B1 - Cell Type-Informed and Spatially Resolved Subtyping Reveals Mechanistic Diversity and Therapeutic Opportunities in Triple-Negative Breast Cancer

Introduction: : Triple-negative breast cancer (TNBC) is a clinically aggressive and biologically diverse disease, with disproportionate incidence and poor outcomes among African American (AA) women. While existing gene expression-based TNBC subtyping frameworks provide insights into molecular heterogeneity, they lack consideration of the effect of tumor microenvironment (TME) and cellular composition. Our primary hypothesis is that the tumor heterogeneity should be reflected in the tissue and cell heterogeneity and if we can use the latter through cell-specific markers, we can capture not only tumor heterogeneity but also the underlying cellular mechanisms driving it. To this end, we developed a cell type-informed subtyping model using bulk RNA-sequencing and validated it with high-resolution spatial transcriptomic (GeoMx) profiling of 76 tumors from a racially diverse cohort (250 TNBC patients, 46% AA). This integrative framework refines current TNBC classification and reveals spatially organized heterogeneity of TNBC with potential therapeutic implications.

Materials and

Methods: : Bulk RNA-seq data from 250 TNBC patients (123 EA, 127 AA) were classified using hierarchical (HC) and consensus clustering (CC) on 261 curated breast cell type markers. Differentially expressed genes (DEGs) were identified at p-adj ≤ 0.05. Spatial transcriptomic profiling was performed on FFPE tissue from 76/250 TNBC patients using GeoMX® Digital Spatial Profiler. Regions of interest were segmented on pan-cytokeratin and analyzed for tumor and stromal components, probed using GeoMx human whole transcriptome atlas (WTA) panel. Data quality control, normalization, and differential analysis was performed in R using GeomxTools and linear mixed models. Subtypes (CC1–3) were characterized by differential expression analysis (FDR ≤ 0.05), transcription factor (TF)-target enrichment, protein-protein interaction (PPI) networks, and log-rank survival analysis.

Results, Conclusions, and Discussions:: : Our integrative analysis of bulk and spatial transcriptomics defined three mechanistic TNBC subtypes (CC1–3), each with distinct biomarkers and cellular programs. Spatial profiling across 76 tumors validated and refined these subtypes by revealing intra-tumoral heterogeneity and spatially organized immune, stromal, and epithelial features. Patients in CC1 were enriched for interferon signaling, CD8+ T cell activation, and increased exhaustion markers in bulk analysis. This subtype exhibited compartment-specific activation including IFN-γ signaling and MHC-I antigen presentation in the tumor center (IRF1, STAT1), while type I IFN signaling, innate/adaptive immunity (IRF7, ISGs, PDCD1, CD274), and elevated exhaustion markers at the edge—suggesting potential responsiveness to immune checkpoint blockade plus chemotherapy. Patients in CC2 were enriched for ECM organization and characterized as immune excluded in bulk analysis. This subtype showed a fibrotic center with increased collagen markers, while TGF-β signaling and matrix remodeling (THBS1/2, SPON1/2, integrins) were at the edge, aligning with a fibrotic, immune-excluded phenotype. We observed progressive mesenchymal/fibroblast cells from center to invasive edge to foci. Patients in CC3 showed enrichment for fatty acid, hormonal, and inflammatory signaling in bulk analysis. This subtype showed enrichment for Wnt signaling and EMT (AXIN2, TGFBR1) at the center, while EMT, angiogenesis towards the edge with infiltration of fibroblast/mesenchymal and myeloid cell types. Foci showed activation of DNA replication, OXPHOS, and stress-adaptive programs (CHEK1, MRPLs, NFKBIA) alongside fibroblast/mesenchymal and myeloid cell types. Pairwise comparisons confirmed that CC3 sits on a biological spectrum between immune-active (CC1) and ECM-dominant (CC2) subtypes with unique features including hormonal/nuclear receptor and fatty acid metabolism. Survival analysis demonstrated favorable outcomes for CC1 and less favorable outcomes for CC2. We developed a user-friendly Shiny App for our subtyping method and validated it using external datasets including TCGA.
This integrative approach reveals that TNBC subtypes are defined not only by dominant gene programs but also by their spatial organization within tumors. Importantly, CC3 represents a transitional, mixed phenotype, potentially responsive to combination therapies. These insights underscore the need for spatially informed subtyping frameworks to guide TNBC classification, prognosis, and treatment strategies. Our method can be applied to other cohorts using the Shiny application developed.

Acknowledgements and/or References (Optional):: Supported by a grant from the Wellcome Leap Foundation Delta Tissue Project.