This report describes a novel DNA methylation assay called COMPARE-MS that greatly enriches for methylated DNA in heterogeneous samples by combining two independent and complementary strategies: (i) digestion with methylation-sensitive restriction enzymes and (ii) specific capture of methylated DNA by binding to MBD polypeptides immobilized on a magnetic solid matrix. The enriched methylated DNA is then subjected to gene-specific quantitative PCR to determine quantities of methylated CGIs in DNA from heterogeneous samples. One significant advantage of COMPARE-MS over previous methods is that it does not require sodium bisulfite modification. For the methylation-sensitive restriction enzyme step of COMPARE-MS, we chose to use HpaII because its recognition site is abundant in most CGIs. Other methylation-sensitive restriction enzymes may be equally effective. The second step of COMPARE-MS, capture and enrichment of methylated DNA fragments, is similar to the previously reported MIRA (26
), MeDIP (32
) and MECP2-MBD column (23
) based assays. In particular, the methylated-DNA capture step of COMPARE-MS is similar to the MIRA assay, which uses full-length GST-tagged-MBD2-bound magnetic beads for affinity purification of methylated DNA fragments. However, one notable difference between the methylated-DNA capture step of COMPARE-MS and these previous techniques is that we used the small, ~10 kDa MBD fragment of MBD2 for affinity capture and enrichment of methylated DNA fragments, while the MIRA, MeDIP and MECP2-MBD column-based assays used full-length MBD2, α5mC-Abs or the MBD fragment of MECP2, respectively. The use of MBD2-MBD in COMPARE-MS may provide significant advantages over these other reagents for the capture and enrichment of methylated DNA. First, we show that the MBD2-MBD has high affinity and specificity for symmetrically methylated DNA templates. Previous studies have also shown that of all the MBD proteins, MBD2 has the highest affinity for a wide range of methylated DNA sequences (37
), while MECP2 may selectively bind to CpG dinucleotides adjacent to A/T-rich sequences (38
). Second, because it binds double-stranded methylated DNA, MBD2-MBD may be a better candidate for enrichment of CGIs than the α5mC-Abs which only bind single-stranded DNA (32
). This is an especially important consideration since the high G/C content of CGIs may make these sequences resistant to denaturing and prone to forming secondary structures even after denaturing. Finally, using just the small ~10 kDa MBD portion of the MBD2 protein, as opposed to the full-length protein, could eliminate unwanted interactions between unmethylated DNA and other domains on the MBD2 protein. In this study, enrichment of methylated DNA by the combination of digestion with HpaII and capture with the MBD2-MBD minimized the rate of false positives, while maintaining exquisite sensitivity. Furthermore, these processes involve minimal ‘hands-on’ time and small reaction volumes, making COMPARE-MS highly compatible with automated, high-throughput, micro-titer plate analysis. After the initial assay development and optimization stages, we determined the methylation pattern of >160 prostate tissue and cell line samples at multiple CGIs in a single day.
This analysis showed that the CGI hypermethylation pattern at GSTP1, PTGS2 and MDR1 could identify prostate cancer with sensitivities >95% and specificities approaching 100%. Furthermore, we showed that the sensitivity, specificity and dynamic range achieved by COMPARE-MS are highly comparable with those reported for MSP, MethyLight and HeavyMethyl. However, unlike these other techniques, COMPARE-MS is not encumbered with the disadvantages of sodium bisulfite modification. This may allow for higher compatibility with high-throughput, automated, micro-titre-based platforms, and greater ease in the design of real-time PCR primers since there is no reduction in genome sequence complexity. Also, typically, MSP and MethyLight identify the prevalence of a single pattern of methylation at the CpG dinucleotides interrogated by the primers and probes. Although theoretically it may be possible to carry out multiple reactions, each interrogating a different pattern and different set of CpGs, the low sequence complexity of bisulfite-treated DNA limits the application of such strategies. COMPARE-MS, on the other hand, was designed to detect a broader range of abnormal methylation patterns across a large set of CpG dinucleotides without significant design limitations. This feature of COMPARE-MS can be viewed as both a strength and a weakness.
On the one hand, this difference between COMPARE-MS versus MSP and MethyLight may be the reason that four out of the five primary prostate cancer cases in which MethyLight could not detect any GSTP1
CGI hypermethylation were detected by the COMPARE-MS assay. More provocatively, a larger fraction of the tumor-adjacent benign prostate tissues had a small, but significant, amount of methylated CGIs at the GSTP1
genes when analyzed by COMPARE-MS than when analyzed by MethyLight. This finding is in agreement with a recent study showing that the normal epithelia and stroma in tumor-adjacent benign tissues in breast cancers displayed significant hypermethylation of CpG islands (39
). Most of the tumor-adjacent benign tissues in this study had some prostatic intraepithelial neoplasia (PIN) and/or proliferative inflammatory atrophy lesions, which have been shown to have some methylation at the GSTP1
CGI by MSP, but only after rigorous purification of these cells by laser capture microdissection (LCM) (34
). In this study, even without LCM, we were able to quantitatively detect trace amounts of hypermethylation at these CGIs by COMPARE-MS, illustrating the utility of this technique in highly heterogeneous tissues containing only a small amount of methylated DNA. However, since LCM was not used, we could not rule out the possibility that the detected DNA hypermethylation was due to trace contamination by cancer cells. Larger studies, using LCM in at least a subset of samples, would have to be performed to verify this finding.
On the other hand, the ability to detect a broader range of methylation patterns is also a limitation of COMPARE-MS, because it would not be able to quantitate the prevalence of specific patterns of CpG dinucleotide methylation in a given sample. The role of COMPARE-MS would therefore be to quantitate the degree of aberrant hypermethylation with high sensitivity, specificity and rapidity. If necessary, specific samples of interest, as identified by COMPARE-MS, can then be studied by the use of other techniques such as bisulfite genomic sequencing (36
) for determination of the prevalence of specific patterns of DNA methylation.
The COMPARE-MS assay, or components of it, may also be useful for many different methylation detection applications that were not explored in this study. First of all, its sensitivity, specificity and ease of use would allow for design of large clinical trials testing the efficacy of DNA methylation patterns in the diagnosis and risk stratification of human diseases, including cancer. Additionally, since COMPARE-MS does not involve bisulfite modification of the DNA, the original DNA sequence complexity is preserved. Consequently, multiplexing methylation detection at several CGIs in a single assay may be possible. Also, COMPARE-MS may be modified slightly to allow unbiased detection of novel methylated regions by a number of strategies including microarray hybridization (32
), digital karyotyping (39
), restriction fragment length genome scanning (42
), etc. The use of COMPARE-MS could not only enhance our understanding of the role of DNA methylation in health and disease, but improve our ability to sensitively, specifically and rapidly detect CGI hypermethylation.