A Progression-Associated Tumor-Reactive T-Cell Signature Predicts Immunotherapy Response and Reveals ICOS-High CD8 T-Cell Functional States in Head and Neck Squamous Cell Carcinoma.

Immune checkpoint blockade has improved outcomes for patients with head and neck squamous cell carcinoma (HNSCC), but reliable biomarkers for predicting therapeutic response remain limited. Tumor-reactive T-cell states are increasingly recognized as critical determinants of antitumor immunity; however, progression-associated T-cell programs that reflect tumor evolution and predict immunotherapy benefit remain insufficiently characterized. We integrated single-cell transcriptomic and T-cell receptor (TCR) sequencing data from HPV-negative HNSCC progression models to identify T-cell genes associated with tumor progression and TCR-supported immune states. Candidate genes were further refined using pretreatment immunotherapy cohorts, and a five-gene signature (HNSC-TstcSig) was established using survival and response-associated analyses. The predictive performance of HNSC-TstcSig was evaluated across multiple immunotherapy cohorts, including HNSCC and melanoma data sets. Single-cell transcriptomic analyses, spatial transcriptomic analysis, and multiplex immunofluorescence were performed to characterize the biological features of HNSC-TstcSig-associated immune states. In addition, machine learning-based survival modeling was conducted to identify genes shared between immunotherapy response and natural prognosis. HNSC-TstcSig, consisting of GBP2, HAVCR2, ETV7, ICOS, and IL2RA, showed consistent associations with immunotherapy response across multiple cohorts and reflected an immune-inflamed tumor microenvironment characterized by enhanced immune infiltration, T-cell activation, and immune checkpoint-related programs. Single-cell analyses revealed that HNSC-TstcSig genes were predominantly enriched in activated CD8 T-cell populations and were associated with cytotoxic and inflammatory immune states. Among the signature genes, ICOS showed consistent associations with CD8 T-cell activation, cytotoxicity, and immune checkpoint-related features. Network perturbation analysis further suggested that ICOS was connected with transcriptional programs involved in T-cell functional regulation. Spatial transcriptomic analysis and multiplex immunofluorescence further supported the association between ICOS-related immune features and tumor immune activity at the tissue level. Furthermore, a machine learning-derived natural prognostic model identified ICOS as a shared gene linking immunotherapy response and survival outcomes. We developed a progression-associated T-cell signature, HNSC-TstcSig, that predicts immunotherapy response and reflects immune-active tumor states in HNSCC. Our findings suggest that ICOS is associated with a tumor-reactive CD8 T-cell state characterized by cytotoxic activity and immune checkpoint-related features, providing a potential biomarker for immune stratification and a framework for understanding T-cell states associated with therapeutic benefit.