9 cell ranges between 18 and 22 h. 7. Infect the cell at a cell density of approximately 1.0  106 cells/ mL with a multiplicity of infection of 1 for each baculovirus. 8. Harvest the Erlenmeyers at 96 h post infection. 3.2 Clarification, Concentration, and Buffer Exchange

Clarification is usually performed with microfiltration at constant flux. An efficient clarification step should deliver a low turbidity solution with minimal impact on product recovery, while also removing process and product-related impurities. Tangential flow ultrafiltration for virus concentration is typically conducted at

Continuous Chromatography Purification of Virus-Based Biopharmaceuticals. . . Pressure transducer

a)

P

P

Clarified Material

Harvested Material 5 µm filter

b)

375

0.3 µm filter

Retentate

Crossflow Pump

Clarified Material

Permeate

Diafiltration Buffer Recirculation Vessel

Ultrafiltration cassette

Fig. 4 Assembly of clarification train (a). Illustration of tangential flow filtration setup (b) for ultrafiltration and buffer exchange

constant transmembrane pressure. Ideally the permeate flux should be maintained at an optimum value in order to reduce concentration polarization and membrane fouling which will affect process time and efficiency. The main steps of clarification and tangential flow filtration to prepare the VLP feedstock are the following: 1. Assemble filters and pressure transducers in series as shown in Fig. 4a. 2. Rinse the filters with ultrapure water according to the manufacturer’s instructions. 3. Equilibrate the filters with the buffer solution for a load of 100 L·m2 at a flux of 100 L·m2·h1. 4. Filter the harvested material at a constant flux 100 L·m2·h1, and proceed to tangential flow filtration.

of

5. Assemble the ultrafiltration cassette according to Fig. 4b. 6. Rinse the membrane with ultrapure water according to the manufacturer’s instructions and equilibrate the device with the buffer solution. Discard these solutions. 7. Using a 1 L Schott flask as a recirculation vessel, add 600 mL of the clarified material and draw a line in the flask with a permanent pen in order to keep the liquid level constant in the flask in the next operations.

376

Ricardo J. S. Silva et al.

8. Place the pump inlet in the recirculation vessel, and place the permeate line in the same vessel. Set a recirculation flux of 50 L·m2·h1 and keep the system in full recirculation for approximately 5 min to start membrane polarization. Do not apply pressure to the retentate line. 9. After this period, place the permeate line in a separate vessel and ramp up the cross flow flux in small increments until reaching 300 L·m2·h1. Apply pressure to the permeate line in order to maintain a stable permeate flux of 30 L·m2·h1. After reaching the desired cross flow flux, start increasing the transmembrane pressure (TMP ¼ (P1 + P2)/2 – P3) until reaching 0.4 bar. 10. Start the feed addition pump, and set a flow rate that matches the permeate velocity in order to maintain a constant volume on the recirculation vessel. 11. Continue operation until all clarified material is used. 12. Start diafiltration by placing the feed line in a flask containing 2400 mL of buffer. 13. After all the buffer is consumed, ramp down the cross flow until full stop in small intervals. Drain the system to recover the sample. 3.3 Chromatography Method Development and Purification 3.3.1 Capacity Determination: Breakthrough Experiments

1. Connect the columns to the chromatography system and thoroughly prime the system and columns with binding buffer (see Notes 12 and 13). 2. Set the UV monitor to 280 and 260 nm and perform an autozero. 3. Fill the superloop with 20 ml of sample. Place the columns in bypass and inject the sample directly to the UV detector at 0.5 mL/min. The absorbance signal should achieve a stable level. Record the absorbance values for each wavelength monitored. These values will later be used to assess column saturation (see Fig. 5a). 4. Wash the sample from the system with the binding buffer. Afterward, place one of the columns in line and equilibrate with binding buffer for approximately 10 column volumes (CV), or until UV, pH, and/or conductivity signals are stable (see Note 14). 5. Fill the superloop with the virus feedstock. Perform an autozero to the UV monitor. Set the flow rate to 100 cm·h1 and inject the sample. Set the fraction collector to recover the flow through in 2–5 mL fractions (see Note 15). 6. Perform a column wash with binding buffer for 10 CV, or until UV, pH, and/or conductivity signals are stable.

1000

UVsample

UV signal (mAU)

a) 1500

500

0 0

10 Eluted volume (mL)

20

1000

v0

v* UVsample

UV signal (mAU)

b) 1500

500 UVNonBind

0 0

10 20 Eluted volume (mL)

30

c) 1.00

UV signal (mAU)

0.75

0.50

0.25

v0

v10%

0.00 0

10 20 Eluted volume (mL)

30

Fig. 5 UV signal of sample feedstock during column bypass (a). Impurities and product breakthrough curves (b); v0 and v∗ are the retention volumes of the non-binding impurities and product, respectively. Normalized product breakthrough curve (c)

378

Ricardo J. S. Silva et al.

7. Elute the sample with a linear gradient from 0% to 100% with elution buffer using 20 CV, and hold the gradient at 100% for 5 CV. 8. Rinse the column with water and afterward, regenerate the column with 1M NaOH at a flow rate of 0.1 CV during 30 min. Rinse the column with water followed by binding buffer until UV, pH, and/or conductivity signals are stable. 9. Repeat steps 4–7 with the unused column. 10. Open the evaluation software and in each experiment, determine v0, UVsample, and UVNon Bind as seen in Fig. 5b. v0 and UVNon Bind correspond to the elution volume and UV intensity of the non-binding impurities and UVsample to total intensity of VLPs and impurities (see Note 16). 11. The breakthrough percentage can be calculated using the following relationship: BT% ¼ (UV∗  UVNon Bind)/(UVsam∗ ple  UVNon Bind), where UV is an absorbance value selected in the second breakthrough observed. To calculate dynamic binding capacity (DBC) associated to UV∗, the following relationship should be used: DBC ¼ CFEED. (v∗  v0)/VC, where v∗ relates to UV∗, CFEED is the feed concentration of the VLPs, and VC is the column volume. Figure 5b can be further simplified by offsetting the absorbance values by UVNon Bind, followed by a normalization dividing the obtained absorbance values by (UVsample  UVNon Bind). This will result on a breakthrough curve that varies between 0 and 1 as seen in Fig. 5c. 3.3.2 Screening of Elution Conditions

From the gradient elution experiments in Subheading 3.3.1, two or more peaks should be identifiable. By quantifying VLPs in the recovered elution fractions, it is possible to identify the peak that corresponds to this species. Additionally, the conductivity trace allows to recover the composition of the elution buffer that corresponds to each peak. For the present case, VLPs should be the first eluting species with a conductivity window of 45 to 55 mS.cm1. Elution can be optimized by moving from gradient to step elution, in order to concentrate the product of interest and rapidly remove them from unwanted substances. This will reduce separation time and buffer consumption. Composition of the steps should be selected based on the elution range of each group. In the present case, after loading and column washing, the first elution step should be performed at a composition that provides 5–10 mS·cm1 above the elution conductivity of the group of interest. In addition, the composition selected should be low enough to avoid co-elution of the other protein groups. The main steps to optimize the composition of the elution step are the following:

Continuous Chromatography Purification of Virus-Based Biopharmaceuticals. . .

