Discussion
Herein we present a novel approach to utilize amino acid stoichiometric balances as inputs into MVDA models such as the orthogonal partial least squares algorithm to balance key amino acids in chemically-defined cell culture media supplements to increase CHO cell growth and productivity. However, the utilization of stoichiometric balances in chemometric approaches is not novel. Experimental stoichiometric models date as early as 1994 where Xie and Wang designed a cell culture medium development strategy in which the key nutrients required in biomass and product generation could be described mathematically by stoichiometric coefficients (Xie & Wang, 1994b; Xie & Wang, 1994c). Since increased cell mass is a necessary requirement to produce the desired concentration of key biologic products such as mAbs, the specific cell growth rate (ugrowth) and specific productivity (qP) become two key governing parameters within the stoichiometric equation. The resulting stoichiometric balance would help derive the theoretical demand of each nutrient, and thus, Xie and Wang showed that various nutrients could be balanced within complex CD media without the necessity of multi-factorial DoEs.
For instance, in a follow up study, Xie and Wang used their stoichiometric model to decrease ammonia and lactate formation in hybridoma cells by balancing the basal medium nutrient levels. In this model, they analyzed the relationship between glutamine and nonessential amino acids as a function of lactate and ammonia production rates. The modified medium resulted in a 4-fold decrease in ammonia production and almost a 10-fold decrease in lactate formation from the hybridoma cells with the added benefit of increased cell growth and product formation (Xie & Wang, 1994a). Val et al utilized the stoichiometric model to balance nucleotide sugars as a function of cell growth and specific productivity to understand nucleotide sugar demand towards glycosylation profiles in CHO cells (Del Val, Polizzi, & Kontoravdi, 2016). Xie and Nyberg et al further utilized the stoichiometric modeling approach in CHO cells by designing a serum-free feed medium with balanced nutrients to improve cell growth and IFNy productivity as well as glycosylation efficiency for IFNy (Xie et al., 1997).
Similarly, we derived stoichiometric balances to improve cell growth and productivity as a function of all 20 amino acids. In agreement with the original method, the stoichiometric coefficients or theoretical demands for amino acids derived in our model were based on their contributions towards biomass and mAb productivity as a function of specific growth rate (µgrowth) and specific productivity (qP). Accordingly, Xie and Wang’s original method also considered the cell death rate, thus resulting in gross growth rate (Xie & Wang, 1994c). Cell death rate becomes particularly important during the late-stationary and death phases of the CHO cell culture where viable cell densities and viabilities start to decline (Balandras et al., 2011). The increased death rate during the latter half of the culture is attributed to several factors including increased concentration of harmful byproducts such as NH3, increased packed cell volume, and rapid nutrient elimination (Pan, Dalm, Wijffels, & Martens, 2017b; Pascoe, Arnott, Papoutsakis, Miller, & Andersen, 2007). Death rates as a result, are either calculated quantitatively by measuring differential levels of extracellular LDH accumulation that is released from lysed or apoptotic cells or theoretically by adjusting the net growth rate by a factor correlating to viability decline (Martínez, Bulté, Contreras, Palomares, & Ramírez, 2020; Martins Conde Pdo, Sauter, & Pfau, 2016; Templeton et al., 2017). Although the net growth rate was used in our calculation for theoretical demand, the growth model only predicted important variables up to day 9 in the culture at which point the viabilities of the cultures were relatively similar suggesting minute differences between gross and net growth rates. The resulting stoichiometric balance solution was obtainable by taking the difference between the empirical consumption and the theoretical demand of every amino acid. In all cases however, the underlaying assumption behind stoichiometric balances was that the intracellular metabolism and complex gene expression machinery could be generalized into a “black box” in which extracellular fluxes of nutrients from the media could be directly related to biomass and product generation.
