Abstract A037: Single-cell spatial immune architectures shape ctDNA release and prognosis in luminal breast cancer

Abstract Background The release of circulating tumor DNA (ctDNA) is modulated not only by intrinsic tumor features but also by the tumor microenvironment (TME). In luminal breast cancer, distinct microenvironment niches may impact ctDNA release, yet the cellular and molecular mechanism that link TME composition to ctDNA dynamics remain poorly defined. Methods We integrated single-cell full-length transcriptomic, mutational and spatial data from tumor tissue with mutation profiles from paired plasma of ten treatment-naïve early-stage HR+/HER2- breast cancers. A deep-learning multi-omics framework mapped ctDNA release at single-cell resolution by tracing mutations shared between tumor subclones and plasma ctDNA. First, we found microenvironmental niches linked with ctDNA release. Then for every malignant cell, we calculated the immune neighborhood and performed unsupervised clustering of epithelial cells, yielding five epithelial clusters. Afterwards, five TME-neighborhood signatures derived from these clusters were applied to METABRIC and TCGA cohorts to stratify patients into five microenvironment-defined groups. Normalized ctDNA release scores, mutation burden, and functional signatures were compared across groups; prognostic value was evaluated with 21-gene and 70-gene assays and 5-year survival. Results Microenvironmental niche analysis identified three TME architectures associated with ctDNA release: an intra-ductal luminal-progenitor niche attracting B cells, a hot immune niche enriched for T cells and macophages, and an immunomodulated, stemness niche featured by M2 macrophages. Furthermore, integration of single-cell neighborhoods with bulk transcriptomes aligned five microenvironment-defined patient groups (G1–G5) to the underlying cell clusters (C1–C5). G1/C1—enriched for intra-ductal luminal progenitors, B cells, and DCs —showed the highest per-epithelial-cell ctDNA release, yet the lowest total ctDNA burden because of its small malignant compartment. G2 and G3, linked to C2/C3, displayed antigen-presentation and estrogen-response phenotypes enriched for T cells and macrophages. G4/C4 formed an immune-quiescent compartment. G5/C5 combined the largest malignant compartment with prominent M2-TAM, DC3 and pDC infiltration plus high CD274 expression, producing the highest ctDNA release and the most immunomodulatory microenvironment. Clinically, the gradient from G1 to G5 paralleled increasing 21-gene and 70-gene recurrence scores and decreasing survival (p < 0.001). Elevated ctDNA-release signatures independently associated with worse 5-year survival (HR = 1.63, 95 % CI 1.2–2.2), confirming microenvironment-shaped ctDNA dynamics as a non-invasive prognostic biomarker. Conclusion TME architecture impacts ctDNA release in luminal breast cancer. Immune-rich, B-cell/DC niches limit total shedding despite high single-cell release efficiency, whereas immunomodulatory, PD-L1-high niches maximize ctDNA output and portend poor prognosis. TME-informed ctDNA metrics provide a non-invasive tool for immune monitoring and precision risk stratification. Citation Format: Hengyi Xu, Pengming Pu, Jiaqi Liu. Single-cell spatial immune architectures shape ctDNA release and prognosis in luminal breast cancer [abstract]. In: Proceedings of the AACR Immuno-Oncology Conference (AACR IO): Discovery and Innovation in Cancer Immunology: Revolutionizing Treatment through Immunotherapy; 2026 Feb 18-21; Los Angeles, CA. Philadelphia (PA): AACR; Cancer Immunol Res 2026;14(2 Suppl):Abstract nr A037.