Purpose: To use gene expression profiling of formalin-fixed primary melanoma samples to detect expression patterns that are predictive of relapse and response to chemotherapy.
Experimental design: Gene expression profiles were identified in samples from two studies (472 tumors). Gene expression data for 502 cancer-related genes from these studies were combined for analysis.
Results: Increased expression of DNA repair genes most strongly predicted relapse, and was associated with thicker tumors. Increased expression of RAD51 was the most predictive of relapse-free survival (RFS) in unadjusted analysis (HR 2.98, p=8.80×10−6). RAD52 (HR 4.73, p=0.0004) and TOP2A (HR 3.06, p=0.009) were independent predictors of RFS in multivariable analysis. These associations persisted when the analysis was further adjusted for demographic and histological features of prognostic importance (RAD52 p=0.01; TOP2A p=0.02). Using principal component analysis, expression of DNA repair genes was summarized into one variable. Genes whose expression correlated with this variable were predominantly associated with the cell cycle and DNA repair. In forty-two patients treated with chemotherapy, DNA repair gene expression was greater in tumors from patients who progressed on treatment. Further data supportive of a role for increased expression of DNA repair genes as predictive biomarkers are reported, which were generated using multiplex polymerase chain reaction (PCR).
Conclusions: Over-expression of DNA repair genes (predominantly those involved in double-strand break repair) was associated with relapse. These data support the hypothesis that melanoma progression requires maintenance of genetic stability and give insight into mechanisms of melanoma drug resistance and potential therapies.