Abstract
BACKGROUND: The need for new prognostic factors in breast cancer is ever increasing as breast cancer management evolves. Aberrant DNA methylation plays a pivotal role in cancer development and progression; DNA methylation-based biomarkers may provide independent prognostic information. We used pyrosequencing to investigate the prognostic potential of quantitative DNA methylation of a large set of candidate genes in a Korean single-institution series of operable breast cancer. METHODS: Absolute DNA methylation in 20 candidate genes from an initial set of 30 genes was measured by pyrosequencing of bisulfite converted DNA in 121 fresh frozen breast cancer cases. Survival analyses used continuous and categorized (quintile-based) gene methylation data with time to recurrence (TTR) as an endpoint. Prognostic abilities of gene-only and risk-score models were explored. RESULTS: Median follow-up was 5.1 years; 25 recurrences (21%) were observed. Nodal status, methylation of TWIST1, SLIT2 (both as continuous and categorized variables) and APC, HLA-A, NKX2-5, SERPINB5, SFN (as categorized variables) were significantly prognostic; grade showed a prognostic trend. A multivariate model containing nodal status, grade and TWIST1 was a best fit (p< 0.001) in stepwise regression; risk-score based on this model separated patients into 3 distinct risk-groups (p< 0.001). A gene-only model based on TWIST1 and SFN also classified patients into distinct risk-groups (p=0.009). CONCLUSIONS: This study shows that accurate quantitative measurement of DNA methylation by pyrosequencing identifies a small set of genes with independent prognostic potential in breast cancer. These genes complement the current clinico-pathological prognostic factors and appear to be potential biomarkers that warrant further validation.