Metabolic-flux analysis of mammalian-cell culture.

In the biopharmaceutical industry mammalian cells are cultivated for the production of recombinant glycoproteins, vaccines, and monoclonal antibodies. In contrast to other expression systems, such as prokaryotes or yeasts, mammalian cells are able to glycosylate and fold therapeutic proteins correctly, and therefore the only possible production system for many (recombinant) therapeutics.Cultivated mammalian cells are similar to tumor cells: in contrast to normal cells in mammalian tissue they can proliferate continuously and are not differentiated to fulfill tissue-specific tasks. Cultivated cells and tumor cells also share other characteristics, for example in their metabolism. In general the metabolism of continuously proliferating cells is not or only poorly regulated and controlled, and therefore inefficient. Cultivated mammalian cells show a high metabolic activity, and waste large amounts of nutrients and energy. Instead of tuning the consumption of glucose and certain amino acids to the requirements for growth, these nutrients are taken up whenever they are available. As a result waste products such as lactic acid, carbon dioxide, and bicarbonate accumulate, acidify the culture medium, and inhibit cell growth and protein production.Another shared characteristic of tumor cells and cultured mammalian cells is the production of ammonia. Mammals normally produce urea, a waste product of the endogenuous metabolism, in the liver. Mammalian cells that are cultivated for glycoprotein production do not possess the machinery for the production of urea. Instead they secrete ammonia into the culture medium, which accumulates at toxic levels.It is thus apparent that the metabolism of mammalian cells is suboptimal for an efficient energy- and nutrient supply. To quantify the exact nutrient requirements for growth and energy, and to investigate which metabolic pathways should be optimized to reduce waste-product synthesis to increase production yields, the intracellular reaction rates, i.e. "the metabolic fluxes", have to be determined. Intracellular fluxes can be quantitated by incubating cells with isotope-labeled nutrients and measurement of the isotope distributions of end-products. This method (i) has practical limitations as it is limited to the analysis of single metabolic pathways per tracer experiment, (ii) is expensive, and (iii) is not feasible at industrial scale. An alternative method is based on solving the linear set of equations that is determined by the mass balances of the relevant metabolites. In this dissertation this new method which is referred to as "metabolic-flux balancing" is applied to mammalian-cell culture.<img src="/wda/abstracts/i2493_1.gif" ALIGN="LEFT" HSPACE=12/>Metabolic-flux balancing techniques are based on relatively simple linear algebra. If the stoichiometry of the relevant intracellular reactions and the cellular composition are known and the uptake- and secretion rates of the relevant metabolites have been measured, the reaction rates can be determined using the appropriate mass-balance equations. Taken together, the mass-balance equations form a set of linear equations that can be solved by linear regression. However in the metabolism of the cell there are a number of cyclic pathways which are linearly dependent within the set of mass-balance equations, which causes the metabolic network to be underdetermined. This is the central problem in this dissertation and is explained in detail in Chapter 1.<img src="/wda/abstracts/i2493_2.gif" ALIGN="LEFT" HSPACE=12/>In most cyclic metabolic pathways certain co-metabolites are consumed or produced. Carbon dioxide for example, is a waste product of the TCA cycle. By measuring the carbon-dioxide production rate it should therefore be possible to estimate the flux through this essential metabolic cycle. However, in mammalian-cell culture the measurement of the carbon-dioxide production rate is troubled by the use of bicarbonate as a buffer system and the accumulation carbon dioxide in the culture medium. In Chapter 2 a method is developed for the determination of the carbon-dioxide production rate in bicarbonate-buffered bioreactor systems.<img src="/wda/abstracts/i2493_3.gif" ALIGN="LEFT" HSPACE=12/>In Chapter 3 it is shown that, if this method is used, it is possible to close the mass balance for total carbon in mammalian-cell culture. Together with the balance for nitrogen it is now possible to demonstrate statistically that there are no gross errors in the measurement data and there are not missing any relevant metabolites. Now, the metabolic-flux analysis can start.....