KLRG1 defines a distinct tumor-infiltrating granzyme K+ CD8 + T cell population

Emerging evidence indicates that granzyme K expressing tumor infiltrating CD8 + T-cells play a critical role in mediating anti-tumor T-cell immunity and immunotherapy treatment response. However, precise characterization and cell surface markers for the viable isolation of these cells from tumor samples are lacking. We constructed a tumor-infiltrating CD8 + T cell subpopulation atlas (n = 286,827 cells) by integrating single-cell RNA sequencing datasets from two immunotherapy-treated patient cohorts with aerodigestive tract malignancies. This unified dataset facilitated detailed mapping of transcriptional trajectories and stringent identification of distinctive cell surface markers across major CD8 + tumor-infiltrating lymphocyte (TIL) compartments. Analysis of the CD3 + CD8+ TIL transcriptional landscape revealed three predominant populations across all tumor types and treatment conditions: stem-like cells, dysfunctional effector cells, and GZMK+ effector cells. Notably, the GZMK+ effector population displayed a distinctive transcriptional profile lacking positive enrichment of pre-defined gene sets, suggesting a previously poorly characterized subpopulation without established cell surface markers. We performed high-dimensional flow cytometry on primary human tumor samples to validate candidate CD8 + TIL subpopulations. Through validation in primary human tumors, we established that cell surface KLRG1 reliably identifies and enables viable isolation of the GZMK+ effector TIL population. Concurrently, we demonstrated that dysfunctional and stem-like populations can be effectively isolated as KLRG1-CD39 + and KLRG1-CD39- CD55 + subsets, respectively. Importantly, accurate enrichment occurs despite decreased starting cell type proportion. Our characterization of CD8 + TIL subpopulations and validation of their distinguishing surface markers establishes a robust platform for isolation and protein-level granzyme assessment of these key cells in patient samples.