Abstract
Esophageal squamous cell carcinoma (ESCC) is a highly lethal malignancy with limited therapeutic progress and a 5-year survival rate below 20%. Normal adjacent-to-tumor (NAT) tissues, widely used as "normal" controls, are increasingly recognized as molecularly distinct from both tumor and healthy tissues, reflecting early carcinogenic alterations rather than a true normal state. Here, we integrated proteomic data from 20 Healthy, 124 NAT, and 124 Tumor tissues to systematically map protein alterations across the full spectrum of ESCC development. Cross-stage analysis identified eight distinct expression modes, capturing stepwise molecular transitions from Healthy to NAT to Tumor. Notably, NAT tissues exhibited extensive early molecular alterations, characterized by pronounced immune activation-particularly in the complement and coagulation cascades-and broad metabolic reprogramming. We further demonstrated that the NAT proteome itself harbors critical clinical information, defining two proteomic subtypes and four immune subtypes that were strongly associated with patient survival and tumor stage. Based on these features, we developed two prognostic models: (i) an integrated NAT-subtype-pTNM model, which outperformed traditional staging, and (ii) a "US" model, built from proteins consistently upregulated from Healthy to NAT and remaining stable in Tumor samples, which achieved superior predictive performance in the independent test set (5-year AUC = 0.849 for overall survival; 3-year AUC = 0.861 for disease-free survival). Together, these findings extend beyond conventional Tumor-NAT comparisons, offering molecular insights and clinically relevant resources for early detection, patient stratification, and therapeutic development in ESCC.