Investigation of the metabolic and energetic states of antibody producing CHO cells for process intensification in fed-batch and perfusion cultivations Dissertation Von der Fakultät Energie-, Verfahrens- und Biotechnik der Universität Stuttgart zur Erlangung der Würde eines Doktor-Ingenieurs (Dr.-Ing.) genehmigte Abhandlung Vorgelegt von: Max Becker aus Münster Hauptberichter: Prof. Dr.-Ing. Ralf Takors Mitberichter: Prof. Dr. Roland Kontermann Tag der mündlichen Prüfung: 05.12.2019 Institut für Bioverfahrenstechnik 2019 Für Maike Danksagung Zuerst möchte ich Herrn Prof. Dr.-Ing. Ralf Takors für die Möglichkeit danken, meine Promotion mit dieser spannenden Aufgabenstellung am IBVT durchzuführen. Darüber hinaus danke ich ihm für die Diskussionen über offene Fragestellungen und Ergebnisse, aber auch die Freiheiten bei der Durchführung der Arbeit. Meinen Dank möchte ich auch Herrn Prof. Dr. Roland Kontermann aussprechen für die Übernahme der Mitberichterschaft. Meinen Ansprechpartnern bei Boehringer Ingelheim Jan Bechmann und Raphael Vo- ges danke ich für die unkomplizierte Zusammenarbeit und die Möglichkeit, mich jed- erzeit mit neuem Medienbedarf melden zu können. Bei allen Festangestellten und technischen Mitarbeitern am Institut, die mit Ihrer Un- terstützung die täglichen Arbeiten erst möglich gemacht haben und mit ihrer positiven Art immer für einen Plausch zu haben waren, bedanke ich mich: Frau Reu im Sekre- tariat sowie Alex, Andreas und Salah in technischen Fragen (inklusive guter Sprüche). Darüber hinaus Mira in Sachen HPLC Analytik und Martin für die vielen Anträge an die Verwaltung. Herzlicher Dank geht an die vielen Kollegen, die zu Freunden geworden sind und die mir die Zeit am Institut über all die Jahre verschönert haben: Für die schönen Zeiten in unserem abgelegenen Zellkulturlabor, gegenseitige Unter- stützung in Sachen Labor und Promotion, sowie Gespräche über jeden Unsinn und das Essen von reichlich Kuchen danke ich all meinen Bürokollegen, aber vor allem Lisa, Vikas, Natascha und Jennifer. Bei Jurek, Andrés, Annette, Salah, Kraml und Attila bedanke ich mich für die Feier- abendbierchen, gemeinsamen Essen oder auch einfach Pausen am Institut, die für die nötige Ablenkung gesorgt haben. I Schwandy, Robert und Sebastian danke ich für die lustigen Abende in der Sportsbar, die Fußballrunden mit dem Institut und Ihren bemühten Einsatz mich beim Manager- spiel zu schlagen. Meinen Studenten Lisa Stepper, Birgit Salzmann und Felix Oster danke ich für ihre erfolgreichen Arbeiten rund um meine Promotion sowie die angenehme Zusammenar- beit. Zuletzt möchte ich meinen Eltern und meinen Brüdern mitsamt ihren Familien für den Zuspruch und die Ablenkung auch in schwierigen Phasen danken. Mein ganz besonderer Dank gilt Maike für Ihre Unterstützung und Motivation, aber auch Ihre Gesellschaft in den vielen gemeinsamen Stunden des Schreibens, die mir vieles erleichtert haben. II Contents Declaration of Originality VII Nomenclature IX List of Figures XV List of Tables XIX Zusammenfassung 1 Abstract 5 1 Introduction and Motivation 7 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2 Theoretical Background 11 2.1 Chinese hamster ovary cells . . . . . . . . . . . . . . . . . . . . . . . . 11 2.2 Metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.2.1 Main carbon metabolism . . . . . . . . . . . . . . . . . . . . . 13 2.2.2 Energy generation and redox balance . . . . . . . . . . . . . . . 15 2.3 Flux analysis in CHO cells . . . . . . . . . . . . . . . . . . . . . . . . 18 2.4 Fed-batch processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.4.1 Process overview and performance . . . . . . . . . . . . . . . . 23 2.4.2 Parameters influencing process performance . . . . . . . . . . . 24 2.5 Perfusion processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3 Materials and Methods 31 3.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.1.1 Chemicals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.1.2 Consumables . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.1.3 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.1.4 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.1.5 Antibodies for ELISA . . . . . . . . . . . . . . . . . . . . . . . 36 3.1.6 Buffers and Solutions . . . . . . . . . . . . . . . . . . . . . . . 36 3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.2.1 Cryoconservation . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.2.2 Seed Train . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.2.3 Bioreactor Cultivation . . . . . . . . . . . . . . . . . . . . . . 40 3.2.4 Analytics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.2.5 Cell specific Rates . . . . . . . . . . . . . . . . . . . . . . . . 50 3.2.6 Flux Balance Analysis . . . . . . . . . . . . . . . . . . . . . . 51 4 Results 53 4.1 Fed-Batch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 4.1.1 Cellular phenotype and product formation under the influence of pH and pCO2 . . . . . . . . . . . . . . . . . . . . . . . . . . 53 4.1.2 Intracellular flux distributions and balancing of ATP and carbon 61 4.1.3 Intracellular pool sizes of glycolysis, TCA and nucleotides in fed-batch phases . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.2 Perfusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.2.1 Growth and metabolic state in reference and glucose-limited steady-states . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.2.2 Flux balance analysis and cellular energetic and redox state . . . 79 4.2.3 Intracellular pool sizes of glycolysis, TCA and adenylate nu- cleotides in perfusion steady-states . . . . . . . . . . . . . . . . 85 5 Discussion 89 5.1 Fed-Batch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 5.1.1 Early elevated CO2 and late decreased pH influence growth and induce changes in production kinetics . . . . . . . . . . . . . . 90 5.1.2 Adjusmtens of intracellular fluxes and energetic state . . . . . . 92 5.1.3 Metabolic adaptations of intracellular pool sizes in different fed- batch phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 5.2 Perfusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 5.2.1 Impact of glucose availability on cell density and main metabolism 97 5.2.2 Adaptations in ATP formation and cellular redox state and effect on productivity . . . . . . . . . . . . . . . . . . . . . . . . . . 100 5.2.3 Changes in perfusion mode affect intracellular pool sizes of gly- colysis, TCA and adenylate nucleotides . . . . . . . . . . . . . 103 5.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 References 109 Appendix A Manuscript I 123 Appendix B Manuscript II 137 Appendix C List of Publications and Author Contribution 149 Declaration of Originality Declaration of Originality I declare that the submitted work has been completed by me and that I have not used any other than permitted reference sources or materials. All references and other sources used by me have been appropriately acknowledged in the work. Hiermit erkläre ich, dass ich die vorliegende Arbeit selbstständig angefertigt habe. Es wurden von mir nur die in der Arbeit ausdrücklich benannten Quellen und Hilfsmit- tel benutzt. Übernommenes Gedankengut wurde von mir als solches kenntlich gemacht. Stuttgart, den 10.03.2020 Ort, Datum Max Becker VII Nomenclature Nomenclature 2PG 2-Phosphoglyceric acid 3PG 3-Phosphoglyceric acid µ Growth rate AA Amino acids Acetyl−CoA Acetyl coenzyme A ADP Adenosine diphosphate Ala Alanine AMP Adenosine monophosphate AMPK AMP-activated protein kinase Asn Asparagine Asp Aspartate AT P Adenosine triphosphate B Bleed rate BHK Baby hamster kidney c Concentration of compound cT Transposed vector of weights CHO Chinese hamster ovary IX Nomenclature cisAco Cis-aconitate Cit Citrate CO2 Carbon dioxide COBRA Constraint-Based Reconstruction and Analysis COP CO2 stressed fed-batch process DHAP Dihydroxyacetone phosphate DHFR Dihydrofolate reductase dMFA Dynamic metabolic flux analysis DNA Deoxyribonucleic acid EC Energy charge ELISA Enzyme-linked immunosorbent assay F6P Fructose 6-phosphate f CO2 Fraction of CO2 FADH2 Flavin adenine dinucleotide FBA Flux balance analysis Fum Fumarate G6P Glucose 6-phosphate GAP Glyceraldehyde 3-phosphate Glc Glucose Gln Glutamine Glu Glutamate Gly Glycine X Nomenclature GRP78 Binding immunoglobulin protein GS Glutamine synthethase GT P Guanosine triphosphate HEK Human embryonic kidney His Histidine HPLC High-performance liquid chromatography IgG Immunoglobulin G Ile Isoleucine Iso Isocitric acid KOH Potassium hydroxide Lac Lactate LC−MS Liquid chromatography–mass spectrometry Leu Leucine Lys Lysine M Molarity in mol/L mAb Monoclonal antibody Mal Malate Met Methionine MFA Metabolic flux analysis Na2CO3 Sodium carbonate NAD Nicotinamide adenine dinucleotide (oxidized) NADH Nicotinamide adenine dinucleotide XI Nomenclature NADPH Nicotinamide adenine dinucleotide phosphate NOB Fed-batch process with no base titration NS0 Mouse myeloma OPA Ortho-phtaldialdehyde Oxal Oxaloacetate P Perfusion rate P/O Ratio Phosphate/Oxygen Ratio pCO2 Partial pressure of carbon dioxide PEP Phosphoenolpyruvic acid Phe Phenylalanine pKS Acidic dissociation constant PPP Pentose phosphate pathway Pyr Pyruvate Q Volumetric production rate q Cell specific production or consumption rate R Redox variable R5P Ribose 5-phosphate REF Reference fed-batch process RNA Ribonucleic acid rpm Revolutions per minute S Stoichiometric matrix Ser Serine XII Nomenclature SS Steady-state Succ Succinate Succinyl−CoA Succinyl coenzyme A TCA Tricarboxylic acid cycle T FF Tangential flow filtration T hr Threonine Trp Tryptophane V Liquid bioreactor volume v Flux vector Val Valine VCD Viable cell density XV Viable cell density Z Objective function XIII List of Figures List of Figures 1.1 Global annual sales for biopharmaceuticals . . . . . . . . . . . . . . . . 8 2.1 Percentage of biopharmaceuticals produced in mammalian systems . . . 12 2.2 Simplified scheme of the main carbon metabolism . . . . . . . . . . . . 17 2.3 Workflow of flux balance analysis . . . . . . . . . . . . . . . . . . . . 21 2.4 Schematic representation of the three main cultivation modes batch, fed- batch and perfusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.5 Different published cell retention devices used for perfusion processes . 29 3.1 Simplified scheme of the perfusion cultivation . . . . . . . . . . . . . . 44 4.1 Profiles of pH, pCO2 and osmolality over the process time for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . . . . 55 4.2 Profiles of viable cell density and growth rate over the process time for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . 56 4.3 Profiles of antibody concentration and cell specific antibody productiv- ity over the process time for the three settings REF, COP and NOB . . . 57 4.4 Profiles of glucose concentration and cell specific glucose uptake rate over the process time for the three settings REF, COP and NOB . . . . . 58 4.5 Profiles of lactate concentration and cell specific lactate production rate over the process time for the three settings REF, COP and NOB . . . . . 59 4.6 Cell specific antibody productivity as a function of growth rate for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . 60 4.7 Profile of the volumetric productivity of the antibody over the process time for the three settings REF, COP and NOB . . . . . . . . . . . . . . 61 XV List of Figures 4.8 Simplified main carbon metabolism representing the results of the flux balance analysis during the growth phase for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 4.9 Simplified main carbon metabolism representing the results of the flux balance analysis during the early stationary phase for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4.10 Simplified main carbon metabolism representing the results of the flux balance analysis during the early decline phase for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.11 Cell specific ATP production rates during growth phase, early stationary phase and early decline phase for the three settings REF, COP and NOB 66 4.12 Cell specific lactate production rate combined for all three settings REF, COP and NOB as function of the redox variable R . . . . . . . . . . . . 