379

1. Fill the superloop with enough volume to reach a breakthrough level of 50%. See Subheading 3.3.1. 2. Equilibrate column with 5–10 CV of binding buffer or until the UV, pH, and conductivity signals are stable. 3. Set the linear velocity to 100 cm·h1 and inject the sample. 4. Wash the column with binding buffer for 5–10 CV or until the UV, pH, and conductivity signals are stable. 5. Set the fraction collector to recover 1 mL fractions and perform a step elution with 45% of elution buffer. After UV and conductivity signals stabilize, perform a second elution step with 100% of elution buffer. Recover the fractions and analyze them for VLP and impurities content. 6. Rinse the column with water and afterward, regenerate the column with 1M NaOH at a flow rate of 0.1 CV during 30 min. Rinse the column with water followed by binding buffer. 7. Repeat steps 1–6 changing the composition of the first elution step from 45 to 50% and on a third experiment to 55% of elution buffer (see Notes 17 and 18). The optimal elution composition will minimize impurities but also maximize product recovery. From these experiments, it is also possible to optimize the length of the washing and elution steps. 3.3.3 Design of Operating Cycle and Semi-continuous Purification Run

The design of the operating cycle for a semi-continuous purification of hepatitis C VLPs requires the knowledge about product dynamic binding capacity and breakthrough profile (from Subheading 3.3.1), an optimized elution buffer, and information on the duration of wash, elution, and regeneration steps (from Subheading 3.3.2). Table 2 summarizes the data required to setup the operating cycle (see Note 19). The distribution and sequence of steps depends not only on the number of columns and pumps available in the chromatographic system to deliver each of the buffers and feedstock, but also on the relative duration of the feed step versus the wash, elution, regeneration, and equilibration. For the present case, the feed step (6 CV) is shorter than all the remaining steps (15 CV). Consequently, the operation needs to be distributed between the two columns as seen in Fig. 6a. The logical temporal arrangement of operation is depicted in Fig. 6b for the first half cycle (the second half cycle has a similar periodic arrangement of the steps, with the inlet and outlet ports exchanging columns) and described as follows: Feed during 5 CV to reach a 10% BT in column 1, connect the two columns in series, and continue loading for 1 CV (column 1 will reach 50% BT approximately). To reduce product loss, the columns are kept connected during the wash step (5 CV). Meanwhile,

380

Ricardo J. S. Silva et al.

Table 2 Summary of the information required for the design of a semi-continuous purification step using a two-column system DBC (10% BT)

5 CV

DBC (50% BT)

6 CV

Wash (Wsh)

5 CV

Elution (ELT)

3 CV

Regeneration (R)

4 CV

Equilibration (Eq.)

3 CV

Equilibration buffer

50 mM HEPES, 150 mM NaCl, pH 7.4

Wash buffer

50 mM HEPES, 150 mM NaCl, pH 7.4

Elution buffer

50 mM HEPES, 550 mM NaCl, pH 7.4

Regeneration buffer

50 mM HEPES, 2000 mM NaCl, pH 7.4

DBC stands for dynamic binding capacity

a)

Feed 5 CV

Column 1

Feed 1 CV

v = 3.35 mL/min

Elution 3 CV

Regeneration 4 CV

Column 2

Wash 5 CV v = 3.35 mL/min

Equilib. 3 CV

v = 6.7 mL/min

v = 3.35 mL/min

interconnected stage

b)

F

E

col.1

col.2

P

W F

R

col.1

col.2

W

W F

Eq

col.1

col.2

W

W F

col.1

col.2

Eq

col.1

col.2

W W Fig. 6 Scheduling of process steps and flow rates for columns 1 and 2 (a); Column 2 is connected in series with column 1 in order to capture the eluting MTZ (interconnected stage). Step temporal arrangement for the first half cycle (b); the second half cycle will be similar with columns exchanging places. F, E, R, and Eq. denote feed, elution, regeneration, and equilibration inlets, respectively; W and P denote waste and product outlets, respectively

Continuous Chromatography Purification of Virus-Based Biopharmaceuticals. . . a)

381

col. 1 col. 2 10% BT

b)

50% BT c)

d)

e)

f)

g)

h)

Feed

Wash

Elution

Regen.

Equilib.

Fig. 7 Visual illustration of a shortcut algorithm for implementing a two-column semi-continuous chromatography process

column 2 was eluted during 3 CV, regenerated during 4 CV, and equilibrated for 3 CV in order to receive the effluent of column 1. The algorithm to reach this sequence of steps is summarized below: 1. Draw a temporal line for columns 1 and 2. The number of subunits should match the sum of number of CV of the feed and wash steps (Fig. 7a).

382

Ricardo J. S. Silva et al.

2. The first interval corresponds to the loading of 5 CV of feedstock, enabling 10% of BT in column 1 (Fig. 7b). Columns 1 and 2 are connected in series and one additional CV is loaded (Fig. 7c). 3. The CV correspondent to the wash step are added (Fig. 7d) with both column kept connected. The connection performed in steps c and d has a practical implication of synchronizing the flow rates in both columns. 4. In order to accommodate the remaining steps, the available space in column 2 is subdivided (Fig. 7e). The proportion chosen will affect the flow rate of column 2. In the present case, the proportion is 2:1, meaning that the flow rate in column 2 will be doubled in these steps. 5. Elution, regeneration, and equilibration are placed by order in column 2 (Fig. 7f–h). 3.4 Analytical Methods

4

Analytical methods for the quantification of ds-DNA, p30 protein (VLPs), total protein, and host cell protein are carried using commercially available kits described in Subheading 2.3, item 3 and should be performed according to the manufacturer’s instructions.

