Mesenchymal stromal cells (MSCs) have been widely used in regenerative medicine applications due to their immunomodulatory properties. However, there are currently no licensed MSC therapies by the Food and Drug Administration (FDA). This can be attributed, in part, to the functional heterogeneity from different MSC donor and tissue sources and lack of critical quality attributes (CQA) for manufacturing MSC therapies. MSC metabolism during the manufacturing process plays a critical role in the potency of these therapies. Metabolites are highly abundant and reflect the cellular phenotype making them ideal candidates for identifying CQAs. This dissertation aims to identify metabolites, both in-process and end of manufacturing, to be used as candidate CQAs that are predictive of MSC immunomodulation. First, three MSC lines were expanded to three passages with metabolic profiling and functional testing (CD4+ and CD8+ T cell proliferation from different two donors and indoleamine-2,3-dehydrogenase (IDO) activity) at the end of each passage. A composite functional score was developed using all five functional metrics for potency prediction. Partial least squares regression identified candidate CQAs predictive of function including several small polar molecules and phosphatidylcholines. Lastly, ten MSC lines were expanded with profiling of non-destructive, in-process media metabolites and end of expansion intracellular metabolites. Using a robust consensus machine learning approach, metabolites predictive of MSC function were identified. Metabolites found in multiple machine learning models were used to build consensus models. Consensus intracellular metabolites included multiple lipid classes while consensus media metabolites included several amino acids and sugars. Pathway enrichment identified metabolic pathways significantly associated with MSC function such as sphingolipid signaling and metabolism, arginine and proline metabolism, and autophagy. Overall, this work establishes a framework for identifying consensus metabolites that predict MSC function and can be used as candidate CQAs to help guide MSC manufacturing.INDEX WORDS: mesenchymal stromal cells, metabolomics, critical quality attribute, potency, immunomodulation, t-cell, machine learning, cell manufacturing
