Exploratory integration of spatial transcriptomics and MRI phenotypes to characterize metastasis-associated microenvironmental features in ipsilateral breast tumor recurrence

Ipsilateral breast tumor recurrence (IBTR) presents a critical clinical challenge, as it is strongly associated with a high risk of subsequent distant metastasis, the primary cause of mortality. Although the spatial organization of the tumor microenvironment (TME) may contribute to metastatic progression, the spatially resolved cellular architecture of IBTR and its relationship with clinically available imaging phenotypes remain insufficiently characterized. We performed an exploratory analysis integrating high-resolution spatial transcriptomics using the CosMx Spatial Molecular Imager (SMI) platform with qualitative breast magnetic resonance imaging (MRI) phenotyping. Spatial transcriptomic profiling was conducted in 21 patients with IBTR, including 8 patients who developed distant metastasis and 13 who remained metastasis-free during follow-up. Among them, 9 patients had available preoperative contrast-enhanced MRI for imaging–spatial association analysis. Spatial transcriptomic data were used to construct a cellular atlas, identify multicellular spatial niches using a latent Dirichlet allocation-based approach, and infer putative ligand–receptor interactions using CellChat. MRI phenotypes, including background parenchymal enhancement (BPE), lesion type, multifocality, and peritumoral edema, were assessed and correlated with spatial cellular and molecular features. Spatial transcriptomic analysis identified diverse immune, epithelial, stromal, and endothelial populations within the IBTR microenvironment. Compared with non-metastatic cases, metastatic cases showed enrichment of selected cancer-associated fibroblast (CAF) subtypes and specific immune cell populations. Nine multicellular spatial niches were identified, among which Niche 1 was more abundant in metastatic cases and was characterized by co-enrichment of epithelial cells, CAFs, and tumor-associated macrophages (TAMs). Within this niche, CellChat analysis suggested putative COL1A1–SDC1 and COL1A1–CD44 ligand–receptor interactions between stromal and epithelial compartments. In the MRI-linked subset, high BPE and peritumoral edema were associated with higher abundance of Niche 1 and related cellular components. Gene set enrichment analysis (GSEA) further suggested that high BPE was associated with IFN-α/β pathway enrichment, whereas EMT-related signatures were observed in low-BPE, mass-enhancement, and no-edema subgroups, indicating that distinct MRI phenotypes may capture different biological programs. This exploratory study provides a spatially resolved overview of the IBTR microenvironment and identifies candidate MRI-associated spatial features linked to metastatic status. The observed associations between BPE, peritumoral edema, and metastasis-associated spatial niches generate hypotheses for future radiogenomic studies, but require validation in larger, independent cohorts before clinical application.