A stratification system for breast cancer based on basoluminal tumor cells and spatial tumor architecture

Rapid recurrence is common in triple-negative breast cancer (TNBC). To better understand drivers of recurrence, we use imaging mass cytometry to characterize the tumor phenotype landscapes of 215 TNBC patients. We observe high intertumor heterogeneity with eleven tumor cell phenotypes, each of which dominates in an individual patient, and identify a tumor cell phenotype with reduced basoluminal lineage fidelity and stem-like traits that is correlated with rapid disease recurrence. Scoring of tumor-CD8⁺ T cell interactions identifies patients with inflamed tumors and high HLADR expression. We combine these features in multi-omics analyses of 8 cohorts with 3737 patients across all molecular subtypes to propose five prognostic breast cancer subtypes distinguished by tumor cytokeratin expression profiles and CD8⁺ T cell spatial patterns. This stratification scheme has direct clinical implications: inflamed tumors show good prognosis and high immunotherapy response rates, whereas patients dominated by basoluminal tumor cells have poor prognosis.