Notes 1. Before starting any experiment, thoroughly familiarize yourself with the chromatographic system and software tools. 2. Previous knowledge about the biological system—species isoelectric point and stability under binding and elution conditions—is critical to set up an efficient purification method. 3. Sufficient amounts of buffers and solutions for column regeneration, cleaning, sanitization, and storage should be prepared beforehand. Using different lots of these solutions might influence the efficiency and reproducibility of the purification. 4. Binding in IEX is usually performed at low salt concentrations, with elution occurring with the increase of this parameter. 5. In IEX, the range of pH values to be used should be compatible with the stability of the product of interest. If the isoelectric point of the target protein is known, the range of pH values can be reduced to one unit above or below this point. 6. IEX column regeneration with high salt concentration might not provide a full capacity recovery. The use of organic solvents such as 30% (v/v) isopropanol may be required. 7. To enhance hydrophobic interaction, the protein mixture is loaded on the column in a buffer with a high concentration of salt. Elution occurs by reduction of the salt concentration.

Continuous Chromatography Purification of Virus-Based Biopharmaceuticals. . .

383

The elution/precipitation strength of an ion is described by the Hofmeister series. 8. Binding of the molecules to the affinity ligand usually is promoted at physiological conditions, and elution happens by an altered conformation via a changed buffer pH or ionic strength or competitive displacement. 9. Due to the isocratic separation principle of SEC, the buffers used in this type of chromatography are usually not limited in terms of ionic strength or pH. 10. Binding to hydroxyapatite is often performed with dilute phosphate buffers. Elution is promoted by increasing the phosphate buffer concentration. 11. Familiarize yourself with the different mechanical limitations of the hardware, stationary phases, and chromatography equipment. Review manufacturer’s indications on pressure limits and chemical compatibility of chromatography media and equipment. 12. Different solutions may present different viscosities and, subsequently, different pressure drops across the chromatographic bed. This is often critical when using storage solutions such as 20% ethanol. 13. When working with buffer solutions, the chromatographic systems should be flushed at the end of the experiments. This is often critical for the elution lines and pumps used that are in contact with high salt solutions. 14. Maintain sample, start and elution buffers, columns, and chromatographic equipment at the same, constant temperature throughout a separation to ensure consistent and reproducible results. 15. A superloop is convenient for sample volumes bellow 50 mL or single injections close to this volume. If multiple injections are to be performed, or the volume to be injected is higher than 50 mL, a sample pump becomes essential. 16. Some chromatography systems allow choosing between different UV cells with different path lengths. Ideally, when performing capacity determinations with breakthrough experiments, the UV signal of the sample injected directly to the UV detector should be below 1000 mAU. 17. In bind-and-elute purification, conditions are usually chosen in such a way that the target product is strongly adsorbed. Therefore, the absolute product quantity loaded determines column capacity rather than product concentration.

384

Ricardo J. S. Silva et al.

18. During the high salt elution and re-equilibration steps, time can be saved by using higher flow rates. However, the maximum recommended flow for the medium should not be exceeded. 19. Column scale-up can be achieved by increasing column diameter while keeping bed height and linear flow rate constant. The linear flow rate can be calculated by dividing the volumetric flow rate by the column’s cross-sectional area.

Acknowledgments The authors would like to acknowledge Ana Coroadinha, Rute Castro, and Joa˜o Clemente (iBET) for the production of the HCV-VLPs and Alex Xenopoulos (EMD Millipore) for the useful discussions. References 1. Deloite (2018) Global life sciences outlook. Innovating life sciences in the fourth industrial revolution: embrace, build, grow. https:// www2.deloitte.com/content/dam/Deloitte/ global/Documents/Life-Sciences-HealthCare/gx-lshc-ls-outlook-2018.pdf 2. Wolf MW, Reichl U (2011) Downstream processing of cell culture-derived virus particles. Expert Rev Vaccines 10:1451–1475 3. Nestola P, Peixoto C, Silva RRJS et al (2015) Improved virus purification processes for vaccines and gene therapy. Biotechnol Bioeng 112:843–857 4. Moleirinho MG, Rosa S, Carrondo MJT et al (2018) Clinical-Grade Oncolytic Adenovirus Purification Using Polysorbate 20 as an alternative for cell lysis. Curr Gene Ther 18:366–374 5. Carvalho SB, Silva RJS, Moreira AS et al (2019) Efficient filtration strategies for the clarification of influenza virus-like particles derived from insect cells. Sep Purif Technol 218:81–88 6. Jungbauer A (2005) Chromatographic media for bioseparation. J Chromatogr A 1065:3–12 7. Zhao M, Vandersluis M, Stout J et al (2018) Affinity chromatography for vaccines manufacturing: finally ready for prime time? Vaccine 37(36):5491–5503 8. Carvalho SB, Fortuna AR, Wolff MW et al (2018) Purification of influenza virus-like particles using sulfated cellulose membrane

adsorbers. J Chem Technol Biotechnol 93:1988–1996 9. Fischer LM, Wolff MW, Reichl U (2018) Purification of cell culture-derived influenza A virus via continuous anion exchange chromatography on monoliths. Vaccine 36:3153–3160 10. Nestola P, Silva RJS, Peixoto C et al (2014) Adenovirus purification by two-column, sizeexclusion, simulated countercurrent chromatography. J Chromatogr A 1347:111–121 11. Nestola P, Silva RJS, Peixoto C et al (2015) Robust design of adenovirus purification by two-column, simulated moving-bed, sizeexclusion chromatography. J Biotechnol 213:109–119 12. Silva RJS, Rodrigues RCR, Mota JPB (2012) Relay simulated moving bed chromatography: concept and design criteria. J Chromatogr A 1260:132–142 13. Arau´jo JMM, Rodrigues RCR, Silva RJS et al (2007) Single-column simulated moving-bed process with recycle lag: analysis and applications. Adsorpt Sci Technol 25:647–659 14. Silva RJS, Mota JPB, Peixoto C et al (2015) Improving the downstream processing of vaccine and gene therapy vectors with continuous chromatography. Pharm Bioprocess 3:489–505 15. Steinebach F, Mu¨ller-Sp€ath T, Morbidelli M (2016) Continuous counter-current chromatography for capture and polishing steps in biopharmaceutical production. Biotechnol J 11:1126–1141

