Exploring CHO cell stability during prolonged passaging via eXplainable AI driven flux balance analysis

Production stability remains a major challenge in Chinese hamster ovary (CHO) cell-based therapeutic protein manufacturing, particularly during extended passaging where the underlying mechanisms of instability are not fully understood. Thus, in this study, we leveraged multivariate data analysis (MVDA) and flux balance analysis (FBA) with explainable AI (xAI) to mechanistically characterize the phenotypic differentiation between early (EP) and late passage (LP) of CHO cultures. Although EP and LP reached comparable peak viable cell densities, LP cultures exhibited a ~35% reduction in peak IgG titers and increased lactate and ammonia accumulation. Subsequent MVDA of temporal exometabolite profiles identified the exponential growth phase as the primary window of divergence, allowing us to interrogate metabolic rewiring via an FBA-xAI approach. This revealed that EP cells preferentially directed acetyl-CoA towards fatty acid biosynthesis to support proliferation. In contrast, LP prioritized oxidative stress mitigation by upregulating the trans-sulfuration pathway for de novo cysteine and glutathione synthesis while exhibiting heightened TCA cycle activity to maintain energy homeostasis. Overall, these mechanistic insights uncover a passage-associated shift from biosynthetic activity toward redox maintenance and identify the cysteine-glutathione axis as a critical metabolic lever for enhancing long-term stability and productivity in CHO cell culture.