Dissecting macrophage heterogeneity in ulcerative colitis: Single-cell analysis and functional validation of S100A4...

Ulcerative colitis (UC) is a chronic inflammatory bowel disease (IBD) with a rising global incidence and significant socioeconomic impact. While macrophages are key drivers of intestinal inflammation, their heterogeneity and molecular characteristics in UC remain poorly defined. This study aimed to dissect macrophage subpopulations and validate potential therapeutic targets through integrated single-cell and experimental analyses.

We analyzed publicly available single-cell RNA sequencing (scRNA-seq) datasets from colonic tissues of healthy controls (HC), non-inflamed UC, and inflamed UC patients. Data preprocessing, clustering, and cell-type annotation were performed using Seurat. Macrophage subsets were identified and characterized through pseudotime trajectory analysis (Monocle3) and intercellular communication profiling (CellChat). High-dimensional weighted gene co-expression network analysis (hdWGCNA) was employed to uncover macrophage-associated gene modules. Candidate genes were evaluated for diagnostic potential in an independent bulk transcriptomic UC cohort using nine machine learning algorithms. Functional validation of the key gene was conducted in vivo using a dextran sulfate sodium (DSS)-induced colitis mouse model and in vitro with LPS-stimulated RAW264.7 macrophages.

scRNA-seq analysis revealed 18 major cell types across all samples, with macrophages notably enriched in inflamed tissues. Five distinct macrophage subsets were identified: APOE+, S100A8+, LSP1+, IGHM+, and IL1B+ macrophages. Among these, S100A8+ and IL1B+ macrophages were significantly expanded in inflamed UC and displayed non-classical M1/M2 polarization patterns. CellChat analysis demonstrated that these subsets were predominant contributors to pro-inflammatory signaling and mucosal injury. hdWGCNA uncovered five macrophage-specific co-expression modules, with the macrophage-M2 module-active in S100A8+ macrophages-enriched in S100A4, S100A6, and VCAN. Among nine machine learning models, the Random Forest algorithm achieved the highest diagnostic accuracy (AUC = 0.89) in predicting UC based on these gene signatures. Functionally, both pharmacological inhibition (niclosamide) and lentiviral shRNA-mediated knock-down of S100A4 in RAW264.7 macrophages significantly attenuated DSS-induced colitis in vivo and suppressed LPS-driven M1 polarization while restoring M2 markers in vitro, validating S100A4 as a critical therapeutic node.

This study provides a comprehensive single-cell landscape of macrophage heterogeneity in UC, identifying S100A8+ and IL1B+ macrophages as key mediators of mucosal inflammation. Integrative analysis pinpointed S100A4 as a robust diagnostic and therapeutic candidate, with experimental validation underscoring its translational potential. These findings offer new insights into macrophage-driven pathogenesis and open avenues for targeted UC therapies.