Chapter 22 Single Step Purification of Glycogen Synthase Kinase Isoforms from Small Scale Transient Expression in HEK293 Cells with a Calcium-Dependent Fragment Complementation System Gavin McGauran, Sara Linse, and David J. O’Connell Abstract Purification of proteins for the biophysical analysis of protein interactions occurring in human cells can benefit from methods that facilitate the capture of small amounts of natively processed protein obtained using transient mammalian expression systems. We have used a novel calcium-dependent fragment complementation-based affinity method to effectively purify full length glycogen synthase kinase 3 (GSK3) α and β isoforms to study their interaction with amyloid β peptide (Aβ42). Using these proteins, purified from 1 mg of total cell lysate, we measured an apparent KD of 100 pM between GSK3α/β and immobilized Aβ42 with surface plasmon resonance technology. This approach can be used to retrieve useful quantities of protein for biophysical experiments with small scale mammalian cell culture. Key words EF hand, Calcium, Glycogen synthase kinase, Human embryonic kidney cells, Single step purification, Biophysical, Surface plasmon resonance, Fragment complementation

1

Introduction Purification of mammalian proteins from mammalian systems is more challenging at small scale than from bacterial systems, while bacterial expression systems lack the requisite translational and post-translational machinery to express natively folded, fully functional human proteins [1]. Ideally, characterization of the interaction between human proteins will be with proteins expressed in a mammalian host system to ensure a fully biologically active protein. However, purification of proteins from mammalian systems often requires multi-step purification procedures, and the steps involved can be expensive, time consuming and with limited yield, where a simple, single step approach to rapidly prepare purified protein from standard adherent cell culture systems would be very beneficial to many laboratories.

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4_22, © Springer Science+Business Media, LLC, part of Springer Nature 2020

385

386

Gavin McGauran et al.

We have developed a novel affinity system based on the reconstitution of a small calcium binding protein, calbindin D9k, from two subdomains (EF1 and EF2). Structural biology studies, where this protein was divided into the two subdomains, revealed a high affinity binding between them in the presence of Ca2+ ions [2]. Fragment complementation between free EF1 and immobilized EF2 was measured to have a KD of approximately 80 pM, and the complementation did not occur in the absence of Ca2+ ions. We have tested this complementation as the basis of an affinity purification method of EF1 tagged recombinant proteins with EF2 functionalized agarose affinity resins. Recently, we reported purification of a membrane-bound G protein-coupled receptor, CB2, and performed biophysical analysis using this method [3]. Here, we describe application of this method to validate an interaction between amyloid β peptide and glycogen synthase kinase 3 α (GSK3α) discovered by protein microarray screening [4]. GSK3 is a proline-directed serine/threonine kinase ubiquitously expressed in mammalian tissues and is involved in multiple signaling pathways [5, 6]. GSK3 has three isozymes, GSK3α (51 kDa), GSK3β (48 kDa), and GSK3β2, a splice variant containing an insertion of 13 amino acids in the catalytic domain [7, 8]. These kinases plays a fundamental role in cell differentiation, cell division, stem cell renewal, proliferation, apoptosis, insulin regulation, synaptic plasticity, and learning and memory [8– 11]. Studies into a potential role of these kinases in the pathology of neurodegenerative diseases including Alzheimer’s disease, schizophrenia, and bipolar disorders have previously been published [12–15]. Here, we describe transient transfection of adherent HEK293T cells to express EF1-GSK3α and EF1-GSK3β with expression for 24 h (Fig. 1) and purification of these proteins with EF2-agarose resin prior to their use in surface plasmon resonance analysis of binding kinetics to Aβ42. This simple methodology is useful for the rapid characterization of protein-protein interactions with human proteins expressed in human cells.

2 2.1

Materials Transfection

1. HEK293T cells grown in 10 cm3 tissue culture dishes. 2. TransIT-2020 lipid-based transfection reagent (Mirus).

2.2 Harvesting of Cells and EF1-Tagged Protein Purification Using EF2 Agarose Beads

1. Lysis buffer: 50 mM Tris–HCl, 200 mM NaCl, 0.5% CHAPS, 0.1% CHS, 1% DDM, 30% glycerol, 2 mM CaCl2, pH 7.5, and protease inhibitors. 2. Elution buffer: 10 mM HEPES, 150 mM NaCl, 10 mM EDTA, pH 8.0. 3. Agarose beads (Pierce Biotechnology, Rockford, IL, USA).

Single Step Purification of Glycogen Synthase Kinase Isoforms. . .

387

Fig. 1 Schematic of small scale expression and purification of EF1-GSK3 protein. (a) EF1 amino acid sequence is expressed at the N terminus of GSK3α. (b) Transient transfection of HEK293 cells in 2  10 cm3 culture dishes. (c) Schematic of calcium-dependent fragment complementation-based affinity capture of EF1-GSK3α with EF2 agarose

2.3 Sodium Dodecyl SulfatePolyacrylamide Gel Electrophoresis (SDS PAGE)

1. 4 Laemmli buffer: 166 mM Tris–HCl pH 6.8, 6.6% (w/v) SDS, 0.33% (w/v) bromophenol blue, 6% (v/v) β-mercaptoethanol. 2. Stacking gel (5%): 6.8 ml deionized H2O, 1.7 ml 30% acrylamide mix, 1.25 ml 1 M Tris (pH 6.8), 0.1 ml 10% SDS, 0.1 ml 10% ammonium persulfate, 0.01 ml tetramethylethylenediamine (TEMED). 3. Resolving gel (7%): 5 ml deionized H2O, 2.35 ml 30% acrylamide mix, 2.5 ml 1 M Tris (pH 6.8), 0.1 ml 10% SDS, 0.1 ml 10% ammonium persulfate, 0.007 ml TEMED). 4. Running buffer: 25 mM Tris, 192 mM glycine, 0.1% (w/v) SDS. 5. Mini-Protean III electrophoresis kit (Bio-Rad).

388

2.4

Gavin McGauran et al.

Western Blotting

1. Transfer buffer: 48 mM Tris, 39 mM glycine, and 20% (v/v) methanol. 2. Tris-buffered saline Tween 20 (TBS-T): 50 mM Tris–HCl, 166 mM NaCl, 0.5% (v/v) Tween-20. 3. Blocking solution of non-fat dry milk (5%) dissolved in TBS-Tween 20. 4. Polyvinylidene difluoride (PVDF) membrane. 5. BM Chemiluminescence Western Blotting Substrate (Roche).

2.5 Colloidal Coomassie

1. Fixation buffer: 47% deionized H2O, 50% ethanol, 3% phosphoric acid. 2. Coomassie brilliant blue G-250 solution: 63% deionized H2O, 34% methanol, 3% phosphoric acid, 15% (w/v) ammonium sulfate, and 0.1% (w/v) Coomassie brilliant blue G-250.

