Purpose: Lung adenocarcinomas comprise the largest fraction of non-small cell lung cancer, which is the leading cause of cancer-related deaths. Seventy-five percent of adenocarcinomas lack targeted therapies because of scarcity of druggable drivers. Here, we classified tumors on the basis of signaling similarities and discovered subgroups within this unmet patient population.
Experimental design: We leveraged transcriptional data from >800 early- and advanced-stage patients.
Results: We identified three robust subtypes dubbed mucinous, proliferative, and mesenchymal with respective pathway phenotypes. These transcriptional states lack discrete and causative mutational etiology as evidenced by similarly distributed oncogenic drivers, including KRAS and EGFR. The subtypes capture heterogeneity even among tumors lacking known oncogenic drivers. Paired multi-regional intratumoral biopsies demonstrated unified subtypes despite divergently evolved prooncogenic mutations, indicating subtype stability during selective pressure. Heterogeneity among in vitro and in vivo preclinical models is expounded by the human lung adenocarcinoma subtypes and can be leveraged to discover subtype-specific vulnerabilities. As proof of concept, we identified differential subtype response to MEK pathway inhibition in a chemical library screen of 89 lung cancer cell lines, which reproduces across model systems and a clinical trial.
Conclusions: Our findings support forward translational relevance of transcriptional subtypes, where further exploration therein may improve lung adenocarcinoma treatment.See related commentary by Skoulidis, p. 913.
©2020 American Association for Cancer Research.