PURPOSE There is currently no reliable biomarker to predict who would benefit from anti-PD-1/PD-L1 inhibitors. We comprehensively analyzed the immunogenomic properties in The Cancer Genome Atlas (TCGA) according to the classification of tumor into four groups based on PD-L1 status and tumor-infiltrating lymphocyte recruitment (TIL), a combination that has been suggested to be a theoretically reliable biomarker of anti-PD-1/PD-L1 inhibitors.

EXPERIMENTAL DESIGN The RNA expression levels of PD-L1 and CD8A in the samples in the pan-cancer database of TCGA (N = 9,677) were analyzed. Based on their median values, the samples were classified into four tumor microenvironment immune types (TMIT). The mutational profiles, PD-L1 amplification, and viral association of the samples were compared according to the four TMITs.

RESULTS The proportions of TMIT I, defined by high PD-L1 and CD8A expression, were high in lung adenocarcinoma (67.1%) and kidney clear cell carcinoma (64.8%) among solid cancers. The number of somatic mutations and the proportion of microsatellite instable-high tumor in TMIT I were significantly higher than those in other TMITs, respectively (P < 0.001). PD-L1 amplification and oncogenic virus infection were significantly associated with TMIT I, respectively (P < 0.001). A multivariate analysis confirmed that the number of somatic mutations, PD-L1 amplification, and Epstein-Barr virus/human papillomavirus infection were independently associated with TMIT I.

CONCLUSIONS TMIT I is associated with a high mutational burden, PD-L1 amplification, and oncogenic viral infection. This integrative analysis highlights the importance of the assessment of both PD-L1 expression and TIL recruitment to predict responders to immune checkpoint inhibitors. Clin Cancer Res; 22(9); 2261-70. ©2016 AACRSee related commentary by Schalper et al., p. 2102.