67 4.13 Cell specific antibody productivity combined for all three settings REF, COP and NOB as function of the redox variable R . . . . . . . . . . . . 68 4.14 Carbon balances showing the main fractions for incoming and outgoing carbon during the growth phase for the three settings REF, COP and NOB 69 4.15 Carbon balances showing the main fractions for incoming and outgoing carbon during the early stationary phase for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.16 Carbon balances showing the main fractions for incoming and outgoing carbon during the early decline phase for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.17 Intracellular pool sizes of intermediates of glycolysis for the three set- tings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . . . . . . 72 4.18 Intracellular pool sizes of intermediates of TCA for the three settings REF, COP and NOB . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 4.19 Intracellular pool sizes of nucleotides and the corresponding adenylate energy charge for the three settings REF, COP and NOB . . . . . . . . 74 4.20 Profile of viable cell density over the process time for the three perfusion steady-states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 4.21 Main cell specific rates for the three perfusion steady-states . . . . . . . 79 XVI List of Figures 4.22 Simplified main carbon metabolism representing the results of the flux balance analysis for the three perfusion steady-states . . . . . . . . . . 80 4.23 Cell specific ATP production rates for the three perfusion steady-states . 81 4.24 Cell specific antibody productivity and volumetric productivity as a func- tion of ATP formation due to oxidative phosphorylation for the three perfusion steady-states . . . . . . . . . . . . . . . . . . . . . . . . . . 82 4.25 Carbon balances showing the main fractions for three perfusion steady- states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 4.26 Cell specific lactate production rate for the three perfusion steady states as function of the redox variable R . . . . . . . . . . . . . . . . . . . . 84 4.27 Cell specific antibody production rate for the three perfusion steady states as function of the redox variable R . . . . . . . . . . . . . . . . . 85 4.28 Intracellular pool sizes of intermediates of glycolysis for the three per- fusion steady-states . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 4.29 Intracellular pool sizes of intermediates of TCA for the three perfusion steady-states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 4.30 Intracellular pool sizes of nucleotides and the corresponding adenylate energy charge for the three perfusion steady-states . . . . . . . . . . . . 88 XVII List of Tables List of Tables 3.1 Chemicals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.2 Consumables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.3 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.4 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.5 Antibodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.6 Buffers and solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.7 Seed train steps and volumes . . . . . . . . . . . . . . . . . . . . . . . 40 3.8 Parameter setpoints Fed-Batch . . . . . . . . . . . . . . . . . . . . . . 41 3.9 Controller settings Fed-Batch . . . . . . . . . . . . . . . . . . . . . . . 42 3.10 Parameter setpoints Perfusion . . . . . . . . . . . . . . . . . . . . . . . 43 3.11 Perfusion settings and flowrates . . . . . . . . . . . . . . . . . . . . . . 45 4.1 Viable cell density, bioreactor antibody and metabolite concentrations during the three perfusion steady states . . . . . . . . . . . . . . . . . . 77 4.2 pH and partial pressure of CO2 during perfusion . . . . . . . . . . . . . 78 XIX Zusammenfassung Zusammenfassung Die heutige biopharmazeutische Industrie verlässt sich hauptsächlich auf großskalige Fed-Batch Prozesse mit tierischen Zellen, um möglichst hohe Produkttiter zu erreichen. Jedoch führen, aufgrund von höheren Mischzeiten in Bioreaktoren im Produktions- maßstab, Inhomogenitäten zu Zonen vermehrten Stresses und damit schlussendlich zu verminderten volumetrischen Produktivitäten. Leistungsdaten der Zellen wie Wachs- tum, Produktivität und Nebenproduktbildung werden dabei durch verschiedene Stress- faktoren beeinflusst. Unter anderem bilden der Partialdruck von CO2 (pCO2) und der pH-Wert solche Inhomogenitäten, die die Leistung von Produktionsprozessen beein- flussen. Die zugrundeliegenden metabolischen Anpassungen, die dabei letztendlich zu einem Verlust der Produktivität führen, sind bisher noch nicht komplett verstanden. Mögliche Alternativen für übliche Fed-Batch Prozesse können Perfusionsprozesse sein. Als kontinuierliche Prozesse mit Zellrückhaltung können sie Zelldichten erreichen, die ein Vielfaches höher liegen als in Fed-Batch Kultivierungen. Darüber hinaus müssen aber auch zellspezifische Produktivitäten maximiert werden, um bessere volumetrische Produktivitäten zu liefern. Nur dann kann Perfusion intensivierte Prozesse hervorbrin- gen, welche ökonomische Effizienz in deutlich kleinerem Maßstab gewährleisten. Diese Arbeit soll die metabolischen Anpassungen von CHO Zellen in sowohl Fed- Batch Kultivierungen als auch Perfusionsprozessen unter variierenden Prozessbedin- gungen aufklären. Die Energieversorgung in Form von ATP wurde dabei als Bindeglied zwischen Metabolismus und zellspezifischer Produktivität genutzt. Für die Fed-Batch Kultivierungen wurde der Einfluss von pCO2 und pH-Veränderungen verdeutlicht. Früher CO2 Stress mit Partialdrücken bis zu 200 mbar führte zu einem zeitweisen Anstieg in der zellspezifischen Produktivität bis hin zu 23 pg/Zelle/Tag. Jedoch zeigten die Zellen kurz darauf nur noch eine verringerte intrazelluläre ATP Poolgröße von 2,5 fmol/Zelle. Gleichzeitig stieg die zellspezifische Laktatbildung deutlich (um 80 %) an, während die Produktivität schnell abfiel, obwohl pCO2 wieder im Bereich der Referenzwerte 1 Zusammenfassung lag. In einem zweiten Prozess resultierte eine pH Absenkung in Richtung 6,70 in kon- tinuierlich ansteigenden zellspezifischen Produktivitäten bis hin zu einem Maximalw- ert von 20 pg/Zelle/Tag. Zudem wurde keine Laktatbildung beobachtet, da sämtlicher Kohlenstoff aus der Glukose in den TCA geleitet wurde. Interessanterweise waren die damit zusammenhängenden intrazellulären ATP Konzentrationen auch in der sta- tionären Phase konstant bei 4 - 5 fmol/Zelle und damit erhöht im Vergleich zu sowohl dem CO2 gestressten Prozess (2,5 fmol/Zelle), als auch dem Referenzprozess (weniger als 1 fmol/Zelle). Ebenso zeigte nur der Prozess mit sinkendem pH-Wert über den ganzen Prozess stabile ATP Bildungsraten von 25.000 fmol/Zelle/Tag wie die Flussanal- ysen zeigten. Für den Perfusionsprozess wurde Glukoselimitierung als Methode zur Verbesserung der metabolischen Effizienz und der damit zusammenhängenden zellspezifischen Pro- duktivitäten verwandt. Insgesamt wurden in der Perfusion drei verschiedene steady- states verglichen, wobei einer ohne offensichtliche Limitierung zwei glukoselimitierten steady-states gegenüberstand. Demzufolge führte die Glukoselimitierung zu einem vorteil- haften metabolischen Zustand, in dem die Zellen verringerte Glukoseaufnahme (30 %) und Laktatbildung (75 %) zeigten. Darüber hinaus waren die zellspezifischen Produk- tivitäten über mehrere Tage konstant bei Werten von 15 pg/Zelle/Tag und damit um 50 % erhöht gegenüber dem nicht-limitierten steady-state. Wie in den Fed-Batch Kul- tivierungen beobachtet, war die zellspezifische Produktivität mit der ATP Versorgung verknüpft. Darüber hinaus konnten aufgrund der verschiedenen steady-state Bedingun- gen noch genauere Erkenntnisse gewonnen werden: Aufgeschlüsselt auf den Ursprung des ATP zeigte die respiratorische ATP Versorgung starke Korrelationen mit der An- tikörperproduktivität. Während der glukoselimitierten steady-states wurde der meiste Kohlenstoff aus der Glukose in den TCA geleitet. Folglich wurde einhergehend mit der niedrigen Laktatbildung mehr NADH in die Mitochondrien für die oxidative Phospho- rylierung transportiert. Im Vergleich mit dem steady-state mit Glukose im Überschuss, war die ATP Versorgung aus der oxidativen Phosphorylierung um fast ein Drittel erhöht. Schlussfolgernd zeigen die Resultate aus Fed-Batch- und Perfusionsprozessen den deutlichen Einfluss der Prozessbedingungen auf den zellulären Metabolismus und die damit zusammenhängende Energieverfügbarkeit. Darüber hinaus ist die resultierende spezifische Produktivität stark abhängig von der ATP Bereitstellung. Wie diese Arbeit zeigt, ist die ATP Versorgung anfällig für Veränderungen in pCO2 und pH-Wert, so- 2 Zusammenfassung dass bereits Zellen, welche nur kurzzeitig Inhomogenitäten nach dem Scale-up erleben, die Prozessleistung beeinflussen können. Auf der anderen Seite kann die Glukoselimi- tierung für die Prozessintensivierung im Perfusionsbetrieb genutzt werden, da sie die Ef- fizienz Kohlenstoff betreffend und damit auch die ATP Bereitstellung aus der oxidativen Phosphorylierung erhöht. Dadurch wird die zellspezifische Produktivität gesteigert, so- dass infolgedessen Perfusionsprozesse mit bereits hohen Zelldichten verbesserte vol- umetrische Produktivitäten erreichen. 3 Abstract Abstract Today’s biopharmaceutical industry mostly relies on large-scale fed-batch processes with mammalian cells to reach highest possible product titers. However, due to higher mixing times in production scale bioreactors, inhomogeneities lead to zones of increased stress inside the bioreactor and consequently to diminished volumetric productivities. Especially cellular performance parameters like growth, productivity and by-product formation are influenced by various stress factors. Among others, the partial pressure of CO2 (pCO2) and pH are known to form such inhomogeneities influencing the perfor- mance of production processes. The underlying metabolic adaptations finally leading to the loss in productivity are not yet fully understood. Possible alternatives for common fed-batch processes can be perfusion processes. As continuous processes using cell re- tention they can reach cell densities multiple times higher than in fed-batch cultivations. But additionally, cell specific productivities have to be maximized to yield superior vol- umetric productivities. Only then, perfusion can provide intensified processes which guarantee economic efficiency in much smaller scale. This thesis aims at unraveling metabolic adaptations made by CHO cells in both fed- batch cultivations and perfusion processes under varying process conditions. The energy supply in form of ATP was used as link from metabolism to cell specific antibody pro- ductivities. For the fed-batch cultivations the impact of pCO2 and pH shifts were eluci- dated, typical for cells experiencing large scale inhomogeneities. Early CO2 stress with partial pressures of up to 200 mbar led to a temporary increase in cell specific producitiv- ity towards 23 pg/cell/day. However, shortly after cells showed diminishing intracellular ATP pool sizes towards 2.5 fmol/cell. Simultaneously, cell specific lactate production increased severely (80 %), while the productivity fell quickly although pCO2 was back to reference values. On the other side, a pH down-shift towards pH 6.70 resulted in continuously rising cell specific productivities to peak values of 20 pg/cell/day. Addi- tionally, no lactate formation was observed since all carbon from glucose was fueling 5 Abstract the TCA cycle. Interestingly, related intracellular ATP concentrations were constant at 4-5 fmol/cell even in the stationary phase and therefore elevated compared to both CO2 stressed (2.5 fmol/cell) and reference conditions (below 1 fmol/cell). Likewise, only the pH shifted process showed stable ATP formation rates of about 25,000 fmol/cell/day throughout the process as flux analysis revealed. For the perfusion glucose limitation was applied as method to improve metabolic effi- ciency and the interrelated cell specific productivities. In total, three different perfusion steady states were compared, where one steady state without any apparent limitation contrasted two glucose-limited steady states. As a result, glucose limitation led to a beneficial metabolic state where cells showed decreased glucose uptake (30 %) and lac- tate formation (75 %). Furthermore, cell specific productivities were constant at about 15 pg/cell/day for several days. Therefore, the values were increased by 50 % com- pared to the non-limited steady-state. Like observed in the fed-batch cultivations, the cell specific productivity was connected to the ATP supply. Moreover, due to the differ- ent steady state conditions even more precise findings could be made: Broken down to the origin of ATP, respiratory ATP supply showed strong correlation to the antibody pro- ductivity. During the glucose limited steady states most of the carbon coming from glu- cose was fueled into TCA. Consequently, along with the lower lactate formation, more NADH was transported into mitochondria for oxidative phosphorylation. In comparison to the steady state with excess glucose ATP supply from oxidative phosphorylation was increased by almost one third. In conclusion, the results from both fed-batch and perfusion processes show the strong influence of process conditions on the cellular metabolism and the interrelated energy availability. Furthermore, the resulting specific productivity is strongly depen- dent on ATP supply. As this work shows, ATP supply is prone to variations in pCO2 and pH, so that cells experiencing only short term inhomogeneities after scale-up can alter the overall process performance. On the other side, glucose limitation can be used for process intensification in perfusion mode as it significantly increases the carbon ef- ficiency and therefore the ATP supply from oxidative phosphorylation. Hereby, the cell specific productivity is boosted and consequently perfusion processes, already having high cell densities, can reach enhanced volumetric productivities. 