Development of chemically defined medium for biopharmaceuticals production using mammalian cell lines guided by metabolic modelling tools and metabolomics measurements

Universidade do Minho Escola de Engenharia Anis Hamdi March 2023 Anis Hamdi UMinho|2023 Development of chemically defined medium for biopharmaceuticals production using mammalian cell lines guided by metabolic modelling tools and metabolomics measurements Development of chemically defined medium for biopharmaceuticals production using mammalian cell lines guided by metabolic modelling tools and metabolomics measurements March 2023 Doctoral Program in Bioengineering Universidade do Minho Escola de Engenharia Anis Hamdi Work developed under supervision of Professor Isabel Cristina A. Pereira da Rocha Doctoral Thesis Universidade do Minho Escola de Engenharia Development of chemically defined medium for biopharmaceuticals production using mammalian cell lines guided by metabolic modelling tools and metabolomics measurements ii DIREITOS DE AUTOR E CONDIÇÕES DE UTILIZAÇÃO DO TRABALHO POR TERCEIROS Este é um trabalho académico que pode ser utilizado por terceiros desde que respeitadas as regras e boas práticas internacionalmente aceites, no que concerne aos direitos de autor e direitos conexos. Assim, o presente trabalho pode ser utilizado nos termos previstos na licença abaixo indicada. Caso o utilizador necessite de permissão para poder fazer um uso do trabalho em condições não previstas no licenciamento indicado, deverá contactar o autor, através do RepositóriUM da Universidade do Minho. Licença concedida aos utilizadores deste trabalho Atribuição-NãoComercial-SemDerivações CC BY-NC-ND https://creativecommons.org/licenses/by-nc-nd/4.0/ iii ACKNOWLEDGMENTS Accomplishing the PhD degree is a dream that became true. After all the 4.5 years working on this project, I want to acknowledge everyone who was part of my life during this period of time, which was in fact very challenging but also exciting, adventurous and unique. First of all, I want to thank my supervisor Professor Isabel Rocha, for her guidance and help throughout the full period of PhD and her support and openness towards new ideas and interesting projects. I want to thank her also for her willingness to help, her positiveness that always boost me to achieve better every time. Second, I want also to thank Professor Juergen Zanghellini for hosting me at his laboratory in Vienna, for his valuable support and help and constant encouragement. Further, I want to thank MIT-Portugal program for the unique learning opportunity that I was part of, all the professors within the program and the administrative. In addition, I want to thank NORTE 2020 regional fund (NORTE-08-5369-FSE-000053), for their support regarding my PhD grant, the center of biological engineering at the University of Minho for being my host institute for the first 2.5 years of the PhD. In addition, I want to thank BOKU and Acib, Vienna, Austria, Austria for being my host institute for 1.5 years of my PhD and finally the University of Vienna, for being my host institute for the last 6 months of my PhD. I want to thank my family for their constant love and support during my few years abroad, especially my mother Neila and sister Jihene, my aunts Nabiha and Najet and my father who left us few years ago, Mouldi. I want to thank my closest people for being such great company along these months/years, and for our priceless relationship, especially Tobias Guy Adams, Klavdja Cesnik, Klaus Leitner, Elena Lascialfari, Xaver Mayr, Carina Koplenig and Guglielmo Papigni and Amine Khammessi. I want to thank deeply my colleagues at BISBII, especially Sophia, Patricia D., Joana A., for being such great colleagues and my colleagues at Juergen Zanghellini and IMICS group especially, Diana S., Nick M., Peter E., Caterina R., Neza N., Guilia B., Alex S., Oliver P., Mathias G., Helena H. and other closest friends for making my PhD much better, especially Raquel C., Raquel M., Martina P., Martina S., Carlotta C., Enmanuel S., Petar K. and Giulia F. iv STATEMENT OF INTEGRITY I hereby declare having conducted this academic work with integrity. I confirm that I have not used plagiarism or any form of undue use of information or falsification of results along the process leading to its elaboration. I further declare that I have fully acknowledged the Code of Ethical Conduct of the University of Minho. v ABSTRACT Development of chemically defined medium for biopharmaceuticals production using mammalian cell lines guided through metabolic modelling tools and metabolomics measurements Systems biology and metabolic engineering tools hold a tremendous promise in improving biomanufacturing attributes. The emergence of omics tools and computational modeling potentiated the development of new approaches to optimize several expression platforms, in particular mammalian cell lines of which Chinese hamster ovary (CHO) cells, the most used platform for recombinant proteins production. This optimization envisions not only growth parameters of CHO, but also the final product titers. In this context, a CHO genome scale metabolic model (iCHO1766) and flux balance analysis (FBA) were used to study metabolic mechanisms in response to variations in environmental constraints (e.g., amino acids levels) aiming at optimizing cell culture medium formulations. Hence, iCHO1766, combined with an in-house developed algorithm (OptiModels) was first used to determine the minimal medium formulation able to sustain growth of both naïve and recombinant CHO cells lines. Subsequently, based on the prediction results, α-ketoglutarate (AKG) was determined as a potential media supplement and its effect on culture was investigated experimentally. Further, spent media analyses were performed to understand the influence of AKG on CHO metabolism and media formulation was optimized based on balancing the levels of non-essential amino acids together with supplementing AKG and ammonium. As a result of adding AKG to the media, growth parameters were improved, and ammonia accumulation during the process was reduced. In addition, recombinant protein titers were increased by 1.9-fold. Following, specific productivities were improved when rebalancing nutrient levels in the media, together with supplementing AKG, leading to more efficient metabolic features of CHO. In conclusion, in silico -based approaches for medium optimization are powerful tools for predicting the metabolic interconnexions within a cell and hold great potential in improving media design and bioprocess optimization. Key words: CHO cells, GSMM, media optimization, α-ketoglutarate (AKG). vi RESUMO Desenvolvimento de meio definido para produção de biofármacos usando células de mamíferos guiado por modelos metabólicos e metabolómica As ferramentas de biologia de sistemas e engenharia metabólica constituem uma grande promessa na melhoria do desempenho da bio-manufactura. As ferramentas “ómicas” e bioinformáticas potencializaram o desenvolvimento de novas abordagens para otimizar os parâmetros de crescimento e o rendimento do produto final em diversas plataformas de expressão, em particular linhas de células de mamíferos, sendo as células de ovário de Hamster Chinês (“CHO”) uma das linhas celulares mais utilizadas para a produção de proteínas recombinantes. Neste contexto, o modelo metabólico à escala do genoma (GSMM) de células CHO iCHO1766 foi utilizado com o objetivo de estudar o comportamento metabólico das células em resposta a variações nas restrições ambientais, por exemplo, níveis de aminoácidos, visando a otimização da formulação do meio de cultura para células CHO. Para estudar essa influência, o modelo, combinado com um algoritmo desenvolvido internamente, foi usado para determinar a formulação de meio mínima para sustentar o crescimento de CHO não recombinantes, bem como de células recombinantes. Portanto, com base nos resultados da previsão, a suplementação de diferentes níveis de α-cetoglutarato (AKG) à composição do meio padrão foi estudada experimentalmente, e foi realizada uma análise do meio resultante para avaliar os efeitos de AKG sobre o metabolismo de CHO. Por fim, a formulação do meio de cultura foi otimizada com base no equilíbrio dos níveis de aminoácidos não essenciais em conjunto com a suplementação de AKG e amónio. A suplementação com diferentes níveis de AKG permitiu melhorar os parâmetros de crescimento e reduzir a acumulação de amónia. Foi ainda observado um aumento nas concentrações e produtividades específicas das células produtoras, sendo que esta foi melhorada em 1,9 vezes. Por conseguinte, ao utilizar a formulação de meio otimizado, observou-se um aumento da produtividade específica das células, bem como características metabólicas mais eficientes. As abordagens in silico para otimização de meio são, assim, ferramentas poderosas para prever a interconexão metabólica na célula e possuem um grande potencial para melhorar o desenho e otimização do meio de culturas. Palavras-chave: células CHO, GSMM, otimização do meio de cultura, α-cetoglutarato (AKG). vii TABLE OF CONTENTS ACKNOWLEDGMENTS ......................................................................................................................... iii RESUMO ................................................................................................................................................ vi TABLE OF CONTENTS ......................................................................................................................... vii LIST OF FIGURES ................................................................................................................................... x LIST OF TABLES .................................................................................................................................. xv LIST OF ABBREVIATIONS AND ACRONYMS ................................................................................... xvi SCIENTIFIC OUTPUT .......................................................................................................................... xix CHAPTER 1 Motivation and outline of the thesis ............................................................................. 1 1.1. Context and motivation ........................................................................................................ 3 1.2. Research aims ..................................................................................................................... 5 1.3. Outline of the thesis ............................................................................................................. 6 1.4. References .......................................................................................................................... 8 CHAPTER 2 State of the art .............................................................................................................. 13 2.1. Mammalian cell factories ................................................................................................... 13 2.2. CHO cells lines: Pioneering the production of recombinant proteins .................................... 13 2.3. CHO compared to other industrially relevant platforms ....................................................... 14 2.4. Cell culture media and its importance in bioprocessing ....................................................... 16 2.5. Highlighting CHO metabolism ............................................................................................ 17 2.5.1. Glycolysis .......................................................................................................................... 18 2.5.2. Glutaminolysis ................................................................................................................... 18 2.5.3. Amino acids and mammalian cell culture ........................................................................... 19 2.5.4. Metabolic flow path in CHO culture: By-products accumulation ........................................... 20 2.5.5. Accumulation of other potential toxic metabolites ............................................................... 22 2.6. Engineering CHO Bioprocess for a better performance ....................................................... 23 2.6.1. Reducing by-products accumulation ................................................................................... 23 2.6.2. Increasing productivity and improving process performance ............................................... 24 2.6.3. Maintaining the balance between growth and productivity ................................................... 25 2.6.4. Ensuring product quality .................................................................................................... 26 2.7. Systems biology for studying and improving CHO ............................................................... 27 2.7.1. Omics picture of CHO ........................................................................................................ 27 xiv Figure 5.8 Comparison of the exchange rates of key metabolites during culture of CHO-HyC cells in standard condition and medium B with different ammonia concentrations. The standard condition is highlighted in dark blue (Standard condition_qM). The negative and positive value indicate, respectively, the uptake and secretion rates of the corresponding metabolite. The values of the exchange rates of metabolites are expressed in mmol/gDW/h. Note that the rates lactate was scaled down and glutamine together with ammonia were scaled-up to fit the plot (indicated by the numbers after “/” for scaling down and “x” for scaling up). .......................................................................................................... 145 xv LIST OF TABLES Table 2.1 Comparison of the three major platforms for biopharmaceuticals production. ..................... 15 Table 2.2 Essential and non-essential amino acids for mammalian cells. Adapted from [64]. ............. 19 Table 2.3 Recent studies focusing on the use of GSMMs for CHO bioprocess optimization in the last 4 years. ............................................................................................................................................... 30 Table 3.1 Computational frameworks used for metabolic engineering. ............................................... 55 Table 3.2 Original model constraints vs in-house optimized constraints for non-producer cell lines. .... 82 Table 3.3 Comparison between the constraints used for predictions using producer cell lines. ........... 82 Table 4.1 Sequential adaptation of CHO-HyC cells to glutamine-free conditions. ................................. 94 Table 4.2 Exchange rates of metabolites for CHO-HyC cells treated or not with AKG (mm/gDW(h). .. 113 Table 4.3 Exchange rates of metabolites for CHO-K1 cells treated or not with AKG (mmol/gDW/h). . 114 Table 5.1 Amino acids used in this study. ....................................................................................... 132 Table 5.2 The levels of essential, non-essential amino acids and metabolites in the prepared cell culture media............................................................................................................................................. 133 Table 5.3 Osmolarity and pH values of the media used in the experiments. ..................................... 134 xvi LIST OF ABBREVIATIONS AND ACRONYMS 1,3BP6 1,3 bisphosphoglycerate 2PG 3-pPhosphoglyceric acid 3PG 2-pPhosphoglyceric acid AA Amino acids ACA Anti-clumping agent AcCoA AcetylCoA AKG α-ketoglutarate ALA Alanine ALAAT Alanine aminotransferase AlaGln l-alanyl-l-glutamine ALT Alanine transaminase AMP Adenosine monophosphate ASP Aspartate ATP Adenosine Triphosphate bp Base pairs BPCY Biomass product coupled yield CCD Cumulative cell density cdkis Cyclin-dependent kinase inhibitors CDM Chemically defined media CHO Chinese hamster ovary CIT Citrate CO2 Carbon dioxide CRISPR Clustered regularly interspaced short palindromic repeats DHAP Dihydroxyacetone phosphate DHFR Dihydrofolate reductase DoE Design of experiment DSP Downstream processing E. coli Escherichia coli EAAs Essential amino acids FBA Flux balance analysis xvii FDA Food and drug administration FMOC Fluorenylmethyloxycarbonyl FRAMED Framework for metabolic engineering and design FRU6P Fructose-6-pPhosphate FVA Flux variability analysis GCL3P Glucose-3-pPhosphate GDH Glutamate dehydrogenase GLC Glucose GLC6P Glucose-6-pPhosphate GLN Glutamine GLU Glutamate GOI Gene of interest GPR Genes proteins and reactions GS Glutamine synthase GSH Glutathione GSMM Genome scale metabolic model HCP Host cell proteins HER-2 Human epidermal growth factor receptor 2 IgG Immunoglobulin G IVCD Integral viable cell density LAC Lactate lncRNAs Long non-coding ribonucleic acid LP Linear programming LTM Logic transformation of model mAbs Monoclonal antibodies MAL Malate Mbp Mega base pairs MeOH Methanol MILP Mixed integer linear programming mM Millimolar MSX Methionine sulfoximine mTOR Mechanistic target of rapamycin xviii MTX Methotrexate NAD Adenine dinucleotide NEAAs Non-essential amino acids OAA Oxaloacetate ODEs Ordinary differential equations OPA O-phthalaldehyde P5P Pyridoxal-5’-phosphate PEP Phosphoenolpyruvate pFBA Parsimonious flux balance analysis PPP Pentose phosphate pathway PTMs Post translational modifications PYR Pyruvate R&D Research and development rcf Relative centrifugal force rpm Rotation per minute SDH Serine ammonia lyase siRNAs Small interfering ribonucleic acids SNPs Single nucleotide polymorphisms TCA Tricarboxylic acid tPA Tissue plasminogen activator USP Upstream processing VCD Viable cell density VCV Viable cell volume µm Micrometer xix SCIENTIFIC OUTPUT According to the 2nd paragraph of the article 8 of the Portuguese Decree-Law no. 388/70, the scientific outputs of this thesis are listed below. The results presented in this thesis have been partially published elsewhere. Peer reviewed journal articles: Hamdi, A.; Széliová, D.; Ruckerbauer, D.E.; Rocha, I.; Borth, N.; Zanghellini, J. Key Challenges in Designing CHO Chassis Platforms. Processes 2020, 8, 643. https://doi.org/10.3390/pr8060643 Oral presentation in conferences: Hamdi, A.; Santos, S.; Rocha, I. Towards metabolic optimization of CHO cells: In silico improvement of culture medium. Foundations of Systems Biology in Engineering conference. Valencia, Spain 2019. Posters in conferences: Hamdi, A.; Santos, S.; Rocha, I; Zanghellini, J. In silico -based approaches towards optimization of CHO cell culture medium. European Summit of Industrial Biotechnology, Graz, Austria, 2019. Hamdi, A.; Santos, S.; Baumann, M.; Borth, N.; Zanghellini, J, Rocha, I. Towards improvement of cho cells culture medium using in silico -based approaches. Cell culture engineering, Tucson, Arizona, U.S, 2020 (Postponed conference). Peer reviewed journal articles in preparation: Hamdi A., Santos S., Borth, N., Széliová, D., Zanghellini, J., Rocha I.; In silico -based approach for medium optimization of CHO cells (To be submitted in 2021). CHAPTER 1 1 1. CHAPTER 1 Motivation and outline of the thesis _____________________________________________________________ Biopharmaceuticals or biologics are large molecules derived from living organisms that, in the correct structure, can be very effective for preventing or treating a wide range of conditions such as infectious diseases and cancer. Examples of such biologics are vaccines, recombinant proteins and growth factors [1]. Along the years, there has been a continuous demand for developing complex biopharmaceuticals and the corresponding manufacturing processes, due to their therapeutic potential and global need in case of epidemics to control potential outbreaks that can be induced by continuously mutating pathogens (e.g., Influenza viruses or SARS-CoV) [2]. Indeed, the increased adoption of bio-based products holds a tremendous promise in improving current prevention and therapeutic procedures, especially in the field of vaccinology, gene therapy and cancer treatment [3]. Long time prior to the emergence of recombinant DNA technology, biotherapeutics were isolated from animals (e.g., Insulin being isolated from cows and pigs) or produced in animal tissues [4,5]. Following the revolutionary discovery of DNA recombination, biologics shifted to be produced in microbes, in vitro . The insertion of the insulin gene into a bacterial genome was an important stepping stone that drove the large scale manufacturing of human-insulin precursors in alternative producers already in the late 1970s, with the major production cell factories being Escherichia coli ( E.coli ) and Saccharomyces cerevisiae [6,7]. These efforts resulted in the approval of the first biopharmaceutical product by the regulatory bodies in the early 80s, Humulin® produced by DNA recombination technology using bacterial cells [8,9]. Due to its rapid growth to high cell densities in cheap and simple media formulations and the ease with which it can be genetically manipulated, E. coli remains a prime production host in the biopharmaceutical and biotechnological industries [10–13]. Bacterial systems are not the only hosts that have been used for insulin production. Other cell factories such as yeast systems have been explored as a result of advancement in genetic engineering [14,15]. For peptides/proteins that require the formation of disulfide bonds, including insulin as the best-known example, several yeast species have been explored and now have well established platform technologies available [16]. However, for the production of high value biotherapeutics that require human-like glycosylation or other types of complex post-translational modifications (PTMs), complex organisms, such as mammalian cells are required [17,18]. CHAPTER 1 2 As a result of these breakthrough technologies over the last 25 years, the pharmaceutical industry invested a great deal of resources into research and development (R&D) [19], aiming at generating groundbreaking complex biologics using heterologous expression in mammalian platforms, mainly the Chinese hamster ovary cells (CHO) due to its potential in producing high-quality biopharmaceuticals [20,21]. Since then, innovation in the biopharmaceutical field triggered the development of various novel compounds that demonstrated great therapeutic potential towards the treatment of both existing and emerging diseases [22]. The spectrum of produced biologics broadened along the years and the focus shifted towards even more sophisticated bio-based therapies, that are unable to be produced by microbial systems. Ergo, mammalian cell factories became one of the most important systems for the manufacturing of complex biopharmaceuticals such as monoclonal antibodies and recombinant therapeutic proteins. Since the introduction of human tissue plasminogen activator (tPA) to the market, the first bio-based therapy generated by mammalian cells, the biopharmaceutical industry continues to generate thumping profits overtime [23]. Currently, 316 biopharmaceuticals are on the market [24]. In fact, about 51% of the total produced biotherapeutic proteins are generated using mammalian cell lines, including 95% of the total produced therapeutic monoclonal antibodies (mAbs) and 83% of the total recombinant blood factors [25]. Emphasizing on monoclonal antibodies-based therapies, the number of commercialized treatments sextupled between 2012 and 2018 [26,27]. A total of 13 mAbs-based drugs were approved in 2018 along with 5 other potential therapies undergoing clinical trials. As an example, Humira® is the most selling drug in the United States, with a profit close to 18 billion dollars in 2017. The latter is a tumor necrosis factor (TNF)-inhibiting and anti-inflammatory drug used for the treatment of many conditions (e.g. plaque psoriasis, rheumatoid arthritis and Crohn’s disease) [28]. As a matter of fact, in 2018, the food and drug administration (FDA) approved 59 novel treatments, especially for cancer and infectious diseases. This is considered a 20 years record of approved biopharmaceuticals after 1996, where FDA authorized 53 novel treatments for various conditions [29,30]. As a consequence of this trend, the market size of the biopharmaceutical sector has greatly expanded overtime, with a total of US$ 228 billion in global sales in 2016 [31]. Simultaneously, the global bioprocess technology market is also expected to achieve 71 billion dollars by 2022 [32]. This indicates the market value of the total material needed for producing biopharmaceuticals (e.g., bioreactors, raw materials, chemicals, etc .). Breakthroughs in the biopharmaceutical R&D technologies allowed the large-scale production of various bio-based therapies and made the manufacturing of these products easier and more profitable. CHAPTER 1 3 Parameters such as process yield and productivity are fundamental for the economic value and profitability of bioprocesses. Due to the importance of biotherapeutics and the large competition between manufacturers, several attempts have been made to “modernize” the process optimization strategies throughout designing innovative engineering approaches, relying for instance on using different omics data sets to boost process titer, product quality and to decrease the production cost of biopharmaceuticals. In the post-genomic era, it became easier to study the genome of several species. Relying on modern sequencing techniques, depicting the genetic information of several organisms became faster and less expensive. Lately, sequencing became affordable and optimized to high-throughput [33] and is evolving to be the base for studying specific traits of industrially important cell lines. Due to the easiness of sequencing, it became more straightforward to think about combining genomics data with several other omics datasets for instance, transcriptomics, proteomics, fluxomics and especially metabolomics, referred to as high‐dimensional biology [34]. The analysis of these data represents the core of computational systems biology, enabling understanding and optimizing cellular machineries, by guiding, simultaneously, the modification of its genetic information and the rewiring of its metabolic flux distribution towards expressing phenotypes of interest. Along with this progress, there is still a need for improving production pipelines and room for further studying the cell at “omics” levels. In this era of big data, mathematical modeling has the potential to integrate multi-omics data sets into a single model and to correlate the observed changes in one dataset to observed changes in another [35]. Systems biology, which uses mathematical modeling to integrate current knowledge in a holistic manner, is a promising approach to optimize the time and resources required in biopharmaceutical production and to improve the industrial phenotypes of interest. While mathematical modeling in the context of systems biology is leading to many great advances in bioprocessing, it is still not fully translated to industry due to lack of expertise in the field [36]. 