Abstract 7489: Archetype analysis links heterogeneity in premalignant lesions to diverging features of lung adenocarcinoma

Abstract Lung cancer is the leading cause of cancer-related death. Adenocarcinoma (LUAD), the most common form of lung cancer, is a heterogeneous disease with high variability in histopathologic features and clinical outcomes. Atypical adenomatous hyperplasia and adenocarcinoma in situ are the premalignant lesions (PML) in LUAD pathogenesis. The molecular features of PML that lead to aggressive LUAD are poorly characterized. We hypothesized that transcriptomic changes in PML are linked to distinct clinicopathologic features that may identify aggressive PML before they become invasive LUAD. We examined lung resections from patients with early-stage LUAD where adjacent tissue contained premalignant histology. Normal, PML, and LUAD areas were laser-capture micro-dissected, and RNA was isolated and then profiled using bulk RNA-sequencing (n=41 patients, n=177 samples). We derived a set of recurrent gene modules (n=9 modules) in LUAD PML using samples from our cohort (n=80 PML samples) and three publicly available data sets (n=57 LUAD PML samples). Using these modules, we identified four PML archetypes, where the vertices of a polytope define each archetype. Based on pathway (Gene Set Enrichment Analysis, FDR <0.01) and signature enrichment, we identified the four archetypes as: normal-like, inflammation, cell adhesion, and proliferation. Analyzing the expression of published gene signatures suggests that the immune environment and epithelial cell plasticity may play different roles in different archetypes. Archetypes also differ with regard to a gene-expression signature for predicted immunotherapy response, with the proliferation archetype predicted to have a poor response, and the inflammation archetype predicted to have a good response to immunotherapy (logistic regression, p<0.05). An advantage of the archetyping approach is that archetype space is continuous, and samples from other cohorts can be placed within the PML-derived polytope based on their transcriptomic profile. We found that the PML-derived archetypes were related to clinical features in studies of LUAD. The proliferation archetype was associated with a decrease in disease free survival in multiple datasets (cox-proportional hazard model, p<0.05 for each tested cohort), and aggressive LUAD histology (linear model, p<0.05 for each tested cohort). While past studies demonstrated transcriptional differences between tumors and PML, our archetyping approach focuses on PML-specific gene expression patterns using multiple cohorts to increase the diversity of PML samples studied. Gene expression differences among archetypes were related to pathway dysregulation and prognosis. Characterizing PML based on transcriptomic features may improve our understanding of divergent mechanisms of carcinogenesis, and enable identification of aggressive PML capable of developing into invasive LUAD. Citation Format: Kelley Anderson, Linh Tran, Kostyantyn Krysan, Gregory A. Fishbein, William D. Wallace, Emily Green, Hanqiao Liu, Xiaohui Xiao, Gang Liu, Erin Kane, Avrum Spira, Steven M. Dubinett, Eric Burks, Sarah Mazzilli, Marc Lenburg, Jennifer Beane. Archetype analysis links heterogeneity in premalignant lesions to diverging features of lung adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 7489.