Immune checkpoint inhibitor (ICI) therapies have shown great promise in cancer treatment. However, the intra-heterogeneity is a major barrier to reasonably classifying the potential benefited patients. Comprehensive heterogeneity analysis is needed to solve these clinical issues. In this study, the samples from pan-cancer and independent breast cancer datasets were divided into four tumor immune microenvironment (TIME) subtypes based on tumor programmed death ligand 1 (PD-L1) expression level and tumor-infiltrating lymphocyte (TIL) state. As the combination of the TIL Z score and PD-L1 expression showed superior prediction of response to ICI in multiple data sets compared to other methods, we used the TIL Z score and PD-L1 to classify samples. Therefore, samples were divided by combined TIL Z score and PD-L1 to identify four TIME subtypes, including type I (3.24%), type II (43.24%), type III (6.76%), and type IV (46.76%). Type I was associated with favorable prognosis with more T and DC cells, while type III had the poorest condition and composed a higher level of activated mast cells. Furthermore, TIME subtypes exhibited a distinct genetic and transcriptional feature: type III was observed to have the highest mutation rate (77.92%), while co-mutations patterns were characteristic in type I, and the PD-L1 positive subgroup showed higher carbohydrates, lipids, and xenobiotics metabolism compared to others. Overall, we developed a robust method to classify TIME and analyze the divergence of prognosis, immune cell composition, genomics, and transcriptomics patterns among TIME subtypes, which potentially provides insight for classification of TIME and a referrable theoretical basis for the screening benefited groups in the ICI immunotherapy.
Keywords: immunotherapy; programmed death ligand 1; the Cancer Genome Atlas; tumor immune microenvironment; tumor-infiltrating lymphocyte.