Abstract / Summary
Abstract Background Esophageal squamous cell carcinoma (ESCC) has a poor prognosis. The heterogeneity of cancer-associated fibroblasts (CAFs) within the tumor microenvironment and their interactions with immune cells warrant further investigation. This study employed single-cell RNA sequencing to analyze the subtypes and differentiation trajectories of CAFs in ESCC, as well as to explore the mechanisms underlying CAF–myeloid cell interactions and their prognostic significance. Methods Single-cell transcriptomic data for ESCC were downloaded from the GEO database, followed by quality control, dimensionality reduction, and clustering. Major cell types were identified using marker genes. CAF subtypes were characterized, differentiation trajectories were constructed, and Scissor was employed to identify cell subpopulations associated with prognosis. CellChat was used to examine receptor-ligand interactions between CAFs and prognosis-associated cell populations, while the M1/M2 polarization score assessed the polarization direction of myeloid cells. For prognostic validation, CIBERSORT estimated the abundance of M2 macrophages and evaluated their association with overall survival. Cross-validation was performed using an M2 marker gene module score independent of CIBERSORT. Finally, the prognostic relevance and cross-cohort consistency of key molecules were validated in the TCGA-ESCC dataset and two independent GEO cohorts. Results A total of 11 major cell types were annotated. Among the 2,749 cancer-associated fibroblast (CAF) cells, extracellular matrix (ECM)-CAFs accounted for 79% of tumor CAFs. Prognostic analysis revealed that 82.3% of the prognosis-associated cells identified by Scissor were myeloid cells. A subpopulation with high HLA expression was present within ECM-CAFs, and it was found that ECM-CAFs communicate with myeloid cells via the HLA-DRA-CD4 axis. Notably, 65.5% of myeloid cells expressed CD4 and exhibited M2 polarization. Batch transcriptomic validation yielded consistent results: HLA-DRA expression was positively correlated with M2 macrophage abundance, CD4 expression was associated with the M2 module score, and both the HLA-DRA and CD4 high-expression groups were linked to poorer prognosis. M2 macrophage abundance was identified as an independent prognostic factor, with consistent risk ratio directions observed across two GEO cohorts. Conclusion ECM-CAFs are the predominant CAF subtype in ESCC tumors. Their internal subpopulation, characterized by high HLA expression, interacts with myeloid cells through the HLA-DRA-CD4 axis, promoting the polarization of CD4-positive myeloid cells toward the M2 phenotype. The abundance of M2 macrophages, confirmed by two independent methods—CIBERSORT deconvolution and signature gene module scoring—is associated with poor prognosis.