Over the past three decades biopharmaceutical sales have continuously increased becoming a 100 billion dollar industry. Around 60% of biopharmaceuticals are produced using mammalian cells, due to their ability to perform complex post-translational modifications. When compared with microorganisms, however, mammalian cells require expensive media and have low productivity, resulting in expensive biopharmaceuticals. The production of growth inhibitory by-products is the main factor affecting product yield. Therefore, a quantitative understanding of mammalian cell metabolism as a whole is crucial to improve biopharmaceutical production. Fluxomics, i.e., the quantification of metabolic fluxes, represents integrative information of cellular regulation at many different levels, thus is the ideal framework to study cellular metabolic phenotypes.
Metabolic flux analysis (MFA) is the approach used to estimate metabolic fluxes from measurement of external metabolites. MFA relies on mass balances around each metabolite in the metabolic network representing all biochemical reactions in a cell. While metabolic steady-state is the main assumption of MFA, the cell metabolism changes over time in industrial processes of biopharmaceuticals production. Hence, a method to unravel the dynamic metabolism of mammalian cells is required in order to improve their productivity. This thesis considers the two essential elements required for performing dynamic MFA (DMFA), namely a metabolic network and a framework to estimate dynamic metabolic fluxes.
The most comprehensive metabolic networks are called genome scale models (GeMs). These models are underdetermined and thermodynamic constraints dictating reaction directions are essential to obtain meaningful metabolic flux estimations. The erroneous assignment of thermodynamic constraints, however, can lead to misleading interpretations of cell metabolism. Using thermodynamic principles the constraints can be validated (review in Chapter 2) and NExT – a Matlab implementation of network-embedded thermodynamic analysis – was developed to validate GeM thermodynamic constraints (Chapter 3 and Appendix A). Two eukaryotic GeMs, Yeast 5 and Human Recon 1, were curated using network thermodynamics. Approximately 70% of all reactions in both models were initially specified as irreversible. The irreversible reactions in Recon 1 were all found to be thermodynamically feasible and further 27 thermodynamic constraints were identified under physiological conditions. For Yeast 5, three thermodynamically infeasible, and thirteen new thermodynamic irreversibility specifications were found. In addition, several unconnected internal loops were identified in both models and differences between yeast and humans were exposed in the compartmentalization of proline metabolism. These results highlight network thermodynamics as a powerful tool for the curation of compartmentalized GeMs.
Internal metabolite measurements can be included with thermodynamic principles to determine the direction of reactions and further constrain metabolic networks (review in Chapter 2). We investigated the minimum number of measurements required to determine the direction of all Recon 1 GeM reactions (Chapter 4). In the process of finding this minimum number of measurements, a new approach to modularize metabolism was developed. Ignoring connections through currency metabolites (e.g., ATP), the metabolism was divided into 58 modules functionally isolated from each other by irreversible reactions. It was finally found that the structurally reversible reactions (the ones not already directionality constrained by surrounding irreversible reactions) operate almost independent of each other, i.e., around 93% of structurally reversible reaction directions need to be known in order to determine the direction of all network reactions.
A semi-dynamic MFA approach was employed to study the metabolism of CHO cells that are able to consume lactate after glucose depletion (Chapter 5). The growth profile was divided in lactate production (exponential growth) and lactate consumption (reduced growth rate); then the fluxes in each of the metabolic phases were calculated by conventional MFA. As expected, the glycolysis flux was reversed to gluconeogenesis under lactate consumption to satisfy biomass requirements. TCA cycle reaction fluxes were similar under both metabolic phases, resulting in a more efficient use of nutrients for energy production in the lactate consumption phase. Although the CHO metabolism was partially unraveled by semi-dynamic MFA, using this approach only produces an average flux values for the 20 hours of lactate consumption phase. In reality metabolism changes gradually, thus to get a better understanding of the metabolic switch a real dynamic approach is essential.
Finally, a DMFA framework based on B-spline fitting (B-DMFA) was developed (Chapter 6). B-DMFA performs smooth fitting of experimental measurements and estimation of mass balanced metabolic fluxes in one step, when the metabolic state transition times are known. Moreover, a fast algorithm to determine the position and number of transition time points that generate a statistically acceptable fit was created. B-DMFA was used to characterize and contrast the metabolism of two CHO cell cultures conditions: cultured at constant 37°C or temperature shifted from 37°C to 32°C. The differences in central metabolism were mainly caused by a bigger cell volume of the temperature shifted cells. Thus, the metabolic fluxes were estimated considering cell volume instead of cell number. B-DMFA revealed that the volume-specific IgG productivity under temperature shifted conditions was close to constant, while the cells at constant temperature had a peak during mid-exponential and a subsequent gradual decrease in productivity. Consequently, a final higher product yield was achieved by the temperature shifted cells.
The tools and framework developed in this thesis set the path to unravel the dynamic metabolism of mammalian cells. The framework is generic, and with adequate organism specific models can be used to study temporal metabolism of any organism.