A Predictive Model for Energy Metabolism and ATP Balance in Mammalian Cells: Towards the Energy-Based Optimization of mAb Production
Introduction
The pharmaceutical industry has developed complex therapeutic drugs to meet the medical needs of patients in recent decades. The leading segment of biologic drugs is monoclonal antibodies. Biopharmaceutical industry forecasts approximately 70 mAbs approved by 2020 with a global value of $125bn (Eckera et al., 2015). About 50% of the approved biopharmaceuticals are produced by mammalian cell culture (Zhu et al., 2012), mainly due their capacity to produce active molecules with humanlike post-translational modifications (PTM) identical to native endogenous proteins, including glycosylation, formation of disulphide bonds and proteolytic processing. All these characteristics are known to be essential for the biological function and pharmacokinetics of the final product (Berlec et al., 2013). Despite their success in industry, mammalian cells cultures present serious disadvantages such as low yield. Model-based optimization techniques could be applied to improve the mAb productivity. Alas, current metabolic models of GS systems do not consider vital metabolites such as amino acids and energy requirements of cell proliferation, maintenance and mAb production. Most recent studies take into account glucose and glutamate requirements but none of them consider the energy requirements and the effect of the depletion of energy sources.
Energy production depends mainly on glycolysis and the TCA cycle. In contrast, cells consume ATP for proliferation (biosynthesis and polymerization reactions) and for non-proliferation-associated processes such as maintenance (concentration and electrical gradients) as well as mAb production. Xie and Wang (1996) extensively studied energy metabolism in mammalian cells, developing a detailed material-balance model of animal cell metabolism based on a stoichiometric reaction network using experimental data from batch and fed-batch cultures of hybridoma cells. In this attempt, the estimated biosynthetic ATP demand was lower than the reciprocal of the maximum ATP yield determined by the relationship between the specific ATP production rate and cell growth rate. Additionally, it missed non-growth-associated ATP demand and so the final ATP balance was not achieved.
This study provides the calculation of the total ATP production and consumption, enabling the development of a novel dynamic model that predicts ATP balance in mammalian cell cultures. Herein, we present the first step towards an energy model based optimization that would allow the derivation of an optimal feeding profile that ensures nutrient and energy supply, eradicating the excessive presence of glucose and the accumulation of lactate, resulting in prolonged culture viability and the maximization of the mAb titer.
Section snippets
Cell line and culture conditions
GS-NS0 mouse myeloma cells were cultured in triplicate 1 L Erlenmeyer flasks (Corning) with 200 mL working volume, agitated at 130 rpm, incubated at 37 °C and 5% CO2. The media contained Advanced-DMEM X1 (Invitrogen Ltd.), MEM Non-essential amino acids (Sigma–Aldrich) X2, MEM-Essential amino acids (Sigma –Aldrich) X2, GS-Supplement (Sigma-Aldrich) X1, MEM-Vitamins (Gibco) X1, Penicillin/Streptomycin (Gibco) X1, 5 mg/L MSX (Sigma-Aldrich) and 10% Dialyzed Fetal bovine serum (Gibco). The monoclonal
Mathematical Model Development
Model categories are structured/unstructured and segregated/non-segregated. The former classification consists in the internal compartmentalization of the cell, the model explain processes that take place in different unit inside the cell. The latter classification is related to the heterogeneity of cell culture population. If a cell culture is segregated, it is composed of cells in different stages of growth. Conversely, an unsegregated model assumes a homogeneous cell population (Sidoli et
Global sensitivity analysis and parameter estimation
The model consists of 21 differential and 66 algebraic equations, 87 variables and 44 parameters. The model was simulated in gPROMS ModelBuilder ® v.4.1.0 and subjected to Global Sensitivity Analysis (GSA) (Li et al., 2002) using a sampling time of 2 hours and a variation of ± 50%. The GSA results indicate 14 significant parameters, which were re-estimated using experimental data (Exp 1).
Results and Discussion
The cultures demonstrated a short lag phase and exponential growth phase that extends until 48 hours of culture time. The model successfully predicted concentrations of viable, apoptotic and dead cells; glucose, lactate, glutamate, arginine, histidine, aspartate, asparagine, lysine, isoleucine, leucine, methionine, valine and threonine, monoclonal antibody and ATP. The exhaustion of glutamate and aspartate (Figure 2 (D)) has a critical effect on the growth rate and the cells start the stationary
Conclusions
We have successfully developed a dynamic novel model that couples dynamic equations and stoichiometric coefficients to calculate the ATP balance in mammalian cells and estimates the consumption of glucose and 13 amino acids. The next step is to include the ATP concentration effect in the viable cells balance and implement a modelbased optimization to design a feeding strategy that maintains cellular metabolism in energy-efficient pathways to maximize the mAb titer.
References (5)
Mammalian cell protein expression for biopharmaceutical production
Biotechnology advances.
(2012)- et al.
Current state and recent advances in biopharmaceutical production in Escherichia coli, yeasts and mammalian cells
Journal of Industrial Microbiology & Biotechnology.
(2013)
Cited by (4)
Sequential and Simultaneous Optimization Strategies for Increased Production of Monoclonal Antibodies
2019, Computer Aided Chemical EngineeringCitation Excerpt :More precisely, by solving the optimization problems locally, we end up with suboptimal solutions, the performance of which (as shown) can be strongly affected by good initial guesses. This work applies dynamic optimization studies to a predictive mathematical model developed by Quiroga et al. (2016) and numerically improved by Kappatou et al. (2018). The main scope of the study is to develop computationally efficient and biologically meaningful optimization strategies for mAb process intensification.
Dynamic Optimization of the Production of Monoclonal Antibodies in Semi-batch Operation
2017, Computer Aided Chemical EngineeringCitation Excerpt :The first one refers to optimization of the culture media and the second one to optimization of the feeding strategy. A model-based medium composition has been developed using the model in Quiroga et al. (2016). Therefore, we focused here on the latter approach, namely the derivation of a feeding profile that leads to optimal operation.
Model-Based Dynamic Optimization of Monoclonal Antibodies Production in Semibatch Operation - Use of Reformulation Techniques
2018, Industrial and Engineering Chemistry Research