TMIC-89. Spatial transcriptomics reveals tumor–brain expression convergence and identifies prognostic gene signatures in breast cancer brain metastases

Abstract BACKGROUND As cancer therapies improve and patients live longer, the incidence of brain metastases (BrMets) continues to rise. Among women with metastatic breast cancer, 10-15% develop BrMets, with rates as high as 30% in HER2+ and 50% in triple-negative. BrMets, represent highly evolved, therapy-resistant tumors with a median survival of 10 months and accompanied by severe neurologic decline. BrMets often emerge years after initial diagnosis, suggesting that metastatic cells must acquire brain-specific traits to survive. This process may involve transcriptional convergence—where tumor cells adopt brain-like programs, and brain-resident cells exhibit tumor-associated changes. Despite its clinical relevance, the molecular basis of this convergence remains poorly defined. Decoding, and targeting, these shared gene programs could reveal new therapeutic vulnerabilities. METHODS We performed spatial transcriptomic profiling on 235 patient tissue cores (BrMets, adjacent normal brain, primary breast tumors, and non-cancer brain). Using NanoString GeoMx, DSP, we measured the expression of 18,677 RNAs across 450 spatially-defined regions enriched for tumor, immune, and brain-resident cells. To interrogate transcriptional convergence, we developed the Equivalent Expression Index, a novel statistical tool to detect genes with biologically meaningful expression similarity across distinct cell populations. RESULTS Our integrated analysis revealed two prognostic gene signatures. The Metastasis-Induced Brain Shared (MIBS-9) signature, comprising brain-like genes upregulated in tumor cells, was associated with longer patient survival. Conversely, the Adjacent brain-Resident Cancer Shared (ARCS-81) signature, defined by tumor-like gene expression in brain-resident cells, correlated with shorter survival. These signatures were externally validated and enriched for pathways related to neuronal remodeling, immune evasion, cellular communication, and cancer/metastasis. CONCLUSION By coupling spatial transcriptomics with a novel equivalence-based algorithm, we uncover distinct, clinically relevant gene programs that mediate bidirectional adaptations between tumor and brain-resident cells. These findings illuminate key mechanisms of metastatic outgrowth and offer new targets for disrupting the breast cancer BrMet microenvironment.