Interaction Between Stoma And Tumor Characteristics As A New Prognostic And Predictive Marker In Breast Carcinomas.

Abstract Background. We recently demonstrated that primary breast tumors can be classified based upon extracellular matrix composition (ECM) (Bergamaschi et al.) suggesting stroma characteristics could influence tumor progression. Aim of this study is to further investigate the robustness of this classification and its impact in tumor progression and response to therapy. Methods. The expression profile of ECM-related genes was analyzed by unsupervised hierarchical clustering in 10 independent datasets of breast tumors, counting more than 1000 samples and the prognostic and predictive value of this signature were evaluated in two datasets of not treated patients and in one dataset of neo-adjuvant treated patients, respectively. Results. Only one of ECM subsets (ECM3) showed a homogeneous gene pattern that consistently allowed the classification of an independent group of tumors in all tested datasets. ECM3 is characterized by highly correlated over expression of 34 ECM genes encoding mainly structural proteins. From 24 to 38% of cases were ECM3-enriched and were mainly estrogen receptor-positive and low grade (p<0.0001). The ECM3 showed inconsistent association with DMFS in the two data sets of not treated patients, whereas the interaction between ECM3 and grade showed strongly and significant association with DMFS (HR=5.35, p=0.0012). In particular, multivariate analysis of covariates available (size, ER, age and ECM) indicated that ECM3 is significantly associated with worse DMFS in grade III patients (HR=2.5, p=0.0386), whereas it is slightly correlated with better DMFS in grade I-II patients (HR=0.6, p=0.088). It is noteworthy that the probability of 10-year DMFS was 90% in ECM3 versus 74% in non-ECM3 patients in differentiated tumors, and 75% in non-ECM3 versus 41 % in ECM3 grade III tumors. The interaction between ECM3 and grade remained significantly associated with DMFS also in ER negative (HR=14.2, p=0.03) and ER positive (HR=3.34, p=0.079) subgroups. ECM3, indeed, was correlated with good and bad prognosis in differentiated and grade III tumors, respectively. Moreover analysis of a data set of patients treated neoadjuvantly with chemotherapy revealed that, among the grade III tumors, pathological complete response was reached by 9% of ECM3 versus 74% of non-ECM3 tumors (p=0.04). In differentiated tumors no response difference was observed according to ECM. Conclusions. Our results provide evidence that breast carcinoma progression and response to therapy are influenced by the interaction between tumor and stromal characteristics. (Partially supported by AIRC). Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P5-14-06.