rights / license: research collection in copyright - non ...14495/... · impact of process...

Research Collection

Doctoral Thesis

Impact of Process Parameters on Cell Growth, Metabolism andAntibody Glycosylation

Author(s): Ivarsson, Marija

Publication Date: 2014

Permanent Link: https://doi.org/10.3929/ethz-a-010239626

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

https://doi.org/10.3929/ethz-a-010239626

http://rightsstatements.org/page/InC-NC/1.0/

https://www.research-collection.ethz.ch

https://www.research-collection.ethz.ch/terms-of-use

DISS. ETH NO. 22069

Impact of Process Parameters on Cell Growth,

Metabolism and Antibody Glycosylation

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH

(Dr. sc. ETH Zurich)

presented by

MARIJA IVARSSON

Dipl.-Ing. University of Natural Resources and Applied Life Sciences, Vienna

born on April 4th, 1986

citizen of Croatia

accepted on the recommendation of

Prof. Massimo Morbidelli, examiner

Prof. Markus Aebi, co-examiner

Dr. Miroslav Soos, co-examiner

2014

“The process is the product.”

I

Acknowledgements

My gratitude goes to Prof. Massimo Morbidelli who gave me the opportunity to work

in his group. I appreciate the freedom he gave me with my research. I wish to thank

him for his trust and the scientific guidance and support. Also I would like to thank

Prof. Markus Aebi for agreeing to be co-examiner, and him and his group, in particular

Dr. Robert Gauss and Chia-wei Lin, for useful scientific discussions.

From the scientific point of view, my greatest thanks go to Dr. Miroslav Soos for his

supervision over the last 4 years. I am impressed by his dedication to research and his

indefatigable patience. Likewise, I would like to thank all my colleagues from

upstream, in particular Benjamin Neunstöcklin, Thomas Villiger, Jochen Sieck, Daniel

Karst and Vania Bertrand for sharing their knowledge with me and for fruitful

discussions. Also my colleagues from downstream, especially Thomas Müller-Späth,

Martin Krättli, David Getaz, Bertrand de Neuville and Bertrand Guelat, who always

gave me the feeling of being part in their team and helped me solving technical

problems with my HPLC.

I am indebted to my Master thesis students Robin, Teja, Heeju and Tobias. My thanks

go to the Functional Genomic Center Zurich for MALDI TOF measurements and Dr.

Peter Gehrig for sharing his expertise with me. I would further like to thank Adriana

Primorac-Sliwa for her help with the real-time PCR measurements. I also acknowledge

the funding of the European Union in the frame of the OPTICO project and the

resulting collaboration with Prof. Mantalaris’ group from Imperial College London. I

particularly enjoyed the collaborative work with David Garcia-Münzer and Chon

Usaku.

Of course, a big thank you goes to all my colleagues, co-workers and lab mates from

the Morbidelli group. I spent a wonderful time in the group and we shared great

moments together, not only in the lab but also at conferences and outside the lab.

Thank you for providing me with an amazing working atmosphere, which made

coming to Hönggerberg every day something to look forward to. Especially, I would

like to thank Stefano Lazzari, Bertrand de Neuville, Fabio Codari, Paolo Arosio and

Benjamin Neunstöcklin, who, apart from being great colleagues, also became great

friends.

I am very grateful to my parents and my sister for their everlasting and doubtless

support. I thank them for the education that I received and for standing behind all my

decisions.

A very special thank you goes to my husband, Mattias, for his love and his support, the

patience and encouragement he gives me every day. I am very lucky to have you by

my side.

III

Abstract

Bioreactor process parameters influence the growth behavior and metabolism of

mammalian cell culture, and the quality of the produced monoclonal antibodies

(mAbs). A systematic assessment of their impact allows for a better link between

process-related parameters and cellular processes, such as N-linked glycosylation,

lactate metabolism and cell cycle transition. Thus, understanding the effect of process

parameters is a prerequisite for providing better controllability over cell growth,

productivity and mAb critical product quality.

Initially, we investigated the effect of single and combined process parameters on the

glycan microheterogeneity of an IgG1 antibody using a shift-experiment procedure in

batch culture of a hybridoma cell line. The N-linked glycosylation profile of the

murine IgG1 was found to be highly complex since it included terminal galactosylation

and sialylation, as well as variable core-fucosylation. Within a pH range of 6.8 to 8.0,

differences in galactosylation and sialylation of approximately 50% were obtained.

Variation of dissolved O2 (10 – 90% air saturation) resulted in a maximum variability

of 20% in galactosylation and 30% in sialylation. In contrast, no significant effect on

the glycosylation profile was observed when osmolarity increased from 320 to 420

mOsm/kg and sparging from 0.05 to 0.2 vvm.

Apart from its role in glycosylation, pH alteration during the early exponential growth

phase was found to control lactate formation and consumption. In particular, lactate

consumption was induced even at high glucose concentrations at pH 6.8, whereas

highly increased production of lactate was obtained at pH 7.8. Lactate accumulation in

mammalian cell culture is known to impede cellular growth and productivity.

Consequently, constraint-based metabolic flux analysis was used to examine pH-

induced metabolic states in the same growth state. We demonstrated that lactate influx

at pH 6.8 led cells to maintain high fluxes in the TCA cycle and malate-aspartate

shuttle resulting in a high ATP production rate. In contrast, under increased pH

conditions, less ATP was generated and different ATP sources were utilized. Gene

expression analysis led to the conclusion that lactate formation at high pH was enabled

by gluconeogenic pathways in addition to facilitated glucose uptake. The obtained

IV ABSTRACT

results provide new insights into the influence of pH on cellular metabolism, and are

of importance when considering pH heterogeneities typically present in large-scale

industrial bioreactors.

On industrial scale, fed-batch processes are mainly used for mammalian cell

cultivation and protein production. Cell productivity in fed-batch processes can be

increased by cell cycle arrest through mild hypothermia. Since hypothermia can

simultaneously reduce cell growth, which is regulated by the cell cycle, temperature

shifts have to be considered on the cell cycle level. Consequently, the time point for

the temperature shift is important and requires optimization. An unstructured cell cycle

model including the distribution of proliferating (G1, S, G2/M) and arrested cells (G0)

has been proposed in order to predict the time point of temperature shift in fed-batch

culture. The mathematical model is generally applicable and a procedure on how to

evaluate the required model parameters is described. The parameters are estimated

from batch and fed-batch cultivations, each carried out once at 37 °C and once at 33

°C in order to characterize temperature dependency. The batch cultures are also used

to evaluate substrate depletion and fed-batch cultures are used to study the impact of

metabolite accumulation on the cell cycle, in particular on the quiescent cell cycle

phase. The reliability of the proposed procedure for parameter estimation is validated

using a mAb-producing hybridoma cell culture and the model predicts hypothermic

transitions within the cell population at different shift time points. Thus, this

framework can be used to optimize the time point of the temperature shift, which is

commonly adjusted in industrial fed-batch processes in order to obtain a good balance

between temperature induced growth limitation and cell cycle specific enhanced

productivity.

In this work, a better understanding of bioprocess-related parameters affecting growth,

metabolism and critical quality attributes under the scope of quality by design (QbD)

is provided and can bring us one step closer towards stable process performance while

ensuring desired glycosylation for therapeutic proteins.

V

Zusammenfassung

Prozessparameter in Bioreaktoren bestimmen das Wachstumsverhalten sowie den

Stoffwechsel tierischer Zellen und beeinflussen gleichzeitig die Qualität der

produzierten monoklonalen Antikörper (mAb). Für eine bessere Verknüpfung von

prozessabhängigen Parametern und zellulären Prozessen, wie Glykosylierung,

Laktatstoffwechsel und Zellzyklus, bedarf es einer systematischen Untersuchung der

einzelnen Parametereinflüsse. Eine solche Untersuchung ist daher eine Voraussetzung

für die bessere Kontrolle des Zellwachstums, der Produktivität und der kritischen mAb

Qualtität.

In dieser Arbeit wurden HFN 7.1 Hybridomazellen zunächst im Batch-Modus

kultiviert und während der exponentiellen Wachstumsphase die Veränderung einzelner

sowie kombinierter Prozessparameter vorgenommen. Auf diese Weise konnten die

Parametereinflüsse auf die Verteilung der Glykanstrukturen eines IgG1 Moleküles

untersucht werden. Aufgrund terminaler Galaktosylierung and Sialylierung, sowie

variabler Fukosylierung, stellte sich das N-Glykosylierungprofil als äussert komplex

dar. Die Unterschiede in terminaler Galaktosylierung und Sialylierung innerhalb eines

pH-Bereiches von 6.8 bis 8.0 lagen bei ungefähr 50%, wohingegen die Variation des

Gelöstsauerstoffes (10 – 90% Luftsättigung) in einer Änderung von 20% in

Galaktosylierung und 30% in Sialylierung resultierte. Im Gegensatz dazu, führte eine

abrupte Veränderung der Osmolarität in einem Bereich zwischen 320 und 420

mOsm/kg, sowie eine Änderung der Begasungsraten von 0.05 bis 0.2 vvm zu keinem

Unterschied im Glykosylierungsprofil.

Neben dem signifikanten Einfluss auf die N-Glykosylierung, konnte gezeigt werden,

dass gezielte Veränderungen des pH Wertes während der früh-exponentiellen

Wachstumsphase zur Kontrolle der Laktatproduktion und der Laktataufnahme

verwendet werden können. Genauer gesagt, bewirkte eine Erniedrigung des pH Wertes

auf 6.8, selbst bei hoher Glukose Konzentration die Aufnahme von Laktat. Eine

Erhöhung des pH Wertes auf 7.8 führte hingegen zu einer deutlich gesteigerten

Laktatproduktion. In diesem Zusammenhang ist zu erwähnen, dass Laktat-

VI ZUSAMMENFASSUNG

Akkumulation in der tierischen Zellkultur für ein schlechteres Zellwachstum und eine

erniedrigte mAb Produktivität verantwortlich ist. Folglich wurden die durch pH

Modifikation hervorgerufenen Stoffwechselzustände mit Hilfe einer Stoffflussanalyse

untersucht. Die Berechnung der intrazellulären Flüsse zeigte deutlich, dass der

Einstrom von Laktat bei pH 6.8 zu hohen Flussraten innerhalb des Citratzyklus und

des Malat-Aspartat-Shuttles führte und vermehrt ATP produziert wurde. Bei hohen pH

Werten hingegen wurde unter der Verwendung von anderen Stoffwechselwegen,

verhältnismäßig weniger ATP generiert. Ferner, konnte durch eine

Genexpressionsanalyse aufgezeigt werden, dass die erhöhte Laktatproduktion bei

hohem pH Wert sowohl aufgrund der Aktivierung des gluconeogenetischen

Stoffwechselweges, als auch durch gesteigerte Glukoseaufnahme bedingt war. Diese

Ergebnisse liefern neue Einblicke in die Wirkungsweise von unterschiedlichen pH

Werten auf den Zellstoffwechsel und sind hinsichtlich der in industriellen

Produktionsbioreaktoren vorkommenden pH Heterogenitäten von Bedeutung.

Im industriellen Maßstab werden in der tierischen Zellkultur für die Proteinherstellung

überwiegend Fed-Batch-Prozesse eingesetzt. Die Kultivierung der in dieser Studie

eingesetzten Hybridomazellen im Fed-Batch-Modus, führte zu einer Verdoppelung der

Zellkonzentration, sowie des Produkttiters. Des weiteren wurde gezeigt, dass

Hypothermie während einer Fed-Batch Kultivierung Auswirkungen auf den Zellzyklus

und somit auf das Zellwachstum und die Produktivität haben kann. Zellwachstum und

Zellteilung sind Prozesse, die durch den Zellzyklus reguliert sind. Aus diesem Grund

wurde ein mathematisches Model des Zellzykluses entwickelt, welches die Verteilung

in proliferierende (G1, S, G2/M) und ruhende Zellen (G0) berücksichtigt. Hiermit

konnte der Einfluss von Temperaturänderungen von 37 °C auf 33 °C in Batch und

Fed-Batch Kulturen vorhergesagen werden. Das Model ist fähig hypothermische

Übergänge innerhalb einer Zellpopulation zu beschreiben und Auswirkungen einer

Temperaturveränderung zu unterschiedlichen Zeitpunkten vorherzusagen. Überdies

können Einflüsse von Substraterschöpfung und Metabolitakkumulation vorausgesagt

werden. Die lebende Zellkonzentrationen, der wesentliche Stoffwechsel sowie die

resultierenden Antikörperkonzentrationen in Batch und Fed-Batch wurden bei zwei

unterschiedlichen Temperaturen modelliert und experimentell validiert. Zielsetzung ist

eine Optimierung des Zeitpunktes, an dem eine Temperaturänderung vorzunehmen ist

um ein gutes Gleichgewicht zwischen temperaturbedingter Wachstumslimitation und

erhöhter zellzyklusspezifischer Produktivität zu erhalten.

Durch diese Arbeit konnte ein besseres Verständnis prozessabhängiger Parameter und

deren Auswirkung auf das Wachstum, den Stoffwechsel und die kritische

Produktqualität gewonnen werden. Diese Arbeit ist ein weiterer Schritt in Richtung

verbesserter Prozessstabilität mit dem Ziel die Glykosylierung von zukünftigen

therapeutische Proteinen besser kontrollieren zu können.

IX

Table of Contents

Acknowledgements .......................................................................................................... I

Abstract ......................................................................................................................... III

Zusammenfassung .......................................................................................................... V

1 Introduction ............................................................................................................. 1

1.1 Mammalian Cell Culture Technology ............................................................... 1

1.2 Operating Conditions in Bioreactor Systems .................................................... 3

1.3 Glycosylation of Monoclonal Antibodies ......................................................... 4

2 Evaluating the Impact of Cell Culture Process Parameters on Monoclonal

Antibody N-Glycosylation ............................................................................................. 7

2.1 Introduction ....................................................................................................... 7

2.2 Material and Methods ........................................................................................ 9

2.2.1 Cell Line and Cell Culture .......................................................................... 9

2.2.2 Monoclonal Antibody Purification ........................................................... 10

2.2.3 Calculations .............................................................................................. 11

2.3 Results and Discussion .................................................................................... 13

2.3.1 Process Parameter Shift Experiments ....................................................... 13

2.3.2 Growth Rate Dependency on Bioreactor Process Parameters .................. 14

2.3.3 Analysis of N-linked Glycan Microheterogeneity ................................... 17

2.3.4 Variation in N-linked Glycan Microheterogeneity .................................. 18

2.3.5 Impact of Chemical and Mechanical Stress Parameters on N-linked

Glycan Microheterogeneity ................................................................................... 21

2.4 Conclusions ..................................................................................................... 26

2.5 Remark ............................................................................................................ 26

3 Insights into pH-induced Metabolic Switch by Flux Balance Analysis ................ 27

X TABLE OF CONTENTS

3.1 Introduction ..................................................................................................... 27

3.2 Material and Methods ...................................................................................... 29

3.2.1 Cell Line and Media ................................................................................. 29

3.2.2 Bioreactor Setup and pH Shift Experiments ............................................ 29

3.2.3 TaqMan® Gene Expression Assay ........................................................... 30

3.2.4 Flux Balance Analysis (FBA) .................................................................. 31

3.3 Results and Discussion .................................................................................... 33

3.3.1 Characteristics of Cell Growth and Metabolism ...................................... 33

3.3.2 Metabolic Flux Distribution of Standard Cultivation during Exponential

Growth Phase ......................................................................................................... 37

3.3.3 Metabolic Flux Distribution in Lactate Production and Lactate

Consumption State ................................................................................................. 38

3.3.4 Metabolic Flux Distribution in Unfavorable Lactate Over-Production

State 40

3.3.5 Consequences of pH-Shifts on the Energy Metabolism ........................... 43

3.3.6 Relevance of pH-induced Metabolic States for the Design and Scale-up of

Industrial Bioprocesses .......................................................................................... 45

3.4 Conclusions ..................................................................................................... 46

3.5 Remark ............................................................................................................ 47

4 An Unstructured Model of Metabolite and Temperature Dependent Cell Cycle

Arrest in Hybridoma Batch and Fed-Batch Cultures .................................................... 49

4.1 Introduction ..................................................................................................... 49

4.2 Materials and Methods .................................................................................... 51

4.2.1 Cell line and Media................................................................................... 51

4.2.2 Shake Flask Batch Culture ....................................................................... 51

4.2.3 Fed-Batch Cell Culture ............................................................................. 52

TABLE OF CONTENTS XI

4.2.4 Cell Cycle Analysis .................................................................................. 53

4.2.5 Metabolic Profiling ................................................................................... 53

4.2.6 Extracellular Antibody Quantification ..................................................... 54

4.3 Modeling Aspects ............................................................................................ 54

4.3.1 Model Structure ........................................................................................ 54

4.3.2 Temperature Dependent Population Transition ....................................... 58

4.3.3 Global Sensitivity Analysis and Parameter Estimation ............................ 59

4.4 Results ............................................................................................................. 60

4.4.1 Global Sensitivity Analysis and Parameter Estimation ............................ 60

4.4.2 Temperature Shifted Batch Cultures ........................................................ 63

4.4.3 Temperature Shifted Fed-Batch Cultures ................................................. 66

4.4.4 Prediction of Temperature Shift Time in Fed-Batch Cultures ................. 70

4.5 Discussion ....................................................................................................... 74

4.6 Conclusions ..................................................................................................... 75

4.7 Remark ............................................................................................................ 76

5 Concluding Remarks ............................................................................................. 77

6 Outlook .................................................................................................................. 79

7 Appendix ............................................................................................................... 81

7.1 Appendix for “Evaluating the Impact of Cell Culture Process Parameters on

Monoclonal Antibody N-Glycosylation” .................................................................. 81

7.1.1 Deglycosylation by PNGase F and 2-AB Labeling .................................. 81

7.1.2 HPLC Separation of 2-AB Labeled Oligosaccharides ............................. 81

7.1.3 N-linked Oligosaccharide Analysis by Matrix-Assisted Laser

Desorption/Ionization Time-Of-Flight (MALDI-TOF) Mass Spectrometry (MS)82

7.2 Appendix for “Insights into pH induced Metabolic Switch by Flux Balance

Analysis” ................................................................................................................... 86

XII TABLE OF CONTENTS

7.2.1 Gene Expression Analysis ........................................................................ 86

7.2.2 Metabolic Network for Flux Balance Analysis ........................................ 88

7.2.3 Calculated Intracellular Fluxes from Flux Balance Model ...................... 96

7.3 Appendix for “An Unstructured Model of Metabolite and Temperature

Dependent Cell Cycle Arrest in Hybridoma Batch and Fed-Batch Cultures” .......... 99

7.3.1 Model Parameter Estimation .................................................................... 99

8 References ........................................................................................................... 101

9 Figure and Table Index ........................................................................................ 123

9.1 List of Figures ............................................................................................... 123

9.2 List of Tables ................................................................................................. 127

1

1 Introduction

Around 30% of all drugs approved by the European Medical Agency (EMA) up to

2012 were biologically-derived. Among them glycoproteins are the largest and fastest

growing group [1]. Mammalian cells are the predominant production system of

glycoproteins due to their ability to express large proteins and perform post-

translational modifications, such as glycosylation [2]. Therapeutic glycoproteins

require glycosylation for their biological function, and mammalian cells possess the

machinery to provide proper glycosylation that is human-like and therefore shows a

reduced risk of immunogenicity [3]. Only recent advances in yeast and plant

technology have led to the first approvals of a P. pastoris-derived glycoprotein [4],

and a genetically modified plant cell-derived glycoprotein [5] by the US Food and

Drug Administration (FDA). Nonetheless, the use of mammalian cell culture for the

production of complex, large, recombinant proteins for therapeutic or diagnostic

purpose remains unparalleled to date.

1.1 Mammalian Cell Culture Technology

Mammalian cell culture technology has evolved dramatically since its first use for

large-scale vaccine production in the 1960s [6]. The biggest advances in mammalian

cell culture technology included the establishment of continuous, immortal cell lines

that could be grown in suspension in contrast to earlier adherent primary cell lines [7].

Thus, immortal cells, such as baby hamster kidney cells (BHK) for veterinary use, and

later Chinese hamster ovary (CHO), myeloma (Sp2/0, NS0) and human embryonic

kidney (HEK) cells, facilitated the production in large-scale bioreactor systems. Due to

the simpler scale up by volume instead of surface, reactor scales could be extended to

10 000 L systems, such as for the production of interferon alpha in the 1980s [8]. In

addition, immortal cell lines could be genetically modified with emerging recombinant

DNA technology. As a result, tissue-type plasminogen activator protein was the first

commercial recombinant protein produced in a CHO cell line [9].

Over the years a variety of recombinant proteins was generated from mammalian cell

culture such as cytokines, hormones, growth factors, clotting factors, vaccines, fusion

2 CHAPTER 1: INTRODUCTION

proteins and monoclonal antibodies (mAbs) [10], [11]. Since the first approval of a

monoclonal antibody drug in 1987 [12], therapeutic mAbs mainly used for the

treatment of cancer, anti-immune and inflammatory diseases, have witnessed an

incredible growth. Nowadays, they represent the most important drug product class

with currently 40 FDA-approved mAbs, and annual sales of around $ 50 billion [13],

which represents around 40% of the total biologically-derived drug market [11]. The

increase in monoclonal antibody drugs also fostered a corresponding improvement of

the production processes to meet the high product demand of several tons/year [14].

Such a development was achieved by the generation of high-producer cell lines,

mostly CHO cells, using optimal gene amplification systems [15] and by including

growth-promoting genes [16]. Such genetic engineering approaches resulted in a 2-

fold increase in specific productivity [17]. More interestingly, a 20-fold increase in

volumetric productivity could be obtained in the same time frame through

optimization of bioprocess conditions including the control of the culture environment

and the development of chemically defined media and optimized feeding techniques

[7], [17]. As a consequence, newly developed fed-batch production processes are

reported to produce mAb titers of more than 10 g/L [15], [18].

Considering the recent emergence of biosimilar drugs [10], the focus in mammalian

cell culture technology and bioprocess development is shifting increasingly towards

matching given product attributes, rather than aiming exclusively at increased growth

and productivity. Small differences in the three-dimensional structure, charge variants

profile or post-translational modifications of the protein can arise from minor

variations in the production process and can result in differences in drug efficacy in

comparison to the originator drug [19], [20]. Considering post-translational

modifications, different levels of core-fucosylation in the glycosylation pattern have

been shown to have severe consequences on the effector function of a mAb [21]. Thus

the process as a whole has to be put into focus in order to qualify for biosimilarity

[10]. Moreover, methods that can link individual process steps to product quality

require development.

CHAPTER 1: INTRODUCTION 3

1.2 Operating Conditions in Bioreactor Systems

Mammalian cells for the production of recombinant proteins are typically cultured in

stirred tank bioreactors in a scale from 10 L to 20 000 L. Such reactors require a

suitable design to provide proper mixing and mass transfer of oxygen and carbon

dioxide (CO2) [22]. Physical control parameters are the inlet gas flow rate, stirring

speed and temperature, whereas pH, osmolarity, dissolved oxygen (DO) and dissolved

C02 are considered chemical control parameters. Process parameters are used to define

the optimal culture environment that is intended to provide homogenous conditions to

the cultured cells. Sensors and control systems are used to monitor and adjust the

culture environment. pH is controlled through CO2 sparging and base addition, while

constant dissolved oxygen levels are provided by sparging air or oxygen into the

culture at constant gas flow rates. While initial limitations regarding shear stress

sensitivity of mammalian cells by agitation and aeration have been mostly overcome

[23–25], other process parameters are still known to cause variations in growth,

productivity and product quality, in particular during scale-up. pH heterogeneity can

arise from base addition in poorly mixed bioreactors [26]. Poor mixing can possibly

also lead to gradients in temperature and osmolarity due to the addition of base and

feed close to the surface [27]. Insufficient CO2 removal in large-scale bioreactors is

likewise a common problem and is connected to low agitation rates, insufficient gas

transfer and small bubble size [28]. Since high levels of dissolved CO2 negatively

influence cell growth and product quality [27], [29], [30], the presence of different

dissolved CO2 levels in different reactor scales has to be avoided. In this context,

variability in the culture environment and thus in single process parameters should be

linked to cellular processes that are related to growth, productivity and product quality.

By understanding the impact of relevant parameter ranges it should be possible to

better provide optimal operating conditions at all bioreactor scales.

Apart from process parameter control, improvement of the culture media played an

important role in the development of cell culture processes. The adaptation of

mammalian cells to serum-free conditions and the replacement of serum proteins by

chemically defined, protein- and peptide-free media increased process reproducibility,

facilitated downstream processing, excluded the risk of contamination by animal-


derived components and led to a general cost reduction [22], [28]. The main

components of chemically defined, commercial or proprietary media are glucose,

amino acids, vitamins, trace elements, inorganic salts and lipids. Determining the

optimal ratio of components requires systematic studies using different available

techniques [31] and is often dependent on the cell line. Likewise the composition of

feed solutions and the feeding strategies need to be investigated with the additional

aim of reducing by-product accumulation (lactate and ammonia). Optimization of

media and feeding strategies has proved to have a dramatic benefit on culture

longevity and specific productivity. More recently, control of glycosylation has also

been achieved by the addition of certain media components [32–37].

1.3 Glycosylation of Monoclonal Antibodies

Monoclonal antibodies of the immunoglobulin G (IgG) isotype are the most common

therapeutic antibodies [38]. An IgG molecule consists of 4 polypeptide chains (2

heavy chains and 2 light chains) that are bound covalently through disulfide bridges.

The target specificity, and thus the antigen binding, of IgG arises from the

hypervariable region formed by the variable domains of the heavy and the light chain.

Two constant domains of the heavy chains form the crystallizable fragment (Fc)-

region that is responsible for the immune effector functions [39]. Activation of

immune effector functions such as antibody-dependent cellular cytotoxicity (ADCC)

or complement-dependent cytotoxicity (CDC) is mostly dependent on the binding to

the FcγRIII receptor and to the C1q component of complement, and triggers a cascade

that ultimately leads to the death of target cells [40]. This mode of action is used by

monoclonal antibodies for treatment of cancer and infectious diseases [41]. Instead,

binding of IgG-Fc to the FcγRIIb receptor is exploited for the treatment of

autoimmune diseases as it leads to a modulation of the immune response [42], [43].

The binding of the Fc region to Fc receptors or the C1q component of complement is

critically dependent on N-linked glycosylation [3], [44], [45]. N-linked glycosylation

of therapeutic monoclonal antibodies implies the presence of oligosaccharide

structures, so called glycans, in the Fc region of the IgG molecule [3]. Glycosylation

of the Fab-fragment which has an involvement in antigen binding affinity [46] and

CHAPTER 1: INTRODUCTION 5

antibody half-life [47], is less common for therapeutic antibodies but has also been

reported [48].

Monoclonal antibodies were initially produced in hybridoma cells formed by the

fusion of a mouse B cell and a human cancer cell [49]. The resulting murine antibodies

are specific to a single antigen, but have the disadvantage of carrying murine-type

glycosylation, which can induce immune reactions in humans. This stimulated further

developments including the generation of chimeric mouse-human antibodies and

humanized antibodies that could reduce immunogenicity [50]. Later, the development

of transgenic mice [51], [52] as well as the use of phage display technology [53]

allowed the production of fully human antibodies, which have replaced hybridoma

cells for monoclonal antibody generation. Nowadays, monoclonal antibodies are

further engineered to overcome immunogenicity, improve their biological function and

their biophysical properties such as solubility and half-life [54–58]. At the same time

bispecific antibodies, as well as antibody-drug conjugates have been developed and are

reviewed elsewhere [59], [60].

In this work, we have cultivated HFN 7.1 hybridoma cells in batch and fed-batch

culture. HFN 7.1 cells produce a murine IgG1 antibody against human fibronectin,

commonly used in diagnostics. The antibody is characterized by high oligosaccharide

content and represents a good model protein for the study of antibody glycosylation. In

Chapter 2, we present a methodology to study the influence of different bioreactor

process parameter ranges on glycan microheterogeneity, in particular on terminal

galactosylation, sialylation and core-fucosylation. The degrees of glycosylation

obtained within relevant process parameter ranges are compared to simultaneous

variations of multiple process parameters and to literature. The study revealed the

importance of pH for IgG glycosylation.

In Chapter 3, the impact of pH alterations on the cellular metabolism was further

investigated. We have identified high pH to cause overexpression of metabolic

enzymes, which results in changes in cell metabolism, in particular the lactate

metabolism. Additionally, high pH showed negative effects on cell growth and

antibody production, whereas low pH improved antibody productivity. At the same


time, low pH resulted in reduced lactate accumulation. Flux balance analysis was

applied to give more insight into the redistribution of the main energy metabolism at

different pH values that mimic heterogeneities in large-scale bioreactor systems.

Apart from the implication of process parameters in scale-up, they can be actively used

to optimize process conditions. In this context, hypothermia is known to prolong

culture longevity through its action on the cell cycle. In Chapter 4, an unstructured

model based on cell cycle transition and quiescence that was developed and employed

to predict temperature-shifts in fed-batch culture is presented. At the same time,

metabolic cell cycle arrest, induced by nutrient depletion and production of toxic

metabolites was evaluated and included in the model. The model was able to capture

growth, basic metabolism and mAb production characteristics. Following the

described procedure it can be employed to optimize the temperature shift time point in

order to obtain a good balance between cell growth and productivity.

7

2 Evaluating the Impact of Cell Culture Process

Parameters on Monoclonal Antibody

N-Glycosylation

2.1 Introduction

The relevance of monoclonal antibodies as therapeutic agents against cancer,

autoimmune and inflammatory diseases has been widely recognized. Most of the

mAbs approved to date are based on the immunoglobulin (IgG) isotype, carrying a

consensus N-linked glycosylation site on the CH2 domain of each heavy chain [61]

and further possible, but less common, N-linked glycosylation sites on the antigen-

binding fragment (Fab) [48]. Glycosylation is a complex process of oligosaccharide

attachment to the polypeptide backbone of a protein taking place in the endoplasmic

reticulum (ER) and Golgi apparatus. Presence or absence of certain oligosaccharides

can critically impact mAb stability [62], effector functions [38], immunogenicity [63],

[64], and clearance rate[65], [66], through a clearly defined structure/function

relationship. More precisely, reduced terminal galactosylation decreases complement-

dependent cytotoxicity (CDC) [67–69], absence of core-fucosylation results in

increased antibody-dependent cytotoxicity (ADCC) [70–72], and high sialylation

levels reduce ADCC activity and impact inflammatory responses [73], [74].

Furthermore, the use of CHO cells, mouse NS0 cells or Sp2/0 mouse myeloma cells

for mAb production can introduce non-human epitopes such as galactose-alpha-1,3-

galactose (α-gal) and N-glycolylneuraminic acid (NGNA) residues that act

immunogenically [75], [76].

Moreover, it is understood that variations in the N-linked glycan profile

(microheterogeneity) can occur during the mAb production process and therefore

glycosylation is considered a critical quality attribute of the end product. Studying the

impact of various environmental factors and process conditions can enlarge the

understanding of the glycosylation machinery and at the same time provide potential to

control and target desired mAb glycoforms. Multiple cell culture conditions during

8 CHAPTER 2: ANTIBODY GLYCOSYLATION

upstream processing such as different feeding regimes [32–35], media

supplementation [36], [37], as well as waste metabolite accumulation [77–79] can

affect glycosylation. Bo et al. (2013) recently showed a reduction in terminal

galactosylation for a chimeric heavy chain antibody at limiting glucose concentrations

in fed-batch culture. Moreover, supplementation with manganese chloride, galactose

and uridine can provide control over galactosylation during fed-batch cultivation [80],

[81]. Apart from metabolic control, bioreactor process parameters can lead to

variability in the glycoform profile and thus, play an important role in manipulating

the product quality. pH, temperature, dissolved O2 and CO2 levels, and shear stress

through agitation and sparging are classical process parameters. On the one hand they

can be applied to actively control the glycan microheterogeneity; on the other hand

such parameters are partially bioreactor scale-dependent and can inadvertently lead to

inconsistent glycosylation during scale-up. The impact of process parameters on N-

linked glycosylation has been summarized in literature [82], [83], however the effects

shown are often incoherent and incomplete, and therefore do not allow general

conclusions. This is partly due to the fact that glycosylation is dependent on the cell

line [84], [85] and the structure of the glycoprotein [86], but it is also influenced by the

experimental methodology that is used to study process-related impact factors.

