High-Throughput On-Chip Human Mesenchymal Stromal Cell Potency Prediction

Abstract

Human mesenchymal stromal cells (hMSCs) are a promising source for regenerative cell therapy. However, hMSC clinical use has been stymied by product variability across hMSC donors and manufacturing practices resulting in inconsistent clinical outcomes. The inability to predict hMSC clinical efficacy, or potency, is a major limitation for market penetration. Standard metrics of hMSC potency employ hMSCs and third-party immune cell co-cultures, however, these assays face translational challenges due to third-party donor variability and lack of scalability. While surrogate markers of hMSC potency have been suggested, none have yet had translational success. To address this, a high-throughput, scalable, low-cost, on-chip microfluidic potency assay is presented with improved functional predictive power and recapitulation of in vivo secretory responses compared to traditional approaches. Comparison of hMSC secretory responses to functional hMSC-medicated immune cell suppression demonstrates shortcomings of current surrogate potency markers and identifies on-chip microfluidic potency markers with improved functional predictive power compared to traditional planar methods. Furthermore, hMSC secretory performance achieved in the on-chip microfluidic system has improved similarity compared to an in vivo model. The results underscore the shortcomings of current culture practices and present a novel system with improved functional predictive power and hMSC physiological responses.

Keywords: biomaterials; cell therapies; mesenchymal stem/stromal cells; microfluidics; on-chip technologies.

Figure 1.. Healthy hMSC donors demonstrate various T cell suppressive ability.

(a) Nine donors from three manufactures were expanded as per manufacturer protocol. (b) Population doubling level (PDL) reported from time of receipt of cell line. (c) Microscope phase contrast images of each hMSC donor line. (d) Schematic of T cell suppression protocol. (e) CD3⁺ T cell suppression for various hMSC dose normalized to activated PBMC-only control (dotted line). One PBMC donor shown across three repeated experiments. Two-way ANOVA comparing cell ratio (or dose) mean to activated PBMC-only control. ****P<0.0001; n≥38 across 9 hMSC donors. All data represented as means ± SEM.

Figure 2.. Evaluation of IDO as a potency metric in 2D IFN-γ system.

(a) Schematic of 2D experimental design. (b) Image flow cytometry of brightfield, intracellular IDO, and surface bound PD-L1. (c) IDO expression (left) and IDO activity (right) upregulated from 2D Ctrl. (d) Linear regression analysis of suppression index to IDO expression (left) and IDO activity (right). (e) Summary linear regression R2 of suppression index to IDO (left) and PD-L1 (right) metrics. (f) Linear regression of 2D IFN-γ IDO expression and activity. Mean fluorescent intensity (MFI) normalized to 2D ctrl (MFI MFI⁻¹) for comparison across repeated experiments. Bar graphs: Two-tailed unpaired t test with Welch’s correction used. ***P<0.0005, ****P<0.0001, dotted red line as 2D ctrl mean. Linear regression: significant for P<0.05. Solid black lines are best fit lines, dotted black lines as 95% confidence bands. Summary linear regression: Corresponding *P<0.0332. **P<0.0021; n=44 across 9 donors. All data represented as means ± SEM.

Figure 3.. Microfluidic system informed from secretion of hMSCs delivered in vivo.

(a) Schematic of microfluidic synthesis including hydrogel crosslinking, cell encapsulation, and pressure-driven media perfusion. (b) Image of microfluidic chip indicating inlet and outlet channels, PDMS base, and hydrogel location. (c) Iterative approach for development of analyte panel and design parameters informed from subcutaneous delivered hMSC-laden hydrogel mouse model.

Figure 4.. Evaluation of IDO and PD-L1 as a potency metric in microfluidic IFN-γ system.

(a) Schematic of microfluidic IFN-γ experimental design. (b) Heat map and two-way hierarchal clustering of analyte secretion across 2D IFN-γ and microfluidic IFN-γ culture systems. Clustering by Ward Method (pg pg⁻¹). (c) IDO expression (left) and IDO activity (right) compared to 2D Ctrl. (d) Linear regression analysis of suppression index to IDO expression (left) and IDO activity (right). (e) Summary linear regression R² of suppression index and IDO (left) and PD-L1 (right) metrics. (f) Linear regression of microfluidic IFN-γ IDO expression and activity. Mean fluorescent intensity (MFI) normalized to 2D ctrl (MFI MFI⁻¹) for comparison across repeated experiments. Bar graphs: Two-tailed unpaired t test with Welch’s correction used. ****P<0.0001, dotted red line as 2D Ctrl mean. Linear regression: significant for P<0.05. Summary linear regression: Corresponding *P<0.0332; n=38 across 9 donors. All data represented as means ± SEM.

Figure 5.. Comparison of analyte secretion in 2D IFN-γ and microfluidic IFN-γ systems.

(a) Summary linear regression R² for suppression index values, of various T cell subsets and hMSC dose, and secretion levels of 20 analytes across both 2D IFN-γ and microfluidic IFN-γ culture systems. (b) Summary linear regression R² for suppression index and MMP-13 secretion. (c) Top five R² values for CD3⁺ 1:2 suppression index for all metrics both 2D IFN-γ and microfluidic IFN-γ systems. (d) Contrasting correlative trends of suppression index (CD4⁺ 1:4) and TNF R1 secretion in microfluidic IFN-γ or 2D IFN-γ systems. (e) Summary linear regression R² (left) and corresponding best fit slopes (right) showing variable correlative confidence but consistent contradicting trends between culture systems. Analyte secretion (pg mL⁻¹) normalized to 2D Ctrl (pg pg⁻¹) for comparison across repeated experiments. Summary graph: *P<0.0332, **P<0.0021, ****P<0.0001 for corresponding P values; (+) positive slope, (−) negative slope. Linear regression analysis: black solid lines as best fit lines, dotted lines as 95% confidence bands; correlation significant for P<0.05. Microfluidic IFN-γ n=38, 2D IFN-γ n=44, across 9 donors. All data represented as means ± SEM.

Figure 6.. In vivo signaling pathways recapitulated in microfluidic IFN-γ but not 2D IFN-γ systems.

(a, b) Positive correlation in vivo between local inflammatory signals IFN-γ and (a) TNF-α and (b) TNF R1. (c-e) Inflammatory signaling pathways recapitulated in microfluidic IFN-γ system, but not 2D IFN-γ system, compared to in vivo-delivered hMSCs for TNF-α and (c) TNF R1 (d) IL-17E and (e) MMP-13. Solid black lines as best fit lines, dotted lines as 95% confidence bands; correlation significant for P<0.05. Microfluidic IFN-γ n=38 and 2D IFN-γ n=44 across 9 donors; in vivo n≥30 across 3 donors.