Gastric cancer (GC) is a leading cause of cancer death. While CD8 + T cell infiltration correlates with better prognosis, few CD8 + T based subtypes and predictive models exist for GC. Using scRNA-seq (GSE163558) and bulk RNA-seq (TCGA-STAD) data, hub CD8 + T genes were identified. Samples were clustered into two subtypes via NMF based on prognostic hub genes. Clinical features, molecular traits, immune infiltration, and treatment response were compared to define the two subtypes. Subsequently, Multiple machine learning methods were employed to select prognostic differentially expressed genes to build a gene signature. A nomogram integrated the signature and clinical characteristics was established. Finally, flow cytometry was conducted to evaluate the influence of BMP inhibitor on GC tumor microenvironment and tumor infiltrating lymphocytes. 153 CD8 + T genes were identified, while six genes (RNF19B, IRF1, TAP1, STK17A, CXCR4, SELL) had prognostic value in GC. Patients were stratified into two subtypes with distinct survival. C1 showed higher immune activity and better immunotherapy response than C2. Besides, C1 had elevated calcium/chemokine signaling and reduced BMP pathway. A CD8 + T-related gene signature was developed to predict the hot-like tumor. The final nomogram accurately stratified the risk in GC. Flow cytometry showed the BMP inhibitor could switch cold-like GC tumor to hot-like phenotype. This study identified prognostic CD8 + T hub genes, defined CD8 + T based subtypes in GC which had differing features/therapy responses, and established a CD8 + T gene signature and nomogram for risk stratification, offering new insights for GC management.
Li et al. (Mon,) studied this question.