Gene expression-based predictors of chemotherapy response in basal-like breast cancer.
10500 Background: The Basal-like subtype is generally associated with high chemo-sensitivity, but not all tumors respond and/or benefit to the same extend. In this study, we sought to identify gene expression predictors of neoadjuvant chemotherapy sensitivity in Basal-like breast cancer. Methods: Expression of 542 genes was measured using the Nanostring nCounter platform from 69 FFPE pre-treated samples of the GEICAM/2006-03 phase II trial, which were treated with epirrubicin/cyclophosphamide followed by docetaxel+/-carboplatin. Research-based PAM50 and Claudin-low predictors were also evaluated. The association between response (Miller-Payne criteria) and gene/signature expression was assessed by multivariable ordinal logistic regression. Significant findings were evaluated in 109 independent triple-negative and Basal-like tumors treated with anthracycline/taxane-based chemotherapy (Hatzis et al.). Finally, interaction tests were performed to identify genes/signatures associated with carboplatin response. Results: In GEICAM/2006-03, 61/69 (88%) tumors were identified as Basal-like by PAM50. High correlation to the Basal-like centroid, or high expression of proliferation-related genes (i.e. FANCA), were found to be significantly associated with high chemo-sensitivity, whereas high expression of genes associated with mesenchymal/stem cell biological processes (i.e. SNAI1 and IL6) and/or luminal differentiation (i.e. MUC1 and FOXA1) were significantly associated with chemo-resistance; similar findings were observed in Hatzis et al. Finally, high expression of genes associated with proliferation/DNA-repair (i.e. ATR) and tight junctions (i.e. CLDN3/4/7) were found associated with carboplatin response, whereas expression of the Claudin-low signature was found associated with carboplatin resistance. Conclusions: High expression of Basal-like and/or proliferation-related genes and low expression of luminal/mesenchymal/stem cell-like biological processes were consistently identified as predictive of chemotherapy response. Our data suggests that gene expression profiling might help shed light into the biological and clinical heterogeneity of Basal-like breast cancer.