A dynamic metabolic flux analysis (DMFA) model for performance predictions of diverse CHO cell culture process modes and conditions

Abstract

Bioreactor pH can significantly affect Chinese Hamster Ovary (CHO) cell metabolism, thus impacting glycoprotein titers. However, there is very limited literature on incorporating pH in mathematical models for CHO cell metabolism. To address this limitation, guided by recently published experimental data, we have curated a stoichiometric network and formulated phenotype-driven kinetic expressions to develop a Dynamic Metabolic Flux Analysis (DMFA) model. The DMFA model incorporates Critical Process Parameters (CPPs), notably bioreactor pH, basal and feed media nutrient composition, feeding times, and inoculation cell densities to predict bioreactor performance: cell growth rates, antibody titers, and nutrient and metabolite profiles. The DMFA model was trained on diverse fed-batch data of the CHO VRC01 cell line to regress the kinetic parameters. The models utility was demonstrated through experimentally validated model predictions of CHO-cell performance in intensified fed-batch cultures, perfusion cultures, and cultures with different media. Experimentally validated predictions of a culture with high initial cell density and increased feed addition (intensified fed-batch culture) showed that mAb titers similar to fed-batch culture can be achieved with shorter culture durations. Similarly, experimentally validated predictions of perfusion bioreactor performance showed that coupling historical fed-batch data with computational tools can be leveraged to predict continuous biomanufacturing performance. We thus demonstrate that the developed mathematical model can simulate culture performance outside of the training data set. This supports the predictive robustness of the framework and provides a valuable tool for bioprocess development of diverse culture modes. HighlightsO_LIExperimentally measured fed-batch cell culture data was used to curate a reaction network. This reaction network was integrated with phenotypically driven kinetic expressions to yield a dynamic metabolic flux analysis (DMFA) model. C_LIO_LIThe DMFA model can predict bioprocess performance indicators such as concentration of viable cells, mAb, amino acids, glucose, lactate, and ammonia. C_LIO_LIThe model was developed to make these predictions under various process conditions such as bioreactor pH, media concentrations, feed supplementation schedule, and initial cell densities. C_LIO_LIPredicting and experimentally validating the impact of high initial cell density and increased feed media supplementation yielded in mAb titers similar to traditional fed-batch processes with much shorter culture durations. C_LIO_LIThe application of the DMFA model trained on data from a traditional fed-batch process to predict perfusion bioreactor culture performance was successfully demonstrated and experimentally verified. C_LIO_LIThe impact of AMBIC reference media on cell culture process performance was also predicted and experimentally validated. The predictions of amino acid metabolism yielded insights into improving the media. C_LI