1.1. Context and motivation While optimizing heterologous protein expression of important biopharmaceuticals, it is necessary to focus on studying the cellular metabolism in order to determine the metabolic bottlenecks of the cell and to optimize its machinery. The latter is controlled by multiple genes and interconnected metabolic pathways. CHAPTER 1 4 Cell culture medium contains the most important components for CHO cell lines growth, as well as the main fuel for production of recombinant proteins. Several efforts have been performed to optimize the nutrient levels in these formulations and customize them to the cell’s need. This approach can help improving growth parameters of the cells as well as boosting the production of high titers of recombinant proteins, improving also its glycosylation patterns [37,38]. Glucose, amino acids and vitamins are the most important nutrients of cell culture medium. These metabolites are the main providers of carbon, nitrogen and other elements, crucial for proteins synthesis and essential for various biochemical reactions in the cell. Therefore, several media formulations have already been tested by various manufacturers and research groups to evaluate their production potential, especially by employing CHO cells [39]. These formulations have been improved over the years (e.g., the current use of protein-free chemically defined media) but are nowhere close to optimal. In the past, time-consuming, laborious and relatively inaccurate methodologies have been used for media optimization (e.g., strategies based on varying one factor at a time) [40]. However, modern approaches based on the use of high-throughput strategies together with deterministic and mechanistic modeling approaches, for instance relying on the use of genome scale reconstructs and various omics data, improved the robustness of medium design and provided a solid base to overcome the use of the classical optimization methodologies [41,42]. Refining cell culture media formulation to the metabolic requirements of the cells can solve several bioprocessing problems, for instance, decreasing byproducts levels during production. These improvements can be made by assessing metabolic requirements through studying, for instance, nutrient uptake rates and transporters capacity. Tools such as constraint-based modeling are very important to study the metabolic behavior of the cells in answer to variation in environmental conditions, in silico . In general, using in silico -based strategies and genome-scale metabolic models for studying CHO metabolism is a powerful approach to optimize growth parameters of CHO. These models not only provide a holistic overview about CHO metabolic network, but also can define robust strategies for predicting the cellular phenotypic changes in response to environmental adjustments [43], which in long term will reduce R&D time and costs. Subsequently, developing a refined formulation of cell culture media based on CHO cells specific needs, based on studying the effects of different nutrients such as amino acids on the metabolism of CHO cells is a groundbreaking ambition that will subdue many biomanufacturing problems such as low cell growth, low cell densities, low product yields and the accumulation of cell culture by-products. CHAPTER 1 11 https://doi.org/10.1016/j.mec.2020.e00149. [36] A. Richelle, B. David, D. Demaegd, M. Dewerchin, R. Kinet, A. Morreale, R. Portela, Q. Zune, M. von Stosch, Towards a widespread adoption of metabolic modeling tools in biopharmaceutical industry: a process systems biology engineering perspective, Npj Syst. Biol. Appl. (2020). https://doi.org/10.1038/s41540-020-0127-y. [37] M. Gawlitzek, M. Estacio, T. Fürch, R. Kiss, Identification of cell culture conditions to control Nglycosylation site-occupancy of recombinant glycoproteins expressed in CHO cells, Biotechnol. Bioeng. (2009). https://doi.org/10.1002/bit.22348. [38] D.Y. Kim, J.C. Lee, H.N. Chang, D.J. Oh, Effects of supplementation of various medium components on Chinese hamster ovary cell cultures producing recombinant antibody, in: Cytotechnology, 2005. https://doi.org/10.1007/s10616-005-3775-2. [39] D. Reinhart, L. Damjanovic, C. Kaisermayer, R. Kunert, Benchmarking of commercially available CHO cell culture media for antibody production, Appl. Microbiol. Biotechnol. 99 (2015) 4645– 4657. https://doi.org/10.1007/s00253-015-6514-4. [40] Y. Rouiller, A. Périlleux, N. Collet, M. Jordan, M. Stettler, H. Broly, A high-throughput media design approach for high performance mammalian fed-batch cultures, MAbs. (2013). https://doi.org/10.4161/mabs.23942. [41] S.N. Galleguillos, D. Ruckerbauer, M.P. Gerstl, N. Borth, M. Hanscho, J. Zanghellini, What can mathematical modelling say about CHO metabolism and protein glycosylation ?, Comput. Struct. Biotechnol. J. 15 (2017) 212–221. https://doi.org/10.1016/j.csbj.2017.01.005. [42] V. Singh, S. Haque, R. Niwas, A. Srivastava, M. Pasupuleti, C.K.M. Tripathi, Strategies for fermentation medium optimization: An in-depth review, Front. Microbiol. (2017). https://doi.org/10.3389/fmicb.2016.02087. [43] D. Széliová, D.E. Ruckerbauer, S.N. Galleguillos, L.B. Petersen, K. Natter, M. Hanscho, C. Troyer, T. Causon, H. Schoeny, H.B. Christensen, D.Y. Lee, N.E. Lewis, G. Koellensperger, S. Hann, L.K. Nielsen, N. Borth, J. Zanghellini, What CHO is made of: Variations in the biomass composition of Chinese hamster ovary cell lines, Metab. Eng. (2020). https://doi.org/10.1016/j.ymben.2020.06.002. [44] H. Hefzi, K.S. Ang, M. Hanscho, A. Bordbar, D. Ruckerbauer, M. Lakshmanan, C.A. Orellana, D. Baycin-Hizal, Y. Huang, D. Ley, V.S. Martinez, S. Kyriakopoulos, N.E. Jiménez, D.C. Zielinski, L.E. Quek, T. Wulff, J. Arnsdorf, S. Li, J.S. Lee, G. Paglia, N. Loira, P.N. Spahn, L.E. Pedersen, J.M. Gutierrez, Z.A. King, A.M. Lund, H. Nagarajan, A. Thomas, A.M. Abdel-Haleem, J. CHAPTER 1 12 Zanghellini, H.F. Kildegaard, B.G. Voldborg, Z.P. Gerdtzen, M.J. Betenbaugh, B.O. Palsson, M.R. Andersen, L.K. Nielsen, N. Borth, D.Y. Lee, N.E. Lewis, A Consensus Genome-scale Reconstruction of Chinese Hamster Ovary Cell Metabolism, Cell Syst. 3 (2016) 434-443.e8. https://doi.org/10.1016/j.cels.2016.10.020. CHAPTER 2 13 2. CHAPTER 2 State of the art Partially adapted from: A. Hamdi, D. Széliová, D.E. Ruckerbauer, I. Rocha, N. Borth, J. Zanghellini, Key challenges in designing CHO chassis platforms, Processes. (2020). https://doi.org/10.3390/PR8060643. ------------------------------------------------------------------------------------------------- 2.1. Mammalian cell factories Biopharmaceuticals are mainly produced using heterologous expression in recombinant cells or microorganisms [1]. Mammalian cell factories are successful platforms for the production of recombinant proteins, especially monoclonal antibodies (mAbs), where Chinese hamster ovary (CHO) are predominant hosts [2]. This success is not only linked to their production capacity, but also to their history of safety, regarding the low susceptibility to viral infections [3]. These cells hold various unique features, namely, their adaptation capacity to high density suspension cultures, their easy scale-up and also effortless compliance to serum-free medium conditions, the most preferred formulations for biomanufacturing nowadays [4]. Alongside, one of the main features of mammalian platforms is linked to their ability to perform complex human-like Post-Translational Modifications (PTMs) [5]. Among these, glycosylation represents one of the most important attributes [6,7] and the most common structurally diversified modification in secreted proteins [8]. Correct glycosylation is required to sustain optimal pharmacokinetic and pharmacodynamic properties of biopharmaceuticals, since it affects the efficacy and in vivo turnover rate of therapeutics, and prevents immune responses triggered by non-human glycans [9,10]. 2.2. CHO cells lines: Pioneering the production of recombinant proteins Within mammalian platforms, CHO cell factories dominate the production of recombinant proteins in today’s biopharmaceutical industry. CHO cells, as indicated by their name, were derived from the ovary of the Chinese hamster. They are in fact mainly of epithelial phenotype [11]. Originally, CHO cells were established in the late 1950s by Theodore T. Puck [12]. Since then, the family of CHO cells has expanded, giving rise to various new lineages such as CHO-K1, CHO-S, CHO-GS-, CHO-DG44, etc . CHAPTER 2 14 These lineages were developed to fulfill specific industrial requirements, such as suspension culture or specific gene deficiencies that enabled selection, and are the result of genetic modifications via chemical and radiation mutagenesis [13], targeted gene knockouts and adaptation to new culture conditions [14,15]. Thereupon, due to their comparatively simple handling, CHO cell lines have proven to be crucial for the industrial manufacturing of recombinant proteins [16]. However, despite several decades of research and process design, the productivity remains low compared to the theoretical maximum productivity predicted in silico by a genome-scale metabolic model of CHO [17]. Hence, along with the increasing demand for biopharmaceutical products, there is a growing need to optimize CHO’s production yield and to fast track the development of newly optimized production cell lines, in order to satisfy the large demands for complex biotherapeutics nowadays. In view of the many different aspects of mammalian platforms, several engineering approaches have been developed to address, typically individually, the many challenges encountered during cell line development and manufacturing of highly complex biotherapeutics. Medium optimization and highthroughput screening for good producers were previously described [18,19]. Alternative optimization strategies based on modular design, synthetic biology and systems metabolic engineering, hold also tremendous promise to study the metabolic network of the cells and further improve productivity, yield, product quality and to reduce the time and cost of cell line development. Nonetheless, applying such rational engineering tools to mammalian cells is more difficult compared to other platforms, due to the complexity of the system. Many important aspects need to be considered, namely, the large genome, sophisticated regulatory, signaling, and metabolic networks, genome instability and epigenetic regulation [20]. 2.3. CHO compared to other industrially relevant platforms From a genetic perspective, mammalian cells are considered more complex than microbial systems as their genome is by far larger than that of E. coli and S. cerevisiae . The assembly of the sequenced CHOK1 genome comprises 2.45 Gb with 24,383 predicted genes [3], while microbial cells used in the biotechnological industry have a smaller genome size by one to two orders of magnitude. A comparison of E. coli , S. cerevisiae and CHO platforms is summarized in Table 1. Even though having a larger genome does not necessarily relate to the cell’s morphological complexity, it can be an indication of the intricacy of its proteome, fluxome, transcriptome and metabolome and, in particular, of its regulatory CHAPTER 2 15 capacities. Even from the viewpoint of the proteins that are encoded in these genomes, the proteins constituting prokaryotic cells are considered less complex than those of eukaryotic cells. The latter idea was claimed by different researchers such as Zhang et al and Wang et al [21,22] stating that the organism’s protein structural complexity (e.g., length) can directly affect the growth performance of cells. It was demonstrated that, when optimizing for growth, a higher growth rate was observed for cell types containing smaller proteins. This is due to the cell tendency to increase its mass-normalized kinetic efficiencies during growth [23]. Table 2.1 Comparison of the three major platforms for biopharmaceuticals production. Characteristic E. coli S. cerevisiae CHO References Genome size (Mbp) 4.6 12.1 2450 [3], [24], [25] Cell size (µm) <1 3-5 12-24 [26–28] Cell volume (µm3) 0.3-3 30-100 900-7200 [27–29] Doubling time (h) Fast (0.5-4) Fast (1.5-6) Slow (18-48) [30–33] N-Linked Glycosylation No High mannose Complex [34] Gene length (bp) ~1000 ~1000 ~1300/18000 † [29,35] Promoter length (bp) ~100 ~1000 ~104-105 * [29,36] Number of protein coding genes ~4300 ~5300-5400 ~24000 [35,37,38] Proteins per cell ~106 ~108 ~1010 [29,33] Cell culture medium complexity Low Low High [34] Cost of cell culture medium Low Low High [34] †Coding/transcript; * HeLa cell line In addition, cultivating mammalian cells is considered more demanding compared to microbial organisms, especially when focusing on their bioprocessing requirements. Due to the lack of a cell wall, there is significantly higher shear sensitivity and the cell culture medium must contain a higher number of essential nutrients compared to microbial systems. Bacteria (e.g., E. coli ) and yeast (e.g., S. cerevisiae ) can grow in a simple medium containing solely basic elements (e.g., glucose and salts) and usually only in specific cases a few amino acids (AA) or vitamins are added. In contrast, mammalian cells require a larger and more complex set of nutrients, including amino acids, organic acids, vitamins, CHAPTER 2 16 cofactors, carbohydrates and salts. This complexity of the growth medium reveals the strict nutritional demand of mammalian cells. A major difference between microbial and mammalian cells is the fact that the genome of the latter actually encodes many different types of cells and developmental stages, namely more than 100 different types of tissues that are part of a mammalian body. To ensure correct expression of the required genes at the necessary level in each of these different tissue types, a much more complex regulatory network is required that includes highly sophisticated mechanisms such as epigenetics and chromatin remodeling that simply are not necessary for microbial cells and therefore are not present or are only at immature levels of development [39]. Apart from these chromatin state and epigenetic mechanisms, other regulatory factors are abundant in mammalian cells, such as microRNAs or longnon-coding RNAs (lncRNAs), which are transcribed in large numbers [40,41]. Over the last years, scientists are moving forward to employ optimization strategies that have been successfully used to study simpler organisms to other less explored systems. For that, synthetic biology, systems biology and metabolic engineering have been employed towards this goal. The use of these tools is facilitated by the availability of the genetic information [42] of different organisms. While many of these tools are already widely applied in the field of recombinant protein production or strain engineering in microbial research, up to the level of design of chassis strains, its application to mammalian production hosts is still fragmentary and lagging far behind. 2.4. Cell culture media and its importance in bioprocessing Cell culture media is a complex mixture of nutrients, energy sources and trace elements, essential for the growth and maintenance of the cells ex-vivo (figure 1). This concept was first described in the groundbreaking work of Eagle Dulbecco and Freeman, in the early fifties, stating that amino acids combined with other nutrients such as vitamins, are the core base of cell line cultivation, especially in adherent mode [43]. Nowadays due to regulatory guidelines, using animal derived serum poses various problems from bioprocessing standpoint (lot-to-lot variation), safety (contamination with viruses, mycoplasma, prions, etc. ) and also from ethical point of view [44–46]. To overcome the use of serum in bioproduction, hydrolysates were employed in various processes to improve growth and productivity of the cells by developing cell culture media containing animal free components. Both plant and yeast hydrolysates CHAPTER 2 17 were tested with mammalian cells, especially using CHO cell lines [47,48]. However, hydrolysates also contain undefined concentrations of components, which can impair the consistency of both upstream processing (USP) and also downstream processing (DSP) [49]. Figure 2.1 Mammalian cell culture media components. Currently, industry is trying to veer the attention towards the use of chemically defined media (CDM) deprived from animal components to design solid processes with consistent product titers among batches, together with producing potent and clinically safe biopharmaceuticals. In culture, additives such as Pluronic F68 is very useful for CDM, since it can reduce the shear generated by the hydrodynamic motion during mixing [50]. 2.5. Highlighting CHO metabolism For growth, CHO cells require different nutrients (Nitrogen and carbon sources), energy carrying molecules (Adenosine Triphosphate ((ATP)), and other cofactors (e.g., adenine dinucleotide NAD+). These molecules are fundamental for sustaining the basic metabolic functions of mammalian cells [51]. Among the most important nutrients, glucose and glutamine represent a prime energy sources to the cell [52]. They are usually provided in vitro throughout the cell culture medium delivering carbon and nitrogen atoms to the cells in order to support its basic mechanisms such as proliferation [53]. In cell culture media, a balance between glucose and glutamine levels is essential, not only for maintaining optimal cell growth but also for sustaining the glycosylation of the produced recombinant proteins [54]. Tight regulation between glucose and glutamine metabolism was previously discussed by •Amino acids •Lipids •Trace elements •Vitamins •Glucose •Glutamine •TCA cycle metabolites (e.g., pyruvate, alphaketoglutarate, etc.) •Bicarbonate •Pluronic F68 •Salts pH regulation and sheer stress Energy Proteins production Process efficiency and product quality CHAPTER 2 18 Zeng and Deker [55]. That being said, glycolysis, as part of the central carbon metabolism and glutaminolysis are the major metabolic pathways for mammalian cells [56]. 2.5.1. Glycolysis The central carbon metabolism in mammalian cells is a complex set of biochemical reactions, transforming glucose into different metabolites, generating cell’s biomass and various metabolic precursors, essential for various metabolic reactions [57]. This system is composed of three main pathways, including glycolysis, pentose phosphate pathway (PPP) and tricarboxylic acid cycle (TCA cycle) [52]. Glucose is the major player in the central metabolism since it is the main carbon and energy source in mammalian cells. Glycolysis is one of the most important pathways in the cell, where glucose is phosphorylated to glucose-6-phosphate and finally oxidized to 2 molecules of pyruvate, that are channeled into the mitochondria in order to enter the TCA cycle [58] (Figure 2). 2.5.2. Glutaminolysis Glutamine represents a versatile donor and the main provider of nitrogen to the cells, together with several other amino acids. It is also the main substrate of the glutaminolytic pathway in mammalian cells (Figure 2). Glutamine degradation fuels the TCA cycle, generating cellular energy and boosting the biosynthetic pathways [59,60]. Figure 2.2 Representation of the glycolysis and glutaminolysis in CHO metabolism. CHAPTER 2 19 Legend: GLC (Glucose); GLC6P (Glucose-6-Phosphate); FRU6P (Fructose-6-Phosphate); DHAP (Dihydroxyacetone phosphate); GLC3P (Glucose-3-Phosphate); 1,3BP6 (1,3 bisphosphoglycerate); 3PG (3-Phosphoglyceric acid); 2PG (2Phosphoglyceric acid); PEP (Phosphoenolpyruvate); PYR (Pyruvate); LAC (Lactate); AcCoA (AcetylCoA); CIT (Citrate); AKG ( α-ketoglutarate); MAL (Malate); OAA (Oxaloacetate); ASP (Aspartate); ALA (Alanine); GLN (Glutamine); GLU (Glutamate). On the one hand, glutamine degrades to glutamate via phosphate - dependent glutaminase and then to α-ketoglutarate via enzymatic reactions involving glutamate dehydrogenase (GDH). Following this metabolic path, the carbon backbone of glutamine is oxidized to CO2 and malate that will be converted to pyruvate molecules [61,62]. On the other hand, the conversion of glutamate to α-ketoglutarate via aspartate/alanine transaminase is usually activated to overcome the overproduction of ammonium by the cells involving either pyruvate or oxaloacetate molecules [63]. The latter mechanism is important to overcome the overproduction of the toxic ammonium during glutaminolysis. 2.5.3. Amino acids and mammalian cell culture Amino acids are vital for mammalian cells cultivated in vitro . They are divided into two categories: essential and non-essential. Essential amino acids (EAAs) cannot be synthesized de novo by the cells and have to be provided in the cell culture media. On the other hand, non-essential amino acids (NEAAs) can be synthesized by the cells in order to sustain and support growth. The latter are produced relying on several metabolites such as TCA cycle intermediates or others generated through the glycolytic pathway. A list of different essential and non-essential amino acids for mammalian cell culture are described in Table 2 and the work published by Salazar et al., 2016 [64]. Table 2.2 Essential and non-essential amino acids for mammalian cells. Adapted from [64]. Essential amino acids Non-essential amino acids Arginine, histidine, isoleucine, leucine, lysine, methionine, phenylalanine, threonine Tryptophan and valine Alanine, asparagine, aspartate, cysteine, glutamate, glutamine, hydroxyproline, proline, serine and tyrosine The chemical and biological attributes of these metabolites are crucial for cell growth and they constitute the building blocks for naïve/recombinant proteins synthesis. They are precursors for important metabolic pathways and source of nitrogen and carbon atoms [65]. The catabolism of several CHAPTER 2 20 amino acids is performed either via transamination (e.g., glutamate) or direct deamination (e.g., serine and threonine) [66]. Several studies aimed at adjusting the levels of several amino acids in the media due to their importance for obtaining high growth and yield of recombinant proteins simultaneously [67]. The levels are adjusted according to the metabolic requirements of cells [64]. A lack of specific amino acids can starve the cells; however, an excess of other amino acids, for instance lysine can hamper cell growth [68]. In fact, the consumption of amino acids during culture is directly dependent on many factors such as the culture environment or the cell cycle [69,70]. Furthermore, amino acids transporters play a role in sensing the different levels of amino acids outside of the cell. As a consequence, any change in the levels of amino acids in the extracellular environment is able to influence the activity of AA transporters within the cell [71]. Finally, it is important to highlight that the combination of the essential and non-essential amino acids in the cell culture medium is important for triggering various reactions and complexes in the cell, as for example the mechanistic target of rapamycin (mTOR), being mTORC1 one of the most studied pathways [72,73]. The latter is strongly dependent on amino acids presence, especially leucine [74]. It represents the penstock of cellular growth and metabolism [72]. Activating this sophisticated signaling network is the basis for cell proliferation, lipid synthesis, protein synthesis and also mitochondrial metabolism and biogenesis [75]. 2.5.4. Metabolic flow path in CHO culture: By-products accumulation The metabolism of mammalian cells, especially CHO is very complex and far away from being optimized for bioprocessing [20,58,66]. In order to achieve healthy proliferation of the cells and generate high yields of biotherapeutics with correct PTMs, it is important to control the cell culture environmental conditions to limit the secretion of toxic by-products that can hamper cell growth and alter product quality attributes [76]. In the cell, metabolic pathways are interconnected, for instance, glucose and amino acids metabolism. Any variation regarding the consumption rate of one of these metabolites during culture, can influence the metabolic homeostasis and can lead to accumulation of unwanted products. As a matter of fact, the central carbon metabolism changes depending on the culture phase, cells metabolic need and the availability of nutrients in the cell culture media. In experimental conditions, CHAPTER 2 27 Steps towards developing custom and consistent glycosylation profiles have already been taken. For example, a panel of cell lines expressing custom glycosylation patterns was created with the use of CRISPR/Cas9 technology [149]. In another study, the level of galactosylation was manipulated based on predictions from a kinetic model, leading to a reduction in glycan heterogeneity [150]. While for monoclonal antibodies, with their relatively simple glycosylation pattern, work on detailed control has already been initiated, the field is still open for more complex proteins bearing multiple glycosylation structures with high prevalence of tetra-antennary structure and the need for full terminal sialylation [151]. 