2.6 Surface Plasmon Resonance (SPR)

1. Running buffer: 10 mM HEPES/NaOH, 0.15 M NaCl, 3 mM EDTA, 0.005% Tween 20, pH 7.4. 2. Biacore 3000 (GE Healthcare, Uppsala, Sweden). 3. CM5 (carboxymethylated dextran sensor chips), 0.25 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide, 0.05 M N-hydroxysuccinimide (GE Healthcare, Uppsala, Sweden). 4. Lyophilized monomeric Aβ42 was dissolved in 10 mM sodium acetate buffer pH 3 at a concentration of 10 μM monomers. 5. EF1-GSK3α and EF2-GSK3β diluted in running buffer.

3 3.1

Methods Transfection

1. HEK293T cells are seeded at 800,000 cells per 10 cm3 dish with Dulbecco’s Modified Eagle Media containing 10% fetal bovine serum (FBS) supplemented with penicillin and streptomycin. 2. Cells are grown to 60–80% confluency prior to transfection. 3. Prior to transfection, remove media and wash cells with phosphate buffered saline (PBS). 4. 10% FBS containing media supplemented without penicillin and streptomycin is added and cells incubated at 37  C. 5. Transfection is performed as per manufacturer’s protocol using 5 μg DNA and 2.4 μl TransIT-2020 per μg of DNA made up in 1ml serum free media (see Note 1). 6. The TransIT-2020 reagent-DNA complex mix is incubated for 20 min at room temperature.

Single Step Purification of Glycogen Synthase Kinase Isoforms. . .

389

7. Samples are then added drop-wise to pre-warmed media containing 10% FBS without penicillin and streptomycin and agitated (see Note 2). Incubate cells at 37  C for 24 h. 3.2 Harvesting and Lysis of Cells

1. Discard media and wash cells twice with PBS (see Note 3). Cells are gently harvested in 750 μl of lysis buffer per 10 cm3 dish using a cell scraper. 2. Cell pellets are mechanically disrupted via aspiration through a 21-gauge needle in the lysis buffer (see Notes 4 and 5). 3. Collect cell pellet in a microcentrifuge at 10,000  g for 10 min at 4  C, retaining the supernatant. 4. The total protein concentration is determined using bicinchoninic acid (BCA) assay (Pierce) according to the manufacturer’s protocol. 5. Expression of full length EF1-GSK3α and EF1-GSK3β is confirmed by Western blot (Fig. 2).

3.3 Purification of EF1-GSK3α and EF1GSK3β Using EF2 Agarose Beads

1. EF2 agarose and control beads are collected at 1000  g for 1 min at 4  C. 2. Beads are washed five times in 1 ml lysis buffer. 3. 1.5 mg of lysate samples are pre-cleared with incubation at 4  C for 1 h using non-conjugated beads. 4. The agarose beads are collected at 5000  g at 4  C for 1 min and the supernatant retained (see Note 6). 5. The pre-cleared lysate is then incubated overnight at 4  C with EF2 agarose beads or non-conjugated beads as a negative control. 6. Following incubation, samples are collected at 5000  g for 1 min at 4  C. Samples are washed five times in lysis buffer (see Note 7). 7. Following the wash steps, EF1-GSK3α and EF1-GSK3β are eluted from the EF2-agarose beads with elution buffer by re-suspension, and samples are incubated for 5 min at room temperature. 8. Beads are then collected at 5000  g for 1 min and the supernatant harvested (see Note 8). 9. Repeat the elution 5 times. 10. Samples are stored at 80  C prior to further analysis by colloidal Coomassie or surface plasmon resonance.

3.4 Sodium Dodecyl SulfatePolyacrylamide Gel Electrophoresis (SDS-PAGE)

1. The protein concentration of each sample is determined using a BCA assay and equal amounts of protein (10 μg) added to Laemmli buffer (4) and boiled for 5 min. 2. Proteins are separated by mass with a stacking gel (5%) and a resolving gel (7%).

Fig. 2 Expression of full length EF1-GSK3α and β isoforms. (a) Pileup of the amino acid sequence of GSK3 isoforms with the N-terminal amino acids of GSK3β aligned to the glycine-rich N terminus of GSK3α highlighted (red box) (b) (i) detection of EF1 affinity tag with anti-EF1 antibody, (ii) detection of EF1-GSK3α with anti-GSK3α antibody, and (iii) detection of EF1-GSK3β with anti-GSK3β antibody. Detection of levels of GAPDH with anti-GAPDH antibody shown as loading control in panels below

Single Step Purification of Glycogen Synthase Kinase Isoforms. . .

391

3. Electrophoresis is carried out in running buffer at constant voltage of 90V through the stacking and 120V through the resolving gel with a Mini-Protean III apparatus. 4. When the proteins have migrated sufficiently through the resolving gel, stop the current and transfer the gels to polyvinylidene difluoride (PVDF) membrane for blotting or stain the gel with colloidal Coomassie. 3.5

Western Blotting

1. Following SDS-PAGE, transfer the gels to a dish along with sponges and Whatman filter paper containing ice cold transfer buffer. 2. Prior to protein transfer, the PVDF membrane must be activated in methanol (100%) for 5–20 s, washed in deionized H2O for between 0.5 and 2 min, and equilibrated in transfer buffer for 3–5 min. 3. The gel is then placed on two sheets of Whatman filter paper that are placed on a sponge, and the gel then overlaid with the PVDF membrane. Two more sheets of Whatman filter paper are placed on top of the membrane, and then a second sponge and the assembly is placed in a transfer cassette with the gel closest to the black side of the cassette (cathode). 4. Transfer proteins to the PVDF membrane using wet transfer for 1.5 h at a constant current of 350 mA. 5. Following transfer, the PVDF membrane is blocked in a blocking solution for 1 h at room temperature. 6. Next incubate the membrane in the primary antibody at a dilution of approximately 1 μg/ml in blocking solution overnight at 4  C. 7. After overnight incubation, wash the membrane in TBS-Tween for 15 min three times, rocking at room temperature. 8. Incubate the membrane with the appropriate HRP-conjugated secondary antibody at approximately 0.1 μg/ml for 1 h at room temperature. 9. After 1 h of secondary incubation, wash the membrane in TBS-Tween for 15 min three times, rocking at room temperature. 10. The protein bands on the PVDF membrane can then be visualized using a chemiluminescence reagent and exposure of the membrane to an X-ray film in the darkroom.

3.6 Colloidal Coomassie

1. Following SDS-PAGE, gels are fixed in fixation buffer by rocking at room temperature for 2 h. 2. Wash gels three times in deionized H2O for 15 min and stain in Coomassie brilliant blue G-250 solution overnight (see Notes 9 and 10). 3. Wash the gels three times in deionized H2O for 15 min and visualized (Fig. 3).

