Identification of recurrence-associated gene signature and tumor immune microenvironment features in resected stage I NSCLC.

Abstract

9080 Background: Surgery is the primary treatment for stage I NSCLC, but postoperative recurrence leads to poor prognosis. Alterations of tumor genes and immune microenvironment may be crucial factors for tumor recurrence; however, the detailed mechanisms remain unclear. Methods: A total of 130 resected stage I NSCLC patients were enrolled, 69 developed recurrence within three years and 61 without recurrence over five years. Whole exome sequencing (WES) was performed to evaluate genomic alterations. Immunohistochemistry was carried out to assess the expression of PD-L1, CD3 and CD8. We calculated density of CD3+ and CD8+ T cells in the center of tumors (CT) and invasive margins (IM), defined six immunoscore types based on the location and density of both cells, and performed ROC analysis to evaluate prognostic value of them. We further verified our results using stage I NSCLC cohorts from the Cancer Genome Atlas (TCGA) and Tumor immune estimation resource (TIMER) database. Results: In univariate analysis, lung adenocarcinoma (LUAD) patients showed significantly higher risk of recurrence (p = 0.008). There was no statistically significant correlation between recurrence and other clinical factors, including TNM stage. Although driver gene mutations, such as those of EGFR, had no correlation with recurrence, MUC4 mutation and high tumor mutation burden (TMB) were significantly associated with higher risk of recurrence (p = 0.001 and 0.0032, respectively). Enrichment analysis of KEGG pathways showed that Ras pathway mutations were significantly enriched in MUC4 mutant group and recurrence group (p = 0.02 and 0.05, respectively). 9.6% patients had PD-L1 positive expression (TPS≥1%), but showed no association with recurrence. Recurrence group had much lower density of CD8CT, CD8IM and CD3CT +T cells(p = 0.0026, 0.0022 and 0.0308, respectively). Immunoscore type V, based on the average of CD8CT, CD8IM and CD3CT + T cells, had the highest prognostic value (AUC = 0.764) and was used as the final immunoscore in our study. In multivariate analysis, we found MUC4 mutation and low immunoscore were independent predictors of higher risk of recurrence. Smoking history was also an independent prognostic factor in LUAD. While in LUSC, only immunoscore correlated with recurrence. In TCGA cohort, MUC4 mutation rate was significantly lower (3.6% vs. 24.3%, p < 0.001) and had no correlation with risk of recurrence (p = 0.765). Besides, the tumor infiltrating CD8+ T cells also had no correlation with risk of recurrence (p = 0.469). Conclusions: In this study, we established a refined immunoscore with high prognostic value for tumor recurrence in stage I NSCLC. In addition, we showed for the first time a strong association between MUC4 mutation and recurrence, which might be mediated by the Ras pathway. Finally, the recurrence mechanisms might vary among different histological subtypes.