Compared to perfusion or fed-batch modes, operating a bioreactor in a continuous, extended perfusion mode can enhance the productivity and efficiency of mammalian cell culture processes. Continuous bioreactor operation does, however, increase operational complexity. In this example, the bioreactor process had a growth phase, followed by a production phase, in which the material was continuously harvested from the bioreactor. During development, several run failures were observed in small-scale bioreactors, which were preceded by lactate spikes. The unpredictable yet frequent (&gt;50%) lactate spikes represented a complex multivariable problem that was difficult to troubleshoot.<br/><br/>Partial least squares models were created using cell culture data from 50 small-scale runs to understand the root cause of lactate spikes. Several variables (green) were strongly correlated with lactate level (blue) in the growth phase, as shown in the multivariate plot. One of these was a controllable process input, cell-specific perfusion rate (CSPR) (red). The perfusion rate was decreased to reduce CSPR of the culture. System-induced oscillations in dissolved oxygen (DO) were also correlated with lactate spikes. Analysis demonstrated that the minimum value of DO was lower (&lt;30 mmHg) in the failed runs than the successful runs. The DO setpoint was increased to avoid hypoxia and the bioreactor control system was upgraded to reduce DO oscillations. These changes to perfusion and DO control eliminated lactate spikes and established a reliable small-scale model for future process characterization studies.

Multivariate methods for troubleshooting continuous bioreactor process

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.