The tumor microenvironment of 14,837 breast cancers is associated with clinical outcome independently of genomic subtypes.

The tumor microenvironment (TME) contributes to breast cancer heterogeneity and outcome but is rarely considered in clinical decision-making. We address this gap by systematically characterizing the TME's cellular composition to establish its independent clinical utility across intrinsic and genomic subtypes. We first compare 15 TME profiling methods in 693 samples and then apply the deconvolution algorithm InstaPrism to a meta-dataset of 14,837 expression profiles. We identify seven distinct TME patterns that associate with disease-free survival independently of intrinsic subtype. We also identify TME features that modulate chemotherapy response, relapse, and metastatic risk, with divergent patterns observed across estrogen receptor subtypes. Notably, long-term recurrence was regulated by vascular stromal cells and the innate immune response. Furthermore, the depletion of B cell lineage derivatives in metastatic lesions suggests an opportunity for therapeutic intervention. These results provide evidence for using TME characterization as a prognostic and predictive biomarker and identify potential targets for TME-based intervention.