Multi-omics characterization of mesenchymal stem/stromal cells for the identification of putative critical quality attributes

Maughon TS, Shen X, Huang D, Michael AOA, Shockey WA, Andrews SH, McRae JM, Platt MO, Fernández FM, Edison AS, Stice SL, Marklein RA.
Preprint from
11 May 2021


Mesenchymal stromal cells (MSCs) have shown great promise in the field of regenerative medicine as many studies have shown that MSCs possess immunomodulatory function. Despite this promise, no MSC therapies have been granted licensure from the FDA. This lack of successful clinical translation is due in part to MSC heterogeneity and a lack of critical quality attributes (CQAs). While MSC Indoleamine 2,3-dioxygnease (IDO) activity has been shown to correlate with MSC function, multiple CQAs may be needed to better predict MSC function.


Three MSC lines (two bone marrow, one iPSC) were expanded to three passages. At the time of harvest for each passage, cell pellets were collected for nuclear magnetic resonance (NMR) and ultra-performance liquid chromatography mass spectrometry (UPLC-MS), and media was collected for cytokine profiling. Harvested cells were also cryopreserved for assessing function using T cell proliferation and IDO activity assays. Linear regression was performed on functional and multiomics data to reduce the number of important features, and partial least squares regression (PLSR) was used to obtain putative CQAs based on variable importance in projection (VIP) scores.


Significant functional heterogeneity (in terms of T cell suppression and IDO activity) was observed between the three MSC lines, as well as donor-dependent differences based on passage. Omics characterization revealed distinct differences between cell lines using principal component analysis (PCA). Cell lines separated along principal component 1 based on tissue source (bone marrow vs. iPSC-derived) for NMR, MS, and cytokine profiles. PLSR modeling of important features predicts MSC functional capacity with NMR (R 2 =0.86), MS (R 2 =0.83), cytokines (R 2 =0.70), and a combination of all features (R 2 =0.88).


The work described here provides a platform for identifying putative CQAs for predicting MSC functional capacity using PLSR modeling that could be used as release criteria and guide future manufacturing strategies for MSCs and other cell therapies.