In this work, we systematically investigated multiple process parameters using shift-

experiments in batch culture which allowed us to study each process parameter

individually. We focused on pH, dissolved O2 (DO) and osmolarity as chemical stress

parameters and sparging as a mechanical stress parameter assessing their impact on

cellular level and protein quality level. In addition, we compared the impact of single

parameters to combined effects (pH/ osmolarity). Using hydrophilic interaction

chromatography (HILIC) and MALDI TOF mass spectrometry we were able to

identify the impact of process parameters on N-linked glycosylation of a murine IgG1.

The present study provides a systematic procedure carried out in controlled

bioreactors, thus minimizing interaction effects, which can help to identify most

profound process-related factors affecting mAb glycosylation. Apart from commonly

assessed variations in terminal galactosylation, sialylation and core-fucosylation, less

common structures carrying α-gal and NGNA residues were investigated.

CHAPTER 2: ANTIBODY GLYCOSYLATION 9

2.2 Material and Methods

2.2.1 Cell Line and Cell Culture

The murine hybridoma cell line HFN 7.1 [87], producing an immunoglobulin G1

(IgG1) antibody against human fibronectin was obtained from the American Type

Culture Collection (ATCC CRL-1606) and adapted to the protein- and peptide-free

culture media Turbodoma® TP6 (Cell Culture Technologies) supplemented with 4.5

g/L D-glucose, 4 mmol/L L-glutamine and 0.1% (w/v) pluronic F-68.

Cells were expanded for 14 days in suspension in a humidified atmosphere (5% CO2)

at 37 °C and cultivated in a DasGip bioreactor system (DasGip) equipped with a

pitched-blade impeller and a porous sparger (10 µm). Exponentially growing cells

from the expansion were inoculated into the reactor at a seeding concentration of 0.6 x

106 cells/ mL. Standard batch cultures were carried out at a working volume of 1 L,

temperature equal to 37 °C, dissolved O2 (DO) set to 50% air saturation, stirring speed

of 150 rpm and an aeration rate of 0.05 vvm. pH was controlled at 7.2 with CO2

sparging.

A shift in one of the process parameters pH, DO, sparging or osmolarity was

performed during batch cultivation once the viable cell concentration reached 1.5 x 106

cells/mL. After the process parameter was shifted to the new setpoint it was kept

constant until the end of the cultivation. pH was shifted by base (1 M sodium

hydroxide) or acid (1 M hydrochloric acid) addition. During the shift, the pH

controller was turned off and only after the new setpoint was reached the pH was again

controlled by base addition and CO2 sparging. The increase in osmolarity due to acid

or base addition was below 5% and this was considered to have a negligible effect on

cells. The DO shift was performed by changing the inlet gas composition through the

DO controller. The time needed for the adjustment of the new pH and DO setpoints

was in the range of a few minutes. The osmolarity shift was performed by addition of

6 M sodium chloride (NaCl) and monitored by offline measurements using an

OsmoLab One osmometer (LLA Instruments GmbH). The sparging shift was

performed by changing the gas flow rate in the controlled bioreactor system.

Perturbations of the system, in particular DO and pH, due to the new gas flow rate


were negligible. In this manner an operating range of pH from 6.5 to 8.5, DO from

10% to 90% air saturation, sparging from 0.05 vvm to 0.2 vvm and osmolarity from

320 mOsm/kg to 450 mOsm/kg was investigated. Moreover, the combined effect of

pH and osmolarity was studied in cross experiments by performing a shift of both

parameters at the same time point following the same procedure as described above.

Cell culture samples were taken twice a day and in addition right after the parameter

shift. Cell number and viability were determined by the trypan blue exclusion method

[88] using a CedeX cell counter (Innovatis). Glucose and lactate concentrations were

determined enzymatically using a Super GL compact instrument (Hitado). Ammonium

and glutamine concentrations were determined using the L-Glutamine/Ammonia

(Rapid) Assay Kit (Megazyme). Samples were spun down and supernatants were

stored at -20 °C and later used for measurement of antibody concentration (PA

ImmunoDetection® Sensor Cartridge, Applied Biosystems) and N-linked

glycosylation analysis.

Profiles of the viable cell concentration were calculated using a generalized logistic

fitting equation containing 4 non-negative model parameters A, B, C and D (Eq. 2-1)

[89]. The specific growth rate µ was calculated by differentiating the fitted viable cell

concentration profile X according to Eq. 2-2.

X =A

exp(B𝑑) + C exp(-Dt) (2-1)

µ =

1𝑉

d𝑉d𝑑

(2-2)

2.2.2 Monoclonal Antibody Purification

200 µg of mAb was purified from the supernatant of each batch cultivation using the

Vivapure® miniprepG purification kit (Sartorius Stedim Biotech) according to the

manufacturer’s protocol. The kit contains pre-packed Protein G resin plugs and utilizes

affinity purification to separate the monoclonal antibody from the supernatant. Protein


G affinity purification was used for the mouse IgG1, as it only showed weak binding

affinities to Protein A. The purified monoclonal antibody was concentrated to 20 µL

with a Vivaspin® 500 (30 kDa) centrifugation filter (Sartorius Stedim Biotech) at

1800 g (4 °C) and used for analysis of the glycosylation profile as described in the

Appendix (6.1).

2.2.3 Calculations

Monoclonal antibody samples were taken from each batch culture at the end of

cultivation (tend) and prior to the parameter shift (tshift). Glycosylation profiles were

determined at both time points by HPLC. The relative peak area (%Area) of the

detected glycoforms was extracted from the HPLC analysis and was equal to the mole

fraction xi of each species. The aim was to identify the change in glycosylation of each

species xi produced between the shift time point (tshift) and the end time point (tend). In

order to account for differences in cell productivity under different operating

conditions, xi was normalized to the difference in mAb concentration (cmAb) between

the two time points, calculated using the molar mass of IgG1 of 150 kDa. Assuming a

constant macroheterogeneity during the culture, the change in each glycoform fraction

∆Fi was calculated according to Eq. 2-3.

∆𝐹𝑖 =𝑐𝑚𝐴𝑏,𝑒𝑛𝑑 × 𝑥𝑖,𝑒𝑛𝑑 − 𝑐𝑚𝐴𝑏,𝑠ℎ𝑖𝑓𝑡 × 𝑥𝑖,𝑠ℎ𝑖𝑓𝑡

𝑐𝑚𝐴𝑏,𝑒𝑛𝑑 − 𝑐𝑚𝐴𝑏,𝑠ℎ𝑖𝑓𝑡 (2-3)

The degree of galactosylation (GI) [90] was calculated according to Eq. 2-4 as the

fraction of tri-galactosylated (G3), di-galactosylated (G2), mono-galactosylated (G1)

and non-galactosylated (G0) structures:

𝐺𝐼 =3 × 𝐺3 + 2 × 𝐺2 + 𝐺1

(𝐺0 + 𝐺1 + 𝐺2 + 𝐺3) × 3 (2-4)

In the same manner the sialylation (SI) [90] and fucosylation (FI) indices were

calculated (Eq. 2-5 and Eq. 2-6), where only di-sialylated (S2), mono-sialylated (S1)


and non-sialylated (S0) structures, and mono-fucosylated (F1) and non-fucosylated

(F0) structures respectively, were observed:

𝑆𝐼 =2 × 𝑆2 + 𝑆1

(𝑆0 + 𝑆1 + 𝑆2) × 2 (2-5)

𝐹𝐼 =𝐹1

𝐹0 + 𝐹1 (2-6)


2.3 Results and Discussion

2.3.1 Process Parameter Shift Experiments

Murine hybridoma cells (ATCC CRL 1606) were adapted to a commercial, protein-

and peptide-free media (Turbodoma® TP6) and showed a similar growth and

metabolic behavior as when cultivated with zinc supplemented IMDM media as

reported elsewhere [91]. In

Figure 2-1 a typical batch culture averaged from four independent experiments is

shown. The maximum viable cell concentration of 3.1 (± 0.2) x 106 cells/mL was

reached after 54 hours (Figure 2-1A) coinciding with glutamine depletion (Figure

2-1D), which led to a stop in glucose consumption (Figure 2-1B). The culture was

continued until 80 hours and reached a final monoclonal antibody concentration of

98.9 (± 7.2) mg/L (Figure 2-1C). The production of by-products (lactate and ammonia)

did not exceed growth-inhibiting concentrations of 40 mM for lactate and 4 mM for

ammonia [92] (Figure 2-1B, D).

A shift in one of the investigated process parameters (pH, osmolarity, dissolved O2 and

sparging) was performed during the early exponential growth phase of a standard

batch culture in order to systematically study the impact of chemical and mechanical

stress parameters on cell growth, productivity and N-linked glycosylation. Due to the

controlled bioreactor environment, one parameter at a time could be shifted and

subsequently kept constant until the end of the cultivation.

For illustration, Figure 2-2 shows the influence of pH on viable cell concentration,

viability and monoclonal antibody concentration in 7 shift-experiments. Each

experiment was started at standard pH 7.2 and shifted at the indicated time point

(dashed line). While under standard conditions the culture reached the highest viable

cell concentration, all pH shifts led to reduced growth profiles (Figure 2-2A). In the

case of pH 6.5 and 8.5 the culture immediately entered death phase, as evident from

the viability profiles (Figure 2-2B). In comparison, a shift to pH 6.8 resulted in a

beneficial behavior regarding antibody concentration, with an 11% increase in final

mAb concentration compared to standard conditions (Figure 2-2C).


Due to the complexity of the shift experiments described above, in the following we

will analyze the corresponding results in a more synthetic way.

Figure 2-1: Standard batch cultivation of murine hybridoma cells in a controlled

bioreactor system (n=4). Profile of (A) viable cell concentration, (B) glucose ()

and lactate (), (C) monoclonal antibody and (D) glutamine () and ammonia

() concentrations.

2.3.2 Growth Rate Dependency on Bioreactor Process Parameters

The specific growth rate µ was calculated during the exponential growth phase of the

batch culture 10 hours after the respective parameter shift according to Eq.2-2. As

evident from Figure 2-3, the highest growth rate was reached under standard

conditions. Consequently, this value was used for normalizing all other specific

growth rate values. The most severe reduction of the specific growth rate was induced

with a shift in pH. A drop to pH 6.8 resulted in 32% reduction of the specific growth

rate whereas an increase to pH 8.0 led to a reduction of 53% (Figure 2-3).

0 10 20 30 40 50 60 70 80 900.00.51.01.52.02.53.03.5

0 10 20 30 40 50 60 70 80 908

12

16

20

24

0 10 20 30 40 50 60 70 80 90020406080

100120

0 10 20 30 40 50 60 70 80 900

1

2

3

4DC

B

Viab

le Ce

ll Co

nc. (

106 ce

ll/m

L)

Time (h)

Gluc

ose C

on. (

mm

ol/L

)

Time (h)

681012141618

Lacta

te Co

nc. (

mm

ol/L

)

mAb

Con

c. (m

g/L)

Time (h)

A

Glut

amin

e Con

c. (m

mol

/L)

Time (h)

0

1

2

3

4

Amm

oniu

m C

onc.

(mm

ol/L

)


Figure 2-2: pH shift experiments during batch cultivation of murine hybridoma

cells. Starting at standard pH condition 7.2 a pH shift was performed in the early

exponential phase indicated by a dashed line and covered a pH range from 6.5 –

8.5. Profile of (A) viable cell concentration, (B) viability and (C) mAb

concentration at different pH setpoints.

0 10 20 30 40 50 60 70 80 900.00.51.01.52.02.53.03.5

0 10 20 30 40 50 60 70 80 9020

40

60

80

100

0 10 20 30 40 50 60 70 80 90020406080

100120140

pH 6.5 pH 6.8 pH 7.0 pH 7.2 pH 7.5 pH 7.8 pH 8.0 pH 8.5

Viab

le Ce

ll Co

nc. (

106 ce

ll/m

L)

Time (h)

Viab

ility

(%)

Time (h)C

B

A

mAb

Con

cent

ratio

n (m

g/L)

Time (h)


Exceeding these limits by further decreasing or increasing pH (pH 6.5 and pH 8.5) led

to immediate cell death (Figure 2-2A) and is therefore excluded from the analysis.

Furthermore, an increase in osmolarity to 420 mOsm/kg resulted in 45% reduction in µ

and is therefore ranked as the second most important parameter affecting cell growth.

Similarly to the effect with pH shifts, a further increase in osmolarity to 450 mOsm/kg

resulted in immediate cell death (data not shown).

Figure 2-3. Impact of process parameter shifts (pH, osmolarity, dissolved O2 and

sparging) on the specific growth rate 10 hours after the parameter shift. The

specific growth rate was calculated for each condition according to Eq. 2-2 and

normalized to the specific growth rate obtained under standard conditions. The

standard deviation was calculated from four independent standard cultivations.

The two process parameters dissolved O2 (DO) and sparging gave a maximum growth

rate reduction of 27% and 15% (Figure 2-3), therefore ranking 3rd and 4th respectively.

Since the investigated range of DO and sparging already accounted for the entire

6.8 7.0 7.2 7.4 7.6 7.8 8.00.0

0.2

0.4

0.6

0.8

1.0

320 340 360 380 400 4200.0

0.2

0.4

0.6

0.8

1.0

10 20 30 40 50 60 70 80 900.0

0.2

0.4

0.6

0.8

1.0

0.05 0.10 0.15 0.200.0

0.2

0.4

0.6

0.8

1.0

pH (-)

Osmolarity (mOsm/kg)

Norm

alize

d Gr

owth

Rate

(-)

Dissolved O2 (%)

Sparging (vvm)


operating range in common mammalian cell bioprocesses [93], a further extension was

not considered.

Growth rate dependencies were determined in order to define the parameter ranges in

which the change in glycosylation should be investigated. In these parameter ranges,

cells are in the growth state and produce IgG1, crucial for the subsequent analysis of

glycosylation which focuses on glycoforms produced between the time of the

parameter shift and the end of the cultivation. The parameter range could vary

depending on the cell line and the experimental procedure used to introduce

environmental perturbations. It is worth noting that the obtained specific growth rate

dependency was comparable with previous literature reports for pH [93–96],

osmolarity [93], [97], hyperoxia [98] as well as sparging [96], [99].

2.3.3 Analysis of N-linked Glycan Microheterogeneity

In order to evaluate the differences in the glycosylation profiles (microheterogeneity)

induced by process parameter shifts, oligosaccharides were enzymatically released by

PNGase F from affinity-purified IgG1 antibody, fluorescently labeled with 2-AB and

characterized by MALDI TOF mass spectrometry. All determined oligosaccharide

masses obtained by MALDI-TOF MS in reflectron positive and negative ion mode

using DHB matrix are shown in the Appendix (Figure A 1). In positive ion mode,

oligosaccharides formed sodium or potassium adducts with sodium adducts having the

higher intensity. Structures of oligosaccharides were assigned either through database

comparison (GlycoWorkBench2) [100] or by comparison to the glycosylation profile

of Cetuximab, a commercially available antibody with known glycosylation profile,

obtained by hydrophilic interaction chromatography (HILIC) [48]. In some cases,

structures were further determined from the fragmentation pattern obtained by

MALDI-TOF MS/MS, as shown in the Appendix (Figure A 1C) for a core-

fucosylated, bi-antennary structure carrying one terminal galactose (G1F). All clearly

identifiable glycan structures are of the complex type containing terminal galactose, α-

linked galactose and/or sialic acid residues. Both types of terminal sialic acid residues

(NGNA and NANA) were detected, whereas terminal NGNA residues were

predominant. The proposed glycoforms are given in the Appendix (Table A 1).


In addition, fluorescently labeled glycoforms were separated by HILIC

chromatography and the mass of each well separated fraction was likewise determined

by MALDI-TOF MS and compared to the characterized oligosaccharide structures

given in the Appendix (Table A 1). The glycan structures could subsequently be

assigned to each well separated fraction in the HILIC glycosylation profile (Figure

2-4). The glycosylation profile demonstrated the separation of oligosaccharides

between 35 and 84 minutes. The profile is characterized by three major structures, one

eluting after 39 and two structures co-eluting after 46 minutes. These three glycan

structures contributed to 57% of the whole glycosylation profile. Through MALDI-

TOF MS/MS, it was found that the major glycans represent core-fucosylated, bi-

antennary structures with none or one terminal galactose, G0F and G1F respectively,

where the two co-eluting glycans are assumed to be G1F carrying the galactose either

on the α(1→6) or the α(1→3) arm. All characterized glycan structures that could be

identified through MALDI-TOF MS and assigned to the HILIC profile are shown in

Figure 2-4. They represent 90% of the whole glycosylation profile of the HFN 7.1

antibody.

2.3.4 Variation in N-linked Glycan Microheterogeneity

Under standard cultivation conditions, the N-linked glycosylation profile of HFN 7.1

is dominated by a bi-antennary, fucosylated structure with one terminal galactose

(G1F), which comprised 31% (± 2%) of the overall oligosaccharides, followed by the

G0F structure, which accounted for 26% (± 2%). The fully galactosylated and

fucosylated bi-antennary structure G2F only contributed to 8% of the total glycan

pool. Other oligosaccharide structures released from HFN 7.1, apart from the non-

fucosylated and non-galactosylated G0 structure, showed higher complexity, carrying

mostly terminal N-glycolylneuraminic acid residues and/or α-linked galactose residues

(Figure 2-4), both typical for murine-derived IgG proteins [48], [64], [101].


Figure 2-4: N-linked glycosylation profile of HFN 7.1 antibody (IgG1). The

antibody was purified at the end of a standard batch culture, glycans were

released by PNGase F, labeled with 2-AB and fluorescently detected by HILIC.

The predominant peaks were fractionated and the masses were measured by

MALDI TOF MS and compared to characterized oligosaccharide structures.

Sugar residues are as reported in the Appendix (Table A 1).

Before investigating the impact of bioprocess parameters on N-linked glycosylation of

IgG1, variations in the glycosylation profile over the course of a standard batch culture

were considered [102] in order to rule out the presence of degradative enzymes in the

supernatant. Released glycosidases and sialidases are known to cleave terminal sugar

residues (galactose and sialic acids) particularly towards the end of the culture due to

cytolysis [103–105]. Simultaneously, nutrient limitations, such as glucose and

glutamine towards the late-stage of the culture, can potentially affect nucleotide sugar

substrate availability [106–108]. Furthermore, the accumulation of ammonia can lead

to variations in the glycoform profile [78]. To check whether post-secretory

degradation, substrate limitation or waste metabolite accumulation affect glycosylation

of HFN 7.1 over time, the secreted monoclonal antibody was purified from three

repeated batch cultures at three different time points and the glycosylation profile was

determined (Figure 2-5A). The relative peak area of nine glycoforms was compared

after 26 hours corresponding to early exponential growth phase tshift, after 54 hours

30 40 50 60 70 80 900

2

4

6

8

G2F-NGNA2

G2F-GalNGNA

G2F-NGNA

G0F-GalNGNA

G3FG2F

G1F

G0F

Fluo

resc

ent i

nten

sity

(LU)

Retention time (min)

G0


when maximum viable cell concentration was typically reached tXv and at the end of

the culture after 80 hours tend.

Figure 2-5: N-linked glycosylation profile of purified monoclonal antibody

analyzed at three different time points throughout a batch culture [t (shift) = 26h,

t (max. Xv) = 54 hours, t (end) = 80 hours]. Relative peak areas of (A) a standard

batch culture without process parameter shift were compared to (B) pH-shifted

culture to pH 7.8.

Throughout the course of standard batch cultures the relative peak area remained

constant (Figure 2-5A) and therefore we concluded that N-linked glycosylation did not

vary within the time frame of a standard batch culture. In contrast, when pH was

shifted to 7.8, the glycosylation profile changed over the time course of the

G0G0F G1F G2F

G0F-GalNGNA G3F

G2F-NGNA

G2F-GalNGNA

G2F-NGNA205

101520253035

B

Peak

area

%

t (shift) t (Xv) t (end)

G0G0F G1F G2F

G0F-GalNGNA G3F

G2F-NGNA

G2F-GalNGNA

G2F-NGNA205

101520253035

A

Peak

area

%

t (shift) t (Xv) t (end)


exponential growth phase, as is most evident from the decrease of terminal

galactosylation (G1F and G2F) and the increase of the relative peak area of G0F by

6% from tshift to tXv (Figure 2-5B). However, the glycosylation profile remained

unchanged within the last 26 culture hours from tXv to tend indicating that post-secretory

degradation was not observed in the shifted culture even though culture viability was

low (see Figure 2-2B). To conclude, changes in the glycan microheterogeneity evident

from the relative peak area of the most abundant glycoforms were induced solely by

process parameter shifts.

2.3.5 Impact of Chemical and Mechanical Stress Parameters on N-linked

Glycan Microheterogeneity

In order to evaluate the effect of multiple process parameters on glycosylation of a

monoclonal antibody, the glycoform profile of the investigated IgG1 protein was

determined prior to the parameter shift and at the end of the culture. Since cell

productivity was influenced differently by each process parameter shift, changes in the

glycan profile are further expressed relative to changes in mAb concentration

produced within the same time frame, in order to provide better comparability. In this

way the fraction of each glycoform could be determined as a function of process

parameter and mAb concentration according to Eq. 2-3. Figure 2-6 shows the impact

of pH and DO, Figure 2-7 the impact of sparging and osmolarity on the proportion of

the most abundant glycoforms. In order to summarize the observed effects the

galactosylation (GI), sialylation (SI) and fucosylation (FI) indices [90] were calculated

according to Eq. 2-4, Eq. 2-5 and Eq. 2-6.

A shift in pH in the early exponential growth phase led to most profound changes in

glycan microheterogeneity of the IgG1 antibody. Terminal galactosylation and

sialylation (NGNA) decreased steadily with increased pH (see Figure 2-6A), which is

further reflected in the decrease of the GI from 0.43 to 0.24 and of the SI from 0.15 to

0.08 between pH 6.8 and 8.0 (Table 2-1), leading to a gradual increase in the G0F

fraction from 16 to 35% (Figure 2-6A). Simultaneously, the afucosylated structure G0

increased with increasing pH, resulting in a reduction of the FI from 0.99 to 0.94 (see

Table 2-1). It is worth noting that the range of variability in galactosylation reached in

the studied pH region is significant, but remarkably smaller compared to a GI range


reported for glucose limited cultures (GI = 0.35 – 0.72) [32]. Nevertheless, both

process conditions could be applied to control galactosylation in the determined range.

The effect of culture pH on glycosylation has been investigated earlier for different

glycoproteins produced in a variety of cell lines. While Muthing et al. (2003) reported

an increase in galactosylation (G2F) with increased culture pH for an IgG3 protein

produced in a hybridoma cell line, Seo et al. (2013) showed the opposite behavior for

a mAb produced in a human cell line. At the same time, Muthing et al. (2003)

observed an increase in afucosylation. In terms of sialylation, Yoon et al. (2005)

reported decreased sialic acid content in Epo with increasing culture pH (pH 7.8).

Other reports showed differences in the NANA/NGNA ratio at different pH values

[109], as well as no effect of pH 6.8–7.3 on the NANA content [95]. A juxtaposition of

the obtained results to earlier literature reports revealed that, as a process parameter,

pH can dramatically affect glycosylation of multiple glycoproteins produced in various

cell lines, including HFN 7.1 produced in murine hybridoma cells. Nevertheless, a

generally valid conclusion on the degree and direction of pH-induced variation of N-

linked glycosylation cannot be drawn. Indeed, the impact of pH as well as other

process parameters has to be assessed case-by-case depending both on the cell line and

the expressed protein. The mechanism of glycosylation is in fact at the same time

influenced by cell line-dependent differences, such as the expression levels of

glycosylation genes [82], and by the accessibility of the glycosylation site which is

protein-specific [110]. In addition, different experimental procedures were previously

applied to study the effect of culture pH on glycosylation and it has been shown that

the procedure itself can influence glycosylation [111]. Nevertheless, the obtained

results extend the knowledge of pH-induced variations in N-glycosylation, using a

model system that is characterized by a complex glycosylation profile.

Osmolarity shifts, while ranked second in the cell growth analysis, did not affect

glycosylation of the investigated IgG1 protein (see Figure 2-7B), as likewise reported

for β-IFN [112]. However, increased osmolarity was shown to increase the antibody

Man5 level [113], a glycoform not detected in the HFN 7.1 glycosylation profile.


Figure 2-6: Change in each glycoform fraction ∆Fi induced by process parameter

(A) pH and (B) dissolved O2 (DO). The change in glycoform fraction is expressed

as difference of the produced glycoform between the parameter shift time and the

end of the culture, normalized to the change in antibody concentration and was

calculated according to Eq. 2-3. () G0F, () G1F, () G2F, () G3F, () G0,

() G2F-NGNA, () G2F-NGNA2, () G2F-GalNGNA, () G0F-GalNGNA.

Sugar residues are as reported in the Appendix (Table A 1).

Apart from single parameter changes, osmolarity and pH were investigated in cross

experiments in order to evaluate their combined effect on glycosylation. pH and

osmolarity are important during fed-batch cultivation where increases in osmolarity

can be related to pH control by base addition or to feed addition. At the same time it

has been shown that low pH and high osmolarity can increase the mAb production rate

in hybridoma cell culture [114]. Thus, both parameters were shifted simultaneously in

the early exponential growth phase combining low pH with high osmolarity as well as

high pH with high osmolarity. For comparison an intermediate pH and osmolarity

cross-condition was included (Table 2-1). The combined effect on glycosylation was

compared to the GI, SI and FI reached from single process parameter shifts. As evident

10

20

30

40

10

20

30

40

6.8 7.0 7.2 7.4 7.6 7.8 8.00

2

4

6

8

10

0 20 40 60 80 1000

2

4

6

8

10

∆ F i (%

)

pH

BA

DO (%)


from Table 2-1, in the low pH 6.8 and high osmolarity (420 mOsm/kg) cross

experiment the effect of pH was predominant and the degree of galactosylation,

sialylation and fucosylation was comparable to a pH-shift to pH 6.8 alone. In the other

two cross conditions where pH and osmolarity were increased (pH 7.5/ 380 mOsm/kg;

pH 7.8/ 420 mOsm/kg) the GI, SI and FI followed the osmolarity behavior with

consistent glycosylation.

Table 2-1: Degree of galactosylation (GI), sialylation (SI) and fucosylation (FI) as

a function of the single shift parameter pH or osmolarity (mOsm/kg), and as a

function of the combined effect of pH and osmolarity.

pH Osmolarity GI SI FI

6.8 320 0.43 0.15 0.99

7.0 320 0.37 0.13 0.99

7.2 320 0.35 0.11 0.98

7.5 320 0.31 0.10 0.97

7.8 320 0.25 0.10 0.95

8.0 320 0.24 0.08 0.94

7.2 350 0.34 0.11 0.99

7.2 380 0.32 0.10 0.98

7.2 420 0.34 0.09 0.99

6.8 420 0.43 0.13 1.00

7.5 380 0.38 0.14 0.99

7.8 420 0.37 0.13 0.98

Regarding other chemical stress parameters, DO shifts resulted in slight changes of the

glycoform fractions (Figure 2-6B). In particular, with DO of 10% and 90% air

saturation the GI increased to 0.44 and 0.42 and the SI increased to 0.16 and 0.13

respectively, whereas at 50% air saturation, both indices remained slightly lower (GI =

0.35; SI = 0.11). Earlier literature reports demonstrate rather consistent glycosylation


at DO levels from 10 – 100% air saturation with small increases in sialylation of the

follicle-stimulating hormone (FSH) [115] and in terminal galactosylation (G2) of an

IgG [116], both at increased DO levels (90 – 100 % air saturation). Therefore, when

assessing chemical stress parameters and their effect on glycosylation, DO is ranked as

second most important single parameter that causes variations in HFN 7.1

glycosylation.

Finally, the mechanical stress parameter of sparging (in a range of 0.05 – 0.2 vvm) was

studied as an impact parameter of glycosylation. Our results suggest no significant

difference in any of the glycoforms within this sparging range (Figure 2-7A), whereas

hydrodynamic stress induced by sparging was earlier reported to result in a minor

increase of the G0F fraction [25].

Figure 2-7: Change in each glycoform fraction ∆Fi induced by process parameter

(A) sparging and (B) osmolarity. The change in glycoform fraction is expressed as

difference of the produced glycoform between the parameter shift time and the

end of the culture, normalized to the change in antibody concentration and was

calculated according to Eq. 2-3. () G0F, () G1F, () G2F, () G3F, () G0,

() G2F-NGNA, () G2F-NGNA2, () G2F-GalNGNA, () G0F-GalNGNA.

10

20

30

40

10

20

30

40

0.05 0.10 0.15 0.200

2

4

6

8

10

320 340 360 380 400 4200

2

4

6

8

10

∆ F i (%

)

Sparging (vvm)

BA

Osmolarity (mOsm/kg)


2.4 Conclusions

We propose a shift-experiment methodology that can be applied to systematically

investigate the effect of varying chemical and mechanical stress parameters that

constitute common bioreactor process parameters, on cellular growth and

glycosylation of a monoclonal antibody. Applying this methodology, we could show a

comprehensive picture of variations in complex N-linked glycosylation of one

particular glycoprotein, HFN 7.1, characterized in this study. An initial assessment of

the cell growth rate dependency on process parameters was used to define an operating

range in which cells were viable and produced monoclonal antibodies. As a result, pH

and osmolarity were considered as critical parameters in terms of growth, whereas

sparging and DO did not impact the specific growth rate in the relevant range.

Regarding the effect of process parameters on glycosylation, pH showed the most

profound impact on the variation of glycan microheterogeneity, followed by DO,

which had a slight impact on galactosylation and sialylation. Osmolarity and sparging

did not affect glycan microheterogeneity significantly. However, high osmolarity

prevailed regarding GI, SI and FI in combination with high pH, whereas the combined

effect of low pH with high osmolarity showed pH predominance on glycan

microheterogeneity.

Monitoring of process parameters and their impact on N-linked glycosylation is crucial

to provide desired as well as consistent product quality throughout a bioprocess. The

methodology we have presented here can provide an initial ranking of the impact of

environmental parameters that have to be considered under the Quality by Design

(QbD) scope. Furthermore, data obtained can be used for validation of recently

available mathematical models of mAb glycosylation [117], [118] and improve their

ability to capture the effect of process conditions. Ideally, this methodology can be

further applied in a high-throughput setting and connected to high-throughput

analytical techniques [119].

2.5 Remark

The work presented in this chapter has been partially submitted for publication to the

Journal of Biotechnology.

27

3 Insights into pH-induced Metabolic Switch by

Flux Balance Analysis

3.1 Introduction

Mammalian cells are used for the production of recombinant proteins that are large,

complex and require post-translational modifications [7]. With the increased medical

and economic demand for recombinant proteins, primarily mAbs, particular attention

has been paid to mammalian cell culture processes with the aim to consistently

produce high quality material. Over the past 25 years, bioprocesses have been

prolonged through process modifications and media optimization as well as through

genetic engineering, resulting in substantial increases of maximum viable cell

concentrations and product concentrations [17].

Nevertheless, mammalian cells are characterized by an inefficient metabolism [120].

With the increase in glycolytic rate observed under growth conditions, and with excess

glucose [121], pyruvate is preferentially reduced to lactate by lactate dehydrogenase A

(LDH A), instead of entering the mitochondria and being further metabolized in the

tricarboxylic acid (TCA) cycle. This has a negative impact on the energy yield which

can be achieved from glucose. Furthermore, it has been shown that lactate itself can

downregulate enzymes in the glycolytic process, namely hexokinase (HK) and

phosphofructokinase (PFK), thereby altering the cellular metabolism during fed-batch

cultivation [122], [123]. In addition, lactate accumulation results in reduced cell

growth in batch and fed-batch cultures [124], [125] and has a negative impact on cell

productivity [126]. Different strategies have been developed in an attempt to limit the

secretion of lactate and the associated adverse effects. Apart from approaches where

metabolic pathways were genetically modified [127], [128], several strategies for

operating the bioreactor were proposed that aim to reduce lactate accumulation and

often even result in lactate consumption. The latter are centered on the optimization of

the media and feeding strategies by reducing the initial glucose concentration [129],

[130] or feeding galactose [131] or even lactate directly [132] as alternative carbon

sources. Finally, lactate secretion can also be affected externally by pH alteration. As

28 CHAPTER 3: LACTATE METABOLISM

demonstrated by Osman et al. (2001), lactate consumption can be induced in GS-NS0

cells by reducing pH below 7.5, with concomitant decreases in cell growth and glucose

uptake. They also showed that the specific lactate consumption rate increased as the

pH value was further reduced, while the specific lactate production rate increased

when pH was set above 8.0. A similar behavior was observed in a different study using

a CHO cell line [95].

