Introduction
Immune checkpoint inhibitors (ICI) provide limited benefit in patients with EGFR-, ALK- or ROS1-positive lung adenocarcinoma (LUAD), while patients with BRAF- and KRAS-positive LUAD demonstrate better responses. To understand the differential response to ICI in driver gene-altered LUAD, we performed a comparative analysis of their tumor neoepitope load and immune environment.
Methods
De-identified molecular data from 2,740 LUAD patients previously sequenced with Tempus|xT, a 648-gene DNA panel (Tempus Labs, Chicago, IL) were analyzed. Whole-exome capture RNAseq was performed on 1,454 of these samples. Driver mutational status was correlated with neoepitope load and immune infiltration enrichment. Neoantigen prediction was performed on all non-silent mutations using tumor only or tumor/normal matched samples, and immune infiltration scores were predicted using a machine-learning model trained on mRNA expression and pathologist-scored IHC data (PMID:31570899). Distributions of predicted neoantigens and immune infiltration, including CD8+/CD4+ T-cells, B-cells, NK cells, and macrophages, between driver-gene altered LUAD (defined as patients with alterations in KRAS, EGFR, ALK, BRAF, MET, PIK3CA, ROS1, RET, or ERBB2) and driver-gene wild type (WT) LUAD were compared using two-sided Wilcoxon tests.
Results
Non-synonymous antigenic mutations were predicted for all DNA-sequenced tumors. The median neoepitope load was 13 (range 1-406) and WT tumors had the highest median neoepitope load (Table). EGFR-, MET-, RET-, ALK-, ROS1– and BRAF-positive tumors had significantly lower neoepitope load than WT, with BRAF-positive LUAD having the highest neoepitope load in this group (Table). Neoepitope loads of KRAS-, ERBB2- and PIK3CA-positive tumors were not significantly different from WT. PD-L1 status was not associated with neoantigen load after controlling for driver mutation status. Overall immune infiltration was higher in MET– and BRAF-positive LUAD compared to WT (Table). Relative to WT, KRAS– and PIK3CA-positive tumors had higher CD4+ T-cell infiltration, while EGFR-positive had significantly lower CD8+ T-cell infiltration (P<0.001, P=0.03 and P<0.0001, respectively).
Conclusion
The high neoepitope load and unique immune composition in KRAS- and BRAF-positive LUAD may be associated with better response to ICI typically observed in these tumors. High immune infiltration in MET-positive LUAD suggests that ICI may be more beneficial in these tumors. Cumulatively, our findings suggest that the significant variability in neoepitope load and immune environment of driver-gene altered LUAD might explain their differential sensitivity to ICI. Ongoing studies are evaluating the impact of neoepitope load on clinical characteristics and outcomes.