Abstract
Aim: Immune checkpoint inhibitors (ICIs) have transformed cancer therapy; however, their efficacy in head and neck cancer (HNC) remains limited, with only a minority of patients achieving durable responses. Understanding the molecular mechanisms underlying ICI resistance in HNC is therefore crucial. Methods: We conducted an integrative analysis of genomic, transcriptomic, and clinical data from 139 ICI-treated HNC patients (MSKCC cohort) and 502 treatment-naïve HNC cases (TCGA cohort). ROS1 mutation status, tumor mutational burden (TMB), neoantigen load, immune cell infiltration (via CIBERSORT), and immune-related gene expression were evaluated. Gene set enrichment analysis (GSEA) was performed to identify dysregulated pathways. Survival outcomes were assessed using Kaplan-Meier analysis and Cox regression, with statistical significance defined as P < 0.05. Results: Patients harboring ROS1 mutations exhibited significantly poorer outcomes following ICI therapy, with shorter median overall survival [OS: 5.0 vs. 11.0 months, hazard ratio (HR) = 3.22, 95%CI: 1.26-8.19, P = 0.011] compared to ROS1 wild-type counterparts. Multivariate analysis confirmed ROS1 mutation as an independent predictor of poor OS in ICI-treated patients (HR = 4.78, 95%CI: 1.70-13.43, P = 0.003). In contrast, ROS1 mutations showed no prognostic significance in the treatment-naïve TCGA-HNC cohort (P = 0.26), confirming their role as a predictive (not prognostic) biomarker for ICI response. Interestingly, despite exhibiting higher TMB and neoantigen levels, ROS1-mutant patients showed inferior survival, underscoring the context-dependent limitations of TMB as a predictive biomarker. Mechanistically, ROS1-mutant tumors displayed an immunosuppressive tumor microenvironment characterized by diminished CD8(+) T cell infiltration, attenuated interferon-γ signaling, and downregulation of immune-related genes (CXCL9, CXCL10, IFNG, PD-L1). GSEA revealed enrichment of MYC pathway activity in ROS1-mutant tumors, which suppressed antigen presentation and T cell activation pathways. Conclusion: ROS1 mutations drive ICI resistance in HNC by promoting an immunosuppressive TME via MYC-mediated transcriptional reprogramming, impairing antigen presentation and T cell function. Incorporating ROS1 status into biomarker panels may improve patient stratification and guide combinatorial therapies targeting both immune evasion and oncogenic pathways.