Efforts to characterize the cell metabolism have included 13C metabolic flux analysis

(MFA) as well as constraint-based flux balance analysis (FBA). 13C-MFA is a

powerful method which can be used to identify the split ratio of fluxes at branch points

or in cyclic pathways, such as the diversion of glucose to the partially cyclic pentose

phosphate pathway [134]. However, this technique relies on isotopomer measurements

and requires the use of 13C-labelled substrates in order to investigate intracellular flux

distributions. Despite the accuracy of MFA estimations, this method is experimentally

tedious and expensive, in particular when working on bioreactor scale. FBA on the

other hand, fully depends on the mass balance and can be applied effectively to

underdetermined systems by including thermodynamic and regulatory constraints.

These constraints play a key role in reducing the solution space [135]. Thus, the

intracellular flux distribution can be obtained from optimization towards a given

objective. FBA has been used to calculate the flow of metabolites through primary

metabolic pathways [136], [137]. The accuracy of FBA estimations using the objective

function ‘maximize ATP’ has been successfully validated in hybridoma cell culture

using experimental 13C-determined flux data [138]. In recent studies, both 13C-MFA

and FBA have been applied as tools to study lactate metabolism under glucose

depletion [139], and to compare different growth phases during fed-batch cultivation

[121], [123], [140], and different cell lines [141].

In this work, flux balance analysis was used to investigate the effect of pH on the

lactate metabolism of a murine hybridoma cell line during the exponential growth

phase of the culture. Lactate consumption was induced even at high glucose

concentrations by altering the pH from standard pH 7.2 to a lower value (pH 6.8),

whereas lactate production was highly increased once pH was set to a higher value (pH

7.8). While in other studies metabolic states were compared before and after the lactate

CHAPTER 3: LACTATE METABOLISM 29

switch [123], [139], we altered the pH in parallel controlled batch cultures and

analyzed the metabolic states considering the effect of culture time and thus the

growth phase of the cells. Furthermore, gene expression analysis by real-time PCR

was implemented to capture the pH-induced changes on the enzymes involved in the

central metabolism of murine hybridoma cells at the transcriptional level. The

information obtained thereby led to the identification of a regeneration pathway that

was activated at pH 7.8, and on this basis the metabolic network was enlarged. Finally,

simulation results were used to identify the consequences of pH alterations on energy

metabolism.

3.2 Material and Methods

3.2.1 Cell Line and Media

The hybridoma cell line HFN 7.1 (ATCC CRL-1606) [87], producing an

immunoglobulin G1 (IgG1) antibody against human fibronectin was used in this study.

The initially serum-dependent cells were sequentially adapted to the chemically

defined protein- and peptide-free culture media Turbodoma® TP6 (Cell Culture

Technologies) supplemented with 4.5 g/L D-glucose, 4 mM L-glutamine and 0.1%

(w/v) pluronic F-68. Cells were expanded as suspension cultures for 2 weeks in 50 mL

spinner flasks and 2 L roller bottles in a humidified atmosphere containing 5% carbon

dioxide (CO2) at constant temperature of 37 °C.

3.2.2 Bioreactor Setup and pH Shift Experiments

A DasGip bioreactor system (DasGip) was used for the cultivation of HFN 7.1 cells

under controlled culture conditions. The exponentially growing cells were transferred

to the bioreactor at a concentration of 2 x 106 cells/mL and diluted to a working

volume of 1 L to result in a seeding cell concentration of 0.6 x 106 cells/mL. Batch

cultivation was performed under standard conditions with temperature equal to 37 °C,

dissolved oxygen set to 50% air saturation and stirrer speed equal to 150 rpm. The pH

was controlled at 7.2 with CO2 sparging. Once viable cell concentration reached

1 x 106 cells/mL, a single pH shift was performed by the addition of a defined volume

of 1 M sodium hydroxide (NaOH) or 1 M hydrochloric acid (HCl). Due to the small

base or acid amount used, the time needed to shift the pH to the new setpoint was in


the range of a few minutes. During the shift, the pH controller was turned off and only

after the new setpoint was reached the pH was again controlled by base and CO2

sparging in order to keep the pH constant at the new setpoint until the end of

cultivation. The osmolarity of each culture was measured with an OsmoLab One

osmometer (LLA Instruments). The increase in osmolarity due to acid or base addition

was below 5% and this was considered to have a negligible effect on the cells.

Cell culture samples were taken twice a day and additionally before and after the pH

shift. Cell number and viability were determined using a CedeX cell counter

(Innovatis). Glucose and lactate concentrations were determined using a Hitado Super

GL compact instrument (Hitado). Samples were spun down and supernatants were

stored at -20 °C and later used for measurement of antibody concentration (PA

ImmunoDetection® Sensor Cartridge, Applied Biosystems) and amino acid

concentration (ZORBAX Eclipse Plus C18 4.6mm x 150mm, Agilent Technologies)

by HPLC. Prior to amino acid measurement, supernatant samples were filtrated using a

Vivaspin® 500 (5kDa) centrifugation filter (Satorius Stedim Biotech) at 4000 rcf at 4

°C for 25 minutes. Derivatization of amino acids with ortho-phthalaldehyde (OPA)

was performed with an online injector program according to Agilent’s application note

[142].

3.2.3 TaqMan® Gene Expression Assay

Cell samples lysed in TRIzol® reagent (Life Technologies) at day 1.1 and 1.9 from

each culture were stored at -80 °C. RNA was extracted from cell samples according to

the manufacturer’s recommendations. Quantity and quality of isolated RNA was

assessed by spectrophotometry with a NanoDrop 1000 (Thermo Scientific). All RNA

samples had an A260/A280 ratio within 2.00 to 2.03, which represents high purity of

RNA.

For each reaction, cDNA was prepared from 2 µg of RNA using the High Capacity

cDNA Reverse Transcription Kit (Applied Biosystems) according to the

manufacturer’s recommendations. A TaqMan® gene expression assay containing a

predesigned array of murine oligonucleotide primers and probes for target genes

involved in murine central metabolism (Appendix Table A 2) was purchased from Life

Technologies. Quantitative PCR reactions were performed in triplicates according to


supplier’s protocol using TaqMan® Gene Expression Master Mix (2x) (Rox; Life

Technologies) in an AB 7900 HT Real-Time PCR System (Life Technologies, USA).

The results were analyzed by applying the comparative CT method (2−ΔΔCT method)

[143] with glyceraldehyde-3-phosphate dehydrogenase Gapdh as an internal control

gene for the normalization of mRNA expression levels. The expression of each target

gene relative to the internal control gene after pH-shift (at day 1.9) was compared to

the relative expression before pH-shift (at day 1.1), for both pH-shifted conditions

respectively.

3.2.4 Flux Balance Analysis (FBA)

Flux balance analysis as described previously [144] was applied to the central

metabolic network based on Mulukutla et al. (2012). The network is

compartmentalized into mitochondria and cytosol and contains reactions of the main

energy metabolism of glycolysis, glutaminolysis and TCA cycle, as well as biomass

and antibody synthesis. The carbon shunt into the pentose phosphate pathway (PPP)

cannot be resolved and was assumed to only result in DNA and RNA synthesis.

Amino acid degradation and synthesis pathways have been adjusted according to the

KEGG database for Mus musculus [145]. Apart from the main energy metabolism, the

malate-aspartate shuttle has been included in the metabolic network to capture the

regeneration of the cofactor NAD+ in the cytosol. Since the cofactor requirement of

some enzymatic reactions is still under investigation and due to possible mitochondrial

transhydrogenase action, the cofactors NADH and NADPH were lumped together. The

total production of NAD(P)H and FADH2 was calculated from the network and a

theoretical P/O value of 2.5 and 1.5 respectively was considered for ATP production in

oxidative phosphorylation. Further, reactions that can restore the phosphoenolpyruvate

and therefore pyruvate level in the cytosol have been added to the metabolic network

after identification of high expression levels of the mitochondrial

phosphoenolpyruvate carboxykinase (PEPCK2) at culture pH 7.8 from gene

expression data. The reactions and metabolites considered in the metabolic network

are listed in the Appendix (Table A 3), together with the scheme of the metabolic

network (Figure A 2).


Metabolite consumption and production profiles were fitted using a logistic decline

equation (Eq. 3-1) to describe metabolite consumption and a logistic growth function

(Eq. 3-2) to describe metabolite production. Viable cell concentration data can be

fitted using a generalized logistic equation containing non-negative model parameters

A, B, C, D and E (Eq. 3-3).[89] The parameter E used in all equations has been added

to account for data that do not start from zero or converge to zero, as follows:

𝑁 =A

exp(B𝑑) + C + E (3-1)

𝑃 =A

1 + C exp(−D𝑑)+ E (3-2)

X=A

exp(B𝑑) + C exp(-Dt)+ E (3-3)

Specific metabolite consumption rates qN and specific production rates qP were

calculated by differentiating the fitted logistic equations and consequently used as an

input in FBA:

𝑞N =

1𝑉

d𝑁d𝑑

(3-4)

𝑞P =

1𝑉

d𝑃d𝑑

(3-5)

The biomass composition for hybridoma cell lines has been described previously

[146–148] and is provided in the Appendix 6.2.

Flux balance analysis using maximization of ATP production as an objective function

was employed to estimate intracellular fluxes. ‘Maximization of ATP’ has proven to

be a valid objective function for hybridoma cells due to their hyperactive, cancer-like

metabolic behavior [138], [149]. The reactions in the network were programmed in

Matlab using the SimBiology® model object and a sequential quadratic programming


algorithm was utilized to solve the optimization problem. The 95% confidence interval

was calculated as two times the estimated standard error. Finally, the reaction network

and the validity of the objective function were verified using literature data from 13C-

tracer studies [121], [150]. Comparison of the calculated intracellular fluxes with the

experimentally determined values gave an acceptable agreement within the main

energy metabolism with expected discrepancies in the pentose phosphate pathway and

in the anaplerotic reaction of pyruvate to oxaloacetate (data not shown).

3.3 Results and Discussion

3.3.1 Characteristics of Cell Growth and Metabolism

pH-shift experiments to high (pH 7.8) and low pH (pH 6.8) were performed in

duplicates during HFN 7.1 batch cultivation in controlled bioreactors once cells were

growing exponentially and had reached a cell concentration of 1 x 106 cells/mL.

Profiles of viable cell, antibody, glucose and lactate concentrations were compared to

an averaged standard cultivation from four independent experiments with constant pH

7.2 (Figure 3-1). Under standard conditions the peak viable cell concentration of

3.3 (± 0.3) x 106 cells/mL as well as a maximum growth rate of 1.1 d-1 were reached

during the exponential growth phase. In contrast, the highest mAb concentration was

observed in the pH 6.8 shifted culture, even though growth was reduced compared to

standard cultivation conditions.

The lowest growth profile was obtained at pH 7.8, where a growth rate equal to 0.70 d-

1 was measured during the exponential phase, resulting in a maximum viable cell

concentration of 2.1 x 106 cells/mL with a final mAb concentration 27% lower

compared to standard conditions (see Figure 3-1A, B). It is worth noting that the

observed variations in growth rate and productivity are in agreement with previously

published pH-shift experiments using a GS-NS0 mouse myeloma cell line [133].


Figure 3-1: Profiles of (A) viable cell concentration, (B) antibody concentration, (C)

lactate concentration and (D) glucose concentration of a standard cultivation pH 7.2

(n=4) and two pH-shifted cultivations (pH 6.8 and pH 7.8) performed in duplicates. pH

shift was performed once cells reached 1 x 106 cells/mL and is indicated by dashed line.

Measured data points were fitted using logistic equations (3-1), (3-2) and (3-3).

0 1 2 3 40.00.51.01.52.02.53.03.5

0 1 2 3 4020406080

100120140

0 1 2 3 405

1015202530354045

0 1 2 3 40

5

10

15

20

25

standard pH 7.2 shift pH 6.8 shift pH 7.8

D

C

B

Viab

le Ce

ll Co

nc. (

106 ce

lls/m

L)

Time (days)

Antib

ody

Conc

. (m

g/L)

Time (days)

Lacta

te Co

nc. (

mm

ol/L

)

Time (days)

A

Gluc

ose C

onc.

(mm

ol/L

)

Time (days)


More interestingly, lactate and glucose metabolism was instantly affected by pH shift

(Figure 3-1C, D), also as reported earlier [93], [94], [133]. Once pH was reduced to

6.8, the lactate concentration decreased, indicating a lactate consumption state that was

triggered by pH and remained present throughout the cultivation. Concurrently,

glucose uptake was substantially reduced. In contrast, the lactate concentration

increased dramatically as soon as pH was set to 7.8 resulting in a final lactate

concentration of 39.3 (±3.3) mmol/L compared to 15.6 (±3.3) mmol/L under standard

conditions. The increase in lactate concentration was connected to an increase in

glucose consumption (see Figure 3-1 C, D).

Specific consumption and production rates of all measured metabolites were calculated

over the time of the exponential growth phase (WD 1.0, WD 1.5, WD 1.9) for the

averaged standard cultivation, and compared in order to assess the effect of culture

time on specific rates. As can be seen from Figure 3-2A, specific production and

consumption rates of basic metabolites, amino acids, as well as biomass and mAb

decreased over time. Therefore, a fair comparison between different conditions can

only be made at the same time point, so as to exclude metabolic alteration due to

culture time. Consequently, specific consumption and production rates for the pH-

shifted conditions were calculated at working day 1.9 and compared to standard

conditions at the same time point (see Figure 3-2B, C). Apart from changes in glucose

and lactate metabolism, consumption and production of several other amino acids was

clearly altered, e.g. the metabolism of glycine switched from consumption to

production in the high lactate production state and at the same time glutamate

production was significantly increased. Specific rates from Figure 3-2 were used as

input fluxes vm in the subsequent flux balance analysis.

The HFN 7.1 batch cultivation is characterized by glutamine depletion after 2.1 days.

Therefore, data after 2.1 days were not considered in flux balance and gene expression

analysis.


Figure 3-2: Specific consumption and production rates of glucose, lactate, various

amino acids (nmol/106cells/h), biomass and mAb (µg/106 cells/h) under different

pH conditions. (A) Standard batch cultivation (pH 7.2) at three time points (WD

1.0, WD 1.5 and WD 1.9) during exponential growth phase. (B) Comparison of

two metabolic states at pH 7.2 and at pH 6.8 evaluated 17 hours after pH-shift

(WD 1.9). (C) Comparison of two metabolic states at pH 7.2 and at pH 7.8

evaluated 17 hours after pH-shift (WD 1.9). Indicated time points were

considered in FBA.

GLCLAC

GLN-400-300-200-100

0100200300400500

Spe

cific

rate

(nm

ol/1

06 cells

/h)

GLUALA

SER

GLYASPARG HIS CYS

THR ILEVAL

PHE

TYRLYS

LEUPR

OMET

ASN-30-25-20-15-10-505

101520

WD 1.0 WD 1.5 WD 1.9

Biomass mAb

0

5

10

15

20

25

30

Spe

cific

rate

(µg/

106 ce

lls/h

)

GLCLAC

GLN-400-300-200-100

0100200300400500

Spe

cific

rate

(nm

ol/1

06 cells

/h)

GLUALA

SER

GLYASPARG HIS CYS

THR ILEVAL

PHE

TYRLYS

LEUPR

OMET

ASN-20-15-10-505

101520

pH 7.2 pH 6.8

Biomass mAb

0

5

Spe

cific

rate

(µg/

106 ce

lls/h

)

GLCLAC

GLN-400-300-200-100

0100200300400500

Spe

cific

rate

(nm

ol/1

06 cells

/h)

GLU

ALASE

RGLY

ASPARG HIS CYSTHR ILE

VALPH

ETYR

LYSLEU

PRO

METASN

-20-15-10-505

101520

pH 7.2 pH 7.8

Biomass mAb0

2468

101214

C

B

A

Spe

cific

rate

(µg/

106 ce

lls/h

)


3.3.2 Metabolic Flux Distribution of Standard Cultivation during

Exponential Growth Phase

The metabolic model based on Mulukutla et al. (2012) contains 4 degrees of freedom

coming from the pyruvate carboxylation reaction as an alternative entry to the TCA

cycle, from two malate to pyruvate pathways located in the cytosol and in the

mitochondria respectively, and the last one from an amino acid synthesis pathway. The

flux distribution in the central metabolic network was investigated using a flux balance

approach with ‘maximization of ATP production’ as the objective function as

suggested and validated for hybridoma cells [138].

The average of the calculated fluxes is shown in Figure 3-3 with error bars

representing the 95% confidence interval corresponding to the variation of input fluxes

calculated from 4 repeated standard cultivations. For convenience, amino acid

synthesis and degradation pathways are not shown. A representation of all intracellular

fluxes is presented in the Appendix (Figure A 3). As shown in Figure 3-3, intracellular

fluxes decrease over time during exponential growth at pH 7.2, which is most apparent

in the glycolytic pathway. Nevertheless, TCA fluxes remain high over the first 12

hours by reducing the lactate flux and therefore increasing the pyruvate influx.

Consequently, the malate-aspartate flux is retained at high rates in order to maintain

the NAD+/NADH balance in the cytosol. At working day 1.9 intracellular fluxes were

significantly reduced compared to the early exponential phase, corresponding to a

reduction of the growth rate by 36% to 0.70 d-1.

The distribution of intracellular fluxes during the short time frame of the exponential

growth phase furthermore clearly demonstrated the effect of culture time, which

should consequently be considered when comparing distinct metabolic states,

especially under the impact of additional external parameters such as pH. This is also

relevant, but often neglected, when differentiating between lactate consumption states

in the stationary growth phase and lactate production states during exponential growth

of fed-batch cultures.


Figure 3-3: Intracellular flux distribution within glycolysis, TCA cycle, malate-

aspartate shuttle, glutaminolysis, biomass and mAb synthesis of standard batch

cultivation at three time points during exponential growth phase. The pathways

of amino acid metabolism are not shown. Values of all calculated fluxes are

reported in the Appendix (Figure A 3).

3.3.3 Metabolic Flux Distribution in Lactate Production and Lactate

Consumption State

The two metabolic states of lactate production under standard conditions and lactate

consumption triggered by pH 6.8 are compared in Figure 3-4. While the glucose

uptake rate at low pH (32 nmol/106 cells/h) was reduced compared to standard

conditions (88 nmol/106 cells/h), pH-induced lactate consumption contributed to a

comparable influx into the TCA cycle through pyruvate and triggered a 37% increase

in glutamine uptake. Similarly, high fluxes through the malate-aspartate shuttle were

maintained, presumably due to the requirement of cofactor balancing in the cytosol

[123], [151]. While under standard conditions cytosolic NADH produced by

glyceraldehyde phosphate dehydrogenase (GAPDH) in the glycolytic process is


primarily regenerated by lactate dehydrogenase A (LDH-A) reducing pyruvate to

lactate, a switch to lactate consumption causes an additional generation of NADH, and

therefore requires an increased activation of the malate-aspartate shuttle to regenerate

the cofactor NAD+. Thus, the influx into mitochondria was preserved through the

pyruvate node, glutaminolysis and the malate-aspartate shuttle, and resulted only in an

insignificant reduction of the final TCA flux. Therefore, pH-triggered lactate

consumption had no severe consequences on the flux distribution within the main

metabolic network compared to pH 7.2. Nevertheless a 35% increased mAb synthesis

rate was obtained at pH 6.8 mainly due to altered amino acid utilization (Appendix

Figure A 4).


aspartate shuttle, glutaminolysis, biomass and mAb synthesis in lactate

consumption state (pH 6.8) compared to lactate production state at standard

conditions (pH 7.2) at working day 1.9. The pathways of amino acid metabolism

are not shown. Values of all calculated fluxes are reported in the Appendix

(Figure A 4).


Mechanistic explanations of the pH effect on cell metabolism, in particular glycolysis

and lactate production/consumption, have been proposed earlier [93], [133]. As

discussed by these authors, lactate consumption might be triggered by a self-regulation

mechanism of the cell trying to counteract the extracellular pH by importing lactate

through the monocarboxylate transporter (MCT), which requires the symport of H+

ions [152]. The authors also argue that pH might alter the activity of glycolytic

enzymes or the membrane potential, which affects the glucose uptake rate and

therefore lactate production and consumption. Even though gene expression of

glycolytic enzymes at pH 6.8 and pH 7.2 does not differ (see Figure 3-5), enzyme

activity can still be affected by pH and is not reflected within gene expression analysis.

At this point, we do not provide any further explanation of the mechanism of action of

pH reduction on the cell metabolism. However, we were able to reveal the

consequences of pH reduction on the intracellular metabolism by FBA.

3.3.4 Metabolic Flux Distribution in Unfavorable Lactate Over-Production

State

The metabolic state of cells at pH 6.8 was the opposite to that observed in cultures at

pH 7.8. Lactate was overproduced, again being triggered by either a self-regulation

mechanism of the cell, by altered membrane potential or by higher enzyme activity in

glycolysis. However, the latter cause of lactate overproduction is yet again not

apparent from gene expression analysis (see Figure 3-5).

Figure 3-5: Gene expression of 32 genes involved in the central murine

metabolism. Data were measured in triplicates and mRNA expression of the two

pH-shifted conditions was compared relative to mRNA expression at the time of

pH shift. 50% over- or under-expression is considered as not significant.

Hk1Hk2 Pfkl

Pgam

2Pk

m2Pep

ck1Pep

ck2Ldha PcxPdha1

Pdha2Dlat

Pdk1

Pdk2

Pdk3 Cs

Idh3aOgdhSd

ha Fh1Mdh1Mdh2AclyMe1Me2Me3

Gpt2Got1Got2Glud

1 GlsG6pd

x-1.0-0.50.00.51.01.52.0

Fold

Cha

nge

pH6.8 pH7.8


Together with the high lactate production, an increase in the glucose uptake rate was

observed (196 nmol/106 cells/h compared to 88 nmol/106 cells/h) which was necessary

to keep up with the carbon requirements to produce lactate. A closer look further

revealed that the ratio of total carbon channeled to lactate could not be met solely by

the measured glucose uptake rate (∆L/∆G=2.2) and therefore an additional pathway

had to be active, which could provide more pyruvate for lactate formation. Two PEP

regeneration pathways (OAA → PEP, OAAm → PEPm), that are part of the

gluconeogenic pathway, were considered to fulfill the pyruvate balance. However, due

to the presence of two cyclic pathways and the thereby increased complexity around

the pyruvate node, a unique optimized solution could not be obtained. To identify the

active pathway, mRNA expression of the involved enzymes has hence been taken into

consideration (see Figure 3-5). Considering the conversion of oxaloacetate to

phosphoenolpyruvate, Pepck1 and Pepck2 are the two genes encoding either the

cytosolic or mitochondrial enzyme phosphoenolpyruvate carboxykinase. According to

gene expression analysis, mRNA level of the cytosolic form Pepck1 did not exceed the

threshold value in the qPCR reaction after 36 PCR cycles and therefore was

considered as not expressed. This led to the conclusion that the cytosolic reaction of

oxaloacetate to phosphoenolpyruvate was inactive under all pH conditions and was

therefore not considered in the metabolic network. In contrast, the mitochondrial form

Pepck2 showed a 1.2-fold higher expression at pH 7.8. Thus, under the assumption

that gene expression correlates with enzyme activity, the mitochondrial reaction

OAAm → PEPm has been included in the metabolic network. Furthermore, the

phosphoenolpyruvate transport flux from mitochondria to cytosol was unbound. Other

theoretically possible pathways that could replenish the pyruvate requirements [141],

including the malate to pyruvate conversion by malic enzymes (ME1, ME2 or ME3)

and the cytosolic malate to oxaloacetate conversion by malate dehydrogenase (MDH1)

were excluded after consideration of mRNA expression data (see Figure 3-5). Hence, a

validation of the metabolic network was performed based on gene expression data.

Finally, the identified PEP regeneration reactions were constrained to zero at all pH

conditions other than pH 7.8. However, even without constraining them, this pathway

was not chosen under standard and low pH conditions due to the objective function

requirement ‘maximization of ATP production’, as the mitochondrial PEP


regeneration requires ATP and additionally takes away flux from the TCA cycle (data

not shown). Even though this metabolic state appears highly energy-inefficient, it is

further confirmed by a 1.2 fold increase in Pdk2 expression. The Pdk2-encoded protein

phosphorylates pyruvate dehydrogenase, down-regulating the activity of the

mitochondrial pyruvate dehydrogenase complex that drives pyruvate decarboxylation

(PYRm → AcCoAm). The calculated flux distribution is presented in Figure 3-6. By

considering the above described regulatory constraints, the tremendous lactate

formation was fulfilled using PEP regeneration in addition to increased glycolytic

fluxes. As a result, the TCA cycle and malate-aspartate shuttle were almost shut off.

Only an increased glutamate flux by amino acid degradation pathways being converted

to α-ketoglutarate (AKGm) resulted in minor fluxes in the lower part of TCA.

To conclude, the unfavorable pH of 7.8 led to the expression of enzymes that are

traditionally considered to be involved in gluconeogenesis but were recently found to

play an important role in the transition of CHO cells from growth to stationary phase

of fed-batch production [141]. Further, such genes are involved in the regulation of the

TCA cycle activity and concurrently the energy metabolism in different cell types

[153], [154].

Finally, a detailed understanding of the pH effect on intracellular metabolism was

provided, even though the signaling pathway on regulatory level remained unknown. It

is assumed that pH-induced changes were guided by the combination of proton

gradient maintenance, redox balancing and pathway regulation, e.g. inhibition of

glycolytic enzymes. In order to gain a mechanistic understanding of the effect of pH

on cell metabolism, activities of glycolytic enzymes and expression of signaling and

regulatory factors as well as transporters should be further investigated. Likewise, the

intracellular flux redistribution through phosphoenolpyruvate should be confirmed by

isotope measurements using 13C-MFA in an appropriate experimental setup in order to

provide additional evidence to transcript data.


Figure 3-6: Intracellular flux distribution including additional mitochondrial

PEP recycle pathway of high pH condition (pH 7.8) compared to lactate

production state at standard conditions (pH 7.2) at working day 1.9. The

pathways of amino acid metabolism are not shown. Values of all calculated fluxes

are reported in the Appendix (Figure A 5).

3.3.5 Consequences of pH-Shifts on the Energy Metabolism

In order to gain more insight into the energy metabolism at different pH conditions,

the ATP production rates were evaluated from the calculated fluxes by applying the

theoretical P/O ratio of 2.5 for NAD(P)H and 1.5 for FADH2 respectively, and are

represented in Figure 3-7. Throughout the exponential growth phase of the standard

batch culture (WD 1.0, WD 1.5 and WD 1.9), the total ATP production qATP decreased

from 5530 (±574) nmol/106cells/h to 2168 (±702) nmol/106cells/h within almost 24

hours and thereby followed the same trend as the intracellular flux distribution (Figure

3-3). Even though the total ATP production rate decreased over time, the ratio between

the different originating pathways remained equal between working day 1.0 and

working day 1.9. The main substance contributing to energy generation was


NAD(P)H, which, together with FADH2, was utilized to produce up to 84-87% of the

total ATP by oxidative phosphorylation. The redox cofactors, i.e. NAD(P)H and

FADH2, involved in oxidative phosphorylation can originate either from the TCA

cycle (65-72%), or from other reactions such as the pyruvate decarboxylation reaction

(PYRm → AcCoAm) and reactions within the amino acid metabolism (14-19%) (see

Figure 3-7). Glycolysis accounted for 7-9% of the total ATP production rate and

succinyl CoA synthetase in the TCA cycle contributed to 7% of the ATP production

(Figure 3-7: WD 1.0, WD 1.5, WD 1.9). Throughout the exponential growth phase of

HFN 7.1 cells, the majority of the energy (72% to 79%) was generated in the TCA

cycle either directly in the form of ATP or through oxidative phosphorylation. ATP

utilization for biomass formation, mAb synthesis, amino acid synthesis and

degradation was not considered in this analysis.

Figure 3-7: Comparison of energy metabolism of lactate consumption state (pH

6.8) and lactate overproduction state (pH 7.8) with lactate production state

during the exponential growth phase at standard conditions (pH 7.2). Total ATP

production rate by ATP, NADH and FADH2 was calculated considering different

originating pathways. A P/O ratio of 2.5 for NADH and 1.5 for FADH2 was

assumed.

Comparison of the ATP production rates and the originating pathways was made

between the standard culture on working day 1.9 and pH-shifted conditions (Figure

3-7: pH 6.8, pH 7.8). The metabolic state at pH 6.8 resulted in a slightly lower qATP of

0

1000

2000

3000

4000

5000

6000

pH 7.8pH 6.8 WD 1.9WD 1.5

ATP

flux

(nm

ol/1

06 cells

/h)

Oxi-P(others) TCA Glycolysis Oxi-P(TCA)

WD 1.0


1752 (±770) nmol/106cells/h. Energy in the form of ATP was likewise primarily

generated by oxidative phosphorylation (TCA: 73%, others: 17%) and further by

succinyl CoA synthetase in the lower part of the TCA cycle (7%). Furthermore, energy

efficiency of the lactate production and consumption state was compared by

calculating the total ATP production per total carbon moles of substrate. The pH-

triggered lactate consumption state corresponded to an energy efficiency of 4.8 mol

ATP/C-mol, whereas under standard conditions the energy equivalent was equal to 4.1

mol ATP/C-mol. Hence, pH reduction induced a slightly more energy efficient

metabolic state and should therefore be considered similar to glucose depletion in

CHO fed-batch cultures [139], as a trigger for the initiation of energy-efficient

metabolic states.

In the lactate overproduction state at pH 7.8 the low fluxes in TCA and malate-

aspartate shuttle led to an overall decrease in total ATP production of 71% (634 (±172)

nmol/106 cells/h ATP) compared to standard conditions. At the same time NAD(P)H

was no longer the main contributor to the overall ATP production (Figure 3-7).

Instead, more than two-thirds of the produced energy came from ATP directly

generated in glycolysis (68% of total ATP production).

3.3.6 Relevance of pH-induced Metabolic States for the Design and Scale-

up of Industrial Bioprocesses

It is well known that pH has a significant influence on cell performance [97], [155]

and product quality [82], [156] during mammalian cell cultivation. Therefore, pH is

commonly investigated in design of experiments (DOE) approaches and tightly

controlled in an optimal range during cultivation. However, it has been shown that pH

heterogeneities within the bioreactor are present due to carbon dioxide accumulation

or base addition [26]. This is particularly significant when comparing small-scale

bioreactors used for process development and large-scale reactors applied in

production. Therefore, pH heterogeneities should be kept in mind when alterations in

cellular metabolism, particularly in lactate, appear throughout process development

and scale-up. The present study showed that a pH reduction to pH 6.8, which in a

large-scale bioreactor could be caused by CO2 accumulation, instantly induced lactate

consumption and a metabolic state that was energetically more efficient with increased


cell productivity. In contrast, at pH 7.8, a pH region that cells could easily experience

locally during alkali addition in large-scale bioreactors, less ATP was generated due to

the fact that glucose was used for lactate formation instead of entering the TCA cycle.

Furthermore, the unusually high lactate formation was afforded not only by increased

glucose uptake, but also by a PEP regeneration pathway. The presented approach can

be applied to increase large-scale process understanding (e.g. heterogeneous culture

conditions), as well as to deduce recommendations for bioprocess control strategies,

such as avoiding the use of base for pH control and rather accept shifts to lower pH

values (up to pH 6.8). According to our results, low pH showed no negative effect on

the culture in terms of energy metabolism but, in contrast, resulted in increased mAb

production.

Hence, by focusing on the whole cellular metabolism, and in particular on the related

energy requirements, rather than just on purely macroscopic variables, we became

aware of additional possibilities for process optimization in murine hybridoma cell

culture.

3.4 Conclusions

This work provides the first example of using a flux balance model containing the

major metabolic pathways of glycolysis, glutaminolysis and the TCA cycle, as well as

the malate-aspartate shuttle for cofactor balancing, to analyze the influence of pH on

cell metabolism during mammalian cell cultivation. Due to a residual degree of

freedom in the model, ‘maximization of ATP’ has been chosen as objective function

and the calculated fluxes were sufficiently tight to describe the metabolic behavior of

murine hybridoma cells. Furthermore, gene expression analysis of metabolic enzymes

was employed for validation of the proposed metabolic network. The analysis

additionally revealed the presence of a pathway that was only activated under high pH

conditions.

