The Identification of Novel Prognostic and Predictive Biomarkers in Breast Cancer via the Elucidation of Tumor Ecotypes Using Ecotyper.

Background: Breast cancer is a highly heterogeneous disease, characterized by tumor and nontumor cells at various cell states. Ecotyper is an innovative machine learning framework that quantifies the tumor microenvironment and delineates the tumor ecosystem, demonstrating clinical significance. However, further validation is needed in breast cancer.

Methods: Ecotyper was applied to identify multiple cellular states and tumor ecotypes using large-scale breast cancer bulk sequencing data, followed by a detailed analysis of their associations with clinical classification, molecular subtypes, survival prognosis, and immunotherapy response. Identified subtypes were further characterized using single-cell and spatial data sets to reveal molecular profiles.

Results: In a comprehensive analysis of 6578 breast cancer samples from four data sets, Ecotyper identified 69 cellular states and 10 tumor ecotypes. Of these, 37 cellular states significantly correlated with overall survival. Notably, specific states within epithelial cells, macrophages/monocytes, and fibroblasts were linked to a worse prognosis. CE2 abundance was identified as the most significant marker indicating unfavorable prognosis and was further validated in an additional data set of 116 HER2-negative patients. These biomarkers also indicated the efficacy of neoadjuvant immunotherapy in breast cancer. CE2-high cancers were characterized by an abundance of basal-like epithelial cells, scant lymphocytic infiltration, and activation of hypoxia signaling. Single-cell analysis showed that CE2-high areas were rich in SPP1-positive tumor-associated macrophages(TAM), basal-like epithelial cells, and hypoxic cancer-associated fibroblasts(CAF). Spatially, these regions were often peripheral in triple-negative breast cancer, adjacent to fibrotic/necrotic zones. Multiplex immunofluorescence confirmed the enrichment of SPP1+CD68+TAM and HIF1A+SMA+CAF in hypoxic triple-negative breast cancer (TNBC) regions.

Conclusions: Ecotyper identified novel biomarkers for breast cancer prognosis and treatment prediction. The CE2-high region may represent a hypoxic immune-suppressive niche.