Abstract
Background: Diabetic foot ulcers (DFUs) are chronic wounds with high morbidity and mortality. Secretory proteins coordinate intercellular communication and may regulate inflammation, tissue repair and regeneration, but their contributions to DFU pathogenesis remain unclear. Aim: To discover and validate secretory protein–linked biomarkers that distinguish non-healing DFUs and to explore their potential utility for diagnosis and therapy. Methods: We integrated bulk RNA-sequencing datasets (GSE199939 training set; GSE80178 and GSE143735 validation sets) and a single-cell RNA-sequencing dataset (GSE223964). Differentially expressed genes, secretory protein lists, and weighted gene co-expression networks were intersected to select candidates. Functional enrichment, protein interaction networks and support vector machine–recursive feature elimination identified key markers. We visualized their cell-type distribution at single-cell resolution and validated their expression in external cohorts. Pathway enrichment, gene co-expression networks, ceRNA regulatory analysis and qRT-PCR in patient samples were used for further characterization. Results: Among 4803 differentially expressed genes, 743 overlapped with known secretory proteins. WGCNA highlighted modules strongly associated with DFUs, yielding 386 candidates. SVM-RFE combined with protein interaction analysis pinpointed four secretory proteins—LYZ, MPO, SLCO2B1 and TREM2—as putative biomarkers. Single-cell analyses showed that MPO, LYZ, SLCO2B1 and TREM2 transcripts are detectable in multiple skin-resident and immune cell populations, and that the DFU-associated upregulation of MPO and LYZ is most pronounced within keratinocyte clusters, while MPO and TREM2 remained consistently dysregulated in independent bulk cohorts. MPO-associated genes were enriched for immune and inflammatory pathways, whereas TREM2-linked genes implicated cell cycle and cytoskeletal regulation. GeneMANIA and ceRNA analyses revealed extensive interaction networks. qRT-PCR confirmed differential expression of MPO and TREM2 in clinical DFU tissues. Conclusions: Integrated multi-modal analysis identifies secretory proteins, particularly MPO and TREM2, as central determinants of impaired healing in DFUs. These molecules and their regulatory networks represent promising biomarkers and therapeutic targets for precision management of diabetic wounds.