2.7. Systems biology for studying and improving CHO The need for developing new methodologies for process optimization drove scientists to combine novel omics technologies with bioinformatics to understand biological systems [152]. Due to the high complexity of cells, mainly mammalian cells, it is mandatory to focus on developing robust models that, combined with omics data, can mathematically describe the metabolism of the cell, allowing the prediction of the effect of different culture conditions, as well as the determination of optimization strategies, which are particularly relevant for industrially valuable platforms such as microbial and mammalian cell factories. Systems biology and metabolic engineering represent some of the most promising tools for process improvement. These technologies target the amelioration of cellular phenotypes via the manipulation of their biochemical pathways, for instance, through metabolic flux optimization. Optimization goals include increasing cell specific productivity by maintaining the balance between the competing interests of growth and productivity, generating an efficient and targeted metabolism to enhance product quality, and decreasing the level of process by-products [153]. Subsequently, following the success of using metabolic modeling strategies in microbial cells, it is promising to apply these tools to metabolically optimize mammalian cells [154]. Designing rational engineering approaches to enhance CHO-based bioprocesses via modulating cell metabolism is a promising strategy [17]. 2.7.1. Omics picture of CHO The CHO-K1 genome sequence, published in 2011 [3], was an important stepping stone towards the application of systems biology methods to CHO. However, in contrast to microbial systems, the genome of CHO cells often contains various chromosomal abnormalities caused by its genetic instability CHAPTER 2 28 [155,156]. Thus, cells belonging to the same lineage within a CHO family can have distinct genetic information and phenotypes. Differences in phenotype can even be observed among cells that normally belong to the same lineage but were grown in different laboratories and under various culture conditions. This can in part be explained by the structural variations in the genome and the accumulated genetic changes such as single nucleotide polymorphisms (SNPs), transgene copy number variations and chromosomal rearrangements [157–160]. Vcelar et al showed that chromosomal rearrangements within the genome of a population are observed during subcloning, adaptation of the cells to a new medium or simply during long-term cultivation [161,162]. On top of this, variations in phenotypes that cannot be explained by genomic diversity and variation alone, are frequently observed, even in subclones of subclones [163]. These facts suggested that the genetic information of the CHO-K1 cells sequenced in 2011 [3] are not representative of all CHO cell lines and subclones, so a reliable and stable reference genome was needed. This was addressed by the generation of a common reference genome of the Chinese hamster Cricetulus griseus [5,164] , which was further improved by a more complete genome assembly in 2018 [35]. These, along with the genomes of other cell lines sequenced in the meantime [5,159,160] serve as basic datasets for in silico studies of CHO via integrative analyses of omics data relying, for example, on the use of genome scale metabolic models (GSMMs) which can support genetic and metabolic engineering studies. 2.7.2. Genome scale metabolic models Genome scale metabolic models (GSMMs) are novel tools for systems biology that carry information regarding different genes, proteins, and reactions (GPR) being an integral representation of the cellular metabolism. The availability of full genomic sequences, as well as omics data of various organisms enabled researchers to reconstruct cellular metabolism in silico , linking different level of information to calculate metabolic fluxes [165]. Indeed, GSMMs are powerful tools in systems biology and can help predicting the changes in metabolic behavior of the studied organisms, not only in response to changes in environmental constraints (e.g., medium formulation) but also to genetic manipulation of the cells (e.g., gene knockouts) [166–168]. These strategies allow to mechanistically link organism’s genetic information and phenotype [92, 93] and can be very useful for rational identification of engineering targets [167,171]. GSMMs are emerging as a common practice in metabolic engineering. Therefore, several reports are available nowadays describing, in a detailed manner, the protocol for reconstructing GSMMs CHAPTER 2 29 [165,172,173]. This process relies heavily on the information regarding genes, metabolites and enzymes activity of the specific organism being investigated. In fact, following the development of the genome scale model for H. influenzae [174], various genome scale reconstructions were made available to the scientific community [175]. Due to the importance of CHO cells in the biopharma industry, scientists joined efforts to develop a CHO genome scale metabolic model [17]. Therefore, researchers are veering the attention towards implementing computational models for cell culture improvement, but still few successful studies were focused on employing CHO GSMM to improve growth and productivity of the cells [176,177]. The universal metabolic model iCHO1766 was built jointly with various research groups and comprises the most complete representation of CHO so far. The metabolic network was reconstructed and associated with over 1700 genes, 2300 metabolites and over 6000 reactions in the Cricetulus griseus genome. The model was built based on the information described in the global human metabolic network (Recon 1) [178], knowledge from the updated version Recon 2 [179] and a curated version of the Recon 2 model [180]. Combining these reconstructions, Cricetulus griseus homologs were determined. Further details are described in Hefzi et al, 2016 [17]. Accordingly, CHO model represents a powerful tool for untangling the complexity of CHO, investigating bioprocess capabilities and improving cell line development strategies [17], which can be further improved by integrating additional biochemical information to understand cellular processes beyond metabolism [181]. Recently, several efforts were performed to curate GSMMs and to improve its prediction accuracy via modifying it [177]. These efforts generated, for instance, the updated models iCHO2101 [182] and iCHO2291 [183]. Clearly, refining our knowledge about the metabolism of CHO cells allows scientists to continue optimizing the available genome scale metabolic models by complementing genetic and metabolic information of CHO [177]. Prediction tools based on metabolic modeling have only recently started to be applied to the design of engineering strategies in mammalian cells, mainly CHO cells (Table 2). One of the first examples where the GSMM of CHO was applied to an industrial process, is the work performed by Calmels et al, where the genome-scale metabolic model [17] was curated and tailored to a CHO-DG44 producer cell line. They performed corrections, such as modifying 601 reactions (for example silencing of 537 amino acids transporters), which led to an improvement of the growth rate and exometabolome predictions in silico [177]. CHAPTER 2 30 In addition, the secretory pathway was integrated into the GSMM of CHO by Gutierrez et al, to enable predictions of energetic and machinery demands of secreted proteins [184], which might lead to better predictions of engineering targets that aim at improving protein production, as shown for example in the work of Kol et al [115]. On the other hand, many studies based on mapping the intracellular fluxes throughout metabolomics studies can improve cell culture parameters [185]. New perspectives aiming at overcoming the redundancy of the large scale metabolic models and decreasing the computational time are moving towards the reconstruction of minimal metabolic networks that contain the most essential genetic and metabolic information needed for predictions [186]. Table 2.3 Recent studies focusing on the use of GSMMs for CHO bioprocess optimization in the last 4 years. Publication Aim of the study Reference Calmels et al., 2018 Curation of genome scale metabolic model to construct CHO-DG44 specific model. [177] Szeliova et al., 2020 Determination of CHO biomass variations among different strains [33] Szeliova et al., 2020 (2) Experimental measurement errors and its impact on in silico -based predictions [154] Huang et al., 2020 Improvement of cell’s productivity relying on the use of CHO GSMM. [187] Gutierrez et al., 2020 Generation of CHO species-specific secretory pathway reconstructions. [184] Schinn et al., 2021 Prediction of amino acid concentrations in cultures and predict nutrient feeding strategy. [188] Pérez‐Fernández et al., 2021 Media optimization for continuous CHO-K1 cultures. [189] Szeliova et al., 2021 Inclusion of maintenance energy improves the intracellular flux predictions of CHO [190] As a matter of fact, due to the size and complexity of the metabolic models, mainly mammalian cells models, computational support is necessary in order to predict optimal intervention strategies from the combinatorial universe of possible modifications. Computational strain design methods, many of which are based on constraint-based analysis of cellular metabolism [168,191–193] are available towards this end [194] and are continuously being refined. Combined with bioinformatics, GSMMs are very effective CHAPTER 2 31 in studying metabolic responses of different stimuli and in determining the metabolic bottlenecks of the cells. In fact, mathematical modeling strategies emerged as a very promising advancement to analyze different biochemical complex networks reflecting a complete picture that draws the different reactions in the studied organism and the specific genes that encode for them [192]. The mathematical representation of these metabolic reactions considering steady state forms a system of linear equations [192]. Tools such as flux balance analysis (FBA), play an important role in solving these equations and help in quantifying the level of contribution of different reactions to a specific target phenotype in response to environmental and genetic changes [195]. 2.7.3. Flux Balance Analysis Flux Balance Analysis (FBA) is a widely used mathematical tool to study metabolic networks and to predict phenotypes [196] which can be easily applied to genome-scale metabolic networks. FBA helps understanding the metabolic distribution inside of a complex network, relying on the use of linear optimization, assuming steady state conditions. FBA targets the maximization of a specific objective function, usually defined as biomass formation [197]. One of the advantages of FBA is to calculate different metabolic flux distributions when varying environmental constraints (e.g., medium formulations) or when performing gene knockouts. Prediction assuming steady state means to consider that the sum of the rates of formation of all internal metabolites is equal to the sum of their production rates. In general, the modeling approach consists of deducing a stochiometric matrix (S). Within this matrix, the rows and columns represent respectively, the metabolites and the reactions. Assuming the steady state conditions is explained by (S.v = 0), where (v) represents the flux vector indicating the specific rates for each reaction, and where every mass balance in the system is represented by a linear equation. This computational approach is considered very useful in systems biology and its prediction accuracy can increase when feeding these models with experimental omics data, for example by means of specifying the flux boundaries. The upper and lower limits of fluxes can be used for every reaction inside of the model where (vlower ≤ v ≤ vupper). From a bioprocessing standpoint, this mathematical tool can be employed for studying metabolic pathways in the cell targeting either maximizing growth or maximizing product formation in industrial cell lines [192]. Inserting experimental data and calculating fluxes taking into consideration the thermodynamic capability of some pathways can improve the prediction accuracy. In addition, other factors such as experimental data quality can imperatively influence the prediction results since the analytical error can CHAPTER 2 32 propagate through FBA. Therefore, it is important to establish solid experimental protocols to overcome possible prediction inaccuracies [154]. Another mathematical tool is parsimonious flux balance analysis (pFBA). The latter is a variant of FBA, and assumes the usage of the minimal amount of metabolic fluxes within a metabolic network in order to sustain an objective (e.g., maximal growth), always taking into consideration steady state assumption [177,186,198]. The latter is considered as an improved version of standard FBA, since in 2 prediction steps, this optimization method predicts the most adequate pathways reflecting, in a different mathematical representation, the idea of “maximum biomass per number of fluxes” described by Schuetz et al [198,199]. In a nutshell, this tool minimizes the total sum of fluxes in the network by removing futile loops [198]. 2.7.4. Defining the objective function: The biomass function The biomass function is a very important parameter for in silico -based predictions using genome scale metabolic models. When maximizing for growth, the biomass function is usually described as the objective function [197]. In this case, it describes the rate at which the different metabolic components are converted to biomass elements. The biomass function can be formulated at different levels of detail. Typically, it contains information regarding different components in the cell such as proteins, RNA, DNA and lipids. It can be further detailed by adding different layers of information (e.g., cofactors, vitamins and elements) [200]. The biomass equation is described as following, where 𝐶𝑖 represents the coefficient of each biomass component 𝑋𝑖 : ∑𝐶𝑖𝑋𝑖 𝑛 𝑖=1 →𝐵𝑖𝑜𝑚𝑎𝑠𝑠 As previously described, different strains/cell lines hold different genetic and metabolic features. It is ideal to build a specific biomass function for each cell type. In silico metabolic phenotypes can vary also according to the biomass objective function used for each cell line [201,202]. Subsequently, it is often important to build strain and condition specific biomass functions, given that they can improve the model capabilities to predict metabolic fluxes [33,203]. CHAPTER 2 33 As an example, it was previously discussed that yeast biomass composition can vary depending on the physiological conditions were it is grown [204]. A similar study concluded the same results for CHO cells [33]. Besides, it was also described that an accurate estimation of cell biomass composition is needed for robust predictions using GSMMs, not only when using microbial models but also for mammalian metabolic models (CHO) [33]. In iCHO1766 [17], two different biomass functions were included in the model. The first biomass equation (R_biomass_cho) is employed in order to predict growth of naïve or (non-recombinant) CHO cells, whereas the second biomass function (R_biomass_cho_producing) is used to predict the growth for recombinant CHO cells targeting the production of monoclonal antibodies. These two different biomass equations were built due to the differences between calculated gross cell composition in nonrecombinant CHO cell lines and measured values for IgG-producing hybridoma lines (e.g., the amount of protein fraction constitute more than 70 % of cell dry weight in a producing cell, while it was calculated to be 55% in a non-producing cell). Using a specific biomass function for each cell type can significantly improve prediction accuracy. It also facilitates understanding of the metabolic features of these cell types in different environmental conditions, for instance, media formulations. In a nutshell, strategies based on in silico predictions using cell line specific genome scale models can play a role in decreasing the experimental workload as well as in avoiding needless laboratory costs by defining the most important parameters in silico without recurring to test all conditions experimentally [205]. Henceforth, in silico approaches hold a tremendous promise in improving bioprocesses relying on the study of cell culture attributes (e.g., cell growth and exchange rates of metabolites) towards improving both final product titer and quality attributes throughout predicting its post-translational modification patterns [206]. 2.8. References [1] T. Dingermann, Recombinant therapeutic proteins: Production platforms and challenges, Biotechnol. J. (2008). https://doi.org/10.1002/biot.200700214. [2] M. Butler, M. Spearman, The choice of mammalian cell host and possibilities for glycosylation engineering, Curr. Opin. Biotechnol. (2014). https://doi.org/10.1016/j.copbio.2014.06.010. [3] X. Xu, H. Nagarajan, N.E. Lewis, S. Pan, Z. Cai, X. Liu, W. Chen, M. Xie, W. Wang, S. Hammond, M.R. Andersen, N. Neff, B. Passarelli, W. Koh, H.C. Fan, J. Wang, Y. Gui, K.H. Lee, M.J. Betenbaugh, S.R. Quake, I. Famili, B.O. Palsson, J. Wang, The genomic sequence of the CHAPTER 2 34 Chinese hamster ovary (CHO)-K1 cell line, Nat. Biotechnol. (2011). https://doi.org/10.1038/nbt.1932. [4] A.D. Bandaranayake, S.C. Almo, Recent advances in mammalian protein production., FEBS Lett. (2014). https://doi.org/10.1016/j.febslet.2013.11.035. [5] N.E. Lewis, X. Liu, Y. Li, H. Nagarajan, G. Yerganian, E. O’Brien, A. Bordbar, A.M. Roth, J. Rosenbloom, C. Bian, M. Xie, W. Chen, N. Li, D. Baycin-Hizal, H. Latif, J. Forster, M.J. Betenbaugh, I. Famili, X. Xu, J. Wang, B.O. Palsson, Genomic landscapes of Chinese hamster ovary cell lines as revealed by the Cricetulus griseus draft genome, Nat. Biotechnol. (2013). https://doi.org/10.1038/nbt.2624. [6] J. Batra, A.S. Rathore, Glycosylation of monoclonal antibody products: Current status and future prospects, Biotechnol. Prog. (2016). https://doi.org/10.1002/btpr.2366. [7] R. Jefferis, Glycosylation of recombinant antibody therapeutics, Biotechnol. Prog. (2005). https://doi.org/10.1021/bp040016j. [8] E. Higgins, Carbohydrate analysis throughout the development of a protein therapeutic, Glycoconj. J. (2010). https://doi.org/10.1007/s10719-009-9261-x. [9] R.D. Cummings, A.G. Darvill, M.E. Etzler, M.G. Hahn, Glycan-Recognizing Probes as Tools, Essentials Glycobiol. (2015). https://doi.org/10.1101/glycobiology.3e.048. [10] M.A. Meehl, T.A. Stadheim, Biopharmaceutical discovery and production in yeast, Curr. Opin. Biotechnol. (2014). https://doi.org/10.1016/j.copbio.2014.06.007. [11] G. Walsh, R. Jefferis, Post-translational modifications in the context of therapeutic proteins, Nat. Biotechnol. (2006). https://doi.org/10.1038/nbt1252. [12] T.T. PUCK, The genetics of somatic mammalian cells., Adv. Biol. Med. Phys. (1957). https://doi.org/10.1016/b978-1-4832-3111-2.50006-7. [13] G. Urlaub, E. Käs, A.M. Carothers, L.A. Chasin, Deletion of the diploid dihydrofolate reductase locus from cultured mammalian cells, Cell. (1983). https://doi.org/10.1016/00928674(83)90422-1. [14] L. Fan, I. Kadura, L.E. Krebs, C.C. Hatfield, M.M. Shaw, C.C. Frye, Improving the efficiency of CHO cell line generation using glutamine synthetase gene knockout cells, Biotechnol. Bioeng. (2012). https://doi.org/10.1002/bit.24365. [15] J.A.H. Bort, B. Stern, N. Borth, CHO-K1 host cells adapted to growth in glutamine-free medium by FACS-assisted evolution, Biotechnol. J. (2010). https://doi.org/10.1002/biot.201000095. [16] T. Omasa, M. Onitsuka, W.-D. Kim, Cell Engineering and Cultivation of Chinese Hamster Ovary CHAPTER 2 35 (CHO) Cells, Curr. Pharm. Biotechnol. (2010). https://doi.org/10.2174/138920110791111960. [17] H. Hefzi, K.S. Ang, M. Hanscho, A. Bordbar, D. Ruckerbauer, M. Lakshmanan, C.A. Orellana, D. Baycin-Hizal, Y. Huang, D. Ley, V.S. Martinez, S. Kyriakopoulos, N.E. Jiménez, D.C. Zielinski, L.E. Quek, T. Wulff, J. Arnsdorf, S. Li, J.S. Lee, G. Paglia, N. Loira, P.N. Spahn, L.E. Pedersen, J.M. Gutierrez, Z.A. King, A.M. Lund, H. Nagarajan, A. Thomas, A.M. Abdel-Haleem, J. Zanghellini, H.F. Kildegaard, B.G. Voldborg, Z.P. Gerdtzen, M.J. Betenbaugh, B.O. Palsson, M.R. Andersen, L.K. Nielsen, N. Borth, D.Y. Lee, N.E. Lewis, A Consensus Genome-scale Reconstruction of Chinese Hamster Ovary Cell Metabolism, Cell Syst. 3 (2016) 434-443.e8. https://doi.org/10.1016/j.cels.2016.10.020. [18] J. Gifford, A. Albee, Z. Deeds, B. Delong, K. Kao, S. Ross, M. Caple, An Efficient Approach to Cell Culture Medium Optimization — A Statistical Method to Medium Mixing, in: Anim. Cell Technol. Meets Genomics, 2005. https://doi.org/10.1007/1-4020-3103-3_108. [19] E. Puente-Massaguer, L. Badiella, S. Gutiérrez-Granados, L. Cervera, F. Gòdia, A statistical approach to improve compound screening in cell culture media, Eng. Life Sci. (2019). https://doi.org/10.1002/elsc.201800168. [20] A. Hamdi, D. Széliová, D.E. Ruckerbauer, I. Rocha, N. Borth, J. Zanghellini, Key challenges in designing CHO chassis platforms, Processes. (2020). https://doi.org/10.3390/PR8060643. [21] J. Zhang, Protein-length distributions for the three domains of life, Trends Genet. (2000). https://doi.org/10.1016/S0168-9525(99)01922-8. [22] M. Wang, C.G. Kurland, G. Caetano-Anollés, Reductive evolution of proteomes and protein structures, Proc. Natl. Acad. Sci. U. S. A. (2011). https://doi.org/10.1073/pnas.1017361108. [23] C.G. Kurland, B. Canbäck, O.G. Berg, The origins of modern proteomes, Biochimie. (2007). https://doi.org/10.1016/j.biochi.2007.09.004. [24] F.R. Blattner, G. Plunkett, C.A. Bloch, N.T. Perna, V. Burland, M. Riley, J. Collado-Vides, J.D. Glasner, C.K. Rode, G.F. Mayhew, J. Gregor, N.W. Davis, H.A. Kirkpatrick, M.A. Goeden, D.J. Rose, B. Mau, Y. Shao, The complete genome sequence of Escherichia coli K-12, Science (80-. ). (1997). https://doi.org/10.1126/science.277.5331.1453. [25] G.G. Stewart, SACCHAROMYCES | Saccharomyces cerevisiae , in: C.A. Batt, M.L.B.T.-E. of F.M. (Second E. Tortorello (Eds.), Academic Press, Oxford, 2014: pp. 309–315. https://doi.org/https://doi.org/10.1016/B978-0-12-384730-0.00292-5. [26] J. Xu, N. Zhang, On the way to commercializing plant cell culture platform for CHAPTER 2 36 biopharmaceuticals: present status and prospect, Pharm. Bioprocess. (2014). https://doi.org/10.4155/pbp.14.32. [27] V.S. Martínez, M. Buchsteiner, P. Gray, L.K. Nielsen, L.E. Quek, Dynamic metabolic flux analysis using B-splines to study the effects of temperature shift on CHO cell metabolism, Metab. Eng. Commun. (2015). https://doi.org/10.1016/j.meteno.2015.06.001. [28] X. Pan, C. Dalm, R.H. Wijffels, D.E. Martens, Metabolic characterization of a CHO cell size increase phase in fed-batch cultures, Appl. Microbiol. Biotechnol. (2017). https://doi.org/10.1007/s00253-017-8531-y. [29] R. Milo, R. Phillips, Cell Biology by the Numbers, 2015. https://doi.org/10.1201/9780429258770. [30] G. Pósfai, G. Plunkett, T. Fehér, D. Frisch, G.M. Keil, K. Umenhoffer, V. Kolisnychenko, B. Stahl, S.S. Sharma, M. De Arruda, V. Burland, S.W. Harcum, F.R. Blattner, Emergent properties of reduced-genome Escherichia coli , Science (80-. ). (2006). https://doi.org/10.1126/science.1126439. [31] G.A. Bitter, K.M. Egan, W.N. Burnette, B. Samal, J.C. Fieschko, D.L. Peterson, M.R. Downing, J. Wypych, K.E. Langley, Hepatitis B vaccine produced in yeast, J. Med. Virol. (1988). https://doi.org/10.1002/jmv.1890250202. [32] R. Salari, R. Salari, Investigation of the Best Saccharomyces cerevisiae Growth Condition, Electron. Physician. (2017). https://doi.org/10.19082/3592. [33] D. Széliová, D.E. Ruckerbauer, S.N. Galleguillos, L.B. Petersen, K. Natter, M. Hanscho, C. Troyer, T. Causon, H. Schoeny, H.B. Christensen, D.Y. Lee, N.E. Lewis, G. Koellensperger, S. Hann, L.K. Nielsen, N. Borth, J. Zanghellini, What CHO is made of: Variations in the biomass composition of Chinese hamster ovary cell lines, Metab. Eng. (2020). https://doi.org/10.1016/j.ymben.2020.06.002. [34] E. Fernandez, J.M. & Hoeffler, J.P., Gene Expression Systems. Using nature for the art of expression, Acad. Press. San Diego. (1999). [35] O. Rupp, M.L. MacDonald, S. Li, H. Dhiman, S. Polson, S. Griep, K. Heffner, I. Hernandez, K. Brinkrolf, V. Jadhav, M. Samoudi, H. Hao, B. Kingham, A. Goesmann, M.J. Betenbaugh, N.E. Lewis, N. Borth, K.H. Lee, A reference genome of the Chinese hamster based on a hybrid assembly strategy, Biotechnol. Bioeng. (2018). https://doi.org/10.1002/bit.26722. [36] L.N. Nguyen, N. Novak, M. Baumann, J. Koehn, N. Borth, Bioinformatic Identification of Chinese Hamster Ovary (CHO) Cold-Shock Genes and Biological Evidence of their Cold-Inducible CHAPTER 2 43 (2018). https://doi.org/10.1002/bit.26724. [104] M. Taschwer, M. Hackl, J.A. Hernández Bort, C. Leitner, N. Kumar, U. Puc, J. Grass, M. Papst, R. Kunert, F. Altmann, N. Borth, Growth, productivity and protein glycosylation in a CHO EpoFc producer cell line adapted to glutamine-free growth, J. Biotechnol. (2012). https://doi.org/10.1016/j.jbiotec.2011.11.014. [105] Y.H. Jeong, S.S. Wang, Role of glutamine in hybridoma cell culture: Effects on cell growth, antibody production, and cell metabolism, Enzyme Microb. Technol. (1995). https://doi.org/10.1016/0141-0229(94)00041-O. [106] J.M. Matés, C. Pérez-Gómez, I.N. De Castro, M. Asenjo, J. Márquez, Glutamine and its relationship with intracellular redox status, oxidative stress and cell proliferation/death, Int. J. Biochem. Cell Biol. (2002). https://doi.org/10.1016/S1357-2725(01)00143-1. [107] Y. Imamoto, H. Tanaka, K. Takahashi, Y. Konno, T. Suzawa, Advantages of AlaGln as an additive to cell culture medium: Use with anti-CD20 chimeric antibody-producing POTELLIGENTTM CHO cell lines, Cytotechnology. (2013). https://doi.org/10.1007/s10616-012-9468-8. [108] T.K. Ha, G.M. Lee, Effect of glutamine substitution by TCA cycle intermediates on the production and sialylation of Fc-fusion protein in Chinese hamster ovary cell culture, J. Biotechnol. (2014). https://doi.org/10.1016/j.jbiotec.2014.04.002. [109] D. Geoghegan, C. Arnall, D. Hatton, J. Noble-Longster, C. Sellick, T. Senussi, D.C. James, Control of amino acid transport into Chinese hamster ovary cells, Biotechnol. Bioeng. (2018). https://doi.org/10.1002/bit.26794. [110] E. Baek, S.M. Noh, G.M. Lee, Anti-apoptosis engineering for improved