Before applying FBA to pH-shifted cultures, the intracellular flux distribution of a

standard cultivation with constant pH was calculated at different time points to

emphasize the often neglected effect of time on cell metabolism. Thus, a comparative

analysis of the two distinct pH-induced metabolic states of lactate consumption and


lactate overproduction was performed at the same time point. The presented results

demonstrate the ability of mammalian cells to switch to more energy-efficient states at

low pH and additionally led to the discovery of undesirable changes in metabolic

pathways required to fulfill an increased lactate production at high pH. Finally,

implications of such metabolic states in process control and optimization are provided.

3.5 Remark

The work presented in this chapter has been partially submitted for publication to

Biotechnology Progress.

49

4 An Unstructured Model of Metabolite and

Temperature Dependent Cell Cycle Arrest in

Hybridoma Batch and Fed-Batch Cultures

4.1 Introduction

Fed-batch processes for the production of monoclonal antibodies are well established

at industrial scale. The complexity of these proteins, in particular the requirement of

post-translational modifications, necessitates the use of mammalian cell culture for

their expression. Since the initial use of cell culture technology for mAb production

[12], the biggest achievement in mammalian cell culture over the last three decades

has been the increase in cell productivity [17]. This has on the one hand been achieved

by the use of advanced production cell lines, originating from CHO cells [157]. On the

other hand, optimized culture media [16] and a better comprehension of fed-batch

processes have fostered this improvement. Regarding the latter, process parameters

used to provide control over the culture environment can be tuned to maximize cell

growth as well as productivity [95]. A decrease in the bioreactor pH setpoint, for

instance, increased productivity in multiple mAb producing cell lines [95], [158],

[159].

Temperature, a process parameter that can be rather easily controlled, has been

reported to similarly influence the mAb production rate, particularly when applying

mild hypothermic conditions. NS0 cells, when grown at 34 °C accumulated in the G1

phase and achieved higher antibody production rates, while reaching higher maximal

viable cell concentration with prolonged cell viability [160]. Similarly, CHO cells

showed cell arrest in G1 with a 1.7-fold increase in specific productivity at 30 °C

accompanied by a reduced cell growth rate [161]. The reported effect of temperature is

not limited to the G1 phase, as others have reported cell blockage in the G2/M phase

post temperature shift [162], or even a rapid decrease in the S phase [163]. Moreover,

Ducommun et al. [164] reported a 6-fold increase in the specific production rate and a

stabilization of the viable cell concentration in continuous CHO cell culture with

50 CHAPTER 4: CELL CYCLE TRANSITION

temperature decrease to 32 °C. Similarly, Schatz et al. [165] showed a 38-fold increase

in product yield with a temperature shift from 37 °C to 28 °C. Nevertheless,

temperature shifts do not always result in an increased product yield. Some reports

showed a prolonged cell viability [166] or higher maximum viable cell concentration

[167] at lower temperatures, accompanied by a lower antibody yield. Others yet have

reported a longer lag phase at 33 °C but achieved comparable cell densities to 37 °C,

together with reduced productivity [168]. Moreover, temperature influenced the cell

cycle, as evident from an increased G1 phase fraction [169], though this was not linked

to a positive effect on mAb titer.

Literature reports are variable and the results seem to depend on the cell lines, the

expression systems, and the expressed recombinant product [170], [171]. To some

extent, the differences can be attributed to the complex link between the cell cycle

being at the center of cellular growth, death, and productivity [172], [173]. In order to

find a good balance between temperature-induced growth limitation (and consequently

the integral viable cell density – IVCD [174]) and cell-specific productivity [171],

temperature dependency has to be investigated and the time point of the temperature

shift needs to be optimized with the aim to increase the final antibody titer.

A quantitative understanding of the effect of temperature can aid bioprocess

development to improve product titer and product quality. Different temperature

conditions and shift strategies are investigated during industrial bioprocess

development. However, these investigations are empirical and lack a systematic,

model-based approach. The development of relevant mathematical models can help

define optimal temperature profiles and is therefore of practical importance. Early

contributions [175], [176], initiated the study and analysis of the temperature effect on

the cell cycle by employing and formulating mathematical models. Further

developments of cell cycle models to study and analyze conditions that enhanced

antibody production have been reported [177–180], although without accounting for

temperature effects. Temperature was however considered in the study of Fox et al.

[181] with the aim to optimize temperature shift times, albeit without considering the

cell cycle. Recently, Karra et al. [96] proposed a detailed mass-based population

balance model with cell cycle segregation and linked it to an unstructured model

CHAPTER 4: CELL CYCLE TRANSITION 51

including the temperature effect. Unfortunately, the model was validated only during

batch culture, which lacks industrial relevance.

Herein we formulate a temperature-dependent cell cycle model for a mAb producing

mammalian cell line. The model is developed for cultures in batch and fed-batch mode

– covering a range of nutrient conditions – while capturing the distribution of cells in

different cell cycle phases, including the quiescent state (G0). We validate the model

experimentally and use it to predict the time points for temperature shifts in fed-batch

culture. The overall methodology maps the path to systematically develop industrially

relevant mathematical models with the aim to optimize cultivation temperature

profiles.

4.2 Materials and Methods

4.2.1 Cell line and Media

HFN 7.1 murine hybridoma cells (ATCC CRL-1606) [87] adapted to chemically

defined, protein- and peptide-free culture media Turbodoma® TP6 (Cell Culture

Technologies) were used in this study. The media was prepared following the vendor’s

instruction and supplemented with 4.5 g/L D-glucose, 4 mM L-glutamine and 0.1%

(w/v) pluronic F-68. Cells were sub-cultured in suspension for 14 days at 37 °C in a

humidified incubator containing 5% CO2 and subsequently used for batch and fed-

batch cultivations.

4.2.2 Shake Flask Batch Culture

HFN 7.1 cells were cultured in triplicates in 1 L Erlenmeyer flasks (Corning) with 240

mL working volume at a seeding cell concentration of 0.6 x 106 cells/mL. At the start

of each batch experiment the temperature was set to either 37 °C or 33 °C. Mixing was

accomplished using a Stuart SSL1 orbital shaker (Bibby Scientific) at 125 rpm. Viable

cell concentration and viability were evaluated manually by a dye-exclusion method

using Erythrosin-B stain solution. Samples were counted using a haemocytometer

under the Leica DM-IL inverted phase microscope (Leica). Samples were taken twice

a day and centrifuged at 660rpm for 3 minutes. The supernatant was stored at -20 °C

for metabolic profiling and antibody titer determination. 1 x 106 cells were fixed by


drop wise addition of 1 mL of 75% v/v ice-cold ethanol for DNA analysis. The fixed

cells were stored at -20 °C prior to cell cycle analysis.

4.2.3 Fed-Batch Cell Culture

Fed-batch cultures were performed in a parallel bioreactor system (DasGip) equipped

with online temperature, pH and dissolved oxygen measurement probes (Mettler-

Toledo), a pitched-blade impeller and a porous sparger (10 µm). Cells were seeded

into the bioreactor at a cell concentration of 0.6 x 106 cells/mL in a working volume of

1 L. Dissolved oxygen was controlled at 50% air saturation with a constant airflow of

3 L/h and an oxygen on demand strategy. pH setpoint was set equal to 7.2 over the

whole culture and controlled initially by CO2 sparging and in some cases by base

addition (2 mol/L sodium hydroxide) depending on the lactate production of the

growing cells. Temperature was either set to 37 °C and kept constant throughout the

culture or shifted to mild hypothermia (33 °C) at three different time points (t = 0 h,

t = 30 h, t = 76 h) by changing the controller setpoint. The time required for the reactor

to reach the new temperature setpoint was negligible short.

A continuous feed of amino acids (RPMI-1640 50x, Sigma), vitamins (RPMI-1640

100x, Sigma), trace elements (trace element mix 1000x, Bioconcept) and L-glutamine

(64 mmol/L) was initiated 30 hours after inoculation with a constant pump rate of 2.33

mL/h. D-glucose feeding solution (1 mol/L) was fed from the beginning of the culture

in order to keep the glucose concentration in the bioreactor at a setpoint of 20 mM.

The pump rate of the glucose solution was adjusted every 12 hours according to the

measured viable cell concentration and specific glucose consumption rate resulting in

a semi-continuous feeding mode.

Samples were taken twice a day and the viable cell concentration was determined by a

CedeX cell counter (Innovatis) using the trypan blue exclusion method [88].

Concentrations of glucose and lactate were measured immediately using a Super GL

compact instrument (Hitado). Samples were stored at -20 °C for later analysis of

metabolites and IgG1 concentration.

Once a day 2 x 106 cells were fixed by drop wise addition of 2 mL of 75% v/v ice-cold

ethanol. The fixed cells were stored at -20 °C prior to cell cycle analysis.


4.2.4 Cell Cycle Analysis

Prior to flow cytometric analysis, fixed cell samples from the batch culture were

washed with a rinsing solution with 1% w/v bovine serum albumin (Sigma) and 0.01%

w/v sodium azide (Sigma Aldrich) in PBS [182]. DNA staining was performed for 30

minutes in the dark, followed by the addition of 1mL of 50 μg/mL propidium iodide

(PI; Sigma, P4170) and of 100 μg/mL ribonuclease A (Sigma, R4875) in PBS at room

temperature. Fixed fed-batch cell samples were washed twice with 1% w/v fetal

bovine serum (PAN Biotech GmbH) in PBS and re-suspended in 200 µL washing

solution. 100 µL of the cell suspension was incubated with 5 µL FITC conjugated

Mouse Anti-Human Ki-67 antibody (556026, BD Pharmingen) for 30 minutes in the

dark at room temperature. Ki-67 staining is used to determine the proliferating cell

population (G1, S, G2/M). After incubation with Ki-67 antibody, the excess stain was

washed with 1% w/v fetal bovine serum (PAN Biotech GmbH) in PBS. DNA was

stained by the addition of 0.5 mL propidium iodide (1 μg/ml) (P4864, Sigma) and 100

μg/ml ribonuclease A (EN053, Thermo Scientific) in PBS and incubated for 30

minutes in the dark at room temperature.

Flow cytometry was performed on a LSRFortessa (BD) with an excitation wavelength

of 561 nm, or on a Calibur flow cytometer (BD) with an excitation wavelength of 488

nm. The data were collected with a 582/15 filter for PI on FACS LSRFortessa, or a

585/42 filter for PI on FACS Calibur and a 530/30 filter for FITC on FACS Calibur.

To reduce the spillover of FITC into the PI channel, compensation was performed

using a FITC single color control sample. Between 10,000-20,000 events/sample were

collected. The data were analyzed using FlowJo software (Tree Star Inc).

4.2.5 Metabolic Profiling

The extracellular concentrations of ammonium, lactate, glutamine and glucose in the

batch culture were measured from the supernatant samples using a Nova BioProfile

400 Analyzer (Nova Biomedical).

Glutamine concentrations in the fed-batch culture were measured from the supernatant

by HPLC (ZORBAX Eclipse Plus C18 4.6 mm x 150 mm, Agilent Technologies)

according to Agilent’s application note (5990-3283EN). Prior to the HPLC


measurement, supernatant samples were filtered at 4000 rcf, 4 °C for 25 minutes using

a Vivaspin500 (5 kDa) centrifugation filter (Satorius Stedim). Ammonium

concentrations were determined using the L-Glutamine/Ammonia (Rapid) Assay Kit

(K-GLNAM, Megazyme) following the vendor’s protocol.

4.2.6 Extracellular Antibody Quantification

Monoclonal antibody concentrations were determined from the supernatant by

standard affinity chromatography using a PA ImmunoDetection® Sensor Cartridge

(Applied Biosystems) containing immobilized protein G.

4.3 Modeling Aspects

4.3.1 Model Structure

The model was developed on basis of previous models [183–185] that included the

description of the cell cycle phases (XG0, XG1, XS, XG2M). The model is stated in the

form of ordinary differential equations (Eq.4-1 – Eq. 4-4). The transition between the

phases is schematically presented in Figure 4-1.

G1 phase:

𝑑(𝑉𝐺1𝑉)𝑑𝑑

= 2𝑘𝐺2/𝑀−𝐺1𝑉𝐺2/𝑀𝑉 − 𝑘𝐺1−𝑆𝑉𝐺1𝑉 − 𝑘𝐺1−𝐺0𝑉𝐺1𝑉

− 𝑘𝑑𝑉𝐺1𝑉 − 𝐹𝑂𝑈𝑇𝑉𝐺1 (4-1)

S phase: 𝑑(𝑉𝑆𝑉)𝑑𝑑

= 𝑘𝐺1−𝑆𝑚𝑠𝑡𝑟𝑒𝑠𝑠𝑉𝐺1𝑉 − 𝑘𝑆−𝐺2/𝑀𝑉𝑆𝑉 − 𝑘𝑑𝑉𝑆𝑉 − 𝐹𝑂𝑈𝑇𝑉𝑆 (4-2)

G2/M

phase:

𝑑�𝑉𝐺2/𝑀𝑉�𝑑𝑑

= 𝑘𝑆−𝐺2/𝑀𝑉𝑆𝑉 − 𝑘𝐺2/𝑀−𝐺1𝑉𝐺2/𝑀𝑉 − 𝑘𝑑𝑉𝐺2/𝑀𝑉

− 𝐹𝑂𝑈𝑇𝑉𝐺2/𝑀 (4-3)

G0 phase:

𝑑(𝑉𝐺0𝑉)𝑑𝑑

= 𝑘𝐺1−𝑆(1 −𝑚𝑠𝑡𝑟𝑒𝑠𝑠)𝑉𝐺1𝑉 + 𝑘𝐺1−𝐺0𝑉𝐺1𝑉 − 𝑘𝑑𝑉𝐺0𝑉

− 𝐹𝑂𝑈𝑇𝑉𝐺0 (4-4)

Viable

cells: 𝑑(𝑉𝑉𝑉)𝑑𝑑

=𝑑(𝑉𝐺0𝑉)

𝑑𝑑+𝑑(𝑉𝐺1𝑉)

𝑑𝑑+𝑑(𝑉𝑆𝑉)𝑑𝑑

+𝑑(𝑉𝐺2/𝑀𝑉)

𝑑𝑑 (4-5)


Dead cells: 𝑑(𝑉𝑉𝐷)𝑑𝑑

= 𝑉𝑉𝑘𝑑𝑉 − 𝑉𝐷𝑘𝑙𝑦𝑠𝑉 (4-6)

Cell Cycle

fractions: 𝑓𝑖=𝐺0,𝐺1,𝑆,𝐺2/𝑀 =

𝑉𝑖𝑉𝑉

(4-7)

The viable cell population (XV) is calculated as the sum of all cell cycle phases (Eq. 4-

5). The dead cell population (XD) can be estimated from the viable cells (Eq. 4-6).

Thus, the cell cycle fractions (fi) can be calculated according to equation 4-7.

The transition rates between the subpopulations XG1-XS (k1G1-S) (Eq. 4-8), XS-XG2/M (kS-

G2M) (Eq. 4-9), and XG2/M-XG1 (kG2/M-G1) (Eq. 4-10) are stated as previously derived

[183]. The transition to the G0 phase is triggered by two types of stresses: temperature-

dependent stress (kG1-G0) (Eq. 4-11), which is expected to have a constant rate rT when

lowered from 37 °C to 33 °C and/or metabolic stresses (mstress,) (Eq. 4-12). The

metabolic stress is defined as the product of the substrate limitation, toxic metabolite

inhibition, and a certain resistance to such stresses (fres).

G1 to S: kG1−S = �µ +FOUT

V� �

2 − fG1fG1

� (4-8)

S to G2/M: kS−G2/M = �µ +FOUT

V� �

1 + fG2/M

fS� (4-9)

G2/M to

G1: kG2/M−G1 = �µ +

FOUTV

� �1

fG2/M� (4-10)

G1 to G0: kG1−G0 = �T = 33 °C = rTT = 37 °C = 0 � (4-11)

Metabolic

stresses: mstress = � if finh1finh2flim1fres > 1 = 1

else = finh1finh2flim1fres (4-12)

As can be seen from equations 4-8 – 4-10, the growth-associated transitions are

dependent on the growth rate (μ) (Eq. 4-13), which is formulated as a function of the

characteristic times that are required to complete every cell cycle phase (tG1, tS, tG2/M,)

(Eq. 4-14). Several biological contributions were considered to affect these


characteristic times including temperature (kT) (Eq. 4-15) [186], the availability of

substrate (flim1) (Eq. 4-16), and the toxic effect of metabolites (finh1/inh2) (Eq. 4-17)

[125]. In the case of the last two, a Monod kinetics form was assumed [187]. The

death rate (kd,) (Eq. 4-18) is formulated based on previous reports [187], [188],

including glutamine limitation (flim1), and it is considered that death can occur at any of

the cell cycle phases.

Growth

rate: 𝜇 =

ln (2)𝑑𝐺1 + 𝑑𝑆 + 𝑑𝐺2/𝑀

(4-13)

Cell cycle

times: 𝑑𝐺1,𝐺2/𝑀 =

𝑑𝐺1,𝐺2/𝑀 𝑖𝑛𝑖𝑡𝑖𝑎𝑙𝑘𝑇flim1finh2finh1

𝑛 𝑑𝑆 =𝑑𝑆 𝑖𝑛𝑖𝑡𝑖𝑎𝑙 + 𝑘𝑇

flim1finh2finh12 (4-14)

Cell cycle

temperature

factor

𝑘𝑇 = ��𝑇 = 33°𝐶 = 𝑘𝑇𝑇 = 37°𝐶 = 1 � (4-15)

Limiting

functions: flim1 =

[Gln]𝐾𝐺𝑙𝑛 + [Gln] (4-16)

Inhibition

functions: finh1 =𝐾𝐿𝑎𝑐

𝐾𝐿𝑎𝑐 + [𝐿𝑎𝑐] finh2 =𝐾𝐴𝑚𝑛

𝐾𝐴𝑚𝑛 + [𝐴𝑚𝑛] (4-17)

Death rate: 𝑘𝑑 =𝑘𝑑𝑀𝐴𝑋

�1 + flim1𝐾𝑑𝐴𝑚𝑚[𝐴𝑀𝑀]�

(4-18)

The changes in substrate and metabolite concentrations are calculated by performing a

mass balance around the bioreactor while accounting for the cell metabolism [187],

[188]. The bioreactor functions under the assumptions of perfect mixing and operates

under batch and fed-batch mode. The change in volume (V) is therefore related to

feeding (FGlc, FGln) or sampling (Fout) (Eq. 4-19). The change in glutamine

concentration [Gln] is dependent on the glutamine uptake (QGln), glutamine

degradation into ammonia (Kdeg), and the presence of glutamine in the feed ([Glnfeed])

(Eq. 4-20). The glutamine uptake depends on the cell growth rate considering a

specific yield on glutamine (YGln), an uptake limitation function following Monod

kinetics (fupt), and a maintenance term (mGln) [189] (Eq. 4-21). The ammonia


concentration [Amm] is closely linked to the glutamine changes by the yield of

ammonia on the glutamine consumption (YAmm) and the degradation of glutamine (Eq.

4-22), as described above.

Reactor

volume: 𝑑𝑉𝑑𝑑

= 𝐹𝐺𝑙𝑐 + 𝐹𝐺𝑙𝑛 − 𝐹𝑂𝑈𝑇 (4-19)

Glutamine:

𝑑([𝐺𝑙𝑛]𝑉)𝑑𝑑

= −𝑄𝐺𝑙𝑛𝑉𝑉𝑉 − 𝐾𝑑𝑒𝑔[𝐺𝑙𝑛]𝑉 + 𝐹𝐺𝑙𝑛[𝐺𝑙𝑛𝐹𝑒𝑒𝑑]

− 𝐹𝑂𝑈𝑇[𝐺𝑙𝑛] (4-20)

𝑄𝐺𝑙𝑛 = �𝜇𝑌𝐺𝑙𝑛

� fupt + 𝑚𝐺𝑙𝑛 fupt =[Gln]

𝐾𝐺𝑙𝑛 + [Gln] (4-21)

mGln =α1[Gln]α2 + [Gln]

Ammonium: 𝑑([𝐴𝑚𝑛]𝑉)𝑑𝑑

= 𝑄𝐺𝑙𝑛𝑌𝐴𝑚𝑛𝑉𝑉𝑉 + 𝐾𝑑𝑒𝑔[𝐺𝑙𝑛]𝑉 − 𝐹𝑂𝑈𝑇[𝐴𝑚𝑛] (4-22)

The glucose concentration [Glc] (Eq. 4-23) depends on the glucose uptake rate (QGlc),

the glucose maintenance consumption (mGlc), and the glucose concentration in the feed

([Glcfeed]). QGlc has been formulated considering glutamine being an essential

biosynthetic precursor for cells [190], and glucose consumption being temperature-

dependent [166], as well as inhibited by lactate [151]. Equation 4-24 combines those

three effects on glucose uptake (Eq. 4-24). Moreover, it is assumed that cells in the

quiescent state G0 will consume glucose at a maintenance rate, as in this phase there is

minimal cell growth [191]. The lactate balance [Lac] (Eq. 4-25) is connected to

glucose consumption (QGlc, mGlc) by a variable yield of lactate on glucose (YLac)

induced by a feedback inhibition as previously reported [125] (Eq. 4-26).

Glucose:

𝑑([𝐺𝑙𝑐]𝑉)𝑑𝑑

= −𝑄𝐺𝑙𝑐𝑉𝑉(1 − 𝑓𝐺0)𝑉 −𝑚𝐺𝑙𝑐𝑉𝑉𝑓𝐺0𝑉

+ 𝐹𝐺𝑙𝑐[𝐺𝑙𝑐𝐹𝑒𝑒𝑑] − 𝐹𝑂𝑈𝑇[𝐺𝑙𝑐] (4-23)

𝑄𝐺𝑙𝑐 = �𝜇𝑌𝐺𝑙𝑐

� fuptfinh1𝑘𝑇 (4-24)


Lactate:

𝑑([𝐿𝑎𝑐]𝑉)𝑑𝑑

= 𝑌𝐿𝑎𝑐𝑄𝐺𝑙𝑐𝑉𝑉(1 − 𝑓𝐺0)𝑉 − 𝑌𝐿𝑎𝑐𝑚𝐺𝑙𝑐𝑉𝑉𝑓𝐺0𝑉

− 𝐹𝑂𝑈𝑇[𝐿𝑎𝑐] (4-25)

𝑌𝐿𝑎𝑐 = 𝑌𝐿𝑎𝑐 𝑖𝑛𝑖𝑡𝑖𝑎𝑙finh1(𝑚+𝑘𝑇) (4-26)

The change in monoclonal antibody concentration ([mAb]) (Eq. 4-27) is growth

associated and separated in antibody specific cell cycle productivity rates (qi).

Antibody:

𝑑([𝑚𝐴𝑏]𝑉)𝑑𝑑

= 𝜇�𝑞𝐺1/𝐺0(𝑉𝐺1 + 𝑉𝐺0) + 𝑞𝑆𝑉𝑆 + 𝑞𝐺2/𝑀𝑉𝐺2/𝑀�

− 𝐹𝑂𝑈𝑇[𝑚𝐴𝑏] (4-27)

4.3.2 Temperature Dependent Population Transition

An important aspect of the proposed population modelling is the cell behavior at the

time of temperature transition. Given that the cell growth (μ) has been formulated as

being temperature-dependent, cell growth transition takes place when the reactor

temperature is switched (i.e. from normal conditions at 37 °C to mild hypothermia at

33 °C). This formulation reflects the dynamic and adaptive process of cell growth.

Therefore, the temperature shift is not expected to be reflected instantaneously in the

cell growth. Such behavior has been previously reported after a temperature shift as a

gradual decrease in the S phase [163], a partial blockage in the G2/M [162] and an

eventual increase in the G1 cell phase [160]. In order to capture such dynamics, it is

herein assumed that after a temperature shift the existing cell population will

proliferate at a transitory rate, where the G1 and S phase will remain at a fast rate (i.e.

37 °C) and G2/M will be affected immediately (i.e. 33 °C). However, every new born

cell of the existing population will proliferate under new reactor temperature (Figure

4-1). Therefore, the existing population will be metabolically active during transition

until this population disappears either because it completes the cell cycle, gets arrested

at G0 or dies. The cells in G0 will continue to consume glutamine at the transition

conditions until the end of the culture. Thereby cells in G0 that are still active in

biosynthesis – since glutamine plays a key role in cell metabolism [192–194] – but do

not grow or commit to apoptosis (an energy intensive process) [195], are considered.


On contrary, the metabolism of all newly born cells will be immediately affected and

is therefore calculated considering the new reactor temperature.

Figure 4-1: Schematic representation of the cell cycle model.

4.3.3 Global Sensitivity Analysis and Parameter Estimation

After defining the model type and structure, the uncertainty of the parameters in the

model output was studied as outlined in the mathematical modelling development

framework [196–198]. In order to systematically evaluate the output variability, global

sensitivity analysis (GSA) was performed. The full range of 27 parameters was varied

±50% from their nominal value and evaluated over 40 000 scenarios generated using

the sobolset function in Matlab. Each scenario was evaluated using the go:MATLAB

tool from gPROMs® providing the link between both softwares. The sensitivity

indexes (SIs) were calculated using the GUI-HDMR software [199].

Six outputs were evaluated (XV, fG1, fS, fG2M, fG0 and mAb) at five time points (12, 24,

36, 48, 96 and 144h) of a fed-batch culture. The SIs are the result of the GSA and vary

between 0 and 1, with 0 being not significant. The cut-off value to determine the

significant parameters was set at 0.1 since 10% can be attributed to experimental errors

[200]. Therefore, parameters that are not identified as significant can be fixed at their

nominal value and the significant parameters should be re-estimated using

experimental data to minimize the output uncertainty.


4.4 Results

4.4.1 Global Sensitivity Analysis and Parameter Estimation

Global sensitivity analysis allows studying the output sensitivity to the variation of the

model input parameters. This process can aid the model development as the most

significant parameters are identified (and re-estimated), whereas the non-significant

parameters can be fixed at their initial value (from literature or assumed). Herein, the

majority of the nominal values were set to literature values as previously reported

[201]. The time points for the GSA were selected in order to evaluate different phases

of a fed-batch process including a temperature shift during the exponential growth

phase at 40 hours. Specifically the time points from the lag phase (day 0.5), early

exponential (day 1), exponential phase (day 1.5), exponential after the temperature

shift (day 2), stationary phase (day 4) and decline phase (day 6) were evaluated. GSA

results in Figure 4-2 are represented at 3 time points to illustrate the time evolution of

the sensitivity index (SI) using six output parameters (XV, mAb, fG0, fG1, fS and fG2/M).

The analysis resulted in the identification of 9 significant parameters (i.e. SI above 0.1

as defined). The SIs for the output of viable cell concentration (XV) for the different

time points are shown (Figure 4-2A). XV was sensitive to the variation of the cell cycle

times (tSinitial, tG2/Minitial) at early stages of the culture and its sensitivity decreased in

time (though remaining significant). This can be attributed to the connection of

characteristic cell cycle times to the cell growth rate. In contrast, the significance of

the resistance to the metabolic stresses (fres) increased along the culture time, reaching

significance at the stationary growth phase. The sensitivity towards the yield of

biomass on glutamine (YGln) increased at early stages, peaked at day 1.5 (data point not

shown), and decreased towards the end of the culture. Comparable results were

observed for the output considering mAb concentration (Figure 4-2B). In addition to

the cell cycle times, the mAb-specific productivity for the G1/G0 phase (qG1/G0) and S

phase (qS) were found to be significant at early stages of the culture, thereby showing

the importance of defining cell cycle specific production rates. The cell cycle

temperature factor (kT) and the sensitivity of cell cycle times to lactate inhibition (n)

were significant during the stationary and decline growth phase. In these phases lactate

is likely to reach inhibitory concentrations. The G0 cell fraction (fG0) output was highly


sensitive to the resistance to metabolic stresses throughout the culture (Figure 4-2C).

Similarly, other cell cycle fractions showed an increasing sensitivity towards this

parameter along the culture. The G1 cell fraction (fG1) was sensitive at early stages to

the initial S phase cell cycle time and at late stages to the G2/M cell cycle time, as well

as to the initial G1 cell cycle time (tG1initial) throughout culture time (Figure 4-2D). The

S phase cell fraction (fS) was sensitive to the initial S phase cell cycle time and to the

G1 cell cycle time (Figure 4-2E), and the G2/M fraction (fG2/M) was sensitive to the

initial S phase cell cycle time and to the G2/M cell cycle time (Figure 4-2F). The three

cell cycle fractions (fG1, fS, fG2/M) associated to proliferation showed sensitivity to the

cell cycle temperature factor (kT) at the time points after the temperature shift.

The association of cell cycle time estimates to all studied GSA outputs stresses the

importance of considering the cell cycle heterogeneity. Similarly, the cell cycle

temperature factor was identified significant for the outputs of proliferating cell cycle

phases, XV and the antibody production. Such a dependency supports the importance

and critical aspect of finding a balance between growth and productivity when using

temperature to optimize fed-batch processes. Particularly, if one or more of the cell

cycle phases are associated to productivity, as obtained from GSA.

Based on these results, significant parameters were re-estimated with the gPROMS

parameter estimator (based on the maximum likelihood formulation), using

experimental data from the 37 °C and 33 °C batch and 37 °C and 33 °C fed-batch

cultivations. The chosen experimental data sets assess apart from two temperature

setpoints (37 °C and 33 °C), the effect of metabolic stresses, which are partially also

identified as sensitive parameters. Batch cultures cover nutrient depletion, that is

responsible for triggering changes in the cell behavior [202]. Similarly, fed-batch

cultures capture the dynamics of toxic metabolite accumulation due to feeding

strategies which are of industrial relevance. The new parameter set, estimated from 4

experimental data sets is reported with the 95% confidence intervals in the Appendix

(Table A 6). The largest confidence intervals are reported for cell cycle specific

productivities (qG1/G0, qS) and reflect a weak estimation of these parameters which is

due to the growth-associated antibody production of HFN 7.1 cells.


Figure 4-2: Sensitivity indices for parameters at simulation times (day 1, 4 and 6).

(A) viable cell concentration (Xv), (B) mAb titer, (C) G0 cell fraction (fG0), (D) G1

cell fraction (fG1) (E) S cell fraction (fS) (F) G2/M cell fraction (fG2/M). The

nomenclature is equal as in Eq. 4-1 – 4-27.

