Clinical Subtype-Derived p53 Gene Signature Is Predictive of Prognosis and Response to Chemotherapy in ER-Positive but Not in ER-Negative Breast Cancers.

Abstract

Abstract BackgroundBreast cancer is a collection of molecularly distinct neoplastic diseases and therefore, we hypothesized that p53 gene mutations may lead to different transcriptional changes in the different molecular subtypes and these may translate into subtype-dependent prognostic and predictive values.MethodsWe developed gene expression-based predictors of p53 status separately for estrogen receptor-positive (ER+) and -negative (ER-) breast cancers from a publicly available data set with known p53 mutation status (n=251). We validated the two signatures on an independent cohort of cancers (n=103) with known p53 functional status and tested their prognostic and predictive values on two other cohorts of breast cancers that received no systemic adjuvant therapy (n=255; n=198), and on one cohort of ER+ patients treated with adjuvant tamoxifen (n=277). We also examined if the p53 signatures were associated with chemotherapy sensitivity in ER+ and ER- cancers, respectively in two separate neoadjuvant data sets (n=233; n=103).ResultsWe developed a 39-gene p53 signature derived from 213 ER+ and a 30-gene p53 signature derived from 38 ER- breast cancers with no overlapping genes. External validation showed a sensitivity and specificity of 89% and 54%, respectively for the 39-gene signature in ER+ breast cancers; and 82% and 61%, respectively for the 30-gene signature in ER- cancers. The 39-gene signature was predictive of worse distant metastasis free survival (DMFS) in ER+ cancers with p53 dysfunction in both prognostic data sets (Hazard ratio (HR): 2.3 (95% confidence interval (CI):1.25-4.23, p=0.005 and HR:2.17 (95%CI:0.85-5.56, p=0.09). It remained predictive of worse DMFS even after tamoxifen adjuvant therapy (HR=2.43, 95%CI: 1.35-4.38, p<0.0001). In contrast it was associated with higher chemotherapy sensitivity in ER+ cancers. Its predictive accuracy for pathologic complete response was of 68% (95%CI: 64-70%), sensitivity 89% (95%CI: 58-98%), specificity 67% (95%CI: 65-68%), positive predictive value 15% (95%CI: 10-17%), and negative predictive value 99% (95%CI: 96-100%) in ER+ cancers. The prognostic and predictive values remained significant in multivariate analysis. The same 39-gene signature was not prognostic or predictive in ER- cancers. The 30-gene signature derived from ER- tumors had no chemotherapy response predictive value in either ER- or ER+ cancers. The p53 dysfunctional cases showed better survival in the absence of any adjuvant therapy among ER- cancers. It had no prognostic value in ER+ cancers.ConclusionThese observations support the hypothesis that predictive or prognostic biomarkers may be best developed separately for different clinical and molecular subsets of breast cancer. P53 dysfunction is clinically most relevant in ER+ breast cancers. Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 6122.