Abstract
PURPOSE: The current study evaluated associative effects of breast cancer cells with the tumor microenvironment and its influence on tumor behavior. EXPERIMENTAL DESIGN: Formalin-fixed, paraffin-embedded tissue and matched protein lysates were evaluated from two independent breast cancer patient datasets (TCGA and MD Anderson). Reverse-phase protein arrays (RPPA) were utilized to create a proteomics signature to define breast tumor subtypes. Expression patterns of cell lines and normal breast tissues were utilized to determine markers that were differentially expressed in stroma and cancer cells. Protein localization and stromal contents were evaluated for matched cases by imaging. RESULTS: A subtype of breast cancers designated "Reactive," previously identified by RPPA that was not predicted by mRNA profiling, was extensively characterized. These tumors were primarily estrogen receptor (ER)-positive/human EGF receptor (HER)2-negative, low-risk cancers as determined by enrichment of low-grade nuclei, lobular or tubular histopathology, and the luminal A subtype by PAM50. Reactive breast cancers contained high numbers of stromal cells and the highest extracellular matrix content typically without infiltration of immune cells. For ER-positive/HER2-negative cancers, the Reactive classification predicted favorable clinical outcomes in the TCGA cohort (HR, 0.36; P < 0.05). CONCLUSIONS: A protein stromal signature in breast cancers is associated with a highly differentiated phenotype. The stromal compartment content and proteins are an extended phenotype not predicted by mRNA expression that could be utilized to subclassify ER-positive/HER2-negative breast cancers. Clin Cancer Res; 22(20); 5068-78. ©2016 AACR.