Kd (max

)K (ly

s)

Kd (Amm)

k (T)r (T

)

K (deg

)

K (Gln)

K (Lac

)

K (Amm)

K (upt)

Y (Glc)

Y (Gln)

Y (Lac

)

Y (Amm)

m (Glc)

alpha

1

alpha

2 n mf (r

es)

t (G1 i

nitial)

t (G2/M

initia

l)

t (S in

itial)

q (G1/G

0)q (

S)

q (G2/M

)0.00.10.20.30.40.50.60.70.80.9

day 1 day 4 day 6

Sens

itivity

Inde

x

A

Kd (max

)K (ly

s)

Kd (Amm)

k (T)r (T

)

K (deg

)

K (Gln)

K (Lac

)

K (Amm)

K (upt)

Y (Glc)

Y (Gln)

Y (Lac

)

Y (Amm)

m (Glc)

alpha

1

alpha

2 n mf (r

es)

t (G1 i

nitial)

t (G2/M

initia

l)

t (S in

itial)

q (G1/G

0)q (

S)

q (G2/M

)0.00.10.20.30.40.50.60.70.80.9B

day 1 day 4 day 6

Se

nsitiv

ity In

dex

Kd (max

)K (ly

s)

Kd (Amm)

k (T)r (T

)

K (deg

)

K (Gln)

K (Lac

)

K (Amm)

K (upt)

Y (Glc)

Y (Gln)

Y (Lac

)

Y (Amm)

m (Glc)

alpha

1

alpha

2 n mf (r

es)

t (G1 i

nitial)

t (G2/M

initia

l)

t (S in

itial)

q (G1/G

0)q (

S)

q (G2/M

)0.00.10.20.30.40.50.60.70.80.9C

day 1 day 4 day 6

Sens

itivity

Inde

x

Kd (max

)K (ly

s)

Kd (Amm)

k (T)r (T

)

K (deg

)

K (Gln)

K (Lac

)

K (Amm)

K (upt)

Y (Glc)

Y (Gln)

Y (Lac

)

Y (Amm)

m (Glc)

alpha

1

alpha

2 n mf (r

es)

t (G1 i

nitial)

t (G2/M

initia

l)

t (S in

itial)

q (G1/G

0)q (

S)

q (G2/M

)0.00.10.20.30.40.50.60.70.80.9D

day 1 day 4 day 6

Sens

itivity

Inde

x

Kd (max

)K (ly

s)

Kd (Amm)

k (T)r (T

)

K (deg

)

K (Gln)

K (Lac

)

K (Amm)

K (upt)

Y (Glc)

Y (Gln)

Y (Lac

)

Y (Amm)

m (Glc)

alpha

1

alpha

2 n mf (r

es)

t (G1 i

nitial)

t (G2/M

initia

l)

t (S in

itial)

q (G1/G

0)q (

S)

q (G2/M

)0.00.10.20.30.40.50.60.70.80.9E

day 1 day 4 day 6

Sens

itivity

Inde

x

Kd (max

)K (ly

s)

Kd (Amm)

k (T)r (T

)

K (deg

)

K (Gln)

K (Lac

)

K (Amm)

K (upt)

Y (Glc)

Y (Gln)

Y (Lac

)

Y (Amm)

m (Glc)

alpha

1

alpha

2 n mf (r

es)

t (G1 i

nitial)

t (G2/M

initia

l)

t (S in

itial)

q (G1/G

0)q (

S)

q (G2/M

)0.00.10.20.30.40.50.60.70.80.9F

day 1 day 4 day 6

Sens

itivity

Inde

x


4.4.2 Temperature Shifted Batch Cultures

The 37 °C batch culture performed in shake flasks under uncontrolled pH conditions

shows the typical batch operation profiles (Figure 4-3). Briefly, the viable cell

concentration (XV) increases exponentially after a short period of adaption (lag phase),

and peaks around the time when one of the key substrates (in this case glutamine GLN)

is exhausted. After glutamine depletion, the proliferation slows down and a significant

death rate is observed (decline phase), coinciding with the increase of the lumped

G1/G0 population and the corresponding decrease of the S and G2/M fractions. The

toxic metabolite production (LAC, AMM) is correlated to the substrate consumption

(GLC, GLN), as well as the antibody concentration (mAb) to the viable cell

concentration.

The experimental data of the 37 °C batch culture are used for parameter estimation in

the formulated model. The fit of the 37 °C batch culture is shown in Figure 4-3. The

model describes the viable cell concentration and viability trends, although there are

small deviations between absolute values of viability in the model and the experiment.

Regarding cell cycle fractions, the change in the cell cycle trend after the exhaustion of

glutamine was captured. At the same time, an initial discrepancy in the model

description for the S and G2/M cell cycle fractions is observed at early stages (~0.5

days). Substrate profiles and lactate production are well predicted by the model,

whereas the ammonia profile is consistently over-estimated. The yield of ammonia on

glutamine (Yamm) is considered constant in the model. However, it has been shown that

Yamm depends on temperature and pH [93], [203]. This can serve as a reason for the

observed discrepancies in the ammonia profile particularly in pH-uncontrolled shake-

flask cultures. Nevertheless, the overall trend describing ammonia production is

correct. The antibody titer is satisfactorily described up to around 2 days of culture.

After this point the model starts to slightly deviate from the experimental data.


Figure 4-3: Shake flask batch culture at 37 °C. Experimental results of viable cell

concentration (Xv), viability, antibody concentration (mAb), glucose (GLC),

lactate (LAC), glutamine (GLN) and ammonia (AMM) concentrations, and cell

cycle fractions G0/G1, S, G2/M compared to model results (solid line: Xv, LAC,

AMM, G0/G1, S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN).

The 33 °C batch cultivation includes the effect of temperature reduction at the

beginning of the culture (0 days) (Figure 4-4). The trends observed for viable cell

concentration XV, viability, substrates (GLC, GLN) and toxic metabolites (LAC, AMM)

are comparable to the 37 °C culture. However, a significantly lower maximum XV is

achieved (1.8-fold smaller compared to the 37 °C batch), with a corresponding lower

final antibody concentration (1.4-fold lower compared to the 37 °C batch) despite

0

20

40

60

80

100

0.0

0.2

0.4

0.6

10

20

30

0.0

0.2

0.4

0.6

0 1 2 30

1

2

3

0 1 2 30.0

0.2

0.4

0.6

Xv viability mAb GLC LAC GLN AMMG0/G1 S G2/M

Xv (1

05 cel

ls/ m

L), V

iabi

lity (%

)

0.0

0.2

0.4

0.6

0.8

1.0

Antib

ody

Conc

. (g/

L)

G0/

G1

Frac

tion

GLC

, LAC

(mm

ol/ L

)

S Fr

actio

n

Time (d)

GLN

, AM

M (m

mol

/ L)

Time (d)

G2/

M F

ract

ion


having a 1.5-fold longer culture time. The substrate/toxic metabolite profiles still have

a good correlation, although the lactate production was further increased at late stages

of the culture even at low glucose consumption. The cell cycle fractions showed minor

changes during the first 0.5 day of culture. Around day 1 a significant change in the

cell cycle distribution was observed, which coincided with the initiation of exponential

growth, and was followed by rather constant cell cycle fractions until the end of the

culture. After glutamine depletion, an increased dispersion in the G1/G0 and G2/M

fractions was observed.

The experimental trends of the 33 °C batch culture that were likewise used for

parameter estimation are fitted by the proposed model (Figure 4-4). Maximum XV and

the timing at which it was achieved (day 3) are properly captured.

However, some differences are observed at early stages corresponding to a pronounced

lag phase, as well as during the decline phase. The latter is also reflected in the over-

estimation of viability compared to the experimental data. The substrate profiles (GLC,

GLN) together with the AMM and antibody profiles are satisfactorily modeled

throughout the culture. The lactate profile shows a good fit up to 3 days, even though

the experimentally observed constant increase cannot be captured by the model. It is

assumed that uncontrolled pH-conditions in the shake flask system are responsible for

the observed differences. The sharp changes in the cell cycle distribution observed at

day 1 are reflected in the model simulations, however appear earlier compared to the

experimental cell cycle fractions. The model formulation is unable to capture the

prolonged lag phase that seemed to take place while the cells were adapting to the new

temperature condition. This is due to the fact that cells are considered homogenous

within a cell cycle phase. Such differences can be attributed to specific cell cycle

transitions of cohorts or partially synchronized populations. In order to model these

particular cell cycle changes, another internal coordinate needs to be employed in a

segregated formulation such as population balance models. Additional discrepancies

appear in the simulations of lumped G1/G0 and G2/M fractions at the late stage of the

cell culture.





cycle fractions G0/G1, S, G2/M compared to model results (solid line: Xv, LAC,

AMM, G0/G1, S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN).

4.4.3 Temperature Shifted Fed-Batch Cultures

Standard 37 °C fed-batch cultures were carried out as duplicates in controlled

bioreactors, and were used for parameter estimation in addition to batch cultures

(Figure 4-5). The increase in viable cell concentration by 2.1-fold compared to a

standard batch culture is attributed to a continuous feed of glucose, amino acids

0

20

40

60

80

100

0.0

0.2

0.4

0.6

10

20

30

0.0

0.2

0.4

0.6

0 1 2 3 4 50

1

2

3

0 1 2 3 4 50.0

0.2

0.4

0.6

Xv viability mAb GLC LAC GLN AMM G0/G1 S G2/M

Xv (1

05 cel

ls/ m

L), V

iabi

lity (%

)

0.0

0.2

0.4

0.6

0.8

1.0

Antib

ody

Conc

. (g/

L)

G0/

G1

Frac

tion

GLC

, LAC

(mm

ol/ L

)

S Fr

actio

n

Time (d)

GLN

, AM

M (m

mol

/ L)

Time (d)

G2/

M F

ract

ion


(including glutamine), vitamins and trace elements. Avoidance of key substrate

exhaustion (GLC, GLN) results in a prolonged stationary phase for several days at

maintained high viability. Glucose and glutamine are consumed throughout the

culture, although at a slower rate towards the end. The feeding strategy results in high

toxic metabolite (LAC, AMM) levels [125]. In particular, ammonia reaches 5.5-fold

higher concentrations compared to the 37 °C batch. Around day 3, ammonia exceeds

toxic concentrations and simultaneously lactate concentration plateaus. As a result, an

arrest in the G0 cell cycle phase is triggered. The G0 fraction increases steadily until

the end of the cultivation with the accumulation of lactate and ammonia over this

period. The observed experimental variation in the lactate profile is attributed to pH

control by CO2 sparging and varying amount of base (sodium hydroxide) added in the

bioreactor control strategy. Similarly the relationship of lactate production and base

addition in pH-controlled bioreactors accompanied by high osmolarity, has been

shown in industrial fed-batch processes to result in big differences in the final lactate

profile, without affecting the growth profile [204].

The model fit for the fed-batch culture at 37 °C is provided in Figure 4-5. The

increase, peak and decline of XV are closely fitted. While the viability decline in the

late stage of the culture is underestimated in batch, it is better captured in fed-batch.

This difference in the model performance can be explained by the different

experimental setups used for batch and fed-batch cultures. Initial comparative 37 °C

batch cultures performed in both cultivation systems showed good agreement, except

for cell viability profiles (data not shown). In general, viability remains higher in the

controlled bioreactor. This observation is attributed to the different cell counting

methods used to determine viable cell concentrations. In batch culture, viable cells

were counted manually, whereas in fed-batch cells they were determined automatically

with a lower cell cut-off size set at 9 µm. Moreover, hydromechanical stress rates are

considerably smaller in shake flask cultures [205] and therefore the variations in

viability can also be attributed to differences in cell lysis within the two experimental

setups, which are not considered in the model.


Figure 4-5: Controlled fed-batch culture at 37 °C. Experimental results of viable

cell concentration (Xv), viability, antibody concentration (mAb), glucose (GLC),


cycle fractions G0, G1, S, G2/M compared to model results (solid line: Xv, LAC,

AMM, G1, S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN, G0

fraction).

Regarding the trend of cell cycle fractions (fG0, fG1, fS, fG2/M) including the timing of G0

increase triggered by metabolite accumulation (LAC, AMM), it is well described in the

model. Taking decreased substrate uptake rates along the culture into consideration, a

good description of the glutamine profile was achieved, whereas the glucose

0

20

40

60

80

100

0.0

0.2

0.4

0.6

10

20

30

40

50

60

0.0

0.2

0.4

0.6

0 1 2 3 4 5 6 7 802468

101214

0 1 2 3 4 5 6 7 80.0

0.2

0.4

0.6

Xv viability mAb GLC LAC GLN AMMG0 G1 S G2/M

Xv (1

05 cel

ls/ m

L), V

iabi

lity (%

)

0.0

0.2

0.4

0.6

0.8

1.0

Antib

ody

Conc

. (g/

L)

G0,

G1

Frac

tion

GLC

, LAC

(mm

ol/ L

)

S Fr

actio

n

Time (d)

GLN

, AM

M (m

mol

/ L)

Time (d)

G2/

M F

ract

ion


consumption seems to be even further decreased in the last two days of culture.

Profiles of AMM and antibody show a good fit, with some differences in antibody

concentration on the last culture day. The decrease in titer can be attributed to protein

degradation due to increased proteolytic activity towards the end of the culture [206].

The averaged lactate profile is also adequately modeled. Since experimentally

observed differences in the lactate profile did not significantly affect other model

outputs (i.e. viable cells, viability, mAb titer, cell cycle distribution), it was sufficient

to use the averaged lactate concentration as a triggering range.

A 33 °C fed-batch culture with initial temperature shift (at time point 0) represents a

further experimental data set for parameter estimation (Figure 4-6). A significantly

lower cell concentration XV is reached (2.6-fold lower when compared to the 37 °C

fed-batch culture). Due to decreased glucose consumption rates, especially towards the

end of the culture, a lower lactate concentration is obtained. However, ammonia

concentrations reach levels comparable to those at 37 °C, while less glutamine is

consumed. Most importantly, a cell arrest towards the G0 cell cycle fraction is evident

from the beginning of the culture with a steady increase over the remaining cultivation

time. Thus, cell cycle arrest is, in addition to metabolic control, also triggered by

hypothermia.

The fit of the 33 °C fed-batch culture is shown in Figure 4-6. XV is accurately fitted and

only slightly over-estimated during the decline phase. Cell viability, antibody

concentration and cell cycle fractions are well described. Substrates (GLC, GLN) as

well as the LAC profile show acceptable agreement, while the ammonia simulation

follows the data trend but is highly under-estimated. Ammonia production was

assumed to be solely dependent on glutamine. However, ammonia is also a product of

asparagine consumption [37], an amino acid present in high concentration in the feed.

In addition, amino acid consumption has been shown to be affected by temperature

[207], thus providing an explanation for the discrepancies.


Figure 4-6: Controlled fed-batch culture shifted to 33 °C at culture start.

Experimental results of viable cell concentration (Xv), viability, antibody

concentration (mAb), glucose (GLC), lactate (LAC), glutamine (GLN) and

ammonia (AMM) concentrations, and cell cycle fractions G0, G1, S, G2/M

compared to model results (solid line: Xv, LAC, AMM, G1, S, G2/M fractions;

dashed lines: viability, mAb, GLC, GLN, G0 fraction).

4.4.4 Prediction of Temperature Shift Time in Fed-Batch Cultures

The model was used to predict two fed-batch experiments with temperature shifts from

37 °C to 33 °C during the exponential growth phase (30 hours) and in the early

stationary phase (76 hours).

0

20

40

60

80

100

0.0

0.2

0.4

0.6

10

20

30

0.0

0.2

0.4

0.6

0 1 2 3 4 5 6 7 802468

101214

0 1 2 3 4 5 6 7 80.0

0.2

0.4

0.6


Xv (1

05 cel

ls/ m

L), V

iabi

lity (%

)

0.0

0.2

0.4

0.6

0.8

1.0

Antib

ody

Conc

. (g/

L)

G0,

G1

Frac

tion

GLC

, LAC

(mm

ol/ L

)

S Fr

actio

n

Time (d)

GLN

, AM

M (m

mol

/ L)

Time (d)

G2/

M F

ract

ion


Figure 4-7: Controlled fed-batch culture with temperature shift to 33 °C after 30

h. Shift is indicated by dotted line. Experimental results of viable cell



cycle fractions G0, G1, S, G2/M compared to model prediction (solid line: Xv,

LAC, AMM, G1, S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN, G0).

Temperature reduction after 30 hours is overall well predicted by the model (Figure

4-7). The profile of Xv follows the trend with minor discrepancies. The exponential

phase, during which the temperature shift is performed, shows an over-estimation,

whereas the prolonged stationary/decline phase is slightly under-estimated. In fact, the

resulting viable cell concentration is comparable to the 37 °C fed-batch culture (85%).

0

20

40

60

80

100

0.0

0.2

0.4

0.6

10

20

30

0.0

0.2

0.4

0.6

0 1 2 3 4 5 6 7 802468

101214

0 1 2 3 4 5 6 7 80.0

0.2

0.4

0.6


Xv (1

05 cel

ls/ m

L), V

iabi

lity (%

)

0.0

0.2

0.4

0.6

0.8

1.0

Antib

ody

Conc

. (g/

L)

G0,

G1

Frac

tion

GLC

, LAC

(mm

ol/ L

)

S Fr

actio

n

Time (d)

GLN

, AM

M (m

mol

/ L)

Time (d)

G2/

M F

ract

ion


The antibody concentration is slightly underestimated after day 4.5, but the prediction

of decreased final mAb concentration compared to 37 °C (90%) is met. Likewise, the

decrease in GLC consumption accompanied by a lower LAC production, as well as

ammonia production expected to remain high as in previous fed-batch cultures, is well

predicted. Glutamine consumption properly predicted until day 2, shows a systematic

difference between the model and the data suggesting that glutamine is consumed at a

higher rate than predicted. The increase in quiescent cells due to toxic metabolite

accumulation and temperature shift is represented by the increase of the G0 fraction at

the time of the temperature shift and further rise throughout the culture. Furthermore,

the shift in temperature is followed by changes in the proliferating cell cycle fractions

such as an increase in the G1 cell fraction. Initial discrepancies in the S and G2/M

fraction are in the same range as the ones observed in the 37 °C standard fed-batch

culture, even though the general trend is captured.

The second model prediction considers a fed-batch culture with temperature shift from

37 °C to 33 °C at 76 hours (Figure 4-8). A further delay of the temperature shift time is

expected to result in higher cell concentrations compared to the earlier shift time

points. Experimental data show that the viable cell concentration is affected instantly

by the temperature shift, reaching 95% of XV at standard 37 °C, and the model matches

the profile during all growth phases to satisfaction. Viability and antibody

concentration profiles show good agreement between the simulation and the data.

Final antibody titer remains at 90% of the 37 °C standard fed-batch culture and does

not show further improvement. The glucose consumption trend is well predicted, even

though after the temperature shift glucose consumption seems to be systematically

under-predicted. On the contrary, the glutamine profile shows systematic differences

by under-estimating glutamine consumption after the temperature shift. Toxic

metabolite profiles followed the prediction with similar differences as observed before.

The G0 fraction starts to increase at day 2 due to metabolic effectors and is further

triggered by the temperature shift. The model is able to predict the behavior of cell

arrest. Solely an experimentally observed decrease in G0 during the last 2 day of the

cultivation accompanied by an increase in G1 fraction is not represented in the model.


As in the previous cases, soon after the temperature shift the sharp change in the cell

cycle fractions is in good agreement to the model prediction.

Figure 4-8: Temperature shifted controlled fed-batch culture to 33 °C after 76 h.

Shift in indicated by a dotted line. Experimental results of viable cell



cycle fractions G0, G1, S, G2/M compared to model prediction (solid line: Xv,

LAC, AMM, G1, S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN, G0

fraction).

0

20

40

60

80

100

0.0

0.2

0.4

0.6

10

20

30

40

0.0

0.2

0.4

0.6

0 1 2 3 4 5 6 7 802468

101214

0 1 2 3 4 5 6 7 80.0

0.2

0.4

0.6


Xv (1

05 cel

ls/ m

L), V

iabi

lity (%

)

0.0

0.2

0.4

0.6

0.8

1.0

Antib

ody

Conc

. (g/

L)

G0,

G1

Frac

tion

GLC

, LAC

(mm

ol/ L

)

S Fr

actio

n

Time (d)

GLN

, AM

M (m

mol

/ L)

Time (d)

G2/

M F

ract

ion


4.5 Discussion

The presented model serves as a tool to study the temperature effect in mammalian cell

culture systems. The impact of temperature shifts is addressed by considering an

unstructured cell cycle model that captures the basic heterogeneity of the system. In

addition to temperature effects, metabolic stresses, such as substrate depletion or toxic

metabolite accumulation, are considered on the cell cycle level to account for different

process modes (batch and fed-batch). This could be accomplished by providing

experimental measurements of the quiescent cell cycle phase in addition to commonly

measured proliferating cell cycle phases for model development. Thus, the model was

formulated using first principles and further developed by using the existing model

development framework [196–198]. The framework describes a procedure that

includes the identification of sensitive parameters and their re-estimation using

experimental data from batch and fed-batch cultures at two different temperature

setpoints. When applying this model to other cell lines, such parameters can easily be

estimated from small-scale cultures used for pre-studies in bioprocess development.

The proposed procedure, validated in HFN 7.1 hybridoma cell culture was able to

capture the experimental trends of batch and fed-batch cultures, especially considering

the relationship between temperature, growth and productivity as the main focus.

Minor discrepancies are mostly observed on metabolic level. In particular, the glucose

uptake rate seemed to be even further decreased along the culture than predicted by the

model. Similarly, ammonia production was considered to only depend on glutamine

consumption and degradation, which resulted in an underestimation of experimental

results. In order to achieve a better representation of substrate and metabolite profiles,

metabolic relations need to be further redefined. Nonetheless, the model was able to

successfully predict two temperature shift scenarios and to capture most of the

experimentally observed trends, not only on growth and metabolism, but also at the

cell cycle level. Most importantly, the model correctly captured the G0 cell cycle phase

arrest. The importance of the cell arrest is reflected in a number of studies [95], [161],

[163], [173], [208] that have used this phase to increase productivity in commercially

relevant cell lines. Moreover, the relevance of considering the cell cycle segregation


and the temperature growth regulation is of industrial interest as this is often the

method of choice for improving productivity [186].

HFN 7.1 cells are characterized by growth-associated antibody production reflected in

a weak estimation of cell cycle specific productivities (qG1/G0, qS). On the contrary,

productivity in other cell lines has been associated to specific cell cycle phases [172],

[173], [209]. In such cases, the developed model is readily applicable and provides a

means to systematically optimize the cell cycle-to-productivity-relationship. However,

the formulation of an industrially significant cell cycle model is not limited to

temperature dependency, but also to the inclusion of biologically relevant events (e.g.

apoptosis, product quality).

This model approach can be employed to systematically study and optimize

temperature shift strategies. Particularly, the use of computational modeling is a

complementary tool that needs to be adopted in biochemical processes [210]. The

advantage of the presented approach over the commonly adopted design of experiment

(DoE) approach relies on the ability to capture the complex dynamic relations of the

mammalian culture system.

4.6 Conclusions

The importance of the results presented herein lies in the development of relevant

modeling approaches and a procedure to assist mammalian cell culture process

development. In particular, time points of temperature shifts can be predicted in fed-

batch culture in order to optimize productivity. The development of an unstructured

model of metabolic- and temperature-dependent cell cycle arrest is validated in

hybridoma batch and fed-batch cultures. Through sensitivity analysis and parameter

estimation using two batch and two fed-batch experimental data sets 9 sensitive

parameters are re-estimated and used to predict different temperature shift time points

in fed-batch culture. This model formulation allows studying complex processes of

biological systems with the experimental tools currently available.


4.7 Remark

The work presented in this chapter has been performed in collaboration with Prof.

Mantalaris’ group from Imperial College London. The model has been developed in

the Mantalaris group and batch cultivations have been performed in their group. Fed-

batch experiments including cell cycle analysis have been performed in the Morbidelli

group at ETH Zurich.

77

5 Concluding Remarks

Through this work, a better understanding of the influence of process parameters on

cultured cells and their expressed proteins has been obtained.

The first two studies focused on experimental approaches that can be applied to

optimize process conditions with the aim of obtaining a comparable cell behavior

regardless of bioreactor scale or aiming in desired product quality. In particular, pH

was identified as a crucial parameter, as it can be used to modify mAb glycosylation

and control lactate production and consumption. The effect of pH on glycosylation

was evaluated in controlled batch culture using MALDI-TOF and HPLC-based

glycosylation analysis. The redistribution of the cell metabolism due to pH alteration

was identified by flux balance analysis and validated by gene expression analysis. The

calculated flux distribution provides an insight into pathway utilization and energy

generation in cell metabolism. Due to the difficulties regarding pH control in large-

scale bioreactors, the findings of both studies explain observed changes in

glycosylation and metabolism during scale-up and can be further exploited for process

control and optimization.

In addition to experimental investigations, mathematical models can support process

optimization by linking complex biological system to process conditions such as

hypothermia. The last study described a mathematical model that predicts the time

point of temperature shifts on the level of cell proliferation and arrest. In addition to

temperature-dependent cell cycle arrest, the model is validated against metabolite-

dependent arrest and is applicable to different culture modes. The aim of such a model

is to achieve a good balance of cell growth and productivity in fed-batch culture.

The presented studies fit well in the scope of Quality by Design (QbD) principles,

where process parameters are investigated in order to identify their impact on critical

quality attributes. In this context, the first study assessed ranges of different physical

and chemical stress parameters representing common bioprocess conditions and

defined their ability to maintain a given glycosylation pattern. Through this analysis,

certain parameters such as sparging rate could be excluded from being qualified as


critical, whereas osmolarity only became important once combined with pH. The

presented methodology could in future be further applied in a high-throughput setting

for studying additional parameters such as CO2, as well as for experiments where

multiple process parameters are simultaneously changed. In this way a broad range of

process parameters could be defined as truly critical for mAb glycosylation.

The use of a QbD-based approach to assess environmental conditions is gaining

increasing importance due to the emergence of biosimilar drugs. In the case of

biosimilars, the originator drug defines the reference quality attributes such as the

glycosylation pattern. Experimental or model-based approaches such as the ones

presented here are therefore becoming an integral part of bioprocess development.

79

6 Outlook

The experimental and model-based approaches that have been presented in this thesis

were tested with one cellular system, a murine hybridoma cell line producing a

monoclonal antibody. In this way the impact of multiple process parameters on mAb

glycosylation, cellular growth and metabolism could be compared among each other.

In order to test for the generality of the approaches, the developed methodologies

require further application on multiple cellular systems, as well as on one defined cell

line producing different glycoproteins. In addition to hybridoma cells, it would be

beneficial to apply the presented procedures to CHO cells, as the most commonly used

mammalian host cells for monoclonal antibody production in industry.

In this way more general conclusions on process parameters affecting glycosylation

and on the degree of the observed changes in glycosylation, as presented in chapter 2

could be drawn. In addition, an extension of the study could include a separate

investigation of glycosylation control by nutrients and feeds using the same cellular

system. The outcome of this study could then be compared to the impact of process

parameters. This would help defining general and complete process conditions and

process parameters that can be applied as control variables to obtain the desired

glycosylation profile following the QbD approach.

The methodology used to study pH alterations on cellular metabolism, as described in

chapter 3, would also benefit from a general approach including different cell lines.

This would provide an insight into whether the flux balance model together with the

chosen objective function hold for the cultivation of multiple mammalian cell lines in

bioreactor systems. In order to confirm the calculated pathway utilization at different

pH conditions, the flux distribution requires experimental validation by isotope

measurements using 13C-metabolic flux analysis. Such measurements should however

be performed in a different, more appropriate experimental setup such as a shake flask

system, since they are not feasible in bioreactors.

The cell cycle model framework together with the described procedure to re-estimate

model parameters presented in chapter 4 will ideally be tested using a cell line that

80 CHAPTER 6: APPENDIX

shows growth-independent productivity. Such cell lines are characterized by a

production phase that appears during stationary growth and would therefore be highly

related to specific cell cycle phases. Due to the different intrinsic cellular

characteristics, such a system would provide an ideal validation of the developed

model framework.

To conclude, QbD-based studies profit from high-throughput and automatization

allowing the simultaneous investigation of multiple process conditions with high

reproducibility. Thus, it will be beneficial to perform future studies in novel high-

throughput bioreactor systems.

81

7 Appendix

7.1 Appendix for “Evaluating the Impact of Cell Culture Process

Parameters on Monoclonal Antibody N-Glycosylation”

7.1.1 Deglycosylation by PNGase F and 2-AB Labeling

N-linked oligosaccharides of each mAb preparation were enzymatically released by

addition of 2 µL of N-glycosidase F (PNGase F glycerol free) (New England Biolabs)

in the supplied buffer for 14 hours at 37 °C. Prior to deglycosylation each mAb

preparation was denatured with the supplied glycoprotein denaturing buffer at 37 °C

for 10 minutes. After deglycosylation the oligosaccharides were bound to a Supelclean

ENVI-Carb column tube (Sigma Aldrich) in 2% acetonitrile / 0.1 mol/L ammonium

acetate solution, washed and eluted with 50% acetonitrile in H2O. The eluent was

collected and evaporated to dryness in a Vacufuge® plus (Eppendorf).

N-linked oligosaccharides were fluorescently labeled with 20 µL of 0.35 mol/L 2-

aminobenzamide (2-AB) (Sigma Aldrich) in a solution of 70% DMSO, 30% glacial

acetic acid, 1 mol/L sodium cyanoborohydride according to Bigge et al. (1995) for 2

hours at 65 °C. The reaction mixture was cooled down and diluted with 380 µL

acetonitrile. Labeled oligosaccharides were purified with Whatman 3MM filter paper

disks according to Merry et al. (2002). Briefly, the reaction mixture was loaded on the

paper disk and excess 2-AB dye was washed 3 times with 95% acetonitrile. The 2-AB

labeled oligosaccharides were eluted 3 times with Milli-Q water (Millipore) and stored

at -20 °C.

7.1.2 HPLC Separation of 2-AB Labeled Oligosaccharides

30 µL of the 2-AB labeled oligosaccharides sample was diluted in 70 µL acetonitrile.

90 µL of the prepared sample was injected at a flow rate of 0.6 mL/min on an Agilent

1200 HPLC system (Agilent Technologies) equipped with a fluorescence detector for

analysis by hydrophilic interaction chromatography (HILIC) using a GlykoSep N-Plus

column (4.6 mm x 150 mm, Prozyme). N-linked oligosaccharides were separated

based on their hydrodynamic volume using a gradient system of buffer A (10 mmol/L


formic acid, 80% acetonitrile) and buffer B (30 mmol/L formic acid, 40% acetonitrile)

at pH 4.4 with a linear increase from 30 – 90% of buffer B over 90 min. The separated

oligosaccharides were fluorescently detected with an excitation wavelength of 250 nm

and an emission wavelength of 410 nm. The peak fractions were collected every

minute with a fraction collector FC 203B (Gilson), concentrated to approximately 10

µL in a Vacufuge® plus (Eppendorf), and stored at -20 °C for mass spectrometry

(MS) analysis.

7.1.3 N-linked Oligosaccharide Analysis by Matrix-Assisted Laser

Desorption/Ionization Time-Of-Flight (MALDI-TOF) Mass

Spectrometry (MS)

MS analysis was performed at the Functional Genomic Center Zurich using a 4800

Plus MALDI TOF/TOF system (AB SCIEX) equipped with a 200 Hz Nd:YAG laser.

2-AB labeled N-linked oligosaccharides were desalted by ZipTip µ-C18 pipette tips

(Millipore). ZipTips were wetted with 80% acetonitrile and twice equilibrated with

Milli-Q water (Millipore), 10 µL of sample was bound by 5 times pipetting up and

down and washed twice with 0.1% formic acid. 0.9 µL of 2,5-dihydroxybenzoic acid

(DHB) matrix (10 g/L) in 70% acetonitrile was pipetted 3 times up and down with the

previously prepared ZipTip µ-C18, spotted onto a 384 Opti-TOF MALDI target plate

and dried at room temperature. The MALDI-TOF mass spectrometer was operated in

reflectron positive and negative ion mode with a laser intensity of 5400 – 5600 arb.

Mass spectra were acquired by accumulating 1000 – 4000 subspectra in a mass range

from 750 m/z – 4000 m/z. Fragment ions of oligosaccharides were detected by MS/MS

with collision-induced dissociation (CID) in positive and negative ion mode. The

following parameter settings were applied in the MS/MS mode: Laser intensity: 5700 –

6000 arb, collision gas pressure: 2.5 x 10-6 Torr, timed ion selector (TIS) resolution:

150, accumulation of 2000 – 4000 subspectra.

Proposed glycan structures were either deduced from the fragmentation pattern

obtained from MS/MS analysis, by comparison of obtained masses with databases

provided in GlycoWorkBench2 [100] or by comparison of HPLC profiles with

commercial Cetuximab antibody with a known glycan profile [48].

CHAPTER 7: APPENDIX 83

Figure A 1: MALDI-TOF MS spectra of the 2-AB labeled oligosaccharide

released from HFN 7.1 antibody by PNGase F in (A) positive ion mode and (B)

negative ion mode. Peaks in positive ion mode are either detected as Na+ or K+

adducts. (C) Fragmentation pattern of G1F (m/z 1767.70) obtained by MALDI-

TOF MS/MS in positive ion mode. Sugar residues are N-acetylglucosamine

(GlcNAc, ), fucose (F, ), mannose (M, ), galactose (G, ).

1500 2000 2500 30000

20406080

100

1500 2000 2500 30000

20406080

100

2253.83

C

2091.79

1945.721929.75

1799.64

1783.67

1637.59

1621.62

1767.70

1605.65

1565.66

1402.57In

tensit

y %

m/z

1459.59

2578.15

B

2522.10

2902.30

2359.072376.08

2740.20

2197.03

2213.022050.96

2034.97

1958.841905.78

1847.87

1796.77

1733.03

1743.27

1634.70

1581.10

Inten

sity

%

m/z

A

204.12

366.19

388.17

510.25

550.26

713.35

741.33

915.39

1077.46

1118.50

1253.59

1298.55

1280.49

1402.64

1384.59

1564.69

1605.67

1621.70

500 1000 15000

20

40

60

80

100

Inten

sity

%

m/z

MS/MS at m/z 1767.70 [M+Na]+

(G1F)


Table A 1: Measured masses of reductively aminated oligosaccharides

determined by MALDI-TOF MS and compared to theoretical masses. Proposed

glycoforms were assigned to the glycosylation profile in Figure 2-4.