protein production from CHO cells, in: Methods Mol. Biol., 2017. https://doi.org/10.1007/978-1-4939-6972-2_5. [111] S. Lee, G.M. Lee, Bcl-2 overexpression in CHO cells improves polyethylenimine-mediated gene transfection, Process Biochem. (2013). https://doi.org/10.1016/j.procbio.2013.03.021. [112] P. Kroll, K. Eilers, J. Fricke, C. Herwig, Impact of cell lysis on the description of cell growth and death in cell culture, Eng. Life Sci. (2017). https://doi.org/10.1002/elsc.201600088. [113] T. Klein, N. Heinzel, P. Kroll, M. Brunner, C. Herwig, L. Neutsch, Quantification of cell lysis during CHO bioprocesses: Impact on cell count, growth kinetics and productivity, J. Biotechnol. (2015). https://doi.org/10.1016/j.jbiotec.2015.04.021. [114] N. Fukuda, Y. Senga, S. Honda, Anxa2and Ctsd-knockout CHO cell lines to diminish the risk of contamination with host cell proteins, Biotechnol. Prog. (2019). https://doi.org/10.1002/btpr.2820. CHAPTER 2 44 [115] S. Kol, D. Ley, T. Wulff, M. Decker, J. Arnsdorf, S. Schoffelen, A.H. Hansen, T.L. Jensen, J.M. Gutierrez, A.W.T. Chiang, H.O. Masson, B.O. Palsson, B.G. Voldborg, L.E. Pedersen, H.F. Kildegaard, G.M. Lee, N.E. Lewis, Multiplex secretome engineering enhances recombinant protein production and purity, Nat. Commun. (2020). https://doi.org/10.1038/s41467-02015866-w. [116] J. Chiu, K.N. Valente, N.E. Levy, L. Min, A.M. Lenhoff, K.H. Lee, Knockout of a difficult-to-remove CHO host cell protein, lipoprotein lipase, for improved polysorbate stability in monoclonal antibody formulations, Biotechnol. Bioeng. (2017). https://doi.org/10.1002/bit.26237. [117] S. Fischer, R. Handrick, K. Otte, The art of CHO cell engineering: A comprehensive retrospect and future perspectives, Biotechnol. Adv. (2015). https://doi.org/10.1016/j.biotechadv.2015.10.015. [118] C. Monger, P.S. Kelly, C. Gallagher, M. Clynes, N. Barron, C. Clarke, Towards next generation CHO cell biology: Bioinformatics methods for RNA-Seq-based expression profiling, Biotechnol. J. (2015). https://doi.org/10.1002/biot.201500107. [119] H.F. Kildegaard, D. Baycin-Hizal, N.E. Lewis, M.J. Betenbaugh, The emerging CHO systems biology era: Harnessing the ’omics revolution for biotechnology, Curr. Opin. Biotechnol. (2013). https://doi.org/10.1016/j.copbio.2013.02.007. [120] N. Borth, D. Mattanovich, R. Kunert, H. Katinger, Effect of increased expression of protein disulfide isomerase and heavy chain binding protein on antibody secretion in a recombinant CHO cell line, Biotechnol. Prog. (2005). https://doi.org/10.1021/bp0498241. [121] C. Mohan, H.P. Soon, Y.C. Joo, G.M. Lee, Effect of doxycycline-regulated protein disulfide isomerase expression on the specific productivity of recombinant CHO cells: Thrombopoietin and antibody, Biotechnol. Bioeng. (2007). https://doi.org/10.1002/bit.21453. [122] M. Tigges, M. Fussenegger, Xbp1-based engineering of secretory capacity enhances the productivity of Chinese hamster ovary cells, Metab. Eng. (2006). https://doi.org/10.1016/j.ymben.2006.01.006. [123] E. Harreither, M. Hackl, J. Pichler, S. Shridhar, N. Auer, P.P. Łabaj, M. Scheideler, M. Karbiener, J. Grillari, D.P. Kreil, N. Borth, Microarray profiling of preselected CHO host cell subclones identifies gene expression patterns associated with in-creased production capacity, Biotechnol. J. (2015). https://doi.org/10.1002/biot.201400857. [124] N. Vishwanathan, A. Yongky, K.C. Johnson, H.Y. Fu, N.M. Jacob, H. Le, F.N.K. Yusufi, D.Y. Lee, W.S. Hu, Global insights into the Chinese hamster and CHO cell transcriptomes, Biotechnol. CHAPTER 2 45 Bioeng. (2015). https://doi.org/10.1002/bit.25513. [125] L. Zhao, H.Y. Fu, R. Raju, N. Vishwanathan, W.S. Hu, Unveiling gene trait relationship by crossplatform meta-analysis on Chinese hamster ovary cell transcriptome, Biotechnol. Bioeng. (2017). https://doi.org/10.1002/bit.26272. [126] A.B. Diendorfer, M. Hackl, G. Klanert, V. Jadhav, M. Reithofer, F. Stiefel, F. Hesse, J. Grillari, N. Borth, Annotation of additional evolutionary conserved microRNAs in CHO cells from updated genomic data, Biotechnol. Bioeng. (2015). https://doi.org/10.1002/bit.25539. [127] M. Hackl, V. Jadhav, T. Jakobi, O. Rupp, K. Brinkrolf, A. Goesmann, A. Pühler, T. Noll, N. Borth, J. Grillari, Computational identification of microRNA gene loci and precursor microRNA sequences in CHO cell lines, J. Biotechnol. (2012). https://doi.org/10.1016/j.jbiotec.2012.01.019. [128] M. Hackl, T. Jakobi, J. Blom, D. Doppmeier, K. Brinkrolf, R. Szczepanowski, S.H. Bernhart, C.H. zu Siederdissen, J.A.H. Bort, M. Wieser, R. Kunert, S. Jeffs, I.L. Hofacker, A. Goesmann, A. Pühler, N. Borth, J. Grillari, Next-generation sequencing of the Chinese hamster ovary microRNA transcriptome: Identification, annotation and profiling of microRNAs as targets for cellular engineering, J. Biotechnol. (2011). https://doi.org/10.1016/j.jbiotec.2011.02.011. [129] W.P. Loh, B. Loo, L. Zhou, P. Zhang, D.Y. Lee, Y. Yang, K.P. Lam, Overexpression of microRNAs enhances recombinant protein production in Chinese hamster ovary cells, Biotechnol. J. (2014). https://doi.org/10.1002/biot.201400050. [130] A. Druz, Y.J. Son, M. Betenbaugh, J. Shiloach, Stable inhibition of mmu-miR-466h-5p improves apoptosis resistance and protein production in CHO cells, Metab. Eng. (2013). https://doi.org/10.1016/j.ymben.2012.12.004. [131] V. Jadhav, M. Hackl, G. Klanert, J.A. Hernandez Bort, R. Kunert, J. Grillari, N. Borth, Stable overexpression of miR-17 enhances recombinant protein production of CHO cells, J. Biotechnol. (2014). https://doi.org/10.1016/j.jbiotec.2014.01.032. [132] P.S. Kelly, L. Breen, C. Gallagher, S. Kelly, M. Henry, N.T. Lao, P. Meleady, D. O’Gorman, M. Clynes, N. Barron, Re-programming CHO cell metabolism using miR-23 tips the balance towards a highly productive phenotype, Biotechnol. J. (2015). https://doi.org/10.1002/biot.201500101. [133] S. Fischer, K.F. Marquart, L.A. Pieper, J. Fieder, M. Gamer, I. Gorr, P. Schulz, H. Bradl, miRNA engineering of CHO cells facilitates production of difficult-to-express proteins and increases success in cell line development, Biotechnol. Bioeng. (2017). CHAPTER 2 46 https://doi.org/10.1002/bit.26280. [134] N.H. Simpson, R.P. Singh, A.N. Emery, M. Al-Rubeai, Bcl-2 over-expression reduces growth rate and prolongs G1 phase in continuous chemostat cultures of hybridoma cells, Biotechnol. Bioeng. (1999). https://doi.org/10.1002/(SICI)1097-0290(19990720)64:2<174::AIDBIT6>3.0.CO;2-D. [135] H. Kaufmann, X. Mazur, M. Fussenegger, J.E. Bailey, Influence of low temperature on productivity, proteome and protein phosphorylation of CHO cells, Biotechnol. Bioeng. (1999). https://doi.org/10.1002/(SICI)1097-0290(19990605)63:5<573::AID-BIT7>3.0.CO;2-Y. [136] J. Xu, P. Tang, A. Yongky, B. Drew, M.C. Borys, S. Liu, Z.J. Li, Systematic development of temperature shift strategies for Chinese hamster ovary cells based on short duration cultures and kinetic modeling, MAbs. (2019). https://doi.org/10.1080/19420862.2018.1525262. [137] A. Bedoya-López, K. Estrada, A. Sanchez-Flores, O.T. Ramírez, C. Altamirano, L. Segovia, J. Miranda-Ríos, M.A. Trujillo-Roldán, N.A. Valdez-Cruz, Effect of temperature downshift on the transcriptomic responses of Chinese hamster ovary cells using recombinant human tissue plasminogen activator production culture, PLoS One. (2016). https://doi.org/10.1371/journal.pone.0151529. [138] Z. Jiang, S.T. Sharfstein, Sodium butyrate stimulates monoclonal antibody over-expression in CHO cells by improving gene accessibility, Biotechnol. Bioeng. (2008). https://doi.org/10.1002/bit.21726. [139] J.X. Bi, J. Shuttleworth, M. Al-Rubeai, Uncoupling of Cell Growth and Proliferation Results in Enhancement of Productivity in p21CIP1-Arrested CHO Cells, Biotechnol. Bioeng. (2004). https://doi.org/10.1002/bit.20025. [140] M. Fussenegger, S. Schlatter, D. Dätwyler, X. Mazur, J.E. Bailey, Controlled proliferation by multigene metabolic engineering enhances the productivity of Chinese hamster ovary cells, Nat. Biotechnol. (1998). https://doi.org/10.1038/nbt0598-468. [141] S. Klamt, R. Mahadevan, O. Hädicke, When Do Two-Stage Processes Outperform One-Stage Processes?, Biotechnol. J. (2018). https://doi.org/10.1002/biot.201700539. [142] L. Liu, Antibody glycosylation and its impact on the pharmacokinetics and pharmacodynamics of monoclonal antibodies and Fc-fusion proteins, J. Pharm. Sci. (2015). https://doi.org/10.1002/jps.24444. [143] M. Spearman, M. Butler, Glycosylation in Cell Culture, in: 2015. https://doi.org/10.1007/9783-319-10320-4_9. CHAPTER 2 47 [144] J. Ehret, M. Zimmermann, T. Eichhorn, A. Zimmer, Impact of cell culture media additives on IgG glycosylation produced in Chinese hamster ovary cells, Biotechnol. Bioeng. (2019). https://doi.org/10.1002/bit.26904. [145] R. Jefferis, Glycosylation as a strategy to improve antibody-based therapeutics, Nat. Rev. Drug Discov. (2009). https://doi.org/10.1038/nrd2804. [146] L. Malphettes, Y. Freyvert, J. Chang, P.Q. Liu, E. Chan, J.C. Miller, Z. Zhou, T. Nguyen, C. Tsai, A.W. Snowden, T.N. Collingwood, P.D. Gregory, G.J. Cost, Highly efficient deletion of FUT8 in CHO cell lines using zinc-finger nucleases yields cells that produce completely nonfucosylated antibodies, Biotechnol. Bioeng. (2010). https://doi.org/10.1002/bit.22751. [147] N. Lin, J. Mascarenhas, N.R. Sealover, H.J. George, J. Brooks, K.J. Kayser, B. Gau, I. Yasa, P. Azadi, S. Archer-Hartmann, Chinese hamster ovary (CHO) host cell engineering to increase sialylation of recombinant therapeutic proteins by modulating sialyltransferase expression, Biotechnol. Prog. (2015). https://doi.org/10.1002/btpr.2038. [148] M.T. Coats, N. Bydlinski, D. Maresch, A. Diendorfer, G. Klanert, N. Borth, mRNA Transfection into CHO-Cells Reveals Production Bottlenecks, Biotechnol. J. (2020). https://doi.org/10.1002/biot.201900198. [149] W. Tian, Z. Ye, S. Wang, M.A. Schulz, J. Van Coillie, L. Sun, Y.H. Chen, Y. Narimatsu, L. Hansen, C. Kristensen, U. Mandel, E.P. Bennett, S. Jabbarzadeh-Tabrizi, R. Schiffmann, J.S. Shen, S.Y. Vakhrushev, H. Clausen, Z. Yang, The glycosylation design space for recombinant lysosomal replacement enzymes produced in CHO cells, Nat. Commun. (2019). https://doi.org/10.1038/s41467-019-09809-3. [150] C.S. Stach, M.G. McCann, C.M. O’Brien, T.S. Le, N. Somia, X. Chen, K. Lee, H.Y. Fu, P. Daoutidis, L. Zhao, W.S. Hu, M. Smanski, Model-Driven Engineering of N-Linked Glycosylation in Chinese Hamster Ovary Cells, ACS Synth. Biol. (2019). https://doi.org/10.1021/acssynbio.9b00215. [151] P. Kotidis, P. Jedrzejewski, S.N. Sou, C. Sellick, K. Polizzi, I.J. del Val, C. Kontoravdi, Modelbased optimization of antibody galactosylation in CHO cell culture, Biotechnol. Bioeng. (2019). https://doi.org/10.1002/bit.26960. [152] A.M. Lewis, N.R. Abu-Absi, M.C. Borys, Z.J. Li, The use of ’Omics technology to rationally improve industrial mammalian cell line performance, Biotechnol. Bioeng. (2016). https://doi.org/10.1002/bit.25673. [153] Y.T. Yang, G.N. Bennett, K.Y. San, Genetic and metabolic engineering, Electron. J. Biotechnol. CHAPTER 2 48 (1998). [154] D. Széliová, D. Iurashev, D.E. Ruckerbauer, G. Koellensperger, N. Borth, M. Melcher, J. Zanghellini, Error propagation in constraint-based modeling of Chinese hamster ovary cells, Biotechnol. J. (2021). https://doi.org/10.1002/biot.202000320. [155] Y. Cao, S. Kimura, T. Itoi, K. Honda, H. Ohtake, T. Omasa, Construction of BAC-based physical map and analysis of chromosome rearrangement in chinese hamster ovary cell lines, Biotechnol. Bioeng. (2012). https://doi.org/10.1002/bit.24347. [156] M. Derouazi, D. Martinet, N. Besuchet Schmutz, R. Flaction, M. Wicht, M. Bertschinger, D.L. Hacker, J.S. Beckmann, F.M. Wurm, Genetic characterization of CHO production host DG44 and derivative recombinant cell lines, Biochem. Biophys. Res. Commun. (2006). https://doi.org/10.1016/j.bbrc.2005.12.111. [157] J.S. Lee, H.F. Kildegaard, N.E. Lewis, G.M. Lee, Mitigating Clonal Variation in Recombinant Mammalian Cell Lines, Trends Biotechnol. (2019). https://doi.org/10.1016/j.tibtech.2019.02.007. [158] H. Dhiman, M.P. Gerstl, D. Ruckerbauer, M. Hanscho, H. Himmelbauer, C. Clarke, N. Barron, J. Zanghellini, N. Borth, Genetic and Epigenetic Variation across Genes Involved in Energy Metabolism and Mitochondria of Chinese Hamster Ovary Cell Lines, Biotechnol. J. (2019). https://doi.org/10.1002/biot.201800681. [159] J. Feichtinger, I. Hernández, C. Fischer, M. Hanscho, N. Auer, M. Hackl, V. Jadhav, M. Baumann, P.M. Krempl, C. Schmidl, M. Farlik, M. Schuster, A. Merkel, A. Sommer, S. Heath, D. Rico, C. Bock, G.G. Thallinger, N. Borth, Comprehensive genome and epigenome characterization of CHO cells in response to evolutionary pressures and over time, Biotechnol. Bioeng. (2016). https://doi.org/10.1002/bit.25990. [160] C.S. Kaas, C. Kristensen, M.J. Betenbaugh, M.R. Andersen, Sequencing the CHO DXB11 genome reveals regional variations in genomic stability and haploidy, BMC Genomics. (2015). https://doi.org/10.1186/s12864-015-1391-x. [161] S. Vcelar, M. Melcher, N. Auer, A. Hrdina, A. Puklowski, F. Leisch, V. Jadhav, T. Wenger, M. Baumann, N. Borth, Changes in Chromosome Counts and Patterns in CHO Cell Lines upon Generation of Recombinant Cell Lines and Subcloning, Biotechnol. J. (2018). https://doi.org/10.1002/biot.201700495. [162] S. Vcelar, V. Jadhav, M. Melcher, N. Auer, A. Hrdina, R. Sagmeister, K. Heffner, A. Puklowski, M. Betenbaugh, T. Wenger, F. Leisch, M. Baumann, N. Borth, Karyotype variation of CHO host cell CHAPTER 2 49 lines over time in culture characterized by chromosome counting and chromosome painting, Biotechnol. Bioeng. (2018). https://doi.org/10.1002/bit.26453. [163] W. Pilbrough, T.P. Munro, P. Gray, Intraclonal protein expression heterogeneity in recombinant CHO cells, PLoS One. (2009). https://doi.org/10.1371/journal.pone.0008432. [164] K. Brinkrolf, O. Rupp, H. Laux, F. Kollin, W. Ernst, B. Linke, R. Kofler, S. Romand, F. Hesse, W.E. Budach, S. Galosy, D. Müller, T. Noll, J. Wienberg, T. Jostock, M. Leonard, J. Grillari, A. Tauch, A. Goesmann, B. Helk, J.E. Mott, A. Pühler, N. Borth, Chinese hamster genome sequenced from sorted chromosomes, Nat. Biotechnol. (2013). https://doi.org/10.1038/nbt.2645. [165] I. Thiele, B. Palsson, A protocol for generating a high-quality genome-scale metabolic reconstruction, Nat. Protoc. (2010). https://doi.org/10.1038/nprot.2009.203. [166] A. Bordbar, J.M. Monk, Z.A. King, B.O. Palsson, Constraint-based models predict metabolic and associated cellular functions, Nat. Rev. Genet. (2014). https://doi.org/10.1038/nrg3643. [167] A.M. Feist, B. Palsson, The growing scope of applications of genome-scale metabolic reconstructions using Escherichia coli , Nat. Biotechnol. (2008). https://doi.org/10.1038/nbt1401. [168] N.E. Lewis, H. Nagarajan, B.O. Palsson, Constraining the metabolic genotype-phenotype relationship using a phylogeny of in silico methods, Nat. Rev. Microbiol. (2012). https://doi.org/10.1038/nrmicro2737. [169] D. Machado, M.J. Herrgård, I. Rocha, Stoichiometric Representation of Gene–Protein–Reaction Associations Leverages Constraint-Based Analysis from Reaction to Gene-Level Phenotype Prediction, PLoS Comput. Biol. (2016). https://doi.org/10.1371/journal.pcbi.1005140. [170] Ž. Rejc, L. Magdevska, T. Tršelič, T. Osolin, R. Vodopivec, J. Mraz, E. Pavliha, N. Zimic, T. Cvitanović, D. Rozman, M. Moškon, M. Mraz, Computational modelling of genome-scale metabolic networks and its application to CHO cell cultures, Comput. Biol. Med. (2017). https://doi.org/10.1016/j.compbiomed.2017.07.005. [171] C. Gu, G.B. Kim, W.J. Kim, H.U. Kim, S.Y. Lee, Current status and applications of genome-scale metabolic models, Genome Biol. (2019). https://doi.org/10.1186/s13059-019-1730-3. [172] I. Rocha, J. Förster, J. Nielsen, Design and Application of Genome-Scale Reconstructed Metabolic Models, in: 2008. https://doi.org/10.1007/978-1-59745-321-9_29. [173] A.M. Feist, M.J. Herrgård, I. Thiele, J.L. Reed, B. Palsson, Reconstruction of biochemical networks in microorganisms, Nat. Rev. Microbiol. (2009). CHAPTER 2 50 https://doi.org/10.1038/nrmicro1949. [174] J.S. Edwards, B.O. Palsson, Systems properties of the Haemophilus influenzae Rd metabolic genotype, J. Biol. Chem. (1999). https://doi.org/10.1074/jbc.274.25.17410. [175] M.A. Oberhardt, B. Palsson, J.A. Papin, Applications of genome-scale metabolic reconstructions, Mol. Syst. Biol. (2009). https://doi.org/10.1038/msb.2009.77. [176] B. Traustason, Amino Acid Requirements of the Chinese Hamster Ovary Cell Metabolism during Recombinant Protein Production, BioRxiv. (2019). https://doi.org/10.1101/796490. [177] C. Calmels, A. McCann, L. Malphettes, M.R. Andersen, Application of a curated genome-scale metabolic model of CHO DG44 to an industrial fed-batch process, Metab. Eng. (2019). https://doi.org/10.1016/j.ymben.2018.09.009. [178] N.C. Duarte, S.A. Becker, N. Jamshidi, I. Thiele, M.L. Mo, T.D. Vo, R. Srivas, B. Palsson, Global reconstruction of the human metabolic network based on genomic and bibliomic data, Proc. Natl. Acad. Sci. U. S. A. (2007). https://doi.org/10.1073/pnas.0610772104. [179] L.E. Quek, S. Dietmair, M. Hanscho, V.S. Martínez, N. Borth, L.K. Nielsen, Reducing Recon 2 for steady-state flux analysis of HEK cell culture, J. Biotechnol. (2014). https://doi.org/10.1016/j.jbiotec.2014.05.021. [180] N. Swainston, K. Smallbone, H. Hefzi, P.D. Dobson, J. Brewer, M. Hanscho, D.C. Zielinski, K.S. Ang, N.J. Gardiner, J.M. Gutierrez, S. Kyriakopoulos, M. Lakshmanan, S. Li, J.K. Liu, V.S. Martínez, C.A. Orellana, L.E. Quek, A. Thomas, J. Zanghellini, N. Borth, D.Y. Lee, L.K. Nielsen, D.B. Kell, N.E. Lewis, P. Mendes, Recon 2.2: from reconstruction to model of human metabolism, Metabolomics. (2016). https://doi.org/10.1007/s11306-016-1051-4. [181] Z.A. King, C.J. Lloyd, A.M. Feist, B.O. Palsson, Next-generation genome-scale models for metabolic engineering, Curr. Opin. Biotechnol. (2015). https://doi.org/10.1016/j.copbio.2014.12.016. [182] H. Fouladiha, S.A. Marashi, S. Li, B. Vaziri, N.E. Lewis, Systematically gap-filling the genomescale model of CHO cells, BioRxiv. (2020). https://doi.org/10.1101/2020.01.27.921296. [183] H.C. Yeo, J. Hong, M. Lakshmanan, D.Y. Lee, Enzyme capacity-based genome scale modelling of CHO cells, Metab. Eng. (2020). https://doi.org/10.1016/j.ymben.2020.04.005. [184] J.M. Gutierrez, A. Feizi, S. Li, T.B. Kallehauge, H. Hefzi, L.M. Grav, D. Ley, D. Baycin Hizal, M.J. Betenbaugh, B. Voldborg, H. Faustrup Kildegaard, G. Min Lee, B.O. Palsson, J. Nielsen, N.E. Lewis, Genome-scale reconstructions of the mammalian secretory pathway predict metabolic costs and limitations of protein secretion, Nat. Commun. (2020). CHAPTER 2 51 https://doi.org/10.1038/s41467-019-13867-y. [185] Z. Huang, D.Y. Lee, S. Yoon, Quantitative intracellular flux modeling and applications in biotherapeutic development and production using CHO cell cultures, Biotechnol. Bioeng. (2017). https://doi.org/10.1002/bit.26384. [186] G. Sambamoorthy, K. Raman, MinReact: a systematic approach for identifying minimal metabolic networks, Bioinformatics. (2020). https://doi.org/10.1093/bioinformatics/btaa497. [187] Z. Huang, J. Xu, A. Yongky, C.S. Morris, A.L. Polanco, M. Reily, M.C. Borys, Z.J. Li, S. Yoon, CHO cell productivity improvement by genome-scale modeling and pathway analysis: Application to feed supplements, Biochem. Eng. J. (2020). https://doi.org/10.1016/j.bej.2020.107638. [188] S.M. Schinn, C. Morrison, W. Wei, L. Zhang, N.E. Lewis, A genome-scale metabolic network model and machine learning predict amino acid concentrations in Chinese Hamster Ovary cell cultures, Biotechnol. Bioeng. (2021). https://doi.org/10.1002/bit.27714. [189] B.A. Pérez-Fernández, J. Fernandez-de-Cossio-Diaz, T. Boggiano, K. León, R. Mulet, In-silico media optimization for continuous cultures using genome scale metabolic networks: The case of CHO-K1, Biotechnol. Bioeng. (2021). https://doi.org/10.1002/bit.27704. [190] D. Széliová, J. Štor, I. Thiel, M. Weinguny, M. Hanscho, G. Lhota, N. Borth, J. Zanghellini, D.E. Ruckerbauer, I. Rocha, Inclusion of maintenance energy improves the intracellular flux predictions of CHO, BioRxiv. (2020). https://doi.org/10.1101/2020.12.21.423738. [191] J. Zanghellini, D.E. Ruckerbauer, M. Hanscho, C. Jungreuthmayer, Elementary flux modes in a nutshell: Properties, calculation and applications, Biotechnol. J. (2013). https://doi.org/10.1002/biot.201200269. [192] J.D. Orth, I. Thiele, B.O. Palsson, What is flux balance analysis?, Nat. Biotechnol. (2010). https://doi.org/10.1038/nbt.1614. [193] J. Gagneur, S. Klamt, Computation of elementary modes: A unifying framework and the new binary approach, BMC Bioinformatics. (2004). https://doi.org/10.1186/1471-2105-5-175. [194] P. Maia, M. Rocha, I. Rocha, In Silico Constraint-Based Strain Optimization Methods: the Quest for Optimal Cell Factories , Microbiol. Mol. Biol. Rev. (2016). https://doi.org/10.1128/mmbr.00014-15. [195] I. Rocha, P. Maia, P. Evangelista, P. Vilaça, S. Soares, J.P. Pinto, J. Nielsen, K.R. Patil, E.C. Ferreira, M. Rocha, OptFlux: An open-source software platform for in silico metabolic engineering, BMC Syst. Biol. (2010). https://doi.org/10.1186/1752-0509-4-45. [196] T. Shlomi, M.N. Cabili, M.J. Herrgård, B. Palsson, E. Ruppin, Network-based prediction of CHAPTER 2 52 human tissue-specific metabolism, Nat. Biotechnol. (2008). https://doi.org/10.1038/nbt.1487. [197] S. Santos, I. Rocha, Estimation of biomass composition from genomic and transcriptomic information, J. Integr. Bioinform. (2017). https://doi.org/10.1515/jib-2016-285. [198] N.E. Lewis, K.K. Hixson, T.M. Conrad, J.A. Lerman, P. Charusanti, A.D. Polpitiya, J.N. Adkins, G. Schramm, S.O. Purvine, D. Lopez-Ferrer, K.K. Weitz, R. Eils, R. König, R.D. Smith, B. Palsson, Omic data from evolved E. coli are consistent with computed optimal growth from genome-scale models, Mol. Syst. Biol. (2010). https://doi.org/10.1038/msb.2010.47. [199] R. Schuetz, L. Kuepfer, U. Sauer, Systematic evaluation of objective functions for predicting intracellular fluxes in Escherichia coli , Mol. Syst. Biol. (2007). https://doi.org/10.1038/msb4100162. [200] A.M. Feist, B.O. Palsson, The biomass objective function, Curr. Opin. Microbiol. (2010). https://doi.org/10.1016/j.mib.2010.03.003. [201] J.C. Lachance, C.J. Lloyd, J.M. Monk, L. Yang, A. V. Sastry, Y. Seif, B.O. Palsson, S. Rodrigue, A.M. Feist, Z.A. King, P.É. Jacques, BOFDAT: Generating biomass objective functions for genome-scale metabolic models from experimental data, PLoS Comput. Biol. (2019). https://doi.org/10.1371/journal.pcbi.1006971. [202] D. Dikicioglu, B. Kırdar, S.G. Oliver, Biomass composition: the “elephant in the room” of metabolic modelling, Metabolomics. (2015). https://doi.org/10.1007/s11306-015-0819-2. [203] E. Almaas, Z.N. Oltvai, A.-L. Barabási, The Activity Reaction Core and Plasticity of Metabolic Networks, PLoS Comput. Biol. (2005). https://doi.org/10.1371/journal.pcbi.0010068. [204] C. Verduyn, A.H. Stouthamer, W.A. Scheffers, J.P. van Dijken, A theoretical evaluation of growth yields of yeasts, Antonie Van Leeuwenhoek. (1991). https://doi.org/10.1007/BF00582119. [205] F.R. Sidoli, A. Mantalaris, S.P. Asprey, Modelling of mammalian cells and cell culture processes, Cytotechnology. (2004). https://doi.org/10.1023/B:CYTO.0000043397.94527.84. [206] F.J. Krambeck, S. V. Bennun, M.R. Andersen, M.J. Betenbaugh, Model-based analysis of Nglycosylation in Chinese hamster ovary cells, PLoS One. (2017). https://doi.org/10.1371/journal.pone.0175376. CHAPTER 3 59 by-products. In these predictions, the constraints of key metabolites were based on data in Hefzi et al, 2016 [35] combined with potential candidates predicted by optiModels. In addition, two different biomass equations were employed, and various environmental conditions were tested. The latter are expressed in mmol gDW-1 hr-1 and refer to the exchange rates values in steady state. Figure 3.1 Medium Optimization: Prediction workflow. GSMM (Genome scale metabolic model), pFBA (Parsimonious flux balance analysis). Multiple optimizations were performed and the values of growth rate and by-products (ammonium) were compared to the results with original constraints of the model which was constructed based on different simulations and omics data described in previous published reports [35,37,38]. This optimization pipeline can be very helpful to screen the effect of balancing the levels of different compounds in the growth medium, through pFBA. Based on the literature, the amino acids uptake levels were optimized to maximize cell growth and particularly to overcome the secretion of process by-products such as ammonium. To better compare/understand the results, the values of both growth rate and growth yield were calculated in order to assess/compare the prediction results. Growth yield values were calculated according to the following equation where 𝜇 represents the predicted specific growth rate and 𝑞𝐺𝑙𝑐 is the predicted specific consumption rate of glucose. CHAPTER 3 60 𝐺𝑟𝑜𝑤𝑡ℎ 𝑦𝑖𝑒𝑙𝑑 = 𝜇 𝑞𝐺𝑙𝑐 3.2.2.2. Maximizing growth and recombinant protein production using FVA In order to predict the maximum antibody production capacity using the original model constraints (Hefzi et al., 2016) and the in-house optimized constraints, flux variability analysis (FVA) was used. The latter tool was used to calculate the maximum possible value of a selected flux, in our case, the specific reaction for the production of IgG described in the model as (DM_igg[g]), for a range of fixed values of the biomass reaction. 