392

Gavin McGauran et al.

Fig. 3 Purification of EF1-GSK3 isoforms with EF2-agarose. (a) Control purification of EF1-GSK3α with control agarose beads lacking EF2 ligand. (i) Colloidal Coomassie of EF1-GKS3α samples. (1) Input, (2) wash fraction 1 (W1), (3) W2, (4) W3, (5) W4, (6) W5, (7) elution fraction 1 (E1), (8) E2, (9) BSA (100 ng). (ii) Anti-EF1 Western blot of EF1-GKS3α samples. (1) Input, (2) supernatant, (3) W1, (4) W2, (5) W3, (6) W4, (7) W5, (8) E1, (9) E2, (10) E3, (11) E4, (12) E5. (b) Purification of EF1-GSK3α with EF2-agarose beads. (i) Colloidal Coomassie of purified EF1-GSK3α samples. (1) Input, (2) W1, (3) W2, (4) W3, (5) W4, (6) W5, (7) E1, (8) E2, (9) BSA (100 ng). (ii) Anti-EF1 Western blot of purified EF1-GKS3α samples. (1) Input, (2) precleared input, (3) supernatant, (4) W1, (5) W2, (6) W3, (7) W4, (8) W5, (9) E1, (10) E2, (11) E3, (12) E4, (13) E5. (c) Purification of EF1-GSK3β with EF2-agarose beads. (i) Colloidal Coomassie of purified EF1-GSK3β samples. (1) Input, (2) W1, (3) W2, (4) W3, (5) W4, (6) W5, (7) E1, (8) E2, (9) BSA (100 ng). (ii) Anti-EF1 Western blot of purified EF1-GKS3α samples. (1) Input, (2) precleared input, (3) supernatant, (4) W1, (5) W2, (6) W3, (7) W4, (8) W5, (9) E1, (10) E2, (11) E3, (12) E4, (13) E5

Single Step Purification of Glycogen Synthase Kinase Isoforms. . .

3.7 Surface Plasmon Resonance (SPR)

393

1. SPR binding studies are performed using a Biacore 3000 instrument with running buffer at a flow rate of 10 μl/min. 2. Prior to immobilization, activate all four flow channel surfaces of CM5 carboxymethylated dextran sensor chip surfaces with a mixture of 0.25 M 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and 0.05 M N-hydroxysuccinimide in water. 3. Couple Aβ42 to the activated surface by preparing a fresh 100 μl solution of 10 μM monomers in 10 mM sodium acetate buffer pH 3, and inject in channels 2–4 (see Note 11). 4. All surfaces are subsequently blocked by injecting 70 μl of 1 M ethanolamine, with empty, blocked channel 1 surface serving as the negative control. 5. To study the association of EF1-GSK3α or EF1-GSK3β proteins with immobilized Aβ42, inject EF2 agarose purified proteins at concentrations ranging from 1 to 60 nM for 5–20 min, followed by buffer flow from 15 min to 24 h (Fig. 4). 6. Subtract the recorded background signal of the control channel from the values obtained from channels with immobilized Aβ42.

3.8 Fitting of SPR Sensorgrams for Affinity Measurement

1. The dissociation phase data minus baseline are fitted to a single exponential decay: Y ¼ A n˜ exp ðkoff n˜ t ÞÞ where A is the amplitude and koff is the dissociation rate constant. 2. The association phase data minus baseline are fitted to one minus a single exponential decay: Y ¼ A n˜ ð1  exp ððC˜ nkon þ koff Þ˜ nt ÞÞ˜ nC˜ nkon= ðC˜ nkon þ koff Þ where A is the amplitude, C is the GSK3α concentration in the flow, kon is the association rate constant, and koff is the dissociation rate constant. 3. An estimate of KD is obtained from the ratio of the rate constants as K D ¼ koff =kon

4

Notes 1. The transfection conditions are very important to get maximum expression for a small scale purification experiment. It is essential to use high-quality DNA preparations. To achieve

394

Gavin McGauran et al.

Fig. 4 SPR analysis of GSK3α and β binding to immobilized Aβ42. (a) GSK3α association phase following injection of 20 nM (red), 10 nM (orange), 5 nM (green), and 2.5 nM (blue) protein solution for 20 min over a CM5 with immobilized Aβ42. (b) Dissociation phase for 24 h with buffer flow. (c) GSK3β association phase following injection of 20 nM (red), 10 nM (orange), 5 nM (green), and 2.5 nM (blue) protein solution for 20 min over a CM5 with immobilized Aβ42. (d) Dissociation phase for 24 h with buffer flow. Each data set is the average over three repeats

this, plasmid DNA is prepared using the Plasmid Maxiprep kit (Qiagen). Plasmid DNA is always diluted in sterilized, deionized H2O to a concentration of 1 μg/μl. 2. To ensure all cells come into contact with the DNA-transfection complex, add the mixture drop-wise to the cells. It is also important to mix the 10 cm3 plate gently to maximize the distribution of the transfection solution. The transfection efficiency can be monitored and optimized using a GFP reporter plasmid and flow cytometry to determine approximately the yield of transfection efficiency. Usually

Single Step Purification of Glycogen Synthase Kinase Isoforms. . .

395

from this type of an experiment, a transfection efficiency of 60% is achieved. 3. HEK293T cells are loosely attached to the plate surface, and this can lead to loss of material during the harvesting steps. To reduce this, 5 ml of PBS is added to the side of the dish which is tilted at an angle. PBS should not be added directly to the flat surface or this will dislodge cells attached on the surface. Removal of the PBS should be done causing as little disruption to the cells simply by pouring off the PBS. 4. The lysis buffer employed here is adapted from a membrane solubilization buffer used to purify G protein-coupled receptors, and this is found to be particularly suitable for purification of the kinase proteins. 5. Disruption of the cells is performed using a syringe and a 21-gauge needle to break open the cells. Moving the syringe lever up and down in a 1.5-ml microcentrifuge tube should be done slowly so as not to lose material, and this should be repeated at least 5 times. 6. After incubating the lysates with the control agarose resin, it is important to ensure that the supernatant is carefully retained. Before the incubation with the EF2-agarose resin, some of the sample should be kept for analysis to determine if there is loss of the target protein in this step. This sample should be compared against the lysate prior to incubation with the control resin. 7. The supernatant and wash buffers should also be kept to determine how much protein is captured by the resin as this will determine if there has been overloading of the EF1-tagged protein, which may lead to loss of material during the washes or will indicate if more material can be applied to EF2-agarose resin to maximize the amount of material captured. 8. To increase the amount of target protein eluted in the first few elution steps, constantly mix the resin by re-suspension over the 5 min incubation step to avoid agarose settling to the bottom of the 1.5-ml microcentrifuge tube. To prevent uptake of the EF2-agarose resin, use a p10 tip that can be added to the top of a p1000 tip, which due to its small pore size will decrease the chances of aspirating resin. Alternatively, a narrow bore gel-loading tip may be used. 9. When making up the Coomassie brilliant blue G-250 solution, all of the components should be added and dissolved prior to the addition of the methanol, which should be added last, slowly to prevent clumping of the solution. 10. Incubate the gel in staining solution overnight to achieve the best results although signals can generally be seen as little as 1 h after the incubation has started.

