A gene program association study (GPAS) delineates novel patterns of lineage plasticity and aggressiveness in non-small cell lung cancer

Combinations of molecular programs can be altered in cancer cells and contribute to tumor initiation and aggressiveness. We developed a novel framework called a Gene Program Association Study (GPAS) to create an atlas of gene co-expression modules in cancer cells from 127 primary non-small cell lung cancers (NSCLCs) and determine their association with clinical and histopathological features. Broadly expressed housekeeping modules were explained by study-specific effects while other categories of modules were better explained with other biological or clinical factors. Lung adenocarcinoma (LUAD) displayed extensive lineage infidelity characterized by heterogeneity in expression of modules from different alveolar cell types. Modules associated with late-stage LUAD were found in cells from early-stage tumors suggesting that aggressive phenotypes can be observed before clinical progression. Basal- and squamous-like cells were identified as possible intermediates in the formation of invasive mucinous LUAD. In lung squamous cell carcinoma (LUSC), a novel subtype was identified with lower levels of canonical squamous modules and higher levels of fibrinogen. LUSC-associated modules were also observed in normal basal cells exposed to cigarette smoke suggesting that this cell state is the more proximal "cell-state-of-origin" than unexposed basal cells. Overall, this atlas elucidated the combinations of molecular programs in cancer cells that contribute to heterogeneity and aggressiveness in NSCLC.