Mea

sure

d m

/z v

alue

s

[M-H

] -

- - -

1581

.10

1743

.27

1847

.87

1905

.78

-

2034

.97

2050

.96

2197

.03

[M+K

] +

- - -

1621

.62

1783

.67

-

1945

.72

- - - -

[M+N

a] +

1402

.57

1459

.59

1564

.63

1605

.65

1767

.70

-

1929

.75

2091

.79

- - -

The

oret

ical

m

ass (

Da)

1379

.53

1436

.56

1541

.59

1582

.61

1744

.67

1848

.68

1906

.72

2068

.77

2035

.76

2051

.76

2197

.81

Olig

o-sa

ccha

ride

b

Sym

bol

G0

G0F

G1F

G2F

G3F

G0F

-G

alN

GN

A

Com

posi

tiona

(Hex

NA

c2H

ex3)

FucH

exN

Ac

(Hex

NA

c2H

ex3)

Hex

NA

c2

(Hex

NA

c2H

ex3)

FucH

exN

AcH

ex

(Hex

NA

c2H

ex3)

FucH

exN

Ac2

(Hex

NA

c2H

ex3)

FucH

exN

Ac2

Hex

(Hex

NA

c2H

ex3)

FucH

exN

AcH

exN

euG

c

(Hex

NA

c2H

ex3)

FucH

exN

Ac2

Hex

2

(Hex

NA

c2H

ex3)

FucH

exN

Ac2

Hex

3

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Ac)

H

exN

Ac

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Gc)

H

exN

Ac

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Ac)

H

exN

AcH

ex


Mea

sure

d m

/z v

alue

s

[M-H

] -

2213

.02

-

2359

.07

2375

.08

2520

.08

2578

.15

2740

.20

2902

.30

a H

ex, H

exN

Ac,

Fuc

, Neu

Gc,

Neu

Ac

indi

cate

hex

ose,

N-a

cety

lhex

osam

ine,

fuco

se, N

-gly

coly

lneu

ram

inic

aci

d an

d N

-ace

tyln

eura

min

ic

acid

, res

pect

ivel

y.

b Suga

r res

idue

s are

N-a

cety

lglu

cosa

min

e (G

lcN

Ac,

),

fuco

se (F

, )

, man

nose

(M,

), g

alac

tose

(G,

) a

nd N

-gly

coly

lneu

ram

inic

aci

d

(NG

NA

,

)

[M+K

] +

- - - - - - - -

[M+N

a] +

-

2253

.83

- - - - - -

The

oret

ical

m

ass (

Da)

2213

.81

2230

.82

2359

.87

2375

.86

2520

.90

2578

.94

2740

.99

2903

.05

Olig

o-sa

ccha

ride

b

Sym

bol

G2F

-N

GN

A

G2F

-G

alN

GN

A

G2F

-N

GN

A2

Com

posi

tiona

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Gc)

H

exN

AcH

ex

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

2)H

exN

AcH

ex2

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Ac)

H

exN

AcH

ex2

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Gc)

H

exN

AcH

ex2

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Gc)

2

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Gc)

(H

exN

AcH

ex2)

Hex

NA

c

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Gc)

(H

exN

AcH

ex2)

Hex

NA

cHex

(Hex

NA

c2H

ex3)

Fuc(

Hex

NA

cHex

Neu

Gc)

(H

exN

AcH

ex2)

Hex

NA

cHex

2


7.2 Appendix for “Insights into pH induced Metabolic Switch by

Flux Balance Analysis”

7.2.1 Gene Expression Analysis

Table A 2: List of target genes investigated by gene expression analysis.

Gene Symbol Name

1 Hk1 hexokinase 1 2 Hk2 hexokinase 2 3 Pfkl phosphofructokinase, liver 4 Pgam2 phosphoglycerate mutase 2 5 Pkm2 pyruvate kinase muscle 6 Pepck1 phosphoenolpyruvate carboxykinase 1, cytosolic 7 Pepck2 phosphoenolpyruvate carboxykinase 2, mitochondrial 8 Ldha lactate dehydrogenase A 9 Pcx pyruvate carboxylase 10 Pdha1 pyruvate dehydrogenase E1 alpha 1 11 Pdha2 pyruvate dehydrogenase E1 alpha 2 12 Dlat dihydrolipoamide S-acetyltransferase 13 Pdk1 pyruvate dehydrogenase kinase, isoenzyme 1 14 Pdk2 pyruvate dehydrogenase kinase, isoenzyme 2 15 Pdk3 pyruvate dehydrogenase kinase, isoenzyme 3 16 Cs citrate synthase 17 Idh3a isocitrate dehydrogenase 3 alpha 18 Ogdh oxoglutarate dehydrogenase 19 Sdha succinate dehydrogenase complex, subunit A 20 Fh1 fumarate hydratase 1 21 Mdh1 malate dehydrogenase 1, NAD 22 Mdh2 malate dehydrogenase 2, NAD, mitochondrial 23 Acly ATP citrate lyase 24 Me1 malic enzyme 1, cytosolic 25 Me2 malic enzyme 2, mitochondrial NADP dependent 26 Me3 malic enzyme 3, mitochondrial NAD dependent


27 Gpt2 glutamic pyruvic transaminase 2 28 Got1 glutamate oxaloacetate transaminase 1 29 Got2 glutamate oxaloacetate transaminase 2 30 Glud1 glutamate dehydrogenase 1 (=GDH) 31 Gls glutaminase 32 G6pdh glucose-6-phosphate dehydrogenase


7.2.2 Metabolic Network for Flux Balance Analysis

Table A 3: Reactions in the metabolic network used for flux balance analysis.

Glycolysis 1 GLC6P <-> F6P 2 F6P + ATP -> ADP + 2 GAP 3 GAP + NAD + ADP <-> [3PG] + NADH + ATP 4 3PG -> PEP 5 PEP + ADP + H -> PYR + ATP

Pentose Phosphate Pathway

6 GLC6P + 2 NADP + H2O -> RBL5P + 2 NADPH + CO2 7 RBL5P <-> R5P 8 RBL5P <-> X5P 9 2 X5P + R5P <-> 2 F6P + G3P

Pyruvate Metabolism

10 PYR + H -> PYR_m + H_m 11 PYR_m + CoA_m + NAD_m -> CO2 + AcCoA_m + NADH_m + H_m 12 PYR_m + HCO3_m + ATP_m -> OAA_m + ADP_m + Pi 13 MAL_m + NAD_m <-> PYR_m + NADH_m + H_m + CO2 14 MAL + NADP <-> PYR + NADPH + H + CO2

TCA Cycle 15 OAA_m + AcCoA_m + H2O <-> CIT_m + CoA_m 16 CIT_m + NAD_m + H2O <-> AKG_m + NADH_m + CO2 + H_m 17 AKG_m + NAD_m + CoA_m -> SUCCoA_m + NADH_m + H_m + CO2

18 SUCCoA_m + ADP_m + Pi_m + FAD_m <-> FUM_m + ATP_m + CoA_m + FADH2_m

19 FUM_m + H2O <-> MAL_m 20 MAL_m + NAD_m <-> OAA_m + NADH_m + H_m

Malate Aspartate Shuttle 21 OAA_m + GLU_m <-> AKG_m + ASP_m 22 AKG + ASP <-> OAA + GLU 23 ASP_m + GLU -> ASP + GLU_m 24 OAA + NADH + H <-> MAL + NAD 25 MAL + AKG_m <-> MAL_m + AKG


Glutaminolysis 26 GLN + H2O -> GLU + NH4 27 GLU_m + H2O + NAD_m <-> AKG_m + NADH_m + NH4 28 GLU <-> GLU_m

Acetyl CoA Formation 29 CIT + CoA + ATP + H2O -> AcCoA + OAA + ADP + Pi

Amino Acid Synthesis 30 GLU + PYR <-> AKG + ALA 31 3PG + NAD + GLU + H2O -> NADH + H + AKG + Pi + SER

32 GLU + ATP + H + NADPH + H + NADH -> ADP + NADP + Pi + H2O + NAD + PRO

33 SER + THF <-> GLY + MTHF + H2O 34 GLN + ASP + ATP + H2O <-> GLU + ASN + AMP + 2 Pi + H

Amino Acid Degradation 35 SER <-> PYR + NH4 36 THR + NAD + CoA -> GLY + AcCoA + NADH + H

37 ARG_m + H2O + AKG_m + NAD_m <-> 2 GLU_m + UREA_m + NADH_m + H_m

38 PRO_m + 2 H2O + NAD_m -> GLU_m + NADH_m + H_m 39 HIS_m + 3 H2O -> GLU_m + Formamide_m + NH4

40 LYS_m + NADPH + 2 AKG_m + 4 NAD_m + 2 H2O + 2 CoA_m -> NADP + 2 H_m + 4 NADH_m + 2 GLU_m + 2 CO2 + 2 AcCoA_m

41 PHE + NADPH + H + O2 -> TYR + NADP

42 TYR + AKG + 2 O2 + CoA + H2O <-> GLU + CO2 + 2 H + FUM_m + AcCoA + Acetate

43

VAL_m + AKG_m + CoA_m + 3 NAD_m + FAD_m + 3 H2O + ATP_m -> 2 H_m + GLU_m + SUCCoA_m + CO2 + 3 NADH_m + FADH2_m + ADP_m + Pi_m

44

ILE_m + AKG_m + 2 NAD_m + 2 CoA_m + FAD_m + H2O + HCO3_m + ATP_m -> GLU_m + SUCCoA_m + AcCoA_m + 2 NADH_m + CO2 + FADH2_m + ADP_m + Pi_m

45 LEU_m + AKG_m + NAD_m + 2 CoA_m + ATP_m + HCO3_m + H2O -> GLU_m + NADH_m + CO2 + H_m + Pi_m + ADP_m + Acetate + 2 AcCoA_m

46 CYS + AKG + SO3 + O2 + H2O -> GLU + PYR + S2O3 + 2 H

47 MET + 2 ATP + H2O + SER + CoA + NAD + HCO3 -> NH4 + CYS + CO2 + NADH + SUCCoA_m + ADP + Pi


Transport 48 CIT_m + MAL <-> CIT + MAL_m 49 PRO <-> PRO_m 50 ARG <-> ARG_m

PEP Regeneration 51 OAA_m + ATP_m -> ADP_m + PEP_m + CO2 52 PEP <-> PEP_m

Oxidative Phosphorylation

53 NADH_m + H_m + 0.5 O2 + 2.5 ADP_m + 2.5 Pi_m -> NAD_m + H2O + 2.5 ATP_m

54 FADH2_m + 0.5 O2 + 1.5 ADP_m + 1.5 Pi_m -> FAD_m + H2O + 1.5 ATP_m Nucleotide Synthesis

55 R5P + ATP -> PRPP + AMP

56 PRPP + 2 GLN + GLY + ASP + 4 ATP + CO2 -> IMP + 2 GLU + FUM + 4 ADP + 2 H2O

57 IMP + ASP + 2 ATP + GTP -> dATP + FUM + 2 ADP + GDP

58 IMP + GLN + 3 ATP + NAD + 2 H2O -> dGTP + GLU + 2 ADP + AMP + NADH

59 HCO3 + NH4 + ASP + PRPP + 3 ATP + NAD -> dTTP + 3 ADP + CO2 + NADH 60 dTTP + GLN + ATP -> dCTP + GLU + ADP

Uptake or Production of Substrates (Measured Fluxes) 61 GLC6P + ADP + H -> GLC_ex + ATP 62 PYR + NADH + 2 H -> LAC_ex + NAD + H 63 GLN <-> GLN_ex 64 GLU <-> GLU_ex 65 ALA -> ALA_ex 66 SER -> SER_ex 67 GLY -> GLY_ex 68 ASP -> ASP_ex 69 ARG_m -> ARG_ex 70 HIS_m -> HIS_ex 71 CYS -> CYS_ex 72 THR -> THR_ex 73 ILE_m -> ILE_ex


74 VAL_m -> VAL_ex 75 PHE -> PHE_ex 76 TYR -> TYR_ex 77 LYS_m -> LYS_ex 78 LEU_m -> LEU_ex 79 PRO -> PRO_ex 80 MET -> MET_ex 81 ASN -> ASN_ex

82

0.0148 dATP + 0.0099 dCTP + 0.0099 dGTP + 0.0148 dTTP + 0.07 ATP + 0.033 dATP + 0.0551 dCTP + 0.0624 dGTP + 0.033 dTTP + 0.07 ATP + 0.6 ALA + 0.377 ARG_m + 0.359 ASP + 0.288 ASN + 0.145 CYS + 0.322 GLN + 0.386 GLU + 0.538 GLY + 0.143 HIS_m + 0.324 ILE_m + 0.564 LEU_m + 0.57 LYS_m + 0.138 MET + 0.219 PHE + 0.313 PRO + 0.43 SER + 0.386 THR + 0.044 TRP + 0.182 TYR + 0.416 VAL_m + 29.04 ATP + 0.38 GLC6P + 1.3148 ATP + 0.109 GAP + 1.9184 AcCoA + 1.70476 ATP + 3.40952 NADH + 3.51852 H + 0.109 NADH + 0.008 GAP + 0.0704 AcCoA + 0.06256 ATP + 0.12512 NADH + 0.13312 H + 0.008 NADH + 0.324 AcCoA + 0.324 ATP + 0.288 NADH + 0.288 H -> Biomass + 0.07 ADP + 0.07 Pi + 0.07 ADP + 0.07 Pi + 29.04 ADP + 29.04 Pi + 1.3148 ADP + 1.3148 Pi + 1.9184 CoA + 1.70476 ADP + 1.70476 Pi + 3.40952 NAD + 1.49112 H2O + 0.109 NAD + 0.0704 CoA + 0.06256 ADP + 0.06256 Pi + 0.12512 NAD + 0.05472 H2O + 0.008 NAD + 0.162 CoA + 0.324 ADP + 0.324 Pi + 0.288 NAD + 0.162 CO2

83

0.5768 ALA + 0.411919 ASN + 0.330176 ASP + 0.35901 GLN + 0.464684 GLU + 0.683611 GLY + 0.669275 PRO + 1.010565 SER + 0.198476 ARG_m + 0.232693 CYS + 0.204171 HIS_m + 0.30046 ILE_m + 0.574408 LEU_m + 0.54615 LYS_m + 0.114338 MET + 0.366898 PHE + 0.979228 THR + 0.19333 TRP + 0.251256 TYR + 0.796933 VAL_m + 39.89 ATP -> MAB + 39.89 ADP + 39.89 Pi


The biomass composition of murine hybridoma cells averaged in previous work by

Altamirano et al. (2001), Selvarasu et al. (2010) and Xie and Wang (1994) was

obtained using the relative compositions from Sheikh et al. (2005) (Appendix Table A

4). The mole of each cellular component can be calculated from the biomass weight.

Dry cell weight of ATCC-CRL 1606 cells was considered as 250 pg/cell [215].

Furthermore, diphosphatidyglycerol was not taken into consideration because of its

negligible proportion in cellular composition. Therefore, protein, DNA/RNA,

carbohydrate, cholesterol, and phospholipids (phosphoglycerides and sphingomyelin)

were included in the equations of biomass formation.

In addition to biomass synthesis, monoclonal antibody synthesis was also considered

in our network by utilizing the amino acids composition of IgG1 from Quek et al.

(2010) (Appendix Table A 5). The theoretical amount of 4.3 ATP per mol amino acid

is taken into account for protein synthesis.


Table A 4: Averaged dry cell composition (protein, carbohydrates, nucleotides,

lipids) of murine hybridoma cells derived from literature [148].

Metabolite nmol /μg

Metabolite nmol/μg

Protein

ALA 0.6 Carbohydrates Glycogen 0.279 ARG 0.377

Nucleotides

dAMP 0.0148 ASP 0.359 dCMP 0.0099 ASN 0.288 dGMP 0.0099 CYS 0.145 dTMP 0.0148 GLN 0.322 AMP 0.033 GLU 0.386 CMP 0.0551 GLY 0.538 GMP 0.0624 HIS 0.143 UMP 0.033 ILE 0.324

Lipids

Cholesterol 0.018 LEU 0.564 Phosphatidylcholine 0.069 LYS 0.57 Phosphatidylethanolamine 0.026 MET 0.138 Phosphatidylinostitol 0.01 PHE 0.219 Phosphatidylserine 0.003 PRO 0.313 Phosphatidylglycerol 0.001 SER 0.43 Diphosphatidylglycerol 0.003 THR 0.386 Sphingomyelin 0.008 TRP 0.044 TYR 0.182 VAL 0.416


Table A 5: Amino acid composition of IgG1 derived from an average IgG1 protein

sequence [137].

Mass fraction Molar fraction g AA/g mAB nmol AA/μg mAB

ALA 0.041 0.577 ASN 0.047 0.412 ASP 0.038 0.33 GLN 0.046 0.359 GLU 0.06 0.465 GLY 0.039 0.684 PRO 0.065 0.669 SER 0.088 1.0106 ARG 0.031 0.198 CYS 0.024 0.233 HIS 0.028 0.204 ILE 0.034 0.3 LEU 0.065 0.574 LYS 0.07 0.546 MET 0.015 0.114 PHE 0.054 0.367 THR 0.099 0.979 TRP 0.036 0.193 TYR 0.041 0.251 VAL 0.079 0.797

sum (nmol AA/μg mAB) 9.264


Figure A 2: Flux balance model scheme. PEP regeneration pathway was only

active at pH 7.8 and therefore the fluxes OAAmPEPm and PEPmPEP were

constrained under all other conditions.


7.2.3 Calculated Intracellular Fluxes from Flux Balance Model

Figure A 3: Intracellular fluxes calculated by flux balance analysis at three

different time points (WD 1.0, WD 1.5, WD 1.9) for standard pH 7.2 culture.

Numbers correspond to reactions specified below.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24-200

0200400600800

Intra

cellu

lar fl

ux

(nm

ol/1

06 cells

/h)

WD 1.0 WD 1.5 WD 1.9

1. GLC6P -> F6P 17. MALm -> OAAm 33. THR -> GLY+AcCoAm2. F6P -> 2GAP 18. OAAm+GLUm -> AKGm+ASPm 34. ARGm+AKGm -> 2GLUm3. GAP -> 3PG 19. AKG+ASP -> OAA+GLU 35. PROm -> GLUm4. 3PG -> PEP 20. ASPm+GLU -> ASP+GLUm 36. HISm -> GLUm 5. PEP -> PYR 21. OAA -> MAL 37. LYSm+2AKG_m -> 2GLUm+2AcCoAm6. GLC6P ->RBL5P 22. MAL+AKGm -> MALm+AKG 38. PHE -> TYR7. PYR -> PYRm 23. GLN -> GLU 39. TYR+AKG -> GLU+AcCoA8. PYRm -> AcCoAm 24. GLUm -> AKGm 40. VALm+AKGm -> GLUm+SUCCoAm9. PYRm -> OAAm 25. GLU -> GLUm 41. ILEm+AKGm -> GLUm+SUCCoAm+10. MALm -> PYRm 26. CIT -> AcCoA+OAA AcCoAm11. MAL -> PYR 27. GLU+PYR -> AKG+ALA 42. LEUm+AKGm -> GLUm+2 AcCoAm12. OAAm -> CITm 28. 3PG+GLU -> AKG+SER 43. CYS+AKG -> GLU+PYR13. CITm -> AKGm 29. GLU -> PRO 44. MET+SER -> CYS+SUCCoAm14. AKGm -> SUCCoAm 30. SER -> GLY 45. CITm+MAL -> CIT+MALm15. SUCCoAm -> FUMm 31. GLN+ASP -> GLU+ASN 46. PRO -> PROm16. FUMm -> MALm 32. SER -> PYR 47. ARG ->ARG_m

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47-20

020406080

Intra

cellu

lar fl

ux(n

mol

/106 ce

lls/h

)


Figure A 4: Comparison between intracellular fluxes at culture pH 7.2 and pH 6.8

calculated by flux balance analysis at WD 1.9. Numbers correspond to reactions

specified below.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24-200-100

0100200300400

Intra

cellu

lar fl

ux(n

mol

/106 ce

lls/h

)

pH 7.2 pH 6.8

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47-20-10

0102030

1. GLC6P -> F6P 17. MALm -> OAAm 33. THR -> GLY+AcCoAm2. F6P -> 2GAP 18. OAAm+GLUm -> AKGm+ASPm 34. ARGm+AKGm -> 2GLUm3. GAP -> 3PG 19. AKG+ASP -> OAA+GLU 35. PROm -> GLUm4. 3PG -> PEP 20. ASPm+GLU -> ASP+GLUm 36. HISm -> GLUm 5. PEP -> PYR 21. OAA -> MAL 37. LYSm+2AKG_m -> 2GLUm+2AcCoAm6. GLC6P ->RBL5P 22. MAL+AKGm -> MALm+AKG 38. PHE -> TYR7. PYR -> PYRm 23. GLN -> GLU 39. TYR+AKG -> GLU+AcCoA8. PYRm -> AcCoAm 24. GLUm -> AKGm 40. VALm+AKGm -> GLUm+SUCCoAm9. PYRm -> OAAm 25. GLU -> GLUm 41. ILEm+AKGm -> GLUm+SUCCoAm+10. MALm -> PYRm 26. CIT -> AcCoA+OAA AcCoAm11. MAL -> PYR 27. GLU+PYR -> AKG+ALA 42. LEUm+AKGm -> GLUm+2 AcCoAm12. OAAm -> CITm 28. 3PG+GLU -> AKG+SER 43. CYS+AKG -> GLU+PYR13. CITm -> AKGm 29. GLU -> PRO 44. MET+SER -> CYS+SUCCoAm14. AKGm -> SUCCoAm 30. SER -> GLY 45. CITm+MAL -> CIT+MALm15. SUCCoAm -> FUMm 31. GLN+ASP -> GLU+ASN 46. PRO -> PROm16. FUMm -> MALm 32. SER -> PYR 47. ARG ->ARG_m

Intra

cellu

lar fl

ux(n

mol

/106 ce

lls/h

)




specified below. Reactions 48 - 50 included to metabolic network as additional

constraints according to gene expression data.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25-100

0100200300400500

Intra

cellu

lar fl

ux(n

mol

/106 ce

lls/h

)

pH 7.2 pH 7.8

26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49-20-10

0102030

1. GLC6P -> F6P 18. OAAm+GLUm -> AKGm+ASPm 35. PROm -> GLUm 2. F6P -> 2GAP 19. AKG+ASP -> OAA+GLU 36. HISm -> GLUm3. GAP -> 3PG 20. ASPm+GLU -> ASP+GLUm 37. LYSm+2AKG_m -> 2GLUm+2AcCoAm4. 3PG -> PEP 21. OAA -> MAL 38. PHE -> TYR5. PEP -> PYR 22. MAL+AKGm -> MALm+AKG 39. TYR+AKG -> GLU+AcCoA 6. GLC6P ->RBL5P 23. GLN -> GLU 40. VALm+AKGm -> GLUm+SUCCoAm7. PYR -> PYRm 24. GLUm -> AKGm 41. ILEm+AKGm -> GLUm+SUCCoAm+8. PYRm -> AcCoAm 25. GLU -> GLUm AcCoAm9. PYRm -> OAAm 26. CIT -> AcCoA+OAA 42. LEUm+AKGm -> GLUm+2 AcCoAm 10. MALm -> PYRm 27. GLU+PYR -> AKG+ALA 43. CYS+AKG -> GLU+PYR11. MAL -> PYR 28. 3PG+GLU -> AKG+SER 44. MET+SER -> CYS+SUCCoAm12. OAAm -> CITm 29. GLU -> PRO 45. CITm+MAL -> CIT+MALm13. CITm -> AKGm 30. SER -> GLY 46. PRO -> PROm14. AKGm -> SUCCoAm 31. GLN+ASP -> GLU+ASN 47. ARG ->ARG_m 15. SUCCoAm -> FUMm 32. SER -> PYR 48. OAAm+ATPm -> PEPm+ADPm+CO216. FUMm -> MALm 33. THR -> GLY+AcCoAm 49. PEP -> PEPm 17. MALm -> OAAm 34. ARGm+AKGm -> 2GLUm

Intra

cellu

lar fl

ux(n

mol

/106 ce

lls/h

)


7.3 Appendix for “An Unstructured Model of Metabolite and

Temperature Dependent Cell Cycle Arrest in Hybridoma

Batch and Fed-Batch Cultures”

7.3.1 Model Parameter Estimation

Table A 6: Model parameter values.

Parameter Value Re-estimated value

95% confidence intervals

kdMAX 0.06

klys 0.04

KdAmm 50

kT 3 3.2 0.04

rT 0.01

Kdeg 4.8*10-3

KGln 0.08

KLac 90

KAmm 15

Kupt 0.4

YGlc 1.2*108

YGln 5.6*108 6.1*108 1*107

YLac 2

YAmm 0.6

mGlc 2.8*10-12

α1 3.4*10-13

α2 4

n 2 2

m 2


fres 2 1.5 0.009

tG1initial 3.3 2.7 0.094

tG2/Minitial 2.5 2.8 0.083

tSinitial 6 5.1 0.205

qG1/G0 2.8*10-11 2.0*10-11 1.1*10-11

qS 2.8*10-11 3.0*10-11 1.3*10-11

qG2/M 2.8*10-11

101

8 References

[1] S. Kyriakopoulos and C. Kontoravdi, “Analysis of the landscape of biologically-derived pharmaceuticals in Europe: Dominant production systems, molecule types on the rise and approval trends.,” European Journal of Pharmaceutical Sciences : Official Journal of the European Federation for Pharmaceutical Sciences, vol. 48, no. 3, pp. 428–441, Dec. 2012.

[2] M. Butler, “Animal cell cultures: recent achievements and perspectives in the production of biopharmaceuticals.,” Applied Microbiology and Biotechnology, vol. 68, no. 3, pp. 283–91, Aug. 2005.

[3] R. Jefferis, “Glycosylation of recombinant antibody therapeutics.,” Biotechnology Progress, vol. 21, no. 1, pp. 11–6, 2005.

[4] P. P. Jacobs, M. Inan, N. Festjens, J. Haustraete, A. Van Hecke, R. Contreras, M. M. Meagher, and N. Callewaert, “Fed-batch fermentation of GM-CSF-producing glycoengineered Pichia pastoris under controlled specific growth rate.,” Microbial Cell Factories, vol. 9, no. 1, p. 93, Jan. 2010.

[5] G. A. Grabowski, M. Golembo, and Y. Shaaltiel, “Taliglucerase alfa: An enzyme replacement therapy using plant cell expression technology.,” Molecular genetics and metabolism, vol. 112, no. 1, pp. 1–8, May 2014.

[6] G. N. Mowat and W. G. Chapman, “Growth of Foot-and-Mouth Disease Virus in a Fibroblastic Cell Line Derived from Hamster Kidneys,” Nature, vol. 194, no. 4825, pp. 253–255, Apr. 1962.

[7] S. S. Ozturk and W. Hu, Cell culture technology for pharmaceutical and cell-based therapies. New York: CRC Press, 2006.

[8] Lubiniecki, Large-Scale Mammalian Cell Culture Technology. CRC Press, 1990, p. 656.

[9] R. J. Kaufman, L. C. Wasley, A. J. Spiliotes, S. D. Gossels, S. A. Latt, G. R. Larsen, and R. M. Kay, “Coamplification and coexpression of human tissue-type plasminogen activator and murine dihydrofolate reductase sequences in Chinese hamster ovary cells.,” Molecular and cellular biology, vol. 5, no. 7, pp. 1750–9, Jul. 1985.

[10] F. Braido, S. Holgate, and G. W. Canonica, “From ‘blockbusters’ to ‘biosimilars’: an opportunity for patients, medical specialists and health care providers.,” Pulmonary Pharmacology & Therapeutics, vol. 25, no. 6, pp. 483–6, Dec. 2012.

102 CHAPTER 8: REFERENCES

[11] R. S. Aggarwal, “What’s fueling the biotech engine-2012 to 2013.,” Nature Biotechnology, vol. 32, no. 1, pp. 32–9, Jan. 2014.

[12] D. J. Norman, C. F. Shield, J. M. Barry, K. Henell, M. B. Funnell, and J. Lemon, “Therapeutic use of OKT3 monoclonal antibody for acute renal allograft rejection.,” Nephron, vol. 46 Suppl 1, pp. 41–7, Jan. 1987.

[13] M. Butler, “From the mold in Dr Florey’s coat to blockbuster drugs,” Pharmaceutical Bioprocessing, vol. 1, no. 1, pp. 3–5, Apr. 2013.

[14] B. Kelley, “Industrialization of mAb production technology: the bioprocessing industry at a crossroads.,” mAbs, vol. 1, no. 5, pp. 443–52, 2009.

[15] J. Y. Kim, Y.-G. Kim, and G. M. Lee, “CHO cells in biotechnology for production of recombinant proteins: current state and further potential.,” Applied Microbiology and Biotechnology, vol. 93, no. 3, pp. 917–30, Feb. 2012.

[16] F. M. Wurm, “Production of recombinant protein therapeutics in cultivated mammalian cells.,” Nature Biotechnology, vol. 22, no. 11, pp. 1393–8, Nov. 2004.

[17] M. De Jesus and F. M. Wurm, “Manufacturing recombinant proteins in kg-ton quantities using animal cells in bioreactors.,” European Journal of Pharmaceutics and Biopharmaceutics : Official Journal of Arbeitsgemeinschaft fur Pharmazeutische Verfahrenstechnik e.V, vol. 78, no. 2, pp. 184–188, Jan. 2011.

[18] Y.-M. Huang, W. Hu, E. Rustandi, K. Chang, H. Yusuf-Makagiansar, and T. Ryll, “Maximizing productivity of CHO cell-based fed-batch culture using chemically defined media conditions and typical manufacturing equipment.,” Biotechnology Progress, vol. 26, no. 5, pp. 1400–10, 2010.

[19] H. C. Ebbers, P. J. K. van Meer, E. H. M. Moors, A. K. Mantel-Teeuwisse, H. G. M. Leufkens, and H. Schellekens, “Measures of biosimilarity in monoclonal antibodies in oncology: the case of bevacizumab.,” Drug Discovery Today, vol. 18, no. 17–18, pp. 872–9, Sep. 2013.

[20] R. S. Kenett and D. A. Kenett, “Quality by Design applications in biosimilar pharmaceutical products,” Accreditation and Quality Assurance, vol. 13, no. 12, pp. 681–690, Sep. 2008.

[21] P. Umaña, J. Jean-Mairet, R. Moudry, H. Amstutz, and J. E. Bailey, “Engineered glycoforms of an antineuroblastoma IgG1 with optimized antibody-dependent cellular cytotoxic activity.,” Nature biotechnology, vol. 17, no. 2, pp. 176–80, Feb. 1999.

[22] J. R. Birch and A. J. Racher, “Antibody production.,” Advanced Drug Delivery Reviews, vol. 58, no. 5–6, pp. 671–85, Aug. 2006.

CHAPTER 8: REFERENCES 103

[23] A. W. Nienow, W. H. Scott, C. J. Hewitt, C. R. Thomas, G. Lewis, A. Amanullah, R. Kiss, and S. J. Meier, “Scale-down studies for assessing the impact of different stress parameters on growth and product quality during animal cell culture,” Chemical Engineering Research and Design, vol. 91, no. 11, pp. 2265–2274, Nov. 2013.

[24] R. Godoy-Silva, M. Mollet, and J. J. Chalmers, “Evaluation of the effect of chronic hydrodynamical stresses on cultures of suspensed CHO-6E6 cells.,” Biotechnology and Bioengineering, vol. 102, no. 4, pp. 1119–30, Mar. 2009.

[25] M. L. Velez-Suberbie, R. D. R. Tarrant, A. S. Tait, D. I. R. Spencer, and D. G. Bracewell, “Impact of aeration strategy on CHO cell performance during antibody production.,” Biotechnology Progress, vol. 29, no. 1, pp. 116–26, 2012.

[26] C. Langheinrich and A. W. Nienow, “Control of pH in large-scale, free suspension animal cell bioreactors: alkali addition and pH excursions.,” Biotechnology and Bioengineering, vol. 66, no. 3, pp. 171–9, Jan. 1999.

