A Conserved Macrophage-to-Hepatic Stellate Cell PDGF Axis in Human MASH Identified by Multi-Cohort sc/snRNA-Seq Re-Analysis

Abstract

Metabolic dysfunction-associated steatohepatitis (MASH) progresses through macrophage activation and hepatic stellate cell (HSC)-driven fibrogenesis. As cell-level single-cell discoveries often suffer from pseudoreplication and fail to replicate across cohorts, in this study we critically re-evaluated human liver single-nucleus RNA-sequencing data using a rigorous per-donor statistical framework to dissociate invariant biological signals from technical artifacts. Macrophage and HSC compartments from dataset GSE202379 were profiled across clinical stages (healthy, metabolic dysfunction-associated steatotic liver disease, MASH). Cell-level statistics were aggregated into per-donor metrics, and continuous phenotypic shifts evaluated via the Jonckheere–Terpstra (JT) trend test. Key findings were tested for cross-cohort reproducibility in two independent cohorts (GSE136103 and GSE244832). A 249-gene scar-associated macrophage (SAMac) program showed reproducible, stage-dependent expansion in the discovery cohort (per-donor Jonckheere–Terpstra trend across the disease-only spectrum, MASLD to advanced MASH, z = 2.23, p = 0.026 [primary]; z = 2.74, p = 0.006 across all four stages including the healthy baseline) and directional replication in external data. Cell–cell communication analysis replicated the receptor-side topology of a conserved macrophage-to-HSC PDGF axis in both external cohorts and the ligand-driven PDGFC→PDGFRB edge in the external snRNA-seq cohort (GSE244832), with PDGFRB being a high-degree hub within the HSC fibrogenic interaction network. Importantly, a previously published global HSC activation module failed to validate (JT p = 0.77); however, its core extracellular matrix and RUNX1/2 sub-axis remained robust, whereas the widely claimed AP-1/JUNB regulatory driver was not supported at the transcript level in the single-nucleus RNA-seq cohort data. These findings are consistent with an isolation-associated immediate-early gene response in enzymatically dissociated scRNA-seq data, given the flat-to-downward transcript-level trend in cryopreserved-tissue cohorts; transcript-level data cannot, however, exclude protein- or phosphorylation-level AP-1 activity. Targetable downstream components within the validated core architecture were annotated. By implementing per-donor verification, we successfully separated robust pathological features of MASH from platform-dependent artifacts. The validated SAMac program and the macrophage–HSC PDGF axis, together with an associated RUNX subprogram, represent high-confidence, reproducibility-filtered candidate therapeutic nodes that require functional validation. These findings establish a reproducibility-first standard for single-cell translational research, and the validated programs are consistent with emerging models of inter-organ inflammatory circuitry in MASH.