From prediction to engagement: defining tumor-reactive T cells through biological interaction

Abstract Adoptive T-cell therapy (ACT), particularly tumor-infiltrating lymphocyte (TIL) therapy, demonstrates that durable regression of solid tumors can occur when polyclonal T cells target authentic tumor antigens, including private neoantigens. Yet scalable identification of tumor-reactive clonotypes remains limited because the relevant cells are rare, patient-specific, and poorly captured by indirect markers, expansion behavior, or predictive algorithms. We examined whether productive engagement between T cells and tumor cells or antigen-presenting cells provides a more direct organizing principle for tumor-reactive T-cell discovery. Across published tumor and blood datasets reanalyzed here, T cells recovered through physical clustering, target-cell extraction, or antigen-dependent activation are enriched for clonotypes occupying antigen-experienced, progenitor-like states with restrained cytotoxic differentiation. Productive engagement can therefore serve as a practical enrichment variable, enabling recovery of rare neoantigen-reactive T cells from circulation while preserving developmental states associated with persistence and therapeutic responsiveness. These analyses support a shift in tumor-specific T-cell discovery from prediction-first nomination to interaction-first recovery: enrich for T cells bearing evidence of prior tumor-antigen encounter in vivo, recover clonotypes preserved in a restrained progenitor-like state, expand them before culture competition erases them, and reinforce them therapeutically with matched antigenic information.