Metabolomics study of CHO cell cultures

TABLE OF CONTENTS Acknowledgements 2 Table of contents 4 Summary 9 List of tables 11 List of figures 12 CHAPTER 1 Introduction 14 1.1 Background 14 1.2 Thesis objectives 15 1.3 T

METABOLOMICS STUDY OF CHINESE HAMSTER OVARY CELL CULTURES

CHONG POOI KAT WILLIAM

B.Sc.(Hons.), NUS

A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF PAEDIATRICS YONG LOO LIN SCHOOL OF MEDICINE

NATIONAL UNIVERSITY OF SINGAPORE

2011

ACKNOWLEDGEMENTS

This PhD thesis is a product that emerged from a long and sometimes arduous journey It would not have been completed without the following people whom I would like to thank:

To my supervisors, Dr Heng Chew Kiat, Professor Miranda Yap and

Dr Niki Wong, I would like to extend my deepest gratitude for your guidance, patience and unwavering support You also nudged me to put in the extra effort to make the results that more impactful and better conveyed My thanks and gratefulness also to Dr Ho Ying Swan who performed like a supervisor by being much involved in discussions, and for providing numerous useful advices and leads

Many thanks go to Toh Poh Choo and Dr Lee Yih Yean from the Animal Cell Technology group for teaching me the fine art of setting up and operating a bioreactor culture Janice Tan, Jessna Yeo and Wong Chun Loong were also essential for maintaining the smooth operation of the lab so that I could conduct my experiments successfully (some of the time) I would also like to thank Dr Lee Dong-Yup, Faraaz Yusufi and Satty Reddy for their technical assistance in implementing the bioinformatics software

My appreciation also to the people from Analytics group who helped

me troubleshoot the HPLC systems when the machines didn’t behave, and who assisted me in amino acid analysis They are Corrine Wan, Ong Boon Tee,

Sim Lyn Chin, Gavin Teo, Eddie Tan and Hoi Kong Meng In addition, the colleagues from Microarray and Proteomics groups were wonderful people to let me use their lab consumables without having to replenish them

Two attachment students under my supervision deserved special mentioning Tan Liting from NUS rendered invaluable assistance to clone out the malate dehydrogenase gene Chow Chung Ping from NTU helped to quantify some of the extracellular metabolites Besides the students, Dr Goh Lin-Tang (formerly from Waters Asia Ltd) was instrumental to assist me in the operation of the SYNAPTTM HDMS

Last but not least, I would like to dedicate this thesis to my family My lovely wife, Mandy, who was and will always be there by my side through the good and not so good times My parents, Alan and Esther, who sacrificed a lot

to provide me with a good education and a loving environment to grow up in And to my five months old baby girl, Chloe, you are the icing on my cake at the end of this long and sometimes arduous journey!

This work was supported by the Biomedical Research Council of A*STAR (Agency for Science, Technology and Research), Singapore

TABLE OF CONTENTS

Acknowledgements 2

Table of contents 4

Summary 9

List of tables 11

List of figures 12 CHAPTER 1 Introduction 14

1.1 Background 14

1.2 Thesis objectives 15

1.3 Thesis organization 16

CHAPTER 2 Literature Review 18

2.1 Metabolomics: the latest complement of functional genomics 18 2.1.1 Challenges of metabolomics analysis 20

2.1.2 Mass spectrometry-based metabolomics 21

2.1.3 Metabolite identification 23

2.1.4 Metabolomics to understand recombinant protein-producing 25 cell cultures 2.2 Mammalian cell culture for the production of recombinant 25 proteins 2.2.1 Accumulation of metabolites in mammalian cell culture medium 26 2.2.2 Induction of apoptosis by external stimuli 27

2.3 Metabolic engineering 28

2.3.1 Metabolic engineering in CHO 29

2.3.2 Malate dehydrogenase 30

Page 3.1.1 Bioreactor fed-batch culture operation 32

3.1.3 Enzyme-linked immunosorbent assay for IgG quantification 34

3.2.1 Liquid chromatography – mass spectrometry (LC-MS) analysis 34

3.2.4 MS2-based metabolite identification 39 3.2.5 MS-based targeted metabolite quantification 41

(MDH II) gene

3.3.4 Western blot analysis of MDH II protein level 46

3.4.1 Assessment of enzymatic bottleneck at MDH II 47 3.4.2 Assessment of apoptosis induction 48

4.1 CHO mAb fed-batch bioreactor cultures 52

4.2 LC-MS analysis of CHO mAb culture medium 54

4.3 MS data pre-processing and statistical analysis 54

Page

4.5 LC-MS strategy produced mass accurate and reproducible 60

data suitable for statistical analysis

4.7 Metabolite identity confirmation was accomplished using 64

two state-of-the-art MS instruments

improved cell growth

5.1 Quantification of extracellular metabolites 69 5.2 Determination of the source of excess malate 71 5.3 Creation of CHO mAb clones stably overexpressing MDH II 74

(CHO MDH II)