396

Gavin McGauran et al.

11. Aβ42 is an amphipathic molecule that is highly hydrophobic in its central region and C-terminus and charged at the N-terminus, physicochemical features that drive and modulate its aggregation [16, 17] and surface adsorption. Therefore, to reduce loss on tube walls, Aβ42 should be made up in low-binding microcentrifuge tubes. We recommend the use of recombinant peptide to avoid sequence inhomogeneity. References 1. Phizicky E, Bastiaens PI, Zhu H, Snyder M, Fields S (2003) Protein analysis on a proteomic scale. Nature 422(6928):208 2. Dell’Orco D, Xue W-F, Thulin E, Linse S (2005) Electrostatic contributions to the kinetics and thermodynamics of protein assembly. Biophys J 88(3):1991–2002 3. Mhurchu´ NN, Zoubak L, McGauran G, Linse S, Yeliseev A, O’Connell DJ (2018) Simplifying G protein-coupled receptor isolation with a calcium-dependent fragment complementation affinity system. Biochemistry 57 (30):4383–4390 4. Dunning C, McGauran G, Willen K, Gouras G, O’Connell D, Linse S (2016) Direct high affinity interaction between Abeta42 and GSK3alpha stimulates hyperphosphorylation of tau. A new molecular link in Alzheimer’s disease ACS Chem. Neuro 7(2):161–170 5. Ali A, Hoeflich KP, Woodgett JR (2001) Glycogen synthase kinase-3: properties, functions, and regulation. Chem Rev 01(8):2527–2540 6. Cohen P, Frame S (2001) The renaissance of GSK3. Nat Rev Mol Cell Biol 2(10):769 7. Woodgett JR (1990) Molecular cloning and expression of glycogen synthase kinase-3/ factor A. EMBO J 9(8):2431–2438 8. Mukai F, Ishiguro K, Sano Y, Fujita SC (2002) Alternative splicing isoform of tau protein kinase I/glycogen synthase kinase 3β. J Neurochem 81(5):1073–1083 9. Frame S, Cohen P (2001) GSK3 takes centre stage more than 20 years after its discovery. Biochem J 359(1):1–16

10. Maqbool M, Mobashir M, Hoda N (2016) Pivotal role of glycogen synthase kinase-3: a therapeutic target for Alzheimer’s disease. Eur J Med Chem 107:63–81 11. Turenne GA, Price BD (2001) Glycogen synthase kinase3 beta phosphorylates serine 33 of p53 and activates p53’s transcriptional activity. BMC Cell Biol 2(1):12 12. Ma T (2014) GSK3 in Alzheimer’s disease: mind the isoforms. J Alzheimer’s Dis 39 (4):707–710 13. Beurel E, Grieco SF, Jope RS (2015) Glycogen synthase kinase-3 (GSK3): regulation, actions, and diseases. Pharmacol Therapeut 148:114–131 14. Hurtado DE, Molina-Porcel L, Carroll JC, MacDonald C, Aboagye AK, Trojanowski JQ et al (2012) Selectively silencing GSK-3 isoforms reduces plaques and tangles in mouse models of Alzheimer’s disease. J Neurosci 32 (21):7392–7402 15. Phiel CJ, Wilson CA, Lee VM-Y, Klein PS (2003) GSK-3α regulates production of Alzheimer’s disease amyloid-β peptides. Nature 423(6938):435 16. de Groot NS, Aviles FX, Vendrell J, Ventura S (2006) Mutagenesis of the central hydrophobic cluster in Aβ42 Alzheimer’s peptide: sidechain properties correlate with aggregation propensities. FEBS J 273(3):658–668 17. Yang X, Meisl G, Frohm B, Thulin E, Knowles TPJ, Linse S (2018) On the role of sidechain size and charge in the aggregation of Aβ42 with familial mutations. Proc Natl Acad Sci U S A 115(26):E5849–E5858

INDEX A

Cryopreservation......................................... 18, 19, 21, 23 Culture monitoring....................................................... 319

Alternating tangential flow filtration (ATF) ...............109, 114, 117, 119, 121, 122, 126, 128, 130, 151, 152, 155, 157–160, 162, 163, 166 Ambr® ....................................................... 15, 43–66, 127 Animal cells............................................................ 4, 7, 19, 105–122, 237, 319–332 Assay ready cells ...........................................18, 19, 21–23 Automated bioreactor........................... 43, 156–157, 182

Design of Experiments (DoE) ................................27–39, 43, 45, 46, 49, 50, 56, 57, 61, 62, 228, 235–248 Dielectric spectroscopy ........................................ 335–361 Downstream processing....................................... 185, 368 Drosophila melanogaster S2 cells .................................. 340

B

E

Batch ...............................................................6, 7, 13, 21, 28,51, 71, 73, 76, 85, 105–122, 126, 142–145, 161, 164, 166, 174, 181, 184, 190, 200, 208, 218, 219, 227, 236–239, 243, 255, 257, 265, 273, 286, 330, 331, 341, 351, 352, 371, 373 Bead-based assay .................................................. 271–283 Benchtop bioreactors ........................................... 125–138 Bioassays ....................................................................17–25 Biophysical..................................................................... 386 Bioprocess development ....................... 83–101, 271–273 Bioprocess optimization ........................................ 86, 271 Biosimilar .............................................................. 107, 108

Enzyme based optical biosensor ......................... 319–332 Experimental space.........................................30, 236, 245

