(683c) Using High-Throughput Microbioreactors to Fit Raman Spectroscopy Model for Measuring Key Metabolites in Monoclonal Antibody Cell Culture Manufacturing Process

Raman spectroscopy is a process analytical technology for simultaneous in-line measurement of important process outputs in therapeutic protein production via mammalian cell culture, such as concentrations of key metabolites or even the protein being expressed. However, due to the complex and shifting nature of the various chemical species present in the cell culture system, Raman models will likely need to be at least partially trained on a process-specific basis. They must also be robust enough to maintain accuracy throughout the expected manufacturing process operational ranges. This presentation is a proof-of-concept study to demonstrate a reliable Raman model can be developed in parallel with standard process development experimentation for later deployment in a scaled-up manufacturing process.

Sartorius ambr® 250 high-throughput microbioreactors were used for cell culture process development studies for a CHO cell line expressing an IgG4 monoclonal antibody (“mAb1”). Process parameters including initial cell density, nutrient feed strategy, pH, and dissolved oxygen control for a fourteen-day bioprocess were examined, and standard off-line monitoring of all process outputs were carried out. Concurrently during this high-throughput study, Raman spectra were also periodically generated from bioreactor samples both before and after being spiked with known concentrations of key metabolites. These data were used to train a Raman model and verified against similar spectra obtained from larger bioreactors.

In conclusion, this study generated a scalable Raman spectroscopy model for various metabolites that can be used for monoclonal antibody process control and performance monitoring. A robust model can be trained over a wide range of processing conditions, in a manner that would not interfere with standard process development or require additional experiments. The comparison of the outputs from this Raman model to the standard off-line data will also be presented.