3.3. Optimization results and discussion 3.3.1. Optimization of minimal medium using optiModels The optimal set of reactions that are capable to sustain CHO growth were determined. Several combinations of reactions were determined with a fitness score restrained between 0.7 to 1, as described in materials and methods. Within the solutions obtained using optiModels, we determined the 30 most frequent reactions that resulted from the predictions, relying on both biomass reactions, for producing and non-producing CHO cells. The comparison is highlighted in figure 3.2. In this representation, we observe the percentage of frequency versus the corresponding exchange reaction, where “R_EX” means Exchange reaction and at the end of the naming of the exchange reaction, “_e” means extracellular. Several exchange reactions of amino acids such as tryptophane, valine and isoleucine were part of the prediction solutions. Exploring these results deeper, we noticed that the solutions with the optimal fitness scores also included many reactions involving the catabolism of several complex structures of molecules, mainly vitamins, oligopeptides and polypeptides. Among these solutions, kinetensin 1-8, corresponding to the exchange reaction (R_CE5789_e) appeared in more than 70 % of the optimal results obtained by optiModels. CHAPTER 3 61 Figure 3.2 Frequency of reactions obtained by optiModels algorithm when minimizing the number of exchange reactions ensuring growth. trp_L (Tryptophan), thr_L (Threonine), Ile_L (Isoleucine), val-L (Valine), pydx5p (Pyridoxal-5-Phosphate), met_L (Methionine), cern (Carnitine), biocyt (Biocytin), leugly(Leucylglycine), CE5789 (Kinetensin 18), CE5786 (Kinetensin), debrisoquine (Debrisoquine), no (Nitric oxide), dgsn (Deoxyguanosine), CE0074 (Alloxan), leuktrF4 (Leukotriene F4), 9_cis_retfa (Fatty acid 9-cis-retinol), amp (Adenosine monophosphate), tag_cho (Triacylglycerol (cho)), dgmp (Deoxyguanosine monophosphate), taur (Taurine), CE4723 (Neocasomorphin (1-5)), galfuc12gal14acglcgalgl ((Gal)3 (Glc)1 (GlcNAc)1 (LFuc)1 (Cer)1), retn (Retinoate), sph1p (Sphinganine 1-phosphate), akg (α-ketoglutarate), gltdechol (beta glucan-taurodeoxycholic acid complex), apnnox (Alpha-Pinene-oxide). The latter oligopeptide is chemically composed by several amino acids, in this case L-Isoleucinele-LAlanine-L-Arginine-L-Arginine-L-Histidine-L-Proline-L-Tyrrosine-L-Phenylalanine-L-Leucine. This high molecular weight molecule (C50H76N16O10) [39] is not a potential candidate for cell culture medium component, not only because of its complex chemical structure, but also because its high price in the market and the lack of commercial availability. In fact, these results are expected, since the model does not account for the biological feasibility of these solutions, but calculates, mathematically, the easiest metabolic path through pFBA towards growth and determines the best candidates with minimal medium components. Subsequently, further filtering of the data has to be performed based on literature studies to choose the best set of candidates proposed by the model that can be consumed by CHO. Nevertheless, it is important to emphasize that CHAPTER 3 62 the prediction results are merely an indication of which pathway should be targeted in terms of experimental implementation. Based on the results described in figure 3.2, for example, pyridoxal-5'-phosphate (P5P) appears to be a potential supplement for cell culture medium. This vitamin, B6, is essential for many mammalian metabolic reactions, acting as a coenzyme for several transamination reactions, mainly involved in the decarboxylation of amino acids [40]. Within the optimal solutions, several reactions are appealing from a biological feasibility point of view, namely amino acids and vitamins. In this study, our target is to narrow the solution space and focus on amino acids, together with other nutrients that are directly involved in the central metabolism. Future studies should be focused on exploring the impact of the other candidates obtained by optiModels, for instance vitamins. Studying the impact of changing the levels of vitamins in the media and their influence on metabolism can thus be very interesting to explore in future studies. 3.3.2. In silico CHO cell culture media optimization As described in the previous section, the genome scale metabolic model of CHO constructed by Hefzi et al, 2016, was combined with an in-house developed algorithm in order to determine the minimal and most essential components for CHO cells growth. Two different biomass reactions were used in the in silico predictions, (R_biomass_cho) and (R_biomass_cho_producing), respectively, for producing and non-producing CHO cells as previously described. Several media components were predicted for both cell types. In the following step we used generated data in order to build the optimal amino acids formulation to sustain CHO growth. 3.3.2.1. Medium optimization for non-producing CHO cell lines In this step, various in silico predictions were performed in order to improve the specific growth rate and decrease cell culture by-products using as basis the results from the previous section. The prediction results were based on varying the environmental conditions, i. e., changing the boundaries of uptake fluxes at which certain metabolites are consumed. This strategy can be used to define the components of cell culture media of a specific cell line, in our case, CHO cells. Multiple simulations were performed and the results were compared to the original constraints of the model which was constructed based on different simulations and omics data described in previous published reports [35,38,41]. The in-house simulations were performed in Optflux. The simulations were performed based on the biomass equation CHAPTER 3 63 (R_biomass_cho) specific for the prediction of growth rate of non-producer CHO cell lines. The input data were based on literature studies and also on different data generated using the Python-based algorithm (optiModels) previously described. Using this approach, the optimal combinations of amino acids used for the growth of non-producer CHO cells were determined. The uptake reaction bounds (or constraints) referred in the model developed by Hefzi et al, 2016 and the in-house optimized constraints are referred in Figure 3.3 (Supplementary data in annexes). The values in Table 3.2 in Annexes are expressed in (mmol gDW-1 hr-1) and refer to the rate at which the cell consumes certain metabolites for growth (in steady state). The prediction results account for the consumption (Figure 3.2) and production (Figure 3.4) fluxes of different metabolites, as well as the growth rate value (Figure 3.3). Various differences were observed between the constraints developed in-house and to the original ones. In our work, we increased the uptake for different metabolites such as glucose and other amino acids, since they play a role in increasing cellular growth rate. On the other hand, we limited the consumption of a number of metabolites since their degradation in cell culture may promote the secretion of toxic byproducts that can inhibit cell growth. As an example, in order to decrease ammonium accumulation in cell culture medium, glutamine was removed from the optimized formulation. This amino acid is known to be a major source of ammonium which is considered as a toxic metabolite (at high concentrations: approximately over 4mM) for the cells (as described in chapter 2). Ammonium diffusion across the cell membrane play a major part in disturbing intracellular pH in addition to the electrochemical gradients [42]. Its build-up in the cell culture medium can significantly inhibit cell growth and final cell densities [43,44]. Removal of glutamine from the cell culture medium was previously studied. According to the literature, cells cultured with a glutamine substitute in the feed medium generated a decrease in cell growth and also a decrease in ammonium levels [45,46]. On the one hand, additional amino acids were removed from the in-house medium formulation such as serine and asparagine. Serine is considered also as a potential source of ammonium secretion, since its degradation using serine dehydratase, also called serine ammonia lyase (SDH) can generate pyruvate and also ammonia [47]. Another study proved that serine presence in the cell culture medium did not improve the growth rates of the cells [42].The in-house in silico predictions using Optflux proved also that when forcing the over-consumption of serine, ammonium secretion rates increased significantly, and no improvements were observed regarding cell growth. Subsequently, we can presume that the in silico results are in accordance with the published data in Chen et al 2005 [42]. CHAPTER 3 64 On the other hand, asparagine supplementation and consumption was proven to be linked to ammonia and alanine accumulation in the culture medium during growth [37]. Alanine accumulation during culture can also inhibit cell growth through repressing the TCA cycle, one of the most important pathways of the cellular machinery [48]. Our strategy was based on finding alternatives to glutamine, serine and asparagine that can improve cell growth and mitigate the effect of ammonium. The in silico predictions based on optiModels definitely played a role in facilitating this task. As part of the optimization strategy, the consumption rates of different groups of amino acids were set out to the minimum allowed levels. Tryptophan and methionine consumption rates were also set to the minimum since, as described in previously published reports by Pfzier, their presence in the cell culture medium in higher levels can generate a variety of toxic metabolites, considered as putative growth inhibitors for CHO cells. Examples can be indole 3-lactate and 2-Hydroxybutyrate [49]. On the other hand, disparate amino acids play a major part in improving CHO growth rate, for instance, proline and threonine [42]. Proline supplementation to the cells cultured in proline-free medium was previously studied, and it demonstrated a positive effect on cell growth [50]. Figure 3.3 Predicted optimal metabolites uptake rates. Comparison between the prediction results of Hefzi et al., 2016 and the results of the in-house optimizations for non-producer CHO cell line. Sink Tyr ggn - Sink reaction for Tyr-194 Of Apo-Glycogenin Protein (Primer for Glycogen Synthesis). As commonly known, the genetic makeup of the cell lines, their expression profile as well as the environment in which cells are present, can obviously influence the consumption rates and the metabolic fluxes of amino acids. The rationale behind this optimization strategy is to develop a formulation which includes not only the amino acids that are easily consumed by the cells, but also 0 0.2 0.4 0.6 0.8 1 1.2 Oxygen D-Glucose L-Glutamine L-Serine L-Asparagine L-Arginine L-Leucine L-Lysine L-Valine L-Threonine L-Isoleucine L-Aspartate L-Proline L-Phenylalanine L-Tyrosine L-Methionine L-Histidine L-Cysteine L-Tryptophan Choline Pyridoxal Sink Tyr ggn Myo-Inositol Hypoxanthine Phosphate Consumption rate (mmol gDW-1 hr -1) CHAPTER 3 65 other metabolites that can retrofit the metabolism of the cells towards more efficiency. Subsequently, higher growth and extended viability of the cells is expected. As shown in figure 3.3, higher oxygen and glucose consumption were noticed as part of the predictions results when using our in-house optimized constraints. These results align with the fact that more oxygen and carbon source are needed as a means to achieve higher cell densities. We can also observe that higher amounts of valine, leucine, aspartate, proline and histidine have to be supplemented in the optimized cell culture medium to accommodate higher glucose consumption and the decrease in other amino acids. Concerning vitamins uptake values, a slightly higher uptake rate of choline was observed in the optimized medium formulation. Also, based on the results obtained, pyridoxal, a form of vitamin B6, positively impacts cell growth and plays a role in the decrease in ammonium accumulation in the cell culture medium. Vitamin B6 has a central role in the metabolism of amino acids. As an example, the cofactor Pyridoxal-5’-Phosphate plays an important role in the catalysis of many important steps in the metabolism of amino acids, such as transamination, racemization, decarboxylation, and α,β-elimination reactions [51]. According to the stoichiometric matrix described in the GSMM, pyridoxal plays a role in transamination through pyridoxal kinase and in the use of ammonium to produce pyridoxine. Pyridoxal potential in decreasing ammonium levels in the cell culture medium is promising. Further experimental validations have to be performed in order to ratify this hypothesis. Figure 3.4 Specific growth rate and growth yield values based on the in-house-based constraints for CHO non-producing cell line. Ensuing, myo-inositol was added to the optimized model constraints. In the predictions, myo-inositol was consumed by the cells and its uptake may hold a promise regarding the targeted improvements. CHAPTER 3 66 Finally, higher uptake rates of phosphate and hypoxanthine were observed comparing to the results based on Hefzi et al., 2016 predictions. A higher specific growth rate was observed using the optimized uptake rates, comparing to the values obtained using the constraints of Hefzi et al. , 2016 (figure 3.4). We were able to increase the growth rate up to 5 times comparing to the prediction results based on the constraints of Hefzi et al., 2016. However, since glucose uptake was also increased, the comparison of specific growth rates can be misleading. Therefore, growth yields on glucose were also compared, and a significant increase was also observed, around 3.3 times. Thus, in principle, higher uptake rates of certain amino acids and vitamins can hold a promise for improving the growth rate/yield. However, higher nutrients uptake has to account for transporters capacity within the cell. Using the in-house optimized uptake constraints of various metabolites, interesting results were observed regarding the secretion rate of toxic metabolites such as ammonium (Figure 3.5). Relying on the in-house optimized constraints, ammonium secretion was eliminated comparing to the predicted results of Hefzi et al., 2016. Figure 3.5 Metabolites secretion values for optimum uptake conditions. Comparison between the in-house prediction results and the results obtained using Hefzi et al 2016 constraints. H+ (hydrogen ion), CO2 (Carbon dioxide), H20 (Water). Looking at the in silico results in figure 3.5, ammonium, formate and urea were not secreted by the cells when using the in-house optimized constraints, when maximizing for growth. Serine and asparagine were present in the formulation described in Hefzi et al., 2016, which can explain the Neither Ammonium, nor Formate and Urea were produced using the in-house constraints CHAPTER 3 67 accumulation of these toxic metabolites as a result of amino acids breakdown. Formate secretion is directly linked to serine presence in the constraints of Hefzi et al., 2016 since formate is considered as a product of serine metabolism [48]. Finally, pyridoxamine and pyridoxine were secreted in our results as part of vitamin B6 metabolism. As a conclusion, we can affirm that the improvements made using Optflux based on the minimal components developed in-house were successful in improving the non-producing cell’s growth rate. The biomass yield was more than tripled comparing to the results described in previously published data. Another major improvement is that the toxic by-products such as ammonium were decreased to their lowest levels. In the next section, we will discuss the in silico efforts made to improve the growth parameters for CHO producer cell lines, as well as the improvements made to enhance recombinant proteins production. 3.3.2.2. Optimization results of CHO producing cell lines In order to predict the best medium components that can be used to improve the growth rate of CHO producing cell lines, the uptake fluxes were also optimized based on literature and also the minimal components predictions obtained using optiModels, previously described. The predictions were performed based on the biomass equation (R_biomass_cho_producing) specific for predicting growth rate of producer CHO cell lines. The prediction results account for the consumption and production rates of different metabolites, as well as the growth rate and yield values, specific for each condition tested. In this section, we were able to improve the growth rate by 5 times comparing to the growth value based on the use of Hefzi et al., constraints. Additionally, the growth yield value increased by 2 times. These results are described in figure 3.6. In this part of the work, not only improving the growth rate/yield and reducing cell culture by-products is targeted, but also improving the production of mAbs is a major ambition. The obtained results were used to determine the effect of flux constraints on the production of recombinant proteins in producer cell lines. Two different sets of constraints were obtained depending on the goal of the use of CHO producing cell lines. The first set of results is based on improving the growth rate and decreasing byproducts levels. The second set of constraints was based on improving the mAbs production with maintaining the cell culture by-products at their lowest levels. The reason behind generating two different sets of environmental constraints in this part of the study, is that the cells use their resources CHAPTER 3 68 (Carbon source, amino acids, etc. ) jointly for maximizing growth and maximizing the recombinant proteins production at the same time, as part of the mammalian growth-uncoupled phenomena, previously described in chapter 2. In this prediction studies we were able to improve the production rate of recombinant proteins comparing to different published data, described in Carinhas et al., 2013 and Selvarasu et al., 2012 [38,41]. In table 3.3 (Supplementary data), we can observe the lowest bounds (uptake) of several metabolites referred in the reference model (original model constraints) and also the in-house optimized constraints aiming at increasing growth rate/yield and mAbs production rate. Figure 3.6 Specific growth rate and growth yield values based on the in-house-based constraints for CHO producing cell line. Predictions in Optflux targeting the enhancement of cell growth rate and the decrease of cell culture toxic metabolites were performed using pFBA. For predicting maximum protein production, different solutions were obtained by maximizing the flux through the DM_igg[g] reaction, specific for the production of Immunoglobulin G (IgG) in the model. On the one hand, the same constraints used for predicting the optimal growth rate for CHO nonproducing cell lines, described in the previous section, were used in this study. Some minor changes were performed in order to reduce the predicted secretion rate of ammonium, since the metabolic behavior of the producing cells was different regarding the use/secretion of some amino acids, pyridoxal-5-phosphate (P5P) and other metabolites. For that, hypoxanthine and P5P were removed from the optimized uptake constraints of CHO producing cell lines. Hypoxanthine was linked to ammonium secretion in this part of predictions. Also, in this case, serine uptake showed a positive effect on CHAPTER 3 75 producing and producing CHO cells. 3.11/D and 3.11/E describe the different predicted secretion rates of key metabolites for CHO naïve and producer cells. Looking deeper into the literature, it was previously discussed that alpha-ketoglutarate supplementation can play a role in both decreasing ammonia levels in the culture and increasing productivity of the cells towards the production of recombinant proteins [53,54]. Additionally, cells supplemented with AKG exhibited lower growth rate comparing to glutamine-fed cells. This observation was highlighted in the work published by Tae Kwang Ha, Gyun Min Lee, 2014. In this study, the growth rate was recovered after several passages [54]. This strategy is further investigated along the thesis and described in the following chapters. According to the prediction results, it is relevant to understand experimentally the optimal levels of AKG to be added in the media. Being a very important intermediate in the TCA cycle, understanding AKG metabolism in both producer and non-producer cell lines, how does it impact the different metabolic pathways of the cell and what metabolic mechanisms are behind the boost in productivity will be targeted in the next chapter and consolidated with experimental data. 3.4. Conclusions and following work Evolutionary algorithms hold a potential in improving various industrial organisms through its combination with genome scale metabolic models. On this regard, several efforts have been performed to explore various optimization strategies using several microbial organisms. Nonetheless, few examples employed mammalian production platforms (e.g., CHO) due to their genetic and metabolic complexity. In this study, we were successful in using the GSMM of CHO combined with optiModels, a novel evolutionary algorithm, that aids in exploring CHO metabolic network and understanding how to develop an optimal medium formulation, that can sustain (high) growth of CHO. Then, higher production titers and/or productivity can be achieved. Higher growth yield (3.3 times) was observed when using the inhouse optimized constraints comparing to predictions using the original model constraints (Hefzi et al., 2016). Furthermore, based on FVA predictions, it was demonstrated that when using the optimized media formulation (based on the optimized in-house constraints), higher production rate of IgG was observed comparing to the conditions used for non/producing cells. In the second part of the study, we demonstrated that the supplementation of several candidates, determined using optiModels, might hold a potential in improving both growth and productivity of the CHAPTER 3 76 cells. Theoretically, AKG, as a very important intermediate in the TCA cycle, holds a huge potential in improving growth and productivity. Different uptake rates of AKG were tested in silico and the metabolic flux distribution was mapped based on the use of pFBA for both CHO producing and naive cells. Very interesting results were obtained regarding the supplementation of AKG to the medium, reflected by an increase in growth yield by 1.1 and 1.2 times, respectively, for naïve and producing CHO cells. In addition, a significant decrease in by-products secretion (Ammonia) was observed comparing to the results obtained using Hefzi et al. constraints. As conclusion, using the minimal media formulation described in this chapter as well as supplementing AKG might hold a tremendous potential in improving CHO bioprocesses. However, it should be pointed out that these results are merely an indication of a pathway forward in terms of experimental implementation. In fact, the results obtained using the GSMM do not take into consideration any kinetic limitations of the transporters and/or enzymes involved. This implies that the optimal uptake rates in silico might be not achievable in real conditions due to rate limitations. However, even if this is not fully possible, one can take the obtained results as leads for further metabolic or enzyme engineering approaches. The next steps will be focused on validating some of the in silico results experimentally. 3.5. Annexes A) OptiModels This tool was developed by Sara Correia and Sophia Santos at Minho University. The framework uses three other open-source Python frameworks, namely:  FRAMED (Framework for Metabolic Engineering and Design): A Python package for analysis and simulation of metabolic models that is used to load metabolic models from SBML files (https://github.com/cdanielmachado/framed).  Odespy: offers a unified interface to a large collection of software for solving systems of ordinary differential equations (ODEs) (https://github.com/hplgit/odespy).  Inspyred: an open-source framework for creating biologically-inspired computational intelligence algorithms in Python, including evolutionary computation (https://github.com/aarongarrett/inspyred). CHAPTER 3 77 This framework was implemented to be able to use parallel computing during the optimization tasks taking advantage of high-performance computing resources. Moreover, the user can specify the maximum time allowed for each simulation. The main entities involved in the simulation process are shown in figure 3.12. The framework files are divided into 4 main packages: Model, simulation, optimization and utils, and the complete description of each one of them is described below:  Model: This package contains functions to load and manipulate the models. Loading the used model in SBML format, is based on the methods and classes present in the FRAMED framework.  Simulation: This package contains the classes and functions used to simulate different types of models.  Optimization: This package contains all required entities to perform the targeted optimization based on evolutionary computation. The inspyred framework is used for creating biologically inspired computational intelligence algorithms in Python, including evolutionary computation and simulated annealing.  Utils: This tool holds a set of generic and auxiliary functions, constants and configurations used by the methods developed in the framework. Figure 3.12 OptiModels simulation workflow. CHAPTER 3 78 3.5.1. Model handling and simulation The data obtained from the SBML file is stored as an instance class according to its own model type. The CBModel and ODEModel classes are extensions of the generic class Model . All of these classes are present in the FRAMED framework. To simulate the problem, the different simulation problem classes have all the information required to perform a phenotype simulation. Different parameters are required for each simulation problem class, depending on the model type (stoichiometric, kinetic, etc.). All these classes extend the abstract class SimulationProblem , and implement the abstract methods get_model and simulate. Thus, the implementation of new types of simulation problems must extend the abstract class SimulationProblem and implement the abstract methods. This allows the usage of any simulation problem by the optimization layer since all the classes have the required methods implemented. Following, for the simulation results, the instances of these classes stores the results of phenotype predictions. Depending on the model type, different data should be saved. All classes must extend the abstract class SimulationResults and implement the method get_fluxes_distribution , which return the steady-state flux distribution of the phenotype simulation. The flux distribution values will be used by the objective function to calculate the fitness of each candidate in the optimization process. Meta-heuristics algorithms, including Evolutionary Algorithms and Simulated Annealing, are used by this framework to identify genetic modifications (strain design) and infer minimal medium composition that can improve production yields for relevant industrial compounds. Although these algorithms do not guarantee the convergence to global optima, they have the necessary flexibility, use lower computational power than exact solvers and also provide a family of optimal or sub-optimal solutions that can be further studied in order to determine the optimal one. Figure 3.13 depicts the workflow of the optimization process using Evolutionary Algorithms and the important points of the implementation are explained in the following. CHAPTER 3 79 Figure 3.13 OptiModels optimization pipeline.  