Process scale-up in the biopharmaceutical industry is complex and empirical, because of the presence of living cells used to manufacture the desired product. Experimentation is increasingly difficult as the scale increases, due to regulatory and cost constraints, so there is a heavy reliance on scaling calculations and small-scale experiments. Simplified mixing experiments are often conducted in conjunction with standard bulk calculations to derive physical characteristics of each scale, such as volume-to-volume-per-minute (VVM), Power to Volume (P/V), Tip Speed, mass transfer coefficient (kLa), maximum shear stress, and recirculation time.
We scaled up one of our processes from the 2L to the 10,000L bioreactor initially by maintaining the VVM and P/V ratios, which resulted in a 30% drop in Maximum Viable Cell Density (VCD) with a 35% drop in productivity from the 2L to the 10,000L scale. Since we have chosen the 2L bioreactor to be our main development scale we wanted to increase our understanding about the differences between the 2L and 10,000L scales. In order to assist us, we are beginning to use Computational Fluid Dynamics (CFD) as an additional tool, allowing us to explore the internal fluid dynamics within each tank with more spatial and temporal detail. Our relatively simple CFD models are analyzed against directly-comparable physical experiments (where possible) to validate the results.
Thus far, we have determined that mixing times of liquids at large scale were 8 times longer and exhibited relatively long-lived regions of high concentrations of added components. This result is important for our processes, where significant quantities of non-physiological solutions (e.g., concentrated nutrient feeds and base) are routinely added to the reactor. We believe that the slower mixing in the large-scale bioreactors permits zones of different concentrations of components (“mini –environments”) to exist when liquids are added to the bioreactor. These “mini-environments” certainly contain pH gradients, may exhibit hypoxic characteristics, likely have altered Carbon Dioxide levels, and/or contain elevated Osmotic environments. This indicates that scaling biological processes may require more than just maintaining bulk physical characteristics but instead rely on scaling mini-environments to allow for adequate gas exchange and liquid-liquid mixing. Overall, this work highlights our initial phase of development to enhance biological process scale-up using customizable CFD software.