(582ad) Using Design of Experiment (DOE) Principles Toward Titer Optimization in Perfusion Cell Culture

Over the past several decades tremendous improvements have been made in biotherapeutic protein production through both cell line engineering and media development. Titers of over one gram per liter are expected even for a non-optimized 14-day fed-batch process producing a monoclonal antibody.  However, not all potentially therapeutic molecules are as stable as an IgG, and a continuous perfusion process is required. While the risks and economics of a perfusion process can be debated, perfusion cell culture allows for a unique system to test the effects of specific parameters or media components on culture performance by taking advantage of the steady state nature of perfusion. Unfortunately for process developers, there is no high-throughput scale down model available for perfusion cell culture. In the current study, we work toward developing a valid high-throughput scale down model for specific parameters in perfusion cell culture by evaluating several significant factors.

Initially, batch experiments were conducted using a factorial design for media development. One particular media component, component X, showed a significant improvement in batch titer over a relatively small concentration range.  A dose-response experiment was then conducted to determine the saturation and potential inhibitory or toxic effects of this component. While toxic effects were not seen for concentrations over three logs, saturation was seen at approximately one log increase in starting component concentration, which corresponded to a tripling in product titer.  While an optimum was not determined, a point well into the saturation curve was chosen for further study and use in a perfusion process.  Unlike the batch study, increases in perfusion titer were highly variable.  These variations were later tied to Lot-to-Lot variability in the basal media, which is a major risk factor in perfusion processes.  Not only were the basal levels of component X found to be highly variable (50-150%), it was also determined that the small variability in incoming osmolality and pH correlated well with the variability in titer.  To further explore this correlation, a response surface design was implemented in a batch shake flask model, looking at the impact of starting pH, osmolality and the concentration of component X. Our results indicate a significant impact from both osmolality and the concentration of component X, but not from the starting pH.  To investigate potential mechanisms of increased titer, additional studies were conducted around the mode of osmolality enhancement as well as mock perfusion to better control pH.  Previous studies have shown cell specific productivity to be highly pH sensitive and thus, developing a representative scale down model for perfusion that captures this sensitivity is highly desired.  To that end, a full perfusion study was conducted around pH, osmolality, and the concentration of component X which was then compared with the results from the scale down model.

With this study we have explored the use design of experiment (DOE) techniques to move from component and parameter screening to a scale down model for perfusion processes development.