5.4 Characterisation of CHO mAb clones stably overexpressing 76

MDH II

5.5 MRM quantification revealed malate was most abundant 79

amongst the accumulated metabolites during cell culture

5.6 Malate efflux from CHO mAb cells was due to aspartate in the 81

medium and enzyme bottleneck at MDH II

5.7 CHO mAb cells stably overexpressing MDH II demonstrated 83

improved cell growth

Page

6.1 Profiling caspase activity of CHO mAb cells in triplicate 85

fed-batch cultures 6.2 Correlation analysis of metabolites to the caspase 86

activity profile 6.3 Assessment of apoptosis-induction by extracellular metabolites 89 DISCUSSION 91 6.4 Correlation analysis facilitated the short listing of 91

metabolites of interest, whose trends required verification by absolute quantification 6.5 Nucleotides and nucleosides formed a major group of 92 extracellular metabolites that correlated with caspase activities 6.6 Amino acid derivatives formed another major group of 95 extracellular metabolites that correlated with caspase activities 6.7 Addition of GSSG to CHO mAb cells resulted in the greatest 96 increase in caspase activity

CONCLUSION 98

CHAPTER 7 Conclusions and recommendations 99 7.1 Conclusions 99

7.2 Recommendations 101

7.2.1 Improve throughput of the metabolomics analysis 101

7.2.2 Account for mass peaks with no identities 103

7.2.3 Improvement in CHO MDH II culture performance 104

7.2.4 Intracellular metabolomics to verify mechanisms of 105

apoptosis-inducing metabolites 7.2.5 Integration of “-omics” data 106

Abbreviations 107

Page

with putative metabolite identities

glutamylphenylalanine

equations of the extracellular metabolites

quantified by multiple reactions monitoring scans

displayed strong correlation (R2>0.8) with

intracellular caspase activities

Metabolomics study of CHO cell cultures

SUMMARY

Chinese hamster ovary (CHO) cells are predominantly used for the production of recombinant proteins Currently, with the exception of ammonia and lactate, there is little knowledge of other metabolites that are released by CHO cells during culture This thesis describes the development of a liquid chromatography-mass spectrometry metabolomics strategy to identify extracellular metabolites in recombinant CHO fed-batch cultures 11 out of 12 identified metabolites were reported for the first time in CHO culture medium Amongst the metabolites, malate accumulation was the highest at 329 µM Malate efflux was found to be due to the supply of aspartate and possibly an enzymatic bottleneck at malate dehydrogenase II (MDH II) Subsequent metabolic engineering to overexpress MDH II in CHO cells resulted in 3 fold increases in intracellular ATP and NADH In addition, integral cell number and product titer increased by up to 1.9 and 1.2 fold respectively This is the first report of a metabolic engineering target that was successfully identified using a metabolomics approach In addition, a panel of extracellular metabolites exhibited good correlation with intracellular caspase activity Some of these metabolites, such as oxidized glutathione and AMP, were shown to induce apoptosis This result could offer the first insight into why CHO cells in fed-batch cultures eventually senesce and die, even though nutrients were not limiting and ammonia and lactate were below inhibitory levels These findings demonstrate how a metabolomics-based strategy enabled the identification of key cell growth and death-related metabolites in

recombinant CHO cultures, thereby guiding metabolic engineering efforts to improve culture performance and product titer

LIST OF TABLES

techniques

sample showing reproducible peak areas and retention times

(relative standard deviation <15%) over eight consecutive

LC-MS analysis runs with the Atlantis T3 column

by the 4000 QTrap mass spectrometer to detect metabolites of interest

induction assay, and their effective concentrations in the

CHO mAb 6 well-plate cultures

via MS2 experiments

of CHO mAb fed-batch cultures (n=3) as isolated by metabolomics

analysis, and their average concentrations

levels between parental CHO mAb and CHO MDH II-A cells in

fed-batch bioreactor cultures (n=2)

correlation (R2 > 0.8) with intracellular caspase activities in at

least two of the three CHO mAb fed-batch bioreactor cultures

LIST OF FIGURES

genome, transcriptome, proteome and metabolome to produce a

cellular phenotype

(MDH) I and mitochondrial MDH II

samples generated from the LC programs utilizing the

(A) Hypersil Gold column and (B) Atlantis T3 column

analysis of CHO mAb fed-batch culture medium

bioreactor cultures (B) The transition points of the various phases

of growth (exponential, stationary and death) were determined by

taking the natural log of the viable cell density

sample generated from the LC program utilizing the Atlantis T3

column

data from a CHO mAb fed-batch bioreactor culture

(A) dimethylarginine standard, and (B) Day 6 culture medium

using the LTQ-Orbitrap mass spectrometer

mass selection at 295, for the confirmation of putative identity of

glutamylphenylalanine

molecule (m/z 295.0946) and the putative glutamylphenylalanine

(m/z 295.1292) were separated by ion mobility spectrometry within

the SYNAPTTM HDMS over real time chromatographic separation

of the sample

bioreactor day 8 culture medium

Page

in CHO mAb cell culture

and the amount of malate released by CHO mAb cells one hour

post-inoculation in shake flask cultures (n=2)

