Abstract 3517: Multi-modal single-cell and spatial genomics reveals genomic, adaptive and microenvironmental features of human non-small cell lung cancer brain metastasis

Abstract

Non-small cell lung cancer (NSCLC) accounts for nearly half of all newly diagnosed patients with brain metastasis (BM), followed by melanoma and breast carcinomas. The presence of BM is associated with reduced response to several modern cancer therapies and a poor prognosis, but the underlying molecular underpinnings remain poorly understood. Here, we performed multi-modal single-nucleus RNA, T cell receptor, single-cell spatial, and whole-genome sequencing (WGS) of 44 primary NSCLC tumors (PTs) and BMs. Through combination of WGS with inferred copy-number alterations (CNAs) and gene expression snRNA-seq, we robustly identify malignant cells despite the presence of healthy cell mosaicism. We find a strong association of chromosomal instability (CIN) and brain-metastatic organotropism. Through integration with clinical information and thousands of publicly available whole-exome sequencing (WES) profiles obtained from patients with NSCLC, we validate this observation and show that CIN progressively increases from PTs to extracranial metastases (ECMs) and is the highest in BMs. Using non-negative matrix factorization, we identify recurrent transcriptional hallmarks cancer metastasis, and additionally find that cancer cells from BMs strongly enrich for a neuronal-like cell state. At single-cell resolution, we indeed identify a rare cancer cell population genomically define by very high CIN, and transcriptionally characterized by a program of epithelial-to-mesenchymal transition (EMT), neuronal-like differentiation, and loss of lineage attribution. We show in our data and external scRNA-seq data that this cell state does not ecist in healthy lungs, progressively enriches from PTs to ECMs, and is most abundant in BMs, suggesting that these cells may indeed give rise to BMs. Furthermore, through integration of snRNA/TCR-seq and spatial transcriptomics, we find distinct tumor-microenvironments across disease sites, including, nearly exclusive expansion of tissue-resident myeloid cells in PTs, while BMs are largely dominated by dense infiltration with monocyte-derived macrophages and granulocytes, impaired T cell infiltration, activation and clonal expansion. Lastly, spatial transcriptomics also recurrent, cell-type specific patterns of geographic variability in key pathways, including antigen presentation, EMT, oxidative phosphorylation, and inflammatory response and associated cellular micro-niches. Together, this work identifies cellular, genomic, and transcriptional features of NSCLC BMs and has important therapeutic implications for novel therapies, in particular immunomodulatory approaches targeting cell types/states unique to disease sites. Citation Format: Somnath Tagore, Lindsay Caprio, Amit Dipak Amin, Irving Barrera, Johannes Melms, Karan Luthria, Yiping Wang, Yohanna Georgis, Abhi Jaiswal, Galina G. Lagos, Zachary Walsh, Parin Shah, Jana Biermann, Neha Sheikh, Priyanka Ramaradj, Niroshana Anandasabapathy, Hanina Hibshoosh, Gary Schwartz, Brian Henick, Alison Taylor, Fei Chen, Benjamin Izar. Multi-modal single-cell and spatial genomics reveals genomic, adaptive and microenvironmental features of human non-small cell lung cancer brain metastasis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3517.