(234a) Metabolite Monitoring of Mesenchymal Stromal Cell Cultures By Spontaneous Raman Spectroscopy

Multipotent Mesenchymal Stromal Cells (MSCs) based therapies show great promise in treating a host of diseases and injuries such as GvHD, arthritis, spinal cord injury, and many more [1]. During the past decade, ten MSC therapies have received regulatory approval including recent approvals in Europe for MSC treatment of complex perianal fistulas related to Crohn’s disease [2, 3]. With more therapies emerging from clinical trials and being approved for use, the research focus will eventually shift from their efficacy and safety to their manufacturing process and quality control. The intrinsic heterogeneity of the cells, their biological complexity and the multiple processing steps that are required to guarantee a reliable product hinder the translation of many MSC therapies into approved biologics.

Cell therapy manufacturing processes often include off-line measurements of cell densities and metabolic markers during limited time points. Very few processes include real-time measurements, and those are usually limited to cells such as CAR-T which are grown in suspension and reach fairly high cell densities [4]. MSC cultures with their relatively low cell densities, long doubling time and the resulting low metabolic marker concentrations, present a challenge for on-line monitoring and control. The ability to provide continuous real-time measurements could allow faster screening of donors, improved process control and therapy quality.

In this study, we use Raman spectroscopy as an optical sensor for bioprocess monitoring of a primary MSC culture. Samples of conditioned media were acquired every 2-3 days from MSCs cultured under four different feeding conditions with three technical replicates for each condition. Using reference datasets generated on a standard bioanalyzer for three of the process conditions, we train chemometric models (Partial Least Square’s regression) from the Raman spectra for glucose and lactate and demonstrate prospective prediction on an independent validation data sets with the fourth set of process conditions.

These chemometric models can monitor nutrient consumption and metabolite production with sufficient accuracy to account for small changes in glucose consumption and lactate secretion, despite the fetal bovine serum Raman background which is found in the cells’ media. For three of the four validation sets, the glucose PLS model produced a fit strength (R²) of 0.73-0.85 and a prediction Root Mean Squared Error (RMSE) of 0.09-0.18 [g/l]. For lactate R² was 0.79-0.9 while the RMSE was 0.18-0.26 [g/l]. The last data set, in which feeding protocol and media exchange time point were distinct from the prior three datasets, the PLS models produced inferior results.

These results indicate the possibility of using Raman spectroscopy for control of feeding strategies in both research and manufacturing settings of MSC cell therapies. Implementing real-time adaptive changes in the process based on specific donor performance and culture conditions could provide the next step in personalized cell based therapies.

References

1. Parekkadan, Biju, and Jack M Milwid. “Mesenchymal stem cells as therapeutics.” Annual review of biomedical engineering 12 (2010): 87-117
2. Levy, Oren et al. “Shattering barriers toward clinically meaningful MSC therapies.” Science advances 6,30 eaba6884. 22 Jul. 2020,
3. Gallo, Gaetano et al. “Mesenchymal Stromal Cell Therapy in the Management of Perianal Fistulas in Crohn's Disease: An Up-To-Date Review.” Medicina (Kaunas, Lithuania) 56,11 563. 27 Oct. 2020
4. Baradez, Marc-Olivier et al. “Application of Raman Spectroscopy and Univariate Modelling As a Process Analytical Technology for Cell Therapy Bioprocessing.” Frontiers in medicine vol. 5 47. 5 Mar. 2018