Abstract 5065: A 22-gene signature predicts response to chemotherapy in a broad range of breast cancer subtypes

Abstract

Abstract Statement of Purpose: To test the utility of a novel, multifaceted 22-gene signature to predict response to chemotherapy in different molecular subtypes of breast cancer. Background: Only 15-25% of breast cancer patients from all histopathological subtypes currently respond to primary chemotherapy with a complete pathological response (pCR) and there is currently no test in widespread clinical use to predict whether a patient will respond. Here we present results describing a 22-gene signature that can identify breast cancer patients who will respond to taxol-combination chemotherapy in a wide range of molecular subtypes. The signature was identified using a biology-driven approach. We hypothesized that genes involved in normal human mammary acinar morphogenesis represent biomarkers capable of stratifying breast cancers. Methods: To determine whether genes with differential expression during acinar morphogenesis predict response to chemotherapy, we have analyzed gene expression microarray results. Three preexisting microarray datasets were used. 1) The dataset of Fournier et al (Cancer Res, 2006. 66:7095) obtained from a time course of two different non-malignant human mammary epithelial cell types grown in a laminin-rich extracellular matrix culture (lrECM). 2) The dataset of Hess et al (J Clin Oncol, 2006. 24:4236) including 133 patients with stage I-III breast cancer obtained before primary taxol-combination chemotherapy (TFAC). 3) The dataset of Popovici et al (Breast Cancer Res, 2010. 12:R5) including 278 patient samples that overlaps with the previous dataset. The assignment of molecular subtype of tumor samples was performed using the intrinsic gene set of 300 genes (Hu et al., BMC Genomics, 2006.7:96). Results: The majority of genes (12 out 22) down regulated upon growth arrest and polarity formation during acinar morphogenesis in lrECM were significantly over expressed in patients who responded to chemotherapy with pCR (p<0.05, Student's T-test and Fisher's Exact test). These 22 genes are referred to here as the 22-gene signature. To test this signature's ability to predict response in different molecular subtypes of breast cancer we applied logistic regression with three-fold cross validation. The 22-gene signature accurately predicted response in all samples with AUC 0.75 and in specific breast cancer molecular subtypes including ER-positive, ER-negative, luminal B, Her2-positive, and basal-like with average AUC 0.69. Conclusions: This study is the first example of a single gene signature to predict chemotherapy response in different molecular subtypes of breast cancer. Its predictive value in ER-negative and basal-like breast cancers is notable. Patients identified as non-responsive have the potential to benefit from adding an alternative treatment early in their care or from improved quality of life in forgoing an ineffective treatment. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 5065. doi:10.1158/1538-7445.AM2011-5065