Single cell RNA-seq reveals immunosuppressive gastric stem-like cancer cells as a poor prognostic factor

The tumor microenvironment is characterized by high cellular heterogeneity and a complex network of cell-cell communications. Analysis at the single cell level is required to dissect this complexity. Here we apply single-cell full-length transcriptome sequencing to analyze 3443 cells from paired normal-adjacent and tumor tissues of 15 gastric cancer patients with different histological profiles. Among the epithelial cells profiled, we discovered subsets of proliferative stem-like cells with upregulated pro-tumor pathways and that were associated with poorer prognosis. Stem-like cell enriched tissues also show distinct T cell populations with proportionally more TIM3+ CD8+ cells, while non-enriched tissues had proportionally more KLRC1+ CD8+ cells, offering opportunities for checkpoint inhibitor therapy. In silico predictions of ligand-receptor interactions revealed immune-suppressive interactions between these stem-like cells and immune cells via interacting pairs such as Nectin-2(CD112)-TIGIT, Galectin-9-TIM3 and CXCL16-CXCR6. Using immunohistochemistry, we found CD8+ T cells expressing TIGIT in close proximity to stem-like cells expressing NECTIN2.