[27] M. M. Zhu, A. Goyal, D. L. Rank, S. K. Gupta, T. Vanden Boom, and S. S. Lee, “Effects of elevated pCO2 and osmolality on growth of CHO cells and production of antibody-fusion protein B1: a case study.,” Biotechnology Progress, vol. 21, no. 1, pp. 70–7, 2005.

[28] F. Li, N. Vijayasankaran, A. Y. Shen, R. Kiss, and A. Amanullah, “Cell culture processes for monoclonal antibody production,” mAbs, vol. 2, no. 5, pp. 466–479, Sep. 2010.

[29] R. Kimura and W. Miller, “Glycosylation of CHO-derived recombinant tPA produced under elevated pCO2,” Biotechnology Progress, vol. 15, no. 1, p. 146, Jan. 1999.

[30] V. M. Dezengotita, R. Kimura, and W. M. Miller, “Effects of CO2 and osmolality on hybridoma cells: growth, metabolism and monoclonal antibody production.,” Cytotechnology, vol. 28, no. 1–3, pp. 213–27, Nov. 1998.

[31] N. Vijayasankaran, J. Li, R. Shawley, A. Chen, M. Shiratori, M. Gawlitzek, F. Li, R. Kiss, and A. Amanullah, “Animal Cell Culture Media,” in Encyclopedia of Industrial Biotechnology, Hoboken, NJ, USA: John Wiley & Sons, Inc., 2010.

[32] B. Liu, M. Spearman, J. Doering, E. Lattová, H. Perreault, and M. Butler, “The availability of glucose to CHO cells affects the intracellular lipid-linked oligosaccharide distribution, site occupancy and the N-glycosylation profile of a monoclonal antibody.,” Journal of Biotechnology, vol. 170, pp. 17–27, Jan. 2014.


[33] N. J. Stark and E. C. Heath, “Glucose-dependent glycosylation of secretory glycoprotein in mouse myeloma cells.,” Archives of Biochemistry and Biophysics, vol. 192, no. 2, pp. 599–609, Feb. 1979.

[34] H. Tachibana, K. Taniguchi, Y. Ushio, K. Teruya, K. Osada, and H. Murakami, “Changes of monosaccharide availability of human hybridoma lead to alteration of biological properties of human monoclonal antibody.,” Cytotechnology, vol. 16, no. 3, pp. 151–7, Jan. 1994.

[35] N. S. C. Wong, L. Wati, P. M. Nissom, H. T. Feng, M. M. Lee, and M. G. S. Yap, “An investigation of intracellular glycosylation activities in CHO cells: effects of nucleotide sugar precursor feeding.,” Biotechnology and Bioengineering, vol. 107, no. 2, pp. 321–36, Oct. 2010.

[36] C. K. Crowell, G. E. Grampp, G. N. Rogers, J. Miller, and R. I. Scheinman, “Amino acid and manganese supplementation modulates the glycosylation state of erythropoietin in a CHO culture system.,” Biotechnology and Bioengineering, vol. 96, no. 3, pp. 538–49, Feb. 2007.

[37] N. A. McCracken, R. Kowle, and A. Ouyang, “Control of galactosylated glycoforms distribution in cell culture system.,” Biotechnology Progress, pp. 24–27, Apr. 2014.

[38] R. Jefferis, “Isotype and glycoform selection for antibody therapeutics.,” Archives of Biochemistry and Biophysics, vol. 526, no. 2, pp. 159–66, Oct. 2012.

[39] R. Jefferis, “Glycosylation as a strategy to improve antibody-based therapeutics.,” Nature Reviews. Drug Discovery, vol. 8, no. 3, pp. 226–34, Mar. 2009.

[40] R. Jefferis, “Recombinant antibody therapeutics: the impact of glycosylation on mechanisms of action.,” Trends in Pharmacological Sciences, vol. 30, no. 7, pp. 356–62, Jul. 2009.

[41] I. J. del Val, C. Kontoravdi, and J. M. Nagy, “Towards the implementation of quality by design to the production of therapeutic monoclonal antibodies with desired glycosylation patterns.,” Biotechnology Progress, vol. 26, no. 6, pp. 1505–27, 2010.

[42] F. Nimmerjahn and J. V Ravetch, “Fcgamma receptors: old friends and new family members.,” Immunity, vol. 24, no. 1, pp. 19–28, Jan. 2006.

[43] F. Nimmerjahn and J. V Ravetch, “Anti-inflammatory actions of intravenous immunoglobulin.,” Annual Review of Immunology, vol. 26, pp. 513–33, Jan. 2008.

[44] Y. Mimura, P. Sondermann, R. Ghirlando, J. Lund, S. P. Young, M. Goodall, and R. Jefferis, “Role of oligosaccharide residues of IgG1-Fc in Fc gamma RIIb


binding.,” The Journal of Biological Chemistry, vol. 276, no. 49, pp. 45539–47, Dec. 2001.

[45] S. Krapp, Y. Mimura, R. Jefferis, R. Huber, and P. Sondermann, “Structural Analysis of Human IgG-Fc Glycoforms Reveals a Correlation Between Glycosylation and Structural Integrity,” Journal of Molecular Biology, vol. 325, no. 5, pp. 979–989, Jan. 2003.

[46] S. C. Wallick, E. A. Kabat, and S. L. Morrison, “Glycosylation of a VH residue of a monoclonal antibody against alpha (1-6) dextran increases its affinity for antigen.,” The Journal of Experimental Medicine, vol. 168, no. 3, pp. 1099–109, Sep. 1988.

[47] L. Huang, S. Biolsi, K. R. Bales, and U. Kuchibhotla, “Impact of variable domain glycosylation on antibody clearance: an LC/MS characterization.,” Analytical Biochemistry, vol. 349, no. 2, pp. 197–207, Feb. 2006.

[48] J. Qian, T. Liu, L. Yang, A. Daus, R. Crowley, and Q. Zhou, “Structural characterization of N-linked oligosaccharides on monoclonal antibody cetuximab by the combination of orthogonal matrix-assisted laser desorption/ionization hybrid quadrupole-quadrupole time-of-flight tandem mass spectrometry and sequential enzy,” Analytical Biochemistry, vol. 364, no. 1, pp. 8–18, May 2007.

[49] G. Köhler and C. Milstein, “Continuous cultures of fused cells secreting antibody of predefined specificity.,” Nature, vol. 256, no. 5517, pp. 495–7, Aug. 1975.

[50] A. Nissim and Y. Chernajovsky, “Historical development of monoclonal antibody therapeutics.,” Handbook of Experimental Pharmacology, no. 181, pp. 3–18, Jan. 2008.

[51] N. Lonberg, L. D. Taylor, F. A. Harding, M. Trounstine, K. M. Higgins, S. R. Schramm, C. C. Kuo, R. Mashayekh, K. Wymore, and J. G. McCabe, “Antigen-specific human antibodies from mice comprising four distinct genetic modifications.,” Nature, vol. 368, no. 6474, pp. 856–9, Apr. 1994.

[52] L. L. Green, M. C. Hardy, C. E. Maynard-Currie, H. Tsuda, D. M. Louie, M. J. Mendez, H. Abderrahim, M. Noguchi, D. H. Smith, Y. Zeng, N. E. David, H. Sasai, D. Garza, D. G. Brenner, J. F. Hales, R. P. McGuinness, D. J. Capon, S. Klapholz, and A. Jakobovits, “Antigen-specific human monoclonal antibodies from mice engineered with human Ig heavy and light chain YACs.,” Nature Genetics, vol. 7, no. 1, pp. 13–21, May 1994.

[53] J. McCafferty, A. D. Griffiths, G. Winter, and D. J. Chiswell, “Phage antibodies: filamentous phage displaying antibody variable domains.,” Nature, vol. 348, no. 6301, pp. 552–4, Dec. 1990.


[54] S.-J. Wu, J. Luo, K. T. O’Neil, J. Kang, E. R. Lacy, G. Canziani, A. Baker, M. Huang, Q. M. Tang, T. S. Raju, S. A. Jacobs, A. Teplyakov, G. L. Gilliland, and Y. Feng, “Structure-based engineering of a monoclonal antibody for improved solubility.,” Protein Engineering, Design & Selection : PEDS, vol. 23, no. 8, pp. 643–51, Aug. 2010.

[55] C. Ferrara, P. Brünker, T. Suter, S. Moser, U. Püntener, and P. Umaña, “Modulation of therapeutic antibody effector functions by glycosylation engineering: influence of Golgi enzyme localization domain and co-expression of heterologous beta1, 4-N-acetylglucosaminyltransferase III and Golgi alpha-mannosidase II.,” Biotechnology and Bioengineering, vol. 93, no. 5, pp. 851–61, Apr. 2006.

[56] V. Kayser, N. Chennamsetty, V. Voynov, K. Forrer, B. Helk, and B. L. Trout, “Glycosylation influences on the aggregation propensity of therapeutic monoclonal antibodies.,” Biotechnology Journal, vol. 6, no. 1, pp. 38–44, Jan. 2011.

[57] V. Oganesyan, M. M. Damschroder, R. M. Woods, K. E. Cook, H. Wu, and W. F. Dall’acqua, “Structural characterization of a human Fc fragment engineered for extended serum half-life.,” Molecular Immunology, vol. 46, no. 8–9, pp. 1750–5, May 2009.

[58] J. Zalevsky, A. K. Chamberlain, H. M. Horton, S. Karki, I. W. L. Leung, T. J. Sproule, G. A. Lazar, D. C. Roopenian, and J. R. Desjarlais, “Enhanced antibody half-life improves in vivo activity.,” Nature Biotechnology, vol. 28, no. 2, pp. 157–9, Feb. 2010.

[59] D. Bhutani and U. N. Vaishampayan, “Monoclonal antibodies in oncology therapeutics: present and future indications.,” Expert Opinion on Biological Therapy, vol. 13, no. 2, pp. 269–82, Feb. 2013.

[60] N. A. P. S. Buss, S. J. Henderson, M. McFarlane, J. M. Shenton, and L. de Haan, “Monoclonal antibody therapeutics: history and future.,” Current Opinion in Pharmacology, vol. 12, no. 5, pp. 615–22, Oct. 2012.

[61] B. J. Sutton and D. C. Phillips, “The three-dimensional structure of the carbohydrate within the Fc fragment of immunoglobulin G.,” Biochemical Society Transactions, vol. 11, no. 2, pp. 130–2, Apr. 1983.

[62] K. Zheng, C. Bantog, and R. Bayer, “The impact of glycosylation on monoclonal antibody conformation and stability.,” mAbs, vol. 3, no. 6, pp. 568–76, 2011.

[63] N. Sethuraman and T. A. Stadheim, “Challenges in therapeutic glycoprotein production.,” Current Opinion in Biotechnology, vol. 17, no. 4, pp. 341–6, Aug. 2006.


[64] K. N. Baker, M. H. Rendall, A. E. Hills, M. Hoare, R. B. Freedman, and D. C. James, “Metabolic control of recombinant protein N-glycan processing in NS0 and CHO cells.,” Biotechnology and Bioengineering, vol. 73, no. 3, pp. 188–202, May 2001.

[65] Y. Kanda, T. Yamada, K. Mori, A. Okazaki, M. Inoue, K. Kitajima-Miyama, R. Kuni-Kamochi, R. Nakano, K. Yano, S. Kakita, K. Shitara, and M. Satoh, “Comparison of biological activity among nonfucosylated therapeutic IgG1 antibodies with three different N-linked Fc oligosaccharides: the high-mannose, hybrid, and complex types.,” Glycobiology, vol. 17, no. 1, pp. 104–18, Jan. 2007.

[66] A. Wright and S. L. Morrison, “Effect of C2-associated carbohydrate structure on Ig effector function: studies with chimeric mouse-human IgG1 antibodies in glycosylation mutants of Chinese hamster ovary cells.,” Journal of immunology, vol. 160, no. 7, pp. 3393–402, Apr. 1998.

[67] J. Hodoniczky, Y. Z. Zheng, and D. C. James, “Control of recombinant monoclonal antibody effector functions by Fc N-glycan remodeling in vitro.,” Biotechnology Progress, vol. 21, no. 6, pp. 1644–52, 2005.

[68] P. N. Boyd, A. C. Lines, and A. K. Patel, “The effect of the removal of sialic acid, galactose and total carbohydrate on the functional activity of Campath-1H.,” Molecular Immunology, vol. 32, no. 17–18, pp. 1311–8, Dec. 1995.

[69] N. Tsuchiya, T. Endo, K. Matsuta, S. Yoshinoya, T. Aikawa, E. Kosuge, F. Takeuchi, T. Miyamoto, and A. Kobata, “Effects of galactose depletion from oligosaccharide chains on immunological activities of human IgG.,” The Journal of Rheumatology, vol. 16, no. 3, pp. 285–90, Mar. 1989.

[70] C. Ferrara, S. Grau, C. Jäger, P. Sondermann, P. Brünker, I. Waldhauer, M. Hennig, A. Ruf, A. C. Rufer, M. Stihle, P. Umaña, and J. Benz, “Unique carbohydrate-carbohydrate interactions are required for high affinity binding between FcgammaRIII and antibodies lacking core fucose.,” Proceedings of the National Academy of Sciences of the United States of America, vol. 108, no. 31, pp. 12669–74, Aug. 2011.

[71] R. L. Shields, J. Lai, R. Keck, L. Y. O’Connell, K. Hong, Y. G. Meng, S. H. A. Weikert, and L. G. Presta, “Lack of fucose on human IgG1 N-linked oligosaccharide improves binding to human Fcgamma RIII and antibody-dependent cellular toxicity.,” The Journal of Biological Chemistry, vol. 277, no. 30, pp. 26733–40, Jul. 2002.

[72] T. Shinkawa, K. Nakamura, N. Yamane, E. Shoji-Hosaka, Y. Kanda, M. Sakurada, K. Uchida, H. Anazawa, M. Satoh, M. Yamasaki, N. Hanai, and K. Shitara, “The absence of fucose but not the presence of galactose or bisecting N-acetylglucosamine of human IgG1 complex-type oligosaccharides shows the


critical role of enhancing antibody-dependent cellular cytotoxicity.,” The Journal of Biological Chemistry, vol. 278, no. 5, pp. 3466–73, Jan. 2003.

[73] B. J. Scallon, S. H. Tam, S. G. McCarthy, A. N. Cai, and T. S. Raju, “Higher levels of sialylated Fc glycans in immunoglobulin G molecules can adversely impact functionality.,” Molecular Immunology, vol. 44, no. 7, pp. 1524–34, Mar. 2007.

[74] Y. Kaneko, F. Nimmerjahn, and J. V Ravetch, “Anti-inflammatory activity of immunoglobulin G resulting from Fc sialylation.,” Science (New York, N.Y.), vol. 313, no. 5787, pp. 670–3, Aug. 2006.

[75] J. A. LaVecchio, A. D. Dunne, and A. S. Edge, “Enzymatic removal of alpha-galactosyl epitopes from porcine endothelial cells diminishes the cytotoxic effect of natural antibodies.,” Transplantation, vol. 60, no. 8, pp. 841–7, Oct. 1995.

[76] T. J. Monica, S. B. Williams, C. F. Goochee, and B. L. Maiorella, “Characterization of the glycosylation of a human IgM produced by a human-mouse hybridoma.,” Glycobiology, vol. 5, no. 2, pp. 175–85, Mar. 1995.

[77] D. C. Andersen and C. F. Goochee, “The effect of ammonia on the O-linked glycosylation of granulocyte colony-stimulating factor produced by chinese hamster ovary cells.,” Biotechnology and Bioengineering, vol. 47, no. 1, pp. 96–105, Jul. 1995.

[78] M. C. Borys, D. I. Linzer, and E. T. Papoutsakis, “Ammonia affects the glycosylation patterns of recombinant mouse placental lactogen-I by chinese hamster ovary cells in a pH-dependent manner.,” Biotechnology and Bioengineering, vol. 43, no. 6, pp. 505–14, Mar. 1994.

[79] M. Yang and M. Butler, “Effects of ammonia on CHO cell growth, erythropoietin production, and glycosylation.,” Biotechnology and Bioengineering, vol. 68, no. 4, pp. 370–80, May 2000.

[80] M. J. Gramer, J. J. Eckblad, R. Donahue, J. Brown, C. Shultz, K. Vickerman, P. Priem, E. T. J. van den Bremer, J. Gerritsen, and P. H. C. van Berkel, “Modulation of antibody galactosylation through feeding of uridine, manganese chloride, and galactose.,” Biotechnology and Bioengineering, vol. 108, no. 7, pp. 1591–602, Jul. 2011.

[81] R. K. Grainger and D. C. James, “CHO cell line specific prediction and control of recombinant monoclonal antibody N-glycosylation.,” Biotechnology and Bioengineering, vol. 110, no. 11, pp. 2970–83, Nov. 2013.

[82] P. Hossler, S. F. Khattak, and Z. J. Li, “Optimal and consistent protein glycosylation in mammalian cell culture.,” Glycobiology, vol. 19, no. 9, pp. 936–49, Sep. 2009.


[83] M. Butler, “Optimisation of the cellular metabolism of glycosylation for recombinant proteins produced by mammalian cell systems.,” Cytotechnology, vol. 50, no. 1–3, pp. 57–76, Mar. 2006.

[84] P. H. C. van Berkel, J. Gerritsen, G. Perdok, J. Valbjørn, T. Vink, J. G. J. van de Winkel, and P. W. H. I. Parren, “N-linked glycosylation is an important parameter for optimal selection of cell lines producing biopharmaceutical human IgG.,” Biotechnology Progress, vol. 25, no. 1, pp. 244–51, 2009.

[85] P. Zhang, K. F. Chan, R. Haryadi, M. Bardor, and Z. Song, “CHO glycosylation mutants as potential host cells to produce therapeutic proteins with enhanced efficacy.,” Advances in Biochemical Engineering/Biotechnology, vol. 131, no. 2012, pp. 63–87, Jan. 2013.

[86] M. Thaysen-Andersen and N. H. Packer, “Site-specific glycoproteomics confirms that protein structure dictates formation of N-glycan type, core fucosylation and branching.,” Glycobiology, vol. 22, no. 11, pp. 1440–52, Nov. 2012.

[87] R. C. Schoen, K. L. Bentley, and R. J. Klebe, “Monoclonal antibody against human fibronectin which inhibits cell attachment.,” Hybridoma, vol. 1, no. 2, pp. 99–108, Jan. 1982.

[88] M. K. Patterson, “Measurement of growth and viability of cells in culture.,” Methods in Enzymology, vol. 58, no. 1974, pp. 141–52, Jan. 1979.

[89] C. T. Goudar, K. Joeris, K. B. Konstantinov, and J. M. Piret, “Logistic equations effectively model mammalian cell batch and fed-batch kinetics by logically constraining the fit.,” Biotechnology Progress, vol. 21, no. 4, pp. 1109–18, 2005.

[90] F. A. A. Majid, M. Butler, and M. Al-Rubeai, “Glycosylation of an immunoglobulin produced from a murine hybridoma cell line: the effect of culture mode and the anti-apoptotic gene, bcl-2.,” Biotechnology and Bioengineering, vol. 97, no. 1, pp. 156–69, May 2007.

[91] V. V. T. Wong, P. M. Nissom, S.-L. Sim, J. H. M. Yeo, S.-H. Chuah, and M. G. S. Yap, “Zinc as an insulin replacement in hybridoma cultures.,” Biotechnology and Bioengineering, vol. 93, no. 3, pp. 553–63, Feb. 2006.

[92] M. W. Glacken, “Development of mathematical descriptions of mammalian cell culture kinetics for the optimization of fed-batch bioreactors,” Massachusetts Institute of Technology, 1987.

[93] S. S. Ozturk and B. O. Palsson, “Growth, metabolic, and antibody production kinetics of hybridoma cell culture: 2. Effects of serum concentration, dissolved oxygen concentration, and medium pH in a batch reactor.,” Biotechnology Progress, vol. 7, no. 6, pp. 481–94, 1991.


[94] W. M. Miller, H. W. Blanch, and C. R. Wilke, “A kinetic analysis of hybridoma growth and metabolism in batch and continuous suspension culture: effect of nutrient concentration, dilution rate, and pH.,” Biotechnology and Bioengineering, vol. 32, no. 8, pp. 947–65, Oct. 1988.

[95] E. Trummer, K. Fauland, S. Seidinger, K. Schriebl, C. Lattenmayer, R. Kunert, K. Vorauer-Uhl, R. Weik, N. Borth, H. Katinger, and D. Müller, “Process parameter shifting: Part I. Effect of DOT, pH, and temperature on the performance of Epo-Fc expressing CHO cells cultivated in controlled batch bioreactors.,” Biotechnology and Bioengineering, vol. 94, no. 6, pp. 1033–44, Aug. 2006.

[96] S. Karra, B. Sager, and M. N. Karim, “Multi-Scale Modeling of Heterogeneities in Mammalian Cell Culture Processes,” Industrial & Engineering Chemistry Research, vol. 49, no. 17, pp. 7990–8006, 2010.

[97] J. A. Zanghi, A. E. Schmelzer, T. P. Mendoza, R. H. Knop, and W. M. Miller, “Bicarbonate concentration and osmolality are key determinants in the inhibition of CHO cell polysialylation under elevated pCO(2) or pH.,” Biotechnology and Bioengineering, vol. 65, no. 2, pp. 182–91, Oct. 1999.

[98] H. P. Bonarius, J. H. Houtman, G. Schmid, C. D. de Gooijer, and J. Tramper, “Metabolic-flux analysis of hybridoma cells under oxidative and reductive stress using mass balances.,” Cytotechnology, vol. 32, no. 2, pp. 97–107, Feb. 2000.

[99] S. Lakhotia, K. D. Bauer, and E. T. Papoutsakis, “Damaging agitation intensities increase DNA synthesis rate and alter cell-cycle phase distributions of CHO cells.,” Biotechnology and Bioengineering, vol. 40, no. 8, pp. 978–90, Oct. 1992.

[100] A. Ceroni, K. Maass, H. Geyer, R. Geyer, A. Dell, and S. M. Haslam, “GlycoWorkbench: a tool for the computer-assisted annotation of mass spectra of glycans.,” Journal of Proteome Research, vol. 7, no. 4, pp. 1650–9, Apr. 2008.

[101] D. M. Sheeley, B. M. Merrill, and L. C. Taylor, “Characterization of monoclonal antibody glycosylation: comparison of expression systems and identification of terminal alpha-linked galactose.,” Analytical Biochemistry, vol. 247, no. 1, pp. 102–10, Apr. 1997.

[102] A. D. Hooker, M. H. Goldman, N. H. Markham, D. C. James, A. P. Ison, A. T. Bull, P. G. Strange, I. Salmon, A. J. Baines, and N. Jenkins, “N-glycans of recombinant human interferon-gamma change during batch culture of chinese hamster ovary cells.,” Biotechnology and Bioengineering, vol. 48, no. 6, pp. 639–48, Dec. 1995.

[103] M. Gawlitzek, T. Ryll, J. Lofgren, and M. B. Sliwkowski, “Ammonium alters N-glycan structures of recombinant TNFR-IgG: degradative versus biosynthetic


mechanisms.,” Biotechnology and Bioengineering, vol. 68, no. 6, pp. 637–46, Jun. 2000.

[104] E. Munzert, J. Muthing, H. Büntemeyer, and J. Lehmann, “Sialidase activity in culture fluid of Chinese hamster ovary cells during batch culture and its effects on recombinant human antithrombin III integrity.,” Biotechnology Progress, vol. 12, no. 4, pp. 559–63, 1996.

[105] M. J. Gramer and C. F. Goochee, “Glycosidase activities in Chinese hamster ovary cell lysate and cell culture supernatant.,” Biotechnology Progress, vol. 9, no. 4, pp. 366–73, 1993.

[106] D. C. F. Wong, K. T. K. Wong, L. T. Goh, C. K. Heng, and M. G. S. Yap, “Impact of dynamic online fed-batch strategies on metabolism, productivity and N-glycosylation quality in CHO cell cultures.,” Biotechnology and Bioengineering, vol. 89, no. 2, pp. 164–77, Jan. 2005.

[107] G. B. Nyberg, R. R. Balcarcel, B. D. Follstad, G. Stephanopoulos, and D. I. Wang, “Metabolic effects on recombinant interferon-gamma glycosylation in continuous culture of Chinese hamster ovary cells.,” Biotechnology and Bioengineering, vol. 62, no. 3, pp. 336–47, Feb. 1999.

[108] P. M. Hayter, E. M. Curling, M. L. Gould, A. J. Baines, N. Jenkins, I. Salmon, P. G. Strange, and A. T. Bull, “The effect of the dilution rate on CHO cell physiology and recombinant interferon-gamma production in glucose-limited chemostat culture.,” Biotechnology and Bioengineering, vol. 42, no. 9, pp. 1077–85, Nov. 1993.

[109] M. C. Borys, N. G. Dalal, N. R. Abu-Absi, S. F. Khattak, Y. Jing, Z. Xing, and Z. J. Li, “Effects of culture conditions on N-glycolylneuraminic acid (Neu5Gc) content of a recombinant fusion protein produced in CHO cells.,” Biotechnology and Bioengineering, vol. 105, no. 6, pp. 1048–57, Apr. 2010.

[110] X. Gu, B. J. Harmon, and D. I. Wang, “Site- and branch-specific sialylation of recombinant human interferon-gamma in Chinese hamster ovary cell culture.,” Biotechnology and Bioengineering, vol. 55, no. 2, pp. 390–8, Jul. 1997.

[111] J. P. Kunkel, D. C. Jan, M. Butler, and J. C. Jamieson, “Comparisons of the glycosylation of a monoclonal antibody produced under nominally identical cell culture conditions in two different bioreactors.,” Biotechnology Progress, vol. 16, no. 3, pp. 462–70, 2000.

[112] M. Spearman, J. Rodriguez, N. Huzel, and K. Sunley, “Effect of Culture Conditions on Glycosylation of Recombinant beta-Interferon in CHO Cells,” in Cell Technology for Cell Products, 2007, pp. 71–85.

[113] E. Pacis, M. Yu, J. Autsen, R. Bayer, and F. Li, “Effects of cell culture conditions on antibody N-linked glycosylation-what affects high mannose 5


glycoform.,” Biotechnology and Bioengineering, vol. 108, no. 10, pp. 2348–2358, May 2011.

[114] T. A. Bibila and D. K. Robinson, “In pursuit of the optimal fed-batch process for monoclonal antibody production.,” Biotechnology Progress, vol. 11, no. 1, pp. 1–13, 1995.

[115] W. Chotigeat, Y. Watanapokasin, S. Mahler, and P. P. Gray, “Role of environmental conditions on the expression levels, glycoform pattern and levels of sialyltransferase for hFSH produced by recombinant CHO cells.,” Cytotechnology, vol. 15, no. 1–3, pp. 217–21, Jan. 1994.

[116] J. P. Kunkel, D. C. Jan, J. C. Jamieson, and M. Butler, “Dissolved oxygen concentration in serum-free continuous culture affects N-linked glycosylation of a monoclonal antibody.,” Journal of Biotechnology, vol. 62, no. 1, pp. 55–71, Jun. 1998.

[117] I. J. Del Val, J. M. Nagy, and C. Kontoravdi, “A dynamic mathematical model for monoclonal antibody N-linked glycosylation and nucleotide sugar donor transport within a maturing Golgi apparatus.,” Biotechnology Progress, Sep. 2011.

[118] P. Hossler, B. C. Mulukutla, and W.-S. Hu, “Systems analysis of N-glycan processing in mammalian cells.,” PloS one, vol. 2, no. 8, p. e713, Jan. 2007.

[119] J. Primack, G. C. Flynn, and H. Pan, “A high-throughput microchip-based glycan screening assay for antibody cell culture samples.,” Electrophoresis, vol. 32, no. 10, pp. 1129–32, May 2011.

[120] O. Warburg, “On the origin of cancer cells.,” Science (New York, N.Y.), vol. 123, no. 3191, pp. 309–14, Feb. 1956.

[121] W. S. Ahn and M. R. Antoniewicz, “Metabolic flux analysis of CHO cells at growth and non-growth phases using isotopic tracers and mass spectrometry.,” Metabolic Engineering, vol. 13, no. 5, pp. 598–609, Sep. 2011.

[122] T. C. Leite, R. G. Coelho, D. Da Silva, W. S. Coelho, M. M. Marinho-Carvalho, and M. Sola-Penna, “Lactate downregulates the glycolytic enzymes hexokinase and phosphofructokinase in diverse tissues from mice.,” FEBS Letters, vol. 585, no. 1, pp. 92–8, Jan. 2011.

[123] B. C. Mulukutla, M. Gramer, and W.-S. Hu, “On metabolic shift to lactate consumption in fed-batch culture of mammalian cells.,” Metabolic Engineering, vol. 14, no. 2, pp. 138–149, Jan. 2012.

[124] M. S. Lao and D. Toth, “Effects of ammonium and lactate on growth and metabolism of a recombinant Chinese hamster ovary cell culture.,” Biotechnology Progress, vol. 13, no. 5, pp. 688–91, 1997.


[125] S. S. Ozturk, M. R. Riley, and B. O. Palsson, “Effects of ammonia and lactate on hybridoma growth, metabolism, and antibody production.,” Biotechnology and Bioengineering, vol. 39, no. 4, pp. 418–31, Feb. 1992.

[126] K. Chen, Q. Liu, L. Xie, P. A. Sharp, and D. I. Wang, “Engineering of a mammalian cell line for reduction of lactate formation and high monoclonal antibody production.,” Biotechnology and Bioengineering, vol. 72, no. 1, pp. 55–61, Jan. 2001.

[127] C. Paredes, E. Prats, J. J. Cairó, F. Azorín, L. Cornudella, and F. Gòdia, “Modification of glucose and glutamine metabolism in hybridoma cells through metabolic engineering.,” Cytotechnology, vol. 30, no. 1–3, pp. 85–93, Jul. 1999.

[128] H. Dorai, Y. S. Kyung, D. Ellis, C. Kinney, C. Lin, D. Jan, G. Moore, and M. J. Betenbaugh, “Expression of anti-apoptosis genes alters lactate metabolism of Chinese Hamster Ovary cells in culture.,” Biotechnology and Bioengineering, vol. 103, no. 3, pp. 592–608, Jun. 2009.

[129] H. J. Cruz, A. S. Ferreira, C. M. Freitas, J. L. Moreira, and M. J. Carrondo, “Metabolic responses to different glucose and glutamine levels in baby hamster kidney cell culture.,” Applied Microbiology and Biotechnology, vol. 51, no. 5, pp. 579–85, May 1999.

[130] W. Zhou, J. Rehm, and W. S. Hu, “High viable cell concentration fed-batch cultures of hybridoma cells through on-line nutrient feeding.,” Biotechnology and Bioengineering, vol. 46, no. 6, pp. 579–87, Jun. 1995.

[131] C. Altamirano, A. Illanes, S. Becerra, J. J. Cairó, and F. Gòdia, “Considerations on the lactate consumption by CHO cells in the presence of galactose.,” Journal of Biotechnology, vol. 125, no. 4, pp. 547–556, 2006.

[132] J. Li, C. L. Wong, N. Vijayasankaran, T. Hudson, and A. Amanullah, “Feeding lactate for CHO cell culture processes: impact on culture metabolism and performance.,” Biotechnology and Bioengineering, vol. 109, no. 5, pp. 1173–86, May 2012.

[133] J. Osman, J. Birch, and J. Varley, “The response of GS-NS0 myeloma cells to pH shifts and pH perturbations.,” Biotechnology and Bioengineering, vol. 75, no. 1, pp. 63–73, Oct. 2001.

[134] W. S. Ahn and M. R. Antoniewicz, “Parallel labeling experiments with [1,2-(13)C]glucose and [U-(13)C]glutamine provide new insights into CHO cell metabolism.,” Metabolic Engineering, vol. 15, pp. 34–47, Jan. 2013.

[135] M. W. Covert and B. O. Palsson, “Constraints-based models: Regulation of Gene Expression Reduces the Steady-state Solution Space,” Journal of Theoretical Biology, vol. 221, no. 3, pp. 309–325, Apr. 2003.


[136] H. P. Bonarius, G. Schmid, and J. Tramper, “Flux analysis of underdetermined metabolic networks: the quest for the missing constraints,” Trends in Biotechnology, vol. 15, no. 8, pp. 308–314, 1997.

