Predictive Biomarkers and Personalized Medicine

Purpose:Non–small cell lung cancers (NSCLC) comprisemultiple distinct biologic groupswith different prognoses. For example, patients with epithelial-like tumors have a better prognosis and exhibit greater sensitivity to inhibitors of the epidermal growth factor receptor (EGFR) pathway than patients with mesenchymal-like tumors. Here, we test the hypothesis that epithelial-like NSCLCs can be distinguished from mesenchymal-like NSCLCs on the basis of global DNA methylation patterns. Experimental Design: To determine whether phenotypic subsets of NSCLCs can be defined on the basis of theirDNAmethylationpatterns,we combinedmicrofluidics-based gene expression analysis and genomewide methylation profiling. We derived robust classifiers for both gene expression and methylation in cell lines and tested these classifiers in surgically resected NSCLC tumors. We validate our approach using quantitative reverse transcriptase PCR and methylation-specific PCR in formalin-fixed biopsies from patients with NSCLC who went on to fail front-line chemotherapy. Results:We show that patterns ofmethylation divide NSCLCs into epithelial-like andmesenchymal-like subsets as defined by gene expression and that these signatures are similarly correlated in NSCLC cell lines and tumors. We identify multiple differentially methylated regions, including one in ERBB2 and one in ZEB2, whose methylation status is strongly associated with an epithelial phenotype in NSCLC cell lines, surgically resected tumors, and formalin-fixed biopsies frompatients withNSCLCwhowent on to fail front-