C13-fumarate (Δ) after exposure of CHO mAb cells to C13-aspartate

(arrowed) with expected size of 1131bp was generated

pcDNA3.1(-)/Hygro using Apa I and Hind III restriction enzymes

between CHO mAb parental and two CHO MDH II clones

( ) and CHO MDH II (Clone A: Δ; Clone B: O) cells in fed-batch

shake flask cultures (B) Cell viability comparison of the same cultures

CHO MDH II-A (Δ) cells in fed-batch bioreactor cultures (n=2)

of the triplicate CHO mAb fed-batch bioreactor cultures

standard (top panel) and the corresponding m/z value in day 6

culture medium sample (bottom panel)

the three CHO mAb fed-batch cultures

compared to control cells (n=3), after being exposed for four hours

to various metabolites

and apoptosis by extracellular ATP and adenosine

CHAPTER 1 INTRODUCTION

1.1 BACKGROUND

The biologics industry is facing an increasing global demand for recombinant proteins that are used as biotherapeutics (Maggon, 2007; Kelley, 2009) Therefore, there are constant efforts to devise strategies to achieve high cell density cultures as this is usually accompanied by high recombinant protein yields Fed-batch bioreactor systems are commonly employed for growing mammalian cell cultures to high cell densities (Bibila and Robinson, 1995; Hu and Aunins, 1997) In such systems, parameters like pH and dissolved oxygen level are controlled stringently, and nutrients are fed constantly according to cell demand

Mammalian cells in culture produce lactate and ammonia as by- products in the course of metabolizing glucose and glutamine respectively Lactate and ammonia at high concentrations are to be avoided because they

are detrimental to cell growth and protein production (Singh et al., 1994; Yang and Butler, 2000; Xing et al, 2008) In fed-batch cultures, lactate and ammonia

production are minimised by the controlled feeding of glucose and glutamine

respectively (Glacken et al., 1986; Lee et al., 2003) However, cell

senescence and death are only temporarily delayed, suggesting that other detrimental extracellular metabolites in the culture medium might also be accumulating

Analysis of extracellular metabolites can be used to determine enzyme bottlenecks in biochemical pathways which have impact on cell growth For example, hexokinase, the first enzyme in glycolysis which converts glucose to glucose-6-phosphate, was identified as a bottleneck in Chinese hamster ovary (CHO) cells based on measurements of glucose consumption rate (Neermann and Wagner, 1996)

Apart from lactate and ammonia, there are very few studies on changes

in extracellular metabolites in mammalian cell cultures These studies have been confined to culture systems in medical research such as human hepatoma

cells (Miccheli et al., 2006) and human embryonic stem cells (Cezar et al.,

2007) More recently, studies related to biologics research involved determining metabolites responsible for shift in lactate metabolism in CHO

cells (Ma et al., 2009) and for productivity increase in growth arrested NS0

cells (Khoo and Al-Rubeai, 2009) However, these studies did not result in any applications to improve cell growth or protein production

The main aim of the thesis is to determine new strategies for improving the growth of CHO cells and consequently improve the recombinant protein yield, through a metabolomics approach

The scope of the thesis involved:

(1) Developing a metabolomics strategy based on high performance liquid chromatography-mass spectrometry (LC-MS) for the analysis

of metabolites in culture medium from recombinant CHO fed-batch cultures

(2) Identifying an extracellular metabolite that was associated with growth limitation and performing metabolic engineering in CHO cells for improved cell growth and recombinant protein yield

(3) Establishing a methodology for the evaluation of extracellular metabolites that were associated with cell death, thereby proposing strategies for the improvement of cell growth and recombinant protein yield

There are seven chapters in this thesis Chapter 1 provides a brief introduction and defines the objective and scope of the thesis Chapter 2 presents literature review on (i) metabolomics, (ii) mammalian cell culture and apoptosis, and (iii) metabolic engineering A detailed description of the materials and methods used is covered in Chapter 3, and in Chapter 4, the development of a metabolomics strategy to analyze metabolites in culture medium from fed-batch cultures of CHO expressing monoclonal antibodies (CHO mAb) In Chapter 5, the results of the overexpression of malate dehyrdogenase II in CHO mAb are reported Chapter 6 describes the

identification of extracellular metabolites that induces apoptosis in the CHO mAb cells Chapter 7 summarizes the important conclusions resulting from this study and provides recommendations for future work

CHAPTER 2 LITERATURE REVIEW

In the past decade, functional genomics is increasingly being utilized

to describe the complex relationship between genome and phenotype in a living organism (Evans et al, 1997; Borrebaeck, 1998) In this regard, various

‘omics’ technologies have been employed to support functional genomics by characterizing the genome, transcriptome, proteome and metabolome in cells (Glinski and Weckwerth, 2006; Geschwind and Konopka, 2009) which is also known as the ‘omics’ cascade (Figure 2.1) This enables the understanding of the multi-leveled cellular responses to stimuli and environmental or genetic changes