<img src="/wda/abstracts/i2493_4.gif" ALIGN="LEFT" HSPACE=12/>Unfortunately, the intracellular fluxes cannot be determined solely by flux-balancing techniques even when the mass balances of co-metabolites such as carbon dioxide are included in the metabolic network. This is both a result of the fact that mass balances of particular co-metabolites cannot be closed ( e.g. the mass balance of ATP), and of the fact that co-metabolites are produced or consumed in more than one cyclic pathway. Such cyclic pathways remain therefore linearly dependent, and additional information is required to quantify the flows through these cycles. A solution is proposed in Chapter 4. It is assumed that hybridoma cells are efficient with respect to their metabolism (while taking into account the "inefficient" production and uptake rates mentioned above), and minimize the flow through the metabolic network. Within the set of all admissible solutions, the flux distribution with the minimum sum of squares is chosen. This assumption is referred to as the "minimum-norm constraint". The method is applied to hybridoma cells under both optimal and suboptimal conditions.<img src="/wda/abstracts/i2493_5.gif" ALIGN="LEFT" HSPACE=12/>In Chapter 5 experiments are described that are conducted to test the assumption mentioned above. The metabolic fluxes in the TCA cycle, the pentose-phosphate cycle, and the malate shunt are determined by <sup>13</SUP>C-tracer experiments. Hybridoma cells are cultured on a small scale and <sup>13</SUP>C-labeled glucose is added to the culture medium. Subsequently, the isotopic distribution of lactate is determined by NMR spectrometry and the fraction of <sup>13</SUP>C in carbon dioxide is measured by mass spectrometry. It appears that the actual fluxes in the mentioned cycles are significantly different from the fluxes estimated using the minimum-norm constraint.In addition, it appears that rapidly proliferating hybridoma cells have a higher pentose-shunt activity than previously assumed. The reason for this high activity is probably the high need for NADPH (which is required for biomass synthesis). This also means that a larger part of glucose is consumed more efficiently than previously assumed on the basis of the amount of produced lactate alone. This allows to estimate the optimal amount of glucose that cells should consume per gram produced biomass.<img src="/wda/abstracts/i2493_6.gif" ALIGN="LEFT" HSPACE=12/>In Chapter 6 other theoretical methods to estimate metabolic fluxes in underdetermined networks are used to estimate fluxes in cyclic pathways. Linear optimization techniques are applied to determine solutions that are optimal with respect to particular "metabolic objectives". Various metabolic strategies that may be relevant for hybridoma cells are translated into so-called "linear objective functions" and used to estimate the metabolic flux distribution. It appears that the biochemical objective "maximize NADH-producing fluxes" gives flux values that approximate the values determined experimentally by isotopic-tracer studies (Chapter 5). It is speculated that this objective is in agreement with the uncontrolled oxidation of any available nutrients by continuously-growing mammalian cells, regardless of the need for ATP or NADH.<img src="/wda/abstracts/i2493_7.gif" ALIGN="LEFT" HSPACE=12/>Under certain (extreme) culture conditions cells have to adapt their metabolism to the stress to which they are exposed. For example, cells in a bioreactor can be limited in oxygen supply, or certain toxic components can force the cell to redirect the fluxes into a particular direction. In Chapter 7 several experiments are described in which hybridoma cells are artificially stressed. It is shown that at oxygen limitation certain NAD(P)H-producing fluxes decrease , most likely to restore the disturbed NAD(P) <sup>+</SUP>/NADPH balance. Under oxidative stress the opposite occurs: NAD(P)H-producing fluxes increase. Other fluxes which -strictly speaken- cannot be determined by balancing techniques alone are subsequently estimated with "physiologically meaningful" objective functions. These objectives are associated with the metabolic strategy to adapt to the relevant stress. For example, at oxygen limitation the objective function "minimize NADH-producing fluxes" applies.<img src="/wda/abstracts/i2493_8.gif" ALIGN="LEFT" HSPACE=12/>Ammonia is a waste product that is toxic for mammalian cells at relatively low concentrations and it limits the cell density in bioreactors. In Chapter 8 it is shown by flux-balancing techniques that hybridoma cells can reduce ammonia levels by converting ammonia andα-ketoglutarate into glutamate (a reaction catalyzed by the enzyme glutamate dehydrogenase). This suggests that overexpression of this enzyme may allow mammalian cells to survive higher concentrations of ammonia, which potentially enables high-cell density cultures.<img src="/wda/abstracts/i2493_9.gif" ALIGN="LEFT" HSPACE=12/>It has been demonstrated that an important fraction of fluxes of large metabolic networks such as mammalian-cell metabolism can be estimated with only mass-balancing techniques. For the determination of fluxes in cyclic pathways isotopic-tracer experiments remain indispensable. However, relative trends in intracellular metabolic fluxes upon changes in extracellular conditions can be determined solely by mass-balancing techniques even if the metabolic network is principally underdetermined. The combination of flux-balance models and isotopic-tracer studies, of which an example is given in Chapter 5, will be the future tool of quantitative flux analysis of complex metabolic networks