泛癌单细胞图谱揭示跨癌种保守的肿瘤印记T细胞状态

Tumor immune escape arises not only from intrinsic alterations within tumor cells but also from chronic antigen stimulation, inhibitory checkpoint signaling and an immunosuppressive tumor microenvironment that progressively remodels infiltrating immune cells. In previous single-cell RNA-seq analyses, we repeatedly observed subsets of tumor-infiltrating T cells whose transcriptional profiles closely resemble those of neighboring tumor cells, blurring the boundary between immune and malignant compartments and suggesting that key features of tumor adaptation may be written into T-cell states. However, the shared features of such putative tumor-imprinted T cells across cancer types and tissues, and their systematic differences from healthy T cells, remain poorly defined. To address this gap, we aggregated single-cell RNA-seq data from 371,108 intratumoral T cells spanning 19 cancer types. Matched healthy controls were obtained from Tabula Sapiens across 15 tissue types corresponding to these cancers, from which we identified ~130,000 T cells among 290,334 immune cells. We find that tumor-intrinsic and microenvironmental transcriptional programs are markedly enriched in CD8 terminally exhausted and NK-like CD8 T-cell states in patients with cancer, whereas these signals are only weakly detectable in healthy T cells. In particular, epithelial-mesenchymal transition, TGF-beta signaling, hypoxia, angiogenesis, MYC and E2F target programs and cell cycle-related pathways are consistently upregulated in tumor-imprinted CD8 T cells, accompanied by selective depletion of naïve and central memory signatures. Building on these observations, we derive a per-cell tumor-imprint score that integrates these pathways, delineates subset-specific and cancer type-specific imprinting patterns, and robustly separates tumor from healthy T cells across multiple organs and independent cohorts. In single-cell datasets from immune checkpoint blockade cohorts, baseline tumor-imprint patterns align with T-cell states associated with treatment response, indicating that this score may aid patient stratification and response prediction. Collectively, this pan-cancer single-cell study provides evidence that tumor cells “hijack” the immune system by rewiring the transcriptional and pathway programs of infiltrating T cells. The resulting tumor-imprint score offers a biologically grounded and operational framework to quantify the extent of T-cell tumorization, refine immunotherapy strategies and inform the design of next-generation T-cell-targeted and cell-based therapies.