Metabolomics is the latest ‘omics’ platform in the complement of functional genomics (Griffin 2006; Baran et al, 2009) It is defined as the unbiased qualitative or (semi) quantitative measurement and identification of small molecules with molecular weight of less than 1000 Daltons (Da) in a biological system (Fiehn 2002) These low molecular weight molecules are also known as metabolites and can exist within the cell, as well as in the extracellular environment They take part in biochemical reactions, linking the effects of various enzymatic pathways to produce the observed phenotype Therefore, the metabolome is context dependent, and is representative of the integrative changes in the genome, transcriptome and proteome being experienced by the cell

What can happen

phenotype (adapted from Dettmer et al., 2007)

As the metabolomics technology continues to mature, it has been

successfully applied to bacterial (Villas-Bôas et al., 2006) and plant (Fiehn et

al., 2000) cultures in sample typing and functional genomics studies It has

also enabled the discovery of disease biomarkers in mammalian cells

(Quinones and Kaddurah-Daouk, 2009; Shlomi et al., 2009) and used for the

prognosis of recovery after organ transplants (Wishart, 2006) In addition, it is

widely employed in drug toxicology studies (Robertson 2005; Bando et al.,

2010)

2.1.1 Challenges of metabolomics analysis

Although there are merits to unraveling the metabolome of a living system, a number of challenges have to be overcome that are not always associated with the other ‘omics’ platforms First of all, a single type of metabolite can be involved in numerous biochemical reactions The same metabolite may also exist in various organelles of the cell interacting with a partner specific to the micro-location Therefore it may be difficult to interpret metabolome data and link changes in a metabolite level to a particular gene or protein (family) Furthermore, unlike DNA and proteins which are polymers

of four nucleotides and 22 amino acids respectively, metabolites encompass a wide family with many different chemical groups These chemical groups give rise to diverse metabolite classes such as lipids, fatty acids, sugars, amino acids, amines, nucleosides and alcohols Currently, no single analytical technique is able to detect all known metabolite classes with high efficiency in one operation It is thus not uncommon for more than one analytical technique

to be employed in the analysis of a sample (t’Kindt et al, 2009; Draisma 2010)

Intracellular metabolite concentrations can also vary widely from nano

molar e.g cystathionine ketimine (Ricci et al., 1990) to milli molar e.g ATP (Wang et al., 1997) amounts In addition, the turnover of some metabolite

pools can happen in seconds This presents a challenge to measure molecules, especially those of low abundance, as part of an accurate snap-shot of the metabolome at any given time point Therefore, methods to instantaneously quench cellular metabolism and extract metabolites at high recoveries have

been extensively investigated (Villas-Bôas and Bruheim, 2007; Sellick et al.,

2009; Dietmair et al., 2010)

In order to circumvent the challenges associated with quenching of metabolism and metabolite recoveries, some metabolomics studies were limited to the external milieu of the cells Nonetheless, they yielded useful

information on the cell metabolism (Miccheli et al., 2006; Cezar et al., 2007)

2.1.2 Mass spectrometry-based metabolomics

Metabolomics analysis is typically achieved using nuclear magnetic

resonance (NMR) spectroscopy (Reo, 2002; Macintyre et al., 2010; Schripsema, 2010) or mass spectrometry (MS) (Nordström et al., 2006; Tanaka et al., 2008; Llorach et al., 2009; Mohamed et al., 2009) Amongst the

two platforms, MS-based metabolomics is more commonly used because it has higher sensitivity and is widely available One of the main disadvantages

of using MS is ion suppression whereby the presence of a more abundant

species suppresses the signal of a lower abundant species (Böttcher et al.,

2007) Ion suppression is especially prevalent in complex samples such as the cell matrix To minimize ion suppression, liquid chromatography (LC), gas chromatography (GC) or capillary electrophoresis (CE) is often introduced to allow the separation of molecules in time before they enter the MS Depending

on the column or capillary used, the separation step can also provide additional information regarding the chemical properties of the metabolites

The main advantages and disadvantages of the various techniques used in

metabolomics studies are listed in Table 2.1

Table 2.1 Advantages and disadvantages of various metabolomics

techniques

NMR ƒ Little sample preparation required

ƒ Sample is not destroyed

ƒ Established compounds database

ƒ Lower sensitivity compared to MS detection

LC-MS ƒ Suitable for analysis of a wide

LC/GC/CE-MS based analysis produces a large set of data per sample

comprising three dimensions: m/z values, intensities and elution times In

comparative studies, samples are labeled as ‘treated’ versus ‘non-treated’ or as

a series of time points The nature of the analysis also warrants the use of

biological and technical replicates that needs to be time aligned and

normalized to generate meaningful interpretations All these increase the

complexity of the final total data set Multivariate statistical analysis is thus

often used to reduce data complexity and highlight mass peaks which are

analysis allows for the classification of samples, as well as for the identification of mass peaks exhibiting key changes in time profile One of the commonly used statistic techniques is principal component analysis (PCA) (Joliffe, 1992) It is an unsupervised approach which reduces the dimensionality of a data set by projecting the data onto a new set of coordinates (the principal components) such that variations can be visualized more easily This technique has been applied successfully in a number of

metabolomics studies involving cell cultures (Allen et al., 2003, Miccheli et

al., 2006; Khoo and Al-Rubeai, 2009)

