Abstract B025: Spatially informed profiling of stage I lung adenocarcinoma reveals an extensive gene expression signature of vascular invasion

Abstract

Abstract Microscopic vascular invasion (VI) is predictive of recurrence in stage I lung adenocarcinoma (LUAD) but is difficult to assess in resection specimens and cannot be accurately predicted prior to surgery. Thus, new biomarkers are needed to identify this aggressive subset of stage I LUAD tumors. To assess molecular and tumor microenvironment (TME) features associated with angioinvasive LUAD we profiled 171 resected stage I tumors with and without VI by RNA-seq, including 24 tumors by high-resolution spatial transcriptomics (stRNA-seq, 10x Genomics Visium). Visium capture areas were selected by an experienced thoracic pathologist to include invasive foci, tumor regions distal to foci, and tumors without invasive foci. We identified a molecular signature from the bulk RNA-seq discovery cohort (n=103) containing three gene expression clusters increased in VI+ stage I LUAD including genomic instability (C1), tissue remodeling (C2), and hypoxia (C3), and one increased in VI- (C4, immune surveillance). Analysis of the stRNA-seq revealed high inter-tumor patient heterogeneity, with most spots clustering by tumor identity, suggesting tumor-intrinsic properties. Scoring the stRNA-seq for VI signature clusters revealed that C2 was highly expressed outside the VI focus and declined as a function of distance toward it, while C1 and C3 increased. C2 was most strongly enriched in regions of desmoplastic but not normal stroma while C4 was enriched in normal-appearing adjacent lung and low-grade histologic patterns including lepidic and papillary. Given increased signature expression in regions of invasive tumors at a distance from VI foci, we leveraged the discovery cohort to develop a transcriptomic predictor of VI. We applied a nested cross-validation approach to select a machine learning model, which demonstrated an AUC of 0.86 for VI+ LUAD in our independent test set (n=60). In contrast, the predictor was unable to detect lymphatic invasion (LI) in VI- LUAD, suggesting a different molecular process may differentiate VI from LI. The VI-associated predictor increased across the spectrum of indolent to aggressive stage I LUAD histopathology and was predictive of recurrence-free survival, even in VI- LUAD, implying detection of pre-angioinvasive LUAD or VI foci that were missed during pathology review. Finally, we used RNA-seq data from multi-region sampling of stage I LUAD cases in TRACERx, a longitudinal study of NSCLC evolution, to evaluate the VI predictor’s robustness to intra-tumor heterogeneity (ITH). Across all tumors, we observed strong correlation between scores from randomly sampled tumor-matched regions (n=136; R = 0.91, p < 2.2e-16). The difference of scores between unmatched regions (inter-tumor heterogeneity) was significantly higher than matched regions (ITH) (p < 2.2e-16). Our study suggests that VI-associated gene expression extends from the site of intravasation and can be used to predict the presence of VI. This may enable the prediction of angioinvasive LUAD from biopsy specimens, allowing for more tailored treatment prior to surgery. Citation Format: Dylan Steiner, Lila Sultan, Jason Weis, Travis Sullivan, Emily Green, Hanqiao Liu, Sherry Zhang, Gang Liu, Avrum Spira, Sarah Mazzilli, Kimberly Christ, Eric Burks, Jennifer Beane, Marc Lenburg. Spatially informed profiling of stage I lung adenocarcinoma reveals an extensive gene expression signature of vascular invasion [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Translating Cancer Evolution and Data Science: The Next Frontier; 2023 Dec 3-6; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(3 Suppl_2):Abstract nr B025.