In Luminal A breast cancer, bulk FOXC1 reports the tumour vasculature, while a proliferation-independent basal-lineage axis marks the tumour cells

Abstract

FOXC1 is widely interpreted as a tumour-cell driver of basal identity and immune phenotype in breast cancer, including within the Luminal A (LumA) subtype, where its bulk expression tracks basal-like features and immune infiltration. Because LumA tumours are compositionally heterogeneous, whether this reflects malignant-cell biology or the surrounding microenvironment has not been resolved, and the distinction changes how the biomarker should be read. Combining purity- and compartment-adjusted partial correlations across three LumA cohorts on two platforms (TCGA, n = 571; METABRIC, n = 700; SCAN-B, n = 1,540) with single-cell data from an ER-positive atlas (GSE176078; 100,064 cells), we show that bulk FOXC1 is principally a readout of the tumour vasculature: roughly 90% of FOXC1 transcripts arise from endothelial and perivascular cells and 1.8% from malignant epithelium, and FOXC1 is detected in 28% of endothelial versus 0.5% of malignant cells. Its coupling to adaptive-immune and tertiary-lymphoid-structure programmes collapses under vascular adjustment, whereas basal cytokeratins (KRT5, KRT14, KRT17) and TP63 retain a residual, population-level basal-lineage signal, the Centaur axis, that is orthogonal to proliferation and to established risk tools; an apparent survival association is explained by age. These findings reassign a presumed tumour-cell programme to the vessel wall, recover a genuine basal-lineage axis within LumA tumour cells, and show that compartment-aware analysis is needed to interpret bulk transcriptomic biomarkers in compositionally heterogeneous tumours. Impact statement In Luminal A breast cancer, single-cell and compartment-adjusted analysis shows that bulk FOXC1 reports the tumour vasculature rather than the malignant cells, while a genuine basal-lineage axis persists in the tumour cells.