For time course experiments, a second statistical technique known as

hierarchical clustering (De Souza et al., 2006; Andreopoulos et al., 2009)

serves to group together mass peaks with similar changes in intensities over time In addition, these intensity time profiles can be examined for their degree

of correlation to a specific parameter such as cell growth rate (Takahashi et al., 2008; Glauser et al., 2010) Both PCA and hierarchical clustering are

performed to facilitate the shortlisting of interesting changes in metabolite levels

2.1.3 Metabolite identification

Once a mass peak of interest has been shortlisted, a putative metabolite identity is assigned if available The m/z value of the mass peak can be compared against entries found in public databases such as Kyoto Encyclopedia of Genes and Genome (www.genome.jp/kegg), the Human

Metabolome Database (www.hmdb.ca) (Wishart et al., 2009) and ChemSpider

is critical for the MS to possess a mass accuracy of less than 5 ppm in order to significantly reduce the number of possible identities At this accuracy, the observed m/z value is able to match the theoretical value up to the third decimal place Such accuracy is also important when computing the possible elemental composition of novel mass peaks with no matches to known metabolites MS using Time of Flight (TOF) and Fourier Transformation (FT) mass analyzers are able to routinely produce such mass accuracy In fact, FT based MS can achieve a mass accuracy of less than 1 ppm Another advantage

of FT- over TOF-based MS is the former can operate at up to 100,000 resolution compared to 20,000 for the latter, permitting the distinction of molecules which differ by just m/z 0.001 In recent years, a FT-based MS by Thermo Scientific known as the LTQ-Orbitrap MS has been increasingly employed in metabolomics analysis due to its relative ease of use and high

mass accuracy and resolution (Hu et al., 2005; Koulman et al., 2009; Zhang et

al., 2009)

Following the establishment of a putative metabolite identity, the mass peak of interest is subjected to fragmentation by collision in MS2 experiments The resulting daughter ions can be compared against those from standards or public MS2 libraries (e.g from National Institute of Standards and Technology) to confirm the identity of the metabolite If neither is available, computational software that predicts daughter ions based on pre-defined fragmentation rules in literature can be used to provide some degree of

confirmation Two such predictive software used in this study were Mass Frontier (Thermo Scientific) and MassFragmentTM (Waters)

2.1.4 Metabolomics to understand recombinant protein-producing cell

cultures

It is only in the past few years that metabolomics studies on mammalian cell cultures producing recombinant proteins have been published The first publication utilized 1(H) NMR to analyze metabolic perturbations during the growth arrest of an antibody producing mouse myeloma cell line, thereby causing it to be hyper-productive (Khoo and Al-Rubeai, 2009) Metabolites associated with phosphatidylcholine homeostasis, lipid and fatty acid metabolism as well as ascorbate formation were found to be elevated in the growth arrested cultures, possibly establishing links to vesicle secretory functions and mechanisms to counteract reactive oxygen species In another study, GC- and LC-MS were employed to describe metabolite changes during the shift from lactate producing to lactate consuming state in cultures of NSO

and CHO cells expressing monoclonal antibodies (Ma et al., 2009) Most

recently, metabolic profiling with GC-MS enabled the differentiation of baby hamster kidney cell cultures based on cell source, cell age and bioreactor scale

(Chrysanthopoulos et al., 2010) However, specific applications which

enhanced cell culture performance did not emerge from these three studies

2.2 Mammalian cell culture for the production of recombinant

proteins

In the biologics industry, large scale stirred-tank bioreactors of up to 50,000 L are widely used for the production of recombinant protein from mammalian cells (Chu and Robinson, 2001; Andersen and Krummen, 2002; Wurm, 2004) The cells are in suspension to maximize cell density and cultured with serum free medium in either batch or fed-batch mode

Fed-batch culture entails the periodic addition of feed medium to prevent nutrient depletion Although this requires more resources and effort compared to batch culture, the cell density, culture longevity and recombinant

protein yield can be two to five times higher (Reuveny et al., 1986; Flickinger

et al., 1990) The content of the feed medium, the timing and amount to add

are often determined empirically and via a reiterative process (Bibila and Robinson, 1995)

2.2.1 Accumulation of metabolites in mammalian cell culture medium

As there is no medium exchange in batch and fed-batch cultures, metabolites excreted by the cells will accumulate in the medium One of the metabolites is lactate which is generated as a by-product of glucose metabolism This occurs even under aerobic conditions when glucose is not maintained at low concentrations Excess lactate will increase the culture

osmolarity and decrease the pH (Omasa et al., 1992) Another abundant and

well-characterized metabolite that accumulates in culture medium is ammonia

It is generated from the spontaneous breakdown of glutamine (Ozturk et al.,

1990) or from glutamine metabolism

Both lactate and ammonia have been shown to cause growth inhibition and productivity decline at concentrations ranging from 3-60 mM depending

on cell line (Mirabet et al., 1997; Cruz et al., 2000; Yang and Butler, 2000)

