Poster M12 - Computational Modeling of Dissolved Oxygen Transport in Small-Volume Bioreactor Environments for Biomanufacturing

Introduction: Dramatic reductions in operating costs for industrial-level cell culture are necessary to make tissue engineering for therapeutics and cellular agriculture (lab-grown meat) economically viable. Cell-based bioproduction leverages systems like stir-tank bioreactors, which employ mechanical agitation to regulate media oxygenation. This process has substantial roadblocks to scalability and sustainable production, originating from high operational and R&D costs, energy-inefficiency, and the necessity of continuous media and nutrient input and monitoring of gas exchange. Costs for design, testing, and optimization of bioreactors can be reduced by employing scaled-down bioreactor systems that allow evaluation of different operating parameters in a high-throughput, benchtop setting. However, dramatic differences in operating volume contribute to observed differences in cell viability and oxygenation under otherwise similar growth conditions, which we hypothesized is partly due to differences in oxygen transfer across the air-liquid interface (kLa). Our objective with this project was therefore to model and validate dissolved oxygen transport within systems of two different volumes (specifically, 1mL microtiter plates and 50mL TPP tubes). This study represents a first step toward a computational model of cell growth parameters that can be employed to optimize bioproduction in silico.

Materials and

Methods: To measure air-liquid interface shape, we mounted a deep 96-well plate on an orbital shaker and recorded interfacial areas with a 1000-fps camera (Figure 1A-B). Water, and later basal media (DMEM/F12), was introduced into wells at 3 volumes (200L, 500L, 1000L) and at different shaking velocities (Static, 100, 200, 300, 400, 500 RPM), at a constant throw length of 19 mm. A high-throughput oxygen sensor was similarly mounted on the shaker and used to record oxygen diffusion in and out of water and media in a pure-nitrogen environment. This data was compared directly to predictions from several computational models with a range of fidelities.

Results, Conclusions, and Discussions: Interface vertical height for water did vary significantly with respect to fluid volume (at 500mL) and mechanical agitation (>300RPM). Basal media reflected similar results, but with more variability with respect to fill volume. Oxygen mass transfer rate began to appreciably increase at 300RPM, consistent with interfacial area data (Figure 1C). Computational modeling reflected these results with reasonable accuracy, and ongoing work is focused on refinement of the model. The model system provides a good estimation of kLa across the parameter space (Figure D and E). Computational modeling obtained good agreement with experimental measurements for kLa, but further examination and incorporation of operating conditions (temperature, mechanical vibration, throw-length amplitude, surface tension) and extension toward larger geometries to further refine the model are necessary.

Acknowledgements (Optional):