Hypermethylation in histologically distinct classes of breast cancer.

PURPOSE A number of different genes are known to be inactivated by aberrant hypermethylation in breast cancer, but it is still unknown to what extent these epigenetic alterations differ according to specific breast cancer phenotypes. We sought to determine whether the extent of hypermethylation or defined profiles of gene hypermethylation are associated with biological characteristics of breast cancers. EXPERIMENTAL DESIGN We evaluated methylation status of 12 different genes in a series of 109 invasive breast tumors, representing the ductal, lobular, and mucinous histologic subtypes using methylation-specific PCR. Frequencies of methylation were compared across the recognized histologic classes, and multivariate techniques (latent class analysis, factor analysis, recursive partitioning, and hierarchical clustering) were used to seek patterns of methylation for individual genes that distinguish recognized histologic types of breast cancer or define breast cancer phenotypes on a molecular level. RESULTS All 109 cases studied have aberrant methylation of multiple genes (3 to 10 genes per case), demonstrating that gene hypermethylation is pervasive in breast cancer. Lobular cancers and mucinous cancers, which often have relatively low levels of chromosomal changes, have higher overall frequencies of hypermethylation than ductal cancers (49% in lobular and mucinous versus 40% in ductal), but there is a relatively unimodal distribution of methylation frequency for all three histologic types. Only one of the individual genes studied, BRCA1, has a variable frequency of methylation that is significantly dependent on histologic pattern of tumor growth, with a higher frequency of methylation in mucinous cancers than ductal or lobular cancers. Although some trends of histology-specific gene methylation were seen, methylation patterns could not definitively classify breast cancers according to histologic type. CONCLUSIONS Although a more comprehensive hypermethylation profile could potentially be useful for breast cancer classification and understanding the biology of this disease, it appears that the hypermethylation patterns across various forms of breast cancer are less distinct than those between breast cancer and cancers of different tissue origins. Furthermore, the relatively unimodal distribution of methylation frequency for all three histologic types does not support there being a distinct CpG island methylator phenotype for breast cancer.