Histone chaperone-based stratification combined with two-sample Mendelian randomization identifies ADORA2B and SAPCD2 as prognostic biomarkers in esophageal cancer

Background: Esophageal cancer (EC) lacks robust biomarkers to guide prognosis and therapy. Histone chaperone-related genes (HCRGs) shape chromatin states, but their roles in EC remain unclear. We integrated histone chaperone-based transcriptomic stratification with two-sample Mendelian randomization (MR) to identify genes with genetic evidence consistent with EC susceptibility and clinical relevance. Methods: RNA-seq and clinical data from TCGA-ESCA (184 tumors with available RNA-seq data and 13 normal samples) and an external cohort (GSE53624; 119 tumors, 119 normals) were analyzed. Prognosis-associated HCRGs informed unsupervised clustering. Differentially expressed genes common to cluster and tumor-versus-normal contrasts entered a two-sample MR framework using eQTL instruments and an EC GWAS (998 cases, 475,308 controls; European ancestry). Sensitivity analyses and the Steiger directionality test were then performed, followed by expression analysis to identify hub genes. We assessed survival associations, performed functional enrichment and immune deconvolution (CIBERSORT), modeled drug sensitivity (GDSC-based prediction), built a clinicogenomic nomogram, and validated expression by RT-qPCR in paired tissues. Results: HCRG-based clustering separated patients with distinct overall survival. MR implicated 26 genes in EC risk; among these, ADORA2B and SAPCD2 were consistently overexpressed in tumors (external validation and RT-qPCR) yet higher tumor expression predicted longer survival. Both genes were linked to cell-cycle and spliceosome programs; ADORA2B showed additional immuno-metabolic enrichment, whereas SAPCD2 mapped to metabolic/proteostasis pathways. High-expression groups exhibited reduced regulatory T-cell proportions and broader immune shifts. Predicted drug response suggested greater sensitivity to vinorelbine/etoposide but reduced sensitivity to paclitaxel in ADORA2B/SAPCD2-high tumors. A prognostic nomogram combining gene expression with stage and nodal status achieved 1-3-year AUCs of 0.67-0.75. Conclusions: Integrating HCRG-guided stratification with MR nominates ADORA2B and SAPCD2 as MR-supported biomarkers that are overexpressed in EC yet mark favorable prognosis, a pattern consistent with a less immunosuppressive microenvironment and distinct chemosensitivity. These genes warrant histology-stratified validation and mechanistic studies and may aid risk stratification and therapeutic decision-making.