Statistical models in bioprocess have also relied on the “black box” approach in correlating extracellular metabolites and fluxes to cellular phenotypical behavior (Kroll, Hofer, Ulonska, Kager, & Herwig, 2017; Lee & Gilmore, 2006). In addition, multivariate approaches also offer the added benefit of being able to highlight key variables that contribute towards the prediction of specific response variables (Akarachantachote et al., 2013; Gangadharan et al., 2019; Hassan, Farhan, Mangayil, Huttunen, & Aho, 2013). Applied together with the stoichiometric model, the complexity of balancing all 20 amino acids in CD media can be greatly reduced by balancing only the specific amino acids deemed important by the statistical model. Furthermore, statistical models are easier to develop, do not require systematic constraints or curation of biological parameters, and are computationally efficient in determining key correlations (Kim, Rocha, & Maia, 2018; Martins Conde Pdo et al., 2016). More recently however, there has been keen interest on obtaining a deeper understanding of the effects on the metabolic framework by stoichiometric balancing via mechanistic modeling (Sha, Huang, Wang, & Yoon, 2018). Such models can elucidate the intricacies of the intracellular pathways that are extrapolated within statistical models. Thus, mechanistic models can help identify causal linkages within correlation structures. In a modeling review, Traustason et al described several computational methods aimed at optimizing amino acid concentrations for CD media in CHO cell culture. Among them include stoichiometric models such as metabolic flux analysis (MFA) or flux balance analysis (FBA) as well as kinetic models (Traustason, B., Cheeks, & Dikicioglu, 2019). Robitaille et al developed a dynamic model combining the benefits of MFA and kinetic models to explain the relationship between extracellular fluxes of amino acids, extracellular concentrations, and the intracellular flux responses towards that consumption. With a hybrid steady-state and kinetic model, a dynamic relationship between nutrient uptake and cellular response in terms of biomass generation and mAb production was determined (Robitaille, Chen, & Jolicoeur, 2015b).
In another example, combination of FBA and genome-scale modeling with the utilization of amino acid stoichiometric balancing revealed that the addition of nonessential amino acids (NEAAs) have a positive impact on CHO cell biomass production (Traustason, Bergthor, 2019). CHO cells are reported to contain seven NEAAs including glycine, alanine, asparagine, aspartic acid, glutamic acid, proline, and serine (Fomina-Yadlin et al., 2014). These seven amino acids are nutrients that CHO cells can directly biosynthesize whereas the remaining 13 essential amino acids must be obtained by nutrient supplementation from the CD media (Salazar et al., 2016). Presumably, the supplementation of NEAAs can redirect the energy needed from biosynthesis of amino acids to other biological processes such as cell growth or protein synthesis. Moreover, several essential amino acids are also used in biosynthesis for NEAAs and other metabolic intermediates, suggesting that supplementation of NEAAs could not only free up energy but also free up nutrients to support the biosynthesis of other cellular processes (Duarte et al., 2014; Harcum & Lee, 2016). Accordingly, the significant increases of cell growth and mAb productivity that were observed within moderate and low nutrient conditions for both the growth and production models in Criterion 2 supported the energy conservation theory of NEAAs since two of the three amino acids supplemented for cell growth were NEAAs. Supplementation of alanine and glycine also showed increased consumption in all model conditions which also resulted in a decreased specific consumption of glucose further suggesting that the conserved energy was shifted towards cell growth and increased cellular metabolism. In the case of mAb productivity, methionine and lysine was also supplemented. Although not counted within NEAAs, methionine plays an important role in protein production as it is the universal initiator of protein synthesis and thus, crucial to cellular uptake and metabolism to increase protein production. In addition, methionine has shown to improve the cellular redox state of cells by using the ubiquitous methionine sulfoxide cycle to provide an antioxidant defense system (Lim, Kim, & Levine, 2019). Accordingly, we presume that the supplementation of NEAAs, critical protein synthesis amino acids, and increased cell growth all contributed towards the significant increase in mAb productivity within moderate and low nutrient model conditions.
In the case of high nutrients feeds, although there was a marked increase in cellular consumption of almost all amino acids, there was only a minute increase in cell growth and mAb productivity, and we presumed that a saturation point in the culture was reached. However, the notion of a saturation point with respect to extracellular amino acid concentrations was supported by Salazar et al who described that supplementing concentrations of amino acids that are generally highly consumed in mammalian cell cultivation does not always lead to improvement in cell culture, but rather can lead to undesired effects and the accumulation of harmful waste products (Salazar et al., 2016). Similarly, Templeton et al experimentally validated previous MFA models with 13C labeling studies in vitro and showed that protein expression does not always correspond the activation of specific metabolic pathways, but rather, even fluxes tend to eventually saturate (Templeton et al., 2017b). Taken together, high nutrient conditions resulted in greater consumption but maintained similar specific glucose consumption rates suggesting that cellular energy was not repurposed to increased cell growth or mAb productivity, rather, the cells had reached their maximum capacity to metabolize amino acids.