Abstract PS2-11-25: Paracrine FGFR activation by stromal signals and its association with immune infiltration in breast cancer

Abstract Background: Fibroblast growth factor receptor (FGFR) alterations occur in up to 20% of estrogen receptor-positive (ER+) breast cancers and are associated with endocrine therapy resistance and poor prognosis. These tumors often classify as high-risk integrative cluster subtypes, which account for approximately one-quarter of ER+ tumors, are characterized by complex genomic rearrangements, and are associated with increased and persistent risk of relapse. In prior work, we found that high-risk integrative subtypes of ER+ breast cancer were significantly more likely to exhibit immune depletion compared to typical-risk subtypes, an observation especially pronounced among tumors with putative FGFR pathway alterations. Despite the clinical rationale, therapeutic efforts to directly target FGFR signaling in breast cancer have largely not been successful. To better understand the role of FGFR in shaping the tumor microenvironment and to identify potential therapeutic vulnerabilities, we employed multiplex single-molecule RNA imaging on the Bruker CosMx platform to profile stromal interactions and immune contexture in primary and metastatic ER+ breast tumors. Methods: We constructed a single-cell spatially resolved transcriptomic meta-cohort of ER+ breast cancer, consisting of primary pre-treatment tumors (208 primaries), 69 patients treated per standard of care at Stanford University (27 primaries, 66 metastatic sites), and 4 patients treated on an investigator-sponsored trial of the FGFR inhibitor infigratinib (3 primaries, 6 metastatic sites). We deployed state-of-the-art machine learning algorithms for cell segmentation (CellPose), cell type classification (singleR), and inference of intercellular communication (CellChat). We identified stromal-tumor interactions enriched in high vs. typical-risk primary tumors and compared the microenvironmental features associated with those interactions. Results: We identified recurrent interactions between tumor-expressed FGFR1 and fibroblast-derived FGF1 or FGF7, significantly enriched in high-risk (17/65) compared to typical-risk tumors (2/87, P