6 1 Introduction and Motivation 1.1 Introduction The global market for biopharmaceutical products is continually growing. Concurrently increasing annual sales provide a major reason for the development of optimized pro- duction processes (Gaughan, 2016; Walsh, 2014; Morrison & Lähteenmäki, 2017). From 2010 to 2016 the global annual sales nearly doubled from 107 to 202 Billion US-$, while the sales for the top 10 biopharmaceuticals alone increased by about a third to 76 Billion US-$ (figure 1.1, Walsh (2014); Morrison and Lähteenmäki (2017)). The biopharmaceutical products comprise many categories like antibodies, blood fac- tors, hormones, fusion proteins, vaccines and many more (Aggarwal, 2014). However, monoclonal antibodies produced by Chinese hamster ovary (CHO) cells have a major share of the top-selling drugs in the past decade (Ecker, Jones, & Levine, 2015). In the past years, an increasing number of expiring patent protections and the resulting emer- gence of biosimilar products have put pressure on current first-generation production processes (Mullard, 2012; Gaughan, 2016). As typical large-scale production processes for biopharmaceuticals, fed-batch culti- vations remain in the focus of process optimization. Accordingly, maximum cell den- sities and product titers in current fed-batch processes have increased substantially up to 30x106 cells/mL and more than 10 g/L (Wurm, 2004; Kunert & Reinhart, 2016; Zboray et al., 2015). After successful cell line engineering and media optimization, numerous process parameters can still influence the volumetric productivity (Birch & Racher, 2006; Shukla & Thömmes, 2010; Seth, Hossler, Yee, & Hu, 2006). Especially during large-scale cultivations several parameters can cause inhomogeneities inside the bioreactor, eventually leading to deterioration in overall performance concering cellular growth and productivity. Identification of these parameters and their impact on process performance is therefore crucial for the success of large-scale production processes. 7 Introduction and Motivation Alternatively to large-scale fed-batch processes, an economic production of biophar- maceuticals can also be achieved by the intensification of production processes, aiming at higher volumetric productivities in smaller production scales. By applying cell reten- tion in continuous mode, perfusion processes can provide long-term operation at high cell densities (Clincke, Molleryd, Zhang, et al., 2013; Clincke, Molleryd, Samani, et al., 2013; Voisard, Meuwly, Ruffieux, Baer, & Kadouri, 2003; Bielser, Wolf, Souquet, Broly, & Morbidelli, 2018). In comparison to fed-batch processes, multiple times higher viable cell densities of more than 100x106 cells/mL are possible (Clincke, Molleryd, Zhang, et al., 2013; Clincke, Molleryd, Samani, et al., 2013; Warikoo et al., 2012). Therefore, elevated volumetric productivities make production processes in medium- scale economically feasible (Croughan, Konstantinov, & Cooney, 2015). However, cell specific productivities need to remain high throughout the long-term perfusion to maximize volumetric productivity and to compete with established fed-batch processes (Steinebach et al., 2017; Bausch, Schultheiss, & Sieck, 2018). Optimization strategies for culture conditions which enhance the cellular productivity can therefore contribute to perfusion as an intensified biopharmaceuticals production alternative. Figure 1.1: Global annual sales for biopharmaceuticals for the years 2010, 2013 and 2016 in billion US-$, divided into total sales and sales of the 10 biopharmaceuticals with the highest sales numbers (data from Walsh (2014); Morrison and Lähteenmäki (2017)) . 8 1.2 Motivation 1.2 Motivation Although media and feeding strategies in large-scale fed-batch processes have been sub- ject to optimization (Birch & Racher, 2006; Shukla & Thömmes, 2010), the occurrence of local gradients cannot be completely prevented due to larger volumes and increased mixing times. Accordingly, typical gradients emerging in production scale bioreactors result from gassing, the addition of feed solutions, or titration agents for pH control (Xu et al., 2018). Since the pH in cell culture systems is mostly controlled via the addition of CO2 in the ingas stream as an acid and Na2CO3 solution as base, local accumulation of these agents and resulting inhomogeneities of the pH correlate and have to be consid- ered during scale-up (Xing, Kenty, Li, & Lee, 2009; Sieblist et al., 2011). Presence and negative influence of these gradients have been shown before (Xing et al., 2009; Osman, Birch, & Varley, 2001; Langheinrich & Nienow, 1999). As scale-down models proved, growth and productivity of mammalian cells are often diminished, while by-product for- mation in form of lactate can significantly increase in zones with higher pH or elevated partial pressure of CO2 (Gray, Chen, Howarth, Inlow, & Maiorella, 1996; Goudar et al., 2007; Darja et al., 2016; Brunner, Fricke, Kroll, & Herwig, 2017; Brunner, Doppler, Klein, Herwig, & Fricke, 2018; Ivarsson, Noh, Morbidelli, & Soos, 2015). However, effects on metabolic flux distributions and the underlying energy metabolism have been adressed seldom, although the impact on energy demanding cellular productivity is sub- stantial (Brunner et al., 2018; Russell, 2007; Dickson, 2014). Similarly, cellular productivity is crucial for the performance of perfusion processes (Bausch et al., 2018; Steinebach et al., 2017). However, most studies in the past decade investigated the maximization of cell densities (Clincke, Molleryd, Zhang, et al., 2013) or the characterization of cell retention devices with a recent focus on hollow fiber mod- ules (Karst, Serra, Villiger, Soos, & Morbidelli, 2016; Clincke, Molleryd, Zhang, et al., 2013; Kelly et al., 2014). Individual approaches for the optimization of cellular metabolism and productivity in particular included temperature reduction (Wolf et al., 2018) or glucose limitation (Takuma, Hirashima, & Piret, 2007). Especially glucose limitation can be a potential tool to increase metabolic efficiency and reduce by-product formation by limiting carbon availability. Until now, similar effects were only observed in late batch or fed-batch phases (Mulukutla, Gramer, & Hu, 2012; Ivarsson et al., 2015; 9 Introduction and Motivation Martínez et al., 2013). However, possible effects of glucose limitation on flux distribu- tions and energy availability have not been analyzed for perfusion, yet. Consequently, the scope of this thesis can be divided into two parts, both focusing on the optimization of antibody production processes with CHO cells: • In the fed-batch part the impact of the typical scale-up sensitive parameters Na2CO3 addition and partial pressure of CO2 (pCO2) was investigated in comparison to an industrial reference process of the project partner Boehringer Ingelheim. • In the perfusion part the effect of glucose limitation as a process intensification approach was analyzed during steady state cultivation in comparison to a non glucose limited reference steady state. Main objective for both parts was to unravel the connection of extracellular process settings and intracellular state. Extracellular rates calculation, quantification of intracel- lular metabolite pool sizes and modelling of metabolic fluxes were performed to gain insight into cellular adaptations to the current conditions. Special focus was put on the connection between energy availability in terms of ATP and cell specific antibody pro- ductivity. Consequently, the resulting volumetric productivities were used as indicators for identifying the most efficient cellular state by either avoiding detrimental effects of stress factors in fed-batch mode or by introducing measures to increase efficiency of carbon usage in perfusion mode. 10 2 Theoretical Background 2.1 Chinese hamster ovary cells Several well characterized mammalian cells lines like chinese hamster ovary (CHO), baby hamster kidney (BHK), mouse myeloma (NS0) or human embryonic kidney (HEK) have been studied extensively and are available for use in both research and industry. All of them are capable to express recombinant proteins which are correctly folded and posttranslationaly modified to be applied in the human body (J. Y. Kim, Kim, & Lee, 2012). Therefore, solubility, stability and biologic activity are guaranteed. Typical prod- ucts derived from mammalian cells are antibodies, therapeutic proteins or viral vaccines which cannot be correctly synthesized by bacteria due to the missing glycosylation abil- ities. Despite the broad range of available mammalian cell lines, the main production hosts for biopharmaceuticals nowadays remain chinese hamster ovary cells (figure 2.1). 61 % of biopharmaceuticals with mammalian producer cells are expressed by CHO cells (Ecker et al., 2015; Kantardjieff & Zhou, 2013). First, the chinese hamster (Cricetulus griseus) was only used as laboratory animal since the early 1900s. In the 1950s, cells were isolated from the hamsters’ ovaries and successfully cultivated in vitro for the first time, leading to various further researches in the following decades (Puck, Cieciura, & Robinson, 1958). In the 1980s, CHO cells became the host for the first therapeutic protein approved for use on the pharmaceutical market: the human tissue plasminogen activator (Wurm, 2004). Decisive for the development of such an industrial process was the creation of highly productive cells by the application of gene amplification systems. Gene amplification systems using dihydrofolate reductase (DHFR) or glutamine synthetase (GS) are the most widely used today. Urlaub, Käs, Carothers, and Chasin (1983); Urlaub et al. (1986) generated the precursors of todays main CHO production cell lines by intro- ducing DHFR deficiency into the cells for the first time in the 1980s. The cells are 11 Theoretical Background transfected with the gene for the target product along with an amplifiable gene for ei- ther DHFR or GS, depending on the system. In absence of the according metabolites hypoxanthine and thymidine (DHFR) or glutamine (GS) only transfected cells are ca- pable of surviving (Butler, 2005). Accordingly, the product of interest is expressed in the remaining viable cells. After the selection process, screening is performed to find clones with a stable expression behaviour and the highest cell-specific productivity. In the following process development, the cellular behavior is characterized and optimal production conditions are found regarding pH, dissolved oxygen, nutrient availability and various further parameters. Consequently, not only the mode of cultivation (see sections 2.4 and 2.5) is decisive but also the medium composition and the resulting metabolic adaptations of the cell. Finally, the scale-up in production size bioreactors is the last step in the generation of current biopharmaceutical production processes using CHO cells as manufacturing organisms besides the regulatory requirements for market approval (Wurm, 2004). CHO 61% Human 8% SP2/0 9% NS0 8% Other 3% Hybridoma 8% BHK 3% Figure 2.1: Percentage of biopharmaceuticals produced in mammalian systems sorted by cell line in the year 2012 (data from (Kantardjieff & Zhou, 2013)) . 