Abstract A107: Epigenetic markers for tissue-based breast cancer risk stratification

Effective breast cancer risk management requires accurate breast cancer risk stratification. The Gail model is the most widely used and best validated epidemiological risk assessment model. Though well calibrated, this model is not ideal for individualized risk counseling as it is only 8% better than chance for discriminating between women with breast cancer and unaffected women. Biologically based approaches, such as cytological assessment of nipple aspirate fluid or assessment of mammographic density, only marginally improve the discrimination of the Gail model. Promoter region methylation is one of the earliest identifiable molecular changes during breast carcinogenesis. We have previously shown that RASSF1A methylation, detected in benign breast epithelium obtained by random periareolar fine needle aspiration biopsy (RP-FNA), is associated with increased breast cancer risk, but this was based primarily on correlation with Gail risk and not cancer case prediction. A clinically useful epigenetic risk stratification panel would be ideally suited to benign RP-FNA samples and would provide a high degree of discrimination between benign samples from women with cancer and unaffected women. We performed a genome-wide 5-aza-2’-deoxycytidine (5-aza) induced gene re-expression screen in 6 breast cancer cell lines and 6 early passage primary benign breast epithelial cell cultures to identify potential epigenetic markers of breast cancer risk. MSP assays were designed for the 286 genes that were 1) expressed in benign breast epithelium, 2) not expressed in cancer, and 3) induced by 5-aza in cancer, or 4) had previously shown some potential for breast cancer risk stratification. 20 genes were identified that were frequently methylated in primary breast cancer, rarely methylated in benign RP-FNA samples and never methylated in lymphocytes. Quantitative multiplex methylation-specific PCR assays were designed and optimized for the 17 best genes and then assessed in an archival RP-FNA sample set that included 146 samples from unaffected women, 59 benign samples from breast cancer patients, and 52 primary breast cancers. A three marker panel including PSAT1, HS3ST2, and GNE provided the best discrimination between benign samples from cancer patients and unaffected women. One or more of these markers was scored as positive in 46% of benign samples from cancer patients, 8% of samples from unaffected patients, and 72% of primary cancers. The odds ratio for cancer case prediction based on assessment of benign RP-FNA samples was 9.42 (95% CI 4.31 — 20.59, p Citation Information: Cancer Prev Res 2010;3(12 Suppl):A107.