Abstract 759: Spatial transcriptomics identifies unique tumor and microenvironment pathomic programs that are associated with the lung premalignancy and adenocarcinoma continuum

Abstract Background: Lung adenocarcinoma (LUAD) is one of the most prevalent and lethal cancer types worldwide. However, our understanding of the transition from normal-appearing tissue (NAT) to adenomatous lung premalignant lesions (aPMLs) and LUADs, particularly within a spatial context, remains limited. This study aims to address this gap by systemically analyzing the pathologic continuum of NAT > aPML > LUAD using spatial transcriptomics (ST). Methods: High-resolution spatial profiling was conducted on 56 samples from 25 patients with paired aPMLs and LUADs using the Visium ST platform. Non-negative matrix factorization was conducted to identify transcriptional programs for each sample, following clustering analysis to define consensus metaprograms across the cohort. Additionally, spatial molecular imaging was performed on an expanded cohort of 188 cores arranged into eight tissue microarrays using the Xenium in situ platform with a customized lung cancer gene panel to establish a single-cell spatial atlas of the disease continuum and systemically investigate spatial organization, cellular neighborhoods and interactions. Results: Eight distinct metaprograms (MP1∼8) were identified, distinguishing stromal (MP2), myeloid (MP4), lymphoid (MP6), and epithelial (MP3, MP5) compartments. Additionally, metaprograms were identified for the lung capillary bed (MP7), stressed cellular state (MP8), and mosaic cellular patterns (MP1). These metaprograms correlated strongly with pathological annotations and captured fine tissue structures like lymphoid aggregates. LUADs and aPMLs showed distinct MP profiles, with MP3 and MP6 being highly abundant in LUADs, while MP1, MP4, MP5 and MP7 were more abundant in aPMLs. Specifically, we observed positive correlations between MP3 and MP6, as well as between MP4 and MP5 in sample distributions, indicating co-evolution of tumor and immune microenvironments during disease progression. Xenium data complemented these findings by revealing distinct compositions of microenvironmental cellular states and tumor neighborhood structures across aPML and LUAD lesions. A total of 44 cell states were defined. Specifically, NK cells were depleted in aPMLs and LUADs compared with NAT, accompanied by an enrichment of regulatory T cells and myofibroblastic cancer-associated fibroblasts (CAFs) in the latter. Additionally, pro-inflammatory cellular subsets, including IL1B+ macrophages and IL6+ inflammatory CAFs were more abundant in tumor neighborhoods of aPMLs compared to LUADs, potentially contributing to the pre-malignant to malignant transition. Conclusions: This study offers a comprehensive molecular and cellular landscape of aPML and LUAD, laying a crucial foundation for future mechanistic studies aimed at early disease interception. Citation Format: Yibo Dai, Fuduan Peng, Ansam Sinjab, Sujuan Yang, Minyue Chen, Warapen Treekitkarnmongkol, Lorena I. Gomez Bolanos, Tieling Zhou, Alejandra G. Serrano, Jianlong Liao, Guangsheng Pei, Yunhe Liu, Yang Liu, Jiahui Jiang, Kyung Serk Cho, Yanshuo Chu, Kai Yu, Ruiping Wang, Jiping Feng, Zahraa Rahal, Guangchun Han, Naoe Jimbo, Takuo Hayashi, Satsuki Kishikawa, Kazuya Takamochi, Akshay Basi, Avrum Spira, Steven Dubinett, Tomokazu Itoh, Takashi Yao, Kenji Suzuki, Luisa M. Solis, Stephen Swisher, Mingyao Li, Junya Fujimoto, Ignacio I. Wistuba, Jared Burks, Kadir Akdemir, Hind Refai, Katy Rezvani, Jeffrey Myers, Humam Kadara, Linghua Wang. Spatial transcriptomics identifies unique tumor and microenvironment pathomic programs that are associated with the lung premalignancy and adenocarcinoma continuum [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 759.