Abstract
Esophageal squamous cell carcinoma (ESCA) remains a lethal disease with few reliable biomarkers for early diagnosis or treatment selection. In this study, we combined bulk RNA-seq (TCGA-ESCA, GSE53625) and single-cell RNA-seq (GSE196756) data to derive and characterize an exosome-related gene signature. Using LASSO-Cox regression on 121 literature-curated exosomal genes, we established a nine-gene risk model that stratified ESCA patients into high- and low-risk groups with significantly different overall survival. This signature was validated computationally in two independent bulk cohorts and shown to correlate with established clinical parameters. Functional assays for one model component, HAPLN3, confirmed its tumor-suppressive role: HAPLN3 overexpression inhibited ESCA cell proliferation and promoted apoptosis. To explore immunotherapy relevance, we analyzed anti-PD-L1 response in the IMvigor210 bladder cancer cohort-chosen because of shared squamous histology and PD-L1 pathways-to show that low-risk patients exhibited higher response rates and longer survival. Finally, single-cell analysis mapped each signature gene to specific tumor microenvironment compartments, and integrative genomic profiling linked risk scores to copy number variation and DNA methylation changes. Together, our findings define a functionally informed exosomal biomarker panel for ESCA prognosis and suggest its potential to guide immunotherapy, with further validation in ESCA-specific immunotherapy cohorts underway.