Though a number of studies have been conducted in subjects with established breast cancer, methylation frequencies of genes measured in different labs and in different sample types have been variable and often not reproducible. This is largely due to 4 factors: 1) Variable methods of methylation analysis are used in different studies, 2) Gene panels are not consistent across studies, 3) If the same genes are used, often different promoter CpG sites are used and 4) sources of DNA are variable from study to study (i.e. serum, plasma, tissue, biopsy etc.).
The first issue to be addressed is the selection of the optimal method for methylation analysis. Optimization for small sample volumes (and therefore a small amount of DNA template) should be the focus, to allow for the use of samples obtained from existing prospective studies which are needed to assess the ability of promoter methylation patterns to detect pre-clinical disease. This will require the determination of the absolute sensitivity of the different methods. The absolute sensitivity of an assay is the minimum quantity of target DNA required for successful amplification and detection (77
). To accomplish this, it is suggested here that criteria for publication of methylation data be standardized and include the requirement for confirmation of methylation results from non-sequence based methods (i.e. MSP, and QMSP) by bisulfite sequencing (the gold standard) for a subset of samples. It is also suggested that the amount of DNA used in each assay and the coefficients of variation (CV) for any repeat measures, be reported.
Reproducibility of methylation results is an area of great importance, one that has not been sufficiently addressed in the current literature. Methylation frequencies have largely not been reproducible across studies. This variability may be reduced with the standardization of methods and reporting of results. One study designed to specifically examine the reproducibility of the PMR (percent of fully methylated DNA found in a sample), was based on QMSP analysis of DNA from paraffin-embedded colon cancer samples. This study found the PMR to have high inter-assay CVs with an average of 21% (range 10-38%) (78
). In a recent study, methylation results using a nested QMSP method (QAMA) on DNA obtained from micro-dissected cells from formalin-fixed and paraffin-embedded tumor tissues (n=13) was found to have a good correlation with sequencing results (R=0.982). To our knowledge no studies have reported the reproducibility of measurements obtained from serum or plasma samples.
Because no single gene has been found to be methylated in all breast cancers, it is necessary to use a panel of genes. The variability in the genes included in each panel makes it difficult to compare or combine the results of different studies and to infer how promoter methylation would fare as a screening tool. Further, though two studies may have included the same genes in their panels, they have not necessarily probed for the same CpG sites within the promoter. This adds another layer of variability between studies and there is no concensus or criteria for the selection of CpG sites within a promoter.
Regarding panel selection, up to now it has been based largely on the candidate gene approach, using genes that have a known involvement in carcinogenesis. Methylation of these genes can be found in many other forms of cancer and is not specific to breast cancer. There are genes however, that may have an increased role in breast cancer specifically, such as GSTP1
; which is known to be involved in hormone related cancers (79
; a known player in the family history of breast cancer such that patients with methylated BRCA1
having a similar phenotype to those with BRCA1
) and ERS1
and its associated genes because of the known role of estrogen in breast carcinogenesis (81
). The inclusion of these genes may help improve the specificity of a gene panel for breast cancer.
It will also be important to understand those factors that influence methylation analysis, such as the source of DNA (i.e. serum versus plasma), sample volume, sample handling, storage temperature and duration and freeze/thaw cycles. For example, variation in the amount of DNA obtained from serum and plasma has been shown. The major difference between serum and plasma is the presence of clotting factors (and associated proteins) in plasma. It appears that serum tends to contain approximately 6-fold more DNA than plasma does. Much of this DNA, however, could come from the normal DNA of contaminating leukocytes (82
). Further, large prospective studies needed to test the diagnostic potential of these markers requires long periods of sample storage as cases are accrued through follow-up. The effect of this long-term storage on DNA methylation also needs examining. The Early Detection Research Network (EDRN) of the National Cancer Institute (NCI) has made understanding those factors that may influence methylation analysis, part of their focus (83