Spatial transcriptomics-guided identification of macrophage-fibroblast crosstalk reveals CD47-targeted extracellular vesicles as a selective delivery platform for MASH-associated liver fibrosis

초록

Introduction Metabolic dysfunction-associated steatohepatitis (MASH) affects over 100 million individuals globally and represents a significant unmet need due to the lack of therapies that reverse established fibrosis. Current antifibrotic approaches targeting biochemical pathways such as TGF-β and RTK have shown limited clinical success, with over 83% failing in phase 2-to-3 transitions. These repeated clinical failures indicate that fibrosis progression may be governed not only by dysregulated signaling pathways, but also by biomechanical and spatial cues that are not addressed by conventional pharmacologic strategies. Accordingly, we hypothesized that the physical microenvironment—specifically matrix stiffening and cell-cell interactions—drives the fibrotic "execution" program, and that targeted delivery to the pathogenic cellular niche is essential for therapeutic efficacy. Methods We performed an integrated spatial transcriptomics analysis using Visium and Xenium platforms on human liver tissues across disease stages (Control, MASLD, MASH). In Visium data, cell-cell interaction (CCI) signals were quantified using CellNeighborEX v2, which identifies interaction-induced gene expression by computing residuals between observed spatial expression and single-cell reference-based predictions. Visium-derived CCI genes were categorized using two complementary databases: Adhesome/Reactome (ECM signaling potential) and MatrixDB (physical ECM interactions). Among these, 126 MASH-specific ECM– CCI genes present in the Xenium panel were selected to model fibroblast state transitions by pseudotime analysis, with a focus on interactions with neighboring macrophages. Based on these spatially defined niches, SIRPα-engineered extracellular vesicles (SIRP-EV) were developed from MSCs (mesenchymal stem cells) and labeled with Cy5.5 for biodistribution studies. In vivo validation was performed in diet-induced MASH mouse models using IVIS imaging and flow cytometry. Results CCI analysis revealed a notable dissociation between fibrotic "potential" and "execution": MASLD tissues showed 2,043 ECM signaling genes but zero physical interaction genes, while MASH tissues exhibited 315 MatrixDB-positive genes, suggesting that physical ECM crosslinking may serve as a key determinant of fibrosis execution. Pseudotime modeling identified a fibrosis transition zone, where fibrogenic scores shifted from negative to positive, coinciding with marked increases in fibrotic and lipid-associated macrophage populations. This transition point represents the fibroblast-to-myofibroblast transdifferentiation threshold. At this transition, both fibroblasts and macrophages showed progressive CD47 upregulation from Control to MASH stages. Leveraging this shared upregulation, SIRP-EV biodistribution studies demonstrated significantly higher accumulation in fibrotic versus normal liver tissue (p<0.01). Flow cytometry confirmed selective delivery to fibroblasts and macrophages—the two key fibrosis-driving cell populations. Conclusion Our spatial transcriptomics-guided approach identified macrophage-fibroblast physical crosstalk as a key driver of MASH fibrosis progression and revealed CD47 as a targetable surface marker on both pathogenic cell types. These findings suggest that SIRP-EV may provide a platform for selective therapeutic delivery to the fibrotic niche, potentially enabling disease-modifying interventions that current systemic approaches have not achieved.