[137] L.-E. Quek, S. Dietmair, J. O. Krömer, and L. K. Nielsen, “Metabolic flux analysis in mammalian cell culture.,” Metabolic Engineering, vol. 12, no. 2, pp. 161–71, Mar. 2010.

[138] H. P. Bonarius, B. Timmerarends, C. D. de Gooijer, and J. Tramper, “Metabolite-balancing techniques vs. 13C tracer experiments to determine metabolic fluxes in hybridoma cells.,” Biotechnology and Bioengineering, vol. 58, no. 2–3, pp. 258–62, 1998.

[139] V. S. Martínez, S. Dietmair, L.-E. Quek, M. P. Hodson, P. Gray, and L. K. Nielsen, “Flux balance analysis of CHO cells before and after a metabolic switch from lactate production to consumption.,” Biotechnology and Bioengineering, vol. 110, no. 2, pp. 660–6, Feb. 2013.

[140] N. Sengupta, S. T. Rose, and J. A. Morgan, “Metabolic flux analysis of CHO cell metabolism in the late non-growth phase.,” Biotechnology and Bioengineering, vol. 108, no. 1, pp. 82–92, 2011.

[141] J. Dean and P. Reddy, “Metabolic analysis of antibody producing CHO cells in fed-batch production.,” Biotechnology and Bioengineering, vol. 110, no. 6, pp. 1735–47, Jun. 2013.

[142] J. Greene, J. H. Jr, and J. Wikswo, “Rapid and precise determination of cellular amino acid flux rates using HPLC with automated derivatization with absorbance detection,” Agilent Pub, pp. 1–8, 2009.

[143] T. D. Schmittgen and K. J. Livak, “Analyzing real-time PCR data by the comparative CT method,” Nature Protocols, vol. 3, no. 6, pp. 1101–1108, Jun. 2008.

[144] A. Varma and B. O. Palsson, “Metabolic Flux Balancing: Basic Concepts, Scientific and Practical Use.,” Nature Biotechnology, vol. 12, no. 10, pp. 994–998, Oct. 1994.

[145] M. Kanehisa and S. Goto, “KEGG: kyoto encyclopedia of genes and genomes.,” Nucleic Acids Research, vol. 28, no. 1, pp. 27–30, Jan. 2000.

[146] H. P. Bonarius, V. Hatzimanikatis, K. P. Meesters, C. D. de Gooijer, G. Schmid, and J. Tramper, “Metabolic flux analysis of hybridoma cells in different culture media using mass balances.,” Biotechnology and Bioengineering, vol. 50, no. 3, pp. 299–318, May 1996.


[147] S. Selvarasu, I. A. Karimi, G.-H. Ghim, and D.-Y. Lee, “Genome-scale modeling and in silico analysis of mouse cell metabolic network.,” Molecular Biosystems, vol. 6, no. 1, pp. 152–61, Jan. 2010.

[148] K. Sheikh, J. Förster, and L. K. Nielsen, “Modeling hybridoma cell metabolism using a generic genome-scale metabolic model of Mus musculus.,” Biotechnology Progress, vol. 21, no. 1, pp. 112–21, 2005.

[149] J. M. Savinell and B. O. Palsson, “Network analysis of intermediary metabolism using linear optimization,” Journal of Theoretical Biology, vol. 154, no. 4, pp. 455–473, Feb. 1992.

[150] H. P. Bonarius, A. Ozemre, B. Timmerarends, P. Skrabal, J. Tramper, G. Schmid, and E. Heinzle, “Metabolic-flux analysis of continuously cultured hybridoma cells using (13)CO(2) mass spectrometry in combination with (13)C-lactate nuclear magnetic resonance spectroscopy and metabolite balancing.,” Biotechnology and Bioengineering, vol. 74, no. 6, pp. 528–38, Sep. 2001.

[151] R. P. Nolan and K. Lee, “Dynamic model of CHO cell metabolism,” Metabolic Engineering, vol. 13, no. 1, pp. 108–124, 2011.

[152] C. Pinheiro, A. Longatto-Filho, J. Azevedo-Silva, M. Casal, F. C. Schmitt, and F. Baltazar, “Role of monocarboxylate transporters in human cancers: state of the art.,” Journal of Bioenergetics and Biomembranes, pp. 127–139, Mar. 2012.

[153] S. C. Burgess, N. Hausler, M. Merritt, F. M. H. Jeffrey, C. Storey, A. Milde, S. Koshy, J. Lindner, M. A. Magnuson, C. R. Malloy, and a D. Sherry, “Impaired tricarboxylic acid cycle activity in mouse livers lacking cytosolic phosphoenolpyruvate carboxykinase.,” The Journal of Biological Chemistry, vol. 279, no. 47, pp. 48941–9, Nov. 2004.

[154] P. Hakimi, J. Yang, G. Casadesus, D. Massillon, F. Tolentino-Silva, C. K. Nye, M. E. Cabrera, D. R. Hagen, C. B. Utter, Y. Baghdy, D. H. Johnson, D. L. Wilson, J. P. Kirwan, S. C. Kalhan, and R. W. Hanson, “Overexpression of the cytosolic form of phosphoenolpyruvate carboxykinase (GTP) in skeletal muscle repatterns energy metabolism in the mouse.,” The Journal of Biological Chemistry, vol. 282, no. 45, pp. 32844–55, Nov. 2007.

[155] S. K. Yoon, S. L. Choi, J. Y. Song, and G. M. Lee, “Effect of culture pH on erythropoietin production by Chinese hamster ovary cells grown in suspension at 32.5 and 37.0 degrees C.,” Biotechnology and Bioengineering, vol. 89, no. 3, pp. 345–56, Feb. 2005.

[156] J. Muthing, S. E. Kemminer, H. S. Conradt, D. Sagi, M. Nimtz, U. Kärst, and J. Peter-Katalinić, “Effects of buffering conditions and culture pH on production rates and glycosylation of clinical phase I anti-melanoma mouse IgG3 monoclonal antibody R24.,” Biotechnology and Bioengineering, vol. 83, no. 3, pp. 321–34, Aug. 2003.


[157] T. P. Munro, S. M. Mahler, E. P. Huang, D. Y. Chin, and P. P. Gray, “Bridging the gap: facilities and technologies for development of early stage therapeutic mAb candidates.,” mAbs, vol. 3, no. 5, pp. 440–52, 2011.

[158] P. W. Sauer, J. E. Burky, M. C. Wesson, H. D. Sternard, and L. Qu, “A high-yielding, generic fed-batch cell culture process for production of recombinant antibodies,” Biotechnology and Bioengineering, vol. 67, no. 5, pp. 585–597, 2000.

[159] J. Wayte, R. Boraston, H. Bland, J. Varley, and M. Brown, “pH: Effects on growth and productivity of cell lines producing monoclonal antibodies: Control in large-scale fermenters,” Genetic Engineer and Biotechnologist, vol. 17, no. 2–3, pp. 125–132, 1997.

[160] H. Swiderek and M. Al-Rubeai, “Functional genome-wide analysis of antibody producing NS0 cell line cultivated at different temperatures.,” Biotechnology and Bioengineering, vol. 98, no. 3, pp. 616–30, Oct. 2007.

[161] H. Kaufmann, X. Mazur, M. Fussenegger, and J. E. Bailey, “Influence of low temperature on productivity, proteome and protein phosphorylation of CHO cells.,” Biotechnology and Bioengineering, vol. 63, no. 5, pp. 573–82, Jun. 1999.

[162] M. Shi, Z. G. Xie, M. Yu, B. F. Shen, and N. Guo, “Controlled growth of Chinese hamster ovary cells and high expression of antibody-IL-2 fusion proteins by temperature manipulation,” Biotechnology Letters, vol. 27, no. 23–24, pp. 1879–1884, 2005.

[163] A. Moore, J. Mercer, G. Dutina, C. J. Donahue, K. D. Bauer, J. P. Mather, T. Etcheverry, and T. Ryll, “Effects of temperature shift on cell cycle, apoptosis and nucleotide pools in CHO cell batch cultues.,” Cytotechnology, vol. 23, no. 1–3, pp. 47–54, Jan. 1997.

[164] P. Ducommun, P.-A. Ruffieux, A. Kadouri, U. von Stockar, and I. W. Marison, “Monitoring of temperature effects on animal cell metabolism in a packed bed process.,” Biotechnology and Bioengineering, vol. 77, no. 7, pp. 838–42, Mar. 2002.

[165] S. M. Schatz, R. J. Kerschbaumer, G. Gerstenbauer, M. Kral, F. Dorner, and F. Scheiflinger, “Higher expression of Fab antibody fragments in a CHO cell line at reduced temperature.,” Biotechnology and Bioengineering, vol. 84, no. 4, pp. 433–8, Nov. 2003.

[166] S. Reuveny, D. Velez, J. D. Macmillan, and L. Miller, “Factors affecting monoclonal antibody production in culture.,” Developments in Biological Standardization, vol. 66, pp. 169–75, Jan. 1987.


[167] G. K. Sureshkumar and R. Mutharasan, “The influence of temperature on a mouse-mouse hybridoma growth and monoclonal antibody production.,” Biotechnology and Bioengineering, vol. 37, no. 3, pp. 292–5, Feb. 1991.

[168] N. Barnabé and M. Butler, “Effect of temperature on nucleotide pools and monoclonal antibody production in a mouse hybridoma.,” Biotechnology and Bioengineering, vol. 44, no. 10, pp. 1235–45, Nov. 1994.

[169] J. W. Bloemkolk, M. R. Gray, F. Merchant, and T. R. Mosmann, “Effect of temperature on hybridoma cell cycle and MAb production.,” Biotechnology and Bioengineering, vol. 40, no. 3, pp. 427–31, Jul. 1992.

[170] M. B. Al-Fageeh, R. J. Marchant, M. J. Carden, and C. M. Smales, “The cold-shock response in cultured mammalian cells: Harnessing the response for the improvement of recombinant protein production,” Biotechnology and Bioengineering, vol. 93, no. 5, pp. 829–835, 2006.

[171] K. Sunley and M. Butler, “Strategies for the enhancement of recombinant protein production from mammalian cells by growth arrest,” Biotechnology Advances, vol. 28, no. 3, pp. 385–394, 2010.

[172] M. Alrubeai and A. N. Emery, “Mechanisms and Kinetics of Monoclonal-Antibody Synthesis and Secretion in Synchronous and Asynchronous Hybridoma Cell-Cultures,” Journal of Biotechnology, vol. 16, no. 1–2, pp. 67–86, 1990.

[173] M. Alrubeai, A. N. Emery, S. Chalder, and D. C. Jan, “Specific Monoclonal-Antibody Productivity And The Cell Cycle-Comparisons Of Batch, Continuous And Perfusion Cultures,” Cytotechnology, vol. 9, no. 1–3, pp. 85–97, 1992.

[174] R. L. Dutton, J. M. Scharer, and M. Moo-Young, “Descriptive parameter evaluation in mammalian cell culture,” Cytotechnology, vol. 26, no. 2, pp. 139–152, 1998.

[175] I. Watanabe and S. Okada, “Effects of temperature on growth rate of cultured mammalian cells,” The Journal of Cell Biology, vol. 32, no. 2, pp. 309–23, Feb. 1967.

[176] M. L. Slater, S. O. Sharrow, and J. J. Gart, “Cell cycle of Saccharomycescerevisiae in populations growing at different rates.,” Proceedings of the National Academy of Sciences, vol. 74, no. 9, pp. 3850–3854, Sep. 1977.

[177] E. Suzuki and D. F. Ollis, “Cell cycle model for antibody production kinetics.,” Biotechnology and Bioengineering, vol. 34, no. 11, pp. 1398–402, Dec. 1989.


[178] S. J. Kromenaker and F. Srienc, “Effect of lactic acid on the kinetics of growth and antibody production in a murine hybridoma: secretion patterns during the cell cycle.,” Journal of Biotechnology, vol. 34, no. 1, pp. 13–34, Apr. 1994.

[179] T. I. Linardos, N. Kalogerakis, and L. A. Behie, “Cell cycle model for growth rate and death rate in continuous suspension hybridoma cultures.,” Biotechnology and Bioengineering, vol. 40, no. 3, pp. 359–68, Jul. 1992.

[180] L. Cazzador and L. Mariani, “Growth and Production Modeling in Hybridoma Continuous Cultures,” Biotechnology and Bioengineering, vol. 42, no. 11, pp. 1322–1330, 1993.

[181] S. R. Fox, U. A. Patel, M. G. S. Yap, and D. I. C. Wang, “Maximizing interferon-gamma production by Chinese hamster ovary cells through temperature shift optimization: Experimental and modeling,” Biotechnology and Bioengineering, vol. 85, no. 2, pp. 177–184, 2004.

[182] P. Pozarowski and Z. Darzynkiewicz, “Analysis of cell cycle by flow cytometry.,” Methods in Molecular Biology, vol. 281, pp. 301–11, Jan. 2004.

[183] K. Uchiyama and S. Shioya, “Modeling and optimization of alpha-amylase production in a recombinant yeast fed-batch culture taking account of the cell cycle population distribution,” Journal of Biotechnology, vol. 71, no. 1–3, pp. 133–141, 1999.

[184] C. M. Lam, K. Sriyudthsak, C. Kontoravdi, K. Kothari, H.-H. Park, E. N. Pistikopoulos, and A. Mantalaris, “Cell Cycle Modelling for Off-line Dynamic Optimisation of Mammalian Cultures,” 18th European Symposium on Computer Aided Process Engineering - ESCAPE18, vol. 25, pp. 109–114, 2008.

[185] B. Basse, B. C. Baguley, E. S. Marshall, W. R. Joseph, B. van Brunt, G. Wake, and D. J. N. Wall, “A mathematical model for analysis of the cell cycle in cell lines derived from human tumors,” Journal of Mathematical Biology, vol. 47, no. 4, pp. 295–312, 2003.

[186] N. Kumar, P. Gammell, and M. Clynes, “Proliferation control strategies to improve productivity and survival during CHO based production culture : A summary of recent methods employed and the effects of proliferation control in product secreting CHO cell lines.,” Cytotechnology, vol. 53, no. 1–3, pp. 33–46, Apr. 2007.

[187] C. Kontoravdi, “An integrated modelling/experimental framework for protein-producing cell cultures,” Imperial College London, 2007.

[188] D. J. Jang and J. P. Barford, “An unstructured kinetic model of macromolecular metabolism in batch and fed-batch cultures of hybridoma cells producing monoclonal antibody,” Biochemical Engineering Journal, vol. 4, no. 2, pp. 153–168, 2000.


[189] S. Tatiraju, M. Soroush, and R. Mutharasan, “Multi-rate nonlinear state and parameter estimation in a bioreactor.,” Biotechnology and Bioengineering, vol. 63, no. 1, pp. 22–32, Apr. 1999.

[190] L. Zhao, L. Fan, J. Q. Wang, H. X. Niu, and W. S. Tan, “Responses of GS-NS0 Myeloma Cells to Osmolality: Cell Growth, Intracellular Mass Metabolism, Energy Metabolism, and Antibody Production,” Biotechnology and Bioprocess Engineering, vol. 14, no. 5, pp. 625–632, 2009.

[191] N. Jenkins and A. Hovey, “Temperature control of growth and productivity in mutant Chinese hamster ovary cells synthesizing a recombinant protein.,” Biotechnology and Bioengineering, vol. 42, no. 9, pp. 1029–36, Nov. 1993.

[192] A. Sanfeliu and G. Stephanopoulos, “Effect of glutamine limitation on the death of attached Chinese hamster ovary cells,” Biotechnology and Bioengineering, vol. 64, no. 1, pp. 46–53, 1999.

[193] B. Rossler, H. Lubben, and G. Kretzmer, “Temperature: A simple parameter for process optimization in fed-batch cultures of recombinant Chinese hamster ovary cells,” Enzyme and Microbial Technology, vol. 18, no. 6, pp. 423–427, 1996.

[194] S. Becerra, J. Berrios, N. Osses, and C. Altamirano, “Exploring the effect of mild hypothermia on CHO cell productivity,” Biochemical Engineering Journal, vol. 60, no. 0, pp. 1–8, 2012.

[195] Y. Eguchi, S. Shimizu, and Y. Tsujimoto, “Intracellular ATP levels determine cell death fate by apoptosis or necrosis.,” Cancer Research, vol. 57, no. 10, pp. 1835–40, May 1997.

[196] C. Kontoravdi, S. P. Asprey, E. N. Pistikopoulos, and A. Mantalaris, “Development of a dynamic model of monoclonal antibody production and glycosylation for product quality monitoring,” Computers & Chemical Engineering, vol. 31, no. 5–6, pp. 392–400, 2007.

[197] F. R. Sidoli, A. Mantalaris, and S. P. Asprey, “Modelling of mammalian cells and cell culture processes,” Cytotechnology, vol. 44, no. 1–2, pp. 27–46, 2004.

[198] A. Kiparissides, M. Koutinas, C. Kontoravdi, A. Mantalaris, and E. N. Pistikopoulos, “‘Closing the loop’ in biological systems modeling — From the in silico to the in vitro,” Automatica, vol. 47, no. 6, pp. 1147–1155, Jun. 2011.

[199] T. Ziehn and A. S. Tomlin, “GUI–HDMR – A software tool for global sensitivity analysis of complex models,” Environmental Modelling & Software, vol. 24, no. 7, pp. 775–785, 2009.

[200] F. R. Sidoli, A. Mantalaris, and S. P. Asprey, “Toward global parametric estimability of a large-scale kinetic single-cell model for mammalian cell


cultures,” Industrial & Engineering Chemistry Research, vol. 44, no. 4, pp. 868–878, 2005.

[201] C. Kontoravdi, S. P. Asprey, E. N. Pistikopoulos, and A. Mantalaris, “Application of global sensitivity analysis to determine goals for design of experiments: An example study on antibody-producing cell cultures,” Biotechnology Progress, vol. 21, no. 4, pp. 1128–1135, 2005.

[202] C. A. Sellick, A. S. Croxford, A. R. Maqsood, G. Stephens, H. V Westerhoff, R. Goodacre, and A. J. Dickson, “Metabolite profiling of recombinant CHO cells: designing tailored feeding regimes that enhance recombinant antibody production,” Biotechnology and Bioengineering, vol. 108, no. 12, pp. 3025–3031, 2011.

[203] J. S. Seo, Y. J. Kim, J. M. Cho, E. Baek, and G. M. Lee, “Effect of culture pH on recombinant antibody production by a new human cell line, F2N78, grown in suspension at 33.0 °C and 37.0 °C.,” Applied Microbiology and Biotechnology, vol. 97, no. 12, pp. 5283–91, Jun. 2013.

[204] H. Le, S. Kabbur, L. Pollastrini, Z. Sun, K. Mills, K. Johnson, G. Karypis, and W.-S. Hu, “Multivariate analysis of cell culture bioprocess data--lactate consumption as process indicator.,” Journal of Biotechnology, vol. 162, no. 2–3, pp. 210–23, Dec. 2012.

[205] C. P. Peter, Y. Suzuki, and J. Büchs, “Hydromechanical stress in shake flasks: correlation for the maximum local energy dissipation rate.,” Biotechnology and Bioengineering, vol. 93, no. 6, pp. 1164–76, Apr. 2006.

[206] H. Sandberg, D. Lütkemeyer, S. Kuprin, M. Wrangel, A. Almstedt, P. Persson, V. Ek, and M. Mikaelsson, “Mapping and partial characterization of proteases expressed by a CHO production cell line.,” Biotechnology and Bioengineering, vol. 95, no. 5, pp. 961–71, Dec. 2006.

[207] K. Furukawa and K. Ohsuye, “Effect of culture temperature on a recombinant CHO cell line producing a C-terminal α-amidating enzyme.,” Cytotechnology, vol. 26, no. 2, pp. 153–64, Mar. 1998.

[208] V. Hendrick, P. Winnepenninckx, C. Abdelkafi, O. Vandeputte, M. Cherlet, T. Marique, G. Renemann, A. Loa, G. Kretzmer, and J. Werenne, “Increased productivity of recombinant tissular plasminogen activator (t-PA) by butyrate and shift of temperature: a cell cycle phases analysis,” Cytotechnology, vol. 36, no. 1–3, pp. 71–83, 2001.

[209] R. L. Dutton, J. Scharer, and M. Moo-Young, “Cell cycle phase dependent productivity of a recombinant Chinese hamster ovary cell line,” Cytotechnology, vol. 52, no. 1, pp. 55–69, 2006.


[210] J. E. Bailey, “Mathematical modeling and analysis in biochemical engineering: Past accomplishments and future opportunities,” Biotechnology Progress, vol. 14, no. 1, pp. 8–20, 1998.

[211] J. C. Bigge, T. P. Patel, J. A. Bruce, P. N. Goulding, S. M. Charles, and R. B. Parekh, “Nonselective and efficient fluorescent labeling of glycans using 2-amino benzamide and anthranilic acid.,” Analytical Biochemistry, vol. 230, no. 2, pp. 229–38, Sep. 1995.

[212] A. H. Merry, D. C. A. Neville, L. Royle, B. Matthews, D. J. Harvey, R. A. Dwek, and P. M. Rudd, “Recovery of intact 2-aminobenzamide-labeled O-glycans released from glycoproteins by hydrazinolysis.,” Analytical Biochemistry, vol. 304, no. 1, pp. 91–9, May 2002.

[213] C. Altamirano, A. Illanes, A. Casablancas, X. Gámez, J. J. Cairó, and C. Gòdia, “Analysis of CHO cells metabolic redistribution in a glutamate-based defined medium in continuous culture.,” Biotechnology Progress, vol. 17, no. 6, pp. 1032–41, 2001.

[214] L. Xie and D. I. Wang, “Stoichiometric analysis of animal cell growth and its application in medium design.,” Biotechnology and Bioengineering, vol. 43, no. 11, pp. 1164–74, May 1994.

[215] C. Zupke and G. Stephanopoulos, “Intracellular flux analysis in hybridomas using mass balances and in vitro (13)C NMR.,” Biotechnology and Bioengineering, vol. 45, no. 4, pp. 292–303, Feb. 1995.

123

9 Figure and Table Index

9.1 List of Figures

Figure 2-1: Standard batch cultivation of murine hybridoma cells in a controlled

bioreactor system (n=4). Profile of (A) viable cell concentration, (B) glucose () and

lactate (), (C) monoclonal antibody and (D) glutamine () and ammonia ()

concentrations. .............................................................................................................. 14

Figure 2-2: pH shift experiments during batch cultivation of murine hybridoma cells.

Starting at standard pH condition 7.2 a pH shift was performed in the early exponential

phase indicated by a dashed line and covered a pH range from 6.5 – 8.5. Profile of (A)

viable cell concentration, (B) viability and (C) mAb concentration at different pH

setpoints. ....................................................................................................................... 15

Figure 2-3. Impact of process parameter shifts (pH, osmolarity, dissolved O2 and

sparging) on the specific growth rate 10 hours after the parameter shift. The specific

growth rate was calculated for each condition according to Eq. 2-2 and normalized to

the specific growth rate obtained under standard conditions. The standard deviation

was calculated from four independent standard cultivations. ....................................... 16

Figure 2-4: N-linked glycosylation profile of HFN 7.1 antibody (IgG1). The antibody

was purified at the end of a standard batch culture, glycans were released by PNGase

F, labeled with 2-AB and fluorescently detected by HILIC. The predominant peaks

were fractionated and the masses were measured by MALDI TOF MS and compared

to characterized oligosaccharide structures. Sugar residues are as reported in the

Appendix (Table A 1). .................................................................................................. 19

Figure 2-5: N-linked glycosylation profile of purified monoclonal antibody analyzed at

three different time points throughout a batch culture [t (shift) = 26h, t (max. Xv) = 54

hours, t (end) = 80 hours]. Relative peak areas of (A) a standard batch culture without

process parameter shift were compared to (B) pH-shifted culture to pH 7.8. .............. 20

Figure 2-6: Change in each glycoform fraction ∆Fi induced by process parameter (A)

pH and (B) dissolved O2 (DO). The change in glycoform fraction is expressed as

124 CHAPTER 9: FIGURES AND TABLES

difference of the produced glycoform between the parameter shift time and the end of

the culture, normalized to the change in antibody concentration and was calculated

according to Eq. 2-3. () G0F, () G1F, () G2F, () G3F, () G0, () G2F-

NGNA, () G2F-NGNA2, () G2F-GalNGNA, () G0F-GalNGNA. Sugar residues

are as reported in the Appendix (Table A 1). ............................................................... 23

Figure 2-7: Change in each glycoform fraction ∆Fi induced by process parameter (A)

sparging and (B) osmolarity. The change in glycoform fraction is expressed as

difference of the produced glycoform between the parameter shift time and the end of

the culture, normalized to the change in antibody concentration and was calculated

according to Eq. 2-3. () G0F, () G1F, () G2F, () G3F, () G0, () G2F-

NGNA, () G2F-NGNA2, () G2F-GalNGNA, () G0F-GalNGNA. .................... 25

Figure 3-1: Profiles of (A) viable cell concentration, (B) antibody concentration, (C)

lactate concentration and (D) glucose concentration of a standard cultivation pH 7.2

(n=4) and two pH-shifted cultivations (pH 6.8 and pH 7.8) performed in duplicates. pH

shift was performed once cells reached 1 x 106 cells/mL and is indicated by dashed

line. Measured data points were fitted using logistic equations (3-1), (3-2) and (3-3). 34

Figure 3-2: Specific consumption and production rates of glucose, lactate, various

amino acids (nmol/106cells/h), biomass and mAb (µg/106 cells/h) under different pH

conditions. (A) Standard batch cultivation (pH 7.2) at three time points (WD 1.0, WD

1.5 and WD 1.9) during exponential growth phase. (B) Comparison of two metabolic

states at pH 7.2 and at pH 6.8 evaluated 17 hours after pH-shift (WD 1.9). (C)

Comparison of two metabolic states at pH 7.2 and at pH 7.8 evaluated 17 hours after

pH-shift (WD 1.9). Indicated time points were considered in FBA. ............................ 36


aspartate shuttle, glutaminolysis, biomass and mAb synthesis of standard batch

cultivation at three time points during exponential growth phase. The pathways of

amino acid metabolism are not shown. Values of all calculated fluxes are reported in

the Appendix (Figure A 3). ........................................................................................... 38


aspartate shuttle, glutaminolysis, biomass and mAb synthesis in lactate consumption

CHAPTER 9: FIGURES AND TABLES 125

state (pH 6.8) compared to lactate production state at standard conditions (pH 7.2) at

working day 1.9. The pathways of amino acid metabolism are not shown. Values of all

calculated fluxes are reported in the Appendix (Figure A 4). ...................................... 39

Figure 3-5: Gene expression of 32 genes involved in the central murine metabolism.

Data were measured in triplicates and mRNA expression of the two pH-shifted

conditions was compared relative to mRNA expression at the time of pH shift. 50%

over- or under-expression is considered as not significant. .......................................... 40

Figure 3-6: Intracellular flux distribution including additional mitochondrial PEP

recycle pathway of high pH condition (pH 7.8) compared to lactate production state at

standard conditions (pH 7.2) at working day 1.9. The pathways of amino acid

metabolism are not shown. Values of all calculated fluxes are reported in the Appendix

(Figure A 5). .................................................................................................................. 43

Figure 3-7: Comparison of energy metabolism of lactate consumption state (pH 6.8)

and lactate overproduction state (pH 7.8) with lactate production state during the

exponential growth phase at standard conditions (pH 7.2). Total ATP production rate

by ATP, NADH and FADH2 was calculated considering different originating

pathways. A P/O ratio of 2.5 for NADH and 1.5 for FADH2 was assumed. ................ 44

Figure 4-1: Schematic representation of the cell cycle model. ..................................... 59

Figure 4-2: Sensitivity indices for parameters at simulation times (day 1, 4 and 6). (A)

viable cell concentration (Xv), (B) mAb titer, (C) G0 cell fraction (fG0), (D) G1 cell

fraction (fG1) (E) S cell fraction (fS) (F) G2/M cell fraction (fG2/M). The nomenclature is

equal as in Eq. 4-1 – 4-27. ............................................................................................ 62


concentration (Xv), viability, antibody concentration (mAb), glucose (GLC), lactate

(LAC), glutamine (GLN) and ammonia (AMM) concentrations, and cell cycle

fractions G0/G1, S, G2/M compared to model results (solid line: Xv, LAC, AMM,

G0/G1, S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN). ......................... 64





fractions G0/G1, S, G2/M compared to model results (solid line: Xv, LAC, AMM,

G0/G1, S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN). ......................... 66

Figure 4-5: Controlled fed-batch culture at 37 °C. Experimental results of viable cell



fractions G0, G1, S, G2/M compared to model results (solid line: Xv, LAC, AMM, G1,

S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN, G0 fraction). ................. 68

Figure 4-6: Controlled fed-batch culture shifted to 33 °C at culture start. Experimental

results of viable cell concentration (Xv), viability, antibody concentration (mAb),

glucose (GLC), lactate (LAC), glutamine (GLN) and ammonia (AMM) concentrations,

and cell cycle fractions G0, G1, S, G2/M compared to model results (solid line: Xv,

LAC, AMM, G1, S, G2/M fractions; dashed lines: viability, mAb, GLC, GLN, G0

fraction). ........................................................................................................................ 70

Figure 4-7: Controlled fed-batch culture with temperature shift to 33 °C after 30 h.

Shift is indicated by dotted line. Experimental results of viable cell concentration (Xv),

viability, antibody concentration (mAb), glucose (GLC), lactate (LAC), glutamine

(GLN) and ammonia (AMM) concentrations, and cell cycle fractions G0, G1, S, G2/M

compared to model prediction (solid line: Xv, LAC, AMM, G1, S, G2/M fractions;

dashed lines: viability, mAb, GLC, GLN, G0). ............................................................. 71

Figure 4-8: Temperature shifted controlled fed-batch culture to 33 °C after 76 h. Shift

in indicated by a dotted line. Experimental results of viable cell concentration (Xv),

viability, antibody concentration (mAb), glucose (GLC), lactate (LAC), glutamine

(GLN) and ammonia (AMM) concentrations, and cell cycle fractions G0, G1, S, G2/M

compared to model prediction (solid line: Xv, LAC, AMM, G1, S, G2/M fractions;

dashed lines: viability, mAb, GLC, GLN, G0 fraction). ............................................... 73

Figure A 1: MALDI-TOF MS spectra of the 2-AB labeled oligosaccharide released

from HFN 7.1 antibody by PNGase F in (A) positive ion mode and (B) negative ion

mode. Peaks in positive ion mode are either detected as Na+ or K+ adducts. (C)

Fragmentation pattern of G1F (m/z 1767.70) obtained by MALDI-TOF MS/MS in

CHAPTER 9: FIGURES AND TABLES 127

positive ion mode. Sugar residues are N-acetylglucosamine (GlcNAc, ), fucose (F,

), mannose (M, ), galactose (G, ). ............................................................................ 83

Figure A 2: Flux balance model scheme. PEP regeneration pathway was only active at

pH 7.8 and therefore the fluxes OAAmPEPm and PEPmPEP were constrained

under all other conditions.............................................................................................. 95

Figure A 3: Intracellular fluxes calculated by flux balance analysis at three different

time points (WD 1.0, WD 1.5, WD 1.9) for standard pH 7.2 culture. Numbers

correspond to reactions specified below. ...................................................................... 96



specified below. ............................................................................................................ 97



specified below. Reactions 48 - 50 included to metabolic network as additional

constraints according to gene expression data. ............................................................. 98

9.2 List of Tables

Table 2-1: Degree of galactosylation (GI), sialylation (SI) and fucosylation (FI) as a

function of the single shift parameter pH or osmolarity (mOsm/kg), and as a function

of the combined effect of pH and osmolarity. .............................................................. 24

Table A 1: Measured masses of reductively aminated oligosaccharides determined by

MALDI-TOF MS and compared to theoretical masses. Proposed glycoforms were

assigned to the glycosylation profile in Figure 2-4....................................................... 84

Table A 2: List of target genes investigated by gene expression analysis.................... 86

Table A 3: Reactions in the metabolic network used for flux balance analysis. .......... 88

Table A 4: Averaged dry cell composition (protein, carbohydrates, nucleotides, lipids)

of murine hybridoma cells derived from literature [148]. ........................................... 93


Table A 5: Amino acid composition of IgG1 derived from an average IgG1 protein

sequence [137]. ............................................................................................................. 94

Table A 6: Model parameter values. ............................................................................. 99

rights / license: research collection in copyright - non ...14495/... · impact of process...

Documents