Therefore, various strategies were devised to prevent their excessive accumulation In batch cultures, glucose and glutamine have been replaced by

fructose (Duval et al., 1992) and glutamate (Holmlund et al., 1992)

respectively as alternative carbon sources In fed-batch cultures, glucose and glutamine concentrations are kept low by feeding according to demand

(Glacken et al., 1986; Ljunggren and Haggstrom, 1994; Kurokawa et al.,

1994)

2.2.2 Induction of apoptosis by external stimuli

In a production culture, it is desirable to prolong the culture viability for increased recombinant protein yield However, cells in culture are constantly subjected to various types of stress which act as triggers of cell death Acute physical stress, such as shearing against the bioreactor wall, can result in necrotic death (Senger and Karim, 2003) On the other hand, apoptotic cell death can result from nutrient depletion and accumulation of

inhibitory by-products (Simpson et al., 1998; Laken and Leonard, 2001)

Apoptosis is a form of programmed cell death executed by a family of

cysteine proteases known as caspases (Alnemri, 1997; Degterev et al., 2003)

Upon their activation, caspases cleave a number of substrates causing a chain

of events leading to DNA laddering, cell shrinkage and blebbing (Wyllie et al., 1984; Andrade et al., 2009)

Physical damage can be alleviated by measures such as adjustment of the stirrer speed and addition of surfactants Nutrient depletion and accumulation of by-products such as lactate and ammonia are mitigated by the fed-batch strategies that are mentioned above Despite the best of efforts, even cells in a fed-batch culture will eventually enter a stationary phase after a period of exponential growth The stationary phase is characterized by a period of cessation of cell proliferation which will inevitably lead to cellular apoptosis Numerous efforts have been implemented in CHO cultures to counteract apoptosis Some successful strategies involved genetic engineering

to overexpress apoptosis suppressor genes such as Bcl-2 and Bcl-xL

(Fussenegger et al., 2000; Meents et al., 2002; Chiang et al., 2005)

Metabolic engineering involves the implementation of genetic control

on one or more metabolic pathways to alter cellular phenotype such as increasing the biomass or overproducing a specific industrially relevant metabolite Identifying the point of genetic control is not always obvious as the relationship between genes, enzymes and metabolites is non-linear and complex In addition, the regulation and kinetics of the metabolic pathway are

also important considerations (Stephanopoulos et al., 2004)

2.3.1 Metabolic engineering in CHO

As 70% of approved biotherapeutics are produced by CHO cells

(Jayapal et al., 2007), considerable efforts have been invested to improve

CHO culture performance and productivity via metabolic engineering For

example, genes governing cell cycle metabolism such as p21 (Bi et al., 2004)

and p27 (Mazur et al., 1998) have been overexpressed to induce growth arrest

in CHO cells which was accompanied by increased recombinant protein productivity This strategy allowed a two-stage production process which consisted of a proliferation phase leading to the desired cell density, followed

by an extended production phase during which the cells remain arrested

growth-CHO cells have also been engineered to reduce the production of metabolites which are detrimental to cell growth Carbamoyl phosphate synthetase I and ornithine transcarbamoylase functioning in the urea cycle

have been targeted for overexpression to reduce ammonia production (Park et

al., 2000) Similarly, lactate production was reduced by down regulating

lactate dehydrogenase-A expression using siRNAs (Kim and Lee, 2007) In another study, a fructose transporter, GLUT5, was introduced in CHO cells so that they could metabolize fructose instead of glucose, and this minimized the flow of excess carbon to lactate (Wlaschin and Hu, 2007) Nevertheless, numerous other targets in the glycolysis pathway and tricarboxylic acid (TCA) cycle remained to be exploited in CHO, even though they are involved in important functions relating to glucose metabolism and energy production

2.3.2 Malate dehydrogenase

The mitochondrial malate dehyrdrogenase (MDH II) converts malate

to oxaloacetate as part of the TCA cycle to produce energy metabolites One molecule of NAD+ (oxidized nicotinamide adenine dinucleotide) is reduced to NADH for each malate converted The reducing potential in NADH is subsequently used to generate adenosine 5’-triphosphate (ATP) during oxidative phosphorylation It was established that majority of NADH

formation is connected to biomass production (Bakker et al., 2001) However,

when mitochondrial MDH was overexpressed in yeast for determining its role

in extending life span, there was no biomass improvement (Easlon et al 2008)

In contrast, cytosolic malate dehydrogenase (MDH I) has been the target of metabolic engineering efforts to overproduce industrially useful metabolites in yeast and bacteria It converts oxaloacetate to malate (Figure

2.1) and was thus overexpressed in yeast to produce malate (Pines et al., 1997), and in Escherichia coli to produce succinate (Wang et al., 2009)

Figure 2.2 Relationship between cytosolic malate dehydrogenase (MDH)

I and mitochondrial MDH II (adapted from Garrett and Grisham,

Biochemistry 2 nd edition 1999 page 663)

