Scaling strategy for cell and gene therapy bioreactors based on turbulent parameters.

So far, power input has been used as the main parameter for bioreactor scale-up/-down in upstream process development and manufacturing. The rationale is that maintaining a consistent power input per unit volume should result in comparable mixing times at different scales. However, shear generated from turbulent flow may compromise the integrity of non-robust cells such as those used during the production of cell and gene therapies, which may lead to low product quality and yield. Of particular interest is the Kolmogorov length parameter that characterizes the smallest turbulent eddies in a mixture. To understand its impact on scale-up/-down decisions, the distribution of Kolmogorov length along the trajectory flow of individual particles in bioreactors was estimated in silico with the help of computational fluid dynamics simulations. Specifically, in this study the scalability of iPSC-derived lymphocyte production and the impact of shear stress across various differentiation stages were investigated. The study used bioreactors of volumes from 0.1 to 10 L, which correspond to the scales most used for parameter optimization. Our findings, which align with in vitro runs, help determine optimal agitation speed and shear stress adjustments for process transfer between scales and bioreactor types, using vertically-oriented wheel and pitched-blade impellers. In addition, empirical models specific to the bioreactors used in this study were developed. The provided computational analysis in combination with experimental data supports selection of appropriate bioreactors and operating conditions for various cell and gene therapy process steps.

Error propagation in constraint-based modeling of Chinese hamster ovary cells.

Chinese hamster ovary (CHO) cells are the most popular mammalian cell factories for the production of glycosylated biopharmaceuticals. To further increase titer and productivity and ensure product quality, rational system-level engineering strategies based on constraint-based metabolic modeling, such as flux balance analysis (FBA), have gained strong interest. However, the quality of FBA predictions depends on the accuracy of the experimental input data, especially on the exchange rates of extracellular metabolites. Yet, it is not standard practice to devote sufficient attention to the accurate determination of these rates. In this work, we investigated to what degree the sampling frequency during a batch culture and the measurement errors of metabolite concentrations influence the accuracy of the calculated exchange rates and further, how this error then propagates into FBA predictions of growth rates. We determined that accurate measurements of essential amino acids with low uptake rates are crucial for the accuracy of FBA predictions, followed by a sufficient number of analyzed time points. We observed that the measured difference in growth rates of two cell lines can only be reliably predicted when both high measurement accuracy and sampling frequency are ensured.

Dissecting peak broadening in chromatography columns under non-binding conditions.

Peak broadening in small columns is dominated by spreading in the extra column volume and not by hydrodynamic dispersion or mass transfer resistances. Computational fluid dynamics (CFD) permits to study the influence of these effects separately. Here, peak broadening of three single component solutes - silica nanoparticles, acetone, and lysozyme - was experimentally determined for two different columns (100 mm × 8 mm inner diameter and 10 mm × 5 mm inner diameter) under non-binding conditions. A mass transfer model between mobile and stationary phases as well as a hydrodynamic dispersion model were implemented in the CFD environment STAR-CCM+®. The mass transfer model combines a model of external mass transfer with a model of pore diffusion. The model was validated with experiments performed on the larger column. We find that extra column volume plays an important role in peak broadening of the silica nanoparticles pulse in that column; it is less important for acetone and is weakly pronounced for lysozyme. Hydrodynamic dispersion plays the dominant role at low and medium flow rates for acetone because we are in a regime of 1-10 ReSc. Mass transfer is important for high flow rates of acetone and for all flow rates of lysozyme. Then, peak broadening was predicted in the smaller column with the packed bed parameters taken from larger column. The scalability of the prepacked columns is demonstrated for acetone and silica nanoparticles by excellent agreement with the experimental data. In contrast to the larger column, peak broadening in the smaller column is dominated by extra column volume for all solutes. Peak broadening of lysozyme is controlled only at high flow rates by mass transfer and overrides extra column volume and hydrodynamic dispersion. CFD simulations with implemented mass transfer models successfully model peak broadening in chromatography columns taking all broadening effects into consideration and therefore are a valuable tool for scale up and scale down. Our simulations underscore the importance of extra column volume.