B111-09 Multi-Omic Profiling Defines Distinct Primary Tumor Ecosystems That Predetermine Metastatic Organotropism in Non-Small-Cell Lung Cancer

Abstract Rationale Organ-specific metastasis, such as to bone or pleura, in lung adenocarcinoma (LUAD) drives poor prognosis, yet the mechanisms governing this organotropism remain poorly understood. We hypothesized that the propensity for organ-specific metastasis is predetermined by distinct cellular and molecular features already present in the primary tumor ecosystem, encompassing both malignant cells and the tumor microenvironment (TME). This study aimed to identify these pre-existing signatures within primary LUADs to understand organ-specific dissemination. Methods We performed single-cell RNA sequencing (scRNA-seq) on 19 primary LUAD tumors, stratified by metastatic status: non-distant metastatic Stage III (n = 4), intra-pulmonary metastasis (IM, n = 4), pleural metastasis (PM, n = 3), bone metastasis (BM, n = 5), and combined bone and pleural metastasis (BPM, n = 3). An independent validation cohort (n = 48), including non-metastatic and metachronous metastatic cases, was analyzed by bulk RNA-seq. We applied Generalized Binary Covariance Decomposition (GBCD) to identify tumor-intrinsic gene expression programs (GEPs) and utilized CoVarNet, a computational framework, to deconstruct co-occurring cellular modules within the TME. Results Within the malignant cell compartment, GBCD analysis identified distinct GEP modules strongly correlated with metastatic sites. A specific GEP module was significantly upregulated in the BM group, with EXT1, PVT1, and RANBP17 as the genes with the highest fold-change. Conversely, primary tumors from the PM group were characterized by a different GEP module, with top-ranking genes including SOX4, FGF13, and MACC1. Bulk RNA-seq deconvolution analysis confirmed the differential abundance of these organ-specific GEP signatures in the validation cohort. Regarding the TME, CoVarNet analysis revealed a distinct cellular module, CM04, which was maximally enriched in the BM group. CXCL10+ macrophages were identified as a critical component of this module. Notably, the abundance of CM04 was significantly lower in Stage III and IM samples compared to both BM and BPM groups. Furthermore, pseudotime analysis indicated that the CM04 module score progressively increased along the differentiation trajectory, implying its association with an evolved, pro-metastatic TME phenotype. Conclusion Our study provides high-resolution evidence that primary LUADs harbor pre-existing, organ-tropic signatures. We identified distinct tumor-intrinsic GEPs associated with metastatic destination: one module (incl. EXT1, PVT1, RANBP17) for bone metastasis, and another (incl. SOX4, FGF13, MACC1) for pleural metastasis. We further defined a co-occurring TME module (CM04, centered around CXCL10+ macrophages) supporting the bone-tropic ecosystem. These findings support metastatic predetermination at the primary site and highlight these specific cellular and molecular axes as promising biomarkers for risk stratification and potential therapeutic targets to intercept organ-specific metastatic progression. This abstract is funded by: National Natural Science Foundation of China