Candidate representation In this framework there are two types of candidate representations: - Set representation: this kind of representation can be used to simulate gene/reaction knockouts and infer minimal medium composition. In this case, each candidate element represents the index of the gene/reaction that will be knocked-out in the phenotype simulation. Figure 3.14 Candidates representation – Set representation. - Set of tuples: this representation is used to perform the simulation of under/over expression. Each element of the candidate solution is a tuple of 2 integers. The first identifies the reaction to manipulate and the second the level of expression. CHAPTER 3 80 Figure 3.15 Candidate representation - Set of tuples. These representations are the result of the functions generator_intSetRep and generator_intTupleRep used in the optimization workflow to generate new candidates for the population.  Operators For reproduction purposes within the Evolutionary Algorithm, the following operators have been implemented: - Mutation operations o Grow : insert a new element (integer/tuple) in the candidate solution. o Shrink : remove an element from the candidate solution. o Replace : replace one element for a new one randomly generated. - Crossover: using two candidates (parents) build 2 children: o Elements present in both parents will be present in both children. o Elements present in only one parent have equal probability to be present in child 1 or child 2. o note : children can be equal to the parents  Objective Functions The objective function has the role of evaluating each candidate of the population calculating the corresponding fitness value. The most common objective function is the flux value of a target compound, implemented in the framework as targetFlux class. All objective functions on the framework must extend the abstract class objectiveFunction and implement the methods: - get_name: returns a string with the method name. CHAPTER 3 81 - method_str: returns a string with the method formulation. - get_fitness: returns the fitness value considering the simulation result given as argument.  Decoders The decoders are responsible to convert a candidate representation into an OverrideModel which contains the modification that will be used over the simulation problem in the phenotype simulation. As example, the candidate represented by a set of integers (1,3,5) given as argument to the method get_override_simul_problem of the decoderReactionsKnockouts class, will retrieve a list of modifications that must be imposed to the simulation problem, in order to knockout the reactions represented by 1, 3 and 5 indexes.  Applications o Strain design Strain design through reaction knockouts or under/over expressed enzymes is implemented in our framework for single and multi-organism models. The approach used is the same for the two cases. The solution candidates can be represented as a set representation or as set of tuples, as described above. The operators (mutation and crossover) used in EA are the ones described previously. Two evaluation functions were implemented in the framework under the scope of strain design: - TargetFlux: The fitness is given by the flux value of the target reaction. - BPCY: "Biomass-Product Coupled Yield" objective function (Patil et al ., 2005). The fitness is given by the equation: 𝑓𝑖𝑡𝑛𝑒𝑠𝑠 =𝑏𝑖𝑜𝑚𝑎𝑠𝑠 𝑓𝑙𝑢𝑥 × 𝑝𝑟𝑜𝑑𝑢𝑐𝑡 𝑓𝑙𝑢𝑥 𝑢𝑝𝑡𝑎𝑘𝑒 𝑓𝑙𝑢𝑥 o Minimal Medium optimization The goal of minimal medium optimization is to find the best medium composition for a given objective function, such as growth or the production of a target compound. EA are used by our framework to identify the smallest set of uptake compounds that can improve a given objective function for single and multi-organism models. The solution candidates can be represented as a set, where each element represents an uptake exchange reaction. The operators (mutation and crossover) used in EA are the ones described previously. CHAPTER 3 82 For medium optimization purposes, two evaluation functions are available in the framework: - BP_MinModifications: this evaluation function is based on the "Biomass-Product Coupled Yield" objective function (Patil et al ., 2005) but considering the candidate size. The fitness is given by the equation: - 𝑓𝑖𝑡𝑛𝑒𝑠𝑠 =𝑏𝑖𝑜𝑚𝑎𝑠𝑠 𝑓𝑙𝑢𝑥 × 𝑝𝑟𝑜𝑑𝑢𝑐𝑡 𝑓𝑙𝑢𝑥 𝑢𝑝𝑡𝑎𝑘𝑒 𝑓𝑙𝑢𝑥 - MinNumberReac: this function returns a fitness value between 0 and 1. Higher fitness values correspond to candidates with a smaller size. 𝑓𝑖𝑡𝑛𝑒𝑠𝑠 = 𝑠𝑖𝑧𝑒(𝑐𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒) 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑜𝑓 𝑐𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒 𝑠𝑖𝑧𝑒 The maximum of candidate size is the maximum number of uptake reactions allowed by user. By default, this value is the number of exchange reactions present in the given model. B) Additional Material Table 3.2 Original model constraints vs in-house optimized constraints for non-producer cell lines. Table 3.3 Comparison between the constraints used for predictions using producer cell lines. Reaction ID Metabolite Hefzi et al 2016 In-House R_EX_glc__D_e D-Glucose exchange -0.19835 -0.3 R_EX_his__L_e L-Histidine exchange -0.00330 -0.2 R_EX_trp__L_e L-Tryptophan exchange -0.00408 -0.00408 R_EX_cys__L_e L-Cysteine exchange -0.00522 -0.05 R_EX_met__L_e L-Methionine exchange -0.00604 -0.02 R_EX_phe__L_e L-Phenylalanine exchange -0.00604 -0.1 R_EX_pro__L_e L-Proline exchange -0.00797 -0.7 R_EX_asp__L_e L-Aspartate exchange -0.00934 -0.2 R_EX_tyr__L_e L-Tyrosine exchange -0.00934 -0.05 R_EX_ile__L_e L-Isoleucine exchange -0.01016 -0.05 R_EX_thr__L_e L-Threonine exchange -0.01016 -0.2 R_EX_val__L_e L-Valine exchange -0.01209 -0.1 R_EX_lys__L_e L-Lysine exchange -0.01346 -0.08 R_EX_leu__L_e L-Leucine exchange -0.01484 -0.3 R_EX_arg__L_e L-Arginine exchange -0.01978 -0.1 R_EX_asn__L_e L-Asparagine exchange -0.04038 0 R_EX_ser__L_e L-Serine exchange -0.04780 0 R_EX_gln__L_e L-Glutamine exchange -0.06703 0 R_EX_chol_e Choline exchange -0.02029 -0.05 R_EX_pydxn_e Pyridoxine exchange -0.00017 -0.00017 R_EX_fol_e EX fol e -0.00046 -0.00046 R_EX_pydx_e Pyridoxal exchange 0 -1 R_EX_hxan_e Hypoxanthine exchange -0.00619 -0.1 R_EX_inost_e Myo-Inositol exchange 0 -0.001 R_EX_o2_e O2 exchange -1.12747 -1.12747 R_EX_so4_e Sulfate exchange -1000 -1000 R_EX_h2o_e H2O exchange -1000 -1000 R_EX_pi_e Phosphate exchange -1000 -0.5 R_SK_pre_prot_r Sink pre prot LPAREN er RPAREN -1000 -1000 R_EX_h_e H+ exchange -1000 -1000 R_SK_Ser_Thr_g Sink Ser/Thr[g] -0.1 -0.1 R_SK_Tyr_ggn_c Sink Tyr ggn -0.1 -0.1 R_SK_Asn_X_Ser_Thr_r Sink Asn X Ser/Thr[r] -0.1 -0.1 R_EX_fe2_e Fe2+ exchange -1 -1 R_EX_hco3_e EX hco3 LPAREN e RPAREN -1 -1 CHAPTER 3 83 3.6. References [1] B.O. Palsson, Systems biology: Constraint-based reconstruction and analysis, 2015. https://doi.org/10.1017/CBO9781139854610. [2] G. Stephanopoulos, A. Aristidoi, J. Nielsen, Metabolic engineering: Principles and Methodologies, 1998. [3] D. Machado, M.J. Herrgård, I. Rocha, Stoichiometric Representation of Gene–Protein–Reaction Associations Leverages Constraint-Based Analysis from Reaction to Gene-Level Phenotype Prediction, PLoS Comput. Biol. (2016). https://doi.org/10.1371/journal.pcbi.1005140. [4] A.M. Feist, B. Palsson, The growing scope of applications of genome-scale metabolic reconstructions using Escherichia coli , Nat. Biotechnol. (2008). https://doi.org/10.1038/nbt1401. [5] C. Gu, G.B. Kim, W.J. Kim, H.U. Kim, S.Y. Lee, Current status and applications of genome-scale metabolic models, Genome Biol. (2019). https://doi.org/10.1186/s13059-019-1730-3. [6] J. Stelling, Mathematical models in microbial systems biology, Curr. Opin. Microbiol. (2004). https://doi.org/10.1016/j.mib.2004.08.004. [7] D. Solle, B. Hitzmann, C. Herwig, M. Pereira Remelhe, S. Ulonska, L. Wuerth, A. Prata, T. Reaction ID Metabolite Hefzi et al 2016 Growth rate optimization IgG maximization R_EX_glc__D_e D-Glucose exchange -0.19835 -0.3 -0.3 R_EX_his__L_e L-Histidine exchange -0.00330 -0.2 -0.2 R_EX_trp__L_e L-Tryptophan exchange -0.00408 -0.00408 -0.00408 R_EX_cys__L_e L-Cysteine exchange -0.00522 -0.05 -0.05 R_EX_met__L_e L-Methionine exchange -0.00604 -0.02 -0.00604 R_EX_phe__L_e L-Phenylalanine exchange -0.00604 -0.1 -0.1 R_EX_pro__L_e L-Proline exchange -0.00797 -0.7 -0.7 R_EX_asp__L_e L-Aspartate exchange -0.00934 -0.2 -0.2 R_EX_tyr__L_e L-Tyrosine exchange -0.00934 -0.05 -0.05 R_EX_ile__L_e L-Isoleucine exchange -0.01016 -0.05 -0.05 R_EX_thr__L_e L-Threonine exchange -0.01016 -0.2 -0.2 R_EX_val__L_e L-Valine exchange -0.01209 -0.1 -0.1 R_EX_lys__L_e L-Lysine exchange -0.01346 -0.08 -0.08 R_EX_leu__L_e L-Leucine exchange -0.01484 -0.3 -0.3 R_EX_arg__L_e L-Arginine exchange -0.01978 -0.01978 -0.01978 R_EX_asn__L_e L-Asparagine exchange -0.04038 0 0 R_EX_ser__L_e L-Serine exchange -0.04780 -0.1 -0.04780 R_EX_gln__L_e L-Glutamine exchange -0.06703 0 0 R_EX_chol_e Choline exchange -0.02029 -0.05 -0.05 R_EX_pydxn_e Pyridoxine exchange -0.00017 -0.00017 -0.00017 R_EX_fol_e EX fol e -0.00046 -0.00046 -0.00046 R_EX_pydx_e Pyridoxal exchange 0 0 0 R_EX_hxan_e Hypoxanthine exchange -0.00619 0.00000 0 R_EX_inost_e Myo-Inositol exchange 0 -0.001 -0.001 R_EX_o2_e O2 exchange -1.12747 -1.12747 -1.12747 R_EX_so4_e Sulfate exchange -1000 -1000 -1000 R_EX_h2o_e H2O exchange -1000 -1000 -1000 R_EX_pi_e Phosphate exchange -1000 -1000 -1000 R_SK_pre_prot_r Sink pre prot LPAREN er RPAREN -1000 -1000 -1000 R_EX_h_e H+ exchange -1000 -1000 -1000 R_SK_Ser_Thr_g Sink Ser/Thr[g] -0.1 -0.1 -0.1 R_SK_Tyr_ggn_c Sink Tyr ggn -0.1 -0.1 -0.1 R_SK_Asn_X_Ser_Thr_r Sink Asn X Ser/Thr[r] -0.1 -0.1 -0.1 R_EX_fe2_e Fe2+ exchange -1 -1 -1 R_EX_hco3_e EX hco3 LPAREN e RPAREN -1 -1 -1 CHAPTER 3 84 Steckenreiter, Between the Poles of Data-Driven and Mechanistic Modeling for Process Operation, Chemie-Ingenieur-Technik. (2017). https://doi.org/10.1002/cite.201600175. [8] K. Raman, N. Chandra, Flux balance analysis of biological systems: Applications and challenges, Brief. Bioinform. (2009). https://doi.org/10.1093/bib/bbp011. [9] S. Gudmundsson, I. Thiele, Computationally efficient flux variability analysis, BMC Bioinformatics. (2010). https://doi.org/10.1186/1471-2105-11-489. [10] R. Mahadevan, C.H. Schilling, The effects of alternate optimal solutions in constraint-based genome-scale metabolic models, Metab. Eng. (2003). https://doi.org/10.1016/j.ymben.2003.09.002. [11] J.L. Reed, B. Palsson, Genome-scale in silico models of E. coli have multiple equivalent phenotypic states: Assessment of correlated reaction subsets that comprise network states, Genome Res. (2004). https://doi.org/10.1101/gr.2546004. [12] I. Thiele, R.M.T. Fleming, A. Bordbar, J. Schellenberger, B. Palsson, Functional characterization of alternate optimal solutions of Escherichia coli ’s transcriptional and translational machinery, Biophys. J. (2010). https://doi.org/10.1016/j.bpj.2010.01.060. [13] J.M. Gutierrez, A. Feizi, S. Li, T.B. Kallehauge, H. Hefzi, L.M. Grav, D. Ley, D. Baycin Hizal, M.J. Betenbaugh, B. Voldborg, H. Faustrup Kildegaard, G. Min Lee, B.O. Palsson, J. Nielsen, N.E. Lewis, Genome-scale reconstructions of the mammalian secretory pathway predict metabolic costs and limitations of protein secretion, Nat. Commun. (2020). https://doi.org/10.1038/s41467-019-13867-y. [14] T.Y. Kim, S.B. Sohn, Y. Bin Kim, W.J. Kim, S.Y. Lee, Recent advances in reconstruction and applications of genome-scale metabolic models, Curr. Opin. Biotechnol. (2012). https://doi.org/10.1016/j.copbio.2011.10.007. [15] M.R. Long, W.K. Ong, J.L. Reed, Computational methods in metabolic engineering for strain design, Curr. Opin. Biotechnol. (2015). https://doi.org/10.1016/j.copbio.2014.12.019. [16] I. Rocha, P. Maia, P. Evangelista, P. Vilaça, S. Soares, J.P. Pinto, J. Nielsen, K.R. Patil, E.C. Ferreira, M. Rocha, OptFlux: An open-source software platform for in silico metabolic engineering, BMC Syst. Biol. (2010). https://doi.org/10.1186/1752-0509-4-45. [17] A.P. Burgard, P. Pharkya, C.D. Maranas, OptKnock: A Bilevel Programming Framework for CHAPTER 4 91 (e.g., producing oxaloacetate and α-ketoglutarate) and also in energy formation [15]. However, its degradation yields high levels of free ammonium in the culture, which can heavily impact bioprocesses. Beyond that, glutamine instability in liquid media is very well known and it poses several issues for media manufacturers. Supplementing glutamine to the chemically defined media formulations can radically decrease the shelf life of the produced media [16]. Therefore, finding an alternative to glutamine in the cell culture media can definitely mitigate several bioprocessing problems mainly regarding the media validity and most importantly, ammonia accumulation. Moreover, not only glutamine degradation is a source of high ammonium concentrations, but also other amino acids, for instance, asparagine and serine are associated with this phenomenon [8]. It is very important to highlight that different amino acids in the culture medium are consumed at different rates [17]. These values depend on the CHO platform in use, nutritional needs, culture condition, medium composition, the produced recombinant protein [13] and most importantly the transporters capacity within the cell, towards the exchange of metabolites between the extracellular and the intracellular environment [18]. In order to overcome the overproduction of ammonium and lactate and optimize the culture parameters, several approaches are being employed but few were able to fulfill the desired results [19,20]. One of the most important approaches for optimizing bioprocesses is focused on media optimization. Previously, methodologies were based on using conventional practices, for instance mixing different media formulations or titration of the existing components, which, experimentally represents a hurdle [21]. In fact, blending different media formulations is commonly used. This procedure relies mostly on the use of design of experiment (DoE) strategies that can help depicting the optimal media formulation among the tested ones. This procedure is efficient; however, the chemical composition of the resulting optimal media formulation selected remains unknown. Modern media optimization strategies are focusing mainly on statistical design and on the use of high-throughput screening of different potential media candidates, typically based upon scaling down cell culture models [22], where erlenmeyers and also spin tubes [23] are commonly used as small scale cell culture methods. Alongside, the emergence of novel high throughput automated robots, for instance ambr® microbioreactors and its combination with powerful DoE approaches is a robust tool for rapid and efficient process and media screening and optimization [24]. These approaches hold several advantages, for instance, accelerating the optimization timeline. CHAPTER 4 92 On the other hand, cell line development strategies have been employed to construct highly characterized and efficient stable cell lines for recombinant proteins production [25]. For large scale manufacturing of recombinant biotherapeutics using CHO platforms, several optimized CHO strains have evolved. These platforms are known for increased production titers comparing to the standard CHO cells commonly in use. Among these newly established cell lines, dihydrofolate reductase DHFR (-) and glutamine synthetase GS (-) cell lines are the most known [26]. Using the GS-knockout CHO cell line is very beneficial when it comes to overcoming the accumulation of ammonia in culture. These cells can grow independently of glutamine presence in the medium [26,27], considering that glutamine synthase catalyzes the consumption of glutamate to biosynthesize glutamine in the cell [28,29]. A strategy built on re-channeling ammonium through several pathways within the cell can also be possible, even knowing that the CHO-GS metabolism is poorly known [17]. Indeed, metabolic engineering strategies based on directing the precursors of these byproducts inside the cell through other pathways combined with optimizing flux distributions is a very promising approach, taking advantage of the versatility of the cell’s metabolic network [30]. For that, balancing the levels of metabolites in the culture and controlling their exchange rates is an interesting approach, particularly emphasizing on medium optimization and customizing it to the cells exact nutritional need. In this case, the cells will evolve a more effective metabolism, by reusing the accumulated metabolites to fuel the cells with more energy and/or important metabolites that can improve both growth and quality attributes of products. α-ketoglutarate (AKG) or also called 2-oxoglutarate is a very important TCA cycle intermediate and a crucial compound for cell's metabolism [31]. It is a precursor in amino acids biosynthesis, also involved in signaling processes in the cell, ATP production, generation of reducing equivalents (NAD+/NADH), also playing a role in regulating various epigenetic mechanisms in the mammalian cell [32]. During cell culture, α-ketoglutarate is a potential additive for cell culture medium, since it can mutually replenish the TCA cycle and trigger its different metabolic intermediates to produce energy and replace glutamine, producing cellular glutamate by glutamate dehydrogenase (GDH). In theory, accumulated glutamate can play a role in the de novo biosynthesis of cellular glutamine by glutamine synthase, using the free ammonium in the culture [28,29]. As well, α-ketoglutarate acts as an antioxidant instead of glutamine in many cellular processes and besides, it is more affordable and chemically more stable (in culture and storage) comparing to glutamine [33]. Previous reports studied the effect of supplementing AKG and different TCA cycle CHAPTER 4 93 intermediates, as a replacement of glutamine on CHO cells. In one study, they used CHO-DG44 DHFR (-) cell line where AKG was supplemented at a concentration of 4 mM. In this study, the authors stated the advantages of supplementing AKG to culture. Among these, an increase in productivity as well as a significant decrease in ammonia accumulation in the media were observed [34]. Following this idea, it is not yet clear how the substitution of glutamine by α-ketoglutarate modifies the metabolic network of CHO cells. It is a fact that supplementing α-ketoglutarate to the culture is not enough to channel most of the by-products in the cell, but concurrently balancing the amino acids levels inside of the culture medium, in the presence of α-ketoglutarate, might restructure the cells metabolism and overcome the accumulation of toxic metabolites that can hinder both growth and productivity. This hypothesis drove us to think about engineering strategies based on balancing the levels of nutrients in the medium in order to sustain and rewire the metabolism towards optimality. For that, balancing the amino acids levels together with testing different AKG concentration on CHO cells is very interesting, especially tackling GSneg CHO cell lines, since they hold a huge potential in modern bioproduction processes. It is a hurdle to rely just on experimental setup for optimization. As a solution, metabolic modeling approaches comes to support faster and accurate predictions towards optimizing the culture parameters by studying biochemical pathways in silico and applying the prediction results to be tested experimentally. Based on the results presented in chapter 3, predictions based on the use of GSMM of CHO and the evolutionary algorithm (optiModels) drove us to think about validating those hypotheses experimentally. 4.2. Materials and Methods 4.2.1. Experimental setup 4.2.1.1. Cell culture Three different CHO cell lines (two producer and one non-producer) were used in this work. First, CHOK1 (ECACC 85051005) was employed in this study representing a naïve cell line, non-modified to produce recombinant proteins. Second, CHO-EpoFc represents a CHO DHFR (-) strain, a producer cell line engineered to produce EpoFc fragments (one molecule of erythropoietin joined to each hinge region of huIgG1Fc). CHO-EpoFc is known as a low producer that was established following the protocol previously described in Lattenmayer et al 2007 [35]. These cells were adapted internally to growth in serum-free and glutamine-free medium in the laboratory, prior to the use in this study. Finally, CHO-HyC CHAPTER 4 94 cells correspond to an antibody expressing CHO cell line provided by Cytiva, Uppsala, Sweden. The latter is GSneg cell line and known as a high-producer industrial clone, producing Trastuzumab, a monoclonal antibody under the commercial name of Herceptin. All the CHO clones were cultivated in suspension mode in chemically defined serum-free conditions using CD CHO medium (Gibco, Invitrogen, Carlsbad, CA, USA). After thawing, the cells were routinely cultivated in 50 mL TPP® TubeSpin bioreactors (Techno Plastic Products AG, Trasadingen, Switzerland) at a maximal working volume of 25 mL. The cells were incubated in 37°C in 80 % humidified air with 7 % CO2, shaking at a speed of 220 rpm (rotation per minute). The cells were passaged every 3-4 days and the viable cell concentrations, viabilities and the values of the average cell diameters were determined using Vi-CELLXR (Beckman Coulter, USA). In standard conditions, CHO-K1 cells were grown in glutamine-free CD CHO medium (Gibco TM, MA, USA) and supplemented with 0.2% anti-clumping agent (ACA) (Thermo Fisher Scientific). CHO-EpoFc cells were grown in glutamine-free CD CHO medium and supplemented with 0.096 μM methotrexate (MTX) (Thermo Fisher Scientific). CHO-HyC cells were originally cultivated in CD CHO medium supplemented with 8 mM glutamine, 75 µM of L-Methionine sulfoximine (MSX) (Thermo Fisher Scientific) and 0.2 % ACA (Thermo Fisher Scientific). An adaptation process to glutamine-free conditions was performed and consists of a sequential adaptation of the cells to different ratios of the CD CHO media containing 8mM glutamine and CD CHO glutamine-free medium. The sequential adaptation process is described in table 4.1. Table 4.1 Sequential adaptation of CHO-HyC cells to glutamine-free conditions. Adaptation step Ratio of CD CHO medium containing 8mM glutamine to CD CHO glutamine-free media Criteria to fulfill A 75:25 Viability ≥90% and normal doubling time for 2 passages B 50:50 Viability ≥90% and normal doubling time for 2 passages D 0:100 Viability above 90% of cells grown in glutamine-free medium and doubling time for 2 passages CHAPTER 4 95 As a result of this adaptation strategy, the cells were grown in glutamine-free media. A working cell bank was stored for further experiments. In all the experiments the cells were grown in glutamine-free conditions using CD CHO cell culture media. These cells were supplemented with different concentrations of AKG (4mM, 8mM and 12mM). For the experimental setup, batch cultures were performed in 125 mL non-baffled Erlenmeyer shake flasks at a working volume of 50 mL, incubated in 37°C in 80 % humidified air with 7 % CO2, shaking at a speed of 140 rpm. All the experiments were performed in triplicates, inoculated at the beginning of the experiment, at the same time and cell concentration. Samples were taken every 24h. The culture continued until reaching a cell viability lower than 60 %. 4.2.1.2. Extracellular Metabolites For the analysis of the extracellular metabolites, samples were collected every 24h along the culture period. The cells were removed by centrifugation 10 min at 200 rcf (relative centrifugal force) and the supernatant was stored at -20°C. Metabolomics analyses were performed shortly after sampling, analyzing glucose, lactate and ammonia using Bioprofile 100 Plus (Nova Biomedical, MA, USA). Amino acids levels were quantified using HPLC with fluorescence detection (Dionex 3000 HPLC, Thermo Fisher Scientific, Waltham, Massachusetts, US). The HPLC was equipped with an AdvanceBio AAA column (4.6 x 100 mm, 2.7 µm, Agilent Technologies, Santa Clara, CA, USA) and a pre-column UPLC guard column, AdvanceBio AAA (4.6 x 5 mm, 2.7 µm, Agilent Technologies, Santa Clara, CA, USA). The column temperature was set to 37°C. The mobile phase consisted of A: 40 mM Na2 HPO4 in 0.02 % NaN3 and B: ACN/MeOH/H2O (45:45:10) (v/v). O-phthalaldehyde (OPA)-derivatized amino acids were detected at 340ex and 450em nm and 9-fluorenylmethyloxycarbonyl (FMOC)-derivatized amino acids at 266ex and 305em nm [13]. Data were processed by Chromeleon software (Thermo Fisher Scientific, Waltham, MA, US). Cysteine could not be quantified due to sensitivity issues in the method, so the results are only qualitative. 