New evaluation of the tumor immune microenvironment of non-small cell lung cancer and its association with prognosis

BACKGROUND: A better understanding of the tumor immune microenvironment (TIME) will facilitate the development of prognostic biomarkers and more effective therapeutic strategies in patients with lung cancer. However, little has been reported on the comprehensive evaluation of complex interactions among cancer cells, immune cells, and local immunosuppressive elements in the TIME. METHODS: Whole-exome sequencing and RNA sequencing were carried out on 113 lung cancers. We performed single sample gene set enrichment analysis on TIME-related gene sets to develop a new scoring system (TIME score), consisting of T-score (tumor proliferation), I-score (antitumor immunity) and S-score (immunosuppression). Lung cancers were classified according to a combination of high or low T-score, I-score, and S-scores (eight groups; G1-8). Clinical and genomic features, and immune landscape were investigated among eight groups. The external data sets of 990 lung cancers from The Cancer Genome Atlas and 76 melanomas treated with immune checkpoint inhibitors (ICI) were utilized to evaluate TIME scoring and explore prognostic and predictive accuracy. RESULTS: The representative histological type including adenocarcinoma and squamous cell carcinoma, and driver mutations such as epidermal growth factor receptor and TP53 mutations were different according to the T-score. The numbers of somatic mutations and predicted neoantigens were higher in T(hi) (G5-8) than T(lo) (G1-4) tumors. Immune selection pressure against neoantigen expression occurred only in T(hi) and was dampened in T(hi)/I(lo) (G5-6), possibly due to a reduced number of T cells with a high proportion of tumor specific but exhausted cells. T(hi)/I(lo)/S(hi) (G5) displayed the lowest immune responses by additional immune suppressive mechanisms. The T-score, I-score and S-scores were independent prognostic factors, with survival curves well separated into eight groups with G5 displaying the worst overall survival, while the opposite group T(lo)/I(hi)/S(lo) (G4) had the best prognosis. Several oncogenic signaling pathways influenced on T-score and I-scores but not S-score, and PI3K pathway alteration correlated with poor prognosis in accordance with higher T-score and lower I-score. Moreover, the TIME score predicted the efficacy of ICI in patients with melanoma. CONCLUSION: The TIME score capturing complex interactions among tumor proliferation, antitumor immunity and immunosuppression could be useful for prognostic predictions or selection of treatment strategies in patients with lung cancer.