CHAPTER 3 MATERIALS AND METHODS

3.1 CHO bioreactor cultures

3.1.1 Bioreactor fed-batch culture operation

Recombinant CHO cells were grown in our in-house proprietary protein free chemically defined medium (PFCDM) in double-walled, round-bottom glass vessel bioreactors (B Braun) with an initial seed density of 3×105 cells/ml The working volume of the 2 L and 5 L bioreactor was 1.5 L and 4 L respectively The CHO cell line expressed recombinant IgG monoclonal antibody (mAb) against Rhesus D antigen (CHO mAb)

(Chusainow et al., 2009) The temperature of the cultures was maintained at

37 °C with the heated water jacket surrounding the bioreactor vessel pH was maintained at 7.15 with either carbon dioxide injections and/or periodic addition of 7.5% (w/v) sodium bicarbonate (Sigma) solution A mixture of either air/nitrogen or air/oxygen was used to maintain the partial pressure of oxygen (pO2) at 50 % saturation Aeration was provided via a silicone membrane tubing basket A three-blade impeller set at a stir speed of 120 rpm kept the culture uniformly mixed and prevented the suspended cells from settling

Feeding was performed every 1.5 hr with a proprietary protein free, chemically defined feed formulated from a 20 × DMEM/F12 concentrate based on projected glutamine consumption until the next sampling point, such

and cell density in culture medium were determined twice daily using a YSI

7500 MBS biochemcial analyzer (Yellow Springs Instruments) and a Cedex automated cell counter (Roche Innovatis AG) respectively Projected glutamine consumption was calculated by first obtaining the specific glutamine consumption rate and cell growth rate of the previous time block The cell growth rate was used to predict the integral viable cell number that would be achieved in the next time block Assuming the specific glutamine consumption rate of the previous and next time blocks would not change, it will be multiplied against the integral viable cell number to obtain an estimate

of the amount of glutamine that would be consumed in the next time block

where M1 and M2 denote the amount of metabolite (g) in the media at time

points t1 and t2 (hr) and IVC denotes integral viable cell number within the same time points

where VCN denotes the viable cell number (cells) and N is the total number of

data points

3.1.3 Enzyme-linked immunosorbent assay for IgG quantification

10 μg/ml of a goat anti-human IgG, IgA, and IgM antibody (SPD Scientific) was used as the capture antibody and an alkaline phosphatase-conjugated, Fc-specific antibody (Sigma) at 1:2,000 dilution was used as the detection antibody PBS containing 3% (w/v) bovine serum albumin was used

in the blocking step SIGMAFAST™ p-nitrophenyl phosphate substrate (Sigma) was used to detect alkaline phosphatase activity by reading the absorbance at 405 nm The standard curve for the mAb was created by serial dilution of a purified human IgG (Sigma) All incubation steps were carried out for 1 hr at 37 °C Each sample was measured in triplicates

3.2.1 Liquid chromatography – mass spectrometry (LC-MS) analysis

Culture medium was collected and filtered through a 10,000 molecular weight cut-off (MWCO) device (Vivaspin 500 PES membrane, Sartorius) by centrifugation 2 µl of filtered medium was subsequently injected using Surveyor Plus LC system (Thermo Scientific) into either a Hypersil Gold C18 column (2.1 x 50 mm, 5 μm particle size, Thermo Scientific) or an Atlantis T3 C18 column (4.6 x 100 mm, 3 μm particle size, Waters) maintained at 30 °C

(Merck) in deionized MilliQ water (18.2 mΩ), and B: 0.1% formic acid (Merck) in acetonitrile (gradient grade, Merck) A 20 min gradient was employed (5 % B 2 min, 5-50 % B 6 min, 95 % B 7 min, 5 % B 5 min) at a flow rate of 100 μl/min For the Atlantis T3 column, the LC solvents used were A: 0.1% formic acid (Merck) in deionized MilliQ water (18.2 mΩ), and B: 0.1% formic acid (Merck) in methanol (gradient grade, Merck) A 45 min gradient was employed (5 % B 3 min, 5-50 % B 12 min, 95 % B 17 min, 5 %

B 13 min) at a flow rate of 110 μl/min

The LC eluent was directed to a LTQ-Orbitrap MS (Thermo Scientific) Electron spray ionization was carried out in both positive and negative modes

in full scan with the Orbitrap for masses between m/z 80 to 1000 at 30,000 resolution and three microscans Sheath gas was set at 30 (arbitrary units), aux gas at 10 (arbitrary units) and the capillary temperature was at 300 °C The capillary voltage and electrospray voltage were 40 V and 4.5 kV respectively for positive mode ionization, and -38 V and 3.0 kV respectively for negative mode ionization Mass calibration was done using standard LTQ-Orbitrap calibration solution prior to the injection of the samples

Fresh medium was analyzed using the LC-MS with two different LC programs for a preliminary assessment of the system A relatively short 20 min program with the Hypersil Gold column produced a MS base peak profile with few separated features (Figure 3.1A) When the solvent gradient was reduced by extending the program to 45 min, and the column switched to an Atlantis T3, a better separation of compounds with widely differing polarities was achieved (Figure 3.2B) Eight replicate runs with fresh medium were

subsequently carried out on the Atlantis T3 LC-MS system to assess peak area and retention time reproducibility of the medium components (Table 3.1) The relative standard deviation of the peak areas and retention times were less than

11 % and 4 % respectively This is within the 15 % acceptable variability

range set by the U.S Food and Drug Administration (Sangster et al., 2006)

3.2.2 MS data processing

The raw MS data was pre-processed using an in-house software based

on XCMS (Smith et al., 2006) The data was first de-convoluted for mass peak

detection by dividing the spectra into 0.05 m/z wide bins Within each bin, the signal with the maximum intensity was selected, and signals with signal to noise ratio less than 10 were rejected After this step, each bin was filtered using a second derivative Gaussian model as the peak shape

Lastly, an alignment algorithm was used to identify corresponding mass peaks from different MS scans For a given peak from one experiment, the algorithm attempted to find the closest matching mass peak in terms of mass and elution time within a pre-defined threshold in other experiments After alignment, missing mass peaks that were not identified by the algorithm

or not present in the sample were filled in by reading the raw MS data file and generating an estimated value

(B)

Figure 3.1 Typical MS total ion count of fresh media samples generated from the LC programs utilizing the (A) Hypersil Gold column and (B) Atlantis T3 column

3.2.3 MS data statistical analysis

An unsupervised approach – principal component analysis (PCA) (Joliffe and Morgan, 1992) - was used to reduce the dimensionality of the pre-processed MS data set The data was projected onto a new set of coordinates (known as principal components) to realize variations amongst replicates and between samples The first three principal components that accounts for the largest variation in the data were used in the scores plot

Table 3.1 Peaks of medium components from fresh medium sample showing reproducible peak areas and retention times (relative standard deviation <15%) over eight consecutive LC-MS analysis runs with the Atlantis T3 column 80% of the medium components were identified

Medium

component Average peak area Relative standard

deviation

Average retention time (min)

Relative standard deviation

started as its own cluster, which then merged with mass peaks possessing similar profiles

3.2.4 MS 2

-based metabolite identification

The observed masses of interest were compared with entries from the Kyoto Encyclopedia of Genes and Genome (www.genome.jp/kegg) and the Human Metabolome Database (www.hmdb.ca) Mass peaks which matched within a 10 ppm error window were assigned putative identities For putative metabolites where standards were available, both samples and standards were injected separately into the LC-MS to compare their LC elution times and MS2 fragment ions Collision-induced dissociation (CID) MS2 scans were performed at 7500 resolution and one microscan The optimal collision energy for a particular MS2 scan was pre-determined by continuous direct injection of

1 mM standard into the LTQ-Orbitrap MS and observing the variation in signal intensity with the collision energy All analytical-grade standards were obtained from Sigma Aldrich In the scenario where a standard was not available, the observed MS2 fragment ions of a putative metabolite were compared to the theoretical fragments generated by Mass Frontier (Thermo Scientific) software

Where the HPLC-LTQ-Orbitrap failed to provide unambiguous MS2data for molecular structure elucidation to confirm the putative metabolite identity, the LC-SYNAPT™ HDMS (Waters) system was used to provide distinctive accurate mass MS2 data to fulfill this purpose The schematic layout of the SYNAPT™ HDMS instrumentation and the design principles of the traveling-wave device which enables ion mobility separation within the

mass spectrometer have been previously described by Eckers et al (2007) and Giles et al (2004), respectively

The liquid chromatography was performed on the Acquity UPLC™ (Waters) instrument using the identical column and gradient elution conditions

as described for the profiling task using the LC-LTQ-Orbitrap MS, with the exception that solvent flow rate was elevated to 200 μl/min The initial set up

of the SYNAPT™ HDMS system included the ion source sensitivity optimization in ESI+ mode using a solution of 200 ng/µl of sulfadimethoxine and TOF calibration using sodium formate solution, over the mass range of 50 – 350 m/z, operated under the ‘V’ mode optics The acquired RMS residual of the calibration curve fit error was 0.7 ppm 200 ng/µl of sulfadimethoxine was used as the lockmass solution to ensure full time exact mass acquisition during the analysis The optimized set points for the instrument were: capillary voltage, 2.8 kV; sample cone voltage, 30 V; desolvation temperature, 350 °C; source temperature, 100 °C; desolvation gas flow, 500 L/h; TOF detector voltage, 1675 V The instrument method was set up to acquire data at 0.1 s interval The ion mobility separation (IMS) was facilitated within the center section of the Tri-Wave device by maintaining N2 gas pressure at 0.55 mBar, with IMS wave height and velocity maintained at 8V and 280 m/s, respectively Post IMS CID fragmentation was enabled by setting the trap and transfer collision energy (eV) at 6 and 20, respectively, using argon as collision gas (0.027 mbar) The trap and transfer collision energy (eV) values were set at 6 and 20, respectively