4.2.1.3. α-ketoglutarate quantification α-ketoglutarate levels were quantified using a colorimetric analytical method, following the manufacturer protocol. The quantification kit was purchased from Sigma Aldrich (MAK054). CHAPTER 4 96 The supernatant samples were deproteinized using 10 kDa MWCO Amicon Ultra-0.5 centrifugal filter units (Merck Millipore, MA, USA) and diluted using the assay buffer of the kit. Several dilutions were performed in order to fit our samples concentrations into the standard curve, not forgetting the negative control samples without adding AKG converting enzyme into the reaction mix. The AKG standard curve was established with concentrations ranging from 0 to 10 nmole/well. Every reaction contained 50 µL of reaction/well. The plates were incubated for 30 min at 37ºC with a shaking speed of 330 rpm to homogenize the reaction mix during the incubation time. Ensuing, the absorbance was measured at 570 nm. 4.2.1.4. Product quantification Product concentration was determined using Octet® QKe (Port Washington, NY), equipped with Dip and ReadTM Protein A Biosensors (Pall corporation, Port Washington, NY) according to the manufacturer’s recommendations. The supernatant was diluted with CD CHO medium prior to measurements in order to fit the samples concentrations within the standard curve ranging between 0100 µg/mL of Trastuzumab (BioVision, Milpitas, CA). A negative control consisting of cell culture medium was included. This method is based on biomolecular interactions, measuring the binding intensity of our product of interest to an immobilized ligand. 4.2.1.5. α-ketoglutarate toxicity assay To evaluate the toxicity of supplementing different concentrations of α-ketoglutarate on CHO cells, Cell Titer 96 AQueousOne Solution Cell Proliferation Assay (Promega, Madison, WI, USA) was used. The latter is a colorimetric assay, used to determine the number of viable cells for cytotoxicity assays. It is based on the use of tetrazolium compound called [3-(4,5-dimethylthiazol-2-yl)-5-(3carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt) or simply MTS, which is converted by the cells to generate formazan. This process is mediated by NADPH or NADH produced by dehydrogenase enzymes in viable cells. The levels of formazan determined by absorbance at 490 nm are directly proportional to the number of viable cells in culture. In our case, 100 µL of CHO-HyC cells, at a concentration of 2x105 cells/mL, were inoculated in coated 96 well plate for cell culture. The cells were supplemented with 10 µL of AKG at different concentrations, ranging between 20mM and 100mM. The cells were incubated at 37ºC for 48 h. Ensuing, Cell Titer 96 reagent was added to the samples, and the plate was incubated 1h at 37ºC with CHAPTER 4 97 a shaking speed of 130 rpm. Following, the absorbance was measured at 490 nm in order to evaluate the level of formed formazan and assess the number of viable cells after incubation with AKG. 4.2.2. Culture characterization Growth data were determined based on Vi-CELL XR data and specific growth rates values were calculated as a function of time according to the following equation, knowing that X is the viable cell concentration at a specific time point ( t ), X0 is the initial viable cell concentration and μ represents the cell growth rate. 𝑋 =𝑋0𝑒𝜇𝑡 The viable cell concentration described as viable cell density (VCD) given in viable cells/mL was measured using Vi-CELL XR. The viable cell volume VCV was calculated as follows. First, the values of the volume per cell described as (µm3/cell) were calculated according to equation 1 (Eq 1). 𝑬𝒒 (𝟏): 𝑉𝑜𝑙𝑢𝑚𝑒 𝑝𝑒𝑟 𝑐𝑒𝑙𝑙= 4 3 𝜋 (𝑉𝑖𝑎𝑏𝑙𝑒 𝑐𝑒𝑙𝑙 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 2) 3 The values determined in Eq 1 were used to calculate the VCV values as follow in Eq (2), using the diameters obtained also from Vi-CELL XR data. The VCV values are given in (mm3/mL). 𝑬𝒒 (𝟐): 𝑉𝐶𝑉 =𝑉𝑜𝑙𝑢𝑚𝑒 𝑝𝑒𝑟 𝑐𝑒𝑙𝑙.10−9.𝑉𝐶𝐷 Pearson’s correlation coefficients were determined for linear correlations between Ln-transformed VCD and the culture time, starting from the first time point analyzed (TP00) and including at least 5 time points. For each sample the highest correlation coefficient (rMAX) and the time point (TPXY) of its occurrence were determined. The growth rates were calculated as slopes in simple linear regressions of the ln-transformed VCD (or VCV) versus the interval (TP00–TPXY). Following, cumulative viable cell days (CCDCD), (CCDCV) were calculated based on different values of VCD and VCV respectively and described as (cells*days). This method was previously described at Klanert et al., 2019 [36] and adopted in this study. The different CCDCD and CCDCV values were determined based on the following equations where t represents the hours post-inoculation, and n the number of time points analyzed per batch. CHAPTER 4 98 𝐄𝐪 (𝟑): CCD𝐶𝐷 = ∑ 𝑛−1 𝑖=1 (𝑉𝐶𝐷𝑖+1 −𝑉𝐶𝐷𝑖).(𝑡𝑖+1 −𝑡𝑖 ) (ln(𝑉𝐶𝐷𝑖+1 )−ln(𝑉𝐶𝐷𝑖)).24 𝐄𝐪 (𝟒): CCD𝐶𝑉 = ∑ 𝑛−1 𝑖=1 (𝑉𝐶𝑉𝑖+1 −𝑉𝐶𝑉𝑖).(𝑡𝑖+1 −𝑡𝑖 ).10−3 (ln(𝑉𝐶𝑉𝑖+1 .10−3)−ln(𝑉𝐶𝑉𝑖.10−3)).24 Pearson’s correlation coefficients were determined for linear correlations between CCDCD and CCDCV and the product titers starting from the second measurement (TP02) (Taking into consideration that TP00 is the first measurement of time of inoculation) and including at least 6 time points. For each sample the highest correlation coefficient (rMAX) and the time point of its occurrence (TPXY) were determined. The specific productivities of product (qP) were calculated as slopes in simple linear regressions of the CCDCD (or CCDCV) versus the titers for the interval (TP02 – TPXY), representing the exponential phase of the cultures. 4.2.3. Mathematical fitting of exchange rates of metabolites The calculated specific growth rates and the initial cell concentrations of different experiments were used to calculate the exchange rates of different metabolites, for instance, glucose, lactate, ammonia and other amino acids, relying on the following equation (Eq (5)) described in Széliová et al., 2020 [13]: 𝐄𝐪 (𝟓): [𝑖]=[𝑖]0+ 𝑞𝑀𝐵0 µ(𝑒µ𝑡-1) where [i]0 and [i] are the concentration of metabolite i at the beginning and during the exponential phase, respectively, qM is the specific uptake or secretion rate of metabolites, and B0 is the initial amount of biomass, calculated from the initial cell concentration X0 and the data of dry mass per cell. The latter was obtained internally in the lab and partially adapted from Széliová et al., 2020 [13]. This equation was used to estimate the exchange rates of metabolites in (mmol/gDW/h), using non-linear regression function. 4.3. Results and discussion As previously mentioned in the materials and methods section, three different cell lines were cultivated in CD CHO medium without glutamine supplemented with ACA. Different concentrations of αketoglutarate (4mM, 8mM and 12mM) were added in order to study the effect of α-ketoglutarate (AKG) CHAPTER 4 99 on growth and also its impact on the production of EpoFC recombinant proteins and Trastuzumab. Negative control batches also were performed, where the cells were not supplemented with αketoglutarate but just grown with CD CHO medium without glutamine. Cell density, viability and metabolic profiles of the cells were monitored over time and compared between the different conditions. 4.3.1. Growth characteristics Throughout the results and looking at figure 4.1, we can observe the growth profiles and the viability trends of CHO cell lines tested in this study. A clear difference in growth profiles was observed among the tested conditions, employing both producer and non-producer CHO cell lines. These results were mainly focused on comparing the batch periods, the viability profiles, as well as the maximal cell densities reached in each tested condition when varying the supplemented AKG concentrations. Higher specific growth rate values and maximal cell densities were observed in cultures non-treated with AKG comparing to the treated cells, noticing also a decrease in CHO growth rate when increasing the AKG concentration in the medium. In contrast to these observations, the viability/life span of the cells was extended in all clones treated with AKG comparing to the non-treated cells (Figure 4.1 B/C/D). It is clear that AKG supplementation is associated with increasing the viability of the cells and prolonging the stationary phase of the culture, which might be interesting from a bioprocessing standpoint. In fact, maintaining a high viability of the cells over time is a standard approach to increase the volumetric productivity of the cells, previously highlighted in literature [37]. Furthermore, several studies focused on controlling cells proliferation to improve mammalian cells productivity were previously discussed [38]. Within these approaches, medium optimization and metabolic rewiring of the cells holds a promise in controlling cell’s proliferation and consequently improving the product’s final titer [39,40]. In fact, the increase of production of recombinant proteins in stationary phase was previously observed [41]. To achieve this behavior, several approaches were already used to control the proliferation of the cells as mean of increasing its productivity in bioprocesses [42]. Strategies based on culturing the cells at lower temperatures (e.g., 32º C), chemical treatments (e.g., sodium butyrate) or also cell cycle arrest strategies, were previously described in the literature [43–46]. Ensuing, in figure 4.2, we can observe a summary describing the different values of the specific growth rate of the 3 different CHO cell lines tested in this study. According to these results, no significant differences in growth rate values were observed in non-treated cells and when treated with 4mM AKG. CHAPTER 4 100 When adding higher concentrations (8mM and 12mM of AKG) a more important difference in growth rate was observed among these cell types. These experimental results are comparable to the data previously published in Kwang Ha et al., 2014, where they tested different TCA cycle intermediates that can substitute glutamine in the cell culture medium. They also observed a clear decrease in growth rate of CHO cells with AKG supplementation. The latter was recovered after various passages [34]. Figure 4.1 Growth profiles of three different CHO cell lines used in the study. (A), (C) and (E) represent respectively the growth profiles of CHO-K1, CHO-HyC and CHO-EpoFc cell lines non supplemented with AKG (Purple squares), supplemented with 4mM AKG (Blue inclined squares), supplemented with 8mM AKG (Black circles) and supplemented with 12mM AKG (Green triangles). (B), (D) and (F) represent respectively the viability of the cells along the cultures of CHO-K1, CHO-HyC and CHO-EpoFc cell lines non supplemented with AKG (Purple squares), supplemented with 4mM AKG (Blue inclined squares), supplemented with 8mM AKG (Black circles) and supplemented with 12mM AKG (Green triangles). The error bars represent the standard deviation between replicates. CHAPTER 4 107 levels comparing to the treated cells together with a slight consumption of lactate when glucose is drained. Figure 4.6 Metabolic profiles of CHO-HyC cultures treated or non-treated with AKG. (A) represents the glucose profile over time, (B) represents lactate profile over time, (C) represents glutamate profile over time, (D) represents ammonium profile over time, (E) represents AKG profile over time and (F) represents aspartate profile over time. In fact, the high level of AKG in the medium replenishes the TCA cycle of CHO and in parallel seems to trigger the glycolytic pathway. Concerning the metabolic switch that forces lactate consumption, in some cases, this phenomenon can be triggered by glucose depletion in the medium, change in pH of the culture or also the difference in lactate levels in intra and extra cellular environment [47]. In fact, in this process, lactate is oxidized to pyruvate, and can take part of 4 metabolic pathways, being converted CHAPTER 4 108 to acetyl-CoA, oxaloacetate, malate or alanine, generating NADH that can play a role in the oxidative phosphorylation mechanism leading to ATP production [47]. This mechanism of lactate switch in mammalian cultures is, in fact, an efficient metabolic process and was previously discussed in literature [48], knowing that high levels of lactate are toxic to the cells, consuming it can alleviate this toxicity and can be a source of energy to the cells. This phenomenon was previously discussed in various studies, but an exact explanation about the mechanisms that drives this switch are still unknown. According to these results, we can conclude that, especially regarding the high producer cell line (CHOHyC), glucose consumption rate increased during the exponential phase of the culture when increasing the levels of AKG (Table 4.2). Jointly, during the late exponential phase, even though lactate production rate was higher (Table 4.2), lactate switched to be consumed which decreases its amount in culture, overcoming the problem of lactate accumulation and its toxicity. 4.3.3.2. Amino acids and ammonium When supplementing α-ketoglutarate, a major shift in the exchange rates of several amino acids such as glutamine, glutamate, asparagine, aspartate and ammonium was observed comparing to the standard condition (where CHO cells were grown in CD CHO medium without glutamine and without αketoglutarate) (Figure 4.7 and 4.8). The different exchange rates of ammonia and amino acids in all the tested conditions are observed in table 4.2/4.3 in annexes. At first, in the presence of high levels of α-ketoglutarate in the medium, glutamate is produced, and its rate increases when increasing α-ketoglutarate concentration in the medium. Yet, its secretion period was also extended in cultures supplemented with 12mM α-ketoglutarate comparing to the ones with 4mM α-ketoglutarate (Figure 4.5/C and 4.6/C). The amount of glutamate produced in the treated conditions was equivalent to the levels of α-ketoglutarate available in the medium. When reaching a level of (~2mM) of α-ketoglutarate, glutamate production from AKG is halted. For the case of CHO-HyC cells and looking to the flow of these metabolites over time, we observe that aspartate was the main driver of the conversion of α-ketoglutarate to glutamate as its concentration decreases gradually over time in the treated cells comparing to the non-treated ones (Figure 4.10). This was not observed for CHO-K1 cells (Figure 4.9). Several pathways are driving in the conversion α-ketoglutarate to other metabolites such as glutamate. This involves several reactions using specific amino acids such as CHAPTER 4 109 aspartate, tyrosine, phenylalanine, etc . [49,50]. In our case, we suggest that glutamate production from AKG is mediated by two main reactions, the first is the direct conversion of AKG to glutamate, using ammonium and the second is relying on the conversion of AKG using aspartate, secreting oxaloacetate. Since the extracellular ammonium levels are very low at the beginning of culture, direct conversion of AKG to glutamate is metabolically rather difficult. We suggest that AKG is using the available aspartate to produce glutamate. Investigating this hypothesis, we noticed that aspartate is consumed at higher rates in AKG treated cells. In addition, in figure 4.7 and 4.8 we can observe a comparison between the different tested conditions. When cells are not treated with AKG, we observe a different metabolic profile of aspartate where it is being produced, certainly, as a consequence of asparagine degradation [9], then consumed at the early/mid exponential phase. During growth of producer cells, when aspartate is consumed at high rate, hypothetically, asparagine comes to support the production of aspartate. Aspartate consumption rates increased by 10 folds when supplementing 12mM AKG to the cultures. In addition, asparagine consumption rates increased by 3 folds when increasing AKG concentrations in the medium. In the case of non-producer cells, the latter increased by 1.5 folds. As described in the literature, aspartate uses α-ketoglutarate to produce oxaloacetate and glutamate through transamination [9]. According to the experimental data, we observed that α-ketoglutarate metabolism in CHO is directly linked to the aspartate-asparagine metabolism. One of the possible theories can be that glutamate is also produced by direct conversion of αketoglutarate to glutamate using the free ammonium in the culture and the cofactor NADH. This hypothesis is based on the results that show that, although aspartate is depleted, the production of glutamate continues, relying on other metabolites available in the medium. To investigate this hypothesis, looking at the experimental data, we can observe that, when aspartate is exhausted, CHO starts to consume ammonium from the medium. We can observe that ammonium starts to be consumed at ~day 6/7 of the experiment, exactly when aspartate depletes from the medium (Figure 4.6/D, 4.7/D, 4.9 and 4.10). In the case of non-treated cells, the ammonium levels are rather stationary, which is probably related to the low levels of amino acids in the medium (e.g., glutamine, glutamate, asparagine, aspartate and serine). In cultures treated with α-ketoglutarate, we observe a production of α-ketoglutarate in the stationary phase, after it was depleted in the mid exponential phase. This production of α-ketoglutarate in the stationary phase, comes with a high increase of ammonium levels in the medium. The latter CHAPTER 4 110 phenomenon is an indication that the produced glutamate during the exponential phase, degrades to produce α-ketoglutarate, which will fuel the TCA cycle producing energy and boosting the production of recombinant proteins of interest. Finally, we can observe that for the cells treated with α-ketoglutarate, glutamine production rate increases also with the increase of α-ketoglutarate concentration in the medium. Looking to figures 4.7 and 4.8, we can observe that the biosynthesis rate of glutamine is clearly higher in treated cells comparing to the non-treated cells, with a production rate 12 folds higher than the non-treated cells. Looking deeper to the metabolic flow of glutamine during culture, we can observe that glutamine is produced at high rate in the early exponential phase. In the mid/late exponential phase, the produced glutamine is consumed rapidly. The production/consumption rate of glutamine is higher in the case of CHO supplemented with 12mM α-ketoglutarate comparing to the non-treated cells. These results are in accordance with the hypothesis targeting the use of α-ketoglutarate supplementation to overcome the lack of glutamine in the medium. This way, the cells will rely in its metabolic capabilities to biosynthesize the required metabolites. In addition, glutamine production plays a role in detoxifying the cells from the free ammonium in the culture during growth instead of secreting it when glutamine is initially present in the medium. We can affirm that this strategy is more efficient from bioprocess standpoint overcoming the drawbacks of the initial supplementation of glutamine to the medium but not ignoring its essentiality for the cells, as a nitrogen supply, fueling biosynthesis, energy generation [51] and for the produced recombinant proteins. Another interesting result is the high production of glycine when increasing the AKG concentrations in the medium. Usually, glycine is a product of serine catabolism in culture. Its accumulation in culture indicates a positive effect [8]. Furthermore, alanine secretion rates increase when increasing the level of AKG in cultures. Even if its rate increases, alanine levels in CHO-HyC cultures were lower for the treated cells comparing to the non-treated (Figure 12). A different alanine profile was observed in CHO-K1 cultures. On the other hand, serine consumption rates were very high in the case of CHO-HyC cultures while in the case of CHO-K1 cells, rather a small change in its uptake rate was observed (Figure 4.7 and 4.8). Following, the uptake rates of the essential amino acids, for instance leucine, isoleucine, lysine and valine increased when increasing AKG concentrations in CHO-HyC cells. Conversely, these values were not influenced for CHO-K1 cells. This might be related to the fact, that these amino acids play an CHAPTER 4 111 important role in the production of recombinant proteins in the case of CHO-HyC cells. Higher productivity and titers need to be supported by higher levels of amino acids. Figure 4.7 Exchange rates of key metabolites during culture of CHO-HyC cells treated or not with αketoglutarate. The negative and positive value indicate, respectively, the uptake and secretion rates of the corresponding metabolite. The values of the exchange rates of metabolites are expressed in mmol/gDW/h. Note that the rates of glucose, lactate and other metabolites were scaled down to fit the plot (indicated by the numbers after “/”). Regarding ammonium, its levels in culture were lower when increasing α-ketoglutarate concentration in the medium, for both CHO-HyC and CHO-K1 cells (Figure 4.5.D and 4.6.D). Yet, we noticed that the secretion rates of ammonium are higher when increasing AKG concentrations in the medium. For both cell lines, ammonium levels were slightly lower in culture during the early/mid exponential. Following, in the late exponential phase of the culture, ammonium is consumed and afterwards produced in the stationary phase of the culture, probably as a consequence of oxidative deamination of glutamate via GDH, generating α-ketoglutarate, that fuels the TCA cycle and produces ATP. This process also generates NADH or NADPH, important cofactors for oxidative phosphorylation [52]. These results are very interesting from bioprocess standpoint, since adding AKG to the culture acts on 2 different metabolic related bottlenecks at the same time: decreasing lactate concentrations in the culture, triggering its consumption by fueling the TCA pathway generating more ATP, and also detoxifying the cells from the free ammonium accumulated in the culture through two different reactions. CHAPTER 4 112 For CHO-K1 cells, we observed a different metabolic behavior of the cells towards the use of the available α-ketoglutarate in the medium Figure 4.8 Exchange rates of key metabolites during culture of CHO-K1 cells treated or not with αketoglutarate. The negative and positive value indicate, respectively, the uptake and secretion rates of the corresponding metabolite. The values of the exchange rates of metabolites are expressed in mmol/gDW/h. Note that the rates of glucose and lactate were scaled down to fit the plot (indicated by the numbers after “/”). At the stationary phase, we notice that α-ketoglutarate starts being produced, as a result of degradation of glutamate that was produced in the exponential phase. 4.4. Conclusions Based upon the in silico results described in chapter 3, we recurred in this chapter to test experimentally the effect of supplementing different concentrations of AKG to the culture medium, evaluating its effect on growth, productivity and accumulation of by products during culture. This experimental validation was performed using different CHO strains (producer and nonproducer cells). In this context, we can conclude that AKG is a valuable additive to the culture media and can play a role in substituting glutamine in the formulation. Glutamine is generated due to AKG conversion to glutamate and the latter to glutamine. In addition, we were able to prove that the conversion of AKG to glutamate was mainly based on the use of available aspartate in the medium. Besides, we deduce that when increasing the AKG concentrations CHAPTER 4 113 in the medium, the final product titer of increased significantly. The latter is 1.9 folds higher comparing to the titer obtained in standard conditions. Obtaining higher titers may be correlated with the fact of exhibiting a more effective metabolism, fueling the TCA cycle and allowing lower accumulation of ammonia and lactate during culture. In addition, we were able to underline the different metabolic mechanisms involved in converting the supplemented AKG to CHO cultures. This knowledge is a valuable asset for developing optimization strategies, especially targeting cell culture media optimization. One approach can be based on increasing the aspartate levels in the media due to its potential in driving the conversion of AKG to glutamate. Another optimization strategy can be based on supplementing traces of ammonium in the culture in order to boost other metabolic reactions based on converting directly AKG to glutamate. 