12 2.2 Metabolism 2.2 Metabolism 2.2.1 Main carbon metabolism The main carbon metabolism of mammalian cells comprises multiple pathways (com- pare figure 2.2). The main anabolic and catabolic pathways providing cellular build- ing blocks and energy, respectively, are the glycolysis, the pentose phosphate path- way (PPP), the tricarboxylic acid cycle (TCA), the oxidative phosphorylation, the glu- taminolysis and the metabolism for the remaining amino acids. As the starting point of the glycolysis, glucose is the main substrate and the backbone for energy generation (figure 2.2). Glucose uptake is achieved by transporters in the cellular membrane and is based on the present concentration gradient (S. Ozturk & Hu, 2005). The first glycolysis intermediate glucose-6-phosphate is not solely used in the glycolysis. Additionally, it is converted to ribulose-5-phosphate in the first step of the pentose phosphate pathway under the generation of 2 NADPH. The fraction of carbon derived from glucose entering the PPP can vary substantially depending on the cell line and process conditions. For CHO cell lines values between 0.3 and 20 % have been re- ported (Carinhas et al., 2013; Templeton, Dean, Reddy, & Young, 2013; Martínez et al., 2013). All PPP reactions are located in the cytosol. Ribulose-5-phosphate is a precursor for nucleotides such as DNA and RNA. Therefore, the pentose phosphate pathway is especially important for cell replication, biomass generation and growth. Additionally, nucleotides and coenzymes like ATP or NAD are formed out of ribulose-5-phosphate. The remaining glucose-6-phosphate is transformed in the further course of glycolysis which is located in the cytosol and consists of several single reactions. Taken together, reactions of glycolysis yield two molecules of pyruvate out of every single molecule glucose taken up by the cells. During the conversion of glucose to pyruvate 2 molecules of ATP are generated directly and 2 molecules of NAD+ are reduced to NADH (Pörtner, 2009; Berg, Tymoczko, Stryer, et al., 2012). Excess intracellular pyruvate concentrations can result in the by-product formation of lactate, especially under conditions of high glucose availability as part of overflow metabolism (Luo et al., 2012). Lactate is directly formed by reducing pyruvate via the 13 Theoretical Background lactate dehydrogenase under the regeneration of NAD+. Lactate accumulating in the medium can lead to decreased pH and therefore increased base titration and osmolality but also direct inhibition of growth and productivity (K. Chen, Liu, Xie, Sharp, & Wang, 2001). Inhibiting concentrations of more than 60 mM especially occur in fed-batch pro- cesses with prolonged process times and high glucose availabilities (Lao & Toth, 1997). Previously secreted lactate can be taken up by the cells again. The metabolic switch from lactate formation to uptake has been the focus of several studies. Clearly, lac- tate can be consumed as alternative carbon source once glucose is depleted, especially in batch cultures (Martínez et al., 2013; S. S. Ozturk, Riley, & Palsson, 1992). Fur- thermore, changing process conditions like limiting glucose availability (Xie & Wang, 1994), using different carbon sources (Altamirano, Paredes, Illanes, Cairo, & Godia, 2004) or applying pH (Ivarsson et al., 2015) or temperature shifts (Sou et al., 2015) has shown to lead to lactate consumption by the cells although glucose is still present in the medium. The redox (NAD/NADH) balance also plays a significant role in the forma- tion of lactate and is interrelated with the ATP formation (see section 2.2.2). However, not all mechanisms and origins of triggering lactate consumption have been completely elucidated, yet. Despite the possible by-product formation from lactate, the major part of pyruvate resulting from glycolysis is transported into mitochondria and fuels the TCA (figure 2.2). Depending on the cell line and culture conditions, fractions of pyruvate entering the TCA may differ. For CHO cells lowest published values were 20 % (Templeton et al., 2013) but can go up to 100 % or even more if lactate is consumed and thereby ad- ditionally fueling the TCA via pyruvate (Templeton et al., 2013; Carinhas et al., 2013). In the first step of TCA, pyruvate is oxidatively decarbonized to acetyl-CoA under for- mation of NADH. Together with the resulting oxaloacetate from the previous TCA cir- culation acetyl-CoA is transformed to citrate and further to alpha-ketoglutarate yielding NADH. More NAD is reduced to NADH in the transformation of alpha-ketoglutarate to succinyl-CoA. In the next step GTP is formed (depicted as energy equivalent to ATP in figure 2.2) when succinyl-CoA is transformed to succinate. This is oxidized to fu- marate which is hydrated to malate. Eventually, malate is oxidized to form oxaloacetate fueling a new circulation of the TCA in combination with acetyl-CoA. Taken together, the TCA finally yields 1 molecule of ATP and 3 molecules of NADH per circulation of 14 2.2 Metabolism pyruvate and double the amount per molecule of glucose (Berg et al., 2012; S. Ozturk & Hu, 2005). Additionally, one molecule of NADH is generated prior, when one molecule pyruvate is being transformed to acetyl-CoA. The second main substrate for CHO cells is the amino acid glutamine which is both carbon and nitrogen source. Glutamine uptake sums up to about 10 % to 20 % compared to glucose uptake, depending on the concentration in the medium (Wahrheit, Nicolae, & Heinzle, 2014; Nolan & Lee, 2011). Via glutaminolysis glutamine is transformed to alpha-ketoglutarate in two steps and thereby fuels the TCA (figure 2.2). In the first step glutamine is transformed to glutamate under the formation of ammonia followed by the second step in which glutamate is transformed to alpha-ketoglutarate. Alternatively, the glutamate resulting from the first step can be secreted by the cells. Glutamine overflow metabolism can lead to ammonia accumulation due to the formation of ammonia during the first step of glutaminolysis. Ammonia is the second major by-product with inhibiting effects besides lactate. Negative effects on growth and product formation were observed for concentrations of about 6-8 mM or higher (Hassell, Gleave, & Butler, 1991; P. Chen & Harcum, 2006; Schneider, Marison, & von Stockar, 1996). 2.2.2 Energy generation and redox balance The energy generation is most important for both biomass generation and cellular anti- body production. CHO cells can metabolize their main energy source glucose in multi- ple ways with different efficiencies. One is the transformation to lactate via glycolysis. The transformation of glucose to lactate yields 2 molecules of ATP. More efficient is the complete oxidation of glucose via glycolysis, TCA and the subsequent oxidative phos- phorylation to CO2 where 36 molecules of ATP can be produced theoretically (Warburg, 1956; Glacken, 1988). The cofactors NADH and FADH2 contribute to ATP synthesis via electron transport chain and oxidative phosphorylation. The resulting number of molecules of ATP produced per NADH or FADH2, respectively, is defined as P/O ratio. The ratio describes the efficiency of oxidative phosphorylation and the previous electron transport chain and depends on the used cofactor. For NADH and FADH2 P/O ratios of 2.5 and 1.5 were described (Hinkle, 2005). 15 Theoretical Background Accordingly, the redox (NAD/NADH) balance is of major importance for the ATP formation. As introduced in section 2.2.1, pyruvate plays a crucial role in the redox bal- ance since lactate is produced under the regeneration of NAD+, therefore representing a sink for NADH. By contrast, pyruvate fueling the TCA eventually yields 3 molecules of NADH, providing a major source of NADH. Hence, lactate formation prevents larger fractions of NADH from being transported into mitochondria. Under the ideal metabolic state of no lactate formation the complete NADH from glycolysis and TCA is subse- quently used in oxidative phosphorylation for the generation of ATP (Nolan & Lee, 2011). Consequently, interrelation of the redox balance and the cell-specific lactate for- mation has been found for results of intracellular flux analyses (Nolan & Lee, 2011; Brunner et al., 2017). 16 2.2 Metabolism Figure 2.2: Simplified scheme of the main carbon metabolism. Depicted are the main metabolites of glycolysis and tricarboxylic acid cycle including the ATP and NADH generation. Simplifications were made for the pentose phosphate pathway leading to biomass precursors and the amino acid metabolism yielding the monoclonal antibody. Compartmentalization regarding mitochondrial department was neglected. Metabolites with a grey background were quantified extracellularly in the experimental part of this work. 17 Theoretical Background 2.3 Flux analysis in CHO cells The analysis of intracellular fluxes can provide a thorough understanding of metabolic states and adaptations during production processes using both microbial and mammalian host systems. Intracellular flux distributions can hardly be measured and are therefore calculated based on a model representing the cellular metabolism. The reconstruction of the cellular metabolism and implementation into a model structure is mostly done with the help of biochemical databases (Thiele, Price, Vo, & Palsson, 2005). Simplifications are often made to improve clarity of the results and to decrease efforts during both model preparation and flux computation. Two approaches are mostly applied to calculate in- tracellular flux distributions: metabolic flux analysis (MFA) and flux balance analysis (FBA) (Stephanopoulos, Aristidou, & Nielsen, 1998; Orth, Thiele, & Palsson, 2010). In both approaches a pseudo steady-state is assumed for the intracellular metabolic state, equivalent to no accumulation or depletion of metabolites. S · v = 0 The metabolic reactions and the involved metabolites are stored in a stoichiometric matrix S. The rows in S correspond to metabolites and the columns to reactions. The matrix is then combined with a vector v containing the corresponding flux values for measured and non-measured fluxes to be determined For MFA the metabolic model is simplified by focusing on the main pathways and most significant reactions. Therefore, the total number of reactions is assessable to such an extent, that measured extracellular uptake and production rates are sufficient to calculate all intracellular fluxes. Consequently, the system can be described as de- termined, yielding a unique solution of the small-scale metabolic network (Niklas & Heinzle, 2011). Analysis is often performed in distinct states of the process such as the growth or the stationary phase where cell-specific rates stay almost constant and can be averaged over longer periods of time. Alternatively, time resolved flux distributions can be gained by performing dynamic metabolic flux analysis (dMFA) where systems dynamics are taken into account by the combination with either a kinetic model or time- series data analysis (Ahn & Antoniewicz, 2012). 