A genome-scale nutrient minimization forecast algorithm for controlling essential amino acid levels in CHO cell cultures.

Mammalian cell culture processes rely heavily on empirical knowledge in which process control remains a challenge due to the limited characterization/understanding of cell metabolism and inability to predict the cell behaviors. This study facilitates control of Chinese hamster ovary (CHO) processes through a forecast-based feeding approach that predicts multiple essential amino acids levels in the culture from easily acquired viable cell density data. Multiple cell growth behavior forecast extrapolation approaches are considered with logistic curve fitting found to be the most effective. Next, the nutrient-minimized CHO genome-scale model is combined with the growth forecast model to generate essential amino acid forecast profiles of multiple CHO batch cultures. Comparison of the forecast with the measurements suggests that this algorithm can accurately predict the concentration of most essential amino acids from cell density measurement with error mitigated by incorporating off-line amino acids concentration measurements. Finally, the forecast algorithm is applied to CHO fed-batch cultures to support amino acid feeding control to control the concentration of essential amino acids below 1-2 mM for lysine, leucine, and valine as a model over a 9-day fed batch culture while maintaining comparable growth behavior to an empirical-based culture. In turn, glycine production was elevated, alanine reduced and lactate production slightly lower in control cultures due to metabolic shifts in branched-chain amino acid degradation. With the advantage of requiring minimal measurement inputs while providing valuable and in-advance information of the system based on growth measurements, this genome model-based amino acid forecast algorithm represent a powerful and cost-effective tool to facilitate enhanced control over CHO and other mammalian cell-based bioprocesses.

Engineering Improved CAR T Cell Products with A Multi-Cytokine Particle Platform for Hematologic and Solid Tumors.

Despite the remarkable clinical efficacy of chimeric antigen receptor (CAR) T cells in hematological malignancies, only a subset of patients achieves a durable complete response (dCR). DCR has been correlated with CAR T cell products enriched with T cells memory phenotypes. Therefore, reagents that consistently promote memory phenotypes during the manufacturing of CAR T cells have the potential to significantly improve clinical outcomes. A novel modular multi-cytokine particle (MCP) platform is developed that combines the signals necessary for activation, costimulation, and cytokine support into a single "all-in-one" stimulation reagent for CAR T cell manufacturing. This platform allows for the assembly and screening of compositionally diverse MCP libraries to identify formulations tailored to promote specific phenotypes with a high degree of flexibility. The approach is leveraged to identify unique MCP formulations that manufacture CAR T cell products from diffuse large B cell patients   with increased proportions of memory-like phenotypes MCP-manufactured CAR T cells demonstrate superior anti-tumor efficacy in mouse models of lymphoma and ovarian cancer through enhanced persistence. These findings serve as a proof-of-principle of the powerful utility of the MCP platform to identify "all-in-one" stimulation reagents that can improve the effectiveness of cell therapy products through optimal manufacturing.