4.5. Annexes: Table 4.2 Exchange rates of metabolites for CHO-HyC cells treated or not with AKG (mm/gDW(h). CHO-HyC_0mM AKG CHO-HyC_4mM AKG CHO-HyC_8mM AKG CHO-HyC_12mM AKG qM Standard error qM Standard error qM Standard error qM Standard error Alanine 0.01304 0.00133 0.02543 0.00050 0.02892 0.00148 0.03488 0.00430 Arginine -0.00336 0.00015 -0.00586 0.00088 -0.00534 0.00088 -0.00682 0.00096 Asparagine -0.03790 0.00467 -0.07042 0.00446 -0.08159 0.00264 -0.11450 0.00637 Aspartic acid -0.00227 0.00080 -0.00816 0.00055 -0.01488 0.00345 -0.02440 0.00679 Glutamic acid -0.00350 0.00019 0.01314 0.00469 0.02264 0.00486 0.03249 0.00446 Glutamine 0.00612 0.00464 0.00705 0.00056 0.02686 0.01273 0.07362 0.02515 Glycine 0.00914 0.00541 0.00954 0.00209 0.01289 0.00239 0.03061 0.00510 Histidine -0.00148 0.00027 -0.00248 0.00048 -0.00262 0.00095 -0.00209 0.00084 Hydroxy Proline -0.00036 0.00067 -0.00052 0.00118 -0.00191 0.00285 0.00139 0.00087 Isoleucine -0.00498 0.00025 -0.01126 0.00163 -0.01130 0.00220 -0.01507 0.00280 Leucine -0.00895 0.00011 -0.01901 0.00243 -0.01894 0.00269 -0.02084 0.00221 Lysine -0.00247 0.00187 -0.00730 0.00251 -0.00784 0.00262 -0.01062 0.00669 Methionine -0.00227 0.00026 -0.00356 0.00028 -0.00351 0.00039 -0.00418 0.00070 Phenylalanine -0.00347 0.00041 -0.00594 0.00050 -0.00607 0.00088 -0.00618 0.00070 Proline -0.00507 0.00049 -0.00961 0.00193 -0.01000 0.00276 -0.01460 0.00301 Serine -0.01674 0.00297 -0.03266 0.00353 -0.03533 0.00237 -0.04302 0.00826 CHAPTER 4 114 Threonine -0.00436 0.00034 -0.00721 0.00117 -0.00568 0.00238 -0.00545 0.00306 Tryptophan -0.00139 0.00025 -0.00251 0.00060 -0.00208 0.00091 -0.00114 0.00652 Tyrosine -0.00258 0.00027 -0.00466 0.00056 -0.00451 0.00079 -0.00596 0.00091 Valine -0.00741 0.00044 -0.01329 0.00182 -0.01193 0.00265 -0.01525 0.00229 Glucose -0.3380 0.0202 -0.3557 0.0192 -0.3992 0.0154 -0.4620 0.02854 Lactate 0.3545 0.0469 0.4057 0.0424 0.5378 0.0504 0.7030 0.0455 Ammonia 0.0398 0.0029 0.0434 0.0026 0.0573 0.0018 0.0707 0.00136 AKG 0 0 -0.0238 0.0015 -0.0292 0.0075 -0.1228 0.01583 Table 4.3 Exchange rates of metabolites for CHO-K1 cells treated or not with AKG (mmol/gDW/h). CHO-K1_0mM AKG CHO-K1_4mM AKG CHO-K1_8mM AKG CHO-K1_12mM AKG qM Standard error qM Standard error qM Standard error qM Standard error Alanine 0.01938 0.00179 0.02950 0.00164 0.03737 0.00040 0.03831 0.00135 Arginine -0.00587 0.00072 -0.00645 0.00148 -0.00660 0.00050 -0.00725 0.00054 Asparagine -0.04132 0.00645 -0.04456 0.01076 -0.05297 0.00656 -0.06185 0.00663 Aspartic acid -0.01229 0.00020 -0.01201 0.00036 -0.01420 0.00030 -0.01558 0.00046 Glutamic acid -0.01520 0.00024 -0.01737 0.00141 -0.02175 0.00220 -0.01936 0.00201 Glutamine 0.01326 0.00151 0.01624 0.00161 0.02161 0.00255 0.03280 0.00433 Glycine 0.00644 0.00166 0.00810 0.00354 0.00933 0.00208 0.00870 0.00244 Histidine -0.00229 0.00024 -0.00219 0.00054 -0.00248 0.00037 -0.00304 0.00021 Hydroxy Proline 0.00059 0.00030 0.00161 0.00051 0.00123 0.00092 0.00083 0.00052 Isoleucine -0.00879 0.00084 -0.00930 0.00108 -0.01054 0.00026 -0.01169 0.00039 Leucine -0.01477 0.00138 -0.01569 0.00174 -0.01738 0.00072 -0.01911 0.00049 Lysine -0.00634 0.00176 -0.00577 0.00300 -0.00801 0.00145 -0.00934 0.00185 Methionine -0.00428 0.00029 -0.00438 0.00039 -0.00500 0.00019 -0.00578 0.00022 Phenylalanin e -0.00381 0.00042 -0.00405 0.00094 -0.00434 0.00026 -0.00484 0.00034 Proline -0.00662 0.00087 -0.00402 0.00244 -0.00670 0.00112 -0.00881 0.00077 Serine -0.02522 0.00273 -0.02571 0.00380 -0.03006 0.00242 -0.03228 0.00219 Threonine -0.00591 0.00045 -0.00501 0.00090 -0.00679 0.00082 -0.00781 0.00064 Tryptophan -0.00192 0.00024 -0.00158 0.00039 -0.00217 0.00020 -0.00244 0.00019 Tyrosine -0.00317 0.00028 -0.00339 0.00035 -0.00385 0.00025 -0.00423 0.00030 CHAPTER 4 115 Valine -0.00810 0.00093 -0.00771 0.00142 -0.00931 0.00066 -0.00999 0.00077 Glucose -0.27302 0.01453 -0.26835 0.01476 -0.28437 0.01331 -0.29501 0.01855 Lactate 0.22692 0.02506 0.22269 0.02658 0.25120 0.02918 0.26569 0.02838 Ammonia 0.03775 0.00305 0.03802 0.00385 0.04071 0.00366 0.05037 0.00373 AKG 0 0 -0.00874 0.00070 0.01220 0.00592 -0.01874 0.00440 Figure 4.9 Metabolic flow of metabolites directly involved in the conversion of AKG to glutamate for CHO-K1 cells. (A) represents the aspartate profile over time, (B) represents alanine profile over time, (C) represents asparagine profile over time, (D) represents phenylalanine profile over time, (E) represents tyrosine profile over time. CHAPTER 4 116 Figure 4.10 Metabolic flow of metabolites directly involved in the conversion of AKG to glutamate for CHOHyc cells. (A) represents the aspartate profile over time, (B) represents alanine profile over time, (C) represents asparagine profile over time, (D) represents phenylalanine profile over time, (E) represents tyrosine profile over time. 4.6. References: [1] V.S. Martínez, S. Dietmair, L.E. Quek, M.P. Hodson, P. Gray, L.K. Nielsen, Flux balance analysis of CHO cells before and after a metabolic switch from lactate production to consumption, Biotechnol. Bioeng. 110 (2013) 660–666. https://doi.org/10.1002/bit.24728. [2] A. Hamdi, D. Széliová, D.E. Ruckerbauer, I. Rocha, N. Borth, J. Zanghellini, Key challenges in designing CHO chassis platforms, Processes. (2020). https://doi.org/10.3390/PR8060643. [3] G. Vert, J. Chory, Crosstalk in Cellular Signaling: Background Noise or the Real Thing?, Dev. Cell. (2011). https://doi.org/10.1016/j.devcel.2011.11.006. [4] R. Donaldson, M. Calder, Modelling and analysis of biochemical signalling pathway cross-talk, in: CHAPTER 4 123 Rocha-Pizaña, A. Ramírez-Medrano, M.M. Alvarez, Amino acid consumption in naïve and recombinant CHO cell cultures: producers of a monoclonal antibody, Cytotechnology. 67 (2015) 809–820. https://doi.org/10.1007/s10616-014-9720-5. [15] J. Wahrheit, A. Nicolae, E. Heinzle, Dynamics of growth and metabolism controlled by glutamine availability in Chinese hamster ovary cells, Appl. Microbiol. Biotechnol. (2014). https://doi.org/10.1007/s00253-013-5452-2. [16] M. Jagušić, D. Forčić, M. Brgles, L. Kutle, M. Šantak, M. Jergović, L. Kotarski, K. Bendelja, B. Halassy, Stability of Minimum Essential Medium functionality despite l-glutamine decomposition, Cytotechnology. (2016). https://doi.org/10.1007/s10616-015-9875-8. [17] T.M. Duarte, N. Carinhas, L.C. Barreiro, M.J.T. Carrondo, P.M. Alves, A.P. Teixeira, Metabolic responses of CHO cells to limitation of key amino acids, Biotechnol. Bioeng. 111 (2014) 2095– 2106. https://doi.org/10.1002/bit.25266. [18] D. Geoghegan, C. Arnall, D. Hatton, J. Noble-Longster, C. Sellick, T. Senussi, D.C. James, Control of amino acid transport into Chinese hamster ovary cells, Biotechnol. Bioeng. (2018). https://doi.org/10.1002/bit.26794. [19] M. Zhou, Y. Crawford, D. Ng, J. Tung, A.F.J. Pynn, A. Meier, I.H. Yuk, N. Vijayasankaran, K. Leach, J. Joly, B. Snedecor, A. Shen, Decreasing lactate level and increasing antibody production in Chinese Hamster Ovary cells (CHO) by reducing the expression of lactate dehydrogenase and pyruvate dehydrogenase kinases, J. Biotechnol. (2011). https://doi.org/10.1016/j.jbiotec.2011.03.003. [20] S.M. Noh, J.H. Park, M.S. Lim, J.W. Kim, G.M. Lee, Reduction of ammonia and lactate through the coupling of glutamine synthetase selection and downregulation of lactate dehydrogenase-A in CHO cells, Appl. Microbiol. Biotechnol. (2017). https://doi.org/10.1007/s00253-016-7876-y. [21] T. Fletcher, Designing Culture Media for Recombinant Protein Production A Rational Approach, Bioprocess Int. (2005). [22] Z. Xiao, M. Sabourin, G. Piras, S.F. Gorfien, Screening and optimization of chemically defined media and feeds with integrated and statistical approaches, Methods Mol. Biol. (2014). https://doi.org/10.1007/978-1-62703-733-4_9. [23] M.J. De Jesus, P. Girard, M. Bourgeois, G. Baumgartner, B. Jacko, H. Amstutz, F.M. Wurm, TubeSpin satellites: A fast track approach for process development with animal cells using shaking technology, in: Biochem. Eng. J., 2004. https://doi.org/10.1016/S1369703X(03)00180-3. CHAPTER 4 124 [24] C. Grimm, W. Kusser, B. Lee, G. Bremer, A. Bossie, Media optimization for CHO fed-batch processes using a DoE approach in automated high-throughput single use ambr15 bioreactors, BMC Proc. (2015). https://doi.org/10.1186/1753-6561-9-s9-p15. [25] B. Tihanyi, L. Nyitray, Recent advances in CHO cell line development for recombinant protein production, Drug Discov. Today Technol. (2021). https://doi.org/10.1016/j.ddtec.2021.02.003. [26] S.M. Noh, S. Shin, G.M. Lee, Comprehensive characterization of glutamine synthetase-mediated selection for the establishment of recombinant CHO cells producing monoclonal antibodies, Sci. Rep. 8 (2018) 5361. https://doi.org/10.1038/s41598-018-23720-9. [27] G.R. Bebbington, G. Renner, S. Thomson, D. King, D. Abrams, G.T. Yarranton, High-level expression of a Recombinant Antibody from myeloma cells using a glutamine synthetase gene as an amplifiable selectable marker, Bio/Technology. 10 (1992) 169. https://doi.org/10.1038/nbt0292-169. [28] A. Rita Costa, M. Elisa Rodrigues, M. Henriques, J. Azeredo, R. Oliveira, Guidelines to cell engineering for monoclonal antibody production, Eur. J. Pharm. Biopharm. (2010). https://doi.org/10.1016/j.ejpb.2009.10.002. [29] F.M. Wurm, Production of recombinant protein therapeutics in cultivated mammalian cells, Nat. Biotechnol. (2004). https://doi.org/10.1038/nbt1026. [30] B.C. Mulukutla, J. Mitchell, P. Geoffroy, C. Harrington, M. Krishnan, T. Kalomeris, C. Morris, L. Zhang, P. Pegman, G.W. Hiller, Metabolic engineering of Chinese hamster ovary cells towards reduced biosynthesis and accumulation of novel growth inhibitors in fed-batch cultures, Metab. Eng. (2019). https://doi.org/10.1016/j.ymben.2019.03.001. [31] N. Wu, M. Yang, U. Gaur, H. Xu, Y. Yao, D. Li, Alpha-ketoglutarate: Physiological functions and applications, Biomol. Ther. (2016). https://doi.org/10.4062/biomolther.2015.078. [32] B. Zdzisińska, A. Żurek, M. Kandefer-Szerszeń, Alpha-Ketoglutarate as a Molecule with Pleiotropic Activity: Well-Known and Novel Possibilities of Therapeutic Use, Arch. Immunol. Ther. Exp. (Warsz). (2017). https://doi.org/10.1007/s00005-016-0406-x. [33] S. Liu, L. He, K. Yao, The antioxidative function of alpha-ketoglutarate and its applications, Biomed Res. Int. 2018 (2018). https://doi.org/10.1155/2018/3408467. [34] T.K. Ha, G.M. Lee, Effect of glutamine substitution by TCA cycle intermediates on the production and sialylation of Fc-fusion protein in Chinese hamster ovary cell culture, J. Biotechnol. (2014). https://doi.org/10.1016/j.jbiotec.2014.04.002. CHAPTER 4 125 [35] C. Lattenmayer, E. Trummer, K. Schriebl, K. Vorauer-Uhl, D. Mueller, H. Katinger, R. Kunert, Characterisation of recombinant CHO cell lines by investigation of protein productivities and genetic parameters, J. Biotechnol. (2007). https://doi.org/10.1016/j.jbiotec.2006.12.016. [36] G. Klanert, D.J. Fernandez, M. Weinguny, P. Eisenhut, E. Bühler, M. Melcher, S.A. Titus, A.B. Diendorfer, E. Gludovacz, V. Jadhav, S. Xiao, B. Stern, M. Lal, J. Shiloach, N. Borth, A crossspecies whole genome siRNA screen in suspension-cultured Chinese hamster ovary cells identifies novel engineering targets, Sci. Rep. (2019). https://doi.org/10.1038/s41598-01945159-2. [37] X. Pan, C. Dalm, R.H. Wijffels, D.E. Martens, Metabolic characterization of a CHO cell size increase phase in fed-batch cultures, Appl. Microbiol. Biotechnol. (2017). https://doi.org/10.1007/s00253-017-8531-y. [38] N. Kumar, P. Gammell, 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, in: Cytotechnology, 2007. https://doi.org/10.1007/s10616-007-9047-6. [39] C. Altamirano, J.J. Cairó, F. Gòdia, Decoupling cell growth and product formation in Chinese hamster ovary cells through metabolic control, Biotechnol. Bioeng. (2001). https://doi.org/10.1002/bit.10096. [40] K.H. Chang, K.S. Kim, J.H. Kim, N-acetylcysteine increases the biosynthesis of recombinant EPO in apoptotic Chinese hamster ovary cells, Free Radic. Res. (1999). https://doi.org/10.1080/10715769900300091. [41] M.J.B. Matthew P. Zustiak, Yueqing Xie, Jianwei ZhuJianwei Zhu, Cell Cycle Arrest Engineering to Enhance Production Phase in CHO Cell Culture, in: AIChE Annu. Meet., 2010. [42] J.X. Bi, J. Shuttleworth, M. Al-Rubeai, Uncoupling of Cell Growth and Proliferation Results in Enhancement of Productivity in p21CIP1-Arrested CHO Cells, Biotechnol. Bioeng. (2004). https://doi.org/10.1002/bit.20025. [43] H. Kaufmann, X. Mazur, M. Fussenegger, J.E. Bailey, Influence of low temperature on productivity, proteome and protein phosphorylation of CHO cells, Biotechnol. Bioeng. (1999). https://doi.org/10.1002/(SICI)1097-0290(19990605)63:5<573::AID-BIT7>3.0.CO;2-Y. [44] J. Xu, P. Tang, A. Yongky, B. Drew, M.C. Borys, S. Liu, Z.J. Li, Systematic development of temperature shift strategies for Chinese hamster ovary cells based on short duration cultures and kinetic modeling, MAbs. (2019). https://doi.org/10.1080/19420862.2018.1525262. CHAPTER 4 126 [45] Z. Jiang, S.T. Sharfstein, Sodium butyrate stimulates monoclonal antibody over-expression in CHO cells by improving gene accessibility, Biotechnol. Bioeng. (2008). https://doi.org/10.1002/bit.21726. [46] N.H. Simpson, R.P. Singh, A.N. Emery, M. Al-Rubeai, Bcl-2 over-expression reduces growth rate and prolongs G1 phase in continuous chemostat cultures of hybridoma cells, Biotechnol. Bioeng. (1999). https://doi.org/10.1002/(SICI)1097-0290(19990720)64:2<174::AIDBIT6>3.0.CO;2-D. [47] F. Hartley, T. Walker, V. Chung, K. Morten, Mechanisms driving the lactate switch in Chinese hamster ovary cells, Biotechnol. Bioeng. (2018). https://doi.org/10.1002/bit.26603. [48] L. Liste-Calleja, M. Lecina, J. Lopez-Repullo, J. Albiol, C. Solà, J.J. Cairó, Lactate and glucose concomitant consumption as a self-regulated pH detoxification mechanism in HEK293 cell cultures, Appl. Microbiol. Biotechnol. (2015). https://doi.org/10.1007/s00253-015-6855-z. [49] R.P. Nolan, K. Lee, Dynamic model of CHO cell metabolism, Metab. Eng. (2011). https://doi.org/10.1016/j.ymben.2010.09.003. [50] D. Xiao, L. Zeng, K. Yao, X. Kong, G. Wu, Y. Yin, The glutamine-alpha-ketoglutarate (AKG) metabolism and its nutritional implications, Amino Acids. (2016). https://doi.org/10.1007/s00726-016-2254-8. [51] A. Csibi, S.M. Fendt, C. Li, G. Poulogiannis, A.Y. Choo, D.J. Chapski, S.M. Jeong, J.M. Dempsey, A. Parkhitko, T. Morrison, E.P. Henske, M.C. Haigis, L.C. Cantley, G. Stephanopoulos, J. Yu, J. Blenis, The mTORC1 pathway stimulates glutamine metabolism and cell proliferation by repressing SIRT4, Cell. (2013). https://doi.org/10.1016/j.cell.2013.04.023. [52] A. Plaitakis, E. Kalef-Ezra, D. Kotzamani, I. Zaganas, C. Spanaki, The glutamate dehydrogenase pathway and its roles in cell and tissue biology in health and disease, Biology (Basel). (2017). https://doi.org/10.3390/biology6010011. CHAPTER 5 127 5. CHAPTER 5 Cell culture media optimization for CHO cells The information presented in this Chapter is being prepared for submission to a peer reviewed journal: Hamdi A., Borth, N., Zanghellini, J., Rocha I.; In silico -based approach for medium optimization of CHO cells. --------------------------------------------------------------------------------------------------------- 5.1. Introduction Based on the recent advances in metabolic engineering and bioinformatics, several efforts are targeted at improving mammalian bioprocesses [1,2] and translating the predictive pipeline for bioprocess optimization from academic research to industry. Various studies highlighted the benefits of mathematical modeling for bioprocess optimization [3]. To achieve these goals, optimization strategies are being used employing CHO specific genome scale models, to predict cell growth, metabolic features of the cells and optimize the productivity/quality of the produced recombinant proteins (e.g., monoclonal antibodies) [4]. These predictions are based on combining genome scale metabolic models, experimental omics data and constraint-based modeling approaches to consolidate fluxomics studies. Omics data are very important asset to calculate, in silico , cellular flux distributions in an accurate manner [5] and consequently infer about the cell metabolism when varying culture conditions. Assuming steady state or a dynamic mode, these tools can be very useful for metabolic engineering and media design/optimization strategies. Along the years, several approaches for media optimization of mammalian cells, especially for CHO, have been explored, aiming at improving cell culture conditions and production titers. As described in chapter 2 and 4, one of the most important highlights was the development of serum-free, animal-free, chemically defined media. These media were further optimized using different strategies such as media blending and high-throughput screening of the best candidates, relying on deterministic modeling approaches, for instance design of experiments (DoE) [6]. CHAPTER 5 128 Moreover, to further improve the production yield, strategies such as fed-batch and perfusion technologies were used to control the levels of toxic metabolites and improve the production titers of recombinant proteins of interest by balancing nutrients levels in culture and improving feeding strategies [7]. Several efforts were performed for media and feed optimization, improving CHO cells productivity and also decreasing the hurdle of downstream processing [8–14]. Knowing that different strains within CHO have different growth requirements, feeding strategies have to be customized to the cell strain in use [15,16]. It is also relevant to address other topics, for instance, cells heterogeneity and epigenetic regulations. Further, one of the most important aspects of media optimization is based on spent media analysis. This strategy is very useful to understand the flow of metabolites in the culture, which will facilitate designing the best media formulation for the cell line in use, and also determining optimal feeding tactics [17]. Accordingly, regular monitoring of glucose, amino acids, vitamins, etc. is very important in order to understand how the cells consume these nutrients and what are the limiting components that might influence the bioprocess performance. Following this idea, balancing amino acids levels in the medium has a great potential in improving not only growth parameters of CHO, but also the final product titers. Previous studies showed that by optimizing the amino acids levels in the medium, CHO cells were able to increase titers of fusion proteins by almost 25% [18]. Other efforts based on employing statistical experimental design methodologies helped also in improving the titers of recombinant proteins in CHO by 70%. These studies allowed understanding the key amino acids essential for CHO growth and the important elements interfering with the production of recombinant proteins [19,20]. Amino acids are very important metabolites for the mammalian cell [21]. Besides providing essential building blocks molecules for cell metabolism and protein production, amino acids can also act as signaling molecules influencing cellular apoptosis [22] and also regulate the levels of osmolality in the medium and of other metabolites (e.g., ammonium). Therefore, it is mandatory to optimize the amino acids levels in the medium to fulfill the cellular needs. Within mammalian cells, it is important to optimize the concentrations of both essential amino acids (EAA) and non-essential amino acids (NEAA) in the medium, since they control, in a sophisticated manner, growth and recombinant proteins production. In fact, the medium has to contain high levels of EAA since they are consumed usually at very high rates [23]. Low concentrations of some EAA amino acids in culture, for instance, branched CHAPTER 5 129 amino acids (e.g., Leucine, isoleucine and Valine) or also phenylalanine, can heavily influence transporters activity in the cell, for instance, the L-transport system. Studies showed that Na+ transport system increased activity by 3-4 folds when cells starved these amino acids [24]. In media design efforts, different amino acids often have to be added at higher concentrations than the values determined in silico or by fluxomics/metabolomics studies [25] and levels of other EAA (e.g., tryptophan or lysine) have to be optimized since their excess or deficiency can radically influence the process, for example by secreting toxic amino acids derivatives [26,27]. On the other hand, NEAA are also very important nutrients for various metabolic pathways within the cell. Even though NEAA can be synthesized by mammalian cells, they are substantial in the cell culture medium. As previously described (Chapter 4), we noticed that when supplementing AKG into the medium, a very significant shift was observed from the levels of NEAAs standpoint. Glutamine and glutamate were highly produced and aspartic acid was depleted at a very early phase of the culture. An important observation was also regarding the exchange rates of both alanine and glycine, reflecting that high levels of AKG in the medium can rewire the metabolism and alter its behavior. Asparagine and aspartic acid are particularly important in various mechanisms in the mammalian metabolic system. These NEAAs are typically consumed at very high rates, mainly at the exponential phase of the culture, playing an important role in energy and glutamine/glutamate metabolism [12,28]. Lack of asparagine in the cell culture medium can radically impact protein synthesis and also the process by altering the quality of monoclonal antibodies [29]. However, its lack in the medium can be compensated by higher uptake of other amino acids such as aspartate, glutamine or glutamate. Although asparagine is known to be highly aminogenic, ammonium concentrations can be further controlled by balancing the levels of asparagine, glutamine and/or glutamate in the medium [30]. When designing an optimized formulation of the cell culture media, it is important to focus on studying also the recombinant protein being produced. In our case, CHO-HyC cells are producing Trastuzumab, a humanized monoclonal antibody under the commercial name of Herceptin. The latter is an IgG1 kappa molecule, a very important biopharmaceutical for the treatment of HER2-positive metastatic breast cancer [31], developed by Genentech/Roche and targeting specifically the human epidermal growth factor receptor 2 (HER-2/neu) [32,33]. The amino acids sequence of both the heavy chain and the light chain of this molecule is described in the invention patent of Trastuzumab [34]. Supplementing α-ketoglutarate to CHO cultures showed a great potential in improving productivity and decreasing by-products accumulation. Subsequently, according to the metabolomics results obtained in CHAPTER 5 130 chapter 4, we will try to optimize cell culture media and apply it to CHO-HyC. However, there is a wide room for media optimization, since we noticed that in the presence of AKG, the amino acids levels in the standard medium are unbalanced. Aspartate was highly consumed in the presence of AKG in culture, together with a high production of glutamate. Following these observations, we decided to design a medium containing the same levels of essential amino acids as the standard used media CD CHO, together with altering the levels of NEAAs by removing glutamate from the formulation and replacing it by 8 mM AKG. Additionally, higher levels of aspartate were supplemented to boost the production of glutamate by AKG. Additionally, a decrease in asparagine levels was tested in the presence of high values of aspartate in the media. Finally, the effect of supplementing different concentration of ammonium was performed. This strategy is based on triggering the reaction of direct production of glutamate from AKG at the beginning of the culture, relying on ammonia. Since ammonia levels in the beginning of the batch are very low, this strategy was employed in order to force the production of glutamate to fulfill the metabolic need of CHO-HyC cells. Taking into consideration that the latter is a GSneg cell line where the GS gene was inserted with the transgene of interest, glutamine production is expected during culture to overcome its lack in the medium. 5.2. Materials and methods 5.2.1. Experimental setup 5.2.1.1. Cell culture The CHO-HyC cell line was used in this work. These cells correspond to an antibody expressing CHO cell line provided by Cytiva, Uppsala, Sweden. The latter is GSneg cell line and known as a highproducer industrial clone, producing Trastuzumab, a monoclonal antibody under the commercial name of Herceptin. CHO cells were cultivated in suspension mode in chemically defined serum-free conditions using CD CHO medium (Gibco, Invitrogen, Carlsbad, CA, USA). After thawing, the cells were routinely cultivated in 50 mL TPP® TubeSpin bioreactors (Techno Plastic Products AG, Trasadingen, Switzerland) at a maximal working volume of 25 mL. The cells were incubated in 37°C in 80 % humidified air with 7 % CO2, shaking at a speed of 220 rpm (rotation per minute). The cells were passaged every 3-4 days and the viable cell concentrations, viabilities and the values of the average cell diameters were determined using Vi-CELLXR (Beckman Coulter, USA). CHAPTER 5 131 To perform the experiments described in this chapter, we recurred to the use of CD CHO media without amino acids, a pre-customized medium purchased from Thermo-Fisher Scientific (Reference: ME19349L1). The formulation contained the same components with concentrations similar to CD CHO media, used for the negative control experiments. In order to design media formulation, non-essential amino acids levels were optimized. Amino acids stock solutions were prepared in the lab (see next section) and supplemented to the culture medium according to the desired levels in each experiment. Regarding essential amino acids, the levels were unchanged. Its concentration in the optimized media is equivalent to the values described in the CD CHO medium. In these experiments, different levels of glutamate, aspartic acid and asparagine, were tested in this study. After thawing, the cells were passaged 3 times in the corresponding culture conditions. For the experimental setup, batch cultures were performed in triplicates using 50 mL TubeSpin bioreactors at a working volume of 28 mL, incubated in 37°C in 80 % humidified air with 7 % CO2, shaking at a speed of 220 rpm. All the experiments were performed in triplicates, inoculated at the beginning of the experiment, at the same time and seeding density. Cell concentration and viability were monitored every 24h. In addition, samples for metabolomics evaluation were withdrawn every 24h. The culture continued until reaching a cell viability lower than 60 %. 5.2.1.2. Preparation of amino acids solutions All the amino acids used in this study are from non-animal origin and suitable for cell culture experiments. The different amino acids used in these experiments were dissolved in CD CHO medium without amino acids, forming the stock solutions prior to the experiments. The reason behind dissolving these amino acids in the medium is to avoid the dilution of the other nutrients of the medium. The pH of the stock solutions was adjusted to 7.2 (besides some amino acids which are stable in acidic or basic pH). In table 1, we find the list of the used amino acids, its solubility values, the stock concentration prepared and the manufacturer references jointly with the CAS-number. CHAPTER 5 132 Table 5.1 Amino acids used in this study. Amino acid Solubility* Stock concentration Reference** CASNumber L-Arginine H2O: 100 mg/mL 70 mM A8094 74-79-3 L-Asparagine 1 M HCl: 100 mg/mL 100 mM A4159 70-47-3 L-Aspartate 1 M HCl: 100 mg/mL 70 mM A7219 56-84-8 L-Cystin 1 M HCl: 100 mg/mL 30 mM C7602 56-89-3 L-Glutamate 1 M HCl: 100 mg/mL 100 mM G8415 56-86-0 L-Histidine H2O: 50 mg/mL 70 mM H6034 71-00-1 LHydroxyproline H20 70 mM H5534 51-35-4 L-Isoleucine 1 M HCl: 50 mg/mL 100 mM I7403 73-32-5 L-Leucine 1 M HCl: 50 mg/mL 100 mM L8912 61-90-5 L-Lysine H2O: 100 mg/mL 70 mM L8662 657-27-2 L-Methionine H2O: 25 mg/mL 70 mM M5308 63-68-3 L-Phenylalanine 1M HCl; 50 mg/mL 70 mM (Protected from light) P5482 63-91-2 L-Proline H2O: 50 mg/mL 100 mM P5607 147-85-3 L-Serine H2O: 50 mg/mL 100 mM S4311 56-45-1 L-Threonine H2O: 1 g/10 mL, clear, colorless 70 mM T8441 72-19-5 L-Tryptophan 1 M HCl: 10 mg/mL 70 mM (Protected from light) T8941 73-22-3 L-Tyrosine 1 M HCl: 25 mg/mL 70 mM T8566 60-18-4 L-Valine H2O: 25 mg/mL 100 mM V0513 72-18-4 *According to manufacturer recommendations. **Sigma-Aldrich order reference.