18 2.3 Flux analysis in CHO cells For FBA mostly larger models are used, ranging up to genome-scale size with hun- dreds of reactions included (Feist & Palsson, 2008). Therefore, FBA is used for models classified as underdetermined due to the large reaction numbers and comparably small number of measured rates. Instead of simplifying the model by reducing the reaction number, upper and lower flux bounds which are derived from measured extracellular up- take and production rates are introduced as constraints (Z. Huang, Lee, & Yoon, 2017; Orth et al., 2010) (see figure 2.3). Furthermore, an objective function like maximization of growth or ATP production is chosen and imposed on the system based on biological coherency and state of the culture to gain a reasonable solution (Feist & Palsson, 2010; Orth et al., 2010). Z = cT · v The objective function Z is a linear combination of fluxes with the vector of weights c. Thereby, the reactions contributing to the objective are chosen. Hence, FBA can also be used for different approaches like media or strain optimization before experimental validation. However, depending on the size of the solution space and chosen constraints multiple solutions could potentially maximize the objective (Z. Huang et al., 2017; Orth et al., 2010). Software solutions like the Constraint-Based Reconstruction and Analysis (COBRA) toolbox enable quickly implemented flux analysis (Schellenberger et al., 2011). There- fore, multiple cell lines and a wide scope of cultivation settings have been analyzed with flux analysis. Applications in the field of biopharmaceuticals production with CHO cells have been diverse in the past few years but contribute to the understanding of various process optimization approaches. As following examples show, investigations cover different cultivation modes but especially multiple approaches to analyze and intensify production processes on a metabolic level: The investigation of the switch from lactate production to consumption is an impor- tant metabolic event not yet fully understood. However, studies showed higher TCA influx and elevated energetic efficiency for lactate consuming cells using flux analysis (Mulukutla et al., 2012; Martínez et al., 2013). When comparing different cultivation phases increased glycolytic and pentose phosphate fluxes were found in early stages 19 Theoretical Background when growth was high in contrast to later stages when TCA flux was high and growth reduced (Templeton et al., 2013) Decreased temperature led to higher productivity but remarkably decreased fluxes of carbon in all pathways (Sou et al., 2015). Ivarsson et al. (2015) found decreased lactate uptake and increased TCA cycle fluxes for lower pH values during batch fermentation. Consequently, the resulting origin of ATP dif- fered among varying pH values. Another approach included the addition of the histone deacetylase inhibitor butyrate, which has been shown to have positive effect on the specific productivity in CHO cells (Jiang & Sharfstein, 2008). Flux analyses revealed that butyrate treatment led to an overall increase in intracellular fluxes even in station- ary phase when fluxes in the previously mentioned studies were lower (Carinhas et al., 2013). Templeton, Xu, Roush, and Chen (2017) compared industrial fed-batch and per- fusion processes using flux analysis and found minor differences in the intracellular flux distributions but major deviations in underlying protein production concerning the fractionation of biomass and antibody production. 20 2.3 Flux analysis in CHO cells Figure 2.3: Workflow of flux balance analysis including 1) the generation of a stochio- metric model of the cellular metabolism from databases and publications, 2) the transfer of the reactions to the stochiometric matrix S and the imposition of constraints (lower bounds LB and upper bounds UB) for the cell-specific rates v whereby the solution space is constrained and 3) the optimization of the fluxes using an objective function z and identification of an optimal flux distribution solution within the constrained solution space. 21 Theoretical Background 2.4 Fed-batch processes Among the three main cultivation modes (figure 2.4), fed-batch processes are the most commonly used cultivation mode in industrial production of biopharmaceuticals. In contrast to simple batch processes without volume change after inoculation, fed-batch processes show prolonged cultivation times and increasing volume due to the addition of growth supporting feed media during the cultivation. Performance parameters and recent approaches for process optimization are discussed in the following section. Figure 2.4: Schematic representation of the three main cultivation modes. The batch in A) is a closed system with no addition of medium and constant liquid volume after inoc- ulation. The fed-batch in B) is supplied with an enriched feed-medium after inoculation without removal of the spent medium and therefore has an increasing liquid volume. The perfusion setup in C) is characterized by an additional cell retention system and continuous addition and withdrawal of medium and therefore constant liquid volume. Feed-medium is added to the bioreactor while spent medium is pumped out via the cell retention device which retains the cells within the system. Consequently, cell densities can reach much higher levels in perfusion culture. 22 2.4 Fed-batch processes 2.4.1 Process overview and performance Typical values for viable cell densities in published CHO fed-batch cultivations in the past decade range from 20 - 30 x 106 cells/mL (Y.-M. Huang et al., 2010; Reinhart, Kaisermayer, Damjanovic, & Kunert, 2013), whereas maximum cell specific productiv- ities can reach values around 50 pg/cell/day with resulting final titers of up to 10 g/L (Zboray et al., 2015; Reinhart et al., 2013). Simple batch processes are therefore out- performed by approximately the factor of three in terms of viable cell density and by the factor of two in cultivation time. Consequently, titers are significantly increased in fed-batch processes (Kunert & Reinhart, 2016). To avoid excessive increase in liquid volume and therefore unwanted dilution of cell density and the product of interest, the added feed is concentrated multiple times (about 10 - 15 x) compared to the batch medium. Possible feeding strategies may be based on previously determined consumption rates or feedback loops regarding continuously measured bioprocess parameters (Wlaschin & Hu, 2006; L. Zhang, Shen, & Zhang, 2004). The feeding routine can be realized by either bolus addition of the feed medium (mostly done once or twice daily) or a continuous feeding thereof. The first option is simple and based on measured substrate concentrations shortly before the feeding but can contribute to the increased accumulation of inhibitory by-products. Lactate, resulting from carbon metabolism, and ammonia, resulting from nitrogen metabolism (compare section 2.2.1), may be produced in overflow due to the increased nutrient availability to ensure sufficient substrate supply until the next bolus feedings. Bolus feeding is often used in large scale biopharmaceutical production processes because of the ease of operation (Li, Vijayasankaran, Shen, Kiss, & Amanullah, 2010). On the contrary, the introduction of a more sophisticated continuous feeding, respectively the variation of the feed rate, and therefore the minimization of the substrate availability can consequently contribute to the reduction of by-product formation (Luo et al., 2012; Ljunggren & Häggström, 1992). The concentration ranges of half saturation constants for glucose (0.4 mmol/L) and glutamine (0.1 mmol/L) can be used as setpoints to avoid lactate and ammonia accumulation although feeding strategies have to be precise to avoid limitation effects (Chee Furng Wong, Tin Kam Wong, Tang Goh, Kiat Heng, & 23 Theoretical Background Gek Sim Yap, 2005; Abu-Absi et al., 2013). Possible consequences may be losses in cell viability or product quality once nutrient limitation becomes too pronounced. Nevertheless, because of superior volumetric productivities compared to batch pro- cesses and the simplicity of operation compared to perfusion processes (see section 2.5) fed-batch processes are the main biopharmaceutical production mode and remain in the focus of process optimization studies. 2.4.2 Parameters influencing process performance Several process parameters like temperature, pH, osmolality, partial pressure of CO2 (pCO2) or metabolite concentrations may influence the performance of the cultivation by altering growth, productivity or metabolism of the cells. Although specific effects may vary depending on the cell line, especially in large scale cultivations inhomo- geneities can lead to deviations in cellular performance. The partial pressure of CO2 is one of multiple parameters showing both formation of inhomogeneities in large scale and influence on cellular metabolism (Gray et al., 1996; Xu et al., 2018). Furthermore, the pCO2 often increases during the course of a cultiva- tion due to the increasing cell densities but also through the accumulation of inorganic carbon species as a consequence of base addition (Goudar et al., 2007). Increased CO2 stress is critical because it is known to hinder both cellular growth and productivity but is nevertheless necessary in smaller fractions to mimic the physiological environment the cells were originally obtained from (Mostafa & Gu, 2003; Gray et al., 1996; Brunner et al., 2018). Likewise, osmolality rises towards the end of a fed-batch cultivation. While the start- ing values reach up to 300 mOsm/kg and therefore similar to cytoplasmic osmolalities (Mortimer & Müller, 2001), an increase is mainly caused by the accumulation of salts and ions through the feeding but also through base addition. Consequently, with increas- ing process time osmolalities above 450 mOsm/kg can contribute to a loss in growth and viability and an increasing number of lysing cells and therefore the begin of the station- ary or decline phase. Furthermore, glucose uptake and lactate formation is increased (Pfizenmaier, Matuszczyk, & Takors, 2015; Han, Koo, & Lee, 2009; N. S. Kim & Lee, 2002; Zhu et al., 2005). Cell lysis itself increases osmolality further and foaming starts 24 2.5 Perfusion processes due to released cellular proteins. However, beneficial effects on cell specific productiv- ity for osmolalities have been observed for osmolalities of about 350 mOsm/kg coin- ciding with cell cycle arrest (D. Shen et al., 2010; Pfizenmaier et al., 2015; Pfizenmaier, Junghans, Teleki, & Takors, 2016). Temperature is normally held constant at the physiological value of 37.0 ◦C to sup- port optimal growth. However, shifts downwards can be used for cell cycle arrest and prolongation of the stationary phase in late fed-batch stages when the cell den- sity already reached its maximum value. Hereby, growth and by-product formation are substantially slowed down while productivity remains on a constant level or is even increased as shown by several authors (Yoon, Choi, Song, & Lee, 2005; Kaufmann, Mazur, Fussenegger, & Bailey, 1999; Furukawa & Ohsuye, 1999). Interrelated with CO2 and base addition, pH values may be distributed inhomoge- neously in large scale cultivations. Numerous proteins with roles in diverse cellular functions are sensitive to the H+ concentration (Casey, Grinstein, & Orlowski, 2010). Likewise, the pH has been shown to substantially affect the lactate formation of CHO cells. Several authors showed decreasing trends of lactate formation rates with the pH value decreasing towards 6.8. Potential reasons for the reduction in lactate formation at lower pH values were the avoidance of a further decrease in extra- and intracellular pH, different activities of glycolytic enzymes and redox balancing (Ivarsson et al., 2015; Trummer et al., 2006; Liste-Calleja et al., 2015; Yoon et al., 2005). Besides the actual substrate concentrations (see section 2.2.1), substitution of the substrate glucose with alternative sugars such as galactose or mannose can also lead to the reduction of by-product formation. However, overall performance in terms of growth and product formation may be hampered in comparison to the use of glucose (Altamirano et al., 2001, 2004; Berrios, Altamirano, Osses, & Gonzalez, 2011). 2.5 Perfusion processes Besides fed-batch cultivations perfusion cultivations (see figure 2.4) are emerging in biopharmaceuticals production as intensified processes using a smaller volume but higher cell densities. Perfusion describes a continuous culture where the same amount of used medium is withdrawn as fresh medium is added and therefore the cultivation volume remains constant. Additionally, a cell retention device is connected to the bioreactor to 25 Theoretical Background avoid the wash-out of cells as experienced in simple continuous cultures. The perfusion rate can be adjusted in subsequent processes or even within a single process. Hereby, substrate supply and by-product removal can be controlled (Chotteau, 2015). Conse- quently, cell densities much higher than during fed-batch cultivations can be reached. Published values range up to 200 x 106 cells/mL (Clincke, Molleryd, Zhang, et al., 2013). Through the continuous mode of operation the process time can theoretically be extended for weeks to months depending on both biological and technical factors. The cell line has to prove stability in growth and productivity over several generations while the stability of the set-up but especially the cell retention device is crucial for the continuous operation. Although titers are not necessarily higher than in fed-batch mode due to the constant wash out of the product of interest, the volumetric produc- tivity is increased because of the high cell densities. Furthermore, the cell retention devices facilitate the introduction of subsequent continuous downstreaming operation since the product is already separated from the cells (Steinebach et al., 2017; Warikoo et al., 2012). Alternatively, perfusion can also be used for cell bank generation (Clincke, Molleryd, Samani, et al., 2013) or in seed reactors (Pohlscheidt et al., 2013) to quickly reach high inoculation densities for the main reactor. An overview of different cell retention devices is depicted schematically in figure 2.5. Depending on the device it is either placed inside the bioreactor or connected externally. However, the mode of action of all cell retention devices is based on differences in size or density between medium components and cells. Filtration techniques retain cells by their size. Limitations can occur once the pores start clogging with biological material and flow is reduced. However, various differ- ent devices have been characterized and published in the recent decades like external hollow fibers for tangential flow filtration (figure 2.5 A) (Clincke, Molleryd, Zhang, et al., 2013; Clincke, Molleryd, Samani, et al., 2013; Karst et al., 2017; Kelly et al., 2014) or alternating flow filtration (Karst et al., 2016; Clincke, Molleryd, Zhang, et al., 2013; Clincke, Molleryd, Samani, et al., 2013). Here, the pores of the hollow fibers are flushed either tangentially or by introducing a backflush in certain time intervals. This is supposed to ensure long-time application without membrane fouling. Alternatively, spin filters (figure 2.5 B) are predominantly integrated on the agitator shaft, where the 26 2.5 Perfusion processes constant rotational movement is supposed to reduce fouling (Komolpis, Udomchok- mongkol, Phutong, & Palaga, 2010; Vallez-Chetreanu, Ferreira, Rabe, von Stockar, & Marison, 2007). Inclined settlers use the principle of sedimentation for the separation of cells (figure 2.5 C). The cell broth is added to the settler and the sedimenting cells are withdrawn at the bottom and sent back to the bioreactor while the spent medium is slowly pumped from the top. Using sedimentation almost no shear stress acts on the cells. On the other side, long residence times within the settler might limit cells in substrate or oxygen availability (Wen, Teng, & Chen, 2000; Y. Shen & Yanagimachi, 2011; Choo et al., 2007). The labile antihemophilic factor VIII is produced in large-scale using gravita- tional settlers (X. Zhang, Wen, & Yang, 2011). Furthermore, several alternative cell retention devices have been described in litera- ture. Among them are acoustic separators applying an acoustic resonance field (figure 2.5 E). Here, the cells agglomerate in standing acoustic waves and can be led back into the bioreactor while the medium is removed. Drawback may be the increased shear applied to the cells by the acoustic field (Ryll et al., 2000; Shirgaonkar, Lanthier, & Ka- men, 2004; Gorenflo, Smith, Dedinsky, Persson, & Piret, 2002). In Hydrocyclones the cell suspension from the bioreactor is separated into an underflow and an overflow (fig- ure 2.5 D). Since the cells have a higher density they are carried out of the hydrocyclone via the underflow and led back to the reactor. The spent medium is rising inside the hydrocyclone and leaving as cell-free harvest through the overflow. Again, shear sensi- tive cells could be affected by the applied forces (Pinto, Medronho, & Castilho, 2008; Elsayed, Medronho, Wagner, & Deckwer, 2006). Centrifugal forces can be used by connecting a centrifuge to the bioreactor as done by B. J. Kim, Oh, and Chang (2008) (figure 2.5 F). Sterile single-use inlets can be inserted to simplify the handling in be- tween different runs. Although the separation settings can be varied dynamically, the long-term robustness could be decreased due to the mechanical complexity of the cen- trifugal set-up. In the recent years cell retention systems using either sedimentation, tangential flow filtration or alternating flow filtration devices have become the most applied set-ups in 27 Theoretical Background industry. The aforementioned devices are favored due to the scalability of the systems and the low shear stress applied (S. Ozturk & Hu, 2005). Regarding process optimization results from process parameter investigations can only partly be transferred from fed-batch cultivations to perfusion systems (see section 2.4.2). Temperature decrease for example can be used similarly for perfusion process as described in section 2.4.2, as a recent study by Wolf et al. (2018) showed. Partial pressure of CO2 can be more pronounced because of the much higher cell densities (Goudar et al., 2007) whereas osmolalities can increase as a result of highly concen- trated perfusion media but are restricted in maximum values by the constant wash-out. Additionally, when adjusting process parameters long-term stability has to be consid- ered in terms of stable growth respective viability due to the continuous mode of oper- ation. Consequently, perfusion rates may have to be changed after process parameter variation. 28 2.5 Perfusion processes Figure 2.5: Schemes for different published cell retention devices used for perfusion processes. Devices can be divided into bioreactor internal and external. A) tangential flow filtration with an external hollow fiber module, B) internal spin filter, C) external gravity settler, D) external hydrocyclone, E) external ultrasonic separator, F) external centrifuge 29 3 Materials and Methods 3.1 Materials 3.1.1 Chemicals Table 3.1: Chemicals Chemical Manufacturer 3-mercaptopropionic acid Sigma-Aldrich, USA 300 mOsm/kg standard Gonotec, Germany 500 mOsm/kg standard Gonotec, Germany 9-fluorenylmethyl chloroformate Sigma-Aldrich, USA Acetonitrile ≥ 99.9 % VWR, Germany Buffer solution LaboTrace TraceAnalytics, Germany Cis-aconitate ≥ 98 % Sigma-Aldrich, USA Adenosine diphosphate disodium salt Gerbu, Germany Adenosine monophosphate disodium salt Fluka, USA Adenosine triphosphate disodium salt Gerbu, Germany Algal lyophilized cells U-13C ≥ 99 % Sigma-Aldrich, USA Amino acid standard Sigma-Aldrich, USA Antifoam (proprietary) Boehringer Ingelheim AG & Co. KG, Germany Basal medium (proprietary) Boehringer Ingelheim AG & Co. KG, Germany Boric acid Merck, Germany Bovine sermum albumin VWR, Germany Choloroform ≥ 98 % Sigma-Aldrich, USA 31 Materials and Methods Chemicals - continued Chemical Manufacturer Dimethyl sulfxoide Sigma-Aldrich, USA Ethanol ≥ 96 % Carl Roth, Germany Ethylenediaminetetraacetic acid Carl Roth, Germany Feed medium (proprietary) Boehringer Ingelheim AG & Co. KG, Germany Fructose-6-phosphate ≥ 98 % Sigma-Aldrich, USA Glucose-6-phosphate ≥ 98% Sigma-Aldrich Glucose and lactate standard TraceAnalytics, Germany Hydrochloric acid Carl Roth, Germany L-asparagine ≥ 99 % Sigma-Aldrich, USA L-glutamine ≥ 99 % Carl Roth, Germany L-ornithine ≥ 99% Sigma-Aldrich, USA L-tryptophane ≥ 99 % Fluka, USA Methanol ≥ 99.8 % VWR, Germany Ortho-phtaldialdehyde Fluka, USA pH 4.00 buffer solution Carl Roth, Germany pH 7.00 buffer solution Carl Roth, Germany pH 9.00 buffer solution Carl Roth, Germany Phosphoric acid Fluka, USA Potassium hydroxide 45 % Sigma-Aldrich, USA Potassium phosphate dibasic Sigma-Aldrich, USA Potassium phosphate monobasic Sigma-Aldrich, USA Preculture medium (proprietary) Boehringer Ingelheim AG & Co. KG, Germany Pyruvate ≥ 99 % Sigma-Aldrich, USA SeramunBlau Seramun Diagnostica, Germany Sodium azide Sigma-Aldrich, USA Sodium carbonate Carl Roth, Germany Sodium chloride ≥ 99.8 % Carl Roth, Germany 32 3.1 Materials Chemicals - continued Chemical Manufacturer Sodium hydrogen carbonate Carl Roth, Germany Sodium hydroxide Carl Roth, Germany Sodium phosphate dibasic Carl Roth, Germany Sodium tetraborate decahydrate Sigma-Aldrich, USA Sulfuric acid ≥ 98 % Carl Roth, Germany Tetrabutylammonium bisulfate Fluka, USA Tris(hyroxymethyl)aminomethane Carl Roth, Germany Trypan blue solution 0.4 % Sigma-Aldrich, USA Tween 20 Fluka, USA Water LC-MS grade VWR, Germany 3.1.2 Consumables Table 3.2: Consumables Consumable Manufacturer 96 well plate, F-bottom Greiner, Germany Cedex smart slides Roche, Germany Cryo vials Greiner, Germany Filter 0.22 µm Carl Roth, Germany Filter Sartolab P-20 0.2 µm Sartorius, Germany Glass fiber filter A/D Pall, USA Hollow fiber module CFP-4-E-3X2MA GE Healthcare, Germany HPLC glas vials VWR, Germany Luer lock connectors Carl Roth, Germany Metrigard filter Pall, USA Microtiter Plates Greiner, USA Pipet tips Sarstedt, Germany Reaction tubes Sarstedt, Germany 33 Materials and Methods Consumables - continued Consumable Manufacturer Reaction tubes safe lock Eppendorf, Germany Sequant ZIC-pHILIC Di2chrom, Germany Serological pipets Sarstedt, Germany Shake flasks Corning, USA Supelcosil LC18 Sigma-Aldrich, USA Syringe Omnifix with luer lock Braun Melsungen, Germany Tube for Osmomat 030 Gonotec, Germany Vial Inserts VWR, Germany Zorbax Eclipse Plus C18/250 x 4.6 mm Agilent Technologies, USA 3.1.3 Hardware Table 3.3: Hardware Hardware Manufacturer Avanti J-25 centrifuge Beckman Coulter, USA Bioblock DASGIP, Germany Bioreactor DS1500ODSS DASGIP, Germany Cedex XS analyzer Roche, Germany Cryoboy Nalgene, USA Cryostat F3 Haake, Germany Diode array detector Agilent Technologies, USA Fluorescence detector Agilent Technologies, USA Gas mixing module DASGIP, Germany HPLC 1200 Agilent Technologies, USA Incubator Minitron Infors, Switzerland LaboTrace TraceAnalytics, Germany Megafuge 1.0R Heraeus, Germany Microcentrifuge 5417R Eppendorf, Germany 34 3.1 Materials Hardware - continued Hardware Manufacturer Microplate reader Synergy BioTek, USA Nanopure II Barnstead, USA Osmomat 030 Gonotec, Germany Oxygen probe Mettler Toledo, USA Peristaltic pump 101u Watson Marlow, UK Peristaltic pump 505s Watson Marlow, UK pH probe Mettler Toledo, USA pH/DO module DASGIP, Germany Pipet boy acu Integra Biosciences, Switzerland Pump module DASGIP, Germany Research pipets Eppendorf, Germany Temperature/agitation module DASGIP, Germany Thermomixer comfort Eppendorf, Germany Total carbon analyzer Multi N/C 2100s Analytik Jena, Germany Triple Quad mass spectrometer 6410B Agilent Technologies Vacuum incubator RVC 2-33 Martin Christ, Germany Vacuum pump Vaccubrand, Germany 3.1.4 Software Table 3.4: Software Software Manufacturer Chemstation Agilent Technologies, USA DASGIP control 4.0 DASGIP, Germany Mass Hunter B.04.00 Agilent Technolgies, USA Matlab 2013a Mathworks, USA Microsoft Office Microsoft, USA 35 Materials and Methods 3.1.5 Antibodies for ELISA Table 3.5: Antibodies Antibody Manufacturer Anti-human IgG F(c) goat antibody Biomol, Germany Anti-human kappa chain goat antibody peroxi- dase conjugated Biomol, Germany IgG1 standard (proprietary) Boehringer Ingelheim AG & Co. KG, Germany 3.1.6 Buffers and Solutions Table 3.6: Buffers and solutions Buffer or solution Compound Agent for amino acid analysis 2.5 mg 9-fluorenylmethyl chloro- formate 1 mL acetonitrile Blocking solution ELISA 1 L Tris buffered saline 10 g Bovine serum albumine Coating buffer ELISA 