C Calcium................................................................. 385–396 Cell culture model ...................................... 231, 262, 263 Cell cultures.................................................. 4, 18, 27, 43, 70, 84, 108, 126, 142, 177, 190, 215, 251, 272, 296, 298, 321, 336, 373, 385 Cell line development ............................... 46, 69–80, 107 Cell size.............................................................13, 80, 312 Centrifugation .............................................. 8, 10, 20, 21, 23, 25, 33–35, 55, 56, 127, 142, 154 Chinese Hamster Ovary (CHO) cells.......................8, 10, 14, 43–66, 71, 83, 107–108, 110, 113, 114, 122, 174–178, 184, 200, 272, 276, 295–302, 330 CHO cell expansion...................................................... 177 Continuous chromatography .............................. 367–384 Continuous process ...................................................... 237 Controlled rate freezing ................................................. 22 Control strategies..........................................79, 142–146, 152–153, 159, 163, 165 Covariance and correlation......................... 198, 199, 206

D

F Fed-batch...........................................................28, 44, 71, 76, 85, 105–122, 126, 174, 181, 184, 190, 200, 208, 236–239, 243, 255, 265, 273, 286, 330, 341 Fragment complementation ................................ 385–396

G Glucose and lactate co-consumption ................ 84, 96, 98 Glucose monitoring ............................................. 319–332 Glycogen synthase kinase .................................... 385–396 Glycosylation ...................................................83–85, 107, 108, 117, 136–138, 184, 271–283, 285, 286, 296, 297, 300 Good cell culture practice............................................... 18 GS-CHO cells ................................................................. 78

H HEK293 cells ........................................ 83–101, 286, 387 High cell density cultivation................................ 141–166 High-throughput assays ....................................... 28, 142, 161, 271–283, 286 Human embryonic kidney (HEK) cells ....................6, 15, 83–101, 385–396

I Identifiability analysis .......................................... 196, 198, 203, 204, 232, 261 IgG production ................................................... 174–175, 177–179, 181–183

Ralf Po¨rtner (ed.), Animal Cell Biotechnology: Methods and Protocols, Methods in Molecular Biology, vol. 2095, https://doi.org/10.1007/978-1-0716-0191-4, © Springer Science+Business Media, LLC, part of Springer Nature 2020

397

ANIMAL CELL BIOTECHNOLOGY: METHODS

398 Index

AND

PROTOCOLS

Immunoglobulin M (IgM).................................. 296–302 Inoculum train .............................................................. 262

K Kinetic model .............................................. 190, 218, 257

L Label-free analysis ......................................................... 304 Lectins......................................................... 273, 275–278, 280–282, 286, 287, 290–292

M Measles virus............................................... 341, 345–346, 353, 356, 358, 359, 361 Mechanistic model ................................................ 14, 189, 190, 192, 200, 202, 257 Metabolic phases .................................................... 84, 175 Micro-Matrix .............................................................69–80 Model-assisted ...................................................... 235–248 Model implementation ........................................ 218–220 Modeling ............................................................. 6, 14, 29, 36, 43, 56, 189–191, 195, 196, 200, 220, 231, 237, 238, 257, 261–263 Monoclonal antibody............................................ 78, 115, 200, 240, 241, 259, 272, 369

N Non-invasive analysis ........................................... 303–316 Nucleic acid contamination .......................................... 300

O Online process monitoring.................. 83–101, 136, 138, 320, 336, 341 Online sensors ............................................................... 157 Optical density (OD) ........................................... 336, 337 Orbitally shaken bioreactor (OSB) .............105–122, 130 Oxygen transfer rate (OTR)................................... 85, 86, 92, 97, 98, 101, 107, 171

P Parameter error .................................................... 198, 205 Parameter estimation .......................................... 194, 196, 198, 200, 207, 215, 217, 218, 220, 222–224, 226, 228, 230, 232, 253, 255, 263 Perfusion............................................................ 27–39, 85, 105–122, 125–138, 141–166, 190, 341 bioreactor....................................................... 127, 131, 138, 142, 143, 148, 151, 161 culture .......................................... 27–39, 44, 125–138 process development ........................................ 28, 127

rate ........................................................ 116, 126, 129, 130, 134, 136, 142–145, 152–161, 163–166 Polymer distribution ............................................ 297–299 Population heterogeneities ............................................... 4 Principle components analysis (PCA) .........................306, 307, 309, 313 Process analytical technology (PAT) ............................. 84, 85, 335, 336, 340 Process development............................................... 28, 45, 50, 57, 69–80, 83–101, 127, 191, 215, 236, 271–273, 285, 286, 336 Process monitoring .............................................. 177, 320 Product quality..................................................46, 55, 56, 107, 108, 117, 125, 127, 136, 137, 144, 248, 285, 286, 295–302 Product quality assessment ........................................... 286 Product sialylation................................................ 285–292 Programming ....................................................... 218, 262 Pseudo-perfusion .............. 142, 143, 148, 153, 161, 163 Purification .......................................................... 152, 273, 287, 296, 301, 367–396

Q Quality attributes ................ 28, 138, 272, 286, 296, 320

R Raman spectrum ................................................. 303–306, 308–310, 312–314, 316, 337 Raman trapping microscopy (RTM)................... 303–316 Recombinant glycoprotein .................272, 285, 286, 292 Respiration Activity Monitoring System (RAMOS) ................... 85–87, 90–93, 95–98, 101 Response surface methodologies (RSM)...................... 28, 31, 35, 38

S Scale-down model ..........................................43, 107, 153 Scale-up parameters ............................................. 107, 109 Seed train ............................................. 251–267, 353–354 Sensitivity.................................................... 195–198, 203, 205, 206, 210, 228, 233, 272, 279, 301, 321, 327, 331 Separation ...................................................................6, 28, 127, 338, 368, 369, 378, 383 Shake flask reader (SFR) ................................83–101, 322 Shake tubes (ST) ................................................. 127–129, 131–135, 138 Simulation .................................................. 194, 215, 220, 223, 226–228, 230, 237, 238, 247, 253, 254, 261–264, 266 Single step purification ........................................ 385–396

ANIMAL CELL BIOTECHNOLOGY: METHODS

AND

PROTOCOLS Index 399

V

Single-use bag ..................................................... 105, 106, 109, 110, 112, 113 Skin cells products......................................................... 304 Small-scale bioreactor ............................................. 28, 69, 70, 106, 127, 161 Stirred single-use bioreactor................................ 169–185 Surface plasmon resonance (SPR)...................... 285–292, 386, 388, 389, 393, 394

Vero cells.............................................................. 147, 150, 341, 344, 345, 353, 356, 360, 361 Viral vaccine..................................................108–109, 141 Viral vaccines production .................................... 108–109 Virus-like particles................................................ 373, 374

T

Wave-mixed single-use bioreactor.............. 170, 174, 184

Titer assay ...................................................................... 274

W