3.7 g Sodium hydrogen carbon- ate 0.64 g Sodium carbonate ad 1.0 L Nanopure water Cryo medium 10 % Dimethyl sulfxoide 90 % Culture medium Dilution buffer ELISA 1 L Tris buffered saline 10 g Bovine serum albumine 5 mL Tween 20 (10 %) Elution buffer A for HPLC 10 mM Sodium phosphate dibasic (amino acids) 10 mM Sodium tetraborate dec- ahydrate 36 3.1 Materials Buffers and solutions - continued Buffer or solution Compound 0.5 mM Sodium azide Elution buffer A for HPLC (nucleotides) 100 mM Potassium phosphate monobasic 100 mM Potassium phosphate dibasic 4 mM Tetrabutylammonium bisulfate Elution buffer A for LC- MS/MS (polar metabolites) 10 % Ammonium acetate buffer (10 mM) 90 % acetontrile Elution buffer A for LC- MS/MS (alpha keto acids) 0.1 % formic acid Elution buffer B for HPLC 45 % acetonitrile (amino acids) 45 % methanol 10 % Water LC-MS grade Elution buffer B for HPLC (nucleotides) 70 % Elution buffer A (nu- cleotides) 30 % methanol Elution buffer A for LC- MS/MS (polar metabolites) 90 % Ammonium acetate buffer (10 mM) 10 % acetontrile Elution buffer A for LC- MS/MS (alpha keto acids) 0.1 % formic acid 90 % acetontrile Filtration buffer: pH = 6.95, 330 mOsm/kg 3.5 mM Potassium phosphate monobasic 6.7 mM Potassium phosphate dibasic 165 mM Sodium chloride OPA reagant for amino acid 10 mg Ortho-phtaldialdehyde 37 Materials and Methods Buffers and solutions - continued Buffer or solution Compound analysis 1 mL potassium borate (0.4 M) 8.2 L 3-mercaptopropionic acid Phosphate buffered saline 1.0 mM Potassium phosphate monobasic 5.6 mM Sodium phosphate dibasic 154.0 mM Sodium chloride Tris buffered saline for ELISA 6.1 g Tris(hyroxymethyl) aminomethane 8.2 g Sodium Chloride 6.0 mL Hydrochloric acid (6 mM) ad 1.0 L Nanopure water Tween 20 (10%) 10 % Tween 20 Washing solution for ELISA 1 L Tris buffered saline 5 mL Tween 20 (10 %) 38 3.2 Methods 3.2 Methods 3.2.1 Cryoconservation Working cell banks of the proprietary CHO cell line (provided by Boehringer-Ingelheim Pharma GmbH Co. KG, Germany) were prepared using the cryo medium (section 3.1.6) to conserve cells of similar origin as a starting point. Therefore, the cryo medium was mixed together and stored at 2 - 8 ◦C. Cells from the master cell bank were expanded in shake flasks (Corning, USA) and harvested during exponential growth phase with high viabilites above 90 %. After determination of the viable cell density, the cell suspension was transfered to sterile centrifuge tubes (Sarstedt, Germany) and centrifuged at 180 g for 7 minutes in a Megafuge 1.0R (Heraeus, Germany) at room temperature. The old medium was discarded and chilled cryo medium was added to yield a final cell density of 1 x 107 viable cells/mL. Aliquots of 1 mL were filled into each cryo vial (Greiner, Germany). The cryo vials were then placed in a controlled-rate freezing apparatus Cry- oboy (Nalgene, USA) with a temperature decrease of 1 ◦C/min and stored in a -70 ◦C freezer. After 24 h the frozen cryo vials were transferred to the vapor phase of a liquid nitrogen container for long-term storage. 3.2.2 Seed Train For thawing the cells and reaching a cell density of 0.4 x 106 cells/mL, 25 mL of precul- ture medium were heated to 36.5 ◦C and then added to a 125 mL shake flask. The cryo vial was thawed at 36.5 ◦C in water as fast as possible. Once the cells and cryo medium were thawed they were transferred to the prepared shake flask under the hood. The seed train (table 3.7) for the bioreactor cultivations consisted of a sequence of shake flasks with volumes ranging from 125 mL to 1000 mL to expand the biomass until the biore- actors could be inoculated. Each time the inoculation density was 0.4 x 106 cells/mL. Passaging of the cells to fresh medium and/or shake flasks with larger volume was done three times each week. The final step in the seed train for a fourfold bioreactor cultiva- tion were four 1000 mL shake flasks with a culture volume of 250 mL each. Cultivation time was about 72 h prior to bioreactor inoculation to reach cell densities of 3 - 4 x 106 cells/mL. Incubator (Infors, Switzerland) settings during seed train cultivations were 36.5 ◦C at 5 % CO2 and 120 rpm. 39 Materials and Methods Table 3.7: Seed train steps and volumes Time [h] Shake flask volume [mL] Liquid volume [mL] 0 125 25 72 125 50 120 125 50 168 250 100 240 2 x 1000 2 x 250 288 4 x 1000 4 x 250 360 4 x bioreactor 4 x 1000 3.2.3 Bioreactor Cultivation The following subsections describe the procedure of bioreactor cultivation from inocula- tion and controller adjustment to both intra- and extracellular sampling. All cultivations were performed in a fourfold DASGIP DS1500ODSS bioreactor system (DASGIP, Ger- many) equipped with a rushton turbine in the bottom and a pitch-bladed impeller above. Aeration was performed with a gas mixing station (DASGIP, Germany) connected to a L-sparger inside the bioreactor. Temperature, dissolved oxygen and pH were moni- tored by probes (Mettler Toledo, USA). Two pump modules (DASGIP, Germany) with four peristaltic pumps each were connected for the addition of feeds and base. DASGIP control 4.0 (DASGIP, Germany) was used as process control system. 3.2.3.1 Inoculation After running the seed train (section 3.2.2), inoculation was done by transferring expo- nentially growing cells into the bioreactor. The used volume of preculture was deter- mined by calculating a resulting seed density of 0.7 x 106 cells/mL for both fed-batch and perfusion cultivations. The according volume of preculture was then centrifuged at 180 g for 7 minutes in a Megafuge 1.0R (Heraeus, Germany) and the cell pellet resuspended in the batch medium. Sterile addition of the inoculum to the autoclaved bioreactor was done by adding the cell suspension to an inoculum flask connected to the bioreactor under the clean bench. 40 3.2 Methods 3.2.3.2 Fed-Batch Mode The cultivations in fed-batch mode were performed using the DASGIP four parallel bioreactor system and can be divided in three different process types explained later in this section. Starting volume for all processes was 1.0 L with an initial batch phase of 24 h. After that, the constant feed was started, so that the final volume after 14 days of cultivation was about 1.4 - 1.5 L. Glucose stock solution with 350 g/L was autoclaved to serve as bolus addition in case of glucose concentrations falling below 11 mM during cultivation. In this case, the glucose concentration was set to 22 mM by adding the according volume of glucose stock solution additional to the constant feed. Setpoints of temperature, pH, dissolved oxygen and agitation were the same for the three different process types and are listed in table 3.8. Prior to cultivation 1 M sodium carbonate solution was prepared as base and filtered for sterilization while CO2 addition in the ingas was used as acid. Elevated CO2 aeration was performed mainly in the beginning of the process because of the high initial medium pH of 7.2. Due to increasing cell densities and therefore CO2 and lactate formation by the cells, the pH controller switched to base addition while CO2 was held at the overlay value of 3 %. Table 3.8: Parameter setpoints Fed-Batch for the reference process (REF), the CO2 stressed process (COP) and the Process with no base addition (NOB) Process REF COP NOB Temperature [◦C] 36.5 36.5 36.5 pH < 48 h [-] 6.95 6.95 6.95 pH > 48 h [-] 6.80 6.80 6.80 Dissolved oxygen [%] 60 60 60 Agitation [rpm] 200 200 200 CO2 overlay [%] 3 3 3 The industrial reference process (REF) (provided by Boehringer Ingelheim AG Co. KG) was compared to two settings with altered pH control. While the parameters re- mained unchanged (table 3.8) the controller settings were changed in regard to the ref- erence (table 3.9). One process setting should depict the influence of increased CO2 stress (COP), therefore the maximum CO2 fraction in the ingas was increased to 15 % compared to 10 % in the reference. Moreover, the controller’s response was increased 41 Materials and Methods by changing the proportional factor and the reset time to 12.5 and 7200 s, respectively. The controller settings for the other process parameters remained the same as for the ref- erence. The third process setting (NOB) should investigate the effects of base addition, which was therefore suppressed completely resulting in a slow pH shift downwards. The remaining settings were similar to the reference. All three process settings were performed in biological triplicates. Table 3.9: Controller settings Fed-Batch for the reference process (REF), the CO2 stressed process (COP) and the Process with no base addition (NOB) Process REF COP NOB Base addition [-] Yes Yes No Max. fraction CO2 [%] for pH 10 15 10 Deadband for pH [-] 0.05 0.05 0.05 Proportional factor for pH [-] 10 12.5 10 Reset time for pH [s] 9000 7200 9000 Proportional factor for dissolved oxygen [-] 0.1 0.1 0.1 Reset time for dis- solved oxygen [s] 300 300 300 Min. flow ingas for dis- solved oxygen [L/h] 3 3 3 Max. flow ingas for dissolved oxygen [L/h] 18 18 18 Min. fraction O2 for dissolved oxygen [%] 50 50 50 Max. fraction O2 for dissolved oxygen [%] 100 100 100 Proportional factor for temperature [-] 15 15 15 Reset time for temper- ature [s] 1800 1800 1800 42 3.2 Methods 3.2.3.3 Perfusion Mode The perfusion cultivation was done in a single bioreactor of the fourfold system used for the fed-batch cultivations. The setup is depicted simplified in figure 3.1. As cell retention device a hollow fiber module CFP-4-E-3X2MA (GE Healthcare, Germany) with a membrane area of 230 cm2 was connected to the bioreactor. The pore size of the hollow fiber module was 0.45 µm, retaining the cells but no media components. The cells were continuously pumped in a loop between bioreactor and hollow fiber module using a 505s peristaltic pump (Watson Marlow, United Kingdom) with a flow rate of 150 mL/min. Therefore, a tangential flow filtration (TFF) was established to avoid clogging of the membrane surface. The harvest, containing the antibody but no cells, was withdrawn from the hollow fiber module with a 101u peristaltic pump (Watson Marlow, United Kingdom). Furthermore, bleeding of the bioreactor could be installed by taking out cell suspension directly from the bioreactor and replacing the volume with fresh medium. The parameter setpoints for the perfusion can be seen in table 3.10. In contrast to the fed-batch processes the pH setpoint was constantly 6.95 throughout the cultivation and agitation was increased to 250 rpm. The controller settings for the perfusion process were same as for the fed-batch reference process (table 3.9). Table 3.10: Parameter setpoints Perfusion Process Perfusion Temperature [◦C] 36.5 pH [-] 6.95 Dissolved oxygen [%] 60 Agitation [rpm] 250 CO2 overlay [%] 3 The liquid volume in the bioreactor was held constant at 1.0 L. In the first 24 h the cultivation was performed in batch mode with batch medium. After that, perfusion mode was started using again batch medium. In the following, three different subse- quent steady states were reached by changing the flow rates. As depicted in table 3.11 the first perfusion rate was 1.0 L/d, equivalent to 1 reactor volume per day. The bleed rate was kept low at 0.03 L/d to reach the first steady-state (SS1) with growth limita- 43 Materials and Methods Figure 3.1: Simplified scheme of the perfusion cultivation. The bioreactor was con- nected to a hollow fiber module and medium was continuously replaced. Cell-free har- vest was withdrawn through the pores of the hollow fiber module. tion due to glucose concentrations below 1 mM and thus an apparent glucose limitation. The according cell density was 15 x 106 cells/mL. For the second steady-state (SS2) the perfusion rate was increased to 1.25 L/d but still bleed rates were kept low, again resulting in glucose limitation. The final cell density for SS2 was 25 x 106 cells/mL. 44 3.2 Methods For the third steady-state (SS3) both perfusion and bleed rate were increased to 1.50 L/d and 0.25 L/d, respectively. Due to the rise in bleed rate the cell density was artifi- cially constrained at 20 x 106 cells/mL and not by glucose limitation. In SS3 medium mixed with glucose stock solution was used to replace the daily bleed to regain glucose concentrations of 35 mM after bleeding. The main difference between SS1 and SS2 on the one side and SS3 on the other side was the glucose limitation in the first two steady- states. Each steady-state was held for at least four days once constant cell densities were reached. Table 3.11: Perfusion settings and flowrates, modified after Becker, Junghans, Teleki, Bechmann, and Takors (2019b) Steady-state (SS) SS1 SS2 SS3 Viable cell density [106 cells/mL] 15 25 20 Glucose concentration [mM] < 1 < 1 > 5 Perfusion rate [L/d] 1.03 1.28 1.50 Bleed rate [L/d] 0.03 0.03 0.25 Harvest rate [L/d] 1.00 1.25 1.25 3.2.3.4 Extracellular Sampling The sampling for extracellular measurements was done once or twice daily. First, the sample port was flushed with 4 mL of cell suspension from the bioreactor and the sus- pension was discarded. Afterwards, 4 mL of sample were taken and distributed as follows: Two times 100 µL were pipetted into each 1900 µL 0.01 M potassium hy- droxide to shift the carbon equilibrium away from gaseous CO2 and prevent gassing out. This part was frozen at -20 ◦C and used for the determination of partial pressure of CO2 as described in section 3.2.4.7. Further 100 µL were used for the determination of cell density and viability as described in section 3.2.4.1. The remaining sample volume was centrifuged at 800 g and 4 ◦C for 5 min in a Megafuge 1.0R (Heraeus, Germany). 60 µL of the resulting supernatant were taken for the determination of glucose and lac- tate concentrations (section 3.2.4.2). The rest of the supernatant was frozen at -70 ◦C 45 Materials and Methods and thawed later for the quantification of the antibody (section 3.2.4.4) and amino acids (section 3.2.4.3). 3.2.3.5 Intracellular Sampling Intracellular samples were taken at four distinct time points of the fed-batch processes (early growth phase, middle growth phase, early stationary phase and early decline phase) and every second day for the perfusion process. The first step of the sampling procedure was a fast filtration approach (Matuszczyk, Teleki, Pfizenmaier, & Takors, 2015). Here, cell suspension containing 30x106 cells was taken from the reactor and medium was immediately discarded over a moistened glass fiber filter (Pall, USA), which was retaining the cells, while applying a vacuum of 30 - 60 mbar. This was followed by a washing step with ice cold washing buffer (section 3.1.6) and finally the freezing of the cells on the filter in liquid nitrogen and storage in a -70 ◦C freezer. For each sample the filtration was repeated three times. The method by Pfizenmaier et al. (2015) was applied to extract the metabolites us- ing a methanol chloroform extraction as well as evaporation steps. The extracts were then thawed again and used for the quantification of intracellular metabolites (sec- tion 3.2.4.5). 3.2.4 Analytics The following sections describe the analytical methods applied to analyze the metabolic changes due to process conditions. 3.2.4.1 Cell Density and Viability Determination of total cell density and viability was done using Cedex XS (Roche, Germany) with trypan blue staining. Herefore, the cell suspension was mixed 1:1 with a 0.4 % trypan blue solution (Sigma-Aldrich, USA) after sampling and then measured in triplicates. 46 3.2 Methods 3.2.4.2 Glucose and Lactate For measurement of glucose and lactate concentrations in bioreactor samples, 20 µL of the supernatant was pipetted into LaboTrace reaction vessels with buffer solution. After mixing, the concentrations were determined in triplicates in the LaboTrace an- alyzer (TraceAnalytics, Germany). If glucose and lactate concentrations were higher than 9 g/L or 2.7 g/L, respectively, dilutions in the buffer solution were increased ac- cordingly. 3.2.4.3 Amino Acids Extracellular concentrations of amino acids in the supernatant were determined by re- versed phase high performance liquid chromatography as described by Pfizenmaier et al. (2015). An Agilent 1200 HPLC (Agilent Technologies, USA) with a fluorescence detector was used for the quantification. Therefore, the wavelengths were set to 230 nm for excitation and 450 nm for emission. Built in were a Zorbax Eclipse Plus C18 guard column and a Zorbax Eclipse Plus C18 column (Agilent Technologies, USA). The gathered chromatograms were analyzed with the software Chemstation (Agilent Tech- nologies, USA) using standards (Sigma-Aldrich, USA) with seven concentration levels for the quantification of each amino acid. Prior to the measurement of samples, the frozen supernatant was thawed, centrifuged and diluted 1:25. Additionally, L-ornithine (Sigma-Aldrich, USA) was used as an internal standard. All measurements were con- ducted at 40 ◦C and a flow rate of 1.5 mL/min regarding the mobile phase. 3.2.4.4 Antibody The antibody titer in the supernatant of the bioreactor samples was determined with enzyme linked immunosorbent assay (ELISA). First, the anti-human IgG F(c) goat an- tibody (Biomol, Germany) was diluted 1:500 in coating buffer (section 3.1.6). Then the well plates (Greiner, Germany) were coated with the capture antibody and incubated for at least 1 h. After three washing steps with washing buffer (section 3.1.6), the blocking solution (section 3.1.6) containing bovine serum albumin (VWR, Germany) was used to block the non-specific binding sites for 0.5 h. Again, the plate was washed three times. Standards and samples , diluted in dilution buffer (section 3.1.6), were added to the plate in triplicates and incubated for 1 h. Once another threefold washing step was 47 Materials and Methods done, the detection antibody (Biomol, Germany) was diluted 1:90,000 in dilution buffer and added to the wells. After 1 h and five more washing steps, chemiluminescence solution Seramunblau (Seramun Diagnostica, Germany) was added to each well and in- cubated in the dark for 0.5 h. By adding 0.25 M sulfuric acid the reaction was stopped and a Systec microplate reader (BioTek, USA) was used to measure the absorbance at 450 nm and the reference at 620 nm. The concentration values were then calculated by using the correlation gained from the diluted standards. 3.2.4.5 Intracellular Metabolites Intracellular adenosine phosphates were quantified from intracellular cell extracts us- ing ion-pair reversed phase high performance liquid chromatography as described by Pfizenmaier et al. (2015). An Agilent 1200 HPLC (Agilent Technologies, USA) with a diode array detector was used for the measurements with a detection wavelength of 260 nm. Built in were a Hypersil BDS C18 guard column and a Supelcosil LC18-T col- umn (Sigma-Aldrich, USA). Elution buffers were prepared as described in section 3.1.6 and used as mobile phase during analysis. The resulting chromatograms were analyzed with the software Chemstation as for the amino acid analysis. Again, seven standard lev- els for AMP (Fluka, USA), ADP (Gerbu, Germany) and ATP (Gerbu, Germany) were used for calibration and final calculation of sample concentrations. All measurements were conducted at 30 ◦C and a flow rate of 1.0 mL/min regarding the mobile phase. The adenylate energy charge (EC) was calculated from the resulting AMP, ADP and ATP concentrations by dividing the sum of ATP and half of ADP concentrations by the sum of all adenosine phosphate concentrations: EC = cAT P + 0.5 · cADP cAT P + cADP + cAMP Intracellular quantifications of metabolites of glycolysis and tricarboxylic acid cycle were done with an Agilent 1200 HPLC system coupled with an Agilent 6410B quadrupole mass spectrometer with an electrospray ion source (Agilent Technologies, USA). Non-derivatized polar metabolites were quantified as described by Junghans et al. (2019) based on the method described by Teleki, Sánchez-Kopper, and Takors (2015). The ap- paratus was equipped with a Sequant ZIC-pHILIC column with guard column (Di2chrom, Germany). Metabolites were detected with high selectivity in the multiple reaction 48 3.2 Methods monitoring mode. Absolute quantification was done by isotope dilution using constant addition of U13C-labeled algal extracts (Sigma-Aldrich, USA) (Vielhauer, Zakhartsev, Horn, Takors, & Reuss, 2011) as internal standard. Additionally, external calibration was done. The measurements were performed at 40 ◦C and a flow rate of 0.2 mL/min. Alpha keto acids concentrations were determined with a newly developed method by Junghans et al. (2019). Therefore, derivatization steps were performed to conden- sate aldehyde and keto groups by phenylhydrazine. The method was adapted from Zimmermann, Sauer, and Zamboni (2014) to cope with LC-MS based quantification. The apparatus was equipped with a ZORBAX SB-C18 column with guard column (Ag- ilent Technologies, USA). Standards were added to quantify metabolites by comparing peak areas. The measurements were performed at 40 ◦C and a flow rate of 0.3 mL/min. All gathered data was analyzed with the software Masshunter B.05.00 (Agilent Tech- nologies) to calculate final intracellular metabolite concentrations. 3.2.4.6 Osmolality The osmolality for both cultivation media and buffers were measured using the freezing point depression method with an Osmomat 030 Osmometer (Gonotec, Germany). Be- fore, two point calibration was done by using Nanopure water as zero point and a 300 mOsm/kg or 500 mOsm/kg standard solution as second point. Samples were measured in triplicates. 3.2.4.7 Partial Pressure of CO2 The inorganic carbon content of the samples, diluted 1:20 in 0.01 M KOH, were deter- mined with a total carbon analyzer (Analytik Jena, Germany) according to Buchholz, Graf, Blombach, and Takors (2014). The samples were thawed and injected into the total inorganic carbon reactor where at first 10% ortho-phosphoric acid was added to shift the equilibrium and gas out the CO2. In the next step the remaining carbon was determined by combustion at 750 ◦C turning the carbon compounds to CO2. The CO2 from both acidification and combustion was quantified by infrared spectrometry. Stan- dards with defined inorganic carbon content coming from sodium carbonate were used for calibration of the system. All samples were replicated three times. A previously de- 49 Materials and Methods termined Hägg diagramm in combination with measured pH values during cultivation was used for the determination of dissolved CO2 concentrations. fCO2 = 1 1+ 10-pKS 10-pH The CO2 fraction fCO2 was calculated at each distinct pH using the acidic dissociation constant pKS. The fraction was then multiplied with the molar concentration of inor- ganic carbon to yield the dissolved CO2 concentrations. Finally, the henry coefficient of the medium, also determined in previous experiments, was used to calculate the partial pressure of CO2. 3.2.5 Cell specific Rates For all processes the extracellular cell-specific rates were calculated for growth and glu- cose, lactate, antibody and amino acids production or consumption. For the fed-batch processes cell densities and metabolite concentrations except glucose were fitted with splines using the open access Shape Language Modelling tool in Matlab Version 2013a (Mathworks, USA). All resulting functions were continuously derivable. As described by Wahrheit et al. (2014) no more than four splines were used for each concentration profile to avoid overfitting. Furthermore, only positive values were allowed. Glucose concentrations were not fitted since they were not continuously derivable due to the bo- lus additions. For the perfusion process cell specific rates and errors for the above mentioned metabo- lites were calculated in 24 h intervals from the last three days of each steady-state with- out spline fitting using mass balance equations: µ = B V + 1 XV · dXV dt The growth rate µ was calculated using B = bleed rate, V = liquid bioreactor vol- ume, XV = viable cell density in the bioreactor under the assumption of complete cell retention without cells in the harvest stream. 50 3.2 Methods qmAb = 1 XV · ( cmAb ·P V + dcmAb dt ) The cell specific antibody productivity was calculated with cmAb = antibody concen- tration in the bioreactor, P = perfusion rate under the assumption of the same antibody concentrations in bioreactor and harvest stream. qGlc,cons = 1 XV · ( (cGlc,m− cGlc) ·P V